J . Chern. Soc., Furuduy Trans. I, 1989, 85(6), 1267-1278 Influence of Pretreatment on the Properties of Ag/a-Al,O, Catalysts containing Large (& 1 pm) Pure and Cs-promoted Silver Particles Part 2.-CO Oxidation Measurements Garmt R. Meima,*? Mees Hasselaar, Adrianus J. van Dillen, Frederik R. van Buren$ and John W. Geus Department of Inorganic Chemistry, State University of Utrecht, Croesestraat 77A, 3522 AD Utrecht, The Netherlands The catalytic activity for the oxidation of CO with molecular oxygen of silver catalysts containing large silver particles (ca. 1 pm) has been studied. Both pure and caesium-promoted silver catalyst were investigated. The various pretreatments to which the catalysts were subjected, i.e. reduction or oxidation at different temperatures, influenced their activity.The promoted catalyst proved to be somewhat less susceptible to the various pretreatments. Arrhenius plots of the catalysts often exhibited a sharp break, giving rise to two sets of kinetic parameters. An apparent activation energy of 40 kJ mol-I was found at low temperatures, whereas an activation energy of 60 kJ mol-' was observed at high temperatures. It is argued that the activation energy of 40 kJ mol-1 can be ascribed to the reaction of CO with oxygen atoms in the neighbourhood of defects and the activation energy of 60 kJ mol-' to the reaction of CO with oxygen penetrated into the bulk at the same sites. The variation in activity could be attributed to changes in the pre-exponential factor. A separate discussion is devoted to a comparison of the present results with the findings on silver catalysts with a smaller mean particle size we reported on previously.Introduction In Part 1 of this publication we have reported on the interaction of oxygen and hydrogen with silver catalysts containing large silver particles (ca. 1 pm) both with and without a caesium pr0motor.l It was established that the effect of the pretreatment on the extent of sorption of both oxygen and hydrogen was small compared with the amounts taken up by the smaller silver particles studied previously.2,3 It was also found that the fraction of oxygen taken up as adsorbed oxygen atoms did not vary significantly with the pretreatment. Moreover, even after oxidation at elevated temperatures the amounts of oxygen that could penetrate into the silver lattice were relatively small.The overall results indicated that the effect of the pretreatment on the surface structure of large silver particles is restricted to the near surface layers. In this paper the influence of the state of these catalysts and of the various pretreatments on the activity for the oxidation of CO by molecular oxygen will be dealt with. In previous studies on catalysts containing much smaller silver particle^,^? we have shown that this reaction is very sensitive with respect to the surface morphology of the t Present address: Dow Benelux N.V. P.O. Box 48, 4530 AA Terneuzen, The Netherlands; to whom all $ Dow Benelux N.V. correspondence should be addressed. 12671268 Influence of Pretreatment on the Properties of Ag/a-Al,O, silver particles.The pretreatment turned out to have a very strong influence on the activity. The results could be explained satisfactorily by the fact that on the most abundant plane in the silver particles, uiz. the (1 1 1)-plane, oxygen is adsorbed exclusively at defects. The large variation in the activity for the oxidation of carbon monoxide could be ascribed to : (i) a change in the morphology leading to a larger or smaller fraction of (1 1 1)-surfaces, and (ii) a varying number of defects in Ag (1 1 1)-surfaces. Taking into account that with large silver particles complete structural rearrangements are more difficult and with our findings in Part 1 that the effect of the pretreatment is restricted to the near surface layers, the influence of defects and/or added caesium on the activity for CO oxidation can be studied.Our previous results have shown that oxygen penetrated into the silver lattice exhibits an activation energy of 60 kJ mol-', whereas adsorbed oxygen gives rise to an activation energy of 40 kJ mol-1.2T Therefore, CO oxidation measurements will also provide additional information on the influence of the pretreatment on the state of the sorbed oxygen species. Experiment a1 Catalysts The preparation procedure of the catalysts is described in Part 1.l Both catalysts consisted of 19.9 wt % silver supported on a-Al,03. The promoted catalyst contained 190 ppm Cs. The catalysts are designated