Langmuir 1989,5, 1364-1369
1364
Methane Oxidation over Pt on y-Alumina: Kinetics and Structure Sensitivity K. Otto Research Staff, Ford Motor Co., P.O. Box 2053, Dearborn, Michigan 48121 Received March 22, 1989. I n Final Form: June 7, 1989 A recirculation batch reactor and a mass spectrometer were used to measure methane oxidation rates, free of mass-transfer limitations, over Pt. Samples included platinum on y-alumina in the range from 0.03 to 30 wt % and Pt black. The reaction is described by first-order kinetics with respect to methane and is independent of oxygen pressure under the experimental conditions chosen. A maximum in turnover frequency and a decrease in apparent activation energy with increasing Pt concentration are found. These features can be explained by the existence of two distinct platinum entities, Le., dispersed and particulate Pt. A corresponding change in kinetic parameters can also be achieved reversibly at low Pt concentration by sintering in hydrogen and redispersion in oxygen at 500 "C. Calculations of a reaction site density, based on theoretical and empirical kinetics, are inconsistent with an Eley-Rideal reaction mechanism. The reaction is possibly rate controlled by the dissociation of methane.
Introduction As a rule, the activity of a catalyst increases with the number of surface sites. The reaction rate, however, is not merely a function of site density. Usually when the unit turnover frequency (TOF) is used to express specific catalyst activity, it is tacitly assumed that all the reaction sites are equivalent or that, at least, the specific site activity can be represented by a fixed average. However, in general, a catalytic surface is not uniform and varies with particle size and crystal habit. Thus, specific activity per surface site depends on the type of crystallographic plane exposed, which in turn can vary with the reaction conditions.'p2 Differences are also introduced by disparate surface sites located on corners and edges. Furthermore, reactions which are governed by a Langmuir-Hinshelwood mechanism cannot take place on isolated sites and require a minimum site ensemble, which may include sites on the support material, e.g., when a hydrogen-spillover mechanism is involved. In this case, activity measured per site usually decreases with the number of neighboring sites. Finally, support interaction can modify catalytic sites substantially and thus result in an activity change as well. These examples illustrate four distinctly different reasons for TOF modifications caused by changes in catalyst structure. Activity changes of this type, generally lumped together under the term structure ~ensitivity,~ are of importance for the optimization of catalyst formulations with respect to efficiency, selectivity, and cost. Further complications of the concept of structure sensitivity have been pointed out by Carberr~.~ A more complete mathematical description of a catalytic reaction requires activity differences to be taken into account by appropriate activity coefficients. To optimize catalyst performance as a function of particle size, it is necessary to consider all of the structural parameters that cause activity changes. For example, alkane oxidation on automotive catalysts is known to be relatively slow if the catalytic metals are highly d i ~ p e r s e d .On ~ (1) Flytzani-Stephanopoulos, M.; Wong, S.; Schmidt, L. D. J. Catal. 1977,49, 51. (2) Wong, S.; Flytzani-Staphanopoulos, M.; Chen, M.; Hutchison, T. E.; Schmidt, L. D. J. Vac. Sci. Technol. 1977,14,453. (3) Boudart, M. Ind. Eng. Chem.Fundam. 1986,25,656. (4) Carberry, J. J. J. Catal. 1977, 107, 248.
0743-7463/89/2405-1364$01.50/0
the other hand, if the particles are too large, the catalyst becomes less efficient as the fraction of surface atoms decreases. Thus, there exists an optimum particle size for oxidizing the least reactive HCs (hydrocarbons), i.e., alkanes and particularly methane, most efficiently. The relative concentration of these specific HCs increases as the total amount of HC decreases during catalytic oxidation, resulting in residual levels a t the catalyst outlet as high as 25 wt 3' % for methane and 70 wt % for total alkanes.6 The notoriously slow oxidation of methane' was selected as a first model reaction to study changes in reaction parameters as a function of Pt dispersion.
