Langmuir 1994,10, 699-708 699

A Laser-Induced Fluorescence Study of OH Desorption from Pt in H20/02 and H20/H2 Mixtures

Erik Fridell' and Arne Ros6n

Department of Physics, Chalmers University of Technology and University of Giiteborg, S-412 96 Giiteborg, Sweden

Bengt Kasemo

Department of Applied Physics, Chalmers University of Technology and University of Giiteborg, S-412 96 Giiteborg, Sweden

Received December 7,1992. I n Final Form: January 7, 1994"

The desorption of OH from high-temperature-annealed polycrystalline Pt in H20/02, H20/H2 (D2), and HzO/Oz + Hz (D2) mixtures has been investigated using the laser-induced fluorescence (LIF) technique. The measurements were done in the pressure range 1-lo00 mTorr and temperature interval 900-1300 K. The experiments yield new interesting information about water decomposition on Pt, unimolecularly and in the presence of 02 and/or H2, and provide complementary results to earlier studies of water formation from H2 + 0 2 . There is a small but significant OH desorption even in pure H2O. In H20/02 mixtures the OH desorption rate has a maximum at 22-30% 0 2 content. The absolute desorption rate increases nearly linearly with total pressure. The maximum yield occurs at successively lower relative 0 2 concentration, the higher the pressure. In H20/H2 mixtures the OH desorption is strongly suppressed due to reaction of OH with adsorbed hydrogen atoms. H20/D2 mixtures yield OD and OH desorption. The experimental results are well described by a previously developed kinetic model for the H2 + 1/202 - H20 reaction. Approximate analytical expressions are derived for the kinetics and compared with full numerical solutions. Interestingly the dominant path for HzO decomposition seems to be the unimolecular decomposition H2Oa

Ha + OH. rather than the bimolecular reaction HzOa + 0' e 20Ha (a = adsorbed). The promoting effect of oxygen is kinetic, rather than mechanistic; i.e., the presence of 0 atoms moves the surface equilibrium to higher OH concentration, rather than providing a new reaction path via direct H2O' + 0. interaction. An evaluation procedure is briefly outlined on how to extract the true OH desorption rate from the measured LIF' intensities at pressures where collisional gas-phase quenching and pumping speed effects are important.

1. Introduction At low temperature, water adsorbs m~lecular ly~-~ on

platinum surfaces. No decomposition to hydroxyl and hydrogen has been observed with pure H2O. In contrast, the presence of adsorbed oxygen, atomic or molecular, induces the decomposition of water to form stable hydroxyl species,&g as evidenced by electron energy loss spectros- copy (EELS) results a t low temperatures.

At high temperatures, above 900 K, OH has been observed to desorb from Pt due to water dissociation on a partially oxygen covered surface, by using the laser- induced fluorescence (LIF) t e c h n i q ~ e . ~ ~ ~ ~ ~ ~ ~ An inter- esting question is if H2O dissociates a t high temperatures on Pt, even in the absence of adsorbed oxygen.1°J2 (The strong promoting effect of oxygen for OH formation makes it a nontrivial experimental problem to prove that H2O dissociates in the absence of 0 2 . 9

@Abstract published in Aduance ACS Abstracts, February 15,

(1) Creighton, J. R.; White, J. M. Chem. Phys. Lett. 1982,92, 435. (2) Fieher, G. B.; Gland, J. L. Surf. Sci. 1980,94, 446. (3) Fisher, G. B. General Motom Research Publication, GMR-40071

(4) Sexton, B. A. Surf. Sei. 1980,94, 435. (5) Langenbach, E.; Spitzer, A.; Lath, H. Surf. Sci. 1984,147, 179. (6) Ibach, H.; Lehwald, S. Surf. Sci. 1980,91, 187. (7) Fieher, G. B.; Sexton, B. A. Phys. Rev. Lett. 1980,44, 683. @)Melo, A. V.; O'Grady, W. E.; Chottiner, G. 5.; Hoffman, R. W.

(9) Wagner, F. T.; Moylan, T. E. Surf. Sci. 1987, 191, 121. (10) Talley, L. D.; Lin, M. C. Chem. Phys. 1981,61,249. (11) He, J.; Kaeemo, B.; Ljungstrbm, S.; W n , A.; Wahnetrbm, T.

J. Vac. Scr. Technol., A 1987,5, 523. (12) Fridell, E. Chem. Phys. Lett. 1992,188,487. (13) Fridell, E.; H e W i , B.; Kaeemo, B.; LjungetrBm, 5.; RoeBn, A.;

Wahnstrbm, T. J. Vac. Sci. Technol., A 1991, 9, 2322.

1994.

PCP-171,1982.

Appl. Surf. Sci. lSSS,21, 160.

In this work we have investigated the OH formation/ desorption from H2O on Pt at high temperatures (>900 K) both in pure H2O gas and in the presence of 0 2 , H2, or D2. The work represents a considerable extension of the first study presented recently.I2 In that work the focus was on the temperature dependence of the OH desorption rate in different H20/02 mixtures. The maximum in the OH desorption flux (0.04 site-l s-l a t 1200 K and 50 mTorr) occurs a t around 30% oxygen. The "apparent" activation energy for desorption was found to vary from 1.6 to 1.9 eV, depending on the gas mixture. These results were well reproduced by the kinetic modeP employed to analyze the data. In the present work we present more extensive measurements of OH formation/desorption in H2O/O2 mixtures and new results from H20/H2 (D2) and HzO/Oz + H2 (D2) mixtures.

There are several motivations to study the H2O de- composition on Pt: (i) To study H20 decomposition is a natural complement to previous studies of OH formation in the water formation reaction (H2 + l/202 - HzO), to provide further insight into the kinetics and reaction paths. Several different kinetic models have been developed to describe the catalytic water formation reaction (H2 + l/202 - H2O) on Pt.1P22 The present data provide additional tests of some of these models (generally, the models should

(14) Hellsing, B.; Kaeemo, B.; Ljungstrbm, S.; Roe& A.; Wahmtrbm,

(15) Hellsing, B.; Kasemo, B. Chem. Phys. Lett. 1988,148,466. (16) Hellsing, B.; Kasemo, B.; Zhdanov, V. P. J. Catal. 1991,132,210. (17) Hsu, D. S. Y.; Hoffbauer, M. A.; Lm, M. C. Surf. Sci. 1987,184,

T. Surf. Sei. 1987,1891190, 851.

25. _ _ (18) Verheij, L. K.; Hugenachmidt, M. B.; Poeleema, B.; Comea, G.

Surf. Sei. 1990, 233, 209.

