Kinetic studies on the CO oxidation over platinum by means of carbon 13 tracer

Surface Science 79 (1979) 63-75 0 North-Hofland Publishing Company

KINETIC STUDIES ON THE CO OXIDATION OVER PLATINUM BY MEANS OF CARBON 13 TRACER T. MATSUSIIIMA TheReseurchInstitute fir Catalysis. Hokkaido University, Sap.xw MU, Japan Received 8 May ~97S~rn~~s~r~~t received in final form 31 July 197X

The kinetics of the CO oxidation over polycrystaUne platinum was investigated with a bansient isotope tracer method. Transient CO2 production was generated by a dosage of various gas mixtures of ’ 2C0 and 02 to the surface precovered by various ratios of 13C0 to ’ 2CO. The initial ratio of ‘sCO* to I2 CO2 in the CO2 produced transiently equaled that of ’ 3C0 to “CO in the CQ preadsorbed for all the conditions studied; CO must be chemisorbed before being oxidized, i.e. a Langmuir-Hinshelwood mechanism is operative. The ratio of ’ %Y.& to ’ 'CO2 in CQ+J decreased more rapid than the average ratio of t3CO(a) to ‘*CO(a) in CO(a). This apparent inhomogeneity in the reactivity of CO(a) was explained in terms of the boundary reaction model, in which the reaction took place predominantly outside or near the boundary of the CO(a) island.

1. Introduction The CO oxidation over platinum has been the subject of numerous studies [l111. Two reaction paths have been proposed to fit the experimental data under low reactant pressures. The first path involves an irrteraction between an oxygen adatom and a CO molecule in a physically adsorbed mabil precursor state or a gas state (Eley-Rideal process, ER) [3-6,8]. In the second a CO molecule chemisorbed attacks an oxygen adatom (Languor-~~~el~ood process, LH) [3,9,20] _ Recently, certain authors have made attempts to separate the above two processes, or to rule out one mechanism, using a modulated molecular beam technique 11% 14f. Their analysis was limited at high temperatures, and was made on the basis of assumptions that the rates of the ER and LH processes were proportional to the product of the CO collision frequency and the oxygen coverage, and that of the coverages of CO(a) and O(a), respectively, These assumptions, however, have not been verified [7,8]. The purpose of the present work is to rule out one mechanism without any assumptions regarding the rate expressions, and to study the kinetics of another mechanism. These goals were approached by the transient isotope tracer method. 63

2. Experimental The system was a bakeable ultrahigh ~a~uu~~~ apparatus with a base pressure of fess than 3 X 10W7Pa (1 Pa = 7.5 X 10-j Torr). During all the experiments the system was continuously pumped by an ion pump. The catalyst - a polycrystalline platinum foil (30 X 4 X 0.05 mm) - could be heated resistively. Reactant gases, CO, O?,, r3C0, Cl80 (carbon 12 and oxygen 16 are simply designated as C and 0, respective&), and their mixtures were introduced through variable leak valves. The temperature was monitored with a Pt-Pt/Rh thermocouple spot-welded on the substrate. Prior to the experiments the substrate was exposed to oxygen at a pressure of 10e4 Pa and a temperature of 1500 K for several hours, and then flashed to 1700 K in vacuum for a few tens of minutes. Several repetitions of this treatment were sufficient to establish stable catalyst behavior. Po~ycrysta~line Pt foils pretreated similarly [ 151, but heated to 1720 K in O2 for short periods, consisted predominantly of (I 11) oriented crystallites which showed good single crystal grains. Further the individual grains were 0.2 to 0.4 mm across. A Bayard-Alpert gauge was used only for the calibration of the sensitivity of the mass-spectrometer; only a single gas was introduced into the system and the true value of the partial pressure was calculated from reading of the ion gauze and the sensitivity of the gauze in the literature f 161. Aif of these experiments utilize a hot-filament electron supply in the mass-spectrometer. There is some background reaction on the filament which was measured separately with the substrate covered fully by CO at room temperature and subtracted from the measurements made under the same gas compositions.

3, Results and discussion Several preliminary experiments were conducted transient isotope tracer experiments.

to study conditions

suitable for

The isotopic molecules 12C’80 and r3C”60 were coadsorhed on the surface at room temperature, After the evacuation for an hour the substrate was heated slowly (-10 K/s) to 800 K. The mass 31 peak due to the production of ‘3C’80 did not increase during the heating over the background level which was about 2% of the total carbon monoxjde peak. This mixing as the background reaction proceeded on the filament of the mass-spectrometer, because the rate was independent of the substrate temperature in the range 300-1000 K in a steady flow of a gas mixture of 12C’s0 and r3Cr60. That background mixing, however, was negligible in the following tracer experiments.