Experimental Method The catalyst samples were prepared by multiple impregnation of y-alumina (Degussa, aluminum oxide C) with chloroplatinic acid. The lowest concentration of 0.027 wt % Pt/A1,0, was prepared by adding a sufficient amount of a Pt solution to produce half the concentration desired, drying the sample at 125 "C, and repeating the impregnation. Higher concentrations were prepared by successiveimpregnations with the same starting solution, until the number of impregnations became too large to be practical. The same procedure was then repeated with another starting solution of higher Pt concentration and, initially, three successive impregnation steps. Each impregnation of the briskly agitated alumina powder was stopped at incipient wetness, and the total Pt deposit was calculated from the Pt concentration of the solution and the amount of liquid consumed. The samples were calcined for 16 h at 600 "C, which is not sufficient to remove chlorine completely, according to analysis by XPS (X-ray photoelectron spectroscopy). Chlorine was expelled by heating in flowing hydrogen, first at 150 "C for 2 h and then at 450 "C for at least 1h. After this treatment, residual chlorine was always less than 0.1 atom %, according to XPS analysis. Standard sample preparation included exposure to oxygen at 500 "C for 20 h, which followed the initial reduction in hydrogen. For comparison, reaction rates were also measured on platinum black (Alfa Products). The BET surface areas, measured by Ar as well as by N, adsorption, were 85 f 5 and 8.9 f 1m2/gforthe supported and unsupported samples, respectively. Methane, argon, and oxygen in the reaction mix( 5 ) Gandhi, H. S.; Shelef, M. In Catalysis and Automotive Pollution Control; Crucq, A,, Frennet, A,, Eds.; Elsevier: Amsterdam, 1987;
p 199. (6) Sigsby, J. E., Jr.; Tejada, S.; Ray, W. Enuiron. Sci. Technol. 1987, 21(5), 466. (7) Boreskov, G. K. Discuss. Faraday SOC.1966, 41, 263.
0 1989 American Chemical Society
Langmuir, Vol. 5, No. 6,1989 1365
Methane Oxidation Rates ture were of high purity. A vacuum of about 5 X lo4 Torr (1 Torr = 133.3 Pa) was maintained in the glass apparatus by a turbo pump. A differential batch reactor was used to measure reaction rates essentially free of concentration and temperature gradients. The recirculation system, the reactor (made of glass), and the principle of the gas analyais by maw spectrometer have been described before? The volume of the circulation loop was 345 cm3,excluding the reactor, which had a volume of about 130 cm3. The initial reaction mixture in the circulation loop, isolated from the reactor, consisted usually of 10 Torr CH,, 60 Torr O,, and 600 Torr Ar. The oxygen pressure thus was 3 times higher than the pressure required for the complete oxidation of methane to CO, and H,O. As a rule, the reactor charge was fixed at 4 mg Pt, deposited on varied amounts of alumina (0-15 g), to study the dependence on Pt loading. In order to achieve reproducibility, for reasons given below, samples which had been exposed to hydrogen for chlorine removal, were later kept in oxygen at 500 "C for at least 20 h. The reaction was started by opening the evacuated reactor, kept at a predetermined temperature, to the premixed reactants. Sufficiently low reaction temperatures were chosen to keep the reaction rate low enough to prevent local overheating. The gas mixture was recirculated over the catalyst at least twice per minute. A minimum of 2 h was required for each reaction isotherm to avoid rate retardation caused by a pumping speed limitation. A small sample of about 1 cm3 was injected into a mass spectrometer (UTI100C) at predetermined intervals to measure the gas composition as a function of time. The gas analysis was based on the m/e peaks 15 (CH, fragmentation) and 28 (CO parent peak, CO, fragmentation) and the parent peaks of 0,, Ar, and CO, at m/e 32,40, and 44,respectively. The methane parent peak at mle 16 was not used because of interference with fragmentation peaks of H,O, 0,, and CO,. On the basis of the known amount of Ar, the