J. CHEM. SOC. FARADAY TRANS., 1994, 90(6), 905-910 Surface Characterization and Catalytic Activity of Co,Mg, -xAl,O, Solid Solutions Oxidation of Carbon Monoxide by Oxygen Franco Pepe* Dipartimento di Chimica della Terza Universita di Roma, Via Ostiense 159,00154-Roma, ltalia Manlio Occhiuzzi Centro Struttura e Attivita Catalitica di Ossidi (CNR), c/o Dipartimento di Chimica dell'Universita di Roma 'La Sapienza ', P. le A. Moro 5,00185-Roma, ltalia Spinel solid solutions of Co,Mg, -xAl,O, with x = 0-1 have been studied as catalysts for CO oxidation by molec- ular 0,. The experiments were performed either with or without CO, condensation. The apparent activation energies varied depending on the cobalt content and the procedure adopted. The activity per ion (turnover frequency) was fairly constant over the whole range of cobalt contents, when surface cobalt ion concentrations, as measured by X-ray photoelectron spectroscopy (XPS), are considered.It is inferred that each Co2+ ion is active irrespective of its concentration and configuration (isolated, in dimers or in clusters) and that the coordi- nation (tetrahedral or octahedral) plays a minor role in the activity per ion. The reaction appears to be structure insensitive as opposed to the N,O decomposition which appears to be concentration dependent. A mechanism of CO oxidation is proposed in terms of the nature and reactivity of the surface species formed on adsorption of 0, and/or CO. The effect of CO, on the reaction rate is also discussed.In a previous paper' it has been shown that cobalt ions fully Experimentalisolated and dissolved in the MgAl,O, spinel matrix are inactive for dinitrogen oxide decomposition. Comparison Sample Preparation and Characterization with oLer solid solutions containing led to the The catalysts used were those investigated in a previous conclusion that this behaviour had to be correlated with the paper' where the notation adopted, preparation, chemical non-assistance of the matrix MgAl,O, in the kinetically analysis, structural and magnetic properties are described. important step of oxygen migration and desorption. If, Note that the Co,Mg, -xA120, solid solutions are designated however, pairs of cobalt ions were created simply by increas- as SAMCo and that the figure after the symbol SAMCo indi-ing the guest ion concentration, the activity strongly cates the nominal number of cobalt atoms per 100 (Mg + Co) increased with respect to that of the matrix.This finding was atoms. Pure MgAI,O, and CoAl,O, are designated SAM correlated both with (i) the presence of cobalt ions octa- and SACo, respectively. In the present investigation surface hedrally coordinated and (ii) the isolation of pairs of the analysis was performed by XPS. The catalysts were treated +Co2 species. for different times in air at various temperatures (873, 753, The effect on the activity of the different coordinations of 573 K), immediately transferred to the sample holder and transition metal ions (tmi), in the spinel phase (site symmetry) then evacuated at room temperature to better than has been a matter of controversy. Here it is sufficient to recall 1.3 x lo-' N rn-, in the XPS analysis chamber. It was the classical work by Schwab et al.on ferrites for CO oxida- checked that a standard treatment of 3 h at 753 K was suffi- ti~n,~where the divalent ions were found to be completely cient to achieve reproducible results. The spectra were inactive and the normal ferrites more active than the inverse recorded at room temperature on a Leybold Heraeus LHS 10 ones. An opposite result was found by Boreskov et al. for spectrometer (FAT mode) equipped with an HP 2113 com- methane and hydrogen oxidation.6 puter for data analysis. Mg-Ka (1253.6 ev) and Al-Ka (1486.6 As a further aspect, several model reactions appear to be eV) radiations (12 kV, 30 mA) were used.The Co 2p, 0 Is, A1 'structure sensitive' or 'demanding' because the activity per 2s, A1 2p, Mg 2s and Mg 2p peaks were recorded. The 0 1s ion (turnover