0~~~-7~63/94/2410-0699$04.50/0 0 1994 American Chemical Society

700 Langmuir, Vol. 10, No. 3, 1994

equally well apply to the decomposition and to the formation of water). (ii) The water decomposition reaction and how it may be influenced by coadsorbates is of general interest in the search for reaction schemes to produce H2 from water. (iii) OH production by catalytic decompo- sition of water is of interest in catalytic combustion, since it may influence and stabilize gas-phase combustion.23-26

The water formation reaction on Pt is one of the most extensively studied catalytic reactions.14 The reactants, H2 and 0 2 , as well as the product, H20, and the reaction intermediate, OH, have been studied thoroughly. The kinetics has been explored using several experimental techniques such as time-resolved EELS,28 molecular beam technique^,^^^^^^^^ and reactions in isotropic gas mix- tures.2*3032 The importance of OH as a surface interme- diate in this reaction is well established by EELS detection at low temperature~~133 as well as by LIF detection of OH molecules in the gas phase by Lin's p 0 ~ p , ~ ~ J ~ 1 ~ 9 ~ in our l a b ~ r a t o r y ~ ~ J ~ * ~ * ~ ~ and recently by Williams, Marks, and S ~ h m i d t . ~ ~ ~ ~

We have used LIF to monitor OH production on the surface during this reaction and simultaneously studied the water production rate.13J4930996.89 The results include OH desorption and water production rates at different gas mixtures, total pressures, and surface temperatures.30 A maximum in the water production occurs at 1522% H2, while the OH desorption shows a maximum at 3-8% H2. Both the water production rate and the OH desorption rate increase almost linearly with pressure up to 100 mTorr. The water formation rate decreases slowly with increasing temperature between 900 and 1200 K. The temperature dependence of the OH desorption rate shows a complicated dependence on the gas mixture due to kinetic effects. The activation energy for OH desorption was found to be 2.0 & 0.16 eV from a kinetic analysis of experimental data.38 The LIF technique also makes it possible to examine the rotational distribution of the desorbing OH."@t37 In general, the experimental data have been subject to kinetic m0deling.l3-~6~38

(19) Verheij, L. K.; Hugemchmidt, M. B.; Cblln, L.; Poelsema, B.; Comaa, G. Chem. Phys. Lett. 1990,166,523.

(20) Verheij, L. K.; Freitag, M.; Hugenschmidt, M. B.; Kempf, I.; Poelwma, B.; C o w , G. Surf. Sci. 1992,272,276.

(21) Anton, A. B.; Cadogan, D. C. Surf. Sci. Lett. 1990,239, L548. (22) Anton, A. B.; Cadogan, D. C. J. Vac. Sci. Technol., A 1991, 9,

1890. (23) Driscoll, D. J.; Campbell, K. D.; Luneford, J. H. Adu. Catal. 1987,

35,139. (24) Pfefferle, L. D.; Griffin, T. A.; Winter, M.; Crosley, D. R.; Dyer,

M. J. Combust. Flame. 1989, 76,325. (25) Pfefferle, L. D.; Pfefferle, W. C. Catal. Rea-Sci. Eng. 1987,29,

219. (26) Norton, P. R. In The Chemical Physics of Solid Surfaces and

Heterogeneow Catalysis; King, D. A., Woodruff, D. P., Eds.; Elsevier: Amsterdam, 1982; Vol. IV, p 27.

(27) Ertl, G. Catal. Sci. Technol. 1983,4, 209. (28) Germer, T. A.; Ho, W. Chem. Phys. Lett. 1989,163,449. (a) D'Evelyn, M. P.; Madix, R. J. Surf. Sci. Rep. 1984,3,413. (30) LjungetrBm, 5.; Kasemo, B.; Rash, A.; WahnstrBm, T.; Fridell,

(31) Fisher, G. B.: Gland. J. L.: Schmien, S. J. J. Vac. Sci. Technol. E. Surf. Sci. 1989, 216, 63.

Fridell et al.

-. 1982,20, 518..

(32) Gland, J. L.; Fieher, G. B.; Kollin, E. B. J. Catal. 1982, 77, 263. (33) Mitchell. G. E.: White. J. M. Chem. Phvs. Lett. 1987.135. 84. (34) Hsu, D. 8. Y.; Lm, M.'C. J. Chem. Phyk 1988,88, 1.' (36) Talley, L. D.; Saunders, W. A.; Bogan, D. J.; Lm, M. C. Chem.

'

Phva. Lett. 1981. 78. 500: J. Chem. Phvs. 1981. 76. 3107. 736) Ljungs&m, S.; Hall, J.; Kasemo, B.; &en, A.; WahnstrBm, T.

(37) WahnstrBm, T.; Ljungstrdm, S.; Rosh, A.; Kesemo, B. Surf. Sci. J. Catal. 1987, 107, 648.

1990.234. 439. ~ (38) W'ahnatrBm, T.; Fridell, E.; LjungstrBm, S.; Hellsing, B.; Kasemo, B.; W n , A. Surf. Sci. 1989, 223, L905.

(39) Rdn,A.;LjungstrBm, S.; WahnstrBm,T.; Kesemo, B. J.Electron Spectroec. Relat. Phenom. 1986, 39, 15.

(40) Marks, C. M.; Schmidt, L. D. Chem. Phys. Lett. 1991,178, 368.

0.06 -I

c I I 8 3 t . 8 305.9 307.0

5 0 0 2 4 0 ) & I 6 ' A /!L\\ h 1:- I I

307 1 3073 3075 nm Figure 1. A scan of the recorded OH fluorescence intensity va laser wavelength at a catalyst temperature of 1300 K, a total pressure of 50 mTorr, and a relative oxygen concentration of K = 0.42 ( K = pod@% + p ~ p ) ) . Assignments of the peaks are done according to ref 56.

The significant and important difference between the H2O decomposition reaction and the water formation reaction is that the latter is a rapid, essentially irreversible, exothermic reaction converting Hz and 02 to HzO while the former is predominantly areuersible reaction at quasi- equilibrium, where H2O and 0 2 landing on the surface decompose to OH, 0, and H that react again, to form H2O and 0 2 that desorb from the surface, except for a minute fraction that desorbs as OH and H2. Thus, the water formation reaction occurs far from equilibrium, while in the decomposition reaction there is a quasi-equilibrium between the gas phase and the adsorbed phase.

