Catalytic oxidation of methane over palladium ... - Science Direct

Applied Catalysis, 68 (1991) 301-314 Elsevier Science Publishers B .V ., Amsterdam

301

Catalytic oxidation of methane over palladium supported on alumina Effect of aging under reactants Patrick Briot and Michel Primet* Ins titut de Recherches sur la Catalyse, Laboratoire Propre du CNRS, Conventionne a . l'Universite Claude Bernard Lyon 1, 2 Avenue Albert Einstein, 69626 Villeurbanne Cddex (France), tel. (+33-72) 445378, fax . (+33-72) 445399. (Received 7 August 1990, revised manuscript received 8 October 1990)

Abstract A Pd/A12O 3 catalyst was aged at 600 °C under a methane, oxygen and nitrogen mixture with an oxygen to methane ratio of 4 . The aged catalyst was more active than the fresh one, especially at low temperature (below 400'C) . The temperature of half conversion was decreased by WC . Aging led to a decrease of the metal dispersion: the palladium particle size increased from 7-16 nm . The turnover number (activity per surface palladium atom) was strongly enhanced (factor 20 at 400'C) on the aged sample_ In that sense, catalytic combustion of methane must be considered as a structure-sensitive reaction . The reactivity of adsorbed oxygen towardss hydrogen increased with the metal particle size; this parallels the increase in the. rate of methane oxidation with particle size . Temperature-programmed oxidation measurementssuggested that palladium was in the form of bulk palladium oxide for reaction temperatures in excess of 400°C . The formation of bulk PdO appears to be also metal particle size dependent . An instability of the conversion as a function of time at a given temperature was observed. The instability was associated with the slow formation of bulk PdO . Keywords: palladium/alumina, methane oxidation, ageing, particle size .

INTRODUCTION

Supported noble metals are very active catalysts for the total oxidation of hydrocarbons . In the case of the complete oxidation of methane, palladium and platinum are the most widely used because of their high activity [ 11, In a previous paper [2], we have shown that the catalytic combustion of methane over platinum supported on alumina is a structure-sensitive reaction, the activity per platinum surface atom increasing with the size of the metal particles . Similar conclusions have been reached for the catalytic oxidation of propene [3 ] and of n-heptane [41 over alumina supported platinum . Cullis et al . [5] found that the rate of methane oxidation, expressed per gram of catalyst was independent of the metal particle size . Very recently, Hicks and co-workers [6,7] have studied the behaviour of supported platinum and palladium in the oxidation of methane at moderate temperature, i .e ., at ca . 300°C 0166-9834/91/$03 .50

© 1991 Elsevier Science Publishers B .V .

302

and at very low conversion levels . On both metals the reaction was found to be structure-sensitive . In addition, in the case of supported palladium [71, they found that the palladium particles were oxidized during the reaction . The present study is concerned with the catalytic combustion of methane over palladium supported on alumina . Its aim was to give more insight into the state of palladium during the reaction and to correlate the changes in catalytic activity with the reactivity of the various oxygen species present on the catalyst . EXPERIMENTAL

Catalyst preparation

The support was an alumina SCM 129 from Rhone-Poulenc consisting of a mixture of the gamma and delta forms, having a B .E.T. surface area of 107 m 2 g ` and a pore volume of 0 .6 em' g- 1. The palladium (II) chloride precursor was dissolved in an excess of concentrated hydrochloric acid in order to form the H0PdC14 compound . Excess hydrochloric acid and water were removed under reduced pressure until dryness . The solid obtained was finally dissolved in water . The aqueous H 2 PdCi, solution was added to a suspension of the support in water. Water was eliminated by using a rotary evaporator under reduced pressure . The sample was dried overnight at 110°C, then calcined for a few hours under a flow of nitrogen at 500°C and finally reduced by hydrogen for 12 h at the same temperature . The palladium content deduced from chemical analysis was 1 .95 wt.-% . Chemisorption measurements