quantities of CH,, CO, 0,, and CO, were calculated from the recorded peak intensities. A mass balance of the carboncontaining species was accurate to at least 5%, provided the concentrations of the reaction products were sufficiently high. It is estimated that under the highly oxidizing condition employed here less than 2% of the carbon in methane was converted to products other than CO or CO,. A conventional vacuum apparatus consisting of calibrated glass bulbs and a quartz-spiral manometer (Heise Instruments) was used for chemisorption measurements. The apparent volume of the adsorption vessel was obtained from the pressure change of helium expanding from a standard volume. Portions of the same samples prepared for the kinetic measurements were used for chemisorption experiments. These samples were prepared as described before: reduction in flowing hydrogen up to 450 "C for a total of 3 h, followed by exposure to oxygen at 500 "C for 20 h. Preceding each chemisorption measurement, samples were exposed to hydrogen at 400 "C to reduce oxidized Pt to its metallic state, followed by evacuation. Oxygen and, separately, CO were used as adsorbates on reduced Pto at room temperature. Hydrogen chemisorption was found to be less reproducible; completion of the adsorption process was very slow, possibly because of hydrogen spillover? Hydrogen chemisorption data were therefore not included in the TOF evaluation. Results a n d Discussion S t a n d a r d Reaction Rates. Methane oxidation over
Pt is described by first-order kinetics with respect to methane; it is independent of oxygen pressure in the range explored here (2.5 < O,/CH, < 60). The rate data are thus represented by the product of methane concentration cM and rate constant k dcM/dt = - C M k
(1)
k = k , exp(-E/RT)
(2)
The rate constant
(8) Otto, K.; Shelef, M.; Kummer, J. T. J. Phys. Chem. 1970, 74, 2690. (9) Sermon, P. A.; Bond, G. C. Catal. Rev. 1973,8(2), 211.
A
10
0
160
80
240
320
400
480
Time ( min )
Figure 1. Reaction isotherms on 5 wt % Pt/y-alumina: ( 0 )
350.8,(0)399.8,( 0 )448.8 "C.
10
I 0
100
200
300
400
Time ( m i n )
Figure 2. Reproducibility of kinetic data illustrated by five reaction isotherms on 1.4 wt % Pt/y-alumina at 400.0 "C: (0) run 1, (0) run 3, (0) run 4, (A)run 9,( 0 )run 11.
is given by the change in methane concentration with time
k = -d In cM/dt (3) The apparent activation energy E is derived from the temperature dependence of the rate constant E = -RAk/ At (4) where R is the gas constant. Examples of reaction isotherms a t three different temperatures are given in Figure 1. The straight lines in the semilogarithmic coordinates of Figure 1 illustrate that the methane oxidation is adequately described by firstorder kinetics with respect to methane. The straight lines, calculated by least-squares fit, yield a correlation coefficient of a t least 0.997 for essentially all the reaction isotherms measured. By keeping the experimental conditions meticulously constant, excellent reproducibility of the reaction isotherms was achievable, as shown by Figure 2. These isotherms, representing a series of measurements a t 400 "C on a single sample of 1.4 wt % Pt, are described by a rate constant It' = 2.07 X lo9 min-' (It = 2.303k') within a standard deviation f0.14 X min-l. It should be stressed that because of potential changes in Pt dispersion, as discussed below, it is very important to keep the sample pretreatment as constant as possible. Changes i n Reaction Parameters with Pt Concentration. The rate constant k' (k = 2.303k' ) is plotted in Figure 3 as a function of Pt loading. The plotted data are those k' values measured after sample pretreatment in oxygen at 500 "C for at least 20 h. The open circles represent the rate constant measured over 4 mg of Pt. At concentrations below 1.4 wt % Pt, the rate
1366 Langmuir, Vol. 5, No. 6, 1989
Otto
1
0.1Ok
Table I. Oxidation of Methane at 10 Torr (1.3 kPa) and 400
oc ~
1
f
TOF," s-l
0.03-1.4 10-100 5
O.OOO6
0.0023 0.0053
~
remarks dispersed Pt particulate Pt absoluta maximum
Turnover frequency.
