J. Chem. SOC., Faraday Trans. I, 1984,80, 87-97 Interaction of Hydrogen Chloride with a Molybdena-Silica Catalyst BY S. RAZI SEYEDMONIR Department of Chemical Engineering, Pennsylvania State University, University Park, Pennsylvania 16802, U.S.A. AND RUSSELL F. Horn* Department of Chemistry, University of Auckland, New Zealand Received 18th April, 1983 The interaction of HCl with a molybdena-silica catalyst has been investigated using e.p.r. and infrared spectroscopy. Exposure of reduced catalysts to HCl at room temperature causes a large (up to 20-fold) increase in the amount of MoV detected by e.p.r. which is reversed on subsequent outgassing. The accompanying changes in g- and the 9 5 M ~ hyperfine-tensor components indicate that chloride ions replace oxide ions in the coordination sphere of MoV, and the increase in MoV spin concentration is attributed to substitution of bridging oxide ligands between magnetically coupled MoV ions.Oxidized catalysts are reduced by exposure to HC1. The reaction of HC1 with molybdena-silica is compared with that previously reported with molybdena-alumina catalysts. Supported molybdena catalysts are widely used for reactions such as hydrogenation of olefins and aromatics, polymerization, metathesis and oxidation of olefins, dehydrogenation and dehydrocyclization of paraffins, hydrodesulphurization and hydrodenitrogenation of petroleum fractions and conversion of synthesis gas to methane and Fischer-Tropsch products.'q Many characterization studies have been undertaken of such catalysts with the objectives of identifying the molybdenum- containing phases present in the catalysts, determining the coordination and valence states of molybdenum in oxidized and reduced forms of the catalysts and ultimately identifying the active sites for various reactions.E.p.r. spectroscopy is a technique which has been employed to follow the reduction of molybdenum from the + 6 to the + 5 valence state during catalyst activation. In the case of alumina-supported molybdena catalysts, the amounts of MoV measured by e.p.r. are significantly lower than those determined by chemical or X-ray photoelectron spectroscopic method^.^-^ This discrepancy was first attributed by Hall et al. to magnetic coupling between adjacent MoV ions on the catalyst surface, such that only isolated MoV ions contributed to the observed e.p.r.spectrum. We have recently presented direct evidence for coupling between MoV ions involving bridging oxide ligands on alumina-supported molybdena catalysts. Reaction of the reduced catalysts with gaseous hydrogen chloride or hydrogen bromide caused a 3-fold increase in the MoV spin concentration determined by double integration of the first- derivative e.p.r. signal. The accompanying changes in the MoV g and 9 5 M ~ hyperfine- tensor components indicated that halide ions were replacing oxide ions in the coordination sphere of MoV, reducing the magnitude of the MoV: MoV interactions and allowing previously invisible MoV to be detected by e.p.r. 8788 INTERACTION OF HCl WITH A MOLYBDENA-SILICA CATALYST The purpose of the present study was to investigate the magnetic interactions between MoV ions in silica-supported molybdena catalysts, using the reaction with hydrogen chloride as a probe. Silica-supported molybdena catalysts differ from the more widely studied molybdena-alumina catalysts in several important respects.In general, the interaction of molybdena with silica is much weaker than that with alumina.8-11 A 'free' MOO, phase is detected on silica at molybdenum loadings well below that corresponding to monolayer coverage of the support, whereas on alumina the MOO, phase appears only after the monolayer capacity of the support has been exceeded. The various molybdenum phases on both supports have similar structures, according to diffuse reflectance9 and laser Ramanlo studies, but temperature- programmed reduction measurements have shown that at least part of the molybdenum on alumina is more difficult to reduce than that on silica.12 Some differences in the extent and magnitude of MoV : MoV interactions on silica- and alumina-supported catalysts were thus anticipated.EXPERIMENTAL The molybdena-silica catalyst