15802
J. Phys. Chem. C 2008, 112, 15802–15808
Efficient Subpicosecond Photoinduced Surface Chemistry: The Ultrafast Photooxidation of CO on Palladium P. Szymanski,† A. L. Harris, and N. Camillone III* Chemistry Department, BrookhaVen National Laboratory, Upton, New York 11973 ReceiVed: May 20, 2008; ReVised Manuscript ReceiVed: August 5, 2008
Ultrafast near-infrared photoexcitation of mixed adlayers of carbon monoxide and atomic oxygen on the Pd(111) surface results in the efficient associative desorption of carbon dioxide. The oxidized fraction of the desorbed molecules ranges from ∼15% to 40%, approximately an order of magnitude greater than observed from Ru(001) under similar conditions. Time-resolved correlation measurements reveal fast dynamics that indicate the photochemistry is mediated by substrate electrons. The dynamics are strongly dependent on the initial CO coverage in a manner consistent with an activation energy for oxidation that depends on the initial CO binding site. These results, in addition to comparison of the ultrafast photooxidation to thermal oxidation, are consistent with the photooxidation occurring rapidly (within one to several picoseconds) from an initial mixed CO+O adlayer structure without time for diffusional phase segregation. The observed efficient ultrafast photoexcitation of associative desorption of CO2 from Pd(111) thus provides the opportunity to follow the picosecond dynamics of the bimolecular surface chemical reaction of CO+O from a well-defined initial geometry. Introduction The application of ultrafast laser technology to surface chemical dynamics has enabled studies of diffusion,1-3 changes in adsorbate-surface bonding,4-7 desorption,8-16 bimolecular reactions,17-21 and the initial stages of molecular dissociation22 with temporal resolution on time scales of molecular motion. This is possible because the subpicosecond deposition of energy into the adsorbate-substrate complex initiates fast dynamical processes such as DIMET23 (desorption induced by multiple electronic transitions) and substrate-mediated heating via electronic friction.12,24 In addition, the evolution of the excited complex can be followed with probe light pulses delayed on time scales equal to those of molecular vibrations and electron-phonon coupling. Thus, ultrafast laser-based pumpprobe techniques can in principle provide time-resolved snapshots of reaction dynamics starting from well-defined initial surface structures. This is in contrast to slow thermal excitation which gives adsorbates ample time to diffuse, readily allowing adsorbate structures to rearrange throughout the course of the measurement. A primary motivation for employing ultrafast pump-probe techniques in surface chemistry is to elucidate the adsorbate reaction dynamics which determine the efficiency of reactions important to heterogeneous catalysis. It is therefore desirable that the photoinitiated surface reactions probed can be linked to thermal reactions. This may be addressed in part by the use of pump radiation within a wavelength range (∼600-1000 nm) that is not directly absorbed by the adsorbates and does not generally result in electron transfer to dissociative states. Instead, the energy is deposited in the surface, rapidly creating transient hot electron distributions that can effectively couple energy into the reaction coordinate on subpicosecond time scales. Subse* To whom correspondence should be addressed. E-mail: nicholas@ bnl.gov. † Current address: Chemistry Division, Los Alamos National Laboratory, MS J567, Los Alamos, NM 87545.
