Oxidation of carbon monoxide on palladium: role of the alumina

Mark A. Newton , Davide Ferri , Grigory Smolentsev , Valentina Marchionni , and ... Martin A. Röttgen, Stephane Abbet, Ken Judai, Jean-Marie Antoniet...
0 downloads 0 Views 2MB Size
Langmuir 1988, 4, 722-728

722

Adsorbed DOPA and TYR yield nor values a t Pt(ll1) which are less than the limiting values for complete oxidation (to COzand NOz) by 8 or 16 electrons, respectively, while CT and DOPAC yield norvalues which are virtually equal to the limiting values, Table 11. Apparently, the pendant amino acid side chain of DOPA or TYR disconnects from the remainder of the molecule during the oxidation process to yield products not yet directly identified, for example H20C&?

+ 13 H20

--+

6

C02 t H02CCHC02H + 31 H*+ 31 e-

(41)

NH2

P&7-co2H

P

+ 12 H20

-

6 C02

t

H02CCHC02H t 2s H’ t 28 e-

(42)

NH2

Lacking such pendant groups, CT and DOPAC undergo essentially complete oxidation, eq 32 and 34.

In most cases the norvalue obtained a t Pt(100) is substantially smaller than the norvalue a t P t ( l l l ) , Table 11. DOPA, TYR, CYS, PHE, CT, and DOPAC belong to this category; DA and ALA are the only exceptions. The nox values observed a t Pt(100) approach rather closely the values expected for edgewise attached species, although in most cases these adsorbates were present in a horizontal orientation prior to oxidation. Apparently, the affinity of the Pt(100) surface for oxygen species results in adsorption of oxygen species and reorientation or displacement of the initial adsorbed species soon after the onset of the oxidation process.

Acknowledgment. This work was supported by the National Institutes of Health.Instrumentation was funded by the National Science Foundation, the Air Force Office of Scientific Research, and the University of Cincinnati. Registry No. DOPA, 59-92-7; TYR, 60-18-4; CYS, 52-90-4; PHE, 63-91-2;ALA, 56-41-7; DA, 51-61-6;CT, 120-80-9;DOPAC, 102-32-9;Pt, 7440-06-4.

Oxidation of Carbon Monoxide on Palladium: Role of the Alumina Support F. Rumpf, H. Poppa, and M. Boudart* Department of Chemical Engineering, Stanford University, Stanford, California 94305 Received September 28, 1987. In Final Form: January 26, 1988 Model supported metal catalysts were prepared under ultrahigh vacuum by evaporating Pd onto a single crystal of a-A1203.On different samples the aver e Pd particle size d ranged between 1.4 and 10 nm and the particle number density n between 1.6 x 10’Y and 1.3 X cm-2. For each sample, d and n were measured by transmission electron microscopy and the number of surface Pd atoms was counted by temperature-programmed desorption of CO. Between 550 and 650 K and at CO and O2pressures of 1.2 X lo-’ and 1.4 X lo4 Pa, respectively, the turnover rate ut for the oxidation of CO increases with decreasing n and d below a critical size of -5 nm. The extent of this increase in ut diminishes with increasing temperature T. To explain these three observations,it is assumed that each palladium particle is surrounded on the support by an effective collection zone of width XD.Molecules of CO impinging on the support within this zone diffuse across the support surface to the alumina-Pd interface where they are chemisorbed and may be oxidized. First, XD and hence the number of CO molecules reaching the interface by surface diffusion both increase with decreasing T. Second, on a per surface Pd atom basis, the relative contribution of CO molecules reaching the metal by surface diffusion increases w i t h decreasing d. W i d , as n increases, the effective collection zones overlap and ut goes down at constant d and T.

Introduction A good model for a supported metal catalyst consists of metal particles vapor deposited under ultrahigh vacuum onto a nonporous support f h or crystal.l9 With this type of sample, it is possible to control the cleanliness, surface Composition, and surface structure of the metal and of the support. These variables along with the metal deposition conditions strongly influence the nucleation and growth of the metal particles. Thus the metal particle size d , number density n, orientation with respect to the support, and surface facets can in principle be controlled? Values of d and n can be measured by transmission electron microscopy (TEM), and surface-sensitive electron spectroscopies may be used to determine sample composition and electronic structure. Therefore, reaction kinetics can be measured on clean and well-characterized supported metal cataly~ts.’*~.~ *Author to whom correspondence should be sent.

