J. Chem. SOC., Faraday Trans. 1, 1986,82, 3315-3330 Cooperative Effects in Heterogeneous Catalysis Part 1 .-Phenomenology of the Dynamics of Carbon Monoxide Oxidation on Palladium Embedded in a Zeolite Matrix Nils I. Jaeger,* Karin Mollert and Peter J. Plath Institut f u r Angewandte und Physikalische Chemie, FB 2, Universitat Bremen, 0-2800 Bremen 33, Federal Republic of Germany Hysteresis behaviour, simple and complex oscillations showing a fractal pattern, long-term changes of the oscillations as well as periodic multistability have been uncovered during the oxidation of CO. Pd zeolites with different amounts of Pd particles per unit volume were used as a catalyst in order to study the synchronization process during the reaction. Zeolites with the highest Pd content showed the largest capability to oscillate, with amplitudes up to 95% of the conversion.Examination of the catalyst before and after the reaction leads to the assumption of a cyclic oxidation-reduction mechanism. Coupling of the individual particles is suggested by a concen- tration wave travelling across the surface. The analysis of complex temporal behaviour in heterogeneous catalytic reactions has led to new insights into the basic mechanism of catalysed reactions. The oxidation of carbon monoxide on Pt foils and wires and on Pd and Pt supported catalysts has been studied intensely in this context,l-ls beginning with the observations presented by Hugo and Jakubithl and Wicke and coworkers.2 Early models put forward to account for the observed reaction rate oscillations propose phase transitions regarding structure and reactivity of adsorbate layers of reacting molecule^,^^ l 1 9 l2 a point of view which has recently been supported by in-situ infrared spectroscopy on Pd supported catalysts.16 Direct observations of phase tran- sitions travelling across single-crystal surfaces of Pt were reported by Cox et al." under vacuum conditions.The idea of phase transitions involving the catalyst had been proposed by Wagner.18 A mathematical model of ideal storagelg involving a phase transition of the bulk of catalytically active crystallites has been developed that shows the essential dynamic properties observed in experimental systems. In this article we describe the phenomenology of heterogeneously catalysed CO oxidation and propose a reaction mechanism.In a subsequent paper (Part 2) we will use techniques developed in connection with chaos theory to create and study attractors from the computer-stored reaction data. Suggestions will then be made for a chemical interpretation of the observed results. Models often suggest that the bulk of a supported catalyst acts in synchrony with the dynamics of a single active centre undergoing phase transitions. Synchronized behaviour of all parts of the catalyst surface cannot be expected, as was demonstrated recently by highly resolved infrared thermography20 supported by i.r. spectroscopy.14 The coupling mechanism required for the cooperation of large numbers of active centres therefore must be specified. Coupling can be achieved by simple diffusive processes, e.g. heat transfer and/or diffusion of molecules participating in the reaction. Local temperature peaks have to be taken into account, especially in the case of exothermic oxidation reactions, and t Present address : Center for Catalytic Science and Technology, Department of Chemical Engineering, University of Delaware, Newark, Delaware 197 16, USA.33153316 Cooperative Efects in Heterogeneous Catalysis models should not be based on overall isothermal conditions established as a questionable constraint on the reaction. In the present work the phenomenology of the non-isothermal CO oxidation on a highly dispersed palladium phase supported within a zeolite matrix will be presented. The faujasite-type zeolite was chosen for three reasons. (i) It is a suitable carrier for the growth of a Pd phase with a narrow particle size distribution within the channels and cages of the matrix.21 (ii) It allows the preparation of catalysts with varying amounts of metal loading, adjustable by the ion-exchange process. (iii) Different metal particle sizes can be generated within the framework by applying well studied reduction routes.22 With this catalyst system it was possible to study the reaction not only in terms of its dependence on the temperature and the CO concentration in the feed, but also in terms of the number of Pd crystallites per unit volume and different metal particle sizes.The experimental results will be interpreted by phase transitions of the Pd crystallites, and a mechanism for their coupling will be suggested which will be evaluated in Part 2.Experiment a1 Preparation of the Catalyst The sodium form of zeolite X (faujasite) was