J. CHEM. SOC. FARADAY TRANS., 1994, 90(8), 1177-1182 Activation of Surface Lattice Oxygen in the Oxidation of Carbon Monoxide on Silica Yasuyuki Matsumura and John B. Moffat* Department of Chemistry and Guelph- Waterloo Centre for Graduate Work in Chemistry, University of Waterloo, Waterloo, Ontario, Canada N2L 3G1 Keiji Hashimoto Osaka Municipal Technical Research Institute, Joto-ku, Osaka 536,Japan Silica prepared by the sol-gel method from ethyl orthosilicate catalyses carbon monoxide oxidation at 850 K or above while no such activity is found with a commercial silica prepared from inorganic reagents. The activity of the silica prepared by the sol-gel method is increased by repetition of the carbon monoxide oxidation reaction. Labelled reactions with '*O, provide evidence to suggest that most of the lattice oxygen species on the surface of the silica participate in the reaction.After evacuation at 950 K or above, radical species identified as Si-0- -0--Si are generated on the silica. Si-0--Si radical species are apparently formed during the reaction with the radical electron originating from the Si-0-0--Si species in the initial stage of the reaction. Adsorp- tion of carbon monoxide and oxygen on the Si-0--Si species is proposed to cause formation of carbon dioxide and 0-which will react with carbon monoxide and provide the radical electron to the lattice oxygen species, that is, Si-0-Si and Si-0-0-Si. The present work shows that the lattice oxygen of silica can be involved in oxidation if the oxygen is activated by radical species such as 0-, which have been proposed as active species in catalytic oxidation processes.Inorganic substances with high surface areas are frequently employed as supports for catalytically active materials to increase the number of active sites and improve the catalytic properties. Hence, clarification of the effect of the support is of great importance in the design of good catalysts. Amorp- hous silica is a representative support with high surface area and is not infrequently used in studies of oxidation processes, such as the partial oxidation of methane.'-20 It is known that not only catalytic activity but also product selectivity often changes with the nature of the support, especially in methane oxidation where the contribution of the support to the reac- tion is not yet understood2'.'' although the structure of amorphous silica is well known.23 Although there are no effectively strong acid-base sites on its ~urface,'~-~~ silica itself often functions as a ~atalyst.~,~,~~-~~Work in this laboratory and elsewhere has shown that silica can catalyse the oxidation of methane; however, the products generated depend on the oxidant ~~ed.~*~*~~-~~ The product distribu- tion from the oxidation of methane varies with the sample of silica even though the samples are apparently similar in com- p~sition.~,~~-~~Alth ough the surface area of silica has been suggested as an important factor in its catalytic activity,29 changes in the product distribution cannot be accounted for by changes in surface area alone and thus special sites for oxidation apparently exist or are formed during the reaction.Silica prepared by the sol-gel method from ethyl orthosili- cate yields a high selectivity to carbon dioxide in the oxida- tion of methane compared with a commercial silica prepared from inorganic compounds, although the surface areas of the products prepared by these two methods are similar.32 Hence there are features of these two catalysts which are distinguish- able and the identification of such differences may assist in the clarification of the mechanism of oxidation. In the present paper, carbon monoxide oxidation, which is stoichiometrically one of the simplest oxidation processes, has been studied on silica prepared by these two methods to provide information on the differences between these two materials.Somewhat surprisingly, almost all of the lattice oxygen atoms on the surface of silica prepared by the sol-gel method appear to be reactive to carbon monoxide. The lattice oxygen is probably activated by radical oxygen species such as 0-,which has often been suggested as the active oxygen species in oxidation processes. Thus, the lattice oxygen on silica employed as a support for oxidation cata- lysts apparently is an active participant in the oxidation process. Experimental 0.2 mol of ethyl orthosilicate [Kanto Chemical Corp. (G.R. Grade)] was