J. Chem. SOC., Faraday Trans. I , 1989, 85(12), 3939-3952 The Use of Electron Paramagnetic Resonance Techniques in the Molecular Approach to Heterogeneous Catalytic Processes on Oxides Michel Che* and Catherine Louis Laboratoire de Re'activite' de Surface et Structure, UA 1106 CNRS, Universite' P. et M. Curie, 4 place Jussieu, 75252 - Paris Cedex 05, France Zbigniew Soj ka Jagellonian University, ul. Karasia 3, Krakow, Poland This review shows that EPR spectroscopy can be a powerful tool for catalysis research. This is illustrated in the particular case of highly dispersed Mo/SiO, catalysts prepared by grafting. The different steps of the preparation may be identified : e.g. the EPR changes observed in the Mo5+ coordination sphere during the grafting process of MoCl, onto silica, showed that the grafted species is probably SiOMoCl? (8, = 1.952, g,, = I .968).The catalyst was also characterized after activation treatment such as thermal reduction. Three different Mo5+ species have been detected. Their coordination sphere was determined using probe molecules such as water, methanol and natural or 13C-enriched carbon monoxide. These Mo5+ species are molybdenyl compounds surrounded by oxygen ligands in distorted octahedral (Maid) (gl = 1.944, g,, = 1.892), square-pyramidal (MoXd) (8, = 1.957, g,, = 1.866), and distorted tetrahedral symmetry (MO,~;) (8, = 1.926, g,, = 1.755). The subscripts 6c, 5c and 4c stand for hexa-, penta- and tetra-coordinated, respectively. The EPR results obtained on UV-irradiated oxidized Mo/SiO, on which methanol was preadsorbed, are also consistent with the variation of selectivity in methyl formate and formaldehyde with Mo dispersion observed during methanol oxidation. They extend the reaction mechanism previously proposed on the basis of kinetic data.EPR revealed the formation of a hydroxymethyl radical 'CH,OH. The deactivation of this radical by an 'external' redox process is favoured on highly dispersed Mo catalysts such as grafted Mo/SiO,. The process involves radical migration on silica where it can react with SiOCH, groups, forming methyl formate. The main steps in the development of a catalytic process involve the control of all the preparation steps of the catalytic system, the characterization of its bulk and surface physico-chemical properties and finally the measurement of its catalytic properties (essentially activity, selectivity and stability). These steps are usually repeated until the best catalytic system is obtained by trial and error.The present paper will illustrate how the EPR techniques have contributed to improve our understanding, at a molecular level, of these main steps for catalytic reactions occurring on oxide systems. The use of powder catalysts generates a number of difficulties particularly the frequent presence of several EPR-active species.' In order to determine their magnetic parameters, it is often convenient to employ adapted techniques such as spectral simulation,l isotopic labelling,l Q-band2 and third-derivative spectra. This paper is a review of the EPR results obtained on silica-supported molybdenum catalysts prepared in our laboratory, either by the classical impregnation method, or the 39393940 EPR of Catalytic Processes on Oxides grafting method.The latter method involves the reaction of molybdenum pentachloride with the silanol groups of a high-surface-area silica support. The experimental details and the full description of the results, if already published, can be found in the original references. In the present review, the surface and coordination properties of the grafted molybdenum ions using various probe molecules will be emphasized before presenting the main steps of the oxidation of methanol used as the catalytic test reaction. Reduced Mo catalysts exhibit EPR signals owing to the presence of paramagnetic Mo5+ ions (4dl).If the latter are in interaction, the dipolar coupling broadens their signal and prevents their detection. As illustrated below, the Mo5+ signal often exhibits some hyperfine lines on the low-field side. This hyperfine structure is due to the interaction between the unpaired electron and the nuclear spin ( I = 5/2) of the ’,Mo and 9 7 M ~ isotopes which have about the same magnetic moment (natural abundance 15.8 and 9.5 %, respectively). Results and Discussion Catalyst Preparation The first EPR studies of supported Mo catalysts were performed with samples prepared by impregnation of the support with (NH,),Mo,02,.4H20, calcined at 773-873 K and then reduced.