prepared by hydrothermal synthesis, yielding an Si/Al ratio of 1 : 2 and a crystallite size of 10-20 pm. The sodium cations of the zeolite with the overall formula [(A102)86(Si02)l,,6] - nH20 were exchanged with an aqueous [Pd(NH3),]C12 solution ranging from 0.003 to 0.025 mol dm-3 at room temperature under shaking for 24 h. After washing the sample chloride-free it was dried and the Pd content was determined by atomic absorption spectroscopy (degree of exchange 7-43%, corresponding to 2.4-14.7 wt %). The metal phase was then prepared by an autoreduction process during the decomposition of the ammine complex, heating a sample in a fluidized-bed reactor up to 623 K (5 K min-l) for 16 h.According to the desired metal particle size either argon or oxygen was used, with a flow rate of 1.4 cm3 s-l at 1 x lo5 Pa. Details regarding the reduction mechanism have been Characterization of the Catalyst The Pd-loaded zeolite was characterized by X-ray diffraction using a Debye-Scherrer camera and electron diffraction, in addition to transmission electron microscopy prior to and following the catalytic reaction. Both the crystallinity of the zeolite and the reduction of the Pd cations to the metal were detected with these methods. The pore volume of the zeolites was checked in addition by measuring the nitrogen physisorption capacity of the sample, while the accessibility of the metal surface towards the reacting molecules such as CO was examined by the oxygen chemisorption capacity of the metal surface, assuming that both molecules would cover about the same surface area with respect to their cleft fractal The geometrical metal surface was estimated from transmission electron micrographs.The micrographs were taken from cuts of the zeolites embedded in epoxy resin. The metal phase with a narrow particle size distribution was located exclusively within the zeolite matrix even though the size of the Pd single crystals was found to exceed greatly the largest cages of the host lattice. The catalysts with a high Pd loading showed local damage to the zeolite framework after the reaction, which was confirmed by physisorption measurements. Details of the results obtained by transmission electron microscopy and Xps, and a discussion of the growth of large metal single crystals within the channels and cages of the zeolite matrix are given in ref.(21), (25) and (26). Depending on the gas atmosphere used in the autoreduction procedure a narrow palladium crystallite size distribution around 10 nm (in argon) or 4 nm (in oxygen) could be obtained. Plate 1 shows the catalysts A-C treated under argon, always with anJ. Chem. Soc., Faraday Trans. 1, Vol. 82, part I I Plate 1 Plate 1. Electron micrographs of (a) catalyst A with 14.7 wt % Pd, (b) catalyst B with 4.4 wt % Pd, (c) catalyst C with 2.4 wt % Pd and ( d ) catalyst D with 14.7 wt % Pd. The scale bar indicates 50 nm. N. I. Jaeger, K. Moller and P. J. Plath (Facingp.3316)N. I. Jaeger, K. Moller and P. J. Plath 3317 Table 1. Catalysts used in the experiments Pd crystallites average size number/ l0ls catalyst Pd (wt %) support /nm per g catalyst A 14.7 zeolite X 10 B 4.4 zeolite X 10 C 2.4 zeolite X 10 D 14.7 zeolite X 4 E 0.5 Al,O,/SiO, 2 (amorphous) 2.34 0.71 0.38 36.5 9.8 -r gas inlet 7- thermocouple connected with a t silver sieve preheating El! Fig. 1. Differential flow reactor. average Pd size of 10 nm but with a decreasing amount of Pd particles per unit volume. Catalyst D is a highly loaded sample treated under oxygen to give an average size of 4 nm. The catalysts used in the experiments are listed in table 1. Catalytic Measurements The Pd-loaded zeolite was pressed and ground up again. Pellets between 0.06 and 0.1 mm in diameter were used in the experiments.The reaction was carried out in a differential flow reactor under shallow-bed conditions. The catalyst rested either on a silver sieve or on a ceramic support (fig. 1). The temperature of the silver sieve is representative of the average temperature of the catalyst bed under reaction conditions and was measured3318 Cooperative Efects in Heterogeneous Catalysis 1 .o 2.0 co (vol %) Fig. 2. Influence of the reaction temperature on the hysteresis effect for catalyst C (50mg) at reaction temperatures T, = @, 423 ; + ,433 and 0 , 4 4 3 K. by a thermocouple connected to the silver support. The temperature of the feed was varied between 403 and 523K and controlled by a second thermocouple. Most experiments were carried out with either 20 or 50 mg of the catalyst.The carefully dried gas mixture containing between 0.2 and 5.0 vol% CO were prepared from a CO-N, test gas mixture (1 5 % CO) and synthetic air [Messer Griesheim, impurities < 0.1 vpm (volumes per million)]. A linear flow rate of 2.5 cm ~ ~ ~ ( 2 . 