J. Phys. Chem. 1986, 90, 6194-6200
6194
Acknowledgment. This research was supported by both the C N R S and AFME under CNRS-PIRSEM contract no. 2031. The authors also thank Dr. E. Amouyal for performing the laser experiments and Drs. R. Gaboriaud and F. H. Quina for valuable discussions, and acknowledge Mr. T. Ben Chabanne for his participation in part of this work.
The water-cation interaction amplitude, which is certainly an essential factor governing the decay rate, would not be an independent parameter but would be correlated to the micellar electric effective charge. Micelles intervene not only through A$ but are capable also of slowing the micellized cation decay when the two reaction partners are solubilized in two distinct phases. In accordance with ref 27, the above considerations may be thought to apply to other organized systems: artificial or natural vesicles and membranes.
CO
+ O2 Reaction on Rh(ll1):
Regisfry No. TMB', 21296-82-2; NaLS, 151-21-3; Brij3S,9002-92-0; CTAB, 57-09-0; DTAC, 112-00-5; CS,7440-46-2; NH4*, 14798-03-9; Li, 7439-93-2; Na, 7440-23-5.
Steady-State Rates and Adsorbate Coverages
S. B. Schwartz,t L. D. Schmidt,*? and Galen B. Fished Department of Chemical Engineering and Materials Science, University of Minnesota, Minneapolis, Minnesota 55455, and the Physical Chemistry Department, General Motors Research Laboratories, Warren, Michigan 48090- 9055 (Received: February 7, 1986)
Steady-state kinetics of CO oxidation on clean Rh(ll1) were measured over a wide range of CO and O2gas-phase compositions and surface temperatures for pressures between 10" and lo-* Torr. Coverages were measured by XPS during steady-state reaction. Below 425 K the reaction rate increases with temperature with an activation energy of 20 kcal/mol, while above 450 K the rate decreases with temperature with an activation energy of -7 kcal/mol. At low temperatures and in excess CO, the reaction rate is proportional to PozPco-', while in excess O2 and at high temperatures the reaction rate is negative order in Poz and more than first order in Pco Near stoichiometricreactant ratios XPS shows that CO is the dominant adsorbed species below 400 K, while from 425 to 900 K the surface is nearly saturated with oxygen. Experimental rates and coverages are fit qualitatively by using a simple Langmuir-Hinshelwood model assuming competitive adsorption, although adsorption parameters for CO and O2are not in agreement with clean surface values. Modified Langmuir-Hinshelwd models involving strong inhibition of oxygen adsorption by adsorbed CO, either through coverage-dependentsticking coefficients or a dependence of the heat of adsorption of oxygen on the CO coverage, give good agreement with UHV parameters, measured CO and oxygen coverages, and reaction rates.
The oxidation of carbon monoxide on noble metal catalysts at reactant partial pressures near 10 Torr is a major reaction in the automotive catalytic converter. The reaction of CO + O2has been studied on Pd, Pt, and Rh.'-8 The common reaction mechanism is generally believed to be a reaction between adsorbed C O and 0 atoms. Titration experiments on Pd, Pt, and Rh suggest2 that oxygen adsorption is inhibited by CO adsorption much more than CO adsorption is inhibited by oxygen adsorption, because when oxygen is preadsorbed on Pd, Pt, or Rh and then exposed to CO, immediate reaction occurs. However, if the order of exposure is reversed, reaction occurs only after an induction period. The adsorption and desorption of CO, 02,and C02on Rh have been studied extensively. CO adsorbs molecularly on polycrystalline' and (1 1 1) Rh surface^^.'^ and desorbs from Rh( 1 11) at 500 K with a desorption activation energy of 31 k c a l / m ~ l . ~ * ' ~ Oxygen adsorbs dissociatively on polycrystalline Rh,'.'' Rh( and R h ( l 1 l)l3*I4surfaces, and oxygen desorbs from Rh( 11 1) between 800 and 1300 K with a clean surface activation energy of 85 kcal/mol.I0 At high oxygen coverage molecular oxygen desorbs with an activation energy near 10 kcal/mol.2-'0*'' At low temperatures coadsorbed CO and oxygen segregate to separate domains on Pt( 111) and Rh( 11 Although some early studies reported oxygen diffusion into bulk R h , 1 3 g L 5 Fisher and SchmiegI2 have shown that on clean Rh the only mechanism for oxygen removal from the surface is desorption or reaction. C 0 2 does not chemisorb on Rh( 11 1) above room temperature. Steady-state reaction on polycrystalline Rh a t pressures near lo-* Torr' and in a recent study near lo4 Torr16 were found to be qualitatively consistent with a surface reaction between adsorbed CO and oxygen, although further assumptions had to be made to explain University of Minnesota. 'General Motors Research Laboratories.
