Energy partitioning in the reaction 2H2+ O2rarr; 2H2O on Pd(111)

摘要

Faraday Discuss., 1993,96,265-275 Energy Partitioning in the Reaction 2H2 + O2+2H20 on Pd(ll1) Arne de Meijere, Kurt W. Kolasinski and Eckart Hasselbrink Fritz-Haber-lnstitut der Max-Planck-Gesellschaft, Paradayweg 4-6, 14195 Berlin, Germany (2 + 1)-Photon resonance-enhanced multiphoton ionization (REMPI) has been used to detect desorbed H,O molecules in a state-specific manner. We have studied the reaction on Pd(ll1) at temperatures between 500 and 700 K in the regime of low oxygen coverages ( -c0.05 ML) resulting from adsorp- tion of ambient 0,. D, is dosed using a supersonic molecular beam with pulse fluxes resulting in stoichiometric coverages of D and 0 atoms. We find that D,O molecules are desorbed with a translational energy of ca. 300 K with weak, if any, dependence on the surface temperature.The angular dis- tribution is nearly cosine in form. The spectroscopic data indicate that the D,O is desorbed with an average energy in the rotational degree of freedom close to the surface temperature. Experiments probing vibrationally excited states yielded negative results setting an upper limit for their population. These data show that most of the excess energy of the final OD + D reac-tion step of ca. 100 kJ mol-' is not channelled into any of the degrees of freedom of the desorbed molecules. The data are interpreted such that the product water molecules reside on the surface for a time sufficient to dissi- pate most of the energy. The understanding of gas/surface interactions has greatly profited from the application of optical spectroscopy to detect the final state distributions of molecules desorbed from well characterized single crystal surfaces.' However, to date only limited data are avail- able for the products of elementary catalytic reactions.For over a decade the experi- ments by Fenn and co-workers, on CO oxidation utilizing detection of infrared emission, have been a benchmark in this area. Further progress has been hampered, but not entirely hindered, by the lack of suitable detection schemes for, from the surface chemist's point of view, relevant molecules. Such experiments promise insights into the structure of the activated complex and the dynamics of product formation, provided that the product does not equilibrate with the surface prior to desorption.Since CO and H, oxidation are both strongly exothermic reactions, such equilibration is not expected, or, if it occurred, would constitute an interesting result in its own right. For CO oxida- tion it has been shown that CO, is desorbed with non-thermal internal state distribu- tions,, with translational energies far in excess of what is expected for equilibration with the s~rface,~ and that the product angular distribution contains information about the reaction site.5 In studies which date back to the work of Michael Faraday, the catalytic oxidation of hydrogen has, together with CO oxidation, been of extensive Conse-quently, the water formation reaction has been studied by almost every available surface-science technique.Most work has been performed on Pt( 11l), as the most well characterized single crystal plane. However, a significant literature on Pd( 1 11) also exists and we feel that differences are more of a quantitative nature. The reaction has been 265 Energy Partitioning in 2H, + 0, -+ 2H,O studied by temperature-programmed desorption (TPD)," electron energy loss spectros- copy,' ,-14 titration of adsorbed oxygen,15+'6 secondary ion mass spectrometry,' 7918 and laser-induced thermal desorption. " Molecular beam work has been performed by several There is also a large body of literature looking at the reaction at high temperatures where the reaction intermediate OH can be desorbed and be detected by laser induced fluorescence (LIF).24-26 The hydr~gen,,~-~~ oxygen,30 and H2012*31 adsorption systems have also been studied separately in great detail.The major pathway for water formation is believed to be the sequential addition of hydrogen to adsorbed oxygen and hydroxy species. Extensive kinetic modelling has been undertaken by several a~thors.