5 cm3 s-l) was used in most of the experiments (1 x lo5 Pa, residence time 5 s). The CO, and CO concentrations on the outlet of the reactor were continuously monitored by i.r. spectrometers (URAS 2T and URAS 3G, Hartmann und Braun, F.R.G.). Concentration changes with time could be resolved to within 1 s. The reaction data were digitized and stored on floppy discs with a PDPl 1 computer (DEC) for further analysis. For dehydration prior to the catalytic measurements the catalyst was heated to 600 K ( 5 K min-l) within the flow reactor and left at this temperature in streaming dry synthetic air (1 x lo6 Pa) for 16 h.Any appreciable activity of the carrier or the silver sieve in the parameter region where oscillations occur was ruled out by running blank measurements.N . I. Jaeger, K. Moller and P . J. Plath 3319 2oi / 15 i 7 7 /i + + 1.0 2.0 co (vol %) I Fig. 3. Influence of the number of Pd crystallites per g of catalyst on the hysteresis loop at T, =423 K for catalysts A (O), B (0) and C (+) (50 mg). Results Hysteresis of the CO Oxidation The Pd-zeolite system shows hysteretic behaviour during CO oxidation, similar to results from other authors using Pt catalysts for this rea~tion.~ Depending on the CO concentration in the feed a threshold value is found to exist for oscillating CO, production.Beyond this value the reaction rate drops to a low stationary state which does not change, even after 20 h. The high-activity branch can only be recovered by considerably reducing the concentration of the carbon monoxide or by increasing the temperature of the catalyst. Resulting hysteresis loops are depicted for different system temperatures in fig. 2 and for three of the catalysts which have the same Pd particle size of 10 nm but differ in the amount of Pd in fig. 3. The difference, AT, of the average temperature of the supporting silver sieve and the reactor temperature, which is proportional to the reaction rate, is plotted as a function of the CO concentration. Hysteresis behaviour can be observed for system temperatures in the range 403-443 K.The threshold and ignition points are found to rise to higher CO concentrations with increasing system temperature (fig. 2) and for increasing numbers of Pd crystallites per g catalyst (table 1 and fig. 3) under 110 FAR 13320 Cooperative Eflects in Heterogeneous Catalysis Table 2. Threshold value for the extinction of the high-activity branch of the reaction at 423 K as a function of the metal loading of the catalyst threshold value of turnover number Pd surface co (vol %) CO/Pd surface atom per s area catalyst /m2 per g catalyst (4 (4 (4 (4 A 7.38 1.20 1.60 0.15 0.21 B 2.22 0.78 1.20 0.33 0.50 C 1.18 0.78 0.57 otherwise constant conditions. Similar behaviour is found for repeated hysteresis loops at a given temperature, indicating a memory effect of the catalyst.The calculation of turnover numbers at the extinction points leads to the surprising result that the oxidation of a smaller number of CO molecules per Pd surface atom per second is needed to suppress oxidation on catalysts with a larger Pd surface area. These results are summarized in table 2, including the threshold values for succeeding hysteresis loops. The Pd surface area and the number of Pd surface atoms were estimated for Pd spheres with the experimentally observed average diameter of 10 nm by using the specific weight of bulk Pd (12 g cm-9. Running the system below temperatures of 433 K may result in spontaneous quenching of the reaction below the threshold value, which recovers without any parameter manipulation.The catalyst changes autonomously between 95,50 and 30 % conversion, and the different states are maintained unpredictably for between 30 min and several hours. The periodic instabilities are observed in the region of hysteresis behaviour between 403 and 433 K. A stepwise quenching is also found in this region.' Oscillation of the CO Oxidation Temperature oscillations and corresponding concentration oscillations occur along the high-activity branch of the hysteresis. At temperatures > 443 K no threshold value is observed, but stable oscillations persist up to 525 K with a level of > 95% conversion. The CO, oscillations represent depressions in the reaction rate, matching the temperature pattern. Both are inverse to the CO oscillations in the effluent gas.As a characteristic feature of the reaction, a fractal structure2' of the product oscillations could be observed. Fig. 4 depicts results representative of all the catalysts, here shown for catalyst B, at T, = 453 K as a function of the amount of catalyst. The oscillations show self-similarity, which means that the global pattern will repeat inside itself with a similar structure, but always on a smaller scale. A prerequisite