These were performed at room temperature and also between -78 and 25 ° C . The procedure adopted for the room temperature measurements has been described elsewhere [2,8 ] . A Bourdon gauge (from Texas Instruments) was used for pressure measurements in the range 0-15 Torr (1 . Torr = 133 .322 Pa) . The amounts of gases irreversibly adsorbed were determined by extrapolating the adsorption isotherms to zero pressure . In the case of hydrogen adsorption, the formation of palladium hydride could occur in the range of pressure investigated . The amount of hydrogen involved in this process was measured by the back-sorption method [9 ] and subtracted from the amount of hydrogen found in the previous experiment . Titration of oxygen adsorbed onto palladium by gaseous hydrogen was also performed below room temperature . The catalyst (weight close to 500 mg) was introduced in a U-shaped reactor and reduced by hydrogen at 500°C . After cooling to 25°C and removing the gaseous hydrogen, a monolayer of oxygen adsorbed onto the palladium particles was achieved by oxygen treatment at the same temperature . The sample was then cooled under a flow of argon to - 78 0C

3 03

(dry ice-acetone mixture) . Argon was replaced by hydrogen diluted in argon (1 :99, v/v) passing through the catalyst with a flow rate of 1 .1 1 h"' . The composition of the effluents was determined by gas chromatography using a catharometric detector. Hydrogen consumption was measured at -78° C, then the dry ice-acetone trap was removed . The amount of oxygen remaining adsorbed was titrated by hydrogen between -78 and 25°C . Transmission electron microscopy (TEM)

Direct observations of the samples were made using a JEOL 100 CX microscope (resolution ca . 0.3 nm) . Reduced samples were dispersed in anhydrous ethanol by ultrasound . One drop of the suspension was deposited onto a carbon-coated copper grid and the ethanol was evaporated . Temperature-programmed oxidation measurements (TPO)

The catalyst (weight close to 300 mg) was loaded in a U-shaped reactor made of silica . It was reduced by hydrogen at 500'C, then flushed with helium at the same temperature in order to obtain a metal surface free of adsorbed hydrogen. The sample was cooled to 25'C under helium . Helium was replaced at 25'C by diluted oxygen in helium (1 : 99, v/v) and the oxygen consumption was measured using a catharometric detector . The temperature was then increased under the oxygen-helium mixture to 600°C with an heating rate of 8°C min - '. The temperature of the catalyst and oxygen consumption were measured during the heating up to 600°C . Catalytic activity measurements

The conditions used for the catalytic activity measurements have been described in a previous paper [2 ] . The catalyst (weight between 200 and 600 mg) was reduced in situ at 500'C under hydrogen . It was submitted to a 6 .3 1 h - ' flow of methane, oxygen and nitrogen (1 :4 :95, v/v/v) . The catalytic activity was measured for 5 h at each temperature . The temperature was increased by steps of 25 ° C . The effluents were analyzed by gas chromatography after separating the components using a (4 m length, 1/4 in . diameter) Carbosieve S column maintained at 120°C . Hydrogen was chosen as carrier gas ; carbon monoxide eventually formed, and the remaining methane and carbon dioxide were converted into methane by means of a,Ni/MgO methanation catalyst maintained at 420'C . The conversion was deduced from the partial pressures of the products (P,,(, and P,~02 ) and the partial pressure of unconverted methane (PCH,) according to the equation :

3 04

conversion- (Peo+Pco z ) / (Pco+Pco2+Pcu4 ) From the conversion, the rate (r) of methane oxidation (in moles of methane per h and per mole of introduced palladium) was calculated . As previously mentioned [2J, r has to be considered as an "average rate" deduced from the conversion rather than a true rate . Because of the high exothermicity of methane oxidation (AH° 29s = -800 kJ mol'), strong dilutions of methane and oxygen in nitrogen were used . In addition, the influence of the weight of the catalyst was studied in the 200-600 mg range . At a given temperature, the conversion was found to be proportional to the weight of the catalyst, suggesting that the reaction is under kinetic control . RESULTS