*-* The alumina support, which is of interest from a catalytic point of view, gives ~pectra’’~ whose poor resolution may be attributed for the most part to the superhyperfine splitting due to the interaction of the unpaired electron of Mo5+ with 27Al nuclei ( I = 5/2).Thus, most of the EPR characterization work has been performed with silica, which does not lead to this problem. Impregnat ion Method The reduction by H, (100 Torr) at 773 K for 20 min of impregnated Mo/Si02 catalysts prepared by the incipient wetness method, led to an EPR Mo5+ signal at g , = 1.940 and g,/ = 1.882, characteristic of an axial crystal fieldg? lo (table 1). After reduction by CO (100 Torr’f) at 873 K for 3 h, another Mo5+ species was detected in addition to the species described above’’ (table 1). By comparison with the EPR signals of the isopolyanion Mo,O:; (gl = 1.930, g , = 1.9 19), 95Mo-enriched Mo/SiO, catalysts and various molybdenyl compounds, Che et al.9711 showed that both species possess a molybdenyl character.They also showed that the Mo5+ coordination spheres were composed of oxygen ligands, in distorted octahedral (Moi:) and square-pyramidal symmetry (Moi;). In the case of MoZT, the vacancy in the Mo5+ coordination-sphere was located in the trans position with respect to the ‘yl’ oxygen. The following models were proposed for these two species: 0 0 0 Grafting Methods This method involves a chemical reaction, under air- and water-free conditions, between surface hydroxyl groups of silica and MoCl,. In contrast to impregnation, where the Mo becomes bonded to the support during calcination, the grafting reaction produces this t 1 Torr = 101 325/760 Pa.M .Che, C. Louis and Z . Sojka 3941 Table 1. g-values of Mo5+ EPR signals and characteristics of reduced Mo/SiO, catalysts samples Mo (wt %) reduction g, gll ref. impregnated 0.17, 2 H2/773 K 1.940 impregnated 0.17, 2 C0/873 K 1.961 1.942 impregnated 1.03 H,/873 K 1.956 1.945 pseudografted 0.15, 1.56 H2/773 K 1.958 1.941 grafted 0.18, 0.33, 1.05 H,/873 K 1.944 1.957 1.926 1.882 9, 10 1.861 10 1.891 1.859 12, 14, 17 1.889 1.856 10 1.885 1.892 12, 14, 17 1.866 1.755 Experimental error: Agl = k0.002; Agl, = f0.004. bonding directly and is expected to prevent Mo migration and aggregation during calcination and lead to a better dispersed Mo. In the first study," MoCl, was treated with a suspension of silica in chloroform. The catalyst was then washed with chloroform, further hydrolysed with water vapour and finally dried at 393 K.The samples obtained were blue. After thermal reduction under H, or CO, the same Mo5+ species were detected as in the case of impregnated samples reduced under CO (table I). The blue colour of the samples was due to the presence of molybdenum blues (mixed-valence Mo5+-Mo6+ hydroxides) in weak interaction with silica, since these hydroxides could be eliminated by washing with water. As a matter of fact, the molybdenum was shown not to be really grafted,l2>l3 hence the term pseudo- grafted in table 1. The grafting method was improved by the use of MoCl, as vapour12* 14, l5 or dissolved in cy~lohexane,l~*'~ a solvent less polar than chloroform. The choice of MoC1, is particularly attractive since it is monomeric and paramagnetic in the vapour phase and in cyclohexane.16 If not altered during deposition, the Mo5+ oxidation state can be used as a very effective probe to monitor, via its EPR spectrum, the changes in the coordination sphere that occur during the grafting process.14 The results presented below concern the preparation in vapour phase.The reactor used for this kind of preparation was equipped with an EPR tube which permitted the monitoring of the grafting process by EPR.l2l 14* l5 When silica (Spherosil XOA 400, RhGne Poulenc, 400 m2 g-') was heated in the presence of only MoCl, at 473 K, a Mo5+ EPR signal with axial symmetry (gl = 1.952, g,, = 1.968, A , = 