i llilll,
I
10-4
0.01
wt % Pt
0.10
,
ll,li,,
1.o
I
I,1111,
10.0
100
Pt Concentration (weight%)
Figure 3. Change in rate constant as a function of Pt concentration at T = 400 "C. Reactor charge is 4 mg of Pt in each case: (0) per 4 mg of bulk Pt;( 0 )normalized to surface Pt. constant k' remains constant at 1.36 X min-' within a standard deviation of f0.23 X min-l. A maximum of the rate constant exists a t 5 wt %. Above 10 wt % Pt, a sharp decrease of 12' takes place. The observed maximum in k' a t a fixed amount of Pt clearly manifests structure sensitivity. At lower Pt concentrations, the metal is well dispersed. On the basis of chemisorption experiments and other observations, it has been claimed that each Pt atom is essentially a surface atom." At higher concentrations, Pt particles are formed which clearly contain a larger fraction of subsurface Pt atoms. Since the amount of surface Pt a t the fixed 4-mg bulk Pt cannot increase with concentration, the maximum in Figure 3 indicates a distinct increase in specific activity of the catalytic sites. In other words, the turnover frequency increases with increasing particle size. The decrease in Figure 3 above 10 wt % Pt indicates that further loss of surface atoms by growth in crystal size is not compensated by additional gain in site activity. The rate constant of Pt black was also measured and is included in the plot (100 wt % Pt). The rate maximum a t about 5 wt % Pt becomes more pronounced when the number of exposed Pt surface sites is taken into account. This correction is based on data derived from CO chemisorption on Pt, after reduction in hydrogen a t 400 "C, followed by evacuation a t the same temperature. If it is assumed that a t low Pt concentrations (below 1 wt 90)each Pt atom is a surface atom, a chemisorption ratio CO/Pt I1 is expected. Experimentally, a ratio CO/Pt = 0.68 f 0.05 is measured a t room temperature in the range from 0.03 to 1.4 wt %. This ratio was found to agree within the experimental error with the adsorption of atomic oxygen on the same s a m ples. A CO/Pt(surface atom) ratio in the range from 0.7 to 0.9 is commonly found in chemisorption experiments."*" The observation that each surface Pt atom does not adsorb exactly one CO molecule has been explained by sample impurities.ll A small fraction of CO adsorbed in the bridged mode30v31found in the IR spectrum a t about 1850 cm-l could account for some lowering of the CO/Pt(surface) ratio. However, the effect is not sufficient to explain the observed deviation from a ratio of 1 completely.12 Even if the true fraction of Pt surface atoms exposed to the gas phase would be only 0.7, as suggested by the chemisorption data, the average cubic Pt particle still would be very small and contain only about 125 Pt atoms. (10) Yao, H. C.; Sieg, M.; Plummer, H. K., Jr. J. Catal. 1979, 59, 365. (11) Verbeek, H.; Sachtler, W. M. H. J . Catal. 1976,42,257. (12) Haaland, D. M.; Williams, F. L. J. Catal. 1982, 76,450.
The rate data measured on samples containing 4 mg of bulk (open circles in Figure 3) were normalized to the Pt surface by using the chemisorption data and are represented by the solid circles in Figure 3. The assumption that a t the lowest concentration (0.0267 wt 90 Pt) a total of 4 mg of Pt is exposed in the surface amounts to 1.2 X 10'' Pt surface atoms. Accordingly, the correction factor applied to normalize the rate constant with respect to surface Pt is given by [(1.2 X 1O1')0.68]/CO molecules adsorbed. As expected, the correction emphasizes the maximum in the 5-10 wt % range and corrects for the steep decline a t higher Pt concentrations, which is caused by a substantial decrease in the fraction of active Pt surface atoms with increasing particle size. Selected TOFs, expressed in the appropriate unit (s-l), calculated from eq 1 and 4 and from the chemisorption data, are listed in Table I. These values, of course, can also be obtained from the data points, represented by the solid circles in Figure 3, by multiplication with a constant conversion factor. The TOFs show that, per Pt surface site, particulate Pt is a t least 4 times more active than is dispersed Pt. The increase in the TOF a t 5 wt % Pt by a factor of 2 above the values obtained a t higher platinum concentrations is noteworthy. The difference is significantly larger than the experimentalerror, as the rate constant a t 5 wt % Pt is known within a standard deviation of f12%. The result suggests that Pt particles formed in the transition zone between dispersed and particulate Pt are exceptionally active. If the change from dispersed to particulate Pt were the only cause of structure sensitivity, one would expect that the turnover frequency (solid circles in Figure 3) would consist of two constant values, one a t about 1 X min-' a t low Pt concentrations and another min-l a t high Pt concentrations, one a t about 4 X with a transition region in between. The maximum rate constant a t 5 wt % Pt and other observations of the methane oxidation experiments suggest the possibility of an additional type of structure sensitivity, akin to that observed on clusters in the gas phase containing less than 30 atom^.