contained 6.2 % molybdenum by weight on silica gel (Davison grade 950, 700 m2 g-l) and was prepared by impregnation of the support with an aqueous solution of ammonium dimolybdate, drying at 423 K and calcination in air at 773 K for 18 h. A v5Mo-enriched catalyst containing 5.5% molybdenum by weight was prepared as follows: 9 5 M ~ metal (97% enriched, Oak Ridge National Laboratory) was dissolved in 50% nitric acid and the solution evaporated to dryness. The resulting precipitate was dissolved in 5% ammonium hydroxide at 340 K and again evaporated, followed by dissolution in distilled water and addition of the silica support.The catalyst was subsequently dried and calcined as above. Standard pretreatment of the catalysts involved heating in uacuo to 723 K followed by calcination in oxygen at this temperature and further evacuation. Reduced catalysts were prepared by subsequent exposure to hydrogen (100 Ton*) for 30 min at the desired temperature and further evacuation at this temperature for 30 min. Research grade hydrogen was purified by passage through a Pd-Ag thimble at 673 K. Anhydrous hydrogen chloride (Matheson) was subjected to repeated freeze-pumpthaw cycles before use.E.p.r. spectra were recorded (at room temperature unless otherwise stated) with a Varian El 15 spectrometer at 9.5 or 35 GHz. Catalyst samples were prepared in a Pyrex high-vacuum cell fitted with a quartz sidearm for e.p.r. measurements. Spin concentrations were determined by numerical double integrationI3 of the MoV signals and comparison with a Varian strong pitch sample which had in turn been calibrated against a fresh single crystal of CuSO, - 5H,O (using a dual sample cavity). Relative spin concentrations are considered to be accurate to & 10% and absolute spin concentrations to & 30%. Simulated e.p.r. spectra were obtained with the program ~ 1 ~ 1 3 . ' ~ Infrared spectra were recorded with a Nicolet MX1 Fourier-transform infrared spectrophotometer, using a conventional high-vacuum cell allowing in situ treatment of pressed wafers of catalyst.RESULTS Reduction of the molybdena-silica catalyst in hydrogen (100 Torr at 773 K for typically 30 min) produced an intense e.p.r. signal characteristic of MoV.15-17 Fig. 1 (a) shows this signal recorded at 9.5 GHz and 298 K. Identical line shapes were observed in spectra recorded at 77 and 20 K. The integrated intensity of the signal in this particular experiment corresponded to 8 x 10l8 spin g-l of the catalyst. Subsequent exposure to HCl (100 Torr, 298 K) caused no change in the colour of the reduced catalyst (dark grey), but the e.p.r. spectrum gave the new signal shown in fig. l(b). The changes in line shape resulting from HC1 addition were accompanied by an * 1 Torr = 101 325/760 Pa.S.RAZI SEYEDMONIR AND R. F. HOWE 89 Fig. 1. X-band e.p.r. spectra of reduced catalyst: (a) before exposure to HC1 and (b) after exposure to HCl. Relative spectrometer gains are indicated on each spectrum. Table 1. Integrated MoV signal intensities MoV spin concentration catalyst treatment /1Ols spin g-l oxidized catalyst reduced in H,, 773 K + HCl outgassed at: 300 K 373 K 473 K 673 K oxidized catalyst + HCI outgassed at: 300 K 373 K 473 K 0.3 8.2 186 77 38 30 9.1 1.7 1.7 1.8 1.7 approximately 20-fold increase in integrated signal intensity (relative spectrometer gain settings are indicated on each spectrum). The new signal appears to have an almost symmetric line shape at 9.5 GHz (which was unchanged on cooling to 77 K), but a spectrum measured at 35 GHz (not shown) revealed that the g tensor of the new signal is in fact orthorhombic, with principal components 1.967, 1.955 and 1.942.Subsequent outgassing of the catalyst, beginning at a temperature of 298 K and increasing in 100 K steps (30 min at each temperature), caused a gradual return to the original spectrum. Table 1 gives the measured signal intensities (expressed as spin concentrations) at several points in this experiment. A significant decrease in intensity90 INTERACTION OF HCl WITH A MOLYBDENA-SILICA CATALYST 100 G - g= 2.0028 '~ Fig. 2. X-band e.p.r. spectra of reduced g5Mo-enriched catalyst: (a) before exposure to HCI and (b) after exposure to HCl. Relative spectrometer