hydrolysed to silica gel with 0.05 dm3 of 5 mol dm-3 nitric acid solution.The hydrogel obtained was dried at CQ. 400 K in air, washed with water and heated at 720 K in air for 5 h in order to remove residual hydrocarbons or nitric acid. Analysis by atomic absorption spectroscopy showed that this sample (denoted Si02-0) has no observable impu- rities. For comparison purposes, a commercial silica sample prepared from inorganic compounds (Cariact- 15) was obtained from Fuji-Davison Chemical Ltd. The BET surface areas of these samples in powdered form were 245.1 and 177.9 m2 g- ',respectively. The oxidation of carbon monoxide was carried out in a static reactor of dead volume ca. 0.1 dm3 equipped with a Baratron pressure gauge which enabled continuous recording of the pressure in the reactor.The catalysts were held in a quartz tube of 6 mm inner diameter. No reaction was observed without the catalyst even at 950 K. The reactant gas contained carbon monoxide and oxygen in a molar ratio of 2 : 1. Tests showed that the reactant gases were well dispersed in the reactor below a pressure of 100 Pa. The reaction pro- ducts were analysed with a mass spectrometer (ULVAC MSQ-150A). Labelled oxygen (' *02, 98.58%) was obtained from Sen Saclay and was used without further purification. Electron paramagnetic resonance (EPR) spectra were recorded with a JEOL JES-RE2X spectrometer at 9.15 GHz (modulation, 100 kHz, 0.02 mT; time constant, 1 s) usually at 100 K. The microwave power was kept at 0.01 mW to prevent saturation of the spectra.The g value was determined with diphenylpicryl hydrazyl (DPPH, g = 2.0036). The silica 1178 sample (0.05 g) was placed in a quartz EPR tube in which pretreatment and activation of the sample was carried out. Surface analysis by X-ray photoelectron spectroscopy (XPS) was carried out using a Shimadzu ESCA 750 spec- trometer. The sample was mounted with adhesive tape in air just after the pretreatment or activation and set into the spec- trometer. Charge correction of the XPS data was accom-plished by assuming that the binding energy of the C 1s peak is 284.6 eV. IR spectra were recorded with a Nicolet 5DX FTIR spec- trometer. The silica sample (0.01 g) was pressed into trans- lucent self-supporting wafers and placed into an IR cell, equipped with KBr windows, which allowed pretreatment in uucuo and activation of the sample in a separate portion of the cell assembly.Results Oxidationof Carbon Monoxide The experiments involving the oxidation of carbon monoxide were generally performed, under a given set of conditions, in a sequential fashion. Carbon monoxide and oxygen were introduced to Si0,-0 preheated in vacuo (ca. low3Pa) at 950 K for 0.5 h. Although the activity of the catalyst evac- uated at 950 K for 0.5 h was very low in the first carbon monoxide oxidation [Fig. l(a)], the catalyst preheated at 950 K for 5 h in uucuo exhibited significantly higher activity in the initial experiment [Fig. l(b)]. When the catalyst was preheat- ed at 1050 K for 0.5 h in vacuo, the reaction rate increased abruptly after an induction period of 1 min [Fig.l(c)]. Evac- uation at 950 K for 0.5 h followed by preheating at 950 K for 0.5 h under carbon monoxide (50 Pa) or oxygen (25Pa) did not activate the catalyst. On the other hand, Cariact-15 showed a slight activity at 950 K even after evacuation at 1050 K for 0.5 h or after several experiments following pre- heating at 950 K for 0.5 h in uucuo (not shown). Following activation by subjecting the catalysts to use in a minimum of six carbon monoxide oxidation steps at 950 K, Si0,-0 catalysed the reaction even at 850 K (Fig. 2). The reaction rate was initially slow, although subsequently the rate increased abruptly. For example, the initial rate of carbon dioxide formation was 0.02 mmol min-' g-' at 950°C and the rates at 5 and 10% conversion were 0.06 and 0.07 mmol min-' g-', respectively.The rate at a conversion higher than ca. 10% decreased discernibly with decrease in h ,\" 40 v C .