2. Experimental Section 2.1. Setup. The experimental arrangement and procedure

have been described elsewhere." Briefly, the setup consists of a resistively heated Pt foil, 99.95% pure, and 22 mm x 3 mm X 0.025 mm in size, mounted in a turbo or Roots (200 d / h at 100 mTorr) pumped stainless steel vacuum chamber. The temper- ature of the foil is determined by its temperature-dependent resistivity and is kept constant by microcomputer control. The reagent grade water ie kept in a stainlesa steel can, connected to the main chamber via heated regulating valves. It is heated to about 90 O C and kept at constant temperature (f2 O C ) . The relative partial pressures of the hydrogen, oxygen, and water are measured with a quadrupole maas spectrometer, cross-checked with a capacitance manometer. The mass flow of Ha and On are measured and controlled by mass flow meters. The total pressure can be increased by gradually closing a valve to the Roots pump, and thus keeping the mass flow constant, or by increasing the mass flow with approximately constant pumping speed (the pumping speed of the Roots pump is almost independent of pressure in the range of interest here).

A small fraction of the OH molecules which are produced by 0 + H recombination or by HzO decomposition are thermally desorbed from the surface and are detected using the LIF technique. A frequency-doubled, excimer-pumped, dye laser, inducing the 0-0 vibrational transition in the AQ+ - XaII electronic band at 306.3-307.5 nm, probes the OH concentration a few millimeters from the catalyst sample. A part of this band taken at 50 mTorr total pressure, a relative oxygen pressure, K , of 0.42 (K = PO,/@% + p ~ p ) ) , and a surface temperature of 1300 K is shown in Figure 1.

2.2. Rotational Quenching. By measuring the heights of different peaks in the LIF spectra, such as those shown in Figure

Desorption from Pt in HzOIOZ and HzOIHZ Mixtures Langmuir, Vol. 10, No. 3, 1994 701

1, and normalizing them with the corresponding transition probabilities (Einstein coefficients), one obtains the correspond- ing initial-state population. Boltzmann analysis of the scan given in Figure 1 gives a rotational temperature of 428 f 6 K. Since the surface temperature is much higher than the temperature of the reactant gas mixture, the rotational distribution will change, with, for example, increasing pressure, due to collisions with the reactants. At the present pressuresthe gas-phase mean free path between collisions is quite small, for example, a few centimeters at 1 mTorr and only 1W cm at 100 mTorr. Since the temperature of the foil is much higher than the temperature of the reactant gas mixture, there will be temperature gradients in the gas, close to the catalyst, and thus also considerable rotational-state redistribution of the desorbing OH molecules, due to gas-phase collisions. (We have previously shown that it is still possible to derive the true temperature of the desorbing OH molecules by an extrapolation procedure, to zero pressure.s7)

Due to the collisional redistribution of rotational states, the height of a particular peak in the LIF spectra will not correspond to the same fraction of the total number of desorbed OH radicals, when parameters like the mixing ratio, total pressure, and surface temperature are varied. In order to calculate the total number of OH radicals in the LIF detection volume element, one should therefore sum up the contributions from all populated levels. This can, as described in ref 37, be achieved using measured population distributions for the various parameter values. Another method is to probe a particular level (N = 4 in this case), whose fraction of the total number of OH radicals is insensitive to changes in the rotational distribution, over the range of parameter values of interest. Both methods have been used in the experiments reported here and give consistent results.

2.3. Electronic Quenching. The fluorescence light detected in the measurements originates from the spontaneous emission when OH radicals are deexcited back to the ground electronic state after initial laser excitation to the f i t electronically excited state. However, if the excited molecule collides with a gas molecule, before radiative deexcitation, the deexcitation may instead proceed via radiationless energy t rader in the collision." We denote the probability of this electronic quenching as Q. In order to obtain the total number of OH radicale from LIF spectra, one must know the fraction of the laser-excited molecules that fluoresce. Further, to be able to compare the relative number of OH radicals for different pressures and gas mixtures, the quenching constant for each component of the gas mixture must be known.

The total fluorescence yield is determined by the ratio between the radiative deexcitation rate, A (Einstein's spontaneous emis- sion coefficient), and the total deexcitation rate, Q + A

(2.1)

where N a is the number of excited molecules. The value of Q depends on the total pressure and on the gas mixture, and can be written as a sum over all collision partners, i, with quenching rate constants k,(i) times the densities ni of molecules i:

To determine Q, one can measure the lifetime, T , of the excited state. If Nw(0) molecules are laser-excited at time t = 0, the excited state will then decay as

(2.3)

Thus, a measurement of the time-dependent decay curve of the fluorescence yield wil l determine Q + A. Figure 2 shows four such curves at T = 1300 K and three different Hn0/0a mixtures (A-C) and one HJOn mixture (D). T h e decay curves are obtained by averaging over lo00 laser shots using a digital oscilloscope. The lifetime is defined as the time it takes for the intensity

to decrease to l/e of the initial intensity, I(0) (Le., r = (Q + A)-l). Figure 3 shows how T changes with total preseure in H*0/0*

(41) Crcmley, D. R. Opt. Eng. 1981,20,311.

0 0 0.5 1 0 1 5 v

TIME (ps)

Figure 2. Timadependent decay curvesof the fluoregcence from OH: A%(u'=O) - Xan(u"=O), R1(2) for T = 1300 K in H*0/0* mixtures (A) p = 100 mTorr, K = 0.6, (B) p = 100 mTorr, K = 0.2, and (C) p = 26 mTorr, K = 0.2 ( K =PO,/@% + p ~ p ) , and the HdOa mixture (D) p = 25 mTorr, a = 0.1 (a PHJ@H, + PO,)).

I I

K = 0.4 500s14 300 0 2 5 5 0 7 5 100

pressure (mTorr) Figure 3. Lifetime, T = (Q + A P , of laser-excited OH molecules as a function of pressure for Hn0/0a mixtures with K = 0.4 (fiied circles) ( K = p d @ q + p ~ p ) ) and Hd01 mixtures with a = 0.1 (crosses) (a = PHJ@H* + p a ) ) . The solid lines represent the results expected using the quenching rate constants given in the text. mixtures (40% 02) and HJOa mixtures (10% Ha). Using the known value for A (0.71 pa in this caseu), Q is found. From several decay curves, the k,(i) for the individual collision partners have been calculated with the result k,(HaO) (6.82 f 0.39) X 10-l0 cms/s and k,(Oa) = (1.10 f 0.29) X 10-lo cms/s. The lifetimes calculated from these quenching rate constants are ale0 illustrated by the curves in Figure 3. The quenching constants determined here agree fairly well with those reported for high-pressure flames (k,(HaO) = (3.7-6.6) X 10-10 cms/s, k,(01) = (0.814.91) X 10-10 cms/s, k,(H1) = (0.78-1.37) X W0 ~ m ~ / s ) . ~ l The experimental results presented below have been adjusted for electronic quenching to obtain the correct relative OH concentration. The small temperature dependence of the quenching constants was not c0nsidered.a

2.4. Influence of Pumping Speed. To obtain the OH desorption rate, &,, from the measured LIF ~ignal,Zo~, requires a certain evaluation procedure, which becomes increasingly more important as the pressure rises. At suffciently low pressures, where the molecular flow condition prevails, the situation is simple and straightforward. Here IOH is directly proportional to +OH. This proportionality holds as long as the mean free path for molecule-molecule collisions is considerably larger than the distance, z, between the sample and detection volume (a few millimeters), i.e., up to a pressure of at most -10 mTorr. Provided angular and rotational distributions are u n a f f d , no corrections are required in order to compare results at different pressures, gas mixtures, etc., since the measured signals in this regime are unaffected by gas-phase collisions.