Ag-Cs and Ag, respectively. CO Oxidation Measurements The pre-shaped catalyst particles were crushed into smaller particles and a sieve fraction of 0.5-1.0mm used in the experiments.A sample of 1.38 g of the Ag and 1.25 g of the Ag-Cs catalyst were placed into a tubular glass reactor with an internal diameter of 9 mm. The volumes of the catalyst bed were 1.55 cm3 and 1.40 cm3, respectively. The pretreatment of the catalysts (reduction or oxidation) was performed overnight at various temperatures, either in a 10% H,/N, flow or a 100% 0, flow. Unless otherwise stated, the catalyst was cooled down in the same atmosphere as that of the pretreatment. To study the influence of the pretreatment on the catalytic activity, the following sequence was decided on. First pretreatments in inert and oxidizing atmospheres would take place, whereafter the influence of reduction and subsequent oxidation would be studied. As shown in Part 1, oxidation of the catalyst is necessary to remove the carbonaceous impurities present on the fresh catalyst.' During the experiments a gas-flow of 50 cm3 min-l was passed downwards through the catalyst bed.The gas composition by volume was 1 YO CO, 1 YO O,, N, balance. All gasses were dried using standard methods. The temperature was measured and controlled by a thermocouple placed on top of the catalyst bed. The amount of carbon dioxide formed during passage through the reactor was determined conductimetrically . The conversion was calculated by comparing total combustion with the measured conversion. Total combustion was achieved by passing the effluent gas of the reactor through a bed of a supported copper catalyst kept at a temperature of 673 K. Details of this very accurate technique for measuring the concentration of carbon dioxide have been published el~ewhere.~ Results and Discussion The activity measurements for the oxidation of CO with molecular oxygen of both the promoted and unpromoted catalyst showed the same tendency towards the various pretreatments. However, the influence of consecutive pretreatments was smaller for the Cs-promoted catalyst.Only the conversion curves of the unpromoted catalyst will be shown here as they illustrate the general effects.1269 100 80 60 h 5 Y GO 20 c 323 423 523 T/K Fig. 1. CO conversion-us.-temperature curves for the non-promoted silver catalyst after consecutive pretreatments: (a) N,, 523 K; (b) 0,, 523 K; (c) 0,, 598 and 673 K.After loading, the catalysts were subjected to various (subsequent) pretreatments, first in inert and oxidizing atmospheres, thereafter in reducing atmospheres. Stable activities as a function of time were always measured over the complete temperature range, as long as the measuring temperature remained lower than about 520 K. Higher measuring temperatures sometimes led to lower conversion levels than previously measured at lower temperatures. To illustrate the strong influence of the pretreatment on the catalytic activity, the conversion of CO to CO, is plotted as a function of temperature in fig. 1. The following sequence was employed (measurement after each pretreatment): (1) N,, 523 K; (2) 0,, 523 K ; and (3) O,, 598 and 673 K. As can be seen, the catalyst clearly became more active during the sequence. After oxidation at the more elevated temperatures (pretreatment 3, curve 3), additional pretreatments in inert or oxidizing atmospheres, at the same or lower temperatures, did not alter the activity of the catalyst further.Ultimately, the catalyst was about a factor of four more active than the freshly loaded catalyst. The results will now be presented and discussed in more detail. Arrhenius plots as derived from the conversion us. temperature data provide interesting information. The experimental data often indicate a distinct break (giving rise to two activation energies). The pretreatments are only affecting significantly the pre-exponential factors of the Arrhenius equation. This has also been found in our previous investigation^.