gains are indicated on each spectrum. I 1 V I V g ='2.002 8 Fig 3.X-band e.p.r. spectra of reduced g5Mo-enriched catalyst exposed to HCl then evacuated at (a) 300 (b) 473 and (c) 773 K. Relative spectrometer gains are indicated on each spectrum.S. RAZI SEYEDMONIR AND R. F. HOWE 91 was caused immediately by evacuation at 300 K, and the original spin concentration and signal shape were restored after evacuation at 673 K. The spectra obtained at outgassing temperatures between 300 and 673 K showed a signal shape intermediate between those in fig. 1 ( a ) and (b). Spectra obtained from a similar experiment with a 95Mo-enriched catalyst are shown in fig. 2 and 3. The MoV signal of the reduced catalyst shows in this case 2 overlapping sets of 6 lines due to the parallel and perpendicular components of the 9 5 M ~ hyperfine 1 1 1 I I 1 M 11 Fig.4. X-band e.p.r. spectra of oxidized catalyst: (a) before exposure to HC1 and (b) after exposure to HCl. Identical spectrometer gains in both spectra (Mn indicates manganese impurity). tensor. (A similar spectrum of a 95Mo-enriched molybdena-silica catalyst has been published by Che el aZ.17) Addition of HCl to the reduced catalyst at room temperature gave the spectrum in fig. 2(b). The increase in signal intensity appears in the first- derivative spectra to be less than that observed for the normal catalyst, but double integration of the signals revealed that the increase in spin concentration is in fact of similar magnitude to that found for the unenriched catalyst (the increased line width of the new signal partially obscures the intensity increase).Fig. 3 shows a series of spectra recorded following subsequent outgassing at successively higher temperatures. Evacuation at 300 K significantly altered the signal shape [compare fig. 3 ( a ) with fig. 2(b)], and the original shape and intensity were completely restored after evacuation at 673 K. Catalysts given the standard pretreatment and not reduced in hydrogen are white in colour and contain molybdenum almost entirely in the +6 valence state. Fig. 4(a) shows an e.p.r. spectrum of an oxidized catalyst; the intensity of the weak residual MoV signal corresponds to a spin concentration of 3 x IOl7 g-l. Exposure of this sample to 150 Torr of HCl gave the spectrum shown in fig. 4(b) and caused a colour change to orange-yellow. The new signal has identical g-tensor components to that obtained from reduced catalysts treated with HCl, although noticeably better92 INTERACTION OF HCl WITH A MOLYBDENA-SILICA CATALYST Fig.5. X-band e.p.r. spectra of oxidized 95Mo-enriched catalyst: (a) exposed to HCI and evacuated at (b) 300, (c) 373, (d) 473 and (e) 773 K. Relative spectrometer gains are indicated on each spectrum. resolved. The MoV spin concentration increased approximately 6-fold. Subsequent outgassing did not in this case restore the original spectrum. Evacuation at 473 K gave a signal identical in shape to that of a catalyst reduced in hydrogen, without any change in integrated intensity, and no further changes occurred on outgassing at higher temperatures (table 1). Fig. 5 shows spectra obtained in a similar experiment with a 95Mo-enriched catalyst.S.RAZI SEYEDMONIR AND R. F. H O W 93 Exposure of the oxidized catalyst to HCl gave the spectrum in fig. 5(a). This shows 9 5 M ~ hyperfine components at identical positions to those for a reduced catalyst treated with HC1 [fig. 2(b)]; however, the signal shape is significantly different. Subsequent outgassing progressively converted the new signal to that of a reduced catalyst, without causing any change in spin concentration. Fig. 5 (d) illustrates that after outgassing at 473 K the new signal was almost completely removed. The effect of HC1 adsorption on the infrared spectrum of the oxidized catalyst was also investigated. Exposure to HCl at 298 K caused a significant change in the v(0H) region (3800-3400 cm-l).Intense new bands appeared at 3720 and 3560 cm-l, superimposed on the existing bands at 3747 and 3650 cm-l due to hydroxyl groups of the silica support. Similar changes in the v(0H) region could be induced by exposing the oxidized catalyst to water vapour at 298 K. In this case, however, a weak band also