^^^'^ It was found that the reaction rate on such clusters changes strongly, in a nonmonotonic fashion, with the number of metal atoms per cluster. A definite claim for this type of structure sensitivity on supported Pt clusters, however, cannot be made with the available data. A transition in the same concentration range is also found when the apparent activation energy is plotted as a function of Pt concentration, as shown in Figure 4. It should be noted that the evaluation of an activation energy is an exacting task because of subtle and unpredictable changes in the catalyst structure resulting from variations with temperature and gas-phase composition.lB2It is therefore essential to go back and forth between two temperatures to ascertain that the reaction isotherms are (13) Whetten, R. L.; Schriver, K. E. Depsrtment of Chemistry and Biochemistry, University of California, Los Angeles, CA, "Atomic Clusters in the Gas Phase"; January 1988. (14) Cox, D. M.;Zakin, M. R.; Kaldor, A. 'Metal Clusters:Size Dependent Chemical and Electronic Properties";Physics and Chemistry of Small Clusters;Jana, P., Rao, B. K., Khanna, S. N., E%.; NATO ASJ Series C: Physics; Plenum: New York, 1987; Vol. 158, p 741.
Methane Oxidation Rates 50
Langmuir, Vol. 5, No. 6, 1989 1367 I
I
40
T
b
0 3 200 1
x
B
e
, d
lo 0.01
0.10
1.o
10.0
100
Pt Concentration ( weight OIo )
Figure 4. Change in activation energy as a function of Pt concentration.
Figure 5. Chemisorption of CO on Pt/y-alumina at room temperature at a function of Pt concentration.
Table 11. ADDarent Activation Energies Pt loading, wt%
0.027 0.124 0.41 1.40 2.50 5.0 10.0 15.0 30.0 100 (Pt black) 0.027 sintered
apparent activation energy, kcal/mol
standard deviation, kcal/mol
32.6 36.4 35.1 36.6 32.6 28.1 26.3 28.6 26.6
2.5 3.9 2.3 0.3 0.7 1.9 1.3 3.7 2.0
22.5 24.4
2.1 4.6
not subject to uncontrolled changes. To judge the significance of activation energy differences, it was necessary to evaluate the apparent activation energy statistically. For this reason, sometimes 10 or more reaction isotherms were measured a t a given Pt concentration. The standard deviation of each point is indicated in Figure 4 by error bars. Table I1 shows the individual values measured. The lower Pt concentrations yield an apparent activation energy of 35.2 f 1.6 kcal/mol; the value in the 5-30 wt % range is 27.4 f 1.0 kcal/mol. The two energies are significantly different. The apparent activation energy of unsupported Pt, 22.5 kcal/mol, which was also measured, is somewhat smaller; however, it is thought to be in agreement with the value measured on supported Pt a t higher concentrations, considering that an additional experimental error is introduced by a strong sintering tendency of platinum black. Sintering was found todepend on the distribution of the Pt on the fritted disk in the reactor and was noticeable from a persistent decrease of the methane oxidation rate. Most of the activation energies which have been reported for the Pt/y-alumina system are in reasonable agreement within the range 27-35 kcal/mol reported here,15-23 although values as low as 5.7 (ref 18) and as high as 47.8 kcal/mol (ref 20) have been reported. Disregarding these extreme values, examples from the literature are 30.9 (25.1~
~~
Pt Concentration (pmol/m2 (BET))
~~
(15) Drozdov, V. A.; Tsyrul'nikov, P. G.; Popovskii, V. V.; Davydov, A. A.; Moroz, E. M. Kinet. Katal. 1986,27,162(translation). (16) Drozdov, V. A.; Tsyrul'nikov, P. G.; Popovskii, V. V.; Pankrat'ev, Yu. D.; Davydov, A. A.; Moroz, E. M. Kinet. Katal. 1986,27,721. (17) Drozdov, V. A.; Tsyrul'nikov, P. G.; Popovskii, V. V.; Bulgakov, N. N.; Moroz, E. M.; Galeev, T. G. React. Kinet. Catal. Lett. 1985, 27(2),425. (18) Hicks, R. F.; Young, M. L. Synthesis and Properties of a Novel Catalyst for the Combustion of Methane; Gas Research Institute: Chicago, IL, Annual Report 1988.