0022-3654/86/2090-6194$01 .50/0
the reaction order in oxygen. From steady-state kinetics alone, mechanistic details of the reaction cannot be determined. In this study we obtained detailed steady-state kinetics of the C O O2reaction on Rh( 1 11) at pressures where we could also measure adsorbate coverages by XPS during reaction (10"-1 O-* Torr). From these data we have developed models that fit both kinetics and coverages over a wide range of reaction conditions with a minimum number of adjustable parameters.
+
Experimental Section All experiments were carried out at General Motors Research Laboratories in an ultrahigh vacuum (UHV) system described in detail p r e v i o ~ s l y . ' ~ *It' ~ was equipped for quadrcpole mass
(1) Campbell, C. T.; White, J. M. J . Cafal. 1978, 54, 289. (2) Engel, T.; Ertl, G. Adu. Cutal. 1979, 28, 1 . (3) Barteau, M. A.; KO,E. I.; Madix, R. J. Surf. Sci. 1981, 204, 161. (4) Golchet, A.; White, J. M. J . Catal. 1978, 53, 266. (5) Daniel, W. M.; White, J. M. I n f . J. Chem. Kinet. 1985, 17, 413. (6) Matsushima, T.; Matsui, T.; Hashimoto, M. J . Chem. Phys. 1984,81, 11.
(7) Gland, J. L.; Kollin, E. B. J . Chem. Phys. 1983, 78, 963. (8) Norton, P. R.; Creber, D. K.; G d a l e , J. W., personal communication. (9) Thiel, P. A.; Williams, E. D.; Yates, J. T., Jr.; Weinberg, W. H. Surf. Sci. 1979, 84, 54. (10) Root, T. W.; Schmidt, L. D.; Fisher, G. B. Surf. Sci. 1983,134, 30. (11) Matsushima, T. J . Catal. 1984, 85, 98. (12) Fisher, G. B.; Schmieg, S. J. J . Vac. Sci. Techno[.,A 1983, A l , 1064. (13) Thiel, P. A.; Yates, J. T., Jr.; Weinberg, W. H. Surf. Sci. 1979, 82, 22. (14) Root, T. W.; Schmidt, L. D.; Fisher, G. B. Surf. Sci. 1985,150, 173. (15) Campbell, C. T.; Shi, S.; White, J. M. Surf. Sci. 1979, 2, 382. (16) Lintz, H. G.; Weisker, T. Appl. Surj Sci. 1985, 24, 251. 259. (17) Fisher, G. B. Chem. Phys. Left. 1981, 79, 452
0 1986 American Chemical Society
CO
+ O2 Reaction on R h ( l l 1 )
The Journal of Physical Chemistry, Vol. 90, No. 23, 1986 6195
spectrometry, Auger electron spectroscopy (AES), X-ray photoelectron spectroscopy (XPS), ultraviolet photoelectron spectroscopy (UPS), and low-energy electron diffraction (LEED). Ion and titanium sublimation pumps and a turbomolecular pump were separated from the analysis chamber by variable valves to control pumping speeds over a wide range. Crystals were cut to within 1/20 of the appropriate orientation by using standard methods. The crystals had been used extensively for TPD experiments, so the only contaminants observed were traces of S, C, 0, and N, which were removed by annealing in Torr Torr of 02.Oxygen was removed by heating in of CO. Before each reaction AES and occasionally XPS or UPS were used to determine surface cleanliness. After reaction the only species on the surface detectable by AES, XPS, or UPS were C O and/or 0. Crystals were heated by passing current through tantalum leads spot-welded to the edge of the crystal. The temperature was measured with a precision of f l K with a Chromel-Alumel thermocouple spot-welded to the face of the crystal. Rates were measured by using the analysis chamber of the UHV system as a continuous-flow steady-state reactor. Reactants were admitted to the reactor through separate variable valves and were pumped with either the turbomolecular pump or the ion and titanium sublimation pumps to obtain desired residence times between 1 and 12 s. The mass spectrometer partial pressures were calibrated against an ion gauge for each gas. Reaction rates (rate of C 0 2 production) were calculated from product partial pressures through the mixed reactor equation rR = PVNO/rARTg where P is the partial pressure of the reaction product, 7 is the reactor residence time, Tg is the gas temperature (300 K), A is the catalyst surface area (0.90 cm2), and Vis the reactor volume (53 L). The conversion was always maintained below 10% so that differential rates were obtained. The lowest measurable reaction rate was 5 x io9 molecules/(cm2 s). Reaction rates were measured for mixtures from 8/1 to 1/4 Pco/Po,, while partial pressures were typically varied from 2 X to 8 X lo-’ Torr. Rates were measured for Rh temperatures from 300 to 875 K, which is 75 K below the temperature that carbon diffusion into bulk Rh becomes significant.I8 CO and oxygen coverages were determined by measuring the area under the O(1s) XPS peak associated with each adsorbate. O( 1s) XPS spectra were deconvoluted by fitting curves obtained with submonolayer amounts of pure CO or oxygen. Because of the proximity of the O(1s) peak to the much larger Rh(3p) peak a t 520 eV, base-line shifts made computer deconvolution impractical. All XPS spectra were taken at reaction partial pressures and reaction temperatures. XPS spectra taken during reaction had no measurable effect on the reaction rate. Data were recorded digitally with an O( 1s) spectrum typically requiring 15 min. Attempts to use AES to measure adsorbate composition led to changes in the reaction rate and surface composition because the electron beam caused the formation of a graphitic carbon layer on the rhodium surface.