(42) Dimpfl, W. L.; Kineey, J. L. J. Quant. Spectrosc.Radiat. ZYamfer

(43) Gudmundson, F.; Fridell, E.; RosBn, A.; Kaaemo, B. J. Phys. Chem. 1979,21, 233.

1998,97,12828.

702 Langmuir, VoZ. 10, No. 3, 1994

0 - 0 - 0

0 - 0

0 - - 0 0 - -

o o constant pumping speed - - I I

Fridell et aZ.

1200 1 I I o o 0 0

0 0 0

A

d 800

6 400

0 0 5 0 100 150

pressure (mTorr)

I c o n d n t mais flow ' ' I 16000

.g 8000 I 0

0

0 0

0 0

rPo0

0 0

0

0

pressure (mTorr)

Figure 4. (a, top) OH fluorescence signal as a function of pressure at 1200 K and K = 0.5 ( K = po,/<Po, $: PH )) taken with approximately constant pumping speed using tge Roots pump. (b, bottom) Same as (a) but taken with constant mass flow.

As the pressure is increased a number of corrections are called for. The first ones are connected with the rotational population redistribution and electronic quenching of laser-excited OH. These are, as discussed above, well understood and can be corrected for. A third potential factor is reactive quenching of OH by reactive collisions with H2 and 0 2 , which are known to be important in HJ02 flames (e.g., OH + Ho- H2O + H). However, inspection of the cross sections for these processes66 shows that they have negligible effects at the present pressures and sample to detection volume distances.

There is yet another factor, however, which becomes increas- ingly troublesome at increasing pressure, when results at different pressures are to be compared, namely, the influence of pumping speed and mass flow. The problem is illustrated experimentally by two different measurements (Figure 4) of the LIF intensity, IOH, as a function of pressure. The two measurements differ in the way the pressure variation was obtained. In the measurement shown in Figure 4a, the pumping speed, U, Le., the volume flow, was kept constant and the pressure was increased by increasing the mass flow, ip (0 = Up). In the measurement of Figure 4b the pressure was instead varied by keeping the mass flow constant, and successively reducing the pumping speed, to increase the pressure. The constant pumping speed measurement produces an IOH vs p curve with a maximum, which is even qualitatively at odds with the monotonic rise of +OH expected from model calculations (see section 3.2 below). The measurement with constant mass flow, in contrast, seems to reproduce the expected behavior of +OH, at least qualitatively, and as we shall see below even semiquantitatively.

The different behaviors in Figure 4 can be understood as follows: The production of OH at the catalyst surface is only dependent on the gas impingement rates (H2O and 0 2 in the present case), which are proportional to the pressure, but not dependent on ip or U. The same production rate at the surface is thus expected in the two experiments. (Note that the HzO + 1/202. e 20H reaction is an equilibrium reaction, not plagued by gradient problems as the forward reaction might be.39943) The different results, with constant U and ip, respectively, derive from differences in OH transport into and out of the detection volume sampled by the laser beam, a few millimeters away from the surface. As the pressure is increased the transport of OH from the surface to the detection volume becomes more and more hindered by the increasing gas density. Eventually the transport will be purely diffusive and gradients will be established in the OH concentration, from the surface and outward. The diffusive

Figure 5. Reaction steps considered in the kinetic modeling according to (3.1)-(3.7) denoted by 1-7.

flow into the detection volume will slow approximately inversely proportional to the pressure, and the gradients will then be successively stronger. The flow out of the detection volume is a combination of diffusive transport and the shear flow due to the net flow through the system. To first approximation the latter varies as the pumping speed. If now the experiment is performed with constant U, the time constant for exchange of the detection volume is kept constant while the diffusive flow into it slows roughly as p-l, and IOH then underestimates +OH more and more as the pressure becomes higher. The latter effect becomes significant when the molecular mean free path becomes small compared to z, which is at a pressure of about 100 mTorr, i.e., exactly where the experimental IOH, with constant pumping speed, has its maximum.

In the constant mass flow experiment the pumping speed is reduced as p-l as the pressure is increased in order to keep ip constant. Therefore, the reduced diffusive flow into the detection volume is to first approximation compensated by the reduced pumping speed, and IOH vs p therefore more correctly reflects the variation of +OH with pressure in this case. A detailed description of the OH concentration outside the surface and how it is influenced by pumping speed/mass flow is given e1sewhere.a

3. Kinetic Model 3.1. General Kinetics. The reactions taking place

when one or several water, oxygen, and hydrogen molecules impinge onto a platinum surface are illustrated in Figure 5. The total system is assumed to comprise the following steps (special cases will be discussed later):

adsorption/desorption

H,Og e H,Oa

0,B e 20"

H,B s 2H"

(3.1)

(3.2)

(3.3)

surface reactions

OH" s 0" + Ha (3.4)

H,O" e OH" + Ha (3.5)

H,O" + 0" s 20H" (3.6)

desorption of OH OH" - OHg (3.7)

(In the scheme above we have not included some steps that may be required to model reactions involving a hydrogen and deuterium mixture.) The superscript g denotes the gas phase and a the adsorbed phase. Two different reaction steps for water decomposition/forma- tion, (3.5) and (3.6), are considered. Desorption of atomic hydrogen or oxygen is not considered because of the high activation energies for these events. The reaction scheme above follows a recently developed model for the water formation reaction on Pt.l6 It results in the following rate

Desorption from F't in H@/O2 and HzOIH~ Mixtures Langmuir, Vol. 10, No. 3, 1994 703

equations for the surface coverages:

deOH/dt = pwlqoe%o(l - 8) - K"%oeOdH - 2Kff50eOH2 + *qoeqoeo - PwOdoH(1 - e) +

@OHeHeO - K d , H f l O H (3.11)

where Bi is the surface coverage of species i, 0 is the total surface coverage (e = eo + B H p + eOH + OH), Kji are the rate constants for reaction steps 3.1-3.7 above. Super- scripts f, d, and dec denote formation, desorption, and decomposition, respectively. f1, d e s , f2, and dec2 denote formation and decomposition of adsorbed molecules via reaction steps 3.5 and 3.6, respectively. fi(0) are the coverage dependencies of the sticking coefficients.