Catalytic oxidation of methane After reduction under hydrogen at 500'C and purging under a flow of nitrogen at the same temperature, the Pd/AI,O, sample was cooled to between 200 and 300° C, Nitrogen was replaced by the methane, oxygen and nitrogen mixture and the catalytic activity was measured according to the procedure described in the experimental part . Fig . 1A shows the variation of the rate (r) of methane combustion (in moles of methane converted per h and per mole of palladium) as a function of the temperature . In the following sections, the state of the freshly reduced catalyst will be called state I . Measurable conversions were observed above 300 ° C and a 100% conversion level was reached at around 480 ° C . Carbon monoxide was never detected, At a given temperature, the evolution of the catalytic activity with time depends on the temperature of the catalyst . Below 390°C and above 460°C, the activity remains stable during the 5 h interval time . On the contrary, in the 390-460'C interval, the catalytic activity continuously increases as a function of the time of the reaction . The evolution of the activity at a given temperature is illustrated by the vertical segments in Fig . 1A . For clarity, that part of the curve corresponding to an unstable activity with time was plotted by a dotted e. After a 100% conversion level was reached, the temperature of the catalyst was increased to 600°C and the system was maintained under the reactants mixture for 14 h . Then the sample was purged under a flow of nitrogen and cooled to between 200 and 300'C . The state of the catalyst obtained after such a treatment will be called state II in the following parts of the paper . The catalytic activity of Pd/Al2O3 in state II was measured in conditions identical to those used for Pd/A120, in state I . Fig . 1B shows the evolution of

3 05

Fig . 1 . Rate (in mote of methanee converted per hour and per mole of introduced palladium) of methane oxidation over a Pd/Al 2 O, catalysts in state I (A curve) and in state II (B curve) .

the rate of methane oxidation vs . the temperature of reaction . The reaction started at ca . 200°C instead of 300°C for the sample in state 1 . A 100% conversion level was reached at around 480°C . As for the sample in state I, an instability of the catalytic activity as a function of time of reaction was observed in a defined range of temperature ; the catalytic activity did not change with time below 340'C and above 400°C . The increase of the rate of reaction at a given temperature is illustrated by the vertical segments in the curve plotted by a dotted line . From 200 to 450 ° C, i .e., over almost the entire temperature range, the catalyst is more active in state II than in state I . The aging of the Pd/A1201 sample was also performed in the absence of methane, i .e., heating under a mixture of 4 vol .% oxygen in nitrogen at 600 ° C for 14 h . The catalytic activity of the resulting sample was similar to that of the Pd/Al oOa catalyst in state II, no increase in catalytic activity was noticed . The ratio of the catalytic activity in state 11 (r 1,) over the catalytic activity in state I (r 1) as a function of the reaction temperature is given in Fig . 2. Values as high as 35 were observed for low reaction temperatures . By comparison, the

3 06

Fig . 2 . Variations of the r„/r, ratio as a function of the reaction temperature . r1 and r 11 are the rate of methane oxidation for catalysts in state I and state II respectively expressed in moles of methane convertedper hour and per mole of introduced palladium, A -Pd/AI 2 O 3 catalyst ; B=Pt/Al 9 O, catalyst .

evolution of the same ratio is given for a Pt/A1 2 0„ catalyst [2] studied under the same conditions . The values of the apparent energy of activation were deduced from Arrhenius plots for low conversion levels (below 5%) and at constant activity with time . Values of 104 and 74 kJ mol' were found for the sample in state I and state 11, respectively . Hydrogen reduction of the Pd/A1 203 sample in state II at 500`C, prior to the catalytic activity measurements, did not modify the behaviour of the catalyst, i .e ., the rate of reaction was still higher than for state 1 . Dispersion of the metallic phase Hydrogen-oxygen titrations performed at room temperature led to a palladium dispersion of ca . 16% for the Pd/A12 0, catalyst in state I . Assuming a density of 1 .27 10' y palladium atoms per m' [10], a mean diameter was calculated for the palladium particles close to 7 nm . Fig. 3 shows a TEM micrograph of the Pd/A1 2 0 3 sample in state I . The metal particles sizes are between 4 and 8 nm (the majority of them had a diameter in the vicinity of 5 ran) . By comparison, hydrogen-oxygen titrations on the sample in state If yielded a metal dispersion close to 7%, and a corresponding mean diameter close to 16 run . A TEM micrograph of the Pd/A1 2 0 3 sample in state 11 is given in Fig . 3B .