-
Results Kinetics. All rates were measured as described above by establishing reactant flows, ramping the temperature to a specified value, and monitoring the C 0 2 (44 amu) signal from the mass spectrometer. Two typical recorder outputs are shown in Figure 1 for the indicated temperatures. Reaction rates are seen to be identical when the temperature is approached from above or below, and steady states are attained quickly. C 0 2 production rates for various compositions as a function of temperature are shown in Figure 2. Data in Figure 2a are for 1.0 X Torr of O2 with different pressures of CO, while (18) Delouise, L.A,; Winograd, N. Surf. Sci. 1984,138,417; 1985, 259, 199. (19) Matsushima, T. Surf. Sci. 1979, 79, 63.
fi
623K
\ 0
5
298K
15
10
20
TIME (min)
Figure 1. Typical recorder traces of Pco2 vs. time as the surface temperature is increased and decreased in steps for two different reaction compositions. Rates quickly attain steady-state values except near the rate maxima where longer transients exist. The reactor time constant was less than 3.5 s for these measurements.
10’0
, , 1 400
800
600
400
600
800
T (K)
Figure 2. Rate of C 0 2production vs. temperature for (a, left) Po, = 1.0 X Torr and varying Pco and (b, right) Pco = 1.0 X lo-’ Torr and varying Po2.
Figure 2b shows results for 1.0 X Torr of CO with different pressures of 02.Rates are believed to be accurate to within a factor of 2 based on ion gauge and mass spectrometer calibrations. Relative rates were reproducible to within 10% in the high- and low-temperature regions and 20% near the rate maxima. The increased uncertainty near the rate maxima is caused by longer transients and higher conversions. At low temperature the reaction rate increases with temperature with an activation energy E, of 19.9 f 2.0 kcal/mol. As the temperature is increased further, the rate goes through a maximum that shifts to higher temperature with increasing Pco and shifts to lower temperature with increasing Po,. At high temperature the rate decreases with increasing temperature with an E, of -7.0 f 1.0 kcal/mol. Order plots for C O and O2 are shown in Figure 3. The rate is approximately 1.5 order in Pco at low Pco and -1.0 f 0.1 at high Pco. The 1.5 order at low Pco has a higher uncertainty because the order decreases continuously with increasing Pco. It was not possible to study lower partial pressures of CO because it was desired to keep reactant partial pressures at least a factor of 100 higher than background. The order in P92decreases from 1.0 f 0.1 at low Po, to approximately -0.5 at high Pol as shown in Figure 3b. At a given composition, the reaction orders are not only a function of reactant partial pressure but also of temperature. The order plots show that by decreasing the temperature, the order in Pco can be reduced from 1.5 to -1.0, while the order in Po, can be raised from -0.5 to 1.0. Couerages. O( 1s) XPS spectra taken at various temperatures Torr during reaction with 5.0 X Torr of CO and 1.0 X
-
Schwartz et al.
6196 The Journal of Physical Chemistry, Vol. 90, No. 23, 1986
4
1.oL
5x10“
0
I
A’ 0 300
I
I
350
I
400
-
450
T (K)
10’0 10-8
Figure 5. Steady-state CO coverages (eco, filled circles) and oxygen coverages (eo, open circles) vs. temperature during exposure to 5.0 X Torr of 02.The dashed line indicates the Torr of CO and 1.0 X reaction rate for the same conditions. 10-7
10-6 10-8
10-7
10-6
Po2 ( Torr 1
Pco ( Torr)
Figure 3. Rate of C 0 2 production at various temperatures vs. (a, left) Pco with Po2 = 1.0 X loW7 Torr and (b, right) Po2 with PCo = 1.0 X Torr.