frequency) varies with ion concentration. peak at 529.5 eV was taken as reference. The data analysis However, the stucture sensitivity does not necessarily arise procedure involved smoothing, backgound subtraction by a from a particular geometry of the active site. In fact, struc- non-linear integral profile and curve-fitting (DS4X program tural sensitivity may also result if the strengths of bonds by Leybold Heraeus). The surface composition of SAMCo involved in the step controlling the molecular mechanism solid solutions was obtained using the sensitivity factor depend on the active ion concentration or are directly linked appr~ach.~."The atomic sensitivity factors were determined to the reaction mechanism.The N,O decomposition and the on SAC0 pure compound by means of the expression: H,-D, equilibration on Coo-MgO solid solutions can be quoted as examples:' in the former, the steps involved are SCd+ = (&o 2plb 1s)SACo (1) oxygen adsorption-desorption ; in the latter, the proposed The following equation was applied to evaluate the cobalt reaction mechanism requires at least two neighbouring Co2 + surface content in the SAMCos ions. It seemed therefore, of interest to investigate, on the same catalytic system, the roles of the active ion concentration and coordination in a simple reaction, such as CO oxidation, in order to elucidate the nature of the active site, to compare the Catalysis activity of cobalt ions in spinel solid solutions with that in CO oxidation was carried out in a circulating system with a pure oxides and to suggest a reaction mechanism.total volume of 0.36 1 with two traps at 77 or 194 K placed before and after the reactor. The catalysts were initially con- ditioned in uucuo (p = 0.0013 N m-,) at 753 K for 4 h and at the same temperature for 0.5 h between runs. The CO : O2 ratio in the mixture was 2 : 1 and the initial pressure usually around 2 x 10, N m-,. The extent of the reaction was fol- lowed by means of a pressure transducer. The absolute first-order rate constant was calculated by the expression : (3) where po is the initial pressure of the mixture, p the pressure at the time t, I/ the reaction volume and A the catalyst surface area.The assumption of a first-order law was proved satisfactory by plotting ln(p/pO) us. time. In all cases, provided that the reaction extent was mantained below 30%, a straight line was obtained. Two sets of experiments were performed by changing the freezing mixture in the traps. By using liquid nitrogen (LN), the CO, produced was immediately condensed after the reactor: these experiments will be labelled as A. The second set with the traps at 194 K and without CO, condensation is labelled as B. The reaction order, n, with respect to the total pressure was calculated by performing experiments at different initial pres- sures, po, of the stoichiometric mixture and plotting the log of initial velocities as a function of the log of the initial pres- sures. Adsorption Experiments Static adsorption experiments were performed in a conven-tional BET apparatus with an LN trap placed in series with the adsorption loop.A fresh portion of the catalyst was evac- uated at a pretreatment temperature of 753 K for 4 h and oxygen and carbon monoxide were added at a given adsorp- tion temperature T, according to the scheme: (a)evacuation at 753 K; (b) oxygen adsorption at T; (c) evacuation at 7';(d) oxygen readsorption at T; (e) evacuation at T; (f) CO adsorption at T;(g) evacuation at T; (h) measurement of the pressure of the condensed gas; (i) evacuation at T; u)oxygen readsorption at T.The steps (a)and (b) allow measurement of the total amount of oxygen adsorbed. The amount of J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 reversibly adsorbed oxygen was measured uia steps (c) and (d). The steps (a)-(e) were repeated until constant values were obtained at each selected temperature T; then a known volume of CO was admitted and the decrease in pressure monitored. The pressure of the condensed gas