65

The replacement of an isotopic CO molecule adsorbed by another was studied at various CO pressures and temperatures. fn this experiment the substrate was heated to 800 K in carbon monoxide containing 13C0 and then cooled to the desired temperature. After the pressure reached a steady state, the supply of the isotope gas was stopped and pure l*CO was introduced quickly at the same pressure as before. The partial pressure of 13C0 decreased rapidly, but it was still significant at high temperatures being due to the desorption. After desired intervals the substrate was flashed to 800 K while the I360 peak was recorded. The amount of remaining ‘“CO(a) was determined from the peak area. Some typical results are represented in fig. 1. The amount of the whole CO(a) was constant during the replacement and close to the saturation. The quantities, ‘3CO(a)0, ‘%0(a), and 13CO(a), in the figure are the amounts of remaining ‘%0(a) at the replacement time t = 0, t and infinite, respectively. The value of 13CO(a), was estimated from the isotope composition in the gas phase. The value of “‘CO(a) showed a first order: decay. The halflife, T~/~, was determined as the replacement time required to reach a half of the quantity in the ordinate in fig. 1. The reciprocal (the rate of the replacement) was not influenced from the CO pressure, but depended on the temperature. The activation energy was estimated to be 90 kJ/mol from the plot of log(~~~~~~) against f/T (fig. 2). It was close to the heat of adsorption (or desorption) of CO at relativefy high coverages 1IT-2 f 1. The rate constant was deduced to be 8.4 X 3.0’ ’ expf-90 kJ m&‘/R 73 see-’ . The rate agrees in general with that of thermal desor~tion of CO(a) [ 221. Therefore, in the replacement a gaseous moiecule merely occupies a site previously vacated by the thermal desorption, which controlled the rate. The halflife at room

;

1 0

I

1

50

t/min

100

150

Fig. 1. The first order plots for the replacement reaction, ’ 3CO(a) + ‘*CO(g) + 13CO(g) + ’ *CO(a), at various temperatures. The total coverage, “CO(a) + t *CO(a), was constant during the replacement. The C’O(’ 3COfg)/‘2CO(g) < 0.1) pressures (in Pa) are; (=, v, 0, 0) 4.6 X 10-s, (a) 6.9 X lo-+, (L\) 3.1 X lo-‘, (v) 1.5 x lo+.

66

Pt

-2.0 -

-2.51

1 2.8

I 3.2

3.0

I 3.4

lO’K/T

Fig. 2. The te~pe~at~ro at the CO pressure

dependence of the halflife of 4.6 X 10+ Pa.

of the repfacement

of

f 3&O(a) by * zCO(g)

temperature (-10’ min) is long enough to keep the isotopic composition in CO(a) constant until the onset in the transient experiment. And also CO(a) will be desorbed as COz rather than CO in the following transients. 3.3. IFansient CO2 production The transient phenomena of the COz production depended on the temperature, the pressure and the composition of the gas dosed, and the initial CO coverage, 00. Fig. 3 shows a typical transient COz production induced at B0 < 0.01 and mom temperature. The clean surface was exposed to a gas mixture of CO and 02 (CO/ O2 = 0.24) at a total pressure of 1 X 1V4 Pa, while the COz, CO and 02 peaks were recorded. The COz production was maximized appro~mately 50 s after the onset and decreased rapidly to the background level, showing a small shoulder around 350 S. The 02 pressure remained almost constant, while the CO pressure grew up after 200 s since the oxidation and the adsorption were finishing. The amounts of CO(a) and O(a) were determined during the transient CO* prorlllction by flash desorption and CO-titration [3,23,24]. Fig. 4 shows some typical flash desorption spectra of CO and COz, which were generated at the interval td after the onset of the transient induced under the same condition as shown in fig. 3. In the first deso~tion the substrate was heated to 1100 K after the quick

T. Matsushima

/Kinetic

CO+Oz dose, (CO/On =0.24)

studies on CO oxidation

67

over Pt

Ix 10m4Pa

Of3

&
00

0

200

400

600

800

0.0

t/s Fig. 3. The variation of the partial pressures of COz. CO and 02, and the amounts of CO(a) and O(a), during the CO* transient generated by the CO + 02 pressure jump at 80 < 0.01. The latter were determined mainly by flash desorption: (v) CO(a) desorbed as CO, (0) CO(a) and O(a) desorbed as CO*, (D) CO(a) and O(a) desorbed as COzA (see text), (+) O(a) determined by COtitration.