for the unfolding of pronounced oscillations, and specifically of their fractal character, is the use of small amounts of catalyst. The oscillations obtained for charges in the ranges 7-20 mg and 50-100 mg are compared in fig. 4. A well developed fractal pattern can only be observed in the case of small charges of catalyst (7-20 mg).Here the increase in the oscillatory capacity is obvious: on using 100 mg of catalyst the conversion drops by 8 % at 0.40 % CO compared to a decrease of 50 % when using only 20 mg. Going to a slightly higher CO concentration with a small amount of catalyst may even result in a decrease of conversion by 95%. Total quenching was observed at concentrations around 1 % CO. The heights of the amplitudes during the reaction can be used as a relative measure of the system's capability to oscillate and gives an indication of the number of Pd crystallites that are synchronized at the same moment. The amplitudes themselvesN . I. Jaeger, K. Moller and P . J. Plath 0.401 3321 I 30 rnin Fig. 4. Influence of the amount of catalyst on the oscillation patterns of catalyst B at T, = 453 K and 0.40 vol % CO: (a) 7, (b) 20 and (c) 100 mg.I t represent a loss of activity with respect to the reference point of almost total conversion ( > 95 % ). In fig. 5 the maxima of the depression of CO conversion during the oscillations are shown for different catalysts as a function of CO concentration. Following curve A of the catalyst with the highest Pd content indicates that at low CO concentrations, e.g. 0.37 v o l x , only very small amplitudes for the CO, oscillations are visible, and with increasing time these are no longer resolvable by temperature measurements. With increasing CO concentration in the educt stream the amplitudes grow and the fractal structure builds up more clearly. Both the fractal structure of the oscillations and the amplitudes are most pronounced for medium CO partial pressures.For higher CO concentrations the amplitudes decrease again and the visual appearance of the observed pattern seems to be of a chaotic nature. At around 5 vol % CO a steady state of the reaction is established. The patterns stabilize within several minutes following an alteration of the experimental conditions. A characteristic long-term change in the catalyst properties did not interfere with the experiments using 50 mg of catalyst, and a similar succession of patterns could be established in the reverse order by decreasing the CO concentration again. If the different catalysts are now compared, distinguished by their increasing amounts of Pd crystallites per unit volume going from catalyst C to catalyst A (table l), an increase in the amplitudes and the evolution of the fractal structure can be observed.For comparison the data of the amorphous catalyst E are added, which contains even more Pd particles per unit volume, whereby the crystals are also of a smaller size (2 nm) (see table I), resulting in amplitudes of 75 % of the yield even at low CO concentrations. The reproducibility of the curves is within f 3 % conversion for different samples of the same amounts of the chosen catalysts. 110-23322 40- - 20 - - Cooperatbe Eflects in Heterogeneous Catalysis 100- 80- - n 8 60- c .r( h > - s 0 \ E 0-0 l l I ~ I l I I ~ l l l 1 1.0 2.0 co (vol %) Fig. 5. Maximum decrease in conversion during the oscillations as a function of the CO concentration and the number of Pd crystallites per g of catalyst (catalysts A, B, C and E, see table 1) at T, = 453 K (50 mg).The dependence of the largest amplitudes on the CO concentration for catalyst A is plotted for several temperatures in fig. 6. The amplitudes increase with decreasing temperature. Below 453 K hysteresis behaviour is found, and below 433 K the reaction is characterized by the appearance of multistable, aperiodic behaviour. The influence of the amount of the catalyst can again clearly be seen in fig. 7. The maximum depression of the CO conversion during the oscillations is compared between a 50 mg and a 20 mg charge of catalyst A. The formation of large-amplitude oscillations is much more pronounced for the small charges. Fig. 8 summarizes the experimental observations in the temperature-CO-concentration phase plane for 50 mg charges of catalyst A.If the experimental conditions were chosen such that large amplitudes could be expected by using a catalyst with a high Pd particle density and the temperature was near the region where hysteresis occurs, a decrease of the amount of catalyst to 20 mg led to an evolution of the oscillation pattern on a large timescale. Representative results obtained for catalyst A (20 mg, = 453 K, 0.37 vol % CO), left under constant conditions for more than 100 h, are depicted in fig. 9. The different patterns represent situations which remain stable for several hours. The fractal structure is established within 15 min and evolved during the first 20 h. The evolution can be characterized by an increase in the length of the largest amplitudes as well as by an increase of the period.After 20 h the maximum amplitude disappeared abruptly and the system was found to'4 \ y 3 / . 