appeared at around 1624 cm-l, whereas the 1624 cm-l band was not detected following exposure to HC1. Outgassing above 473 K restored the original spectrum in the v(0H) region. In the region 800-1000 cm-l the infrared spectrum of the oxidized catalyst shows three bands due to v(Mo-0) vibrations of the MOO, and surface molybdate phases.18 Exposure to HCl caused the complete removal of all three bands. On subsequent outgassing the Mo-0 bands were only partially restored.DISCUSSION Exposure of reduced catalysts to HCl caused significant changes in both the intensity and e.p.r. parameters of the MoV signal. The g- and 9 5 M ~ hyperfine-tensor components of the observed signals are listed in table 2. The g-tensor components of the new signal were obtained by inspection from the Q-band spectrum. Two of the 9 5 M ~ hyperfine-tensor components of the new signal were obtained by inspection from the X-band spectrum of the 95Mo-enriched catalyst [fig. 2 (b)] and the third determined by computer simulation. Fig. 6 (b) shows a computer-simulated spectrum generated with the parameters listed for the new signal in table 2. Agreement between the observed and calculated peak positions was found to be extremely sensitive to the val- ues given to the 9 5 M ~ hyperfine-tensor components.The perpendicular components (Azz and A,,), which are not resolved in the experimental spectrum, were adjusted to give the best fit of the experimental peak positions. Note, however, that the simulated spectrum in fig. 6(b) does not match at all the observed intensity distribution [fig. 2 (b)]. The observed intensity distribution could be satisfactorily simulated only by adding a second signal to the simulation, as illustrated in fig. 6. The composite spectrum in fig. 6(a) [which should be compared with fig. 2(b)] was obtained by adding the simulated spectrum in fig. 6(b) to the broad symmetric signal in fig. 6(c), in the ratio 1 : 9. The parameters of the broad signal and its quantitative contribution to the total integrated intensity cannot be accurately determined from the simulations, since the broad signal could not be observed independently, and the intensities of simulated signals are extremely sensitive to the line widths and line-shape functions employed.For the purposes of simulation the broad signal was generated with the following parameters: g, = 2.030, g, = 1.952, g, = 1.870 and Lorentzian linewidths of 80 G. Nevertheless, the increase in total integrated signal intensity occurring on exposure to HCl must be attributed largely to the contribution from the broad signal. Comparison of the e.p.r. parameters of the signals observed before and after exposure of reduced catalysts to HCI (table 2) reveals that the g-tensor components are reversed by HCI treatment.The normal MoV signal obtained from thermally re- duced catalysts has g,, (associated with the largest 9 5 M ~ hyperfine-tensor component)94 INTERACTION OF HCl WITH A MOLYBDENA-SILICA CATALYST Table 2. E.p.r. parameters of MoV signals catalyst or A x x l A y y l A z z / compound ref. g x x g y p g,, cm-l cm-l lop4 cm-l reduced molybdena- this work 1.947 1.947 1.892 41 41 85 (19) 1.940 1.940 1.882 41 41 91 silica reduced molybdena- this work 1.954 1.944 1.967 36 30 73 oxidized molybdena- this work 1.954 1.944 1.967 36 30 73 silica + HCl silica + HCl alumina + HCl reduced molybdena- (7) 1.951 1.943 1.962 34 34 76 MoOC1,2- (19) 1.94 1.94 1.963 33 33 75 MoOC14- (19) 1.950 1.950 1.967 35 35 73 Fig. 6. Simulated e.p.r. spectra: (Q) superposition of signals (b) and (c) in the ratio 1 : 9; (b) signal simulated with the parameters given for the HC1-treated catalyst in table 2, and Lorentzian line widths of 12.5 G; (c) broad signal simulated with the parameters described in the text.less than g,, and gyy, whereas this order is reversed in the HCl-treated catalysts. A similar reversal of the g-tensor components was observed previously for alumina- supported catalysts treated with HCl,' and is commonly found in coordination compounds of MoV containing chloride ligands. l9 Manoharan and Rogers20 first attributed the g-tensor reversal to spin-orbit coupling from the ligandsS. RAZI SEYEDMONIR AND R. F. HOWE 95 [A(Cl) = 