closure of the leak valve at td, while the CO peak was monitored. Afterward the system was returned to the same condition and the COz peak was recorded during a second desorption experiment. No O2 was desorbed. For td < 50 s CO2 peaked sharply at 370 K (CO*A) and CO showed a very small peak at higher temperatures. The CO2 desorption from td > 50 s was splitted to two peaks. The first peak (C02A) decreased with td and was diminished around 200 s. The second peak around 480 K (C02B) grew up with td. C02B was always accompanied with the CO desorption. These results are represented in fig. 3. The quantities [CO], [C02A] and [CO,] are the amounts of CO desorbed as CO, C02A, and the COzA + COzB sum. The magnitude was defined as the peak area relative to the maximum peak area of CO which was obtained by flashing from room temperature. Assuming that the substrate used here consists predominantly of (111) oriented crystallites, the absolute CO coverage is a half of the coverages in this work since the saturation concentration of CO at room temperature is 7.5 X 1Or4 molecules/cm’ [19]. Since CO* is not adsorbed under the conditions used here [25], CO* observed is produced through an interaction between an oxygen adatom and a CO molecule.

68

T. Matsushima /Kinetic studies on CO oxidation over Pt

bCO2.B

-

( &COr.B

. *i-^

J

11111.z

Fig. 4. The CO and CO2 flash desorption peaks generated from fd after the onset of the transient started with 0,, < 0.01 at 300 K. No 02 was observed. The substrate was heated quickly after the closure of the leak valve. The final temperature was 1100 K.

All oxygen adatoms were removed as CO* because no O2 was desorbed. This will be confirmed below with CO-titration [3,23,24]. At td in the transient CO2 production the supply of the gas mixture was stopped and then a relatively large dose of CO was applied. This CO dosage gave a sharp pulse of CO*. The area gives the amount of O(a). The results are displayed by closed squares in fig. 3. These are in a good agreement with the total amount of COz desorbed. Therefore, [CO,] and [CO,] + [CO] sum equal the amounts of O(a) and CO(a) during the transient. COzA is produced through the LH process since the CO and O2 pressures are negligible. On the other hand, COzB is accompanied with the CO desorption. Thus, CO(a) is desorbed preferentially as CO* rather than CO, as long as O(a) is significant. The removal temperature of CO(a) as CO2 is much lower than that as CO. At higher temperatures the desorption as CO becomes comparable or faster than the LH process. The amount of CO(a) was quite close to that of O(a) in the initial part of the transient production. After 50 s CO(a) increased sharply and O(a) decreased. The transient phenomena depended strongly on 00. Above B0 = 0.5 the transient CO, nroduction was diminished in the background. The intial rate of the transient production showed a 0.7’h order dependence on the pressure of the gas mixture dosed of a fixed composition. The total amount of CO2 produced was almost independent of it. When the ratio of CO to O2 in the gas mixture increased, the ini-

T. Matsushima/Kinetic studies on CO oxidation over Pt

tial rate decreased and the transient production These phenomena will be discussed later.

69

was completed in a shorter period.

3.4. Transient isotopic CO2 production Fig. 5 shows the variation in the partial pressures of 13C02, 12C02, 12C0, 13C0 and O2 during the transient isotopic carbon dioxide production. That transient production was generated by a dosage of the gas mixture of 12C0 and O2 at an apparent total pressure of low4 Pa to the surface precovered by carbon monoxide of Be = 0.23 containing 13C0 at a ratio 13CO(a)/‘2CO(a) = 1.6. The ratio was determined from the pressures of 13C0 and 12C0 dosed before the transient production and then confirmed by flash desorption. During the transient the pressure of 13C0 was less than 10% of that of 12C0. In the transient production 13C02 decreased exponentially while “CO2 showed two peaks. The second peak around 160 s appeared when the “CO pressure grew up. The 13C02 + 12C02 sum agreed well with the total CO2 production induced at the same B0 in the pure 12C0 system. The kinetic isotope effect in the production of 13C02 and 12C02 was negligible. The initial rate of 13C02 was estimated from the plot of the logarithm of the 13C02 pressure against the dose time and shown by a dotted circle. The initial ratio of 13C02 to r2C02 equaled that of 13CO(a) to “CO(a) preadsorbed. The transients generated at the different conditions are displayed in fig. 6. The variations in the