10.’ TOU 3M)K
co
300K 375K 390K 400K
410K 420K 440K
450K
475K
10.’ TO, o2 300K Clean Rh 300K
-53575
Energy ( e V )
-52s75
Torr of CO and Figure 4. XPS spectra during exposure to 5.0 X 1.0 X Torr of O2 at various temperatures. Also shown are spectra of clean Rh at 300 K and Rh exposed to 1.0 X lo-’ Torr of CO and 0 2 .
of O2are shown in Figure 4. Also shown are spectra for the clean Rh surface before and during exposure to pure CO and O2at 300 K. The O(1s) peak maximum was found to occur at 528.7 eV for adsorbed oxygen and 530.8 eV for CO. The width of the CO peak suggests two binding states of CO. This has been attributed to bridged and linear binding sites on both Rh(l1 1)18and R ( l 1 Delouise and Winograd18 report peak positions of 529.7 eV for adsorbed oxygen on R h ( l l 1 ) and a doublet at 530.7 and 532.1 eV for CO adsorption on Rh( 11 1). The coverages of CO and 0 normalized to the coverages at 300 K and 1.0 X Torr (Figure 5 ) were calculated by integrating the area under their O( 1s) XPS lines from Figure 4. Also shown in Figure 5 is the rate measured for the same conditions. The coverage data indicate that at low temperatures, where CO inhibits
Figure 6. Steady-stateCO coverage determined by XPS during exposure to 1.0 X Torr of CO and during reaction with 1 X lo-’ Torr of CO and 1.0 X lo7 Torr of 02.
the reaction, the surface is saturated with CO, while at high temperatures, where O2 inhibits the reaction, the surface is nearly saturated with oxygen. At low temperature the surface is depleted of CO and enriched in oxygen as the temperature is raised. Note that by the time the reaction rate reaches a maximum, there is less than 1% of a monolayer of CO on the rhodium surface, and at higher temperatures there is only oxygen on the surface. The transition between a nearly saturated CO surface to a nearly saturated oxygen surface occurs within at least a 50 K interval. At 300 and 375 K the XPS peaks in Figure 4 a t 528.7 eV appear to indicate 0.2 monolayers of atomic oxygen on the surface, although this could have been a binding state of CO or a transient state frozen by low temperature. Since this peak was easily removed during cleaning, we do not believe it was associated with an oxide or a contaminant. A CO adsorption isobar determined by XPS is shown in Figure 6 for 1 .O X 10” Torr of CO along with the CO coverage during Torr of CO and 1.0 X torr of 02. The reaction in 1.0 X kink in the isobar for pure CO is quite reproducible and correlates with the two CO binding states observed in XPS and TPD.’ Figure 6 shows that CO is near adsorption4esorption equilibrium under reaction conditions below 450 K. The coverages of CO and oxygen determined by XPS as a function of reactant composition at 450 K are plotted in Figure 7. Also shown are the reaction orders at 450 K with respect to CO and 02.Figure 7 illustrates that the surface composition is changing in the temperature region where the reaction order is changing. It is also evident from Figure 7 that more than 0.5 monolayer of adsorbed oxygen is required to inhibit the reaction rate, whereas CO inhibits oxygen adsorption and reaction at approximately 0.1 5 monolayer. Data taken at different coverages support the conclusion that a lower coverage of CO is required to inhibit CO oxidation rather than oxygen.
CO + O2Reaction on Rh( 11 1)
The Journal of Physical Chemistry, Vol. 90, No. 23, 1986 6197
.
+1.5 I
I
10’4
Transient /
1013
T=450K 0.8
0.6
c
I
\
114 1/2 1/1 2/1 411 811
pco /Po2
Figure 7. Steady-state CO and oxygen coverages and reaction orders at 450 K at several compositions. Note that the pressure scale for the order in O2 is different than that for the CO order since the former is for varying Po2 while the latter is for varying PCO
Reaction on Rh(100). The rate of CO oxidation on Rh(100) was also measured. At 1.0 X lo-’ Torr of C O and O2the rate is shown as a function of temperature in Figure 8. The rate rises with temperature until about 450 K, then drops slightly until about 525 K, and again rises to a maximum a t 625 K. Transients between 450 and 575 K lasted up to 45 min, while transients from 575 to 625 K were so long that the rates were difficult to establish. Long transients and rate discontinuities with pressure variation have recently been reported during C O oxidation on Rh(100) under similar reaction condition^.^ The maximum rate on Rh( 100) is nearly the same as on Rh( 1 11) but is shifted 175 K higher in temperature.
-
Discussion The reaction between CO and O2on noble metal catalysts is typically modeled as a surface reaction between adsorbed oxygen Although evidence for multiple binding staes and and C0.1*839*18 interactions between adsorbed CO and oxygen indicate that strict adherence to the Langmuir-Hinshelwood (L-H) assumptions should not be expected to give an accurate description of the data, we shall first examine this model to determine to what extent these deviations from L-H assumptions affect steady-state kinetics. Since the low-temperature coverage of oxygen is lower than that of CO, even though it has a higher clean surface heat of adsorption, the L-H model must assume unreasonable values of some parameters to fit the coverages at low temperature. We then show how two modifications of this model using coverage-dependent sticking coefficients or coverage-dependent heats of adsorption improve the fit of the model to the coverages. Finally we compare our results with those from other metals and with recent highpressure kinetics on Rh( 1 1 1). A stringent test of L-H models is a comparison of the coverages they predict with experiment because the models assume single binding states. Since we measured both kinetics and coverages, the validity of these models can be examined in detail. Langmuii-Hinshelwood Model. The elementary steps in the reaction could be
’o’\OO
t\ f 400
500
600
700
800
900
T(K) Figure 8. Plot of steady-state and peak transient rate of C 0 2production vs. surface temperature during Rh( 100) exposure to equimolar pressures of 1.0 X lo-’ Torr of CO and 0,along with a plot of the rate as a function of temperature for 1.0 X lo-’ Torr of CO and 0,reaction on Rh( 111). Each experimental run is represented by its own symbol. Rates are quite different on the two crystal planes with the rate being an order of magnitude higher on the Rh(100) surface at high temperatures.