37 G, A , , = 70 G) appeared and increased in intensity (fig.I). The change of the Mo5+ EPR signal before (the monomer MoC1, gives an isotropic signal at g = 1.952), and after grafting indicates that the Mo5+ coordination sphere is affected by the grafting process. The parameters of the EPR signal and the UV-visible spectrum of Mo after grafting were found to be very similar to those of the [MoOClJ ion, suggesting the following grafting reaction :12, 1 4 3 l5 MoCl, + SiOH -+ SiOMoC1, + HCl. (1) The samples turned from red-orange after grafting, to a blue colour when exposed to ambient air. This change arose from the hydrolysis and the partial oxidation of loosely bonded MoCl, into molybdenum b 1 ~ e s . l ~ ~ ~ ~ The latter were eliminated by washing with water or 1 mol dmP3 ammonia solution and the samples turned brown, the colour characteristic of grafted molybdenum.This washing step turned out to be all the more effective because the pH was more basic, increasing the surface negative charge density, thereby removing the negatively charged molybdenum blue species by electronic rep~lsion.'~3942 EPR of Catalytic Processes on Oxides u i/'l Fig. 1. EPR spectra recorded at 77 K of Mo/SiO, catalysts obtained after grafting with MoC1, vapour at 473 K: (a) first-derivative and (b) third-derivative. After standard thermal reduction, i.e. in H, (200 Torr) at 873 K for 2 h and further evacuation at 873 K for 30 min, the grafted samples exhibited a complex asymmetric line with a mean g value of 1.944 [fig. 2(a)]. The third-derivative spectrum [fig. 2(b)] revealed the presence of three signals1' for which g values are listed in table 1.These signals correspond to three Mo5+ species in axial symmetry with g , > g,,. While two signals were found similar to those of Mog,f and Mo:,f of impregnated samples (table l), the third one Mot: was new. Molybdenum Coordination Chemistry In order to characterize the coordination sphere of the Mo:: species and to confirm the symmetry previously attributed to Mot: and Moi,+, the approach was to use probe molecules such as water, carbon monoxide,127 1 4 7 l7 and methanolls to complete the coordination sphere of coordinatively unsaturated Mo5+ ions. As shown below, water and methanol could only provide information on the presence of a coordinatively unsaturated state, whereas l2C0 and 13C0 could give information on the number of coordination vacancies through the superhyperfine structure due to the nuclear spin When adsorbed at room temperature, water, carbon monoxide and methanol act only as ligands and not as redox agents.This was inferred from the constant number of spins before and after adsorption. Methanol is non-dissociatively adsorbed on Mo below 350 K." I = 1/2 of 13C.M. Che, C . Louis and Z . Sojka 3943 Fig. 2. EPR spectra recorded at 77 K of a reduced grafted Mo/SiO, catalyst (0.33 wt YO) : (a) first- derivative ; (b) third-derivative. Fig. 3. EPR spectra recorded at 77 K of a reduced grafted MolSiO, catalyst (0.33 wt YO): (a) after reduction; (b) after water adsorption at 300 K (1 Torr); (c) after water adsorption at 300 K (18 Torr). Water and Methanol Adsorption Water adsorption at room temperature led to several transformations in the EPR spectrum recorded at 77 K:l2v1' first, the MO:: signal disappeared whereas that of MoZZ increased [fig.3(a)], then the MoZZ decreased to disappear while that of Moi: increased [fig. 3(b)]. Water desorption at room temperature led to the reappearance of the Moi: signal intensity. At 773 K, the water desorption was complete since the three initial Mo5+ signals were observed again. Similar results were observed with methanol.l* Variable temperature experiments were performed to monitor the intensity changes of the three signals (fig. 4).3944 1.0 EPR of Catalytic Processes on Oxides 4 77 120 170 210 250 T/K Fig. 4. Temperature dependence of the Mo5' signal intensities after methanol adsorption on reduced grafted Mo/SiO, catalysts (0.33 wt %).The arrows indicate the transformation of less coordinated species to more coordinated ones. Examination of fig. 3 and 4 shows that H20 and methanol adsorption involve a two- step p r o c e s ~ . ~ ~ ~ ~ ~ * ' ~ The introduction of