~~-~~v~~Whereas the initial work by Kasemo's group assumed the 0 + H + OH reaction to be rate limiting, more recent work locates the dominating barrier in the final step, H + OH -+H20, with an activation energy of 63 kJ mol-I. However, some other pathways might be significant and have been considered. In particular, the hydroxy disproportionation reaction might be the preferred reaction pathway at low tem-peratures and lean hydrogen to oxygen coverages.In this report we present experimental data on H,0/D20 product internal state distributions obtained by state-specific detection as well as translational energy distribu- tions. From these data we hope to infer novel information on the reaction transition state and energy partitioning. Experimental The experiments were conducted in an ion/turbo pumped UHV apparatus which has been described in detail elsewhere.33 The apparatus was equipped with a low energy electron diffraction (LEED)/Auger unit, an Ar+ ion sputter gun and a quadrupole mass spectrometer (QMS). The Pd( 1 11)sample was cleaned using standard procedures of Ar + bombardment and oxygen treatment at 800 K sample temperature. H, (Messer-Griesheim, 99.998%) and D, (99.7%) were dosed using a pulsed molecu- lar beam source (General Valve, Series 9) with 250 pm orifice.This device allowed minimal pulse length of 150 ps. Stagnation pressures between 1 and 6 bar and a repeti- tion rate of 5 Hz were used. The beam source chamber is separated from the UHV chamber by two stages of differential pumping. Absolute calibration of the beam flux was not feasible, but it can be estimated that a single pulse corresponds to a dose of 0.1 L. 0, (Messer-Griesheim, 99.99%) was introduced into the UHV system using a leak valve. 0, background pressures of lo-* to lop5mbar were used. H20 desorption yields and time-of-arrival (ToA) spectra were recorded using the QMS for detection. The QMS could be rotated around the sample and the distance between ionization volume and sample could be varied between 25 and 75 mm.In addition the QMS signal could be input into a boxcar averager (SRS 250) to gate signal detection with the pulse repetition of the H, beam. State-resolved detection of H,O/D,O was achieved by (2 + 1)-photon REMPI near 248 nm. The output of an excimer pumped, tunable dye laser (Lambda Physik, EMG 201E/F1 2002), operated on Coumarin 307, was frequency doubled in a B-barium borate (BBO) crystal. Typical pulse powers of 1 mJ were applied. The 248 nm light was directed into the UHV chamber, reflected to pass parallel to the crystal at a distance of 2-5 mm. The laser beam was tightly focused by a 100 mm lens. The resulting ions were acceler- ated by 100 V into a field-free flight tube of 100 mm length, in which different masses are separated, and finally detected by a stack of microchannel plates (Galileo chevron).The signal was fed into a boxcar averager with the gate set to integrate the signal of H,O or D,O ions. For each wavelength, the signal was averaged over 50 laser shots. (2 + 1)-Photon REMPI of H,O at 248 nm utilizes the 'B, +-X 'A, transition. The rotational state-resolved, sensitive detection of H20 via this excitation was first demon- strated by Meijer et aZ.34 using a tunable KrF excimer laser. The e state is heavily A. de Meijere, K. W.Kolasinski and E. Hasselbrink a0400 a0600 80800 two-photon energy/cm-' D,O c 'B, t% 'A, (2 + 1)-photon REMPI spectrum.(a) Experimental spectrum recorded in a flow cell at 300 K; (b)theoretical simulation of the same spectrum. predissociated through homogene_ous, rotation independent, and heterogeneous, rota- tion dependent, coupling to the BIAl state, as has been studied by the Bristol and Nottingham groups using (3 + 1)-photon REMPI.35936Kuge and Kleinermanns found evidence for an additional Coriolis coupling to an unbound state of 'A, symmetry.37 Although predissociation reduces the lifetime of the resonance state to ca. 5 ps, there is still enough rotational structure left to allow for rotational population analysis. Prediss- ociation results in a reduced ionization yield and a broadened linewidth. Because pre- dissociation is significantly slower for D,O than for H,O, spectroscopic detection of D,O is advantageous, and we will exclusively present spectroscopic data for this system.A typical spectrum for D,O in a gas cell at 300 K is displayed in Fig. 1. The water molecule is an asymmetric top resulting in a large manifold of rotational states. Located within the spectral range displayed in Fig. 1 are 3862 individual transitions. This con- siderable number, together with their significant width, rule out assignment of individual lines