1 .O 2 .o co (vol%) Fig. 6. Maximum decrease in conversion during the oscillations for catalyst A for various reaction temperatures between 453 and 523 K. 100- 90- 80- - - - - I 473 '0 i \ a '@ 493 0 0 I \ 1 1 ' " ' ' ~ ~ " " ' ' 1.0 2.0 3.0 co (vol %) Fig, 7. Influence of the reaction temperature T, and the CO concentration in the feed on the mafima of the depression of CO conversion during the oscillations for different amounts of catalyst A: 0, 20 mg; 0, 50 mg. w w td w3324 0.35- 0.25- 0.15- Cooperative Efects in Heterogeneous Catalysis 250 - 240 230 220 210 u 200- iG 190 180 170 160 150- 140 130 - - no conversion Fig.8. Summary of the dynamic behaviour of catalyst A in the temperature-CO-concentration phase plane. I 1 30 min t 25 0.35 n * 3 0.25 6 0 . 3 5 " I 0.25 65 O.2SJ 98 I . I 1 18 min 6 min t Fig. 9. Long-term changes of the oscillation pattern of catalyst A by using 20 mg of the catalyst at T, = 453 K and 0.37 vol % CO. Total time on stream given in h.N. I. Jaeger, K. Moller and P . J. Plath 3325 50. 4 I I 60 min t I I I 60 min t '"1 + I 60 min I i" ( d ) t I 60 min I t Fig. 10. Multiplicity of states as a function of the CO concentration in the feed. 20 mg of catalyst D, T, = 453 K. (a) 0.85 (b) 1.10 and (c) 1.33 vol % CO; ( d ) quenching after concentration change of CO in the feed (arrow).3326 Cooperative Efects in Heterogeneous Catalysis approach a state of apparently chaotic oscillations which could be maintained for several days.The different sections of the time series represent trajectories of the system which remain stable for several hours. The initial pattern and a similar sequence of structure changes could be repeated by subjecting the same catalyst to the standard pretreatment conditions in dry air (600 K, 16 h). Performing the experiments under the same conditions but choosing catalyst D, which has a smaller particle size of 4 nm and a larger number of Pd crystallites per unit volume (see table 1) prevents the system from undergoing these changes on the long timescale. In this case different behaviour with periodic multistability is found at higher CO concentrations (fig.10). Again the formation of self-similar oscillations takes place, which is now interrupted by sharp, long-lasting quenchings of the conversion level starting at a concentration of 0.85 vol % CO [see fig. lO(a)]. Larger amounts of CO in the feed result in broadening of the low conversion level, and at 1.33 vol % CO an autonomous change in the conversion between 98, 50-60 and 10% is obtained [see fig. lO(c)]. The different states are maintained between 30min and several hours, increasing with the CO concentration. A decrease in the conversion to 10% is always found after concentration changes in the feed [see arrow in fig. lO(d)]. Catalyst E, also with a high Pd loading and small particle size, showed similar results between 453 and 407 K at 0.37 vol % CO.With the exception of reaction states with periodic multistability all observed oscillations were found to be stable against small perturbations in the experimental parameters (flow rate, CO concentration and temperature of the reactor). However, the patterns undergo quantitative changes following a mechanical rearrangement of the catalyst bed. Characterization of the Catalyst by X-Ray Diffraction The studied reaction was carried out in an excess of oxygen, and the pretreatment of the zeolite was made under flowing synthetic air at a temperature of 600 K in order to drive the water out of the zeolite cavities. Under these conditions it can be assumed that the metal phase may undergo oxidation to PdO, as was observed by several a ~ t h o r s .~ ~ - ~ ~ To confirm this, small amounts of catalyst A were collected at different steps of the pretreatment procedure in order to perform X-ray studies (fig. 11). The reference (pattern tl) shows the strong (lll), (200) and (220) reflections of the Pd phase, in addition to the zeolite pattern. After 2.5 h at 600 K the PdO phase started to appear (pattern tJ and was found to dominate the Pd intensities after 22 h (pattern t3), with the PdO reflections (1 0 1 ), (1 lo), (1 12), (103) and (202). The catalytic measurements were therefore started with a catalyst containing both a Pd and a PdO phase. From the line-broadening of the reflections in the X-ray diffraction patterns the size of the particles could be estimated, and were in