water or methanol into the MoZ: coordination sphere transforms its EPR signal into one analogous to that for Mo::. The introduction of additional molecules completes the coordination sphere of MO~,;, and of the partially coordinated Moi:. This induces the disappearance of their EPR signals and the increase in intensity of a signal similar to Moi:. These results suggest that Moi: can adsorb more than one molecule of water or methanol and indicate that Moi: contains more than one vacancy in the coordination sphere.The Mot: ion is then the most unsaturated species of the three Mo5+ species. As its signal transforms into signals corresponding to molybdenyl ions, the Moi: species is expected also to possess a molybdenyl character. Carbon Monoxide Adsorption The counting of coordination vacancies in transition-metal ions has been shown to be possible with 13C-enriched C0,3* 12, 17, l9 particularly for ground-state orbitals like d,z or d,g-yl, whose lobes point along the metal-ligand bonds.lg The adsorption of carbon monoxide12* l7 induced the disappearance of the Moi: signal and the appearance of a signal composed of two lines at g = 1.965 and 1.969 whose intensities increased with pressure (up to 100 Torr), suggesting that Mot,+ can also coordinate CO (fig. 5).It was difficult to estimate whether CO could enter in the coordination sphere of Mo:: since some of its signal remained visible after CO adsorption. Note that the perpendicular components of the Mo:: and Moi: signals are less intense than the associated parallel components (fig. 5). This is in contrast to the situation before CO adsorption (fig. 2) or when water was adsorbed (fig. 3). This may be explained by the superimposition of the g,, (or g3) components of the new signal(s) induced by CO adsorption either on the g , components of the Mo;: and Moi: signals,3945 M. Che, C. Louis and Z . Sojka 5 0 G c-, Fig. 5. EPR spectra recorded at 77 K of a reduced grafted Mo/SiO, catalyst (0.33 wt %) after l2C0 adsorption (200 Torr/300 K) : (a) first-derivative ; (b) third-derivative.leading to a decrease in their intensity, or on their g,; components, inducing in this case an increase of intensity. When 13C0 was adsorbed, the signal located at g = 1.965 was broader than with l2C0 and the MoZ,f signal less well resolved [fig. 6(a)]. The third-derivative spectrum shows that, instead of two lines, the spectrum is now composed of a quartet, i.e. four lines of different intensities [fig. 6(b), 7(b)], separated from each other by ca. 7.5 G and located at magnetic fields corresponding to apparent g values of 1.974, 1.969, 1.965 and 1.960. The two central lines of the quartet are located at the same field values as the two lines obtained after l2C0 adsorption. These four lines were not resolved in Q-band,12.17 suggesting that they arise from a superhyperfine coupling between the unpaired electron of Moi: and the nuclear spin ( I = 1/2) of 13C.Indeed, the hyperfine coupling constant is independent of the microwave frequency,2 therefore the splitting observed in X-band (7.5 G) is too weak with respect to the linewidth in Q-band (60 G) to be detected. Assuming similar linewidths, these four lines were decomposed into two superimposed triplets centred at g = 1.965 and 1.969, and with a hyperfine constant of 7.5 G (fig. 7). Using the relative intensities of the two lines obtained after l2C0 adsorption, the superposition of the two triplets with relative inner intensities 1 : 2 : 1 compares well with the intensities of the experimental spectrum (fig.7). Each triplet arises from the interaction between the Mo5+ unpaired electron and the 13C nuclear spin of two apparently equivalent 13C0 molecules. The Moicf coordination sphere is then completed on admission of two CO molecules. It was therefore deduced that Moicf is a tetracoordinated species : 0 +2co - OC-Moi: - co 03946 l 2 C O t EPR of Catalytic Processes on Oxides I- 1.9 6 5 DPPH Fig. 6. EPR spectra recorded at 77 K of a reduced grafted Mo/SiO, catalyst (0.33 wt 70) after I3CO adsorption (200 Torr/300 K) ; (a) first-derivative ; (b) third-derivative. Fig. 7. Third-derivative EPR spectra recorded at 77 K of the Moi,+ ion coordinated by (a) l2C0 and (b) I3CO; ( c ) is the analysis of these signals. The Coordination Sphere of Mo5+ Ions after Adsorption of Gas-phase ligands On the basis of the knowledge of the MoiZ coordination sphere, the mechanism of water adsorption can be discussed.It occurs in two steps: (i) the admission of a first water molecule, probably located in the equatorial plane since the EPR signal of the partially hydrated Mo5+ ion is similar to that of MoZ,+, which possesses a square-pyramidalM . Che, C. Louis and 2. Sojka 3947 structure; (ii) the admission of a second water molecule in axial position, since the EPR signal of the fully hydrated Mo:,f is similar to that of Mo:,f. The following adsorption mechanism was proposed : 1 2 ~ l7 On the other hand, one water molecule can be adsorbed on MoXl: H H \ 0 J \ 0 J T T MO;; MoZ On the same basis, the following mechanism of methanol coordination by reduced Mo/SiO, samples was proposed :I8 H H H CH,OH I CH3OH I I * CH3O-Moz -XH3 T> 190K Moz - CH,O-MO~ 120- 190K H CH,OH I MO: CH3O-Moz T > 190K An excess of H 2 0 or CH,OH may be adsorbed in the second sphere of Mo5+ ions by means of hydrogen bonding or on the silica support by different modes of coordination.The g values of the three initial Mo5+ species and of the hydrated forms can be interpreted on the basis of a g, versus g , , diagram (fig. 8) plotted from the g-tensor components known for a number of molybdenyl compounds of C4" symmetry :11 [MOO L,LJ"-, (x = 4, 5, y = 0, 1, n = 1, 2, with L,L' = monodentate ligands). In line with the theoretical expressions of the g-tensor components, fig. 8 illustrates the dramatic variation of g,i, mainly due to the spin-orbit coupling constant R, of L, in contrast to that observed for gl.ll The following conclusions may be drawn: (i) molybdenyl compounds with the same type of ligand L are found in the same region of the diagram, demonstrating the preponderant influence of AL ; (ii) the molybdenyl halide series expands over both parts of the diagram, with A, varying from 270 cm-l for3948 EPR of Catalytic Processes on Oxides A 2 0 g l 19 18 - ./ / gl ’ gll / gll ’ gl / / I / / oxyqen c o n t a i n i n g / halide liqonds / / / tonic c h a r a c t e r / I / / / / / / / f / / / c o v a l e n t c h a r a c t e r 17 1.8 1.9 2.0 2 1 2 2 gll Fig. 8. Representation of various molybdenyl compounds is the g,, gll plane : (1) Mo,O,,H, ; ( 2 ) MoO(HSO,),; (3) MoOJTeO,; (4) MoO(H,PO,),; (5) MoO(H,AsO,),; (6) M0,O;;; ( 7 ) PMoW,,; (8) MoO(NCS),; (9) MoOFi-; (10) MoOClE-; (11) MoOCl,(H,O)-; (12) MoOC1;; (13) MoOBrE-; (14) MoOIg-; (15) Moil; (16) Moi:; (17) Moil; (18) SiOMoCl,.Compounds 1-14 from ref. (1 1) and references therein. fluorine to 5060 cm-l for iodine; (iii) a comparison between MoOCl,, MoOCC-, and MoOCl,(H,O)- shows that the axial ligand has little influence on the g-tensor components; (iv) the oxygen-containing ligands are characterized by g , > g ; A,, the spin-orbit coupling constant of oxygen is the same whether the oxygen is an oxide 02- ion or belongs to an hydroxyl group, a water or a methanol molecule. This is the reason why the partially and fully coordinated MoZ; species possess EPR signals similar to those of Moi: and Mo::, respectively [eqn (3)-(5)].By contrast, when a ligand is connected to the central ion by a non-oxygen atom, the spin-orbit coupling constant changes and affects the g,, value. This is the case of CO connected to Moi: by the carbon atom (A, = 28 cm-’ instead of A, = 152 cm-l). Moreover, the CO adsorption lowers the Mo symmetry which becomes C,. The two lines at g = 1.965 and 1.969 observed after CO adsorption are therefore attributed to the g, and g, components, respectively, of an orthorhombic signal; the g , component of the Mo5+ carbonyl species, MoZ: (CO),, as stated above overlaps with the MoZ; and Moil signals and is thus ill-defined. Therefore, the magnetic parameters of the Mo:; (CO), are as follows: g, = 1.965, g, = 1.969, g , = ill-defined, A , = 7.5 G, A , = 7.5 G, A , = ill-defined.The hypothesis of two different Mo5+ carbonyl species in axial symmetry may be discarded, since after admission of two CO ligands in its coordination sphere, Moi: no longer preserves its axial symmetry. In addition, the two