is formally assumed to be nonactivated. The rate constants are given by the product of the surface impingement rate, vi, and the initial sticking probabilities Si(0):

Adsorption of H2OP4 O Z , ~ and

p i Visi(0) = piASi(O)/(2*mik~T,,)~'~ (3.12)

where pi is the partial pressure of species i, A is the site area, mi is the molecular mass, k g is the Boltzmann constant, and T,, is the gas temperature. The coverage dependence of the sticking coefficient for water is neg- ligible, Le., fH&(B) = 1, on most metals,M including Pt,3 with a sticking probability of S H ~ = 0.7.8 A coverage dependence for the oxygen sticking coefficient of fo = (1 - 8)2 has been reported by several researcher&47-'~ and has successfully been used to predict the kinetics of water formation on this surface.16 Sticking coefficients of S@(O) = 0.023 and SH,(O) = 0.046 at zero coverage have been determined experimentally for the polycrystalline surface used in this study.g0 The sticking coefficient for hydrogen has been reported to depend on hydrogen coveragem but to be fairly independent of, or even to increase slightly with, oxygen ~ o v e r a g e . ~ ~ * ~ ~ @ In the experiments reported here the hydrogen coverage is very low (d0-4) due to the high temperature. We therefore use f H 2 = 1.

The rate constants for the surface reaction steps are assumed to be given by the Arrhenius expression:

~j~ = Ji exp(-@@) (3.13)

(44) Thiel, P. A.; Madey, T. E. Surf. Sci. Rep. 1987, 7,211. (46) Luntz, A. C.; W W , M. D.; Bethune, D. S. J. Chem. Phye. 1988,

(46) Norton, P. R.; Davies, J. A.; Jackman, T. E. Surf. Sci. 1982,121,

(47) Campbell, C. T.; Ertl, 0.; Kuipere, H.; Segner, J. Surf. Sci. 1981,

(48) Winkler, A.; Guo, X.; Siddiqui, H. R.; Hagans, P. L.; Yates, J. T.

(49) Derry, 0 . N.; Row, P. N. J. Chem. Phye. 1986,82,2772. (60) Hugenechmidt, M. B.; Verheij, L. K.; Freitag, M. K.; Poelsema,

89, 4381.

loa. 107, 220.

Surf. Sei. 1988, 201, 419.

B.; Coma, G. Surf. Sei. Lett. 1991,259, L763.

Table 1. Parameter Valuer Uwd in ths Kinetic Model Calculationr

quantity value ref quantity value ref 0.023 (1 - e)2 so 0.048 30 0.7 3 4Aa 14 2.6 eV 63 4.48 eV 64 6.12 eV 64 4.39 eV 64

where Vii is the preexponential factor, Eii is the activation energy, and @ = (kg7')-1.

In order to compare the model with experiments, we need to evaluate how the OH desorption rate and the apparent activation energy for OH desorption, POH, vary with the different experimental parameters, i.e., with total pressure @),gas mixture (K), and surface temperature (T). The apparent desorption energy is obtained from the temperature dependence of the OH LIF signal; i.e., it is determined from the slope of Arrhenius plots of the desorption rate of OH. The OH desorption rate is given bY

(3.14)

The apparent activation energy for OH desorption is defined as

(3.15)

This quantity is not equal to the true activation energy for OH desorption, E d o ~ , because a change in surface temperature changes all reaction rates on the surface and thus also the OH coverage.12J6J6*88 This also produces different DOH for different parameter values. By solving the steady-state equations (all dei/dt = 0 in (3.8)-(3.11)) for the respective coverages, +OH and DOH can be calculated. This is done numerically, using the same procedure as by Hellsing et aL16 in the analysia of the hydrogen-oxygen reaction. Alternatively, for some special cases, e.g., H2O/O2 and H20/H2, analytical expressions can be derived by making reasonable approximations and thus neglecting a few reaction steps (see below). For the calculations we have used the parameter values listed in Table 1. The enthalpy diagram obtained from these parameter values is shown in Figure 6.

The kinetic model is based on the so-called mean field approximation. As applied here it neglects possible adsorbate rearrangements due to adsorbate-adsorbate interactions and any memory of the adsorption step which could give rise to nonrandom distribution of adsorbates, such as island formation, statisticalvariations in adsorbate density, etc. Of course, it also neglects nonhomogeneities of the surface such as different exposed surface planes, and defects of all kinds (see also discussion about these points in refs 14 and 30). 3.2. HzO/Oz. The reaction H2O + V 2 0 2 ~t 20H ~t 2H

+ 2 0 is essentially in equilibrium, with asmall perturbation due to OH and H2 desorption. Because of the high activation energy for desorption of chemisorbed oxygen (2.1-2.4 eV) and the small activationenergy for desorption of H2O (0.4-0.7 eV), the surface coverage will in this case be dominated by oxygen. Thus, it is quite reasonable to set 0 = Bo.

The overall reaction for this system is thus equal to the rates of adsorption and desorption of oxygen and water except for the minute OH and H2 desorption. The latter

704 Langmuir, Vol. 10, No. 3, 1994 Fridell et al.

E" f

j 20H'

Figure 6. Enthalpy diagram for the water formation and decomposition reactions on Pt. The activation energies for the reactions are chosen according to Table 1.

can be neglected when discussing the overall reaction and in calculations of surface coverages. We may thus in the first step neglect reactions 3.3 and 3.7 to obtain OoH from which &H is obtained. Using the coverage dependencies for adsorption of water and oxygen mentioned above, these approximations lead to the following simple expressions for the HzO and 0 coverages:

e%o = r%dp%o (3.16)

(3.17)

To find an analytical expression for BOH, we must neglect one of the two water formation steps, Le., (3.6) or (3.6). By doing so, we get the following expressions for the OH and H coverages:

8, = peoHeoH(l - ~ / p o d o (3.19)

Exactly the same expressions for BOH and hi, respectively, are obtained considering either step 3.6 or step 3.6 (this is due to the equilibrium nature of the reaction). The calculated surface coverages of 0, HaO, OH, and H as a function of K are shown in Figure 7 (note the logarithmic scale). Obviously, only oxygen contributes significantly to the coverage.

Equations 3.14 and 3.18 give an expression for the OH desorption rate:

rdOH = veoH exp(-POH@) =

OH)@) (3.20)

ADOH= DOH- (1/2)(&,+ D%),whereDiisthedissociation energy of species i. The values of the preexponential factors can be estimatedusing the transition-state theory.16

0

-2 1200 K 1

-8

-10 I I 0 0.2 0.4 0.6 0.8 1

K

Figure 7. Calculated surface coverage8 of 0, H, OH, and HnO ma function of K ( K = p d @ q + p ~ f i ) ) at 1200 K and 50 mTorr, usmg the parameter values luted in Table 1.