30 7

Fig . 3 . T .E .M. micrographs of Pd/A1,O, samples . A=state I ; B=state IL -

The metal particles are not spherical, some of them are faceted. Their diameters vary between 10 and 20 nm. Using the chemisorption data, the catalytic activity per palladium surface



3 08

N h-1 1000

800

600

400

200

200

0

400

500

600

Fig . 4 . Variations of the turnover number (N in h - ') as a function of the reaction to for Pd/AI20 ;, samples in state I (A curve) and in state 11 (11 curve) .

erature

atom or turnover number (T.O.N.) can be calculated . Variations in the T .O.N . of the Pd/Al 2O3 sample in states I and II, given as a function of the reaction temperature are plotted in Fig . 4. Titration measurements at tow temperature

The reduced samples were covered with a monolayer of oxygen at room temperature . They were cooled under Ar to - 78'C. At this temperature, hydrogen diluted in argon was passed through the sample and the hydrogen consumption was measured at -78°C . After recovering the baseline, the cooling mixture was removed and the temperature of the sample was rapidly increased up to room temperature . Hydrogen consumption as well as the temperature of the

30 9

sample were recorded during this process . According to our calibration procedure, the amounts of hydrogen used for the titration of adsorbed oxygen at -78°C and between - 78° and 25°C were determined. In the case of the Pd/A12 03 sample in state I, titration of adsorbed oxygen . During the process of heating to room temperature, a did not occur at -78 8'C peak of hydrogen consumption was observed at 13'C . It corresponded to 44 .6 pmol of hydrogen per gram of sample ; according to the stoichiometry of titration, a palladium dispersion phase close to 16 .2% was observed . For the sample in state II, no hydrogen consumption was detected during heating from -78'C to room temperature . The titration of adsorbed oxygen occurred at -78°C since a single peak was observed at this temperature, its surface corresponded to 22 .05 ymol of hydrogen per gram of sample and to a dispersion of the metallic phase close to 8% . The values of the dispersion deduced from these experiments are in good agreement with those deduced from chemisorption measurements at room temperature. Thus it appears that all the surface palladium atoms were involved in the titration process performed between -78 and 25'C . The reactivity of adsorbed oxygen is strongly enhanced for palladium in state II as compared with state I . In state II, adsorbed oxygen was fully titrated by hydrogen at -78°C whereas no reaction occurs at this temperature for palladium in state I. Temperature programmed oxidation measurements

After reduction at 500°C, adsorbed hydrogen was removed by treatment under helium at the same temperature . At 25°C, a flow of oxygen diluted helium was passed over the sample . The consumption of oxygen was measured at room temperature. After the baseline was recovered, the temperature of the catalyst was increased to 600°C with an heating rate of 8°C min - ' under the oxygen and helium mixture . In addition to the peak observed at room temperature, a second peak of oxygen consumption was observed during heating to 600'C (Fig. 5) . For the sample in state I, the temperature of this peak was close to ca . 390°C, (Fig. 5A) . In state II this peak was observed at a lower temperature, i .e., 340'C (Fig. SB) . In both cases no desorption of oxygen occurred during heating to 600°C . Heating under oxygen and helium at higher temperatures led to a removal of oxygen from the surface; this process was effective for temperatures higher than 700°C in our experimental conditions [111, After calibration, it was found that the amount of oxygen adsorbed at room temperature was in good accordance with the previous chemisorption measurements, leading to a dispersion of the metallic phase close to 17% for the sample in state I and 8% for the sample in state II . For the two catalysts studied, the global amounts of oxygen deduced from



3 10

600 C

Pur

He

\\ 39 0 'C

~~

A

\\

\\

He +1%.0

2

25'C

Pure He

e

0,

B Fig . 5 . Temperature-programmed oxidation experiments performed on Pd/A'201 samples inn state I (A profile) and in state II (B profile) . Curves drawn in an unbroken line correspond to the oxygen consumption, curves drawn in a dotted line correspond to the temperature evolution .

the areas of the peaks at 25°C and between 340 and 390°C was determined and compared to the amount of palladium introduced into the sample . The ratio of the number of oxygen atoms to the number of palladium atoms was then calculated . For the sample in state I, and taking into account the two peaks of oxygen consumption, a stoichiometry Pd/O=0 .95 was found . The corresponding value for the sample in state II was Pd/O=1 .1. Thus within experimental error, the formation of a bulk palladium oxide on both samples can be assumed after heating under oxygen at reduced pressure (P02 ;~- 8 Torr) and at temperatures higher than 400 ° C . The formation of bulk PdO occurs at a lower temperature for the Pd/A1 203 sample in state II in comparison with the same sample in state I .