-E > 2 20 0 0 1 2 3 ti me/mi n Fig. 1 Initial activity of silica prepared by the sol-gel method in the oxidation of carbon monoxide. Catalyst : SO,-0, 0.020 g, evac- uated at (a)950 K for 0.5 h, (b) at 950 K for 5 h, (c) at 1050 K for 0.5 h. initial pressure of carbon monoxide, 50 Pa; oxygen, 25 Pa. Reac- tion temperature, 950 K. J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 2o0 L0 2 4 6 8 ti me/min Fig. 2 Transformation of carbon monoxide to carbon dioxide over silica prepared by the sol-gel method.Catalyst; Si0,-0, 0.0052 g. Initial pressure of carbon monoxide, 50 Pa; oxygen, 25 Pa. Prior to the runs shown here, the catalysts were activated by more than six runs at 950 K. Reaction temperatures of (a) 850, (b) 900 and (c) 950 K are shown. the pressure of the reactant. The apparent activation energy was calculated as ca. 90 kJ mol-' from Arrhenius plots for the rate at 10% conversion. A liquid-nitrogen trap was usually employed to remove carbon dioxide from the reac- tion gas, but no significant change was observed without the trap. The reaction proceeded stoichiometrically while a small amount of hydogen was detected in the reaction products of the initial two reactions. No reaction was observed without oxygen gas at 950 K (Pco = 50 Pa) over the activated cata- lyst.When Si0,-0 was heated with oxygen (Po2 = 25 Pa) for 0.5 h following preheating at 950 K for 0.5 in uacuo, no formation of carbon oxides was observed. Oxidation of Carbon Monoxide with Labelled Oxygen In order to obtain further information on the reaction, carbon monoxide oxidation on Si0,-0 was studied with 1802 contained in the reactant mixture. The catalyst was pre- heated in uucuo for 0.5 h at 950 K. The reaction was repeated on the same aliquot of catalyst after evacuation at 950 K for 0.2 h after each run. The internal pressure was <0.1 Pa when the reaction system was isolated before the reaction. The rate of carbon dioxide production during the first experiment was very low but increased with successive experiments up to a run number of 8 (Fig.3). A small amount of hydrogen was detected in experiments 1 and 2. Note that a considerable quantity of Cl60, was formed in the reaction while the quantity of 1602 in the reactant gas was very small. The accumulated amount of C60,in runs 1-10 was 1.0 mmol g-' and that of Cl8O, and Cl8O was 0.3 mmol g-'. No formation of dioxygen containing l6O was observed in the runs in which 1802was employed as an oxidant. EPR Spectra After evacuation of Si0,-0 at 950 K for 0.5 h, weak but sharp peaks were observed at 326-327 mT in the EPR spec- trum [Fig. 4(a)], and the peak intensity was decreased to ca. 1/3 of the initial intensity after introduction of the reactants (carbon monoxide, 50 Pa; oxygen, 25 Pa) at 950 K for 0.1 h (not shown).The intensity of EPR peaks increased after the next reaction at 950 K and the intensity of the peaks was 4/3 of that in Fig. 4(a) (not shown). The EPR peaks were signifi- cantly intensified with the third reaction at 950 K [Fig. 4(b)]. Before the measurement the sample was evacuated at room J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 0.10. r I n cn 0.08, I .-C E 5 0.06 E E.-. 0.04 C .-4-;0.02 5+ 0 1 2 3 4 5 6 7 8 9 10111 run number Fig. 3 Formation rates for products of the reaction between CO and 1802 over silica prepared by the sol-gel method. Catalyst: Si0,-0, 0.0047 g. Initial pressure of carbon monoxide, 50 Pa; oxygen, 25 Pa.Reaction temperature, 950 K. The rates were deter- mined by division of products yield by period of the reaction; each run was stopped at conversions below 60%. Solid bar, Cl6O,; cross hatched bar, C'60180; hatched bar, CEO,; open bar, C'*O. Run 11 was carried out with 1602 instead of "02. temperature for 0.1 h. Each reaction was carried out after evacuating at 950 K for 0.2 h following the previous reaction. No significant change in the EPR spectrum was observed on the sample either with or without evacuation at 950 K for 1 h after the reactions. The peaks were also evident in the EPR spectrum obtained at 450 K [Fig. qc)], although the shape of the peaks was broader than that observed at 100 K. The peak intensity increased by a factor of ca.two when the sample was evacuated at 1050 K for 0.5 h compared with the sample evacuated at 950 K for 0.5 h, but there was no significant change between the samples evacuated at 950 K for 0.5 h and 5 h (not shown). No peaks were observed in the spectrum for the sample evacuated at 750 h for 1 h. The shape of the spec- trum was reproduced by simulation using a Lorentzian func- tion [the result is shown in Fig. 4(b) with open circles]. Simulation using a Gaussian function was unsuccessful. The values for g1 and gI1were determined as 2.0023 and 2.0027, respectively, by simulation using a Lorentzian function (linewidth for gl, 0.025 mT; gI1, 0.035 mT). The radical species detected was so