For the reaction between CO and O2on model Pd/alumina catalyts, the correlation between d and the turnover rate ut, defined as the number of product molecules produced per second per surface metal atom, was first measured by Ladas et d.’ At 518 K and CO and Oz pressures of 1.2 X lo4 and 1.4 X lo4 Pa, ut increased by a factor of -3 as d decreased from -5 to 1.5 nm. An explanation for the increase of ut was provided. It was based on the idea that the edge and corner Pd surface atoms that predominate on the smaller Pd particles are more accessible to CO molecules impinging from the gas phase then are face Pd atoms, which are embedded in the flat surface of (1)Ladas, 5.;Poppa, H.; Boudart, M.Surf. Sci. 1981,102, 151. (2)Ichikawa, S.;Poppa, H.; Boudart, M.In ACS Advances in Chemistry Series; Whyt8, T. E., Dalla Betta, R. A,, Derouane, E. G., Baker, R. T. K., Eds.; American Chemical Society: Washington,D.C., 1984;p 439. (3) Poppa, H. Vacuum 1984,34(12), 1081. (4) Maeson, A.;Bellamy, B.; Hadj Romdhane, Y.;Che, M.; Roulet, H.; Dufour, G. Surf. Sci. 1986, 173, 479.

0743-7463/88/2404-Q722$01.5Q/O0 1988 American Chemical Society

Langmuir, Vol. 4, No. 3, 1988 723

Oxidation of CO on Pd a low index crystal plane. Since for these reaction conditions ut is proportional to the CO collision rate, the ut vs d behavior may result from the larger effective impingement rate of CO on the smaller Pd particles. This idea has been further elaborated by L a d a ~ .Alternatively, ~ it was proposed that ut is independent of d if the rate of C02 production is normalized by the total number of surface sites rather than the number of Pd surface atoms.2 But the results of the present work suggest a simpler and more convincing explanation. In the original study, it was also noted that the values of ut on the smaller particle samples were actually greater than the impingement rate (per surface Pd atom) of CO from the gas phases1 In other words, the number of C02 molecules being produced in, say, 1 s was greater than the number of CO molecules striking the Pd particles per second. A related observation was made later by the Gillet Group in Marseille. They reported that the initial effective CO sticking coefficient so was greater than unity on small Pd particles! Both of these results seem to indicate that more CO molecules are reaching the Pd particles than can be accounted for by direct impingement from the gas phase. One is reminded of the parallel observation of Volmer and E ~ t e r m a n n . ~ While studying the growth of a flat Hg crystal from the vapor phase, they found the lateral growth rate of the crystal to exceed the rate of impingement of Hg atoms on the lateral crystal planes. They concluded that Hg atoms struck the basal planes and then diffused across the basal surface to stable growing edges of the crystal. Indeed, the Gillet Group has proposed that during CO adsorption and the CO-O2 reaction on supported metal samples a CO molecule may strike the support and then migrate across the support surface to a Pd particle on which it chemisorbs.6-6 The total number of CO molecules reaching a Pd particle is then the s u m of those striking the Pd atoms directly from the gas phase and those reaching the Pd atoms by surface diffusion. In support of this mechanism, they found the areal rate of the CO-O2 reaction to increase with decreasing nag This idea is a good example of the porthole e f f e ~ tIn. ~ 1960 a t the Second International Congress on Catalysis, Sir Hugh Taylor combined the concepts of surface diffusion and active sites to propose that sites of special activity could act as portholes for the release of products from a catalytic cycle.1° At the same congress, Tsu and Boudart provided an example of the porthole effect." They reported that between 200 and 500 K the kinetics of recombination of hydrogen atoms on Pyrex can be described by a mechanism that includes the surface diffusion of hydrogen atoms weakly chemisorbed on the glass surface to active sites where they recombine. The porthole effect is in fact the reverse of a phenomenon that has been studied extensively, namely, spillover, especially hydrogen spillover.12-16 In hydrogen spillover, molecular hydrogen first (5) Ladas, S. Surf. Sci. 1985, 159, L406. (6) GiUet, E.; Channakhone, S.;Matolii, V.; Gillet, M. Surf. Sci. 1988, 1521153. , - - -603. ~. - (7) Volmer, M.; Estermann, I. 2.Phys. 1921, 7 , 1. (8) Gillet, E.; Channakhone, S.;Matolin, V. J. Catal. 1986, 97, 437. (9) Matolii, V.; Gillet, E. Surf. Sci. 1986, 166, L115. (10) Taylor, H. In Actes du Deuzieme Congres International de Catalyse; Editions Technip: Paris, 1961; Vol. 1, p 159. (11) Tau, K.; Boudart, M. In Actes du Deurieme Congres International de Catalyse; Editions Technip: Paris, 1961; Vol. l, p 593. (12) Holatein, W. L.;Boudart, M. J. Catal. 1981, 72, 328. (13) Fujimoto, K.; Toyoshi, S. In Studies in Surface Science and Catalysis; Seiyama, T.; Tanabe, K. E%.; Elsevier: Amsterdam, 1981; Vol. I A , p 235. (14) Khoobiar, S.J. Phys. Chem. 1964,68(2), 411. (15) Boudart, M.; Aldag, A. W.; W.;Vannice, Vannice, M. A. J. Catal. 1970, 18, 46. ~~~

dissociates on the metal component of the catalyst and then the atoms diffuse across the metal-support interface to react on the support. Some authors have suspected that under certain conditions the surface diffusion of reactants on the support may contribute significantly to the total reactant flow through a porous c a t a l y ~ t . ' ~ J Others ~ have presented models to describe the effect of surface diffusion on the kinetics of surface reacti~ns.llJ"~~ Of particular interest is the theoretical treatment of the problem of the "catalytic archipelagos" by Aris and co-workers.%p%They have used the finite element method to model the case where reactants adsorbed on the sea of support diffuse to and react on catalytic islands, i.e., the supported material. In this paper, the surface diffusion explanation for the increase of ut with decreasing d for the CO-O2reaction on Pd/A1203is supported by measuring ut between 525 and 650 K on a series of samples with varying d and n and describing the results with a quantitative kinetic model that includes the reverse spillover of CO from the support to the metal.