Here we have used the same preexponential factor for all surface reactions.16 Further, we have used the following relation which is derived from the enthalpy diagram (Figure 6) of the reaction:

By using (3.16) and (3.20), an expression for the apparent desorption energy is obtained

The calculated E.OH from (3.22) reproduces the experi- mental resulte with regard to both the absolute value of E'OH and the trend that it increases with increasing K.~'

The parameters that are imperfectly known in (3.20) and (3.22) are v, Ed%, S%(O), and S~fl(0). To get a reasonably good agreement with the experimente, these parameters must lie in the following intervals: 2 X 1012 s-l< v < 10" s-l, 2.2 eV < Ed% < 2.6 eV, and 0.003 < S%(O) < 0.1. The sticking coefficient for water, SH&(O), will influence only the absolute OH desorption rate, which is difficult to determine with sufficient accuracy from LIF experimente.1237 The activation energies for the surface reaction steps will not influence the OH desorption yield as long as they are low. (This is of course due to the fact that the system is at equilibrium, and thus not influenced by these activation energies. If, for example, water decomposition steps 3.6 and 3.6 were both so slow due to high activation energies that their rates were comparable to OH desorption, equilibrium would not be established. However, this is beyond reason since it would require that at least one of the activation energies for reaction steps 3.4-3.6 must be 2 2 eV.)

3.3. HaO/Ha. An analytical expression for the OH desorption rate can be derived also for OH desorption in H20/H2 mixtures. In order to do this, we omit water dissociation step 3.6 (the motivation for this is given in section 4.1) and OH desorption step 3.7 in the kinetic calculations. We then obtain the following expression for OH coverage:

By using (3.14) and (3.21)

Desorption from Pt in H~OIOZ and HzOIHZ Mixtures Langmuir, Vol. 10, No. 3, 1994 706

The calculated $OH as a function of the HzO/H2 gas mixture is presented below in connection with the ex- perimental data.

4. Results and Discussion 4.1. HzO/Oz. Experimental 101.1 vs K results at four

different total pressures are shown in Figure 8. In these measurements the mass flow through the chamber was kept constant. IOH has a maximum at around 30 75 oxygen (IC-) which corresponds to the point where the product between 00 and 0Hfl has ita maximum (see Figure 7). The OH desorption rate increases with increasing pressure with a slight shift of the maximum to lower K values.

The maximum OH intensity (from Figure 8) vs p is shown in Figure 9 (squares) together with the calculated result from (3.20). The increase in1oH withp is fairly well reproduced by the kinetic model, but the measured p dependence is closer to linear than the model prediction. (A somewhat weaker coverage dependence for 0 2 sticking would work in the right direction, but is not reported in the literature.) The slight decrease in K- with pressure (shown in Figure 10) is well reproduced by (3.20). This change in position of K- is caused by the coverage dependence of the oxygen sticking coefficient: The increase in pressure is accompanied by an increased surface coverage due to the accompanying increase in the ad- sorption rates of water and oxygen. The adsorption rate of water increases linearly with the water pressure because of the zero-order coverage dependence in the water sticking coefficient. For oxygen, on the other hand, the increase in the adsorption rate with pressure is less than linear because of the second-order dependence on coverage in the sticking coefficient. The optimal product of 00 and dHfl (which is proportional to 0OHZ) therefore occurs at lower K values when the pressure increases.

Since the desorption of OH from the surface in the HzO + 02 reaction is just a minute desorption from an equilibrium coverage of OH, its absolute magnitude will be insensitive to the position of the enthalpy levels of the various intermediate species in the reaction enthalpy diagram. Thisiseasilyseenin (3.20) wheretheonlysurface activation energy that influences $OH is Edo,, which comes into play only because of the coverage dependence of the oxygen sticking coefficient.

In order to obtain the analytical expression (3.20) for the OH desorption rate, we had, as mentioned already, to omit one of the two reaction steps for water formation (reaction step 3.6 or 3.6). In both cases exactly the same expression (3.20) is obtained for the OH desorption rate (and fOrEaOH (3.22)) in H&/Oz mixtures. It thus appears that a study of H2O decomposition to form desorbing OH does not allow a discrimination between the two routes 3.5 and 3.6. However, if we combine the present results with the previous analysis of the Hz oxidation reaction, we can indeed make a discrimination. Hellsing et a2.16 addressed the question of the two routes 3.6 and 3.6 for water formation in their kinetic model study. They found, on the basis of comparison with the e ~ p e r i m e n t s s ~ ~ ~ for the forward reaction HZ + ' / 2 0 2 - HzO, that the route H + OH - Hz0 was fully compatible with most of the data, while the route OH + OH - HzO + 0 could only account for the data if acoverage dependence (on oxygen coverage)

a c. E S 0 .- P 8 a U I 0

0 100 I + + +

++ I++++++ 50

0

0

+ c I

0,O 0,2 0 , 4 0 , 6 0 , 0 1,0

K Figure 8. OH desorption rate as a function of K ( K = p (p + PH )) at 1200 K and four different total preseures (2&, h, an8150 mTorr). The different total pressures are indicated in the figure (mTorr).

h

5 cd

E v

a "

E 0 .- P 8 a U I

0 100

pressure (mTorr)

Figure 9. Maximum OH desorption rate (with respect to K ( K

= p d @ & + ~ H I O ) ) versus total pressure measured with LIF (squares) and the corresponding result of (3.20).

I I

1 1200 K

a E

Y T

0,o 1 0 100 200

pressure (mTorr)

Figure 10. Relative 01 concentration at the maximum in OH desorption, K- ( K = p (pa + p ~ f i ) ) , as a function of total

was introduced for the OH chemisorption energy. They also concluded that the coverage dependence, required to obtain agreement with experiments, was approximately EdoH(8) = Edo~(0) - Be, with a quite large value of B = 0.7 eV. If we use this coverage dependence for the OH desorption activation energy in the numerical calculation, it is impossible to fit the present data. The calculated E*OH for the H20 + 02 reaction with B = 0.7 eV becomes negatiue, i.e., @OH decreases with increasing surface temperature, in obvious contradiction with experiments. The value Of EnoH derived from experiments is 1.6-1.9 eV depending on K, in excellent agreement with (3.20).12 A reasonable agreement between calculations and experi- ments actually requires B < 0.05 eV, in conflict with the corresponding analysis for the forward reaction, if one assumes route 3.6 to be dominant.I6 Together with the

pressure. The solid line orl s o w the corresponding result of (3.20).