stable that the same EPR spectrum was observed after introduction of carbon monoxide, oxygen, hydrogen or air to the silica sample at room temperature, although heating the sample with one of these gases, with the exception of hydrogen, at 950 K for 0.2 h resulted in dimin- ishing the peaks.The EPR peaks appeared again after evac- uation at 950 K for 0.5 h following contact with oxygen or carbon monoxide at 950 K but the peak intensity was similar to that observed in the spectrum of the sample evacuated at 950 K for 0.5 h. No such EPR peaks were detected with Cariact- 15. XPS Analyses In order to obtain information on the air-stable radical species detected by recording EPR spectra, XPS analyses of the silica samples were performed. In the spectrum for Si0,-0 evacuated at 950 K for 0.5 h, the main peaks for Si 2p and 0 1s were present at 104.1 and 533.6 eV, respectively, although small shoulders were also detected [Fig.5(a)]. After three consecutive reactions with 50 Pa of carbon monoxide and 25 Pa of oxygen at 950 K followed by evacuation at 950 K for 0.2 h to ca. Pa, the spectrum for the sample showed a definite shoulder at 101 eV for Si 2p and a small shoulder at 532 eV for 0 Is, while no shift was observed with the main peaks [Fig. 5(b)]. IR Spectra After evacuation of Si0,-0 at 950 K for 0.5 h, the IR absorption band attributed to hydroxy groups was clearly observed at 3747 cm-' (not shown).33 Three consecutive reactions with 50 Pa of carbon monoxide and 25 Pa of oxygen (160,) at 950 K were carried out over the sample.After the final reaction the sample was cooled and evacuated at room temperature for 0.5 h while evacuation at 950 K for 0.2 h was carried out subsequent to the first and second reac- tions. No absorption bands in the range of 1400-4000 cm- attributed to adsorption species were observed after the final reaction and the intensity of the band at 3747 cm-' was almost the same as observed just after the pretreatment. 0 1s Si 2p -1 I I I 326.0 326.2 326.4 326.6 326.8 magnetic field/mT Fig. 4 EPR spectra of the silica prepared by the sol-gel method recorded at 100 K: (a) evacuated at 950 K for 0.5 h; (b)activated by oxidation of carbon monoxide at 950 K three times; (c) recorded at 450 K. Open circles are the results of computer simulations.540 535 530 110 105 100 95 binding energy/eV Fig. 5 XPS bands of the silica prepared by the sol-gel method for 0 1s and Si 2p: (a)evacuated at 950 K for 0.5 h; (b) activated by oxidation of carbon monoxide at 950 K for three times. The inten- sities of the spectra were normalized by using the atomic sensitivity factors of Si 2p (0.17) and 0 1s (0.63). Model and Results of Molecular Orbital Calculations In order to obtain further information on the nature of the active sites, ab initio Hartree-Fock calculations (Gaussian-86)34 were performed using the STO-3G basis set on the model compounds for the surface species of silica. The models used were (HO,),SiOSi(O,H), (labelled SiOSi), (HO,),SiO -Si(O,H), (labelled SiO -Si), (HO,),SiOOSi(O,H), (labelled SiOOSi), and (H0,),SiOO-Si(0,H)3 (labelled SiOO- Si) (Fig.6). Geometry optimization was performed on the Si-0-Si and Si-0-0 angles and the Si-0 and 0-0 bond lengths. The remaining parameters were held constant (Table 1). The internuclear separation between the two Si atoms in model SiOSi was calculated, after optimization, as 0.302 nm while that in model SiO-Si was 0.342 nm (Table 2). On the actual surface of silica, movement of the silicon atoms is restricted by the surrounding atoms. Hence, the SiO-Si model SiOSi model SQSi Fig. 6 Model compounds for Si-0-Si and Si-0-0-Si oxygen bridges Table 1 Structural parameters fixed during optimization bond lengthlnm angleldegrees Si-0, 0.163 0-Si-0, 109.5 0,-H 0.960 Si-0,-H 115.0 Table 2 Structural parameters optimized" bond length/nm andeldegrees model Si-0 Si-Si 0-0 Si-0-Si Si-0-0 SiOSi 0.159 0.302 144.5 SiO-Si 0.184 0.342 136.9 SiO-Si(S) 0.185 (0.302) 109.4 SiOOSi 0.168 0.396 0.141 104.8 SiOO-Si 0.163 0.345 0.197 78.1 SiOO-Si(S) 0.164 (0.396) 0.190 96.2 ~~~~ ~ Values in parentheses were fixed during optimization.J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 radical species would be expected to have a structure similar to that of SiOSi. To provide a more realistic model for the radical species, only the position of the bridge oxygen in SiO-Si was optimized and the positions of the other atoms were fixed as in the model