adsorption and rapid O2dissociation, one obtains a rate expression of the form
with
Oco =
Kcopco (1 + KCOPCO + K02PO2)
and 60 =
KO2PO2 (1 + Kcopco + KO2PO2)
(6)
to yield rR =
kRKC&02PC@02
(1 + Kcopco
+ Ko2P02)2
(7)
In these expressions Kco and Ko2 are the CO and oxygen adsorption-desorption equilibrium constants, Ki = Koi exp(E,/RT) =
Soi (exP(Ei/RT)) v o i ( 2 ~ MTRg ‘1’ )
(8)
where E , S,, and vd are the heat of adsorption, sticking coefficient, and desorption preexponential factor for species i . If O2 dissociation is slow and requires a vacant site, or if two sites are required for adsorption, then KqPo, in the above equation would simpl; be replaced by (KO2Po2)ll2.The first-order adsorption characteristics of oxygen on Rh( 111)20appear to justify the form of eq
7.
OW
++
COW + O(a)
2%) +
(2)
Comparison with Data. W e will now determine what parameters give a best fit of the L-H model to the kinetic data. The
(3)
where the subscript a refers to an adsorbed species and g refers to a gas-phase species. For Langmuir isotherms with competitive
(20) Yates, J. T., Jr.; Thiel, P. A.; Weinberg, W. H. Surf. Sci. 1979, 82, 45. (21) Klein, R. L.; Schwartz, S.; Schmidt, L. D. J . Phys. Chem. 1985,89, 4908.
Schwartz et al.
6198 The Journal of Physical Chemistry, Vol. 90, No. 23, 1986
rf?
(*) 10'2
,ui/
IO"
, , ,
, , ,
,,,,,,
/ , , , , /
T(K)
,
(
TIK)
0.6
10'0
10.'
1o
10-610.8
-~
P (Torr)
Pco (Torr )
02
Figure 9. Langmuir-Hinshelwood model rate vs. (a, left) PCOwith PO, = 1.0 X lo-' Torr and (b, right) Po2 with Pa = 1.0 X lo-' Torr. Good qualitative agreement with the data of Figure 3 is obtained.
unity term in the denominator of eq 7 will only be important at high temperatures when both coverages approach zero. Our results show (Figure 5 ) that there is always more than half a monolayer of CO or atomic oxygen on the surface up to 875 K during reaction. Therefore, our experiments never probe sufficiently high temperature to be in the low coverage regime, and the unity term in the demonimator is neglected. This means that kR, KO, and Kco cannot be determined independently from these experiments, and the appropriate L-H expression is therefore rR =
kR~coKo,~opo2
(Kcopco + Ko,po,)2
(9)
When only kineitic data from Figure 1 were used, the following parameters were found to give a best fit of the data for the form of eq 9: rR = (2.5 X 101se-10/Rr)(5X 10-4e30/Rr)Pco(5 X 105e'2/R3P02 ((5 x
+
I O - ~ ~ ~ O / R ~(5P x~ 1~ 0 5 e 1 2 ~ R ' 3 ~ o , ) 2
(10)
In this expression all activation energies are in kcal/mol, rR and reaction preexponential are in molecules/(cm* s), and the adsorption-desorption equilibrium preexponentials are in Torr-'. The fit to the rate as a function of temperature at lo-' Torr of CO and O2is shown in Figure 11 (LH) and is generally quantitative with the only deviation being near the rate maximum. In fitting the rates with eq 10, one preexponential and one activation energy could not be determined independently because the zero coverage limit was never attained. We assumed KCo = 5 X 10-4e-30000/RTTorr-' based on experimental values obtained and determined the reaction rate constant and by Root et effective oxygen adsorption-desorption equilibrium constant KO, by fitting eq 9 to the kinetic results. The reaction activation energy of 10 kcal/mol is lower than the 30-45 kcal/mol derived from TPD6 but close to the 14 kcal/mol observed during titration studies on polycrystalline Rh.I5 The value of 12 kcal/mol calculated for the heat of adsorption of 0, is very low in comparison to the tightly bound atomic state observed in TPD but is close to the value of 8-10 kcal/mol reported for molecular oxygen on Rh The oxygen heat of adsorption in the L-H model is lower than observed during TPD because CO inhibits reaction at low temperature, and the model fits this by making Eo2 < Eco. As a result of the low heat of adsorption for oxygen, the preexponential is forced to be unrealistically high in order to predict the oxygen inhibition observed
0.4 -
0.2 -
"'i
e t 0.2
0 0
- Y 0
0
Figure 10. Calculated coverages for 5.0 X lo-* Torr of CO and 1.0 X lo-' Torr of oxygen as a function of temperature for (a, top) L-H model, (b, middle) a model assuming that sticking coefficients are functions of coverage, and (c, bottom) a model assuming Eo(6'co). Points are ex-