^? The data, therefore, were analysed using two fixed activation energies, uiz.40 and 60 kJ mol-1 and varying pre-exponential factors, k,. To facilitate comparison, in all Arrhenius plots the In k , values are given. In the figures the values are given in relative units because no correction has been made for the mass of silver. However, all measurements have been performed on the same samples, making comparison of the obtained values between the various pretreatments of each catalyst valid. First, we will present the Arrhenius plots obtained for the Cs-promoted catalyst, as they clearly illustrate the general observations and also will render information on the1270 InJEuence of Pretreatment on the Properties of Ag/cz-Al,O, Fig.2. Arrhenius plot for the fresh Cs-promoted catalyst after thermal treatment in oxygen at 523 K. At low temperatures oxidation with an activation energy of 40 kJ mol-I is exhibited. Pre-exponential factor in relative units. influence of the caesium promotor. Thereafter, we will return to the measurements on the pure silver catalyst of which some of the conversion curves are presented in fig. 1. As much of the same reasoning holds for both catalysts, the results obtained on the pure catalyst will be discussed more briefly. Fig. 2 shows the Arrhenius plot obtained for the fresh Cs-promoted catalyst after treatment in oxygen at 523 K. The catalyst exhibits an activation energy of 60 kJ mol-1 over almost the complete temperature range with only a very small part displaying an activation energy of 40 kJ mol-' at lower temperatures.When the Ag-Cs catalyst was subsequently thermally treated in hydrogen at 523 K, the Arrhenius plot represented in fig. 3 was obtained. The first consecutive pretreatment in nitrogen after the pretreatment in hydrogen gave rise to an identical plot and is also shown in the figure. Again the plot clearly exhibits a distinct break. Interestingly, now a much larger part is exhibiting an activation energy of 40 kJ mol-' at the lower temperatures and only a small part is displaying the activation energy of 60 kJ mol-' at more elevated temperatures. Also, it can be seen that the pre-exponential factors have substantially increased. The oxygen chemisorption and thermal desorption experiments described in Part 1 have unambiguously shown that the fresh catalyst contained carbon.Formation of carbon dioxide during interaction with oxygen and thermal desorption was demonstrated. As established previously,2 and from the it has been demonstrated that the presence of carbon in the surface layer of silver promotes the penetration of oxygen into the bulk. Also, in previous work2'* we have argued that the activation energy of 60 kJ mol-1 for the oxidation of CO with molecular oxygen is due to the more tightly bound penetrated oxygen atoms and that the activation energy of 40 kJ mol-1 can be ascribed to oxidation by adsorbed oxygen. The results presented in fig. 2, therefore, indicate that the presence of carbon causes the oxygen to be initially taken up mainly as bulk oxygen atoms. The thermal pretreatment after the first oxidation run has removed the carbon virtually completely from the surface.As a result, the activation energy of 40 kJ mol-'G. R. Meima et al. 0 . -Y 5 -2 -4- 1271 - T/K 500 400 350 2 I I I 1 1 1 1 1 1 1 l 1 1 I I I t 2.0 2.5 3.0 lo3 K / T Fig. 3. Arrhenius plot for the Cs-promoted catalyst after thermal treatment in hydrogen at 523 K (+) followed by pretreatment in nitrogen 673 K (0) subsequent to the pretreatment mentioned in fig. 2. Pre-exponential factor in relative units. is mainly exhibited in fig. 3. Furthermore, the pre-exponential factors of the reactions displaying the activation energy of 40 and 60 kJ mol-1 have increased also. The rise in the pre-exponential factor of the latter reaction is unexpected.When carbon is removed from the surface, it must be anticipated that the number of sites where oxygen can penetrate into the surface will be diminished. However, a thermal treatment of silver particles in hydrogen has been found to lead to metal particles of an almost spherical shape,8 whereas a thermal treatment in oxygen causes the particles to assume a more faceted shape.’-’l Consequently, thermal treatment in hydrogen will cause an increase in the proportion of crystallographic planes which are atomically more rough than the smooth (1 1 1)-plane in the silver surface. Since dioxygen can adsorb dissociatively on atomically more rough silver surface^,^ the number of sites where oxygen can be adsorbed in a state capable of rapidly reacting with CO will increase by thermal treatment in hydrogen.Kinetic and thermodynamic factors are governing the interaction of oxygen atoms with silver surfaces. On essentially pure silver surfaces, oxygen is dissociative1 y adsorbed onto atomically rough surfaces and defect sites on the silver (1 1 1)-surface. Penetration of oxygen atoms into the surface proceeds slowly at lower temperatures, and more rapidly at higher temperatures. At still higher