SiOSi. The bond length of Si-0 in this model, SiO-Si(S), was approximately equal to that in SiO-Si but significantly longer than that in SiOSi (see Table 2).The total energy of SiO-Si(S) was almost the same as that of SiO-Si (Table 3). The internuclear separation between the two Si atoms in model SiOOSi was 0.396 nm, significantly longer than that in model SiOO-Si (0.345 nm). Hence, a calculation was carried out with model SiOOSi(S) in which only the position of the bridge oxygen atoms was optimized and the positions of the remaining atoms were fixed. The cal- culated bond lengths of Si-0 and 0-0 were very close to those of SiOO-Si (see Table 2). The results of the calcu- lations are summarized in Table 3. The binding energies were obtained from the orbital energies corresponding to the 0 1s and Si 2p orbitals.The atomic spin densities for Si and 0 in model SiO- Si(S) were calculated as 0.33 and 0.80, respec-tively, while those in model SiOO-Si(S) were -0.07 and 0.65, respectively. Discussion Reaction Mechanism of Carbon Monoxide Oxidation In the experiments with labelled oxygen as oxidant significant quantities of C160, (1.0 mmol accumulated in runs 1-10 of Fig. 3) are produced. Since the catalyst is evacuated at 950 K to <0.1 Pa prior to the reaction, the quantity of adsorbed species on the surface is expected to be small and the number of l60included in these residual species should be negligible. Although the formation of Cl6O, could result from the dis- proportionation of carbon monoxide, the total amount of C1802 and Cl80 was only 0.3 mmol and is considerably smaller than that of C1602.In addition, the amount of PO, and Cl80 formed in each run increases with run number and is not proportional to the amount of C160,. Thus, the preponderance of the l60which participates in the oxidation of carbon monoxide evidently originates from the lattice oxygen of Si0,-0. The number of oxygen atoms on the surface of Si0,-0 is estimated as 6 mmol g- based on its surface area and the density of silica; therefore, approx- imately 12% of the lattice l60atoms on the surface are esti- mated to be consumed in the labelled reaction. With the last run (run 11 in Fig. 3), in which the reactant gas did not contain 1802, 8% of the carbon dioxide produced was C160180,showing that the lattice oxygen on the surface of Si0,-0 is replaced with l80.Since dioxygen containing '*O is not formed in the previous runs, the surface oxygen replacement is not the result of exchange between the lattice oxygen and gas-phase oxygen but rather occurs during the oxidation of carbon monoxide.In the early runs in which the Table 3 Calculated parameters for oxygen bridges binding energy/eV charge atomic bond population model energy/eV 0 1s Si 2p 0 Si Si-0 0-0 SiOSi -29707.0 548 111 -0.69 + 1.42 0.282 -SiO -Si -29697.9 539 103 -0.49 + 1.06 0.063 -SiO -Si(S) -29697.9 540 103 -0.44 + 1.03 0.055 -SiOOSi -31714.5 55 1 112 -0.35 + 1.40 0.224 SiOO- Si -31708.8 539 105 -0.53 + 1.30 0.241 -0.001 SiOO-Si(S) -31708.5 539 105 -0.52 + 1.28 0.202 O.OO0 J.CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 number of the lattice lS0on Si0,-0 is considered to be small, ca. one half of the carbon dioxide produced is Cl60, while the oxidant was "O,, suggesting that oxidation of one carbon monoxide with lattice oxygen accompanies oxidation of another carbon monoxide with gas-phase oxygen. Since one half of the carbon dioxide produced evidently results from the reaction between carbon monoxide and the lattice oxygen, it is estimated that 16% of the lattice oxygen atoms which can react with carbon monoxide are replaced with ''0 in the catalyst of the last run. Although the number is higher than the number of surface lattice l60consumed in the labelled reactions (12%), the numbers are approximately identical; therefore, it appears that most of the lattice oxygen atoms on the surface of Si0,-0 participate in the oxidation process.Since lattice oxygen atoms on the surface of silica are gen- erally considered to be inactive, an activation process appar- ently occurs during the reaction. The oxygen atom in carbon monoxide is not exchanged with one from either 0, or the surface of silica because in the labelled reaction formation of C"0 is not accompanied by formation of oxygen molecules containing l6O. Furthermore, no Cl80 is formed by intro- duction of Cl60 and I6O2 on Si0,-0 which contains an appreciable amount of l80as lattice