perimental coverages from Figure 5 . at high temperature. Using the adsorption-desorption equilibrium constant observed during TPD would predict oxygen inhibition from 300 to 900 K. A similar case of an apparent heat of adsorption during steady-state reaction being much lower than measured during TPD has been interpreted as reaction through a precursor intermediate for the NO CO reaction on polycrystalline Pt.22323 The order plots predicted by the L-H model are shown in Figure 9. The model qualitatively predicts the observed order variation with partial pressure and temperature shown in Figure 3. When Pco is raised or the temperature is lowered, the order in CO can be reduced to -1 and the order in O2raised to + 1. In excess CO and high temperatures the experimental rate is 1.5 order in Pa and --0.5 order in Po, whereas the model predicts a rate proportional to Pco/Po2. Note also that the experimental variation in order from 1 to -1 occurs over a smaller pressure change than predicted by the model. Calculated Coverage from L-H Model. The CO and oxygen coverages as a function of temperature calculated from eq 9 for 1.0 X lo-' Torr of 0, and 5.0 X lo-* Torr of CO are shown in Figure loa. The calculated coverages exhibit the same general
+
-
(22) Mummey, M. J.; Schmidt, L. D. Surf. Sci. 1981, 109, 43. Surf.Sci. 1980, 94, 57.
(23) Schmidt, L. D.
CO
+ O2 Reaction on R h ( l l 1 )
The Journal of Physical Chemistry, Vol. 90, No. 23, 1986 6199
behavior as the experimental results but are shifted 20 K higher in temperature. A better fit of coverages would result in a somewhat poorer fit of the rates. Note that the CO coverage data have been corrected, so it is normalized to the maximum coverage of the high-temperature state shown in Figure 6. In summary, the simple L-H model of eq 7 gives a qualitative description of the experimental rates and coverages. The four parameter fit (Ec,, Eo,, vc,, and v0J, while using only the kinetic data for parameter determination described both kinetics and coverages fairly well. This model correctly predicts the changeover from positive- to negative-order kinetics in reactant partial pressure, and changes in adsorbate concentrations as reactant pressure and temperature are changed. The L-H model, however, is inconsistent with UHV experiments in several respects. The low-coverage heat of adsorption of oxygen on R h ( l l 1 ) observed during TPD is nearly 85 kcal/mol, while the L-H model predicts 12 kcal/mol, and the oxygen adsorptiondesorption preexponential is several orders of magnitude higher than observed in UHV experiments. Titration studies indicate a strong dependence on which species is adsorbed first, while the L-H model does not allow for this. The L-H model also does not allow for adsorbate segregation observed by other investigator^.'^ We will now examine two models that address these discrepencies. Coverage-Dependent Sticking Coefficient. One model that fits both the coverages and kinetics but is more consistent with UHV experiments assumes that oxygen adsorption is more strongly inhibited by adsorbed C O than C O adsorption is by oxygen. The differential equations for surface coverage are doc0 -dt
- kaC#CO(
- OCO - POBO) -
kdBCO
- krBCOBO
dB0 _ dt - Kao2P02(l- BCO~~CO - BO) - ~RBCOBO
(
)
(12)
In the CO and O2adsorption terms Po and Pc0 are the number of competitor sites blocked by an adsorbed oxygen atom and CO molecule, respectively. The exact expressions for and PCocan be complicated as they are functions of surface geometry, adsorption mechanisms and energetics, etc. We fit our results by using the functional forms
Bc0 = N / ( 1
+(N-
l)Bco) = 2/(1 Po =
x
+ Bco)
for N = 2 (13) (14)