32.1),l"17 29.0-39.6,18 27.2,2018.1,2124.6,22and 24 kcal/ mol.23 The substantial scatter of the energy values is possibly a manifestation of the changes in Pt structure mentioned above. A substantial decrease in apparent activation energy around 400 "C has been observed in two instances,18s20suggesting a change in catalyst structure. The reaction order has been reported to be zero order with respect to ~ x y g e n ~and . ~ ~first . order in methane," in agreement with the present work. The transition in TOF and apparent activation energy takes place in the range from 1.5 to 5 wt % Pt and seems to be related to two distinct types of Pt, i.e., dispersed and particulate Pt.lo The transition range is equivalent to a concentration interval of 0.8-3.0 pmol Pt/m2 (BET) of the y-alumina. A plot of CO chemisorption vs Pt concentration in logarithmic coordinates is shown in Figure 5. The break point is located a t 1.6 pmol/m2 (BET), which corresponds to 2.7 wt %. From the densit of Pt (21.45 g/cm3), a Pt cross section of 6.1 X lo-'' m Bis calculated. Accordingly, it is estimated that particulate Pt is formed after dispersed Pt has covered approximately 6.0% of this alumina with a surface area of 85 f 5 m2/g. A transition point a t 2.2 pmol/m2 (BET) was reported by Yao et al.10324for a series of Pt deposits on a different y-alumina support with a larger BET surface area (150 m2/g). In this case, the saturation concentration of dispersed Pt was 6.5 wt %, corresponding to 8.1% coverage of the alumina surface. Agreement in saturation concentration with the results reported here is deemed to be satisfactory considering the experimental error and the difference in support material. Effects of Sample Pretreatment. It'was found early in this investigation that sample pretreatment can have a profound influence on the rate of methane oxidation. The pretreatment effect can be linked to the type of gas to which the Pt sample was exposed a t sufficiently high temperature. Hydrogen has the tendency to increase the reaction rate, oxygen to decrease it. In Figure 6, some results are shown for a sample containing 0.4 wt % Pt/ y-alumina. Run 3 (circles), obtained after exposure to standard oxidizing conditions for two runs, yields a rate min-' a t 450 "C;exposure to constant k' = 8.30 x (19) Cullis, C. F.;Williat, B. M. J . Catal. 1983,83,267. (20) Firth, J. G.;Holland, H. B. Trans.Faraday SOC.1969,65,1121. (21) Trimm, D.L.;Lam, C.-W. Znd. Eng. Sci. 1980,35,1405. (22) Anderson, R. B.; Stein, K. C.; Feenan,J. J.; Hofer, L. J. E. Znd. Eng. Chem. 1961,53(10),809. (23) Yao, Y.F. Ind. Eng. Prod. Res. Dev. 1980,19,293. (24) Otto, K.;Yao, H. C. J . Catal. 1980,66,229.
Otto
1368 Langmuir, Vol. 5, No. 6, 1989
Table 111. Calculated Number of Reaction Sites for Dispersed and Particulate Pt"
103
apparent activation energy,
kcd/mol
35.2
21.4
Eley-Rided mechanism 9.1 x 10% 5.1 X lo2' slow CH, dissociation step 8.6 X 10' 5.3 x 1017 ,, Maximum number of Pt surface sites in reactor (4 mg): 1.2 X 1019.