oxygen. This implies that surface atomic oxygen species such as 0-,which readily react with carbon monoxide,' 5*36 are present.The adsorption species appear to be consistent with the product distribution. Assuming that the atomic oxygen species is 0-,an electron must remain on the surface of silica after the reaction between 0-and carbon monoxide to form carbon dioxide. The atomic bond population of Si-0 in Si-O--Si is cal- culated as 0.06, significantly lower than in Si-0-Si (see Table 3), suggesting that the Si-O--Si group, if extant, will be reactive. Hence the reactions : CO + 0-+ Si-0-Si -+ CO, + Si-O--Si (1) CO + 0, + Si-O--Si --r Si-OCO + O,--Si (2) Si-OCO + O,--Si -+ CO, + 0-+ Si-0-Si (3) are expected to occur. In these reactions, one molecule of carbon monoxide reacts with a lattice oxygen and another reacts with oxygen originating from dioxygen.Radical Species on the Surface Since the number of the radical species on Si0,-0 is decreased by contact with carbon monoxide or oxygen at 950 K, the generation of the radical is not due to reduction or oxidation of the surface. The quantity of residual carbon in the silica sample is believed to be very small since carbon oxides were not observed after heating the sample in oxygen. No regeneration of the EPR peaks is expected after heating the sample with oxygen if the residual carbon is responsible for the EPR peaks. Since the EPR peaks appear again after evacuation of the sample at 950 K, residual carbon is evi- dently not the radical site. It is known that y-irradiation of silica can result in the cleavage of Si-0 bonds and consequently the generation of some radical centre^.'^'^^ Griscom et al.reported that annealing of phosphorus-doped silica glass at ca. 800-1100 K generates S centres which are attributed to E' type defects such as (OSi,)Si' and/or (0,Si)Si' irrespective of the irradia- ti~n.~~The centre gives a sharp EPR singlet at a g value of 2.0030. However, no such EPR peak was observed in the spectrum for silica prepared by the sol-gel method from ethyl orthosilicate after annealing at 850 K followed by y-irradiati~n.~'Moreover, the lineshape for the S centre is not L~rentzian.'~The EPR peaks for the E' centre are broad and 1181 also not L~rentzian.~~,~'Thus, the radical species in Si0,-0 cannot be attributed to these radical centres.Since the binding energies for the shoulders attributed to radical species in the XP spectra are lower than those for the main peaks (see Fig. 5), the silicon and oxygen atoms constituting the radical species are presumably more negative than those on the usual silica ~urface.~' This suggests that electrons are captured by the precursors of the radical species. The afore- mentioned Si-O--Si species are consistent with this sug- gestion. It is known that quantitative values for energy cannot be obtained by calculation using the STO-3G basis set, while the result will be qualitatively correct when com- paring similar models.42 The binding energies of 0 1s and Si 2p calculated for Si-0 -'-Si are significantly lower than those for Si-0-Si, while the charge on the oxygen atom is calculated to be less negative than in Si-0-Si (see Table 3).However, the g values for the EPR spectra of the radical species do not correspond to those for the 0-spe~ies.'~.~~.~' It has been reported that Si-0-0 species are formed during y-irradiation on the silica prepared by the sol-gel method, but the EPR spectra observed in such studies bear little or no resemblance to those obtained in the present However, the observation provides evidence for the presence of Si-0-0-Si species on the silica prepared by the sol-gel method. Silica prepared from the sol-gel method has been suggested to contain three-fold siloxane rings.45 Since the ring is strained,46 the Si-0-Si oxygen bridge must be less stable than usual and Si-0-0-Si species replaced with the one-oxygen bridge in the three-fold ring, would be expected to be more stable than those contain- ing the usual oxygen bridge. The total energy calculated for model SiOO-Si(S) is higher than that of model SiOOSi by 6 eV while the energy of model SiO-Si(S) is higher than that of model SiOSi by 9 eV (see Table 3).The result suggests that Si-0-O--Si may be formed in a similar manner to that shown in eqn. (1)during the reaction. Since the binding ener- gies of 0 1s and Si 2p calculated for Si-O-O--Si are significantly lower than those for Si-0-0-Si or