Experimental Section The experimental apparatus and procedures for temperature-programmed desorption (TPD), determination of the steady-state turnover rate, and measurement of d and n were similar to those detailed elsewhere.' The experiments centered around a dual-chamber ultra-high-vacuum (UHV) system. SupportedPd particle sampleswere prepared in the main chamber by electron beam vapor depition of 99.99% pure Pd onto a (10121 a-A1203support crystal from Insaco. The background pressure was about 5 X lO-' Pa during depition. The samples were heated resistively by a tantalum film which had been sputtered onto the backside of the alumina support. Before Pd deposition, surface carbon was reacted off the support crystal by an oxygen plasma.% Auger electron spectroscopy (AES)was used to verify the removal of surface carbon. The AES data showed trace impurities of calcium and potassium. After Pd deposition,the supported Pd sample was transferred to the smaller reaction chamber for TPD and steady-statereaction experiments. This reaction chamber was equipped with a separate ion pump, a quadrupole mass spectrometer, and leak valves for the intrcductionof CO (Matheson,99.99% pure) and O2(Scientific Gas Products Inc., 99.99% pure). Before each experiment the mass spectrometer was calibrated against a nude ionization gauge to indicate the pressure of N2 The sensitivity (Pa/A) of the mass spectrometer to CO, 02,and C02was then calculated from the sensitivity to N2,1and subsequently the mass spectrometer was used to measure the pressures of CO, 02,and COz. The kinetics of steady-state CO oxidation were studied between 525 and 650 K for CO and O2 partial pressures of 1.2 X lo-* and 1.4 X 10"' Pa, respectively. Carbon monoxide adsorption at 320 K with subsequent TPD was used to count the number of surface Pd atom for u8e in calculation of the tumover rate.' The temperature T was measured with a Chromel-Alumelthermocouple clipped to the edge of the support crystal. An infrared pyrometer was used to calibrate the thermocouple and to determine the heating schedule during TPD. (16) Levy,R. B.; Boudart, M. J. Catal. 1974, 32, 304. (17) Riekert, L.MChE J. 1985, 31(5), 863. (18) Tsotais, T. T.; Sane, R. C.; Webster, I. A.; Goddard, J. D. J.Catal. 1986. 101.416. - _ - _--, I

(19) Aris, R. J. J. Catal. 1971, 22, 282. (20) Kung, H. H.; Kung, M. C. Chem. Eng. Sci. 1978,33, 1003. (21) Prager, S.;Frisch, H. L. J. Chem. Phys. 1980, 72(5), 2941. (22) Freeman, D. L.; Doll, J. D. J. Chem. Phys. 1983, 78(10), 6002. (23) Freeman, D. L.; Doll, J. D. J. Chem. Phys. 1983, 79(5), 2343. (24) Kuan, D. Y.; Davis, H. T.; Aris, R. Chem. Eng. Sci. 1983,38(5), 719. (25) Kuan, D. Y.; h i s , R.; Davis, H. T. Chem. Eng. Sci. 1983, 38(9), 1569. (26) Poppa, H.; Moorhead, D.; Heinemann, K. Thin Solid Films 1985, 128, 251.

724 Langmuir, Vol. 4, No. 3,1988

Rumpf et 01.

Table I. Experimental Values of the Mean Pd Particle SI= d , Particle Number Density n, and the Approximate Number of Pd Surface Atoms per Pd Particle Mma dlnm nl10"cm~2 . M dlnm n/10"em~2 M . =" ,1.5 -20. -70 3.7 -13.5 2.5 -8.3 -155 4.8 -9.0 -680 2.6 -2.3 -160 5.7 -8.0 -760 -14.5 -300 10 -2.8 -21 3.6 @Otherparmetam used for the modeling work are a = 0.4 nm, E. - Ed = 23.5 W mol-', and S,(r) = 0.65[(1-1.3) X 1od (T-

--*--600 K

."