temperatures, penetration of oxygen into the silver surface is thermodynamically unfavourable. As a result, only adsorbed atoms are present at high temperatures, and oxygen atoms dissolved into deeper layers of the silver are slowly released. As has to be expected, the reactivity of adsorbed oxygen atoms is higher than that of penetrated oxygen atoms.Accordingly, penetrated oxygen atoms are exhibiting an activation energy of 60 kJ mol-l, and adsorbed oxygen atoms of 40 kJ mol-l. The exact state of the penetrated oxygen is not exactly clear. However, it presumably pertains to a dissolved atomic form without substantial formation of Ag,O. This is especially well supported by the fact that prolonged pretreatments at high temperatures do not eliminate completely the penetrated oxygen. Moreover, Ag,O is thermodynamically unstable at these temperatures. If the adsorbed oxygen atoms would continue to contribute to the oxidation of carbon1272 InJEuence of Pretreatment on the Properties of Agla-Al,O, T/K 2.0 2.5 3 .O lo3 KIT Fig. 4. Arrhenius plot of the Cs-promoted silver catalyst after a second pretreatment in nitrogen at 673 K.Pre-exponential factor in relative units. monoxide, the sharp break in the Arrhenius plot of fig. 3 would not be evident. We therefore, must assume that the oxygen atoms adsorbed at defect sites on the (I 11)- surfaces and on atomically rough planes penetrate much more rapidly into the surface at temperatures above about 450 K. The rapid conversion of adsorbed atoms to atoms penetrated into the surface causes the sharp break in the Arrhenius plot. As stated above, it has been observed that the presence of carbon in the silver surface facilitates appreciably the penetration of oxygen atoms into the surface. The more rapid penetration explains the transition to the activation energy of 60 kJ at already lower temperatures, uiz.at about 370 K. This raises an interesting question about the reversibility of the conversion-us.-temperature plots which we will deal with next. In fig. 4 an Arrhenius plot is presented for the oxidation of carbon monoxide over the Ag-Cs catalyst after a second pretreatment in nitrogen at 673 K. It can be expected that a substantial fraction of the oxygen atoms dissolved in the near surface layer of the silver has been removed by these subsequent treatments in reducing and inert atmospheres. As can be seen, the activation energy of 40 kJ mol-' is displayed up to about 450 K as in fig. 3. However, the range of temperatures where the activation energy of 60 kJ mol-' is exhibited is now very small. At temperatures above about 470 K the conversion predicted by the Arrhenius equation with an activation energy of 40 kJ mol-1 shows up again.The explanation is that at higher temperatures penetration of oxygen atoms into the silver surface is thermodynamically unfavourable. As long as penetrated oxygen atoms are present, sites are provided where more oxygen atoms can still penetrate. When the oxygen content of the silver surface is, however, low, the penetrated oxygen is exhausted fairly rapidly and oxygen adatoms can no longer penetrate into the silver surface. Consequently, CO is oxidized again by adsorbed oxygen atoms. As a result, the course of the Arrhenius plot depends on the amount of oxygen atoms being dissolved into the silver before the start of the catalytic reaction. The small effect of the pretreatment on the catalytic activity of the caesium promoted catalyst, which could be anticipated from the sorption and thermal desorption experiments described in Part 1,l is evident from fig.5. Pretreatment in oxygen or in hydrogen at elevated temperatures followed by cooling in nitrogen leads to the same Arrhenius plot. The fact that no constant activity is obtained at high temperaturesG. R. Meima et al. 1273 500 40 0 350 2 I , , t l l l l I l l l 2.0 2.5 3 .O lo3 KIT Fig. 5. Arrhenius plots for the oxidation of CO over a Cs-promoted silver catalyst. The effects of thermal pretreatment in oxygen at 598 K, (+, x ) and 673 K (u), and in hydrogen at 673 K (0) followed by cooling down in nitrogen are shown. Pre-exponential factors in relative units. ly S -2 - - 1 1 1 1 1 I I I I I I 10’ KIT 2 .o 2.5 3.0 Fig.6. The effect of thermal pretreatment in hydrogen at 673 K on the Arrhenius plot for the oxidation of CO over the Cs-promoted silver catalyst. ( x , 0) cooled in H,; (0, +, a) cooled in N,. Pre-exponential factors in relative units. illustrates the variation in the amount of penetrated oxygen in the different experiments. In