706 Langmuir, Vol. 10, No. 3, 1994

results of ref 16, the present results therefore strongly favor OH + H F! H2O as the most important paths for water formation and decomposition, respectively, under the present experimental conditions. As noted in ref 16 this result does not rule out that route 3.6 could be the more important a t low temperatures. It should also be noted that it is the combination of the analysis of the forward reaction (compatible with both routes if we employ a coverage-dependent and the H2O decomposition reaction (not compatible with a coverage-dependent EdoH) that provides the argument against the OH + OH route. Regarded independently, both works could be made compatible with the OH + OH route.

Also with pure water, Le., without oxygen, there is a significant OH desorption observed in the experiment, demonstrating that a unimolecular decomposition path (3.5) is indeed accessible a t high temperatures in line with the conclusion just arrived at. This is in contrast to the frequently proposed bimolecular reaction 3.6,6-9J'1 for H2O decomposition. The experimental verification of the unimolecular path in pure H2O requires great caution in the experiments, with respect to oxygen traces in the H20 gas since even minute additions of 0 2 rapidly enhance the OH signal as is illustrated by Figure 8. This extreme sensitivity to oxygen is due to the high desorption energy for 0 2 , which causes a high oxygen coverage even at low oxygen partial pressures. I t is tempting to take this large sensitivity of the OH signal to the oxygen coverage as evidence that route 3.6 is more important than route 3.5. This is, however, not a correct conclusion. The numerical simulations show that addition of oxygen to the surface greatly increases the OH coverage even when only .the unimolecular decomposition route 3.5 is open, because oxygen atoms will then react through step 3.4 with the H atoms created through step 3.5, which in turn increases the OH coverage. Additional measurements on the pure H2O decomposition on Pt(lll), performed under ultrahigh vacuum conditions, will be presented in a forthcoming publication.58 4.2. H20/H2 (D2). Some additionalresults, elucidating

the reaction mechanism, were obtained with H2O mixed with H2 or D2 and H2O/O2 + H2 (D2) mixtures. Figure 11 shows IOH vs y (y = P H ~ / @ H ~ + PH~O)) a t 50 mTorr total pressure and a surface temperature of 1300 K. The signal has its maximum at y = 0, corresponding to the unimo- lecular decomposition of H2O just discussed above, and then rapidly decreases to a very low value at y 1 0.05 due to the increasing consumption of OH via the step OH + H - H2O. Also shown in Figure 11 are the corresponding calculated results from (3.24). A full numerical calculation differs only slightly from the analytical approximation at y = 0, where the analytically calculated OH desorption rate is somewhat larger. The calculated result is about 20% too low compared with the experiment a t y = 0 while the fit at y # 0 is quite good (the fit may alternatively be good at y = 0 but then poorer around y = 0.1). Adding a small background of oxygen (in the calculation) does not improve the overall fit as this will give an OH signal even

Fridell et al.

(51) Nyberg, C.; TengstAI, C. G. J. Chem. Phys. 1984,80,3463. (52) Parker, D. H.; Bartram, M. E.; Koel, B. E. Surf. Sci. 1989,217,

(53) Atkine, P. W. Physical Chemistry, 3rd ed.; Oxford University 489.

Preas: Oxford, 1987.

Nostrand Reinhold New York, 1979.

Springer: Berlin, 1984; p 204.

(54) Huber, K. P.; Herzberg, G. Constants of Diatomic Molecules; van

(55) Warnatz, J. In Combustion Chemistry; Gardiner, W. C., Ed.;

(58) Dieke, G. H.: Croaswhite, H. M. J. Quant. Spectrosc. Radiat. Transfer 1962,2, 97.

1992,96,5922. (57) Williams, W. R.; Marks, C. M.; Schmidt, L. D. J. Phys. Chem.

n n7 " . " I , I I I 1 I

1300 K g 0.02 50 mTorr

8 0.01

0

0 0.2 0.4 0.6 0.8 1 Y Figure 11. OH desorption rate 88 a function of y (y = P H J @ H * + ~ ~ 1 0 ) at 1300 K and 50 mTorr together with the corresponding analytical model calculation.

I O I

0 ... I I . I

0,s 1 , o 1,s (mTorr)

PD2

Figure 12. OD and OH desorption rates aa a function of Dz pressure with P H ~ O = 50 mTorr and T = 1300 K.

at y - 1 which is not seen in the experiments. The latter is actually a good control that the oxygen background level is low in the experiment.

If D2 rather than H2 is added at a certain water concentration, both OD and OH will be produced. This is illustrated in Figure 12 which shows the OH and OD signals when D2 is added to 50 mTorr of H2O. The OH signal drops rapidly to zero with increasing D2 pressure and is close to zero a t 0.25 mTorr of D2. The OD signal increases from zero to a maximum a t a low D2 pressure (<0.2 mTorr) and then decreases. OD may be formed from H20 and D2 by water dissociation all the way to atomic oxygen, which then reacts with D* to produce OD. We consider the additional pathways D + OH * H + OD or OH + D e HDO F! DO + H for OD formation (and thus OH suppression) less likely, not because they are mech- anistically improbable, but because the product 6OH6D is much smaller than 6OH. 4.3. H20/02 + H2 (D2). The effect of adding H2 to an

HzO/O~mixture is illustrated in Figures 13 and 14 together with the corresponding calculated curves. The desorbing OH'S are in this case formed both from dissociating water and from the 0 2 + H2 reaction. The discrepancy between experiments and calculations at high pressure is due to the pumping speed effect discussed in section 2.4, since the hydrogen pressure was increased by increasing the hydrogen mass flow. A measurement with constant mass flow, a, would give a higher OH signal a t high H2 pressures.

The effect of adding D2 to a mixture of 50 mTorr of H2O and different poa (0,3,10, and 30 mTorr) is illustrated in Figure 15. The OD desorption rate as a function of P D ~ is shown in Figure 15a. It maximizes a t higher Dz pressures as the 0 2 pressure increases. The observed behavior can be understood as follows: At zero oxygen pressure the result is the same as in Figure 12. Addition of successively

Desorption from Pt in HzOIOz and HzOIHz Mixtures

5 0 0 ) I 1 1 1 1 1 1 1

X x x x .w 0 f ~ ",.

Langmuir, Vol. 10, No. 3, 1994 707 2000 I I

OD water pressure= 50 mTorr - 12d0 K oxygen pressure = 10 mTorr water pressure = 10 mTorr

0 I I I I" l o I O 1 I 0 5 10 1 5 20 2 5 30 35 40

'H2 (mTorr)

Figure 13. OH desorption rate as a function of Hz pressure with p a = p ~ p = 10 mTorr at 1200 K (circles). The solid line represents full numerical model calculations using the parameter values listed in Table 1.