c

2

0.60

5 0.20

~

500)*]. 0 0

2

4

6

8

10

12

dlnm

Flgure 1. Turnover rate 88 a function of Pd panicle size for Pd/(i012)o-Al,02;Pco = 1.2 X 10'' Pa and P2 = 1.4 x 10'' Pa. Data points are the experimental values of uu an the hea connect the calculated values of ut. The mean equivalent diameter d of the Pd particlea and n were determined by bright field transmission electron microscopy ("EM). The TEM specimen. a 3"m disk of (1012)-oriented o-A120,, wan prepared as described elsewherel and then clipped to the main support slab during sample preparation and TPD and reaction experiments. Thia disk, representative of the catalyst sample, was subsequently removed from the UHV chamber and observed in a Hitachi HR-5CQHelectron microscope. For modeling the effect of surface diffusion on the observed reaction kinetics it was necessary to obtain a value for the heat of physisorption of CO on O I - A I ~ OThe ~ Clausius-Clapeyron equation was applied to a set of adsorptionisotherms to obtain the h t e r i c beat of adsorptionn A b u t 2 g of u-A1203powder (Johnson Mathey Inc., 99.99% pure) was placed in a glass cell connected to a conventional volumetric gas adsorption system." The quantity of adsorbed CO (Matheson, 99.9% pure) was calculated from the known volumes and pressures measured with a Teras Instruments precission pressure gauge. Before the adsorption isotherms were recorded the sample was treated in O2 (Matheson, 99.98% pure) at 673 K for 1h. The N2 (Matheson, 99.998% pure) BET specificsurface area of the alumina was 3.0 m2g-'. Carbon monoxide adsorption isotherms were recorded at 297,279,273, and 257 K. These temperatures correspond to room temperature, cyclohexane/liquidnitrogen, ice, and ethylene glycol/dry ice, respectively."

Results Valuea of ut are reported on a per surface Pd atom basis (eq 1)by using the CO TF'D spectrum to count the number of Pd surface atoms NPd(eq 2).'

Table 11. Deposition Procedures and Conditions for Preparation of Pd/ol-A120aSampleso step flux/cm-a 8.' T./K time/s sample A deposit 2.7 X loLa 750 180 B deposit 3.3 X 10l2 910 45 anneal 910 120 deposit 2.2 X 830 90 'A, n = 1.3 X em4; B, n = 1.6 x loLaem-.

....-

.

. . . .-...

... .

Ut

NC&/NPd

(1)

NPd

= NCO/BCO

(2)

Figure 2. TEM of Pd/(l0l2)a-Al2O8showing the difference in Pd particle number density between A and B (A) d = 3.7 nm. ' cm-2, n = 1.3 X lo'* cm-2; (B) d = 3.9, n = 1.6 x 10O

N c is~the total number of COz molecules produced per second, and Oco and Nco are the coverage and number of CO molecules adsorbed in the high-temperature CO/Pd

nation) as on a macroscopic Pd single crystal for which the

adsorption state at saturation. Values of Nco were calculated f"a C02balance on the reaction chamber during the steady-state C M 2 reaction. Similarly, Nco was determined from a CO balance on the reaction chamber during a CO TF'D experiment. A CO TPD spectrum from 1-10-nm supported Pd particles consista of two peaks, one at -360 K and the second near 460 K.1.2 Evidence suggesta that the high-temperature CO adsorption sites on Pd particles are the of same type (2-, 3-, and 4-fold coordi(27) Somorjai, 0.A. F'rincipks of Surface Chemistry; Prentice Hall: Englsaood Cliffn,NJ., 1972; Chaptar 5. (28) H m n , F. V. Ph.D. Dissertation, Stanford. University, 1976, p 80. (29) Gordsn. k J.; Ford, R. A. The Chemist's Companion; W i l w New Yark, 1972; p 451.

saturation coverage Oco is -0.45.'.30 Hence, NPdcan be calculated from the number of CO molecules desorbing from the high-temperature state Nco and the known coverage Ocm Values of d and n for a series of Pd/a-alumina samples are listed in Table I. With these samples and for Thetween 550 and 650 K, ut increases aa d decreases below -5 nm (Figure 1). The experimental points in Figure 1were obtained by measuring ut between -525 and 650 K at intervals of -25 K and then interpolating, if necessary, between the measured valuea of ut to give the experimental values at 550,600, and 650 K. The extent of this increase in ut with decreasing d decreases with increasing temperature. Two Pd/a-A120, samples with similar values of d (84) Conrad, H.; Ertl 0.;Koeh, J.; Lstta, E.E.Surf. Sei. 1974, e, 46,

Langmuir, Vol. 4, No. 3, 1988 725

Oxidation of CO on Pd 1-

A. 4

0.75 -

12 -2 n=1.3x10 cm 10 -2 n=1.6x10 cm

d=3.7 nm

8

B. A d=3.9 nm

A

-

20

3.0

4.0

5.0

6.0

drm

20

0

6.0

4.0

.

I

.

.

I

.

.

.

.

,

.

.

.

.

,

I

.

.

.

,

.

.

.

.

I

.

.

.

.

I

.

.

.

.

I

0.0

d/rm

Figure 3. Palladium particle size distributionsfor the high (A, top) and low (B, bottom) number density samples.