and derivation of a Boltzmann graph. Instead, it is necessary to simulate the spec- trum based on a guess of the rotational state population taking into account the effects of predissociation. Such simulation is feasible using results of sophisticated two-photon line strength factor calculations, which have been carried out by the Bristol spectroscopy The result for a 300 K sample is displayed in Fig.l(b). Comparison of the experimental and the calculated spectra indicates agreement to such a degree that appli- cation of this detection scheme should be feasible. Results The reaction yield has been recorded as a function of the 0,background pressure with a constant stagnation pressure of 3 bar for the H, beam at a sample temperature of 700 K. The reaction yield increases up to an 0, partial pressure of 5 x mbar and then decreases slowly again. Higher 0,background pressures were not advisable due to the risk of oxidizing the Pd sample at a temperature of 700 K. For variation of the H, stagnation pressure we observe a steady increase up to a pressure of 3 bar.Afterwards the signal shows a tendency to fall off again, which might be due to a reduced sticking of Energy Partitioning in 2H, + 0, +2H,O H, and to a reduced beam intensity resulting from excessive background pressures in the beam source chamber. Analysis of the scattered intensity and angular distribution indicated that about two-thirds of the H, sticks to the surface. Under these conditions it is expected that a single H, gas pulse will suffice to react completely with the 0 coverage to give H,O. This assumption is further corroborated by the observation, that if under these conditions the 0, leak valve is suddenly closed, the H,O production ceases within two to three H, pulses.Sub-surface H does not appear to play a major role in the reaction as corroborated by the waveform of the product H,O which closely tracks the incident H, pulse. At a constant H, stagnation pressure of 3 bar and an 0, background pressure of lop6mbar, the reaction yield was recorded as a function of the sample temperature, q(Fig. 2). The measurement was performed by gated integration during the peak of H,O desorption with a width of 180 ps, which is shorter than the H, pulse duration. With this technique the signal is proportional to the rate of the reaction. An exponential increase of the reaction yield is observed between 400 and 530 K, which subsequently saturates. A fit of an exponential function, I(q) = exp(EJ/kT,), to the increase between 430 and 500 K, indicates an apparent activation energy, E,, of 60 kJ mol-l.The angular distribution of the product H,O, recorded at T, = 700 K, is shown in Fig. 3. A fit to the data with a cos" 8 distribution yields n = 1.2 & 0.1. Hence, the angular distribution is only marginally more focused than expected for a thermal, non- activated desorption process. ToA spectra have been recorded in order to derive the translational energy of the product H,O. The shape of the H,O gas pulse traversing the detector is a convolution of the temporal opening profile of the nozzle, with the velocity spread in the molecular beam, the residence time on the sample, and the velocity distribution of the desorbing H,O. The dominant factor contributing to the width is the 400 ps opening time of the nozzle.The actual residence time on the surface is, in comparison, negligible, since based on known rate constants for the H, + 0 reaction and the H,O desorption, character- istic times below 0.4 ps are predicted. The flight time from the sample to the QMS is of the order of 100 ps. Because these times are shorter than the nozzle opening time, a 500 400 100 0 400 500 600 700 TJK Fig. 2 Reaction yield as a function of the surface temperature. The O2 background pressure was mbar, the H, stagnation pressure 3 bar, and the repetition rate 5 Hz. A. de Meijere, K. W.Kolasinski and E. Hasselbrink I .o 0.8 0.6 1 0.4 0.2 0.0 0 30 60 90 desorption angle/degrees Fig. 3 Angular distribution of the H20 product. The solid line represents a fit to the data yielding a COS'.