CO appear equivalent (same superhyperfine constant of 7.5 G) and are thus probably located in the equatorial plane. Two models for 2 CO coordinated to Mo:: can be proposed: 0 0 0 0M. Che, C. Louis and Z. Sojka 3949 Fig. 9. EPR spectra recorded at 77 K of the Mo/SiO, catalysts prepared by grafting with MoC1, vapour at 473 K, followed by evacuation: (a) at 473 K for 30 min; (b) at 573 K for 30 min; (c) at 773 K for 30 min; ( d ) third-derivative spectrum of (c). The cis model is more probable since it is formed from Mo:: bonded to the surface and that attack by two CO is likely to yield the cis isomer for steric reasons.Eflect of Evacuation Temperature directly after Grafting As described above, after grafting in the vapour phase, the sample exhibits an EPR signal with g,l > g , (fig. 1) owing to the formation of the SiOMoC1, species [eqn (l)]. If after grafting, the sample was directly evacuated at increasing temperatures without any intermediate exposure to air, the following changes were observed : (i) at ca. 473 K, the relative values of the g components are reversed with g , > g,l [fig. 9(a)]. The spectrum is similar to that of pure Moi; species obtained after water adsorption on reduced sample [fig. 3(c)]; (ii) at 573 K, the spectrum exhibits the signals of both MoZ: and Moi,f species [fig.9 (b)] ; (iii) at 773 K, the spectrum becomes similar to that of the reduced samples (fig. 2), i.e. with three Mo5+ species [fig. 9(c)].' The spin concentration measurements indicated that the number of Mo5+ ions remains constant during evacuation, suggesting that the spectral changes arise only from modifications within the Mo5+ coordination sphere. Fig. 8 shows that the inversion of the g values on sample evacuation at 473 K is due to the replacement of the chlorine ligands of SiOMoC1, by 0,- ions of hydroxyl groups or water molecules arising from the silica support, leading to the formation of MoZ,+. On subsequent increase of temperature, some oxygen ligands are lost, leading to the formation of vacancies in the Mo5+ coordination sphere, i.e.to Mog; and Mo~:.~~. 1 4 7 l5 Conversely, the Mo:: signal disappeared when the sample was left in static vacuum at room temperature for ca. 24 h12,15 or when water was adsorbed : its signal was first transformed into one similar to Mo;: and then into one similar to Moi:. Upon water adsorption, it was shown previo~slyl~ that these changes are due to the admission of a first then a second water molecule within its coordination sphere. The binding with H,O is via the available doublet of oxygen of the water molecule, the molybdenum remaining paramagnetic. The similarity between the signals of hydrated Moi: species and those of Mo;; and Moil3950 EPR of Catalytic Processes on Oxides means that MoER and Moi: ions are also bound to oxygen ligands via dative bonds. These oxygen ligands can belong to OH hydroxyl groups of silica and act as H20.Methanol Oxidation Experiments have shown that Mo is much better dispersed on grafted samples than on impregnated samples and in strong interaction with silica.12-15 These characteristics are very attractive to investigate the effect of isolated Mo ions which are believed to be the active sites of certain catalytic reactions. The influence of the Mo relative distance for lower Mo dispersions can also be studied. Impregnated and grafted Mo/Si02 samples have been compared in methanol oxidation2' which is known to be structure-sensitive. 21, 22 While on polycrystalline MOO,, the main product is formaldehyde, on orientated crystallites the selectivity to formaldehyde, dimethoxymethane or dimethyl ether depends on the face exposed.In an earlier study20 the selectivity in formaldehyde and methyl formate, the main products on supported Mo, was shown to depend upon the molybdenum content of grafted catalysts. As the Mo content decreases, i.e. when the Mo dispersion increases, the selectivity in methyl formate increases while that in formaldehyde decreases. In contrast, the dependence is not so clear for impregnated catalysts because of the lack of reproducibility in their preparation. However, their main product is formaldehyde, and this appears to be due to a lower Mo dispersion. The study of the formation of methyl formate using different reaction mixtures and kinetic calculations led us to propose a reaction mechanism.On Mo sites, methanol leads to formaldehyde, which spills over to silica, where it further reacts with methoxy groups to form methyl formate via a hemiacetal intermediate. 