Figure 4. CO TPD spectra from samples A and B.

but different values of n were prepared, and for each sample values of ut between 535 and 635 K were measured. The different Pd particle number densities were obtained by varying the Pd vapor deposition conditions (Table II)?l Transmission electron micrographs show that the two samples A and B each have a mean Pd particle size d near 4 nm but that they differ in Pd particle number density by a factor of -100 (Figure 2). Values of d and n were obtained by measuring over lo00 particles for each sample. The corresponding particle size distributions are shown in Figure 3. Carbon monoxide TPD spectra from supported Pd particles are sensitive to the mean Pd particle size.'J The similar shapes of the CO TPD spectra from A and B confirm that the two samples have similar values of d (Figure 4). Near 535 K, ut is significantly larger for the low n sample B than for sample A (Figure 5). With increasing temperature, the difference between the values of ut for A and B decreases.

Discussion Qualitative. The turnover rate for CO oxidation by O2 on our Pd/a-A1203catalysts increases as d and n decrease for reaction conditions of 550-650 K and partial pressures of CO and O2 equal to 1.2 X lo4 and 1.4 X lo4 Pa, respectively. With these reaction conditions, ut is first and zero order in CO and O2 pressure, respe~tively,l-~~ so any

Figure 6. Pd particles (solid circles) and effective collection mnes (open circles). (A) Effect of Pd particle size on area of the collection zone. (B and C) Overlapping and independent effective collection zones. (D) If the Pd particles are considered to be regularly positioned on the support surface, then neareabneighbor particles are a distance 2L apart.

will result in an increase in ut. The observed increase of ut with decreasing d (Figure 1)may be explained by an increase in the rate of arrival of CO per surface Pd atom to the smaller particles as a result of reverse spillover. The reasoning is as follows. Everything happens as if each Pd particle is surrounded by an effective collection zone on the support surface. Carbon monoxide molecules impinging on the support and within this zone diffuse to the support-metal interface where they are chemisorbed. The interface between the A1203 support and a Pd particle is assumed to be circular, but it is not necessary to specify any three-dimensional particle shape. The effective collection zone may then be pictured as circular and extending away from the particle a distance XD which is equal to the mean distance traveled by an adsorbed species before it desorbs, XD = (Figure 6A).33 Here D is the surface diffusion coefficient and r is the mean residence time of CO on the support (eq 3 and 4).34

D = v ( a 2 / 4 )exp(-Ed/Rn 7

= u-l exp(E,/RT)

(3) (4)

In these expressions for D and 7,Y is a frequency factor assumed for simplicity to be the same for desorption and

increase in the rate of arrival of CO per Pd surface atom (31)Poppa, H.In Epitaxial Growth, Part A; Matthews, J. W., Ed.; Acadermc: New York, 1976; Chapter 3. (32)Engel, T.;Ertl, G . J. Chem. Phys. 1978, 69(3),1267.

(33)Kruyer, S.Koninkl. Ned. Akad. Wetenschap Proc. 1953,B56, 274. (34)Gomer, €2. In Surface Mobilities on Solid Materials; Binh, V. T., Ed.;Plenum: New York, 1983; p 7.

726 Langmuir, Vol. 4, No. 3, 1988

Rumpf et al.

surface diffusion, a the surface diffusion hop distance, R the gas constant, and Ed and E, the activation energies for CO surface diffusion on and desorption from the alumina. Combining eq 3 and 4, we get

(5) The radius and area of an effective collection zone are determined by the support-reactant interaction through a, E,, and Ed and decrease with increasing temperature (eq 5). For instance, a t 550 K, with a = 0.4 nm and E, Ed = 23.5 kJ mol-l, XD= 2.65 nm. For a given support and temperature a small Pd particle will receive more CO molecules per surface Pd atom via surface diffusion than a larger particle (Figure 6A). Indeed, the supply of CO by surface diffusion is proportional to the area of the effective collection zone, while the number of surface Pd atoms is approximately proportional to the surface area of the Pd particle. The ratio of these two areas is larger for the smaller Pd particle. Reverse spillover of CO also explains the increase of ut with decreasing n a t constant T and d. Consider Figure 5 qualitatively. Near 530 K, ut is larger on the low n sample ( B ) than on the high n sample (A). This observation is in agreement with that of Matolin and Gillet, according to whom the areal rate of the C0-02 reaction on Pd/mica increases with decreasing n.g A t this temperature, a CO molecule may remain on the support surface before desorbing long enough to diffuse a considerable distance. With sample A, the Pd particles are close together, the collection zones may overlap, and nearly all CO molecules colliding with the support surface may reach the Pd particles (Figure 6B). However, since the collection zones overlap, a Pd particle will receive only some of the CO molecules impinging within its collection zone. In contrast, when n is small (sample B ) , the Pd particles are far apart, the collection zones do not overlap (Figure SC), and all the CO molecules striking the support within a collection zone reach the corresponding Pd particle. The result is as follows: on a per surface Pd atom basis the rate of arrival of CO molecules and therefore ut is larger for sample B than for sample A. Since the area of a collection zone shrinks with increasing temperature, the reverse spillover of CO is also evident from the temperature dependence of ut. With increasing temperature, ut decreases for two reasons. First, under the given reaction conditions, the probability S,(T) that a CO molecule which reaches the Pd is oxidized decreases. Indeed, with increasing temperature, the reaction rate for the CO-O2 reaction on large Pd single-crystal surfaces d e ~ r e a s e s . ~This ~ decrease in S,(T) with increasing temperature occurs because a CO molecule chemisorbed on the Pd is increasingly likely to desorb before it reacts with an oxygen atom. Second, with increasing temperature the rate of desorption of CO from the support increases, and therefore the effective collection zones shrink (eq 5). When n is large (sample A), the collection zones may continue to overlap as the temperature increases, and any decrease in ut corresponds simply to a decrease in S,. On the other hand, a low n (sample B) leads to nonoverlapping collection zones over the full temperature range, and the effect of the shrinking collection zones on ut is important even at the lower temperatures. In the higher temperature region (Figure 5), (35) Ertl, G.;Koch, J . In Proceedings of the Fifth International Congress on Catalysis;Hightower, J., Ed : North-Holland Amsterdam, 1973; p 969.