~ 0 distribution. The dashed lines gives a cos 8 distribution for comparison.deconvolution of the ToA spectra is not feasible. Therefore we resorted to recording ToA spectra for two distances, 25 and 75 mm, from the ionizing volume to the sample. The mean velocity is derived from the delay between the two H,O pulses observed. For data evaluation we compared the lo%, 50% and 90% points of the peak intensity in the spectra. All of these values indicated consistent results. We find that the mean trans- lational energy, (Etran,)/2k,is in all cases smaller than that expected for thermal equi- librium with the sample (Fig. 4). Moreover, (Etran,)/2kis rather constant for sample temperatures between 430 and 710 K.If one calculates the average of these values, then one obtains 284 f27 K. However, it should be noted that this error margin only accounts for the statistical contribution. 800 -I I I I -5 600 --:-3 400 200 -5 !I! 1-0-, I I I - 270 Energy Partitioning in 2H, + 0,+2H,O A (2 + 1)-photon REMPI spectrum of desorbed D,O has been recorded at T, = 700 K [Fig. 5(a)].Simulations are based on a Boltzmann distribution of the rotational state population. Comparison of the experimental spectrum with simulations calculated in steps of 100 K indicates best agreement for an assumed rotational temperature of 800 K [Fig. 5(b)].Owing to the convolution of spectral features and the difficulty in assessing the reliability of the description of the predissociation for higher rotational levels, deter- mination of the rotational temperature to an accuracy better than 100 K is not possible. A similar spectrum has been recorded for T, = 500 K.Here the fit yields good agreement for an assumed rotational temperature of 500 K. Within the spectra, transitions belong- ing to high rotational states of the molecule are located at smaller transition energies. In this area the simulation predicts larger spectral features than we observe. This might well indicate that less population is found in the higher rotational states than expected for a thermal distribution of the chosen temperature. Extensive searches for hot bands of the D,O molecule have been carried out.The three fundamental modes of the gas-phase D,O molecule are: the symmetric stretch (v, = 2668 cm-l), the scissor mode (v, = 1178 cm-') and the asymmetric stretch (v3 = 2788 cm- 1).39 We have scanned all regions belonging to diagonal, e.g. e(Ol0)+-Z(OlO), and off-diagonal transitions, e.g. c(OO0)tx(Ol0). All these transitions are sufficiently separated and are located between 78 000 and 81 000 cm-1.40 We have not found any signals discernible from the noise. Hence, we are restricted to give an estimate for the upper limit of a possible population of excited vibrational states. The sensitivity for the higher vibrational states will unfortunately be lower than for the ground state. First, predissociation in vibrationally excited levels of the e states will be faster, because vibrational excitation affords more efficient coupling to the B state.It is expected that this effect decreases signals by a factor of two to three.34 Hence more signal will be lost at the resonant state. Secondly, we expect the Franck-Condon factors to favour diagonal transition^.^' The x-and e-state potentials are rather similar in the Franck-Condon region.41 The e state is commonly regarded as the first member of an np series of Rydberg states.42 Hence, although the larger predissociation in the resonant 80400 80600 80800 transition energy/cm- ' Fig. 5 (a) 'B, t% 'A, (2 + 1)-photon REMPI spectrum for product D,O at T, = 700 K; (b) spectrum calculated for D,O in thermal equilibrium at 800 K A.de Meijere, K. W. Kolasinski and E. Hasselbrink 27 1 Table 1 Comparison of the energy content in the various degrees of freedom for the CO oxida- tion and water formation reaction CO,/Pt(poly)" 730 1300 f75 1600 f75 1500 f75 1050 f50 CO,/Pt(p~ly)~ 880 3650 D,O/Pt( 111)' 765 305 f12 D,O/Pd( 111) 705 2740 1230 2890 800 f100 284 f100d T,, Tyl, Tvz,TyJ , Totand Transare the sample temperature, the vibrational temperature of the symmetric, bending and antisymmetric modes, the rotational and translational temperature, respectively. All values in units of K. a Ref. 2. Ref. 4. Ref. 22. This value has been obtained for H,O. state could, in principle, be avoided by choosing an off-diagonal transition, this advan- tage will be immediately lost again owing to the smaller Franck-Condon factor in this case.Ashfold et al. display a spectrum [Fig. 2(a) in ref. 351 in which for D,O the c(Ol0)-g(OO0) transition is detected with a factor of five smaller strength than the c(O00)+g(O00)transition. This factor comprises both a reduced Franck-Condon factor and an increased predissociation rate