Time ( min ) Figure 6. Changes in rate constant caused by different preheated treatments of 0.4wt % Pt/y-alumina. Pretreatment: (0) in oxygen at 500 "C for over 20 h; ( 0 ) heated in hydrogen at 500 "C for 20 h; (0) heated in oxygen at 500 "C for 20 h. r
C
*E c
C
0 c
.003
-
cn
0
.002 -
P, c
.001
0 (D
a
1
3
5
7
9
11
13
15
17
19
Run # of Reaction Isotherm Figure 7. Changes in rate constant at 350 "C of sample containing 0.0267 wt % Ptly-alumina caused by heating at 500 "C
in oxygen and hydrogen, respectively.
hydrogen a t 500 "C for 20 h (run 15, squares) increased min-'. Subsequent exposure to oxygen it to 14.6 X a t 500 "C for 20 h (run 16, diamonds) resulted in k' = 10.5 X min-l, a value significantly smaller than that obtained after hydrogen treatment but larger than that of the initial run 3. These results, again, are an indication of a structure-sensitive reaction. In the temperature range 450-550 "C,these reversible changes were found to be noticeable, sometimes even after 24 h. On the other hand, oxidation a t these temperatures is completed within less than 1 h, as observed during chemisorption measurements and sample characterizations by XPS and Raman s p e c t r o s ~ o p i e s .The ~ ~ ~slow ~~ activity changes are therefore thought to be associated with a change in particle size. According to Yao et a1.,l0 the hydrogen treatment causes sintering, while the oxygen treatment results in Pt redispersion. To confirm these results, another series of measurements was carried out to investigate whether particulate Pt could be formed even a t the lowest Pt concentration used (0.0267 wt %), The results are shown in Figure 7 . The rate constant is plotted for all the reaction isotherms measured a t 350 "C on the same sample (circles). The first three reaction isotherms measured a t this temperature (runs 1, 3, and 5) yield the same value, which remains constant after oxygen treatment at 500 "C for 20 h (run 8). Hydrogen treatment under the same conditions increases k' substantially, as shown by runs 11 (25) S h y , J. Z.; Otto, K. Appl. Surf. Sci. 1988, 32, 246. (26) Otto, K.; Weber, W. H.; Graham, G . W.; S h y , J. Z. Appl. Surf. Sci. 1989, 37, 250.
and 12. With exposure again to oxygen in two successive steps, the catalytic activity decreases, as before, and reaches essentially the initial value (run 17). Final exposure to hydrogen, again at 500 "C, but for a longer time (68 h), increases the catalytic activity by 1order of magnitude compared to the initial runs. Attempts to increase the rate constant further by sintering a t 500 "C failed. No Pt particles were visible by electron microscopy in these sintered samples, indicating that the particle size was below 20 A. The interpretation of reactivity changes caused by Pt sintering and redispersion is completely consistent with conclusions of Yao et al." based on chemisorption experiments and microscopic observations. Judging from the methane oxidation experiments, Pt sintering in hydrogen and redispersion in oxygen is very slow and barely noticeable a t 450 "C. These effects become more apparent a t 500 "C. Thus, sintering increases the total activity of the Pt particles in spite of the fact that a fraction of the Pt atoms move from the surface into the bulk. Conversely, redispersion in oxygen, which increases the Pt surface area, lowers the total activity of the 4 mg of Pt in the reactor. It should be noted that, as expected, the sintering of the sample given in Figure 7 is also reflected in a corresponding lowering of the apparent activation energy. This energy value is consistent with particulate Pt; it is listed in the last line of Table I1 and plotted in Figure 4 as a solid circle. Reflections on Reaction Mechanism. The kinetic data reported here can be used to draw some conclusions about the reaction mechanism. On the basis of the rate measurements and fundamental constants, reaction site densities can be calculated for specific reaction mechanisms.24'n.28The basis of the calculation is described in the Appendix. The strength of this method is not so much in proving that a certain reaction mechanism exists but rather in the rejection of a proposed mechanism. Because of the first-order kinetics with respect to methane, an Eley-Rideal mechanism was tested, where a methane molecule striking from the gas phase reacts with oxygen adsorbed on Pt, as proposed for palladium on alumina.29 Depending on the apparent activation energy for dispersed and particulate Pt, the number of sites is calculated to be larger by 4-6 orders of magnitude than the total number of Pt atoms in the 4-mg sample, as shown in Table 111. The same conclusion is reached if it is assumed that the slow step is associated with an oxygen molecule in the gas phase striking an adsorbed methane molecule.20 A better agreement is found if it is assumed that the slow step is the dissociation of an adsorbed methane molecule or a methane-oxygen surface complex. If the apparent activation energy of the dispersed Pt is used, (27) Maatman, R. W. J . Catal. 1976.43, 1. (28) Maatman, R. W. Adv. Catal. 1980,29, 27. (29) Mezaki, R.; Watson, C. C. Ind. Eng. Chem. Process Des. Deu. 1966,5(1), 62. (30) Haaland, D. M. Surf. Sci. 1987, 185, 1. (31) Barth, R.; Pitchai, R.; Anderson, R. L.; Verkios, X. E. J. Catal. 1989, 116, 61.