3 11 DISCUSSION

The aging of a Pd/A12O,, catalyst at 600'C under methane-oxygen--nitrogen mixture rich in oxygen does not lead to a decrease of the catalytic activity . On the contrary, the rate of methane oxidation expressed per gram of catalyst significantly increases . For the aged catalyst, the reaction starts at 200`C instead of 300'C for the freshly reduced sample . Moreover, the temperature of half conversion is decreased by ca . 70°C. In addition, the enhancement of the catalytic activity of supported palladium is much more important than that of supported platinum when using the same conditions of aging (Fig . 3) . Nevertheless, chemisorption measurements show a sintering of the palladium particles since the dispersion of the metallic phase decreases from 16 to 8% after aging under the reactants at 600°C . TEM determinations are in accordance with a sintering of the palladium particles . Consequently the activity per surface metal atom (TOF) increases with the size of the metal particles . The catalytic combustion of methane over supported palladium has to be considered as a structure sensitive-reaction . A parallel study performed on supported platinum leads to the same conclusion [2 ] and this behaviour has already been postulated for the total oxidation of other hydrocarbons [ 3-7 ] . Changes in the reactivity of adsorbed oxygen could be responsible for the enhancement of the catalytic activity after aging under methane-oxygen-nitrogen at 600°C . Low temperature titration experiments show that oxygen adsorbed on palladium in state II is much more reactive to hydrogen than the corresponding form in state I . It can be assumed that the enhanced reactivity to hydrogen at -78°C is also effective with methane at higher temperatures . Similar conclusions have been drawn in the case of supported platinum [2 ] . The increase in the reactivity of adsorbed oxygen to hydrogen after aging at 600°C under the methane-oxygen-nitrogen mixture must be associated with a higher activity in methane oxidation . In this sense, the increase in the rate of methane combustion observed between 200 and 300°C for the Pd/A120, sample in state II has to be associated with the higher reactivity of adsorbed oxygen on such a sample . Changes in the palladium-oxygen bond strength have already been observed by Chou and Vannice [12 ], in a microcalorimetric study of oxygen adsorption on supported palladium . They found that the heat of oxygen adsorption strongly increases when the metal particle size decreases . They concluded that changes in the electronic properties of small palladium crystallites are mainly responsible for the observed increase in the Pd-O bond strength . Temperature-programmed oxidation experiments show that a bulk palladium oxide can be formed in the range of partial oxygen pressure and in the interval of temperature studied for the catalytic combustion of methane . The comparison of the TPO profiles with the catalytic measurements shows that the unstability of the reaction with time occurs during the formation of the

31 2

bulk palladium oxide phase . Thus it is tempting to correlate the development of the catalytic activity with the amount of bulk PdO so obtained . The activity is stable at any given temperature below 390°C and above 480°C for the Pd/ A1.,O3 sample in state I and below 340'C and above 420 ° C for the same sample in state II . TPO measurements show the formation of bulk PdO at ca . 390'C for state I, whereas the temperature of PdO formation is lowered to 340°C for state II, showing that the formation of bulk PdO is favoured on 16 nm diameter palladium particles in comparison with the 7 nm diameter ones . Hicks and co-workers [6,7 ] have already found that the T .O.N. for methane oxidation decreases when the diameter of the palladium particles decreases . They postulate that the reaction occurs on dense faces present on large palladium crystallites . In addition they found that small palladium crystallites are completely oxidized at 300°C under a 110 Torr oxygen pressure . During the oxidation process, new surfaces of NO develop because the PdO lattice obtained from metallic palladium has a porous character . For samples initially reduced, oxidation of the metal occurring during the reaction leads to the formation of new active sites associated with an increase in the rate of methane oxidation . They concluded that all the oxide generated has its surface accessible to the reactants and participates in the methane oxidation . Our results are in agreement with some of the conclusions of Hicks and coworkers [6,7] . Both studies conclude that palladium oxide forms during the catalytic combustion of methane in an oxygen-rich mixture . In addition, Hicks and co-workers surmise the presence of two types of palladium oxide : the first one, poorly active, constitutes palladium oxide dispersed on the alumina support, the second one, very active, constitutes NO deposited on metallic palladium . The TPO experiments performed in the present study do not allow for the discrimination between these two types of palladium oxide . In addition to the sintering of the metal crystallites, some changes in the surface properties of palladium atoms may occur under the conditions of reaction . In fact, on-going from state I to state II, the catalytic activity may be increased by a factor of 35 (at low temperatures) whereas particles size varies from 6-16 nm . Similar aging experiments in the absence of methane do not lead to a significant increase in catalytic activity . Thus the presence of both reactants (methane and oxygen) appears to be required for the enhancement of the activity. Several hypotheses can be advanced to explain such a behaviour: deposition of carbon resulting in a modification of the active sites, surface reconstruction of the crystallites, etc. The behaviour of Pt/A1 2O, catalysts was investigated in the methane oxidation . A similar increase in catalytic activity and in metal particle size was observed after aging at 600'C under a methane, oxygen and nitrogen mixture [2 ] . Nanodiffraction and transmission electron microscopy (TEM) studies [13] have shown that platinum particles grow epitaxially on alumina with Pt (110) planes parallel to y-A1 20n (110) planes and