in the excited state. Hence, the penalty for the off-diagonal transition and the increased predissociation may be estimated as a factor of 2.5 each. We assume for the moment that the same factor will also be valid for other off-diagonal transitions. We expect that a signal level reduced by a factor of 10 compared to that observed for the vibrational ground state should not have escaped our notice.Taking this and the reduced detection sensitivity discussed in the last paragraph into account, we derive upper limits for the population in excited vibrational levels (Table 1). Discussion We have observed the largest signal of product H,O under conditions which are charac- terized by an H, stagnation pressure of 3 bar, an 0, background pressure of 5 x K. These conditions have also been used to mbar, and a sample temperature ~550 obtain (2 + 1)-photon REMPI spectra of product D,O. Hence, it is worthwhile to discuss these reaction conditions. We estimate that the pressure in the pulsed molecular beam corresponds to 2 x mbar. Taking into account the sticking coefficient of H, (for the clean surface) of ca. 0.8,28it is calculated that a single H, pulse leads to adsorp- tion of 0.13 ML atomic hydrogen.If one further considers that H atoms on Pd readily dissolve into the bulk,,' it might be estimated that about 0.09 ML hydrogen are avail- able for the reaction. However, the steady-state coverage during the gas pulse will be much smaller. Since the sample temperature is higher than the hydrogen desorption temperature, excess hydrogen will desorb during and between gas pulses. Thus, the hydrogen coverage at the beginning of each pulse will be zero. The sticking coefficient of 0, at 700 K has been determined to be 0.3.30Assuming that all oxygen is consumed by the reaction, it is calculated that the maximum coverage of 0, for an H, beam repetition rate of 5 Hz is smaller than 0.05 ML (5 x mbar backfill).Further increase of the 0, background pressure results in oxygen poisoning of the surface, since the adsorption sites for hydrogen become blocked. Hence, the signal is maximized if the available coverage is consumed during a single H, pulse. All these observations are in agreement with earlier studies finding that the yield is maximized if conditions are such that the surface coverages are stoichiometric.21 However, it should be borne in mind, that in an experiment with one reaction partner introduced through backfilling and the other through a pulsed beam, the ratio of surface Energy Partitioning in 2H2 + 0,-+ 2H,O coverages changes during the course of the pulse width of the gas source. In our experi- ment the ratio H : 0 will change significantly from lean conditions to unit ratios.However, during the entire period the total surface coverage is small. Since, in particu- lar, the 0 coverage is small, we expect that the reaction proceeds via sequential hydro- gen addition to oxygen. Recent kinetic st~dies~~p~~ indicate, for these conditions, that the rate-limiting barrier (63 kJ mol-') is associated with the final reaction step, 0 + OH +H,O. If one assumes, for the moment, that the final step is the dispro- portionation reaction, it would also be the rate-limiting step. Moreover the height of the barrier for this reaction has recently been determined to have a rather similar value (50.2 kJ mol-I). (In passing we note that these values also fit nicely the apparent activation energy observed in our experiments.) In contrast, we think that we can eliminate the possibility that the rate-limiting step is the formation of hydroxy species because this alternative seems in recent literature only to be discussed for large oxygen coverages.The goal of this study is to examine to what extent and into which degrees of freedom the excess energy of the final reaction step is partitioned. We observed mean translational energies of product H20 which are much smaller than expected for equilibrium of this degree of freedom with the sample temperature. Moreover, in the temperature range studied, the translational energy is only weakly, if at all, dependent on the sample temperature. This finding is consistent with the angular distribution which is only marginally more focused in the direction of the surface normal than a cos 8 distribution.Similar findings have been reported by Ceyer et aL2, for the corresponding Pt(1ll) system, in which case (D,O) a mean translational