586 cm-l compared with A ( 0 ) = 152 cm-l], although it has been suggested more recently that the dominant g-shift mechanism in chloromolybdenum complexes involves low-lying charge-transfer states rather than ligand spin-orbit coupling.21 The evidence against the spin-orbit coupling model is that no correlation exists between g,, and the number of chloride ligands coordinated to Mo,.The reversal of the g tensor of MoV on silica following treatment with HCl is thus due to replacement of oxide ions in the coordination sphere of molybdenum by chloride ions, but the number of chloride ligands cannot be determined from the e.p.r. parameters. The changes in the e.p.r. spectra of reduced catalysts following exposure to HCl indicate that two processes are occurring: ligand substitution around existing paramagnetic MoV ions [reaction (l)] and formation of a new paramagnetic species responsible for the broad symmetric signal.The second process is attributed to the ‘ uncoupling’ of magnetic interactions between adjacent MoV ions through removal of bridging oxide ligands [reaction (2)], as described previously for molybdena-alumina catalysts : (1) (2) No detailed chemical measurements of the MoV concentrations in silica-supported catalysts are available, but X.P.S.~~ and reduction-isothermg data suggest that the total amount of MoV in reduced catalysts considerably exceeds the 3% of the total Mo content observed by e.p.r. prior to HCl treatment. Rapid spin-lattice relaxation of the ‘missing’ MoV species at 77 K or above can be ruled out, since no new signals were observed at 20 K. In particular the second MoV signal reported by Kazanski et aZ.15 and also observed by Che et aZ.16 in catalysts prepared from MoCl, was not detected here.The missing MoV ions must therefore be strongly magnetically ,coupled, evidently via bridging oxide ligands, since reaction with HCl reduces the extent of coupling. The uncoupled MoV ions are still in close proximity, and their e.p.r. signal should be strongly dipolar broadened, as is observed. The increase in MoV spin concentration in reduced catalysts on treatment with HC1 cannot be due to reduction of remaining MoV1, since outgassing restored the original spin concentration. The reversal of reactions (1) and (2)’ i.e. reaction of H,O with halide ligands to restore the original oxomolybdenum species, accounts for the reversibility of the spectral changes in reduced catalysts.The temperature range over which the original spectra are restored on outgassing corresponds to that needed to remove adsorbed water from the catalyst. In contrast, the increase in MoV spin concentration observed on exposure of oxidized catalysts to HC1 was not reversible (table l), and the spectra obtained contained no contribution from the broad symmetric signal. This is particularly clear from the 95Mo-enriched catalyst; the observed signal [fig. 5(a)] could be simulated with the parameters of a single MoV signal [fig. 6(b)]. Since the oxidized catalyst contains molybdenum almost exclusively in the + 6 valence state, the increase in MoV concentration must in this case be due to a redox reaction [reaction (3)]: MoVO, + 2y HCl = M O ~ O , - ~ C12y + y H,O MoVO MoV + 2 HCl = MoVCl C1 MoV + H20.MoV1O + 2HC1= MoVCl + H,O + iC12. (3) MoVr in molybdena-alumina catalysts is known to be capable of oxidizing polynuclear aromatic compounds to the corresponding radical cations.23 The MoV species produced in reaction (3) are magnetically isolated, giving a single well resolved e.p.r. signal. On subsequent outgassing the chloride ligands are replaced by oxide ions without any change in oxidation state and e.p.r. signal intensity. The extent of reduction achieved by treating oxidized catalysts with HCl is low; the MoV spin concentration (table 1) corresponds to 0.5% of the total molybdenum content, and96 INTERACTION OF HCl WITH A MOLYBDENA-SILICA CATALYST no coupled MoV species are produced. The replacement of chloride ligands by oxide on outgassing is consistent with the observations of Che et all6* l7 that silica-supported molybdenum catalysts prepared from MoCl, and (NH,),MoOCl, give MoV signals identical to that of a conventional catalyst prepared from