Si-0-Si, Si-0-O--Si is also consistent with the results from XPS. The sharp EPR peaks observed are well simulated with a Lorentzian function, showing exchange narrowing of the EPR peaks.47 Since the oxygen atoms in the Si-0- -O--Si species are expected to be equivalent (see results of calculations), transposition of the radical, i.e.Si-O-O--Sie Si-0--0-Si can take place. Thus, the radical species observed on the silica sample are most prob- ably Si-0-O--Si species. Activation of the Silica Surface Although the activity of Si0,-0 in the first of a series of consecutive reactions is increased by preheating at 950 K for 5 h (in contrast with the sample pretreated at 950 K for 0.5 h), the intensities of the EPR peaks are not increased by the same treatment. Hydrogen is formed in the early reactions over Si0,-0 pretreated at 950 K for 0.5 h (Fig.3), while there is no change in hydroxy groups on the surface after the reaction. Hence, it is believed that a small quantity of molec-ular water adsorbed on the surface may suppress the reac- tion. Since, after the first reaction the intensities of the EPR peaks are reduced, it is hypothesized that water interacts with and deactivates the surface radical species. In the case of Si0,-0 evacuated at 1050 K for 0.5 h, the rate of the initial reaction increases after a relatively long induction period, suggesting that the water adsorbed still remains on the surface. The appearance of the sharp EPR peaks just after evacuation of Si0,-0 at 950 K provides evidence for the formation of the Si-O-O--Si radical species on the 1182 surface. Since the intensity of the EPR peaks increases after evacuation at 1050 K, the thermal process Si-0-0-Si + e -+ Si-0-O--Si (4) is expected to have occurred.The presence of surface water apparently does not affect this process since Si0,-0 evac-uated at 950 K for 5 h produces EPR peaks which are similar to those observed for the sample evacuated for 0.5 h while the activity under the former conditions is significantly higher. Since the EPR spectrum is regenerated by evacuation at 950 K after diminishing of the spectrum by contact with carbon monoxide or oxygen at 950 K, the radical is appar- ently restored. Hence, the thermal process [eqn. (4)] is expected to occur on the silica during carbon monoxide oxi- dation. The reaction, CO + 0-+ Si-0-0-Si -+ CO, + Si-0-O--Si (5) would lead to an increase in the number of Si-0-O--Si sites, if 0-is generated in reactions (1)-(3).Summary Catalytic carbon monoxide oxidation takes place over silica prepared by the sol-gel method from ethyl orthosilicate at a reaction temperature of 850 K or above, while no such activ- ity is found with a commercial silica prepared from inorganic reagents. When the silica prepared by the sol-gel method is evacuated at 950 K for 0.5 h before the reaction, the catalytic activity in the first reaction of the series is low, but is increased by repetition of the reaction. Labelled reactions with provide evidence to support the contention that most of the lattice oxygen species on the surface of the silica prepared by the sol-gel method participate in the reaction.From the product distribution of the labelled reaction, the formation of Si-O--Si surface species and atomic oxygen species such as 0-during the reaction appears to be pos- sible. After evacuation at 950 K or above, radical species identified as Si-0-O--Si are generated on the silica while the g1 and gll values for the EPR spectrum are 2.0023 and 2.0029, respectively. Although the radical is stable under carbon monoxide, oxygen, hydrogen or air at room tem-perature, its stability diminishes by heating in one of these gases, with the exception of hydrogen, at 950 K. The number of radical species is increased by repeating the reaction, but decreases in the first reaction after evacuating at 950 K for 0.5 h.Activation of Si-0-Si oxygen bridges on the surface of silica to generate Si-O--Si radical species appears to occur during the reaction. It is supposed that in the initial stage of the reaction the radical electron on the oxygen bridge originates from the Si-0-O--Si species formed in the thermal process. The water adsorbed on the surface may trap the electron and prevent the reaction. Adsorption of carbon monoxide and oxygen on the Si-O--Si radical species can cause formation of carbon dioxide and 0-which will react with carbon monoxide to provide the radical elec- tron for the lattice oxygen species, that is, Si-0-Si and Si-0- 0- Si. We cordially thank Prof. Satohiro Yoshida of Kyoto Uni- versity for supplying the EPR simulator program to us.The financial support of the Natural Sciences and Engineering Research Council of Canada is gratefully acknowledged. References 1 N. D. Spencer and C. J. Pereira, AIChE J., 1987,33, 1808. 