agreement with the results obtained by transmission electron microscopy (table 1).X-Ray diffraction patterns obtained for samples of the catalyst taken in the course of the catalytic reaction, on the other hand, showed either the reflections of the metal and the metal oxide simultaneously (although with varying relative intensities) or in several cases the Pd metal phase alone. The latter observation points towards the reducibility of the PdO phase under reaction condition, because decomposition of PdO does not take place below at least 1073 K.31732 The PdO phase can be almost completely reduced by switching from synthetic air in the reaction gas mixture to N, as a carrier. The reduction of PdO starts at relatively low temperatures (370 K), as could be demonstrated by temperature-programmed reduction of the samples.3327 I .. . . 1 . . . , l l l , , r . . . , 1 , , , , 1 , , , I I , . . , I . . , 0 10 20 30 01" Fig. 11. X-Ray diffraction pattern of catalyst A after treatment in synthetic air. (tJ Sample taken during the temperature programme (5 K min-l) at 473 K; (tz) sample taken after 2.5 h at the final temperature (600 K); (t3) sample taken after 22 h at 600 K. x , Pd; 0, PdO. Discussion The first discussions of CO oxidation oscillations mainly concerned kinetic aspects of the adsorbates. Ideas such as the transformation of an inactive linear bonded CO to the reactive bridge-bonded C0,l a switch between the Langmuir-Hinshelwood and Eley- Rideal mechanisms33 or the key role of a slowly desorbing intermediate34 stimulated further investigations.In-situ studies performed with F. tir., u.h.v. measurements on single crystals or i.r. thermography now indicate that the catalyst itself undergoes changes during the oscillations. Adsorbate-induced transitions found on Pd and Pt catalysts,ls* l7 which may spread non-uniformly over the surface,l49 2o have to be taken into account. In the present work we found evidence of an alternating oxidation and reduction of the Pd metal in the zeolite during the reaction. The starting material, consisting of a mixture of Pd and mainly PdO, turned in some cases into a catalyst with only the Pd phase, as examined after the reaction. Both the formation of PdO and its transformation into Pd were shown to be possible under the reaction conditions. This confirmed the in-situ diffraction measurements of Bergeret et al., who observed the complete oxidation3328 Cooperative Eflects in Heterogeneous Catalysis of 2-2.5 nm Pd crystallites supported in a zeolite matrix in the temperature range 450-500 K.35 The experimental results, however, allow no correlation of the timescales of the redox reaction involving the Pd phase with the timescale of the observed oscillations.It is well known that a Pd crystal shows different types of surface transformation, being either covered only with oxygen or at the same time with C0,36 and an enhanced reduction of PdO was in fact observed in the presence of 0,.37 As a result, the redox cycle during the oscillation may proceed much faster than expected from the reduction and oxidation experiments with the coadsorption of both components, CO and 0 on the surface, yielding a higher surface partial pressure.The transmission electron micrographs taken after the reaction support this assumption in another way : only the catalyst that reacted under large oscillations showed blurred images of the metal phase and local distortions of the zeolite. This might be caused by the cyclic change of volume of ca. 60% from Pd to PdO. These observations form the basis of our interpretation of the observed oscillations, where a phase transition (and its reversal) is proposed between a catalytically active Pd metal phase and an inactive PdO phase. Models of cyclic oxidation and reduction have been proposed in order to explain the oscillating oxidation of H, on Pd wires and Ni f0ils,~~,3~ and for the CO oxidation on Pd In the case of Pd-loaded zeolites the bulk of the metal phase can be transformed into stoichiometric PdO in the course of the reaction.No sintering or redispersion of the mixed metal-metal-oxide phase can be observed in the electron micrographs taken from the catalyst after reaction. This is supported by the same line-broadening in the observed X-ray patterns for both the Pd and PdO phases. The possibility of oxidizing highly dispersed Pd to PdO at relatively low temperatures during prolonged pretreatment of the catalyst in air, and the fact that the activity of the catalyst is established under oxidizing conditions, allow the conclusion that the active phase of the catalyst consists of an oxide phase, at least on the surface of the Pd crystallite.The catalyst surface, as well as in our case the bulk of Pd crystallites, can be regarded as a store