313 Pt[111] axes parallel to y-A1 2 0 3 [ 111 ] axes . The preferential exposition of Pt (110) planes in the aged sample as opposed to the freshly reduced one was considered to be responsible for the increase in catalytic activity . A similar behaviour could be considered to explain the enhancement of the aged Pd/ A1 2 0 3 solid in the reaction of methane oxidation . CONCLUSION After aging at 600 ° C under a methane, oxygen and nitrogen mixture, a Pd/ A1203 catalyst was found to be more active in its reaction to methane combustion . A sintering of the metal particles (mean diameter increasing from 7-16 nm) was observed . The catalytic oxidation of methane over Pd/A1 20 3 seems to be dependent on the diameter of the metal particles : the catalytic activity expressed per surface metal atom increases as a result of aging under the reactants mixture at 600'C . The reactivity of oxygen adsorbed at room temperature on the metallic phase is also modified after the aging of the reactants : the reactivity of adsorbed oxygen to hydrogen at -78`C is enhanced on the aged sample . The higher reactivity of adsorbed oxygen is probably the reason behind the increase in the reactivity of the large palladium crystallites in the catalytic oxidation of methane below 340°C . The changes in catalytic activity can not only be connected to the increase in the size of the metal particles . Restructuration of the surface under the reactants has to be considered . In addition TPO measurements show that palladium is in the form of bulky PdO for temperatures of reaction higher than 400'C and for methane-oxygen mixtures rich in oxygen . The temperature for PdO formation appears to be also particle size dependent, it decreases slightly when increasing the metal particle diameter in the 7-16 nm range. The unstability of the reaction rate with time is correlated to the progressive formation of bulk PdO . At any given temperature, the rate of reaction remains constant with time before PdO formation and when the metal particle is fully oxidized into bulk PdO . In contrast, with supported platinum, the catalytic oxidation of methane over supported palladium occurs on a palladium oxide phase in an oxygen-rich reactants mixture, even if the catalyst is initially in a reduced form . ACKNOWLEDGEMENTS Financial support for this work was provided by Gaz de France (Direction des Etudes et Techniques Nouvelles, Centre des Etudes et Recherches sur les Utilisations du Gaz) . The authors thank Mrs . C . Leclercq for TEM determinations, Mr. B . Beguin for assistance with the catalytic activity measurements and Dr . E . Garbowski for stimulating discussions .

314 REFERENCES 1. 2 3 4 5 6 7 8 9 10 11 12 13

D .L . Trimm and C, W. Lam, Chem . Eng, Sci., 35 (1980) 1405 . P . Briot, A . Auroux, D . Jones and M . Primet, Appl . Coral ., 59 (1990) 141 . L .M, Carballo and E .E . Wolf, J . Catal., 53 (1978) 366 . J . Volter, G . Lietz, H . Spindler and L, Lieske, J . Carol., 104 (1987) :375 . C .F . Cullis and B .M . Willat, J . Catal., 83 (1983) 267. R.F . Hicks, H . Qi, M.1 . . Young and R.G . Lee, J . Catal. . 122 (1990) 280. R.F . Hicks, H . Qi, M .L . Young and R.G . Lee, J . Catal., 122 (1990) 295, M . Primet and Y . Ben Taarit, J . Phys . Chem., 81 (1977) 1317 . M . Boudart and H .S . Hwang, J . Catal ., 39 (1975) 44 . J .R. Anderson, Structure of Metallic Catalysts, Academic Press, London, 1975, p . 296 . M . Primet, unpublished results . P . Chou and M .A. Vannice, J . Catal ., 105 (1987) 342 . P . Briot, P . Gallezot, C . Leclercq and M. Primet, Microsc. Microanal . Microst .ruct .. 1 (1990) 149 .