some runs the activity at high temperatures returns to the Arrhenius plot of 40 kJ mol-l, whereas in other experiments the penetrated oxygen is not yet completely consumed during the measurement. This is similar to the effects also previously observed in fig. 4. With catalysts containing much smaller silver particles, it was observed that thermal treatment in hydrogen leads to a rise in activity. However, if the catalyst was cooled down in nitrogen after treatment in hydrogen, an appreciably lower activity was Though the effects are smaller with the caesium-promoted catalyst containing12% Injluence of Pretreatment on the Properties of Aglcc-Al,O, Fig.7. Arrhenius plot for the oxidation of CO over the Cs-promoted silver catalyst. The catalyst was thermally pretreated in hydrogen at 673 K and cooled down in hydrogen, which led to a relatively small amount of dissolved oxygen. 0, increasing temperature ; 0, decreasing temperature. Pre-exponential factors in relative units. larger silver particles, qualitatively the same results were obtained (fig. 6). Cooling in nitrogen led to a significantly lower activity than cooling in hydrogen after pretreatment at 673 K.Since the removal of oxygen continues during cooling in hydrogen, the faceting at the edges and corners will be less extended. At the atomically rough surfaces created during the pretreatment in hydrogen and still remaining after cooling in hydrogen, slightly more adsorption and penetration of oxygen proceeds leading to higher pre- exponential factors. The drop in the pre-exponential factor at more elevated temperatures and the transition to the lower activation energy during some experiments is also evident from fig. 6. The irreversible character of the activity at higher temperatures is well illustrated by the results presented in fig. 7, where the Arrhenius plot obtained after a subsequent pretreatment in hydrogen is shown. The points indicated by filled dots were measured on decreasing the measuring temperature.As to be expected, the activity becomes lower after removal of the penetrated oxygen. The effects of hydrogen-hydrogen, or hydrogen-nitrogen pretreatment on the non- promoted catalyst are evident from fig. 8. As in the experiments dealt with in Part 1 of this paper, the variations are small. The Cs-promoted catalyst exhibited a similar behaviour, though the effects on the pretreatment were generally still smaller. With the non-promoted catalyst the drop in activity at low temperatures, which was observed with the catalyst with silver particles of 70 nm,4 is also found, as can be seen from the marked drop at about 350 K in fig. 9. Arrhenius plots after pretreatments in hydrogen and cooling in hydrogen or nitrogen are collected in fig.10. The lower activity after cooling down from 673 K in nitrogen with respect to the cooling down in hydrogen (fig. 9) can be noted again. The data reproduce well at low temperatures, whereas at high temperatures the drop in pre-exponential factor due to exhaustion of penetrated oxygen and the transition to the lower activation energy is again exhibited. To illustrate the relative stability of the Cs-promoted catalyst, it is interesting to consider the range of pre-exponential factors exhibited by both the promoted and non- promoted catalyst. With the non-promoted catalyst the pre-exponential factor (for the reaction with adsorbed oxygen exhibiting an activation energy of 40 kJ mol-l) varies1275 0' I I I 323 423 523 TIK Fig. 8.CO conversion-us.-temperature after consecutive pretreatments (after prior treatments in oxygen; see fig. 1) for the non-promoted catalyst. (a) H, 523 K, (b) H, 673 K, (c) H, 673 K and subsequent cooling down in nitrogen. TIK 500 400 350 2 .o 2.5 3.0 lo3 KIT Fig. 9. Arrhenius plot for the oxidation of CO over a silver catalyst with particles of about 1 pm after pretreatment in hydrogen at 673 K. No caesium promotion. Note the deactivation of the catalyst at low temperatures. Pre-exponential factors in relative units. from 1.7 x lo6 to 28.5 x lo6 gii min-l, and with the promoted catalyst from 3.85 x lo6 to 9.97 x lo6 g;: min-l. Whereas with the non-promoted catalyst the pre-exponential factor varies by a factor of 20, that of the Cs-promoted catalyst varies only by a factor of 2.5.The reaction with an activation energy of 60 kJ mol-' displays pre-exponential factors ranging from 1 x lo9 to 17 x lo9 g;: min-l for the non-promoted catalyst. The variation1276 Influence of Pretreatment on the Properties of Agla-Al,O, 2 