the collection zones may not overlap in either case A or B, and the ut versus temperature curves for samples A and B converge. Quantitative. The observed behavior of ut and d , n, and T can be described quantitatively by an effective collection zone model which includes the adsorption, desorption, and surface diffusion of CO molecules on the alumina support. In applying this model, the total supply of CO to a Pd particle is calculated, and then it is assumed that for a given temperature ut is proportional to the rate of arrival of CO by the factor S,(T), which is independent of d and n. Only the diffusion of CO will be considered, since ut is zero order with respect to O2 pressure.'^^^ First the situation is idealized by assuming that the interface between the A1203support and a Pd particle is circular with radius R = d / 2 . The total number of CO molecules arriving per unit time a t a Pd particle is denoted by F m It is the sum of those striking the Pd directly from the gas phase and those impinging within the effective collection zone and diffusing to the Pd particle (eq 6). It

Fco = (Mpd/ds)J+ J T [ ( R+ XD)'- R2]

(6)

is assumed that all CO molecules colliding with the alumina adsorb there. The mean distance XDwhich an adsorbed specie diffuses is given by eq 5. In eq 6 MPdis the average number of surface Pd atoms per particle as counted by TPD of CO, J is the number of collisions of CO molecules per unit time per unit surface area, and d, is the surface density of Pd atoms (-1.3 X 10l6cm"). The last term of eq 6 is the area of the effective collection zone multiplied by J . When the effective collection zones do not overlap this term should account for the CO reaching the Pd particle by surface diffusion; however, when the collection zones overlap, this term will overestimate the actual supply of CO by surface diffusion and the model will not apply. With the above equation for Fco, the turnover rate may be expressed as Ut

(T )

=

(7)

MPd

In the equation for Fco, the unknown parameters are a and the quantity E, -Ed. The hop distance a can be estimated in the conventional way from the lattice spacing of the support surface, -0.4 nm.36 For the modeling results presented below, E, -Ed equals 23.5 kJ mol-'. This value will be justified later. To calculate values of ut from our collection zone model, S,(T) must be estimated. This can be obtained from the variation with temperature of the rate of the C M 2reaction on large Pd single ~ r y s t a l and s~ an estimate of S, a t one t e m p e r a t ~ r e .With ~ ~ the single crystal there is no support, and therefore any decrease in the reaction rate with increasing temperature is related totally to a decrease in S,(T) and not partially to reverse spillover as is being proposed in the present case. For example, the values of the rate of the CO-O2 reaction on Pd single crystals a t 500 and 650 K are approximately Therefore, ~ proportional to 100 and 70, r e ~ p e c t i v e l y . ~ S,(650 K) 0.7Sr(5O0 K). For the modeling exercise presented here, S,(500 K) = 0.65 has been used in reasonable agreement with that given in ref 32, and between 500 and 650 K S,(T) has been approximated by S,(T) = 0.65[(1- 1.3) X 10b(T-500)2]. "hiis functionality of S,(T) is in agreement with the experimental form reported in ref 5.

-

(36) Nolder, R.;Cadoff, I. Trans.Metall. AIME 1965, 233,549.