I 06-

:t

I

I

’ ‘2CO+O~

dose,

0.5

pt,

80 =0.23 ‘3CO(o)/

1

I

301K

‘2CO(a) = 1.6

-

04CY 7 0 03> 02-

0.1 -

Fig. 5. The variation in the partial pressures of 1 2C0, ’ 3C0, 02, *‘CO2 and “CO2 during the dosage of the mixture gas (“CO + 02) to the surface precovered by carbon monoxide involvtotal pressure was kept constant at 1 X lo4 Pa. The initial value of ing 13C0. The apparent 13C02 (dotted circle) was estimated (see text).

T. ~~f~us~irna / Kinetic studies on CO oxidation over Pt

70

06-

I-c- '*CO+ 0, dose,

I x 10s4 Pa,

( ‘*CO/O2

= 0 24

1 Pt. 30lK

1 t&=012, ‘3CO(a)/‘2CO(a) = 1.2

Fig. 6. The variaricns in the partial pressures of ’ 3C02 and r *CO2 during the transient isotope tracer experiments. The series represented by circles and triangles were generated separately at different 00 values. The initial values of ’ 3C02 were estimated.

‘*CO 13C0 and O2 pressures are essentially the same as those in fig. 5. The total amount of CO2 produced decreased with an increase in BO, however, the initial ratio of ‘3C02 to 12C02 equaled that of 13CO(a) to 1 “CO(a) in the CO adsorbed in

Table 1 The results of the transient isotopic carbon dioxide production generated by dosing the mixture gas of 1 *CO and O2 at various coverages of (’ 3CO(a) f 1%0(a)) at 301 K; the initial pressures observed were 0.82 X 10m5 Pa for 12C0 and 7.6 X 10-s Pa for 02; the final pressures were 1.7 X lo-’ and 7.2 X lo-’ Pa for “CO and 02 respectively ~--__-.____.Initial isotope Initial isotope No. Ro ~CO2~ 00 (relative ratios in COz, ratios in CO(a), value) ’ 3c02 /l 2co2 l 3CO(a)l’2CO(a) _--..-.. 1 0.05 0.95 0.62 t 0.1 0.75 + 0.1 2 0.12 1.2 * 0.1 1.2 f 0.1 0.96 3 0.18 1.1 1.3 t 0.1 1.4 f 0.2 4 0.23 1.5 ?: 0.2 1.0 1.6 + 0.1 5 0.24 1.5 i- 0.2 1.0 a) 1.8 i 0.1 1.4 t 0.2 6 0.28 1.6 * 0.1 0.99 I 0.30 2.1 t 0.2 2.1 ’ f 0.2 0.92 8 0.32 0.70 1.6 * 0.2 1.8 i 0.1 2.3 * 0.2 9 0.43 0.37 2.2 + 0.2 -_._ --a) Standard.

T. Matsushima /Kinetic

studies on CO oxidation over Pt

71

advance. The equality between them held true for the coverage range studied, 8e = 0.02-0.43. Some typical results are listed in table 1. Accordingly, a CO molecule must be adsorbed before being oxidized, i.e. the LH process is operative. In the case of the ER process the isotope ratio in CO* should equal that in gaseous CO. 3.5. The kinetics of the Langmuir-Hinshelwood process The initial rate of the transient CO* production, R,,(C02), is shown as a function of Be in fig. 7a. It was constant to 19~= 0.3 and thenafter decreased sharply. The total amount of COa produced relative to the maximum of the CO desorption, Q(CO,), is shown in fig. 7b. It went down with an increase in Be and became very small around Be = 0.5. Now, we will discuss on the rapid decrease of 13C02 during the transient. As may be seen in figs. 5 and 6, the isotope ratio in CO* decreased more rapid than that in the adsorbed layer. And further 13C02 went down monotonically while “CO2 showed the two peaks. The second peak was accompanied with the growth of “CO. Those facts suggest that the later CO is adsorbed, the higher the reactivity is. It will be confirmed in the case of 8,, = 0.43 and ‘“CO(a)/ “CO(a) = 2.2 in fig. 6. The total CO coverage was ca. 0.6 when r2C02 peaked. The

A

I

o-

-

‘I

100

(%I (b)

0.0

0.2

_

04

'0 0.6

Fig. 7. (a) The dependence of the initial rate of the CO2 transient, Ro(C02), on the initial CO coverage, 80, at the total pressure of 1 X low4 Pa (CO/O? = 0.24) and room temperature. (b) The variation of the amount of CO2 produced, Q(COz), and the contribution from CO preadsorbed, r(-o(pd), with the initial CO coverage. The data are from table 1.