quent bond cleavage or formation is believed to occur by pathways similar to those in thermally excited surface reactions. In this context, it is also important to note that a clear link between thermal surface chemical reactions and hot electrons has been established by both experiment and theory,25-27 indicating that electronic excitation is a key mechanism for energy transfer. For this reason, ultrafast photoinduced desorption and reaction processes continue to be pursued by us and others as a tool for probing surface reaction dynamics. Furthermore, to conduct time-resolved probing of the adsorbates following the excitation it is ideal to choose adsorbates amenable to high sensitivity vibrational spectroscopic measurements at monolayer and submonolayer surface coverages. The high oscillator strength of CO satisfies this requirement, and CO has been a model system for vibrational-spectroscopy-based surface dynamics studies.1,13,28-30 Many ultrafast photoexcited surface chemistry studies of CO have focused on desorption.9,11-13,16,31,32 By contrast, few studies of prototype bimolecular reaction dynamics such as CO oxidation have been undertaken.18,19,33 This is largely due to the fact that the high degree of electronic excitation employed in these studies— electronic temperatures of 3000-7000 K are typical—generally results in highly efficient desorption of unreacted adsorbates, thus preventing reactions from occurring. An example of the promise and the challenge of following the ultrafast dynamics of a bimolecular surface reaction is the oxidation of CO on Ru(001). For this system, Bonn et al. demonstrated the possibility of subpicosecond-pulse-induced reactions between CO and atomic oxygen.19 However, given the high electronic temperatures required to activate the reaction, simple desorption was the predominant process. The photooxidation efficiency was quite low; with 0.25 ML CO coadsorbed with a saturation coverage of 0.5 ML O, CO2 comprised less than 3% of the photodesorbed products. The remaining >97% of the photodesorbed species were unreacted CO molecules. We present here the results of a study of the ultrafast photooxidation of CO on Pd(111) where the photoinduced
10.1021/jp8044737 CCC: $40.75 2008 American Chemical Society Published on Web 09/11/2008
Efficient Subpicosecond Photoinduced Surface Chemistry oxidation is very efficient; ∼25% of the desorbed CO is oxidized when the reaction is photoinitiated from 0.25 ML CO coadsorbed with a saturation coverage of 0.25 ML O. By using twopulse correlation to determine the rapid time scale of the surface chemical dynamics, we establish the strong role of electronic coupling in driving the photoreaction. We make preliminary conclusions about the nature of the reaction process, and the role of CO binding site in determining the strength of the driving electronic coupling, by comparing the reaction and desorption probabilities as a function of surface reaction conditions (CO coverage and thermal excitation vs photoexcitation). We compare our results to those obtained for (CO+O)/Ru(001), and link the reaction studies to our previous studies of CO desorption from Pd(111).16 Overall, we conclude that the photooxidation of CO on Pd(111) is likely to occur directly from the as-prepared mixed adlayer arrangement, and is therefore a good candidate for future ultrafast spectroscopic determination of the dynamics of a bimolecular surface reaction. We are motivated to investigate the reaction dynamics in the (CO+O)/Pd(111) system because CO+O is the prototypical catalytic oxidation reaction and a wealth of information, including recent experimental surface structure and chemistry studies34,35 and theoretical binding site energies and transition state geometries on Pd(111),36-39 is available. These studies suggest that (CO+O)/Pd(111) is a good candidate in the search for an efficient subpicosecond photoinduced bimolecular reaction. For example, it has been shown that thermal excitation drives the oxidation reaction efficiently over a wide range of coveragesaccessibleunderultrahighvacuum(UHV)conditions.40,41 This distinguishes Pd(111) from Ru(001), which does not catalyze the formation of CO2 from CO+O by thermal heating under UHV conditions.42-44 In addition, the current ultrafast reaction studies utilize as a starting point a mixed CO+O phase in which each O atom is surrounded by three CO molecules34sa geometry that should be favorable for efficient photoinduced chemistry. Experimental Section The experiments are conducted in a UHV chamber (base pressure ∼5 × 10-11 Torr) equipped with a quadrupole mass spectrometer (QMS) for detection of photo- and thermally desorbed products.15 The Pd(111) surface is cleaned by repeated Ar+-sputtering and annealing cycles, followed by oxygen exposure.15 Mixed CO+O adsorbate layers are prepared by exposing the clean surface to O2 at 300 K to deposit 0.25 ML O, followed by CO exposure at 90 K from a directed doser to deposit controlled quantities of CO. In all our measurements, isotopically labeled 12C18O is used to avoid contributions to the QMS signal from background CO. Low-energy electron