Langmuir, Vol. 4, No. 3, 1988 727

Oxidation of CO on Pd

. 8

1

.: 0.20i /: e

0.30

6

0 L

2

0.1

0

0

3.25

0

10

20

30

40

50

CO Pressure/kPa

Figure 7. Adsorption isotherms for CO on or-alumina: ( 0 )257 K (m) 273 K; (A) 279 K; ( 0 , O ) 297 K (A)257 K blank. The experimental and calculated results for ut versus d are compared in Figure 1. The points are the experimental values of ut, and the lines connect the calculated values of ut. The data used to obtain the calculated results are summarized in Table I. Where two experimental points have similar values of d (e.g., 3.6 and 3.7 nm), only one value of ut has been calculated. The agreement between the experimental and calculated results is very good. The model does indicate an increase in ut with decreasing d and predicts that the extent of this increase should decrease with increasing temperature. At 550 K and for d > 3.5 nm, the model overestimates ut. The model may not strictly apply in this region, however, since the effective collection zones may overlap. For example, a measure of the distance 2L between nearesbneighbor Pd particles may be obtained if the particles, which in the actual case are randomly distributed on the support surface, are instead considered to be regularly positioned on the support (Figure 6D). Indeed, a t 550 K XD 2.65 nm while L = 0.5(n-1/2- d ) for, say, the d = 3.6 nm sample is -2.35 nm. Thus the effective collection zones may overlap. Next, consider the dependence of ut on n and t (Figure 5). The model according to which the effective collection zones do not overlap should apply to the low n sample (B) throughout the fulltemperature range and might be valid for the high n sample (A) for temperatures above -580 K. Under these conditions and with the already chosen values of a, E, - Ed, and Sr(T),the effective collection zones do not overlap. As can be seen (Figure 5), the model approximately describes the kinetics of the C M 2reaction on the low n sample (B)both as to the magnitude of ut as well as the monotonic decrease of ut with increasing temperature. An independent estimate of E, - Ed can be made from the set of adsorption isotherms of CO on a-alumina (Figure 7). Clausius-Clapeyron plots derived from the data in Figure 7 at CO surface concentrations corresponding to 0.015 and 0.02 of a monolayer are presented in Figure 8. The slopes of the least-squares fit lines yield an average value of the iaosteric heat of adsorption of CO, namely, E, 28 kJ mol-’. In the absence of measurements of Ed,it may be estimated at 15-50% of E,?7-39 If we assume Ed = 0.16E,,we obtain E, -Ed = 23.5 kJ mol-’, the value used in our modeling exercise. Also, we are assured that any CO physieorbed on the support at the reaction temperature is mobile. If E, 30 kJ/mol, from the rule of thumb of TPDM any CO molecules should readily desorb from a-

-

-

-

(37) De Boer, J. H.The Dynamic Characterof Adsorption; Oxford University Press: London, 1953. (38) Hobson, J. P. In The Solid-Gas Interface; Flood, E. A., Ed.; Marcel Dekker: New York, 1967; p 474. (39) Ehrlich, G.In Chemistry and Physics of Solid Surfaces; Vanselow, R.; England, W., Eds.;CRC: Boca Raton, FL, 1982; Vol. 111.

3.50

3.75

4

T-’/103K-’

Figure 8. Clausius-Clapeyron plots for CO adsorption on aalumina: (A)0 = 0.02; ( 0 )0 = 0.015; AH = -28 kJ mol-’.

Al2O3at T > 120 K. If it desorbs, it must be mobile since < E,.

Ed

Conclusion The increase in ut with decreasing d for the CO-O2 reaction on model Pd/alumina catalysts at low pressure and T > 500 K first reported by Ladas et al.’ and confirmed by his study is now thought to be described best by an enhanced supply of CO molecules to the Pd particles resulting from reverse spillover by surface diffusion of CO on the support to the metal particles. Neither an increased effective impingement rate from the gas nor the presence of additional active sites2 on the smallest Pd particles can explain the dependences of ut on n and T as reported in this paper. The dependence of ut on d can be correctly described by assuming that the reaction probability of CO on palladium, Sr(T),is independent of d. In other words, the CO-OPreaction on supported Pd particles appears to be structure insensitive under the reaction conditions used. The reverse spillover of CO molecules from the surface of the support to the support-metal interface accounts qualitatively for the observed trends of ut with T, d , and n. A very simple model based on the existence of nonoverlapping collection zones that surround the metal particles accounts quantitatively for all observations made under conditions of applicability of the model. The agreement between the model and the observations is obtained with only one adjustable parameter that is in fact the surface diffusivity of CO on the support. The latter clearly must be measured, and work toward that goal has been initiated in the laboratory of Professor S. George at Stanford. When and if the missing quantity becomes available, it will then be fruitful to use more elaborate existing models of the “catalytic archipelagonN%to explain all available observations. Finally, it is of interest to speculate on the importance of the phenomenon discussed in this paper. Reverse spillover of reactanta may affect the rate of reactions occurring on supported metal Catalysts if the rate is a function of the concentration of one or more reactants. At the temperatures used in this work (T> -500 K), the CO-02 reaction is first order with respect to CO, and this accounts for the importance of the reverse spillover of CO. In other cases, however, reverse spillover would not be expected to affect ut. One good example is the CO-O2 reaction in the CO inhibition regime (T< -500 K). Under these conditions>2ut is proportional to [02J/[CO], and any enhancement of ut due to the reverse spillover of O2 is probably canceled by the effect of the reverse spillover of CO. Indeed, ut for the C0-02 reaction is independent of d in the CO inhibition regime.’*2-41 (40) Smith, A.

W.;Queta, J. M.J. Catal. 1965,4,163.