These yield sticking coefficients between 0 and 1 and allow C O to block N sites for oxygen adsorption at low CO coverages. At high CO coverage CO only blocks one oxygen adsorption site. Two oxygen atoms block only one C O adsorption site, allowing CO to adsorb onto a monolayer of oxygen. Several values of N were examined with N = 2 yielding the best fit to the data. A reaction rate constant of 4 X lo2’ exp(-17000/RT) was calculated by fitting the coverages of Figure 5 and kinetics of Figure 2 to an expression of the form rR = kRBcoBo. The co adsorption-desorption equilibrium constant KCo = 3 X lo-’ exp(27000/RT) was determined by first scaling coverages measured during reaction to the maximum coverage of the more tightly bound state in Figure 6 and then finding a best fit to the data. Since the activation energy for oxygen desorption is approximately 85 kcal/mol, oxygen desorption should be slow at the temperatures studied and was neglected in the model. As shown in Figure 10b the coverages are fit quantitatively. Unlike the L-H model, the coverage-dependent sticking coefficient model predicts the slightly lower than unity oxygen coverage at high temperature. The coverage-dependent sticking coefficient model fit the kinetics just as well as the L-H model in Figure 9. If the parameters chosen to predict the coverages in Figure 10b are used, the largest deviation (near the rate maximum) in the predicted rates is less in Figure 11 is than a factor of 2. For example, curve So(B)Lo a fit to the rate as a function of temperature at lo-’ Torr of C O and 02. The model also accurately predicts the positive and
1019
1018
loq7
(e;*) rR
10’6
Low Pressure Model
1015
-
1014
-
1013
10’2
rnl1 I”
300
500
700
900
T(K) Figure 11. Plots of reaction rate vs. temperature for several models: At high pressures, the curve through the data in open circles is based on a model to fit that data and data on supported Rh at high pressures.** The nearly parallel curve, [L-H, S(O)],is the extension of the low-pressure models to high pressures (15 Torr) using eq 17 and 18. The low-pressure data are fit by the simple Langmuir-Hinshelwood model (LH) and by the model modified with coverage-dependent sticking coefficients [SoHigh-pressureexperimental rates on Rh(1 l l) (open circles) are from Fisher et aL2’ for 8 Torr each of CO and 0,; low-pressure rates (filled circles) were measured with lo-’ Torr each of CO and 02. negative activation energy regimes as well as the rate maximum shift with temperature. While the model does not quantitatively predict reaction orders at high temperature in excess oxygen, it does qualitatively fit the observed transition from positive to negative order kinetics in Pco as Pco is increased. This model corrects two aspects of the L-H mechanism that are inconsistent with UHV results. The parameters used in this model are very close to those measured by TPD. When CO adsorption is allowed onto a monolayer of oxygen, but not oxygen adsorption onto a monolayer of CO, the model is consistent with titration data, showing that preadsorbing a monolayer of CO completely inhibits reaction, whereas preadsorbing oxygen does not. Even though the steady-state solution of the coverage-dependent sticking coefficient model is a cubic polynomial, there is only one physically real steady state for the parameters used here. The model may predict multiple steady states under some conditions. This will be discussed in a subsequent paper. Coverage-Dependent Desorption Activation Energy. The order predictions of the L-H model and the coverage-dependent sticking coefficient model are accurate when the coverage of oxygen is below 0.5 monolayer. The breakdown in the models at high oxygen coverage where the kinetics are greater than first order in Pco and negative order in Po, could be due to the formation of segregated domains of CO and oxygen. As the oxygen coverage increases, oxygen islands may grow and cause the kinetics tq deviate from the models. (It is doubtful, however, that islands would persist on the surface at reaction temperatures.) It is impossible to determine from kinetic data alone whether the deviation is due to a change in the heats of adsorption of C O and oxygen, a coverage-dependent reaction activation energy, or adsorbateadsorbate interactions. From TPD results Matsushima et a1.6 postulated that the reaction activation energy decreases due to repulsive CO-oxygen interactions as oxygen islands grow on the surface and CO adsorbs into the islands. This would also cause the heats of adsorption to be functions of coverage.