energy of 305 & 12 K has been found for a sample temperature of 765 K. These results indicate that product water is desorbed with none of the excess energy of the final reaction step partitioned into the translational degree of freedom. The (2 + 1)-photon REMPI spectra obtained for product D,O indicate a mean energy in the rotational degree of freedom of 800 & 100 K. Within the experimental accuracy this is identical to the surface temperature. Measurements for different sample temperatures indicate that the rotational temperature changes in agreement with this assumption.Although one might certainly argue that a rotational temperature is only a poor description of a spectrum, and that, particularly in this case, the high extent of blended lines and the large influence of predissociation allows only a crude reproduction of the spectrum, we can clearly rule out the possibility that the rotational state distribu- tion contains contributions with energies far in excess of the sample temperature. Con- trary to this, we feel that higher rotational states are less populated than calculations of the spectrum indicate for equilibrium with the sample temperature. However, these cal- culations rely on coefficients for the heterogeneous predissociation which have not been tested for such high rotational states.Having made these cautioning remarks, we still feel confident in concluding also that no major fraction of the excess energy from the reaction is partitioned into the rotational degrees of freedom. The search for desorbed molecules in vibrationally excited states has yielded negative results, insofar as we could not detect any signal. Hence, we can only estimate an upper limit for their population. Owing to the large quantum size of the vibrations in general, as well as the limited sensitivity for D,O detection and its vibrationally excited states in particular, the limits obtained are still rather large. We rule out that more than 20% of the desorbed D20molecules have one quantum in their bending mode and that more than 10% have an excitation of one quantum in either the symmetric or asymmetric stretch mode.Assuming Boltzmann distributions for the various degrees of freedom and summing up their different energy contents, we obtain an upper limit for the total energy parti- tioned on average into the desorbed product water molecule of 44 kJ mol-l. This value has to be compared with the excess energy in the final reaction step, where we refer to the potential-energy difference between the peak of the barrier and the level with the 273A. de Meijere, K. W. Kolasinski and E. Hasselbrink water molecule desorbed. If we follow recent work by Williams et a1.,26 this energy is 113 kJ mol-' for OH + H + H,O.If the final reaction step would be OH + OH -+ 0 + H20, as could perhaps be argued for lean H : 0 ratios, then the excess energy would be 88 kJ mol-l. This value establishes a lower limit for the energy available for par- titioning into the various degrees of freedom provided that none of the energy is dissi- pated to the solid prior to desorption. More accurately stated, the energy might well be larger because of thermal population of degrees of freedom not involved in the reaction. However, it would be necessary to correct it for possible changes of zero-point energy in the various vibrational degrees of freedom. Neglecting these effects, we can take our value, 44 kJ mol- ', as an upper limit indicating that not more than 39% or 50% of the excess energy is contained in the desorbed molecules. In comparison, a water molecule in thermal equilibrium with a 700 K sample contains 29 kJ mol-l.It is therefore obvious that the molecules must have dissipated most of the energy prior to desorption. We note that the degree of energy dissipation may, indeed, be much greater than that reported here as a result of the relatively high upper bounds on the vibrational popu- lations. That is, owing to the relative insensitivity to vibrationally excited molecules here, we have set rather generous upper bounds on the excited vibrational state popu- lation. Further studies with improved sensitivity are required in order to reduce these limits and to pin down more accurately the vibrational state population.If we now return to the discussion of the low translational energy observed then it can be reconciled in the following way. The transition state is