for oxygen. At a certain threshold value a new phase is formed with little activity for CO conversion. From the experiments and from the literature data43 the surface of the bulk PdO appears to be the inactive phase, because no further adsorption of oxygen is possible on the surface. Our proposed scheme suggests the diffusion of adsorbed oxygen into a Pd crystallite in the course of the strongly exothermic reaction between the dissociatively adsorbed 0, and the molecularly adsorbed CO. The oxidation of Pd is favoured in the state of high catalyst activity when the local overheating of a metal crystallite supported on a matrix of low heat conductivity can be considerable, according to Ruckenstein and Petty.44 Following the phase transition to PdO, i.e.the transition to a state of low activity, the local temperature drops, the adsorption of CO is favoured and the reduction of the oxide by CO is the only reaction route until the original state of high activity is re-established. This reaction cycle is summarized in the following reaction equations. Overall reaction: CO +to, + CO,. (1) Active phase of a Pd crystallite (adsorption of reactants and reaction): Pd + CO + Pd/CO Pd+$O, + Pd/O Pd/O + Pd/CO + 2Pd + CO,. (2) Oxidation of Pd and transition to an inactive palladium oxide phase: Pd/O + PdO, 0 < x < 1 PdO,+(l-x)O+PdO.N. I. Jaeger, K. Moiler and P. J. Plath 3329 (3) Reduction of the inactive palladium oxide phase and transition back to the active phase of a Pd crystallite: PdO + CO + PdO/CO PdO/CO + Pd + CO,.The outlined model describes a cyclic reaction mechanism for a single catalyst particle. The average number of Pd crystals in the catalysts is ca. 1015, which would yield an apparent steady-state conversion when acting in a totally uncorrelated way, averaging all small-scale oscillations. Almost no discussion can be found in the literature concerning the synchronization of a large number of catalytically active centres. Our results indicate that an internal geometrical factor of the catalyst is responsible for the coordination of events. The application of zeolites as a support has made it possible to vary the amount of Pd particles per unit volume by keeping the particle size constant.In this way the distance between neighbouring crystallites could be decreased, making heat transfer or the diffusion of reactants easier. As a result the observed amplitudes of the oscillations increased, indicating a synchronization of more particles (see fig. 5). The process itself may be pictured as a concentration wave travelling across the catalyst surface : the trigger for the oscillations can be assumed to be one particle or a group of particles being oxidized earlier than others owing to a statistically higher coverage of reactants, a slightly smaller particle size or for other reasons. Their oxidized surface is no longer able to convert the succeeding molecules, since the crystallites are too hot to adsorb CO and are unable to adsorb further oxygen.The resulting local excess of feed molecules begins to diffuse to the still active neighbouring particles, driven by a small-scale concentration gradient. These crystallites can consequently increase their reaction rate until they undergo the same transition to PdO, thus becoming in phase with the timing particle. Depending upon the distance between the particles, more or less time will be taken before the concentration wave becomes diluted by diffusion in space. In the meantime the starting particle is cooled, covered with CO and reduced again, shortly followed by its neighbours. This same mechanism provides an explanation for the observed quenching phenomena, where the catalyst with the highest number of Pd particles per unit volume was quenched by a smaller number of CO molecules per Pd surface atom.The stronger coupling in this case overwhelms the factor of a much higher Pd surface area. The most pronounced decrease in conversion during the oscillations was found with catalyst E (see fig. 5), which contains the largest number of Pd particles among the compared samples. This result may be influenced by (in addition to the short distances between neighbouring particles) the small size of the metal particles, which are much more easily oxidized and reduced. Synchronization of the catalytic activity across the catalyst charge via heat transfer through the supporting disc seems to be less important in this reaction, since the replacement of the highly conducting silver sieve by a ceramic plate did not influence the oscillation patterns.Even a division of the catalyst charge into spatially separated parts showed no effect. The proposed reaction cycle leaves unexplained the long-term changes in the state of the experimental system and the