diffraction (LEED) was used to confirm that the symmetries of the adlayers were consistent with previously reported results.34,36,45 Temperature-programmed reaction/desorption (TPR/D) mass spectrometry is employed to characterize the thermal chemistry and to measure the initial CO coverage. By monitoring the desorption of 12C18O and 12C16O18O we are able to determine the fate of each 12C18O molecule and accurately quantify the oxidized percentage of the total desorption yield, defined as the percentage of the total desorbed yield (CO+CO2) that desorbs as CO2. Accurate CO coverage (θCO) calibration is accomplished by comparison to TPD of 0.33 ML (3×3)R30° CO (no O present). Surface excitation for the ultrafast photodesorption and reaction experiments is achieved using the ∼110 fs (sech2 profile) pulses from a regeneratively amplified Ti:sapphire laser
J. Phys. Chem. C, Vol. 112, No. 40, 2008 15803 operating at 780 nm. The time-resolved dynamics following subpicosecond-pulse irradiation are measured with the two-pulse correlation (2PC) technique; the excitation pulse is divided into two pulses separated by a variable delay time, and the photoyield for a given product is measured as a function of the delay between the pulses. Due to the nonlinear dependence of the photoyield on the laser fluence, this approach permits the measurement of the relaxation time(s) for energy deposited into the reaction coordinate. In the 2PC experiments presented here, the total laser fluence was measured to be ∼18 mJ cm-2. At this fluence, the desorption yield resulting from a single pulse (∼9 mJ cm-2) is typically less than ∼10% of the yield due to a pair of completely overlapped pulses (zero delay time). Desorbed molecules are detected by the QMS on a shot-toshot basis where each shot is comprised of a single pulse pair. Reported photodesorption yields are “first-shot yields”, i.e., the products desorbed by the first pulse pair to impinge upon a freshly prepared adsorbate layer. Because the laser spot is small, several single-shot measurements may be made on a given adlayer preparation by translating the crystal between measurements to expose a fresh area of the surface to the laser. The sensitivity of the mass spectrometer to photodesorbed CO is calibrated against the integrated photodesorption yields measured for the complete depletion of 0.33 ML CO from the surface with multiple laser shots at low temperatures.16 In both the thermal- and photodesorption measurements the CO2 product signal is normalized by accounting for differences in ionization efficiency between CO and CO2 and the cracking pattern of CO2. Results and Discussion The starting point for our ultrafast photooxidation experiments is the θO ) 0.25 ML atomic oxygen adlayer which forms by self-limiting growth during room temperature exposure of Pd(111) to molecular oxygen.45 It has been calculated theoretically36 and shown experimentally34 that the O atoms in the resultant (2×2) structure are bound to face-centered cubic (fcc) sites (Figure 1a). After cooling this O-covered surface to 90 K, we then deposit CO in controlled amounts to produce the desired surface stoichiometry for the reaction. Recent scanning tunneling microscopy (STM)34 and density functional theory (DFT)36 studies of the (CO+O)/Pd(111) surface have shown that at temperatures below ∼110 K, the reaction to give CO2 is completely suppressed; the O atoms remain fixed at the fcc sites during the coadsorption, and the CO populates new sites within the existing (2×2) O unit mesh. We have confirmed with LEED (data not shown here) that at ∼90 K the adlayer maintains the (2×2) symmetry over the full CO coverage range (0 e θCO e 0.5 ML) in our experiments. For a 1:1 mixture of CO and O (θCO ) θO ) 0.25 ML), DFT predicts that in a (2×2) mixed adlayer the CO prefers hexagonal close-packed (hcp) sites (Figure 1b),36 a structure consistent with STM measurements.34 By analogy with the saturated CO-only surface (θCO ) 0.75 ML, Figure 1d), we expect that for θCO > 0.25 ML, CO begins to adsorb at top sites, ultimately resulting in the formation of a “honeycomb” network of CO molecules surrounding the O atoms, maintaining the (2×2) periodicity (Figure 1c). This structure is consistent with our observation that the CO coverage after saturation of the O-precovered (2×2) surface is within experimental error of the expected value of 0.5 ML. Ultrafast photoexcitation of these mixed monolayers results in the efficient desorption of both CO and CO2; O and O2 photodesorption were not observed under the experimental
15804 J. Phys. Chem. C, Vol. 112, No. 40, 2008
Figure 1. (a) Structure of (2×2) O/Pd(111). Possible structures for (CO+O)/Pd(111), after Me´ndez et al.,34 with one oxygen atom and one (b) or two (c) CO molecules per unit cell (θCO ) 0.25 and 0.5 ML, respectively). In the saturated (θCO ) 0.75 ML) CO-only surface (d) the same surface sites are occupied as in (c) but with CO occupying the fcc sites.
Figure 2. Two-pulse correlation measurements of the CO2 photoproduct from initial CO coverages of 0.25 (circles) and 0.4 ML (squares) deposited on the O-(2×2)/Pd(111) (0.25 ML O) surface.