not very much displaced with respect to the H,O-surface bond distance. Therefore, the newly formed water mol- ecule does not experience a repulsive force driving it away from the surface. This propo- sition is also consistent with the nearly cosine angular distribution. Furthermore, the energy contained in the internal degrees of freedom of the molecule does not couple to the translational degree of freedom, or, alternatively, the internal excitation dissipates faster than the molecule is desorbed. Translational energies well below the sample tem- perature have been observed in several examples provided that the sample temperature is much larger than the thermal desorption temperature of the m~lecule.~~~~~~~~ Thermal desorption of H20 from Pd(ll1) is observed at 185 K which corresponds to a binding energy of roughly 45 kJ mol-l.This value contains considerable contributions from adsorbate-adsorbate interactions arising from hydrogen bonding. Hence, the binding energy of an individual molecule is expected to be even smaller than indicated by TPD. Our studies were conducted at T, = 400-700 K. Therefore, it might not be surprising that the mean translational energy has already saturated, since we exceed the thermal desorption temperature by more than a factor of two. The finding of equilibration of the rotations with the sample temperature also fits into this picture.Strong excitation of rotational degrees of freedom of the nascent water molecule is expected to result in rapid desorption uia rotation-translation coupling. Hence, the 'thermal' excitation of rotations is consistent with the low translational energy. Considering these different remarks we are left with the following picture. The only remaining modes into which the excess energy of the final reaction step can initially be partitioned are the internal molecular vibrational modes. However, the residence time must apparently be sufficient to allow most of this energy to dissipate before the mol- ecule is actually desorbed. At the same time, coupling to the molecule-surface vibra- tional modes must be inefficient, because rapid energy transfer into the molecule-surface vibration perpendicular to the surface would efficiently cause desorption.The only other polyatomic system studied in such detail is CO oxidation. The ener- getics of the final reaction step are rather similar for both CO and H, oxidation, with an excess energy of 101 (ref. 10) and 88-113 kJ mol-l, respectively. However, for CO oxida- tion it has been established that higher internal vibrational modes are populated and are not in equilibrium with each other, that the rotational temperature exceeds the sample Energy Partitioning in 2H, + 0, -+2H,O temperature, and that the molecules are desorbed with extreme translational energy (see Table 1). From these findings it is obvious that the CO, molecules dissipate only a minute fraction of the reaction energy.In particular, the high translational energy indi- cates that the molecules are either formed at a location where the molecules experience high repulsive forces, or that the internal modes couple very efficiently to the translation. In the past the appreciable energy content of desorbed product CO, has been reconciled by the geometry of the transition state, which must be bent. The transition to a linear CO, molecule should cause high excitations of the bending vibration which easily couples to the molecule-surface vibrational mode. Such an explanation might also be the clue to understanding the findings for water formation. Since the H,O has an internal angle of 104.5' which is not altered much in the adsorbed state, the geometry of the transition state might well be rather similar to that of the final state.Following this assumption, the bending vibration should not become significantly excited, and the excess energy goes into the 0-H stretch modes. However, it still remains a puzzle as to how the water molecule manages to dispose of, prior to desorption, the energy in its internal vibrational modes originating from the excess energy of the final reaction step, provided such exists, without causing the molecule-surface bond to rupture. This riddle nevertheless leads to the conclusion that the energy dissipation dynamics in the H, oxidation reaction differs fundamentally from that of the CO oxidation reaction.It is a pleasure to acknowledge continuous support and encouragement by G. Ertl. We are indebted to M. N. R. Ashfold and R. N. Dixon for providing us with the line- strength factor calculations which allowed simulation of the H,O and D,O spectra. P. Andresen and B. Kasemo are thanked for valuable suggestions. We appreciate the support of A. Mod1 in constructing the REMPI detector and of W. Nessler during the final stages of the experiments. References 1 M. C. Lin and G. Ertl, Annu. Rev. Phys. Chem., 1986,37,587. 