chlorine-free precursors.The infrared spectra of oxidized catalysts exposed to HCl support the above description of reduction and ligand substitution. Exposure to HCl completely removes the 3 bands due to Mo-0 stretching vibrations.ls This does not necessarily mean that all oxide ions in the catalyst are replaced by chloride, since adsorption of water on the oxidized catalyst also causes a significant reduction in intensity of the Mo-0 bands,24 and Mo-Cl stretching vibrations lie outside the frequency range of the MX1 spectrophotometer used.The effects of chloride substitution and adsorption of water as a reaction product cannot be separated clearly from the infrared result. On subsequent outgassing, the infrared spectrum resembles that of a catalyst slightly reduced in hydrogen, consistent with the e.p.r. observations. Although the reactions of HCl with molybdena-silica catalysts resemble in many respects those with molybdena-alumina, some significant differences also exist between the two systems. The increase in MoV spin concentration obtained by treating reduced molybdena-silica catalysts with HCl is much larger than the threefold increase found with molybdena-alumina,' which suggests that a larger fraction of the MoV produced by reduction is magnetically coupled on silica.The ligand substitution reaction in reduced catalysts is more readily reversed on silica and the fully oxidized silica- supported catalyst is reduced to a lesser extent on exposure to HCl. These differences may reflect differences in the distribution of molybdenum between monomeric and polymeric species on the two supports. This work was undertaken in part in the Chemistry Department of the University of Wisconsin-Milwaukee. Partial support by the Laboratory for Surface Studies is gratefully acknowledged. F. Massoth, Ado. Catal., 1978, 27, 265. G. C. A. Schuit and B. C. Gates, AIChE J., 1973, 19, 417. W. K. Hall and M. LoJacono, Proc. 6th Int. Congr. Catal., London, 1976, ed. G. C. Bond, P. B. Wells and F. C. Tompkins (The Chemical Society, London, 1976), p.246. T. A. Patterson, J. C. Carver, D. E. Layden and D. M. Hercules, J. Phys. Chem., 1976, 80, 1700. L. Petrakis, P. L. Meyer and T. P. Debies, J. Phys. Chem., 1980, 84, 1020. S. Abdo, R. B. Clarkson and W. K. Hall, J. Phys. Chem., 1976, 80, 2431. 'I S. Abdo, A. Kazusaka and R. F. Howe, J. Phys. Chem., 1981,85, 1380. P. Gajardo, P. Grange and B. Delmon, J. Phys. Chem., 1979, 83, 1771. P. Gajardo, D. Pirotte, P. Grange and B. Delmon, J. Phys. Chem., 1979, 83, 1780. lo H. Jeziorowski, H. Knozinger, P. Grange and P. Gajardo, J. Phys. Chem., 1980, 84, 1825. l1 J. Medema, C. van Stam, V. H. J. de Beer, A. J. A. Konings and D. C. Koningsberger, J. Catal., 1978, 53, 386. l2 R. Thomas, M. C. Mittelmeijer-Hazelberger, F. P. J. Kerkhof, J. A. Moulijn, J. Medema and V. H. J. de Beer, Proc. 3rd Inr. Con5 Chemistry and Uses of Molybdenum, 1979, ed. P. C. H. Mitchell and H. F. Barry (Climax Molybdenum Co, Ann Arbor, Michigan, 1979), p. 85. l3 P. B. Ayscough, Electron Spin Resonance in Chemistry (Methuen, London, 1967). l4 G. P. Lozos, B. M. Hoffman and C. G. Franz, Quantum Chemistry Program Exchange, 1973, 265. l5 V. M. Vorotyntshev, V. A. Shvets and V. B. Kazanskii, Kinet. Catal., 1971, 12, 1108. l6 M. Che, F. Figueras, M. Forissier, J. C. McAteer, M. Perrin, J. L. Portelfaix and H. Praliaud, Proc. 6th Int. Congr. Catal., London, 1976, ed. G. C. Bond, P. B. Wells and F. C. Tompkins (The Chemical Society, London, 1976), p. 261. M. Che, J. McAteer and A. J. Tench, J. Chem. Soc., Faraday Trans. I , 1978, 74, 2378. l8 S. R. Seyedmonir, S. Abdo and R. F. Howe, J. Phys. Chem., 1982, 86, 1233. lo M. Che, M. Fournier and J. P. Launay, J. Chem. Phys., 1979, 71, 1954.S. RAZI SEYEDMONIR AND R. F. HOWE 97 2o P. T. Manoharan and M. T. Rogers, J . Chem. Phys., 1968,49, 5510. z1 M. I. Scullane, R. D. Taylor, M. Minelli, J. T. Spence, K. Yamanouchi, J. H. Enemark and N. D. ** M. B. Ward, M. J. Lin and J. H. Lunsford, J . Catal., 1977, 50, 306. 23 C . Naccache, J. Bandiera and M. Dufaux, J . Catal., 1972, 25, 334. 24 S. R. Seyedmonir, Ph.D. Thesis (University of Wisconsin-Milwaukee, 1982). Chasteen, Inorg. Chem., 1979, 18, 3212. (PAPER 3/618)