was then mea- sured [step (h)]. A further admission of oxygen [step (j)] allowed calculation of the amount of oxygen readsorbed. The condensable gas was assumed to comprise CO, only. Results Sample Characterization and Surface Analysis The main features of the sample characterization (experimental cobalt molar fraction, X, , fraction of cobalt ion in tetrahedral sites, a, and BET surface areas, A) are reported in Table 1.The relevant XPS parameters for SAM (1473), SAC0 (1473) and SAMCo (1473) samples are listed in Table 2. In SAMCol the cobalt signal was hardly detectable because its intensity was just a little stronger than the ground noise. However, the peak binding energy (Eb) and the inten- sities of Mg, A1 and 0 were in agreement with those of SAM. For the other cobalt-containing samples, the Co 2p region was very similar in the SAMCos and in SACo. The param- eters are typical of cobalt ions in the 2 + oxidation state.' In the SAMCo samples the 0 1s region showed a small shoul- der at an Eb of ca.2 eV lower than that of the main peak. The shoulder area was estimated by a curve-fitting procedure to be a few per cent of that of the main peak. The cobalt surface contents of the solid solutions were obtained by the areas of the Co 2p and 0 1s peaks reported in Table 3 apply- ing eqn. (1) and (2) quoted in the Experimental section. The results show that with reference to the surface cobalt content of SAC0 (taken equal to l), SAMCo5O has x,, = 0.48, SAMColO has x, = 0.12 and SAMCoS has x, = 0.084. Comparison with the experimental Co content quoted in Table 3 suggests that cobalt enrichment occurs on the dilute samples and the phenomenon is particularly evident on SAMCo5. Catalytic Activity The catalytic activity for carbon monoxide oxidation was investigated in the 420-740 K temperature range on the solid Table 1 Experimental Co2+ mole fraction, fraction of tetrahedral cobalt, BET surface area and apparent activation energies for procedure A and B of Co,Mg, -,Al,O, samples sample a A/m2 g-I SAM (1473) 0.00 1.9 67 f5 113 f5 SAMCol (1473) 0.0033 2.9 38 f5 88 & 5 SAMCo5 (1473) 0.01 6 0.60 1.7 38 f5 52 k5 SAMColO (1473) 0.0323 0.75 1.9 33 f5 50 f5 SAMCo5O (1473) 0.164 0.77 1.o 29 f5 29 & 5 SAC0 (1473) 0.333 0.77 0.7 29 f5 29 f5 SAMCo5 (1073) 0.016 0.69 2.2 42 f5 70 & 5 SAC0 (1073) 0.333 0.82 0.8 33 f5 33 & 5 Table 2 XPSparameters of Co,Mg, -,Al2O, samples equilibrated at 1473 K Sample sat4/eV Mg2p4/eV A1 2s4/eV SAM SAMCo 1 -C -C 537.6 537.4 48.6 48.6 117.6 117.6 SAMCoS 780.2 786.4 0.36 796.1 537.2 48.6 117.6 SAMColO 780.1 785.9 0.43 796.3 537.1 48.5 117.6 SAMCoSO 780.0 785.2 0.36 796.0 536.4 48.3 117.6 SAC0 780.0 785.4 0.39 795.6 536.5 117.6 " Referred to E, (01s) = 529.5 eV.'As observed before charging correction. Hardly detectable. J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 Table 3 XPS intensities of Co,Mg, -xAl,04 samples sample XColl1: I* I,, 2: ico 2pb SAM 18.5 2.09 SAMCo 1 0.010 16.3 2.28 -C SAMCoS 0.048 16.9 2.19 1.65 SAMCo 10 0.097 15.4 1.98 2.17 SAMCoSO 0.49 15.9 2.30 8.8 1 SAC0 1 15.2 2.48 17.7 " Experimental cobalt content in the spinel formula Co,Mg,-,A1,04. Intensities are expressed in lo3 eV counts s-'.Hardly detectable. solutions specified in Table 1, which gives the apparent acti- vation energy, E, ,for both sets A and B. The main results are: (i) the catalytic activity increases with increasing cobalt content irrespective of the procedure adopted and (ii) the E, values are higher in procedure B for SAM as well as for the dilute specimens (SAMCol-lo), whereas difference in E,s between procedures A and B are found for SAMCOSO and SACo. The kinetic constants ktbS and k:bs at 500 and 667 K as a function of cobalt content are listed in Table 4. Comparison shows that: (i) at 667 K, ktbs and k;bS are roughly equal for each cobalt concentration with the exception of SAM and SAMCol and (ii) at 500 K the same behaviour is obeyed by SAC0 and SAMCo5O only, whereas for all the dilute samples, kt&/k;b, x 3.The role of the cation distribution