2o An EPR study was performed with the aim to establish the elementary steps of this mechanism'' and to investigate the cycle of model reactions, which may reproduce the most important features of the real catalytic process. As shown above, methanol adsorbed on reduced grafted Mo/Si02 enters into the coordination sphere of MoZL [eqn (5)]. No EPR signal appeared when methanol was adsorbed on the oxidized sample. However, when the sample was heated above 373 K, an Mo5+ signal corresponding to Moi: species was observed, indicating that Mo6+ ions were reduced. The molybdenyl bond Mo6+=02- of oxidized Mo is known to be activated as [Mo5+-0-]* not only by electron transfer during UV-irradiation at low temperature (77 or 300 K)23324 but also by vibrational excitation during thermal 26 The advantage of the low-temperature UV-irradiation is to produce stable paramagnetic species, which disappear at higher temperature and are therefore undetectable after thermal treatment.When the oxidized grafted Mo/Si02 samples on which methanol was first adsorbed, were UV-irradiated at 77 K, two species were detected by EPR : MoiZ (gl = 1.946, g,, = 1.90) as in the case of thermal activation, and another one as a sharp triplet (gaV = 2.003, A , = 23, A , = 29 and A , = 13 G), attributed to the hydroxymethyl radical 'CH20H (fig. 10). Above 140 K, the triplet signal decreased in intensity while that of Mo:: increased.The same experiments performed on the pure silica support led to the formation of an EPR signal as a doublet g,, = 2.007 and A = 140 G, corresponding to 'CH20 radicals generated from methoxy groups : SiOCH, 2 SiOCH; + H' (7) The EPR results with the Mo/Si02 system are consistent with the photoinduced formation of an excited state ria ligand-to-metal charge transfer. This is in agreement with earlier which showed that methanol reacts with the [Mo5+-O-]* tripletM. Che, C. Louis and Z . Sojka 395 1 A H - @) 4 Fig. 10. EPR spectra recorded at 77 K, obtained after 10 min of UV-irradiation at 77 K with preadsorbed methanol outgassed to Torr: (a) reduced grafted Mo/SiO, catalyst (0.33 wt YO); (b) SiO,. excited state.In our case, methanol was adsorbed first at room temperature and then UV-irradiated at 77 K. The following reaction scheme is proposed : CH,OH CH,OH 'CH,OH The last step is a ligand-to-ligand hydrogen-transfer process. It is in accordance with the increase of the Mo5+ and 'CH,OH signals upon irradiation and their mutual dependence. At temperatures above 140 K, the 'CH,OH radical becomes unstable. Its decay can be explained in two different ways: (i) deactivation by an 'internal non-redox' bi- and/or uni-molecular decay of 'CH,OH radicals resulting in EPR-invisible species. This process, however, does not involve any change in the Mo5+ ion concentration, in contrast to the experimental results observed; (ii) by an 'external' redox decay via interaction with Mo6+ ions resulting in their reduction as evidenced by the increase of the Mo5+ signal intensity and formation of adsorbed formaldehyde by a ligand-to-ligand hydrogen-transfer process : T>140 K Mo6+=02- + 'CH,OH CH,O-Mo5+-0H- (9) This study has shown that the oxidation of methanol is a multi-step process. The order in which these steps occurs, depends on the temperature and the sequence of reactants adsorption, the dispersion and the oxidation state of molybdenum. Other experiments which, for the sake of brevity are not presented here, have confirmed the above processes.18 They involve the adsorption of methanol onto (Mo6+-O-) formed by N,O decomposition onto reduced Mo/SiO,.In the case of thermal activation, the reaction obviously is more complex than for UV-irradiation since methanol can dissociate, leading to the reduction of Mo6+ into Mo5+. This EPR study is consistent with the variation of selectivity in methyl formate and formaldehyde with the Mo dispersion observed during methanol