72

T. Matsushima /Kinetic

studies on CO oxidation over Pt

total amount of CO* produced until that peak is less than 0.1 in the unit for Q(C0,). Thus, the isotope ratio in CO(a) is close to unity at that time, however, the ratio of l3 CO2 to “COz was ca. 0.2. The quantity, rCO(nrea,&), in fig. 7b represents the contribution of CO preadsorbed to the total CO? production, which was calculated from the amount of the 13C02 and the initial isotope ratio in the CO(a). Generally, above Be = 0.2 the rapid decrease of 13C02 became clear, since the consumption of CO preadsorbed was a little. This fact does not suggest that the ER process contributes at higher CO coverages, because the equality between the initial isotope ratio in CO, and that in CO(a) holds true for a wide range of the CO coverage (see table 1). There are two possible reaction models to explain such inhomogeneity in the reactivity of CO(a). In the first model the inhomogeneity is due to the intrinsic heterogeneity of the CO adsorption sites on the surface. The surface model involves strong and weak binding sites. A CO molecule is expected to adsorb preferentially on strong binding sites and to react first from weak binding sites. A high density of defect sites has been observed on a polycrystalline Pt foil [15], but the density decreased with an increase in the annealing temperature. The defects on the surface used here might be low, since the substrate was annealed several times and for long periods at temperatures as high as 1700 K [ 151. And further there is no difference in the heat of desorption of each CO(a) as seen in the replacement experiments in section 2. If the defects like step sites are appreciable on the surface, CO(a) on the defects should be replaced by gaseous CO with the rate of much less than that of CO(a) on the terraces, since the difference of the heat of adsorption between on the step sites and terraces is more than 20 kJ/mol [ 151. Another model is more favorable, where such inhomogeneity of the reactivity of CO(a) is induced during the transient CO2 production. The inhomogeneity is explained in terms of a boundary reaction model, in which an oxygen adatom interacts with a chemisorbed CO molecule outside or near the boundary of the CO(a) island. Oxygen dissociatively adsorbes on vacant Pt sites outside the CO(a) island. ’ 3CO(a) and “CO(a) molecules outside or at the boundary of the CO(a) island migrate and then react with oxygen adatoms to give an initial 13C02/12C02 ratio equal to that of 13CO(a)/12CO(a) in the CO preadsorbed. The isotope ratio in CO(a) outside or at the boundary of CO(a) island equals initially that in the whole CO(a). The former, however, approaches that in gaseous CO faster than the latter, since gaseous CO is adsorbed preferentially outside the CO(a) island and then migrate to the boundary. The CO(a) island has been observed on Pd [26]. The initial CO2 production rate, Ro(C02), depended on the pressure of the mixture gas with a 0.7th order. The initial isotope ratio in CO2 still equaled that in CO(a). Some typical results are listed in table 2. Be was kept around 0.15 and the composition of the dosed gas was fixed at the ratio C3/02 = 0.1-0.07, except No. 6 and 7. Since CO2 is produced via the LH process, the initial rate will be determined by either the O2 adsorption or the LH surface process. The total amount of CO2 produced increased slightly with the pressure of the gas dosed and

2 6 E 8

N

3

a) No.

_II_.-.-~..

5 in

AC319K 1 2 3

_

______-_..-,

0.41 0.2 f. 0.27 0.1X

00

0.20 0.12 0.27 --..table 1 was used as the

_,._. -.._-.---_. At 337 K I 2 3 4

No.

isotope

standard.

1.5 t 0.1 1.2 L 0.1, 1.8 k 0.2

1.9 r 0.2 I,6 Ir 0.2 I.& I!T0.2 1.0 t 0.x

_--..---- ..---11___

i%zQ(a)l”2CO(a)

ratios in cm(a),

Initial Kq.1

1.5 I.2 1.3

0.64 2.5 1.9 1.8

~

Initial isotope ratios in G02, “3CO#V02

___ ___.

______

t: 0.4 L 0.2 : O-2 f 0.1

1.7 +-0.2 1.2 f. 0,l 1.9 * 0.2

2.1 1.9 1.5 1.1

_.__.