800 I , , , I I I

Gi I i ; 700 oxygen pressure = 25 mTorr 2 600 waler pressure 25 mTorr

0 L 0 5 1 0 1 5 20 2 5 3 0

'H2 (mTorr)

Figure 14. Same as Figure 13 but with por = p ~ p = 25 mTorr.

more oxygen gives a much higher peak intensity of OD, primarily because the added D2 and 0 2 react according to the forward reaction route, as described in refs 16 and 30, with water adsorption/decomposition as a small pertur- bation. The OH desorption rate for this experiment is shown in Figure 15b. The experimental points a t p ~ * = 0 correspond to points in Figure 8 at ( K , pht) = (0, 50), (0.06, 53), (0.17,60), and (0.38,90) for pol = 0,3,10, and 30 mTorr, respectively. As P D ~ is increased $ 0 ~ falls rapidly, since the added Da changes the surface equilib- rium. Specifically Oa on the surface, produced through steps 3.2 and to a minor extent (except at poZ = 0) through (3.5) + (3.4), is consumed by Da to produce OD, as soon asp^^ # 0. Further, OHa will react with Da to form HDO. This explains the very rapid decay of OH with increasing p ~ ~ . With increasing pol, the OH signal decreases more slowly with increasing pDZ because there is relatively more 0" available to react with adsorbed Ha, always produced through the decomposition reaction 3.5.

4.4. Comment on the Enthalpy Diagram and Ac- tivation Energies. The enthalpy diagram of Figure 616 and the activation energies used to simulate the kinetics constitute a consistent set of parameters capable of reproducing (i) H2O and OH production/desorption in the forward reaction H2 + l/202 - H2Ol6*m and (ii) OH production/desorption in H2O decomposition reactions (in mixtures of H20, HzO/Oz, H2O/H2, and HzO/Oz + H2 (this work and ref 12)). Since all these studies have been performed under steady-state conditions, there is a certain nonuniqueness in the choice of the absolute values of activation energies and enthalpies. For example, the water decomposition reaction (in H2O and H2O/O2) is, due to its

(58) Fridell, E.; Elg, A.-P.; h B n , A.; Kasemo, B. To be published.

n 0

t I X X

0 L x x x I . . A I x A .

0 10 2 0 3 0

p (mTorr) DZ

OH water pressure = 50 mTorr

v

pressure (mTorr)

a c e 600 C 0 .-

equilibrium nature, only sensitive to the enthalpy values, but not to the activation energies, as long as the OH desorption is a negligible perturbation of the equilibrium. The forward reaction is a nonequilibrium situation and is thus sensitive to both the activation barriers and the enthalpy values,ls but primarily to differences between activation energies rather than to their absolute values.ls@ However, since some of the values are fairly well known (as discussed in ref 161, e.g., the adsorption energies of hydrogen, oxygen, and water, the nonuniqueness problem is mainly concerned with (i) the enthalpy value for OH, i.e., the true desorptionenergy of OH, and (ii) theactivation barriers for OH and H2O formation and/or decomposition, re~pec t ive ly .~~ A more detailed analysis of this non- uniqueness problem is presented elsewhere.s8 It shows that the critical quantity in the numerical simulation of the experimentally measured OH desorption from a Pt- (111) crystals'3 is the difference between the activation energy for OH desorption and the activation energy for H2O formation via route 3.5, Le., E d o ~ - B H ~ . As long as this difference is kept constant (at approximately 1.95 eV), the individual values can vary between 1.9 and 2.6 eV for E ~ O H and between 0 and 0.7 eV for Efi~p. This parameter range also reproduces the results for the backward reaction. This means that the enthalpy diagram of Figure 6 could be modified accordingly. The main reason for our choice of parameter values at the lower end of the range quoted above is the experimental upper limits set by the experiments of Germer and HoB and Fisher et a1.,31 respectively (see discussion in ref 161, for the activation energy of the H20 formation step. If we reject

708 Langmuir, Vol. 10, No. 3, 1994

these upper limit values and rather choose the values reported by Anton and CadoganF1VB we could accept E d o ~ and Eil~p values at the upper end of the range. This is one major remaining obscurity to be resolved for the water formation reaction on Pt.

5. Summary and Conclusions Summarizing these results, we find that addition of

oxygen effectively promotes the decomposition of water on Pt at high temperatures. This is in line with the low- temperature EELS results for coadsorbed HzO and 02 on both Pt7 and Pd.61 At high temperature the primary water decomposition path is the unimolecular reaction HzO - OH + H rather than the bimolecular reaction HzO + 0 - 20H. This may at first sight seem surprising, when oxygen has such a strong promoting effect. However, the explanation is that the role of oxygen is to promote the hydroxyl coverage on the surface indirectly rather than via the direct reaction 3.6. The action of 0 atoms on the surface is thus to trap hydrogen atoms on the surface as OH. (The hydrogen atoms are created through the steps HzO - OH + H and OH - 0 + H.) Thus, the presence of 0 always increases the OH coverage, and simultaneously reduces the forward reaction OH + H - HzO by suppressing the H coverage. Another and equivalent way of looking at the effect of 0 atoms on the surface is that it moves the surface equilibrium from 0 + H to OH by increasing the rate of the OH formation step K f o H 6 0 B H (3.4) and suppressing the rate of OH consumption via

Fridell et al.

KfiH&60H. The influence of (3.4) is effective for both reaction routes 3.5 and 3.6. The conclusion that unimo- lecular water decomposition is the dominant mechanism was arrived at by comparing results from the forward reaction Hz + l / z O ~ - HzO with the present water decomposition results. The conclusion is supported by the observation of OH production in pure HzO. To explain the latter in terms of route 3.6 requires16 that the OH desorption energy have a strong dependence on oxygen coverage, which is contradicted by the present data analysis, using the same model as for the forward reaction. Thus, route 3.5 (OH + H 3 HzO) is the dominant water formation and decomposition route a t high temperature.

The addition of hydrogen at a constant p ~ f i causes a decrease in OH desorption caused by an increased con- sumption of 0 and OH to form water, so that the equilibrium coverage moves from OH and H toward HzO and H. When Dz is used instead of Hz, both OD and OH desorb from the surface, supporting the assumption that the decomposition reaction of HzO goes all the way to 0 + 2H on the surface.

Acknowledgment. We gratefully acknowledge finan- cial support from the Swedish Natural Science Research Council (NFR, Contract No. E-EG 2560-3301, from the National Energy Administration (STEV, Contract No. 276 330-11, and from the Swedish Research Council for Engineering Sciences (TF'R, Contract No. 92-538).