oxidation and extends the reaction mechanism proposed on the basis of kinetic data.20 Indeed, on highly dispersed Mo catalysts such as grafted Mo/SiO,, the decay of the *CH,OH intermediates by an 'external' redox process is favoured.As a consequence, these radicals formed at3952 EPR of Catalytic Processes on Oxides the first stage of the reaction may have enough time to spill over to silica before the final deactivation takes place. This makes it possible for the side reactions with support surface groups, such as SiOCH,, to occur during migration, giving rise to the formation of methyl formate as discussed above. On the other hand, for bulk MOO, or low dispersed Mo/SiO, catalysts prepared by impregnation, where the Mo adsorption sites are in close interaction and involve at least pairs of reducible Mag+ ions, the two deactivation processes (‘internal ’ and ‘external redox ’) are coupled.No migration of ‘CH,OH is required for its deactivation to occur and CH,O becomes the main product of methanol oxidation. Conclusion The present review has shown that EPR can be a powerful tool: during the catalyst preparation, to identify at a molecular level, the different steps of the preparation process. In the case of grafting of MoCl, to silica, the changes in the Mo5+ coordination sphere have been observed and analysed ; to characterize the catalyst after activation treatment such as thermal reduction.Three different Mo5+ species have been detected and their coordination spheres determined using probe molecules such as water, methanol and natural or 13C-enriched carbon monoxide ; to understand the mechanism of the elementary steps that occur during the catalytic reaction, as illustrated here for methanol oxidation. The EPR results have extended the reaction mechanism proposed on the basis of kinetic data. It is, however, wise to remember that other techniques can complement the conclusions based on the EPR technique. Owing to its high sensitivity towards paramagnetic species, EPR may at times give a very incomplete, not to say wrong, picture of the main process(es) occurring on catalytic surfaces.References 1 M. Che and E. Giamello, Spectroscopic Analysis of Heterogeneous Catalysts, ed. J. L. G. Fierro, 2 M. Che and Y. Ben Taarit, A h . Colloid Interface Sci., 1985, 23, 179. 3 M. Che, B. Canosa and A. R. Gonzalez-Elipe, J. Phys. Chem., 1986, 90, 618. 4 G. K. Boreskov, V. A. Dzis’ko, V. M. Emel’yanova, Yu, I. Pecherskaya and V. B. Kazanskii Dokl. 5 J. Masson and J. Nechtschein, Bull. SOC. Chim. Fr. 1968, 3934. 6 K. S. Seshadri and L. Petrakis, J. Phys. Chem., 1970, 74, 4102. 7 K. S. Seshadri and L. Petrakis, J. Catal., 1973, 30, 195. 8 M. Dufaux, M. Che and C. Naccache, J. Chim. Phys., 1970, 67, 527. 9 M. Che, J. C. McAteer and A. J. Tench, J. Chem. Soc., Faraday Trans. I , 1978, 4, 2378. Elsevier, Amsterdam, 1989, in press. Akad. Nauk. SSR, 1963, 150, 829. 10 M. Che. F. Figueras, M. Forissier, J. C. McAteer, M. Perrin, J. L. Portefaix and H. Praliaud. Proc. 6th 11 M. Che, M. Fournier and J. P. Launay, J . Chem. Phys., 1979, 71, 1954. 12 C. Louis, Doctoral Thesis (Paris, 1985). 13 C. Louis, M. Che and F. Bozon-Verduraz, J . Chim. Phys., 1982, 79, 803. 14 M. Che, C. Louis and J. M. Tatibouet, Polyhedron, 1986, 5, 123. 15 C. Louis and M. Che, J. Catal., to be submitted. 16 F. A. Cotton and G. Wilkinson, Advanced Inorganic Chemistry (Wiley, Chichester, 4th edn, 1980), p. 17 C. Louis and M. Che, J. Phys. Chem., 1987, 91, 2876. 18 Z. Sojka and M. Che, J. Phys. Chem., in press. 19 L. Bonneviot, D. Olivier and M. Che, J. Mol. Catal., 1983, 21, 415. 20 C. Louis, J. M. Tatibouet and M. Che, J. Catal., 1988, 109, 354. 21 J. M. Tatibouet and J. E. Germain, J . Catal., 1981, 72, 375. 22 J. M. Tatibouet, J. E. Germain and J. C. Volta, J. Catal., 1983, 82, 240. 23 V. B. Kazansky, Proc. 6th Znt. Cong. Catal., 1976, London, 1977, 1, p. 50. 24 M. Anpo, I. Tanahashi and Y. Kubokawa, J . chem. SOC., Faraday Trans. 1, 1982, 78, 2121. 25 M. Che and A. J. Tench, Adv. Catal., 1982, 31, 77. 26 V. B. Kazansky, Kinet. Katal., 1983, 24, 1338. Int. Congr. Cataf., 1976 (The Chemical Society, London, 1977), vol. I , p. 261. 862. Paper 9/0 16 16C ; Received 17th April, 1989