has been investigated on the catalysts SAC0 and SAMCo5 equilibrated at different temperatures (see Table 1).The results for SAMCo5 are pre- sented in Fig. 1 as Arrhenius plots. It turns out that the speci- mens from both procedures A and B with a larger amount of octahedral cobalt are more active. Also the sample SAC0 equilibrated at 1473 K and, hence, with a larger percentage of octahedral cobalt, is slightly more active. However, for SAC0 no difference either in activity or in E, was observed between procedures A and B. Reaction orders with respect to the total pressure were investigated in the 560-750 K range and calculated from the initial velocities.They ranged from 0.85 to 1 for procedure A. For procedure B the samples SAMCo5O and SAC0 exhibited the same value as for procedure A, whereas for diluted samples and SAM a negative effect of the CO, was estimated as a negative order in CO, in the range 0.3-0.5. 1 I I I 1.5 1.7 1.9 103 KIT Fig. 1 Catalytic activity of SAMCo5 with different cation distribu- tions. (0)SAMCo5 (1473) expt A; (0)SAMCo5 (1473) expt B; (0) SAMCo5 (1073) expt A; (0)SAMCoS (1073) expt B. Adsorption Experiments The adsorption experiments were performed on SAM, SAMCo5 (1473) and SAMCo5O (1473) at 570 and 670 K. No oxygen adsorption was monitored on pure SAM and in Table 5 the results are reported for SAMCo5 and SAMCo5O in terms of atoms of oxygen and molecules of CO or CO, adsorbed per 1 nm2.It turns out that adsorption is activated on both of the samples and that the oxygen is irreversibly adsorbed on SAMCo5, but only partially reversibly adsorbed on SAMCo5O. Assuming that the total number of cations exposed on the averaged (loo), (111) and (110) is 7 x 1Ol8 m-, and that in oxygen adsorption one oxygen atom is adsorbed per cobalt ion, it follows that the number of cobalt ions exposed on SAMCo5 and SAMCo5O would be 0.20 and 1.12 nrn-,, respectively and that two thirds of cobalt ions would be covered by irreversibly adsorbed oxygen atoms at 670 K on SAMCo5; by contrast, only one third of the Co2+ surface in SAMCoSO would be covered at the same tem- perature (Table 5).The subsequent adsorption of CO is in Table 4 Absolute first-order rate constants at 500 and 667 K for both procedures A and B sample ktbS(500)/lO8m s-' k~,,,(500)/108m s-' k,",,(667)/108 m s-' k:b,(667)/108 m s-' SAM (1473) SAMCol (1473) SAMCo5 (1473) SAMColO (1473) SAMCo5O (1473) SAC0 (1473) SAMCo5 (1073) SAC0 (1073) 0.2 2.7 1.8 13.8 14.5 0.9 13.0 0.8 0.6 13.0 14.5 13.0 0.2 1.5 20.0 13.2 63.0 79.4 10.9. 79.b 0.1 0.6 22.8 12.0 62.7 79.4 12.5 79.0 Table 5 sample T/K SAMCo5 (1473) 570 SAMCoS (1473) 670 SAMCo5O (1473) 570 SAMCo5O (1473) 670 ., 0, adsorption is reported as atoms nm-' &0.03. Oxygen and carbon monoxide adsorption at 570 and 670 K" 02.tot 02.rev 02,irrcv co co2 OZ,reads 0.07 0.00 0.07 0.37 0.00 0.00 0.13 0.00 0.13 0.44 0.40 0.47 0.3 1 0.22 0.09 1.03 1.oo 0.9 1 0.43 0.10 0.33 2.27 2.30 2.00 and CO adsorption as molecules nm-'. The experimental error of the measured quantities was large excess compared with the irreversible adsorption of oxygen on both of the samples. It can be inferred that CO adsorption also involves matrix sites, in addition to Coz+ sites. The amount of CO, evolved by SAMCoSO corresponds to the amount of CO admitted at both temperatures, i.e. adsorption was fully reversible in the 570-670 K range. For SAMCo5 no CO, was desorbed at 570 K. If, however, the temperature was raised to 670 K, the amount of CO, desorbed and collected in the LN trap coincided with the amount of CO adsorbed.This suggests the formation of a strong C0,-surface bond, as a consequence of which, at 570 K successive oxygen adsorption is hindered on SAMCoS. The readsorption of oxygen on SAMCoSO, by contrast, seems to replace totally the surface oxygen which reacted with the CO admitted. Discussion Turnover Frequencies and Cobalt Concentration The catalytic activity of the SAMCo solid solutions and SAC0 for CO oxidation is distinctly higher than that of the pure