0 -Y 5 -2 - 4 TIK 500 LOO 35 0 I I I I I I l l 1 I 2.0 2.5 3.0 lo3 KIT Fig. 10. Arrhenius plots for the oxidation of CO over a non-promoted silver-on-alumina catalyst containing silver particles of about 1 pm. The catalyst was thermally pretreated in hydrogen at 523 K and cooled either in hydrogen or nitrogen (0, +), and at 673 K and cooled in nitrogen (0). Pre-exponential factors in relative units. with both the promoted and the non-promoted catalyst involves a factor of 17. This factor is of the same order of magnitude for the reaction with adsorbed oxygen (40 kJ mol-l) for the non-promoted catalyst.We will now compare the present results obtained on catalysts containing large silver particles with our previous findings on catalysts with much smaller particles. In table 1 we have collected the data measured for the different catalysts (also see ref. 2 and 4). The data refer to the reaction of CO with the adsorbed oxygen species (40 kJ mol-l). Clearly, the pre-exponential factors are not related to the specific surface area of the silver particles. Being proportional to the specific surface area of the silver present, the activity of the catalyst with silver particles of 150-200 nm should have been higher than that of the catalyst containing silver particles of 1000 nm by a factor of about six; the pre- exponential factor of the catalyst with the 70 nm particles by a factor of no less than 16.The remarkably small difference observed for the various catalysts in activity per gram of silver can only be explained if the reaction is highly structure-sensitive and thus largely dominated by surface defects. The influence of defects has already been well established with catalytic reactions over oxides. For instance, experimental evidence has been obtained that the catalytic activity per unit surface area of copper oxide decreases with the particle size. This phenomenon was attributed to an easy annihilation of lattice defects in small crystallites.12 With metallic particles the influence of defects on the catalytic activity has been yet less well documented.However, it can be expected that with metallic particles, such as silver, an effect will only be observed at relatively large particle sizes. Generally, the high mobility of metal atoms over metal surfaces will bring about the rapid migration of grain boundaries out of clusters of metal particles, even with particles of, e.g. 50 nm. PashleyG. R. Meima et al. 1277 Table 1. Comparison of the pre-exponential factors of different silver-on-alumina catalysts pre-exponential factor/ lo6 min-l g;: particle size /nm maximum minimum ratio 70 50.4 3.6 16 15&200 10.4 0.5 20 1000 non-promoted 28.5 1.7 17 1000 Cs-promoted 10.0 3.9 2.5 and coworkers13 have demonstrated this to be true for gold particles, where the migration of grain boundaries out of coalesced particles was already observed to proceed smoothly at 573 K.Hence, it can be expected that lattice defects in pure silver particles will anneal rapidly, especially at high temperatures. Interestingly, it has been found that foreign atoms can decrease the rate of anneal of lattice defects in metal surfaces severely. Consequently, we envisage the effect of foreign atoms in silver particles to be twofold, i.e. the production of defects and the decrease in the (surface) mobility of the metal atoms. As a result, a relatively small amount of foreign atoms can severely affect the catalytic properties. This has been demonstrated for the catalyst containing the large silver particles onto which the Cs promotor was added. With silver particles larger than about 100 nm, annihilation of surface defects can be expected to proceed more slowly in view of the number of metal atoms involved and the distances the metal atoms have to cover.We therefore attribute the astonishingly high specific activity of large silver particles to the (surface) stabilization of defects. The minimum specific activity displayed by the particles of 150 to 200 nm is most probably due to the fact that lattice defects can still be annealed with particles of this size. The smaller specific silver surface thus causes the activity to be slightly lower than that of particles of 70 nm. With the particles of 70 nm, defects are clearly much less stable than with the very large silver particles. A break in the Arrhenius plot is only displayed with the fresh catalyst which contains carbonaceous imp~rities.