Langmuir, Vol. 5, No. 6, 1989 1369
Methane Oxidation Rates
(3) The methane oxidation kinetics are consistent with the existence of a t least two different Pt entities, termed dispersed and particulate platinum oxide, which are characterized by a difference in particle size and oxidation state. Unique Pt clusters of extraordinary activity may exist. (4) Larger particles, produced a t higher Pt concentrations, show the kinetic characteristics of unsupported Pt black. (5) Reaction kinetics of methane oxidation depend strongly on the pretreatment of the Pt samples since oxidation tends to disperse Pt, while reduction can cause sintering."
i,
16
4t 0
I
I
I
1
20
40
60
80
I
100
120
140
TIME ( m i n )
Figure 8. Example of half-order plot with respect to methane
concentration for a sample containing 10 wt % Pt/y-alumina, measured at 400 O C . the calculated site density, 8.6 X lo", is too high when compared to the maximum number of Pt sites, 1.2 X lo", given by the total amount of 4-mg Pt. However, the two values are of the same magnitude. If the apparent activation energy of the particulate Pt is selected, the calculated number of 5.3 X lo" Pt sites/sample is reasonable for 4 mg of bulk Pt. If the dissociation were the slow step, one would expect the reaction to be onehalf order in methane. Although the data fit best a firstorder plot (Figures 1 and 2), half-order plots, as exemplified in Figure 8, show that the possibility of a reaction order of ' I 2cannot be completely rejected. Such a mechanism has been proposed by Trimm and Lam?' who concluded that methane oxidation on supported Pt involves the reaction between adsorbed methane and adsorbed oxygen or oxygen from the gas phase. The latter reaction mechanism is in disagreement with the calculated site density.
Conclusions (1)Methane oxidation over Ptly-alumina is found to be structure sensitive, as the turnover frequency increases with Pt concentration. (2) The structure sensitivity is also reflected in a change in the apparent activation energy of the methane oxidation.
Acknowledgment. Extensive work by H. K. Plummer, Jr., to evaluate the size of Pt particles by electron microscopy is gratefully acknowledged.
Site density L for Eley-Rideal mechanism:
L = vFtJmta exp(E/RT)
(5) c,(kT/h) Site density c, when the reactant molecule adsorbs rapidly and the subsequent molecule dissociation is the slow step: c, = L[c,/(FtJr,)l"2
(6)
k = 1.38 X erg K-', h = where A, = 6.023 X 6.63 X erg 8, u = reaction rate (molecules cm-2 s-I), temperature = 297 K, Ft, = [(2amkT)3/21/h3,m = 16/ A, (g), and F,, = [8?r2(8?r31,1,1,)1/2(kT)32]/(ah2). Moment of inertia for CH,: = 5.337 x lo4' (g cm2) =
= I~
R = 8.31 J K-' mol-' The volume of the circulation loop is 327.35 cm3, the methane pressure is 10.45 f 0.20 Torr, and the oxygen pressure is 60 Torr. Registry No. Pt, 7440-06-4; CO, 630-08-0; O,, 7782-44-7;
H,,1333-74-0; CH,, 74-82-8.