matrix SAM. The increase in activity is present over the entire cobalt concentration range and suggests that the active centres responsible for the activity of SAC0 and SAMCos involve cobalt ions. In addition, the presence of cobalt is also responsible for the lower E, compared with that of the pure matrix.Indeed it is possible to envisage three distinct sce- narios: (i) the pure matrix SAM has the highest E, irrespec-tive of the procedure adopted; (ii) SAC0 and SAMCo5O have the lowest E, values irrespective of the procedure adopted; (iii) the dilute samples show decreasing values of E, on increasing the cobalt content (SAMCo1 to SAMColO). However, the variation in E, is quite smooth for procedure A in contrast with procedure B, where the variation is rather drastic (Table 1). Since the reaction order deviates from unity as the cobalt content decreases and since the adsorption experiments showed a larger reversibility of CO, adsorption on SAMCo5O as opposed to SAMCoS, it can be inferred that the variation in E, observed in passing from SAC0 to SAM is due to an increase in the adsorption heat of CO,.Indeed a strong adsorption of CO, can take place if basic sites 0;-are present (see mechanism section) and this situation is shown more markedly for dilute samples and SAM, and for procedure B. Turning now to the specific activities, the values of ksbs indicate that isolated Coz+ ions, such as those in SAMCol, are active per se and it may be asked whether the isolation of the ion is a fundamental prerequisite for the reaction to occur. The problem of identifying the surface configurations required for a given reaction may be investigated by inspect- ing the dependence of the turnover frequencies, PIion, as a function of the tmi ~oncentration.~ In fact, if it is assumed that the same mechanism is operating over the whole range of active ion concentration and that the same active complex is involved, as suggested by the constancy of E, in the reac- tion, then the only parameter affecting kabs is the active site concentration.Now the active site concentration in few cases is accessible experimentally; XPS studies on COO-MgO solid solutions" have supported the assumption that the surface Coz+concentration is equal to the Coz+ bulk concentration and that all the cobalt ions or a constant fraction of them are active sites for CO oxidation. The latter assumption deserves some comment. The real surface concentration of active sites may be less than the surface ion content, because the active site could require a special configuration such as Co2+-Co2+ dimers or Co2+clusters. Such a configurational factor is con- centration dependent and one has to calculate the amount of J.CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 each configuration as a function of the cobalt concentration. However, if it is assumed that the active configuration is simply that containing one cobalt ion, either isolated or belonging to a cluster, the turnover frequency is directly pro- portional to the kabs values divided by the real surface cobalt concentration. An extended study on the quantitative XPS surface analysis in ternary systems 1s in progress and the results presented in the Experimental suggest that the surface cobalt content of SAMCo5 is about twice the analytical one whereas a negligible enrichment is found for SAMColO.For SAMCol it is possible to suggest only that the surface cobalt concentration is similar to the analytical one. This hypothesis is based on the simulation, through the sensitivity factor, of the Co2+signal intensity of samples with x = 0.01 and 0.02. In Fig. 2(a),log Nionvs. log (cobalt surface mole fraction) is reported at two temperatures (500 and 667 K) for procedure A on the assumption that the cobalt surface content of SAC0 is equal to the analytical one. It appears that the trend shown is not influenced by the variation in E,. It may also be added that for procedure B the trend is substantially unmodified. Inspection of Fig.2(a) shows that