conditions. We focus our considerations here on the desorption of the photooxidation product, CO2. Figure 2 shows 2PC measurements of the desorbing CO2 photoproduct from initial CO coverages corresponding to a 1:1 CO:O mixture and to a nearly CO-saturated O-(2×2) surface. There are three remarkable features in this data. First, the 2PC relaxation times are quite short, particularly at the lower coverage, where the response time is subpicosecond, clearly indicating that the desorption is the result of electron-mediated surface-adsorbate energy transfer. Second, the photooxidation percentage yield is high: the first-shot yield at zero delay represents desorption of
Szymanski et al. ∼0.02 ML-equivalents of CO2, compared to a total yield (of desorbed plus reacted CO) of ∼0.09 ML. Thus the oxidized percentage of the total desorption yield is ∼20%, roughly an order of magnitude greater than prior ultrafast reaction measurements observed from Ru(001) under similar conditions19 and approximately one-third to one-half that observed in TPR/D for the very efficient thermally activated oxidation on Pd(111). Finally, we find that the 2PC decay time is strongly coverage dependent, an effect which to the best of our knowledge has not been observed previously and which we attribute to a coverage dependence in the activation energy of the reaction. Each of these features is discussed in further detail below, showing that the ultrafast photooxidation provides the opportunity to take a snapshot of the chemical reaction and observe coverage-dependent dynamics for well-defined initial structures that cannot be isolated during (slow) thermal activation. The 2PC result (Figure 2) gives direct insight into the energy transfer mechanism responsible for the ultrafast photooxidation. We find that the 2PC data are well-described by biexponential decays; the amplitudes (A) and time constants (τ) extracted from biexponential fits to the data are given in Table 1. Comparison of the amplitudes for the fast (Afast) and slow (Aslow) components of the biexponential shows that a very fast process, with empirical time constant τfast, dominates the CO2 2PC decay. For the 1:1 mixed adlayer τfast is subpicosecond, resulting in a full width at half-maximum (fwhm) of 2 ps. Due to the relatively slow electron-phonon coupling in the Pd substrate, only adsorbate coupling to substrate electrons can result in such a fast response in the adsorbate-substrate complex, as has been established in prior ultrafast studies of adsorbate desorption.19,46,47 It would be expected that resonant electron transfer is responsible for this observed strong electronic coupling given the presence of low-lying antibonding states within ∼2 eV both above and below the Fermi level in the O/Pd(111) system.48,49 We also note that the same rapid relaxation was found for ultrafast CO photooxidation on Ru(001) by Bonn et al., who correspondingly attributed the reaction to electron-mediated activation of the oxygen.19 In order to quantify the photooxidation efficiency we measured the oxidized percentage of the desorption yield as a function of initial CO coverage while holding the O coverage fixed at 0.25 ML. Consider first the thermally activated oxidation, which we studied by TPR/D. Figure 3 shows representative CO2 and CO TPR/D spectra recorded for a (2×2) O/Pd(111) adlayer with between 0.06 and 0.49 ML of CO deposited at 90 K. At the larger CO coverages, two main CO2 desorption waves are evident (Figure 3a), the first with a low temperature shoulder showing a gradual onset at ∼150 K, and a steep rise beginning just above 200 K. The latter coincides with a fall off in the CO desorption (Figure 3b). The CO desorption also occurs in two main waves: a low temperature peak at ∼190 K, which, by comparison with the TPD spectrum measured for a saturated CO-only surface (Figure 3b), is thought to represent desorption of top-site CO accompanied by a phase transition;50 and a high temperature peak at ∼480 K, which coincides with desorption of fcc-site CO,36,50-52 observed in TPD of the 0.33 ML (3×3)R30° CO/Pd(111) (data not shown). We interpret these spectra based on previous STM34,35 and combined LEED and TPR/D40,53 studies. These have shown that deposition of CO at temperatures of ∼140 K or above causes compression of the (2×2) oxygen layer into segregated (2×1) domains, with CO adsorbing in areas between the domains. The compression of the O lowers the O-Pd bond strength,36
Efficient Subpicosecond Photoinduced Surface Chemistry
J. Phys. Chem. C, Vol. 112, No. 40, 2008 15805
TABLE 1: Two-Pulse Correlation Dynamics θCO (ML)
θO (ML)
0.25
0.25
0.4
0.25
molecule CO CO2 CO CO2
Afast(ML)
τfast (ps)
Aslow (ML)
τslow (ps)
fwhm (ps)
0.04 ( 0.01 0.015 ( 0.003 0.15 ( 0.01 0.022 ( 0.004
6(1 0.6 ( 0.4 12 ( 1 6(2
∼0.01 0.007 ( 0.003 0.070 ( 0.001 0.008 ( 0.003
∼100 20 ( 10 140 ( 30 g50
13 2 30 14
effectively activating the O, making it reactive toward CO2 production at temperatures as low as ∼140 K.35 Reaction between CO and these activated, compressed (2×1) O domains explains the high CO2 yield observed from high CO coverage (θCO g 0.25 ML) adlayers in the ∼150-250 K range (Figure 3a). The highest coverage data in Figure 3 is consistent with the following considerations: below ∼200 K, at temperatures consistent with CO desorption from top sites (populated only in the highest coverage CO structures), CO desorption dominates, and CO2 production is limited. Desorption of the most weakly bound CO creates vacancies on the surface necessary to accommodate diffusion and phase segregation. At ∼200 K, sufficient CO has desorbed to allow for phase segregation and, therefore, compression of the O atom lattice and concomitant activation of the O atoms and efficient production of CO2. By ∼275 K, roughly half of the O and CO have been consumed. At this stage the O has expanded from the (2×1) to a (3×3)R30° structure,34,40 and is less reactive.40 The adlayer remains phase-segregated, and desorption of both CO and O is prohibited until the temperature reaches ∼400 K. However, in the 275-400 K range, thermal activation is sufficient to cause the oxidation to proceed, though more slowly than at lower temperature (corresponding to higher coverage), when the (2×1) O structure was present. Finally, continued heating beyond ∼400 K causes CO desorption to compete with oxidation. Any remaining atomic oxygen recombinatively desorbs at ∼780 K (data not shown).