___.__~__ __.__. .-- --...

-__--~._I_

-

,.__

QW2

YJ

__

.“__

0.47 0.69 0.87

0.22 0.83 0.73 1.1

.-. ..^--~ll_.~-

-.y_.“~s-e.-------~ -.

,.‘mwwe--~

_ I .“.

10T5

26 22 29

34 25 24 1.5

(%I

‘COfpreads)

the initial ’ zCO and O2 pressures were 0.88 X IO-$ Pa and X.0 x Pa for 0%

X 10n5

(relative valuei a)

Ro

Table 3 Transient isotopic carbon dioxide production at higher temperatures; respectively; the final pressures were 1.8 X 10 -’ Pa for 1 *CO and 7.6 _____,_-_l.---.-- ,.. __~.l____--_-.--

Pa

$

3

0

F: 8 bii: 2. 2

2

r: T: a. 5

h 2 h’

a

. * =,

g

E

.?I 3&Y

T. ~arsMshima i Kinetic studies on CO oxidation over Pt

75

the contribution parameter, rCO(preadS), was almost constant. Table 3 summarizes the transients at 319 and 337 K. The initial isotope ratio in CO2 was still equal to that in CO preadsorbed. Q(C0,) increased and ?CO(preadS) decreased slightly with an increase in the temperature. The fact indicates that the LH process has a significant activation energy [3] and then the role in eliminating CO(a) from the surface increases with the temperature [7,11],

Acknowledgement This work was supported

in part by the Matsunaga Science Foundation.

References [1] I. Langmuir, Trans. Faraday Sot. 17 (1922) 621. [Z] H.P. Bonzel and R. Ku, J. Vacuum Sci. and Technol. 9 (1972) 663. [3] H.P. Bonzel and R. Ku, Surface Sci. 33 (1972) 91; H.P. Bonzel and J.J. Burton, Surface Sci. 52 (1975) 223. [4] Y. Nishiyama and H. Wise, J. Catalysis 32 (1974) 50. 1.51R, Ducros and R.P. Merrill, Surface Sci. 55 (1976) 227. [6] M. Alnot, J. Fusy and A. Cassuto, Surface Sci. 57 (1976) 651. [7] T. Matsushima, D.B. Aimy and J.M. White, Surface Sci. 67 (1977) 89. [S] J.M. White and A. Golchet, J. Chem. Phys. 66 (1977) 5744. [9] R.L. Palmer and J.N. Smith, Jr., J. Chem. Phys. 60 (1974) 1453. [IO] CA. Becker, J.P. Cowin and L. Wharton, J. Chem. Phys. 67 (1977) 3394. [ 11] T. Matsushima, Bull. Chem. Sot. Japan 51 (1978) 1956. [ 121 C. Pacia, A. Cassuto, A. Pentenero and B. Weber, J. Catalysis 41 (1976) 455. [ 13) T. Engel and G. Ertl, in: Proc. 7th Intern. Vacuum Congr. and 3rd Intern. Conf. on Solid Surfaces, Vienna, 1977,~. 1365. [ 141 T. Engel and G. Ertl, Chem. Phys. Letters 54 (1978) 95. [ 1.51D.M. Collins and W.E. Spicer, Surface Sci. 69 (1977) 85. [16] T.A. FIaim and P.D. Ownby, J. Vacuum Sci. and Technol. 8 (1971) 661. [17] CM. Comrie and R.M. Lambert, J. Chem. Sot. Faraday Trans. 1 72 (1976) 1659. [ 18] D.M. Collins, J.B. Lee and W.E. Spicer, Surface Sci. 55 (1976) 389. [ 191 G. Ertl, M. Neumann and K.M. Streit, Surface Sci. 64 (1977) 393. [ 201 R.W. McCabe and L.D. Schmidt, Surface Sci. 65 (1977) 189. [21] R.W. McCabe and L.D. Schmidt, Surface Sci. 66 (1977) 101. [22] W.L. Winterbottom, Surface Sci. 37 (1973) 195. [23] H.P. Bonzel and R. Ku, Surface Sci. 40 (1973) 85. (241 T. Matsushima and J.M. White, J. Catalysis 39 (1975) 265. [25] P.R. Norton, Surface Sci. 44 (1974) 624; P.R. Norton and P.J. Richards, Surface Sci. 49 (1975) 567. [26] G. Ertl and P. Rau, Surface Sci. 15 (1969) 443.