6200 The Journal of Physical Chemistry, Vol. 90, No. 23, 1986
If the heat of adsorption of oxygen is a function of the CO coverage, such as Eo, = Eoo, - aeco then the coverages during reaction can also be fit by using literature values for clean surface heats of adsorption. The w e r a g e s of CO and oxygen are given by eq 5 and 6, but the oxygen adsorptiondesorption equilibrium constant should then be written as The behavior of this model is not as simple as those just discussed. Figure 1Oc shows that when nearly the same clean surface parameters as in the coverage-dependent sticking coefficient model and Koo2= 1.0 X lo4, Eo, = 50 kcal/mol, and a = 40 kcal/mol of monolayer are used, this model can predict a discontinuous coverage vs. temperature curve because of a jump between two steady-state branches. This is the only model of the three examined here that did not show measurable CO coverage above 425 K. The model also exhibits multiple steady states under some circumstances and can predict multipeaked rate vs. temperature curves as seen during CO O2reaction on Rh( 100). This model will also be discussed in a later paper. Comparison of R h ( l l 1 ) with Other Surfaces. Many aspects of this reaction on R h ( l l 1 ) are qualitatively similar to reaction on polycrystalline Rh.1*10*16324 Oxygen inhibition at high temperature, a coverage-dependent reaction activation energy, and variable reaction orders with surface coverage have all been suggested for CO oxidation on polycrystalline Rh (although oxygen inhibition was not observed during steady-state reaction). At high PO, the reaction is zero order in oxygen,'J6 and this has been explained either as a rate limited by lack of adsorbed oxygen16 or reaction through a precursor.' Neither mechanism is consistent with the oxygen inhibition observed in this study on R h ( l l 1 ) and the former is inconsistent with the high coverage of adsorbed oxygen present during reaction on Rh(ll1). During steady-state kinetics at lo4 Torr16 reaction orders in CO similar to those reported here are observed. C O oxidation on Rh( 11 1) also has many similarities with Pt(ll1). On Pt( 111) the reaction activation energy decreases with increasing coverage, and coadsorbed CO and oxygen form segregated domain^.^ One major difference between Rh( 1 11) and Pt(ll1) is that oxygen desorbs from the Pt surface in TPD below 900 K, while it remains on Rh until 1300 K. the kinetic implications are that at high temperatures, where negative one-halforder kinetics in O2are observed on Rh( 11 l ) , zero- or positiveorder kinetics would be expected on P t ( l l 1 ) . Although reaction on Rh(100) has not been observed to exhibit oscillation^,^ it does exhibit reaction pathologies similar to those observed on Pt(100) and polycrystalline Pt.Z5326 Models that predict rate vs. temperature curves such as that in Figure 3 could exhibit oscillations and island formation. Comparison with Rates at High Pressures. Reaction rates as a function of temperature for a 1/1 Pco/Po2 mixture a t a total Torr and at 16 Torr27are shown in Figure pressure of 2 X 11. The experimental rates indicate that the high-pressure rates were measured entirely within the low-temperature regime where rate increases with temperature. Reaction orders further indicate that the surface was always near saturation with CO because rates
+
(24) Kim, Y.; Shi, S.K.; White, J. M. J . Carol. 1980, 61, 374. (25) Ertl, G.; Norton, P. R.; Rustig, J. Phys. Rev. Lett. 1982, 49, 177. (26) Turner, J. E.; Sales, B. C.; Maple, M. B. Surf. Sci. 1981, 103, 63. (27) Fisher, G . B.; Goodman, D. W.; Oh, S. H., to be published.
Schwartz et al. were positive order in O2 and negative order in CO. In the limit of low temperature and high CO coverage the L-H expression derived for the low-pressure rates reduces to kRK02P0,
rR = (17) Kcopco The rates predicted by this model for 2 X l V 7 Torr (LH) by using eq 10 and for 15 Torr (L-H) by using eq 17 are shown in Figure 11 and are seen to give a good fit to the data between lo-' and 10 Torr. If the same assumptions are applied to the sticking coefficient model, the same rate expression derived by Oh et a1.28for the high-pressure data is obtained kao,Po, rR = kdCO kacopco
(18)
Rates calculated by using this expression are shown in Figure 11 for two sets of parameters: those used by Oh et al. to model the high-pressure rates and those used here to model the low-pressure rates. The rates predicted by eq 18 a t 15 Torr with the lowpressure S(e) parameters are nearly identical with the L-H rates but differ from the high-pressure model of Oh et al. because of different sticking coefficients and reaction preexponentials used to model the high-pressure rates. It should be noted that neither model developed to fit the low-pressure rates accurately predicts the high-pressure rates without simplification. This implies that at least one of the parameter choices made at low pressures is not appropriate for high pressure kinetics. The high-pressure model (even before simplification to the low temperature, CO saturated surface form) fails to predict rates and coverages measured at low pressure even qualitatively except at low temperature when the CO coverage is high.
Summary X-ray photoelectron spectroscopy has been used to determine adsorbate coverages and chemical state during the steady-state CO + O2reaction on Rh( 11l), and this information has been used to fit deviations from the L-H model. Rates and coverages could be modeled accurately with modified Langmuir-Hinshelwood models under most reaction conditions using clean surface parameters. In order to fit both kinetics and coverages with parameter values close to those determined during TPD, details of C O and oxygen adsorption must be taken into account. When the effects of CO and oxygen adsorption inhibition are incorporated into the model, reaction parameters determined from isotherm and TPD experiments can be used to qualitatively fit the results. Low-pressure kinetics and models are qualitatively consistent with high-pressure kinetics measured at low-temperature and high-CO coverage. Models developed at both pressure extremes predict the same reaction orders and heats of adsorption in the low-temperature regime. At high temperature, however, only the low-pressure model accurately predicts reaction rates. Differences in sticking coefficients and desorption preexponentials account for differences in predicted rates. This implies that low-pressure kinetics can be very useful in bridging the pressure gap as long as care is taken to reproduce the surface conditions existing at high pressures. Registry No. 02,7782-44-7; CO, 630-08-0; Rh, 7440-16-6. (28) Oh, S. H.; Fisher, G . B.; Carpenter, J. E.; Goodman, D. W. J . Cotol. 1986, 100, 360.