Figure 3. Temperature-programmed desorption and reaction spectra from (CO+O)/Pd(111) as a function of CO coverage measured for (a) CO2 and (b) CO. A spectrum for the CO-only adlayer at saturation coverage (0.75 ML) is shown for comparison. The ramp rate was 2.5 K/s.
The thermally activated oxidation in TPR/D is quite efficient. We have quantified this for the 0.25 ML O adlayer by measuring the oxidized percentage of the yield, as a function of CO coverage (Figure 4). The ideal catalyst would oxidize all CO on the surface, with CO desorbing only if it were in excess. While this limit (indicated by the solid line in Figure 4) is approached at low CO coverage, desorption without reaction is significant at higher coverages. At a stoichiometric ratio (θCO ) θO ) 0.25 ML), where only a small amount of the O adlayer is compressed to the (2×1) structure (based on the intensity of the low-temperature CO2 TPD peak40), ∼65% of the CO is oxidized. This is the maximum deviation of the actual percentage yield from the ideal; as the CO coverage is increased beyond the stoichiometric ratio, the ideal oxidized yield necessarily drops because the CO is in excess. The observed oxidized percentage of the yield does not decrease as quickly, and at saturation coverage (θCO ) 2×θO), the surface acts as an ideal catalyst—essentially all of the CO that can be oxidized (i.e., 50%) is oxidized. Comparison of the photooxidation yield to the thermal oxidation yield reveals a significant difference in the CO oxidation/desorption branching ratio which we attribute to the difference in excitation time scales. We have measured the firstshot yield at zero delay for both the CO and CO2 products to quantify the oxidized percentage of the photoyield as a function of CO coverage. Comparison of these measurements with the TPR/D results shows that the relative efficiency of CO oxidation versus desorption following photoexcitation is coverage-dependent in a manner quite similar to thermal excitation (Figure 4). In fact, the trends nearly parallel each other, with the photooxidation showing a percentage yield that is approximately one-third that of the thermal percentage yield. However, the relative inefficiency of the photooxidation compared to the thermal oxidation indicates that, under ultrafast photoexcitation, desorption competes more effectively with oxidation than during thermal excitation. This is consistent with the fact that the
Figure 4. Oxidized percentage of the yield, defined as the percentage of the total desorbed yield (CO+CO2) that desorbs as CO2, as a function of CO coverage at an initial θO ) 0.25 ML for (i) an ideal catalyst (solid line) where all of the CO is oxidized, (ii) TPR/D (open circles), and (iii) ultrafast photooxidation (filled squares). The yield observed by Bonn et al. for Ru at an initial θO ) 0.50 ML19 is shown for comparison (open square).