possibility of restoring the original pattern by repeating the pretreatment of the aged catalyst in dry air. The ageing of the catalyst apparently involves no irreversible loss of activity. As a possible mechanism the deposition of carbon species on the surface of the Pd crystallites via the Boudouard reaction (2CO + C+CO,) at local overheated sites may be suggested. Precursors for coke deposition due to residual hydrogen in the carbon monoxide may also be taken into account.3330 Cooperative Eflects in Heterogeneous Catalysis A model involving transitions between phases of different catalytic activity yields oscillations in the case of a single oscillating system.This has been demonstrated by the mathematical and numerical analysis of a model of ideal ~t0rage.l~ The model involves transitions in the catalytic activity of an independently reacting crystallite as a function of the amount of chemical species stored. This simple model does not account for the complex structure of the observed oscillations. A numerical analysis of the experimental data, as the basis for the development of a suitable model which accounts for the synergetic behaviour of large numbers of a catalytically active crystallite or ensembles of crystallites, will be presented in Part 2 following proposals by Dress et al.45 We are grateful for financial support from the Stiftung Volkswagenwerk (AZ 1/38 880). References 1 P.Hugo and M. Jakubith, Chem. Ins. Tech. 1972,44, 383. 2 H. Beusch, P. Fieguth and E. Wicke, Chem. Ins. Tech., 1972,44,445. 3 M. Sheintuch and R. A. Schmitz, Catal. Rev.-Sci. Eng., 1977, 15, 107. 4 W. Keil and E. Wicke, Ber. Bunsenges. Phys. Chem. 1980,84, 377. 5 E. Wicke, Chem. Ing. Tech., 1974, 46, 365. 6 J. Rathousky, J. Puszynski and V. Hlavacek, Z . Naturforsch., Teil A , 1980,35, 1238. 7 N. I. Jaeger, K. Moller and P. J. Plath, 2. Naturforsch., Teil A, 1981, 36, 1012. 8 W. Adlhoch, H. G. Lintz and T. Weisker, Surf. Sci., 1981, 103, 576. 9 D. Barkowski, R. Haul and U. Kretschmer, Surf. Sci., 1981, 107, L329.10 P. C. Liao, E. E. Wolf, Chem. Eng. Commun., 1982, 13, 315. 11 G. Ertl, P. R. Norton and J. Rustig, J. Phys. Rev. Lett., 1982, 49, 177. 12 M. P. Cox, G. Ertl, R. Imbihl and J. Riistig, Surf. Sci., 1983, 134, L517. 13 J. Puszynski and V. Hlavacek, Chem. Eng. Sci., 1984,39,681. 14 D. J. Kaul and E. E. Wolf, J. Catal., 1985, 91, 216; 93, 321. 15 N. I. Jaeger, K. Moller and P. J. Plath in Springer Series in Synergetics, Vol. 29, Temporal Order, ed. L. Rensing and N. I. Jaeger, p. 96. (Springer, Berlin, 1985). 16 D. Bocker and E. Wicke, ref. (15), p. 75. 17 M. P. Cox, G. Ertl and R. Imbihl, Phys. Reu. Lett., 1985,54, 1725. 18 C. Wagner, Bet. Bunsenges. Phys. Chem., 1970,74,401. 19 A. Dress, N. I. Jaeger and P. J. Plath, Theor. Chim. Acta, 1982, 61, 437. 20 J. R. Brown, G. A. DNetto and R. A. Schmitz, ref. (15), p. 86. 21 D. Exner, N. I. Jaeger, K. Moller, R. Nowak, H. Schriibbers, G. Schulz-Ekloff and P. Ryder, Stud. 22 D. Exner, N. I. Jaeger and G. Schulz-Ekloff, Chem. Ing. Tech., 1980,52, 734. 23 D. Exner, N. I. Jaeger, K. Moller and G. Schulz-Ekloff, J. Chem. SOC., Faraday Trans. I , 1982,78,3537. 24 P. Pfeifer and D. Avnir, J. Chem. Phys., 1983,79, 3558. 25 G. Schulz-Ekloff, D. Wright and M. Grunze, Zeolites, 1982, 2, 70. 26 D. Exner, N. I. Jaeger, G. Schulz-Ekloff and P. Ryder, Proc. 6th Int. Zeolite Conf. 1983, ed. A. Bisio 27 B. B. Mandelbrot, The Fractal Geometry ofNature (Freeman, New York, 1983). 28 J. M. Guiot, J. Appl. Phys., 1968, 39, 3509. 29 E. Ruckenstein and J. J. Chen, J. Catal., 1981, 70, 233. 30 T. W. Orent and S. D. Bader, Surf. Sci, 1982, 115, 323. 31 S. I. Ginzberg, Analytical Chemistry of Platinum Metals (Keter, Jerusalem, 1975). 32 ci. Bayer and H. G. Wiedemann, Thermochim. Acta, 1975,11, 79. 33 R. Dagonnier and J. J. Nuyts, J. Chem. Phys., 1976, 65, 2061. 34 G. Eigenberger, Chem. Eng. Sci., 1978,33, 1263. 35 G. Bergeret, P. Gallezot and B. Imelik, J. Phys. Chem., 1981,85, 411. 36 T. Engel and G. Ertl, Ado. Catal., 1979, 28, 2. 37 P. V. McKinney, J. Am. Chem. SOC., 1932,54,4498. 38 Z . Kurtanjek, M. Sheintuch and D. Luss, J. Catal., 1980, 66, 11. 39 K. Rajagopalan, M. Sheintuch and D. Luss, Chem. Eng. Commun., 1980, 7, 335. 40 J. E. Turner, B. C. Sales and M. B. Maple, Surf. Sci., 1981, 109, 591. 41 J. E. Turner, B. C. Sales and M. B. Maple, Surf. Sci., 1981, 103, 54. 42 B. C. Sales, J. E. Turner and M. B. Maple, Surf. Sci., 1982, 114, 381. 43 P. V. McKinney, J. Am. Chem. SOC., 1933, 55, 3626. 44 E. Ruckenstein and C. A. Petty, Chem. Eng. Sci., 1972, 27, 937. 45 A. Dress, M. Gerhardt, N. I. Jaeger, P. J. Plath and H. Schuster, ref. (15), p. 67. Surf. Sci. Catal., 1982, 12, 205. and D. H. Olson (Butterworths, Guildford 1984), p. 387. Paper 511840; Received 22nd October, 1985