~ When the carbon has been removed, the break is no longer exhibited.With these small particles, atomically more rough surfaces are dominating the catalytic activity, and not defects present on extended (1 1 1)- surfaces. As has to be expected, the break in the Arrhenius plots of the 70 nm particles is much less sharp. The atomically rough surfaces continue to contribute to the catalytic reaction. Consequently, adsorbed oxygen still contributes to the catalytic reaction, in contrast to the behaviour of the 1 pm particles. As soon as the carbon is removed, the activity due to defects disappears. This is apparent from a drop in the pre-exponential factor, also at lower temperatures. The activity of the freshly oxidized catalyst was higher than that of the catalyst thermally treated for prolonged periods of time.Also the observed instability in the activity at temperatures where annealing of surface defects can be expected to proceed is explained by the presence of defects. This also holds for the irreversible character of the conversion plots observed with measurements at more elevated temperatures. For a good understanding of the activity of silver in oxidation reactions, it is thus highly important to envisage that an undisturbed silver (1 1 1)-surface does not adsorb oxygen dissociatively. Albers et a1.l' have made this important observation on single silver crystals using ellipsometry to assess both the defect density and the oxygen coverage of the surface.Somewhat similar observations have also been made by1278 Influence of Pretreatment on the Properties of Agla-Al,O, Campbell,15 who studied the role of caesium as a promoter in the ethylene epoxidation reaction over a Ag (1 11) single crystal. Under reaction conditions the unpromoted Ag (111)-plane only showed a coverage of a few per cent of oxygen, whereas after Cs-promotion the surface coverage could approach a full monolayer. The promoting effect of caesium on the oxygen uptake is also evident from our results. As discussed above, the fraction of atomically rough surfaces where oxygen can be adsorbed and can penetrate into the silver lattice varies with the pretreatment. Since caesium stabilizes defects and, hence, enhances the ability of the surface to adsorb oxygen, the variation in activity is much smaller with the Cs-promoted catalyst. Any effect of the presence of caesium on the activation energies could not be established.This work indicates that only the pre-exponential factor is affected. Conclusions This study indicates that the activity of silver catalysts for the oxidation of CO with molecular oxygen is governed by the number defects in the silver surface. The pretreatment can influence the surface structure of the silver particles and hence the number of defects. More or less stable defects are established by carbon or the addition of caesium. The latter is not removed during oxidation, rendering the Cs-promoted catalyst more stable towards variations in the pretreatment. References I G. R. Meima, L. M. Knijff, R. J. Vis, A. J. van Dillen, F. R. van Buren and J. W. Geus (part I of this study), J. Chem. SOC., Faraday Trans. 1, 1989, 85, 269. 2 G. R. Meima, L. M. Knijff, A. J. van Dillen, J. W. Geus, J. E. Bongaarts, F. R. van Buren and K. Delcour, in New Developments in Selective Oxidation, European Workshop Meeting, 17-1 8th March 1986, Louvain-la-Neuve, ed. B. Delmon and P. Ruiz; Catal. Today, 1987, 1, 117. 3 G. R. Meima, R. J. Vis, M. G. J. van Leur, A. J. van Dillen, F. R. van Buren and J. W. Geus, J. Chem. SOC., Faraday Trans. 1, 1989, 85, 279. 4 G. R. Meima, L. M. Knijff, A. J. van Dillen, F. R. van Buren and J. W. Geus, J. Chem. Soc., Faraday Trans. I , 1989, 85, 293. 5 A. J. van Dillen, Ph.D. Thesis, Utrecht, 1977, chap. 4. 6 J. J. F. Scholten, J. A. Konvalinka and F. W. Beekman, J. Catal., 1973, 28, 209. 7 A. I. Boronin, V. I. Bukhtiyarov, A. L. Vishnevskii, G. K. Boreskov and V. I. Savchenko, Kinet. Katal., 1984, 25, 1510; Engl. transl., 1984, 25,1301. 8 B. E. Sundquist, Acta Metall., 1964, 12, 67 and 585. 9 G. E. Rhead and M. Mykura, Acta Metall., 1962, 10, 843. 10 R. Shuttleworth, R. King and B. Chalmers, Proc. Roy. SOC. (London), 1948, A193, 465. I1 A. J. W. Moore, Acta Metall., 1958, 6, 293. 12 J. W. Geus, in Preparation of Catalysts UZ, ed. G. Poncelet, P. Grange and P. A. Jacobs (Elsevier, 13 D. W. Pashley, Adv. Phys., 1965, 14, 327. 14 H. Albers, W. J. J. van der Wal, 0. L. J. Gijzeman and G. A. Bootsma, Surf. Sci., 1978, 77, 1. 15 C. T. Campbell, J. Phys. Chem., 1985, 89, 5789. Amsterdam, 1983), p. 1. Paper 8/006 12A ; Received 7th November, 1988