2 D. A. Mantell, S. B. Ryali, B. L. Halpern, G. L. Haller and J. B. Fenn, Chem. Phys. Lett., 1981, 81, 185; D. A. Mantell, K. Kunimori, S. B. Ryali, G. L. Haller and J. B. Fenn, Surf Sci., 1986, 172, 281. 3 H. Zacharias, Int.J. Mod. Phys. B, 1990,4,45. 4 C. A. Becker, J. P. Cowin, L. Wharton and D. J. Auerbach, J. Chem. Phys., 1977,67,3394. 5 T. Matsushima, K. Shobatake, Y. Ohno and K. Tabayashi, J. Chem. Phys., 1992,97,2783. 6 M. Faraday, Experimental Researches in Electricity, Taylor, London, 1844. 7 I. Langmuir, Trans. Faraday. Soc., 1922, 17, 621. 8 P. R. Norton, in The Chemical Physics of Solid Surfaces and Heterogeneous Catalysis, ed. D. A. King and D. P. Woodruff, Elsevier, Amsterdam, 1982, vol. 4, pp. 27-72. 9 P. A. Thiel and T. E. Madey, Surf Sci. Rep., 1987,7,221. 10 C. T. Campbell, G. Ertl, H. Kuipers and J. Segner, J. Chem. Phys., 1980,73,5862. 11 K. M. Ogle and J. M. White, Surf: Sci., 1986, 169,425. 12 G. B. Fisher and J. L. Gland, Surf:Sci., 1980,94,446.13 C. Nyberg and C. G. Tengstll, J. Chem. Phys., 1984,80,3463. 14 T. A. Germer and W. Ho, Chem. Phys. Lett., 1989,163,449. 15 J. L. Gland, G. B. Fisher and E. B. Kolin, J. Catal., 1982,77,263. 16 L. K. Verheij, M. B. Hugenschmidt, B. Poelsema and G. Comsa, Surf: Sci., 1990,233,209. 17 J. R. Creighton and J. M. White, Surf: Sci., 1982, 122, L648. 18 K. M. Ogle and J. M. White, Suif Sci., 1984, 139,43. 19 V. J. Kwasniewski and L. D. Schmidt, J. Phys. Chem., 1992,96,5931. 20 J. N. Smith, Jr., and R. L. Palmer, J. Chem. Phys., 1971,56, 13. 21 T. Engel and H. Kuipers, Surf. Sci., 1979,90, 181. 22 S. T. Ceyer, W. L. Guthrie, T-H. Lin and G. A. Somorjai, J. Chem. Phys., 1983,78,6982. 23 A. B. Anton and D. C. Cadogan, Surf: Sci., 1990,239, L548.24 D. S. Y. Hsu and M. C. Lin, J. Chem. Phys., 1988,88,432. 25 S. Ljungstrom, B. Kasemo, A. RosCn, T. Wahnstrom and E. Fridell, Surf Sci., 1989, 216, 63; T. Wahns-trom, S. Ljungstrom, A. Rosen and B. Kasemo, Surf: Sci., 1990,234,439. A. de Meijere, K. W.Kolasinski and E. Hasselbrink 26 W. R. Williams, C. M. Marks and L. D. Schmidt, J. Phys. Chem., 1992,96,5922. 27 H. Conrad, G. Ertl and E. E. Latta, Surf. Sci., 1974, 41, 435. 28 T. Engel and H. Kuipers, Surf. Sci., 1979,90, 162. 29 G. E. Gdowski, T. E. Felter and R. H. Stulen, Surf: Sci., 1987,181, L147. 30 T. Matsushima, Surf. Sci., 1985, 157, 297. 31 M. Wolf, E. Hasselbrink, J. M. White and G. Ertl, J. Chem. Phys., 1990, 93, 5327; X-Y. Zhu, J. M. Wolf, E. Haselbrink and G.Ertl, J. Phys. Chem., 1991,95, 8395. 32 B. Hellsing, B. Kasemo, S. Ljungstrom, A. Rosen and T. Wahnstrom, Surf. Sci., 1987, 189/190, 851; B. Hellsing and B. Kasemo, Chem. Phys. Lett., 1988, 148,465. 33 T. Engel, J. Chem. Phys., 1978,69,373. 34 G. Meijer, J. J. ter Meulen, P. Andresen and A. Bath, J. Chem. Phys., 1986,85, 6914. 35 M. N. R. Ashfold, J. M. Bayley and R. N. Dixon, Chem. Phys., 1984, 84, 35; M. N. R. Ashfold, J. M. Bayley, R. N. Dixon and J. D. Prince, Ber. Bunsenges. Phys. Chem., 1985,89, 254. 36 A. Hodgson, J. P. Simons, M. N. R. Ashfold, J. M. Bayley and R. N. Dixon, Mol. Phys., 1985, 54, 351; A. Hodgson, J. P. Simons, M. N. R. Ashfold, J. M. Bayley and R. N. Dixon, Ber. Bunsenges. Phys. Chem., 1985,89,251. 37 H-H. Kuge and K. Kleinermanns, J. Chem. Phys., 1989,90,46. 38 M. N. R. Ashfold and R. N. Dixon, personal communication. 39 G. Herzberg, Molecular Spectra and Molecular Structure 11. Infrared and Raman Spectra of Polyatomic Molecules, Reinhold, New York, 2nd edn., 1991. 40 S. Bell, J. Mol. Spectrosc., 1965, 16, 205. 41 G. Theodorakopoulos, I. D. Petsalakis and R. J. Buenker, Chem. Phys., 1985,96,217. 42 D. G. Imre and J. Zhang, Chem. Phys., 1989, 139, 89; J. Z. Zhang, E. H. Abramson and D. G. Imre, J. Chem. Phys., 1991,95,6536. 43 W. L. Guthrie, T-H. Lin, S. T. Ceyer and G. A. Somorjai, J.Chem. Phys., 1982,76,6398. 44 C.T. Rettner, E. K. Schweizer and C. B. Mullins, J. Chem. Phys., 1989,90,3800. Paper 3/02973E; Received 21st May, 1993