Nionvalues are scattered around values of 2 x lo-' and 3 x molecules s-' per ion at the considered temperatures. Then during CO oxida-tion, cobalt ions, either isolated as sn dilute samples or in clusters as in concentrated samples, have constant activity. The trend parallels that previously found for Co2+dispersed either in MgO or in other oxide matrices; in addition, the absolute values of Nionfor the Co2'-containing solid solu- tions previously investigated are in good agreement with the present value^,'^^^^ even if the system COO-MgO was the most active for reasons related to the cobalt coordination and/or to the special ability of the matrix in oxygen transfer. It turns out that the cobalt ion activity has to be attributed to the electronic structure of the pair Co-0 only and the reaction appears to be facile and not configuration depen- dent. Finally, note that the slightly higher activity per ion of SAMCoS is in line with the fact that the SAMCo5 (1473) sample contains the largest amount of octhahedral cobalt (see Table 1) and, hence, a higher activity per ion has to be expected as discussed in the next section.A different situation is envisaged for N20 decomposition as shown in Fig. 2(b). In fact, the inactivity of SAMCol points to non-assistance of the spinel matrix in the oxygen SAMCo 1 5 10 50 SAC0 -1 I I +l t I 1 1 - 16 17 18 19 log[C02+] Fag. 2 Turnover frequency as a function of cobalt concentration for two reaction.(a) CO oxidation at 500 K (a),and at 667 K (m); (b) N,O decomposition at 500 K (a),and at 667 K (a). J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 desorption process, and therefore the active site for N,O decomposition must be a configuration of two or more cobalt ions by which neighbouring adsorbed oxygen atoms can be desorbed. On statistical grounds, isolated cobalt ion pairs are formed in the cobalt ion concentration range 1-10% and the highest values of Nionfor SAMCoS and SAMColO reason- ably suggest that isolated Co2+ pairs are the active surface configuration for the decomposition. The reaction is therefore concentration dependent and hence 'demanding' because it is configuration dependent.Catalytic Activity of Octahedral and Tetrahedral Cobalt For N,O decomposition it was proved that small differences in bulk cation distribution, as exist between the (1073) and the (1473) series, resulted in a large change in activity.' In the reaction investigated here, by contrast, the differences in cata- lytic activity due to different bulk cation distributions are less relevant as shown in Fig. 1. In addition, the E, values in the present case vary from 29 kJ mol-' [SACo (1473)l to 33 kJ mol-' [SACo (1073)l and from 38 kJ mol-' [SAMCoS (1473)l to 42 kJ mol-' [SAMCoS (1073)l. In N,O decompo-sition, the increases for the same pairs of samples were from 84 to 121 kJ mol-' and from 88 to 121 kJ mol-', respec-tively. It can be concluded that in both reactions a larger amount of bulk octahedral cobalt favours a higher activity even if to a different extent.An attempt to explain the different role of the tetrahedral cobalt in the two reactions must consider: (i) the actual tetra- hedral cobalt on the surface; (ii) the role of the tetrahedral cobalt in the specific reaction; and (iii) the presence of carbon monoxide and its effect on the reduction of the surface. For point (i) it was proposed that a randomization effect as a function of the equilibration temperature would render the catalyst surface with a given cobalt concentration more rich in octahedral cobalt with respect to the bulk' and, hence, more active when equilibrated at 1473 K. The present results would be in agreement with this hypothesis.However, while tetrahedral cobalt was found to be completely inactive in N,O decomposition because of a strong cobalt-oxygen bond formed upon decomposition, for the present reaction results published el~ewhere'~ have shown that the cobalt ion tetra- hedrally coordinated in zinc oxide is only slightly less active with respect to the octahedral one. Such a behaviour may find support by considering the reducing power of CO able to overwhelm the difference in oxygen bond stengths deriving from different cation coordinations. A simil-ar explanation was, in fact, proposed for manganese ions dissolved in MgO where both Mn3+ and Mn4+ exhibited the same activity for CO oxidation irrespective of the difference in oxygen bond ~trength.'