Langmuir 1988,4,728-732

728

While the reverse spillover of reactants from the support

to metal may affect the rate of many heterogenous catalytic reactions, it may often be difficult to observe this effect. In the present study, the reaction conditions used resulted in ut being of the same order of magnitude as the CO impingement rate per surface Pd atom. This led to the observation that there was an anomalously large rate of arrival of CO molecules a t the Pd particles and to the original suggestion that surface diffusion was contributing (41) Boudart, M.;Rumpf, F. React. Kin. Catal. Lett. 1987,35/1/2,95.

to the supply of CO to the Pd particle^.^ On the other hand, most reactions on supported metals are studied at higher reactant pressures than those used in the present study and at values of ut considerably less than the reactant impingement rate. Nevertheless, in future work with supported metal catalysts, awareness of the phenomena discussed in this paper may lead to new interesting observations.

Acknowledgment. This work was supported by a continuing grant from NASA/Ames. Registry No. Pd, 7440-05-3; CO, 630-08-0.

In-Plane Structure of Underpotentially Deposited Copper on Gold(11 1) Determined by Surface EXAFS 0. R. Melroy,* M. G. Samant,G. L. Borges, and J. G. Gordon, I1 IBM Almaden Research Center, 650 Harry Road, San Jose, California 95120

L. Blum* Physics Department, College of Natural Sciences, PO Box AT, Rio Piedras, Puerto Rico 00931

J. H. White, M. J. Albarelli, M. McMillan, and H. D. Abruna* Department of Chemistry, Baker Laboratory, Cornell University, Ithaca, New York 14853 Received October 7,1987. In Final Form: February 2,1988 The surface-extended X-ray absorption fine structure (SEXAFS)spectrum of an electrochemically deposited adlayer of copper on a gold(ll1) electrode immersed in solution and under potential control was obtained by employing fluorescence detection and grazing incidence geometry. The X-rays were polarized parallel to the plane of the gold surface. T w o peaks were observed in the radial distribution function, indicating two different near neighbors for copper. The near neighbors were determined to be gold at 2.58 f 0.03 A and copper at 2.92 f 0.03 A. The copper-copper distance is identical with the gold-gold lattice spacing, showing that at full coverage the copper adlayer forms a 1x1commensurate layer on the gold(ll1) substrate. The intensity of the gold-copper scattering relative to the copper-copper scattering suggests the copper atoms sit in %fold hollow sites.

Introduction We have previously reported SEXAFS of both underpotentially deposited metal monolayers and polymer films.'$ Underpotentially deposited layers are good model systems to study at the metal/solution interface because stable monolayer (and submonolayer) coverages can be obtained and because they form well-ordered layer^.^ In situ studies of such systems are relevant because they can provide the structural details necessary for an understanding of the properties of solid/liquid interfaces. In a recent study,' we demonstrated the feasibility of employing SEXAFS as an in situ tool for the structural study of copper monolayers electrodeposited onto gold(1) Blum, L.;Abruna, H. White, J.; Gordon, J.; Borges, G.; Samant, M.; Melroy, 0. J. Chem. Phys. 1986,86,6732. (2) Samant, M.; Borgea, G.; Gordon, J.; Melroy, 0.;Blum, L. J. Am. Chem. SOC.1987,109, 5970. White, J.; Albnrelli, M.; McMillan, M.; A b m a , H. submitted for publication in J. Electroanol. Chem. (3) Kolb, D. in Advances in Electrochemistry and Electrochemical Engineering; Geriacher, H., Tobias, C., E&.; Wiley-Interscience: New York, 1984; Vol. 11, p 125. Adzic, R. In Advances in Electrochemistry and Electrochemical Engineering; Geriacher, H., Tobias, C., Eda.; Wiley-Interscience: New York, 1984; Vol. 13, p 159. Juttner, K.; Lorenz, W.; 2.Phys. Chem. N.F. 1983,152, 211.

0743-7463/88/2404-0728$01.50/0

(111) electrodes. One of the more unexpected and significant findings was the observation of strong scattering from oxygen, which meant that either water or electrolyte ions (sulfate) were adsorbed on the adlayer. It is not surprising that water and/or electrolyte is adsorbed at the interface. Both radiotracel.4 and ex situ experiments6have shown that sulfate is adsorbed on the copper adlayer. The observation of EXAFS from an adsorbed layer is, however, surprising. EXAF'S oscillations of aqueous solutions of metal ions are strongly damped because of their large thermal motions (so-called Debye-Waller factor).6 The observation of well-defined metal-oxygen EXAFS suggests not only that the electrolyte (or water) is adsorbed on the surface but that it is adsorbed at well-defined distances.'S SEXAFS differs from EXAFS of polycrystalline or amorphous materials8 in that the sample is orientated, and hence, the polarization of the incident beam strongly in(4) Hormyi, G.; Rizmayer, E.: Joo, P. J. Electroanol. Chem.1985,162, 211.

0 1988 American Chemical Society