2 S. Kasztelan and J. B. Moffat, J. Catal., 1987, 106, 512. 3 S. Kasztelan and J. B. Moffat, J. Chem. SOC.,Chem. Commun., 1987, 1663. 4 N. D. Spencer, J. Catal., 1988,109, 187. J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 Y. Barbaux, A. R. Elamrani, E. Payen, L. Gengembre, J. P. Bon-nelle and B. Grzybowska, Appl. Catal., 1988,44, 117. 6 J. B. Moffat, in Keynotes in Energy-Related Catalysis, Stud. Surf: Sci. Catal, ed. S. Kaliaguine, Elsevier, Amsterdam, 1988, vol. 35, p. 139. 7 J. B. Moffat, in Methane Conversion, Stud.Surf. Sci. Catal, ed. D. M. Bibby, C. D. Chang and R. F. Howe, Elsevier, Amsterdam, 1988, vol. 36, p. 563. 8 S. Kasztelan and J. B. Moffat, J. Catal., 1988, 109,206. 9 S. Ahmed and J. B. Moffat, Appl. Catal., 1988,40, 101. S. Ahmed and J. B. Moffat, Catal. Lett., 1988,1, 141. 11 S. Kasztelan and J. B. Moffat, J. Catal., 1988, 112, 54. 12 S. Kasztelan, E. Payen and J. B. Moffat, J. Catal., 1988, 112, 320. 13 S. Kasztelan and J. B. Moffat, in Proc. 9th Int. Congr. Catal., Calgary 1988, ed. M. J. Phillips and M. Ternan, Chemical Insti- tute of Canada, Ottawa, 1988. 14 N. D. Spencer and C. J. Pereira, J. Catal., 1989, 116, 399. S. Kasztelan and J. B. Moffat, J. Catal., 1989, 116, 82. 16 S. Ahmed and J. B. Moffat, J.Phys. Chem., 1989,93,2542. 17 S. Ahmed and J. B. Moffat, J. Catal., 1989, 118, 281. 18 S. Ahmed, S. Kaztelan and J. B. Moffat, Faraday Discuss Chem. SOC.,1989,87, 23. 19 N. D. Spencer, C. J. Pereira and R. K. Grasselli, J. Catal., 1990, 126, 546. E. MacGiolla Coda and B. K. Hodnett, in New Developments in Selective Oxidation, ed. G. Centi and F. Trifiro, Elsevier, Amsterdam, 1990, p. 459. 21 J. S. Lee and S. T. Oyama, Catal, Rev.-Sci. Eng., 1988,30,249. 22 Y. Amenomiya, V. I. Birss, M. Goledzinowski, J. Galuszka and A. R. Sanger, Catal. Rev.-Sci. Eng., 1990,32, 163. 23 A. P. Legrand, H. Hommel, A. Tuel, A. Vidal, H. Balard, E. Parirer, R. Levitz, M. Czernichowski, R. Erre, H. Van Damme, J. P. Gallas, J. F. Henidy, J. C. Lavalley, 0.Barres, A.Burneau and Y. Grillet, Ado. Colloid Interface Sci., 1990,33,91. 24 S. Malinowski and S. Szczepanska, J. Catal., 1963,2, 310. H. A. Benesi and B. H. C. Winquist, Adv. Catal. Relat. Subj., 1978,27,97. 26 Y. Matsumura, K. Hashimoto and S. Yoshida, J. Catal., 1989, 117, 35. 27 A. Ogata, A. Kazusaka and M. Enyo, J. Phys. Chem., 1986, 90, 5201. 28 Y. Matsumura, K. Hashimoto and S. Yoshida, J. Mol. Catal., 1991,68,73. 29 G. N. Kastanas, G. A. Tsigdinos and J. Schwank, Appl. Catal., 1988,44,33. G. N. Kastanas, G. A. Tsigdinos and J. Schwank, J. Chem. SOC., Chem Commun., 1988, 1298. 31 S. Ahmed and J. B. Moffat, Appl. Catal., 1990,58, 83. 32 Y. Matsumura, K. Hashimoto and J. B. Moffat, Catal. Lett., 1992, 13, 83. 33 L. H. Little, Infrared Spectra of Adsorbed Species, Academic Press, New York, 1966. 34 M. Frisch, J. S. Binkley, H. B. Schlegel, K. Raghavachari, R. Martin, J. J. P. Stewart, F. Bobrowicz, D. Defrees, R. Seeger, R. Whiteside, D. Fox, E. Fluder, J. A. Pople, Gaussian 86 Release C, Carnegie Mellon University, 1988. V. B. Kazansky, in Proc. 6th Inter. Congr: Catal., The Chemical Society, London, 1976, vol 1, p. 50. 36 M. Che and A. J. Tench, Adv. Catal. Relat. Subj., 1982,31, 77. 37 R. A. Weeks, J. Appl. Phys., 1956,27, 1376. 38 M. Stapelbroek, D. L. Griscom, E. J. Friebele and G. H. Sigel Jr., J. Non-Cryst. Solids, 1979,32, 313. 39 D. L. Griscom, E. J. Friebele and K. J. Long, J. Appl. Phys., 1983,54,3743. A. A. Wolf, E. J. Friebele and D. C. Tran, J. Non-Cryst. Solids, 1985,71, 345. 41 C. D. Wagner, W. M. Riggs, L. E. Davis, J. E. Moulder and G. E. Muilenbergy, Handbook of X-ray Photoelectron Spectroscopy, Perkin-Elmer Corp., Minesota, 1978. 42 J. Sauer, Chem. Rev., 1989,89, 199. 43 J. H. Lunsford, Catal. Rev., 1973,8, 135. 44 G. Kordas and R. A. Weeks, J. Non-Cryst. Solids, 1985,71,327. C. J. Brinker, D. R. Tallant, E. P. Roth and C. S. Ashley, J. Non-Cryst. Solids, 1986,82, 117. 46 T. Kubo and S. Nagase, J. Am. Chem. SOC., 1985,107,2589. 47 J. A. McMillan, Electron Paramagnetism, Reinhold, New York, 1968. Paper 3/004067D; Received 12th July, 1993