~It transpires that the difference in activity between pairs of samples of equal cobalt content and equilibrated at different temperatures is expected to be less for CO oxidation than for N,O decomposition as, in fact, is found.In conclusion, cobalt ion coordination plays a role in the present reaction. However, many factors, such as the presence of a randomization effect, the activity of tetrahedral cobalt and the reducing ability of the CO, tend to smooth the picture. Reaction Mechanism The oxidation of CO on cobalt-based catalysts generally occurs uia a redox mechanism and the oxygen involved in the active surface complex comes from surface lattice ~xygen.'~." The reaction between CO and 0;-in the SAMCo solid solutions takes place at high temperatures only (above 570-670 K depending on cobalt concentration, Table 909 5).Therefore, a mechanism which involves CO oxidation oia the extraction reaction 0:-+ CO+CO, + 0, + 2e (1) followed by oxidative adsorption of O,, thus restoring the 0;-via a Mars-Van Krevelen mechanism, is likely in the temperature range investigated. In addition, the dependence of turnover frequency on the cobalt concentration suggests that a cobalt-containing intermediate is involved in the reac- tion mechanism during sustained catalysis. With reference to an IR study on the spinel CoAl,04 and on NiO/A1,03,'8*19 it is possible to envisage the surface location of Co2 + on the (1 11) plane as: P uAv \IAv \IAV I 9CO Al CO CO Al I CO CO /I\ I In I A\ 1 A\ where Mg2+ can replace tetrahedral cobalt in the present spinel solid solutions.The unsaturated cationic sites would be the adsorption sites for 0, and CO. In fact, the adduct Co3+.* -0;has been identified as an adsorption intermediate in the MgO-CoO system" while combined electron para- magnetic resonance (EPR) and IR experiments" suggested the coordination of a CO molecule to a Co3+*-.0;adduct and the formation of a hydrogencarbonate-like comple~.'~*~~ On cobalt aluminate the CO adsorption gives rise to IR bands assigned to CO on Co2+ centre^.'^*^^ Moreover, on increasing the contact time, bands in the carbonate region grew up and these modifications were associated with redox reactions giving CO, and carbonates." A tentative reaction intermediate would therefore be 0, co The evolution of the suggested intermediate implies the for- +mation of a carboxylate species Co3 * -CO;.However, the formation of a true carbonate species cannot be excluded : co3+. . .co; + 0:--,CoZ+...co2-(11) CO, desorption might take place as follows: CO~+.-CO; -,co, + co2+ (111) and/or as : co~+-*co~-co2++ co, + 0:-(Iv)-b Monodentate and/or bidentate carbonate complexes have frequently been identified by IR on cobalt-containing ~arnples~~*~~-~'and, in general, their stability would depend on the basic character of the 0;-environment. CO, effect on the reaction rate The main result of the adsorption cycles was that CO, adsorption appeared to be reversible on SAMCoSO at all temperatures, but on SAMCoS only at high temperature 670 K.In addition the activity was generally lower for procedure B and deviation from a first-order kinetic law was observed. These findings suggest that strong adsorption of CO, can involve the basic sites 0'-of the matrix MgA1204 and can arise to a greater extent when CO, is not frozen. The two forms of adsorbed CO,, the former produced via reaction and the latter directly adsorbed from the gas phase, may be not in equilibrium, as reported for CO oxidation on Tho, .29 In our case the first species would be that formed on the Co2+. .Oz-centres in concentrated samples and constitutes the active complex. The second species, which inhibits the reaction, would be related to the increasingly basic character of 02-in the Co2+*..02-sites in dilute samples, where, on statistical grounds, the 0’-ions will be shared more with Mg2+ ions.In fact, on MgA120, surface carbonates are found to be more stable than on CoA120, by IR investiga- tion~.~~ Finally, note that the activity of SAC0 in the present work is one order of magnitude less than that of C00.l~ Therefore, the large decay in activity found when surface CoA1204 is f~rmed~l.~~is not confirmed, while the E, values are substan- tially in agreement. Conclusions The cobalt surface analysis proved that a tmi enrichment exists, which is a function of the cobalt concentration. The knowledge of the tmi real surface content has revealed dis- tinctly differance activity patterns for N20 decomposition and CO oxidation. The trend of the turnover frequency as a function of cobalt concentration suggests that the surface configurations active for the specific reactions are different.It turns out that the former reaction is configuration dependent or ‘demanding’ while the latter is a configuration indepen- dent or ‘facile’ reaction. As a consequence in each reaction, both the coordination of the tmi and the matrix environment play a very different role in the activity per ion. References C. Angeletti, F. Pepe and P. Porta, J. Chem. Soc., Faraday Trans. I, 1978,74,1595. A. Cimino and F. Pepe, J. Catal., 1972,25,362. F. Pepe, M. Schiavello and G. Ferraris, 2.Phys. Chem. NF, 1975, %, 297. C. Angeletti, A. Cimino, V. Indovina, F. Pepe and M. Schiavello, J. Chem. SOC., Faraday Trans. I, 1981,77,641. G. M. Schwab, E. Roth, Ch. Grintos and N. Mavrakis, Structure and Properties ofsolid Surfaces, ed. R. Gomer and C. S. Smith, Chicago, 1953, p. 464. J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 6 G. K. Boreskov, V. V. Poposki, N. El. Lebedeva, V. A. Sazonov and T. V. Andruchkevitch, Kinet. Catal., 1970, 11, 1093. 7 V. Indovina, A. Cimino, M. Inversi and F. Pepe, J. Catal., 1979, 58,396. 8 C. Angeletti, F. Pepe and P. Porta, J. Chem. SOC., Faraday Trans. 1, 1977,73, 1972. 9 C. D. Wagner, L. E. Davis, M. V. Zeller, J. A. Taylor, R. H. Raymond and L. H. Gale, SurJ: Interjuce Anal., 1981,3,211.10 B. L. Yang, S. F. Chan, W. S. Chang and Y. Z. Chen J. Catal., 1991, 130, 52. 11 B. A. Sexton, A. E. Hughes and T. W. Turney, J. Catal., 1986, 97, 390. 12 A. Cimino, Mater. Chem. Phys., 1985,13,221. 13 V. Indovina, A. Cimino and F. Pepe, Gazz. Chim Ital., 1980,110, 13. 14 V. Indovina, A. Cimino and F. Pepe, Proc. ZX Zberoamerican Symposium on Catalysis, Lisbon, 1984. 15 A. Cimino and V. Indovina, J. Catal., 1974,33,493. 16 J. K. Dixon and J. E. Longifield, Catalysis, Reynold, New York, 1960, vol. 7, p. 303. 17 J. M. Thomas and W. J. Thomas, Introduction to the Principles of Heterogeneous Catalysis, Academic Press, London, 1967, p. 367. 18 G. Busca, V. Lorenzelli, V. S. Escribano and R. Guidetti, J. Catal., 1991, 131, 167. 19 E. Borello, A. Cimino, G. Ghiotti, M. Lo Jacono, M. Schiavello and A. Zecchina, Discuss Faraday So(:.,1971,52,149. 20 D. Cordischi and V. Indovina, Adsorption and Catalysis on Oxide Surfaces, Elsevier, Amsterdam, 1985, p. 209. 21 V. Indovina, D. Cordischi, M. Occhiuzzi and A. Zecchina, per- sonal communication. 22 V. G. Amerikov and L. A. Kasatkina, Kinet. CataL, 1971, 12, 137. 23 C. Morterra, G. Ghiotti, F. Boccuzzi and S. Coluccia, J. Catal., 1978,51, 299. 24 G. Busca, R. Guidetti and V. Lorenzelli, J. Chem. SOC.,Faraday Trans. I, 1990,86,989. 25 G. Busca, F. Trifiro and A. Vaccari, Langmuir, 1990,6, 1440. 26 W. Hertl, J. Catal., 1973,31,281. 27 A. J. Goodsel, J. Catal., 1973,30, 175. 28 S. Matsushita and T. Tanaka, J. Chem. Phys., 1962,36, 665. 29 M. Breisse, B. Claude1 and J. Veron, Kinet. Katal., 1973, 14, 102. 30 G. Busca, personal communication. 31 Y. Y. Yao, J. Catal., 1974,33, 108. 32 M. A. Wheller and M. Bettman, J. Catal., 1975,40, 124. Paper 3/05781J; Received 24th September, 1993