15806 J. Phys. Chem. C, Vol. 112, No. 40, 2008 photooxidation occurs quickly, on a time scale of ∼1 ps,16,19,46,54 and suggests that under ultrafast photoexcitation the adlayer does not have sufficient time to undergo the phase transition to form the activated oxygen domains, nor adequately sample reaction geometries, in stark contrast to the conditions that prevail during thermal reaction. The implication is that the fast photooxidation provides the opportunity to take a snapshot of the chemical reaction between nearest neighbors by measuring the dynamics from a well-defined initial structure (Figure 1b or c). In contrast, as described above, the (seconds time scale) thermal reaction during TPR for this system is believed to take place from an intermediate state that is substantially rearranged from the lowtemperature mixed structures. It is also instructive to compare the ultrafast photooxidation yield from (CO+O)/Pd(111) with that observed from (CO+O)/ Ru(001). Although the coadsorbates form mixed phases on both surfaces (albeit with different site preferences), we find that ultrafast photoinduced oxidation is far more favorable on Pd(111) (Figure 4). Whereas CO2 comprises from ∼15% to 40% of the yield from (CO+O)/Pd(111), the percentage CO2 yield from a saturation coverage of 0.25 ML CO + 0.5 ML O on Ru(001) is only 3%.19 The increased efficiency on Pd may be understood in the same way that its superiority as a CO oxidation catalyst under UHV conditions is explained, namely that the (second-order) activation energy for desorption of oxygen from Pd (111) (∼2.2 eV) is significantly less than for that from Ru(001) (∼3.6 eV).55 With the oxygen so much more loosely bound to Pd(111), it is more easily activated to react with CO.42 On Ru(001), the strong O-Ru bond makes CO oxidation negligible in TPR/D. Thus, the observation of ultrafast photoinduced CO oxidation on Ru(001) was of great interest as it clearly evidenced the opening of a new, electron-mediated reaction channel. Furthermore, the observation of enhancement of the photooxidation with a decrease in O binding to the metal substrate is consistent with the idea that activation of the oxygen is critical to the ultrafast process.19 Finally, we consider the strong coverage dependence of the oxidation dynamics shown in Figure 2. We observe a factor of 10 increase in τfast when the CO coverage is increased from 0.25 to 0.4 ML (Table 1). A similar, though smaller effect, is also observed for the CO desorption dynamics. We are not aware of any previous reports of such an effect. Furthermore, the magnitude of the effect on the oxidation dynamics suggests a large change in the strength of the substrate-adsorbate coupling that drives the energy transfer responsible for the reaction. The qualitative relationship between the substrate-adsorbate coupling strength and empirical time constant τfast can be understood intuitively; the weaker the coupling to the electronic degrees of freedom of the substrate, the slower the response of the adsorbate to the laser-induced heating. Slower adsorbate heating (and cooling) results in a broader 2PC and thus a larger value for τfast. Quantitatively, based on previous photodesorption simulations employing a two-temperature model for the substrate electron and lattice degrees of freedom56 in conjunction with an “empirical friction” model to describe the substrate-adsorbate energy transfer,12 we expect that the observed increase in the fwhm of the CO2 2PC is consistent with an order of magnitude decrease in the substrate-adsorbate coupling strength.46 All else being equal, we would anticipate such a large decrease in coupling strength to result in a very large (1 or 2 orders of magnitude) decrease in reaction efficiency.46 However, it is reasonable to expect that a decrease in coupling strength is accompanied by a decrease in molecule-surface bond strength, which in turn correlates with a decrease in the desorption
Szymanski et al.
Figure 5. First-shot probability vs initial CO coverage for (a) CO2 formation and (b) CO desorption.
activation energy. In the case of O2/Pd(111) we have seen that it is possible for the decrease in desorption activation energy to more than compensate for the decrease in coupling, resulting in an overall increase in desorption probability.15 However, in the case of (CO+O)/Pd(111), interpretation of the data is further complicated by the availability of two channels, the oxidation reaction and the desorption. For this reason we have recast the reaction efficiency in terms of probabilities, shown in Figure 5, where the reaction (desorption) probability is defined as the ratio of the first-shot yield of CO2 (CO) to the initial CO coverage. These data reveal two striking trends. First, the probability that an adsorbed CO molecule will react to produce CO2 is found to be essentially independent of CO coverage between θCO ) 0.1 and 0.5 ML (Figure 5a). This is surprising, given the order of magnitude change in the fast relaxation time observed in the 2PC measurements (Table 1). Second, the desorption probability for CO increases markedly with coverage. This strongly suggests that the activation energy for CO photodesorption decreases as more CO is added to the system. In fact, all of our observations are simply explained by a decrease in the CO-Pd bond strength with increasing coverage, which in turn is linked to a dependence of the activation energy for oxidation upon the CO adsorption site. The population of increasingly more weakly bound sites is accompanied by a decrease in activation energy for reaction and a decrease in the strength of the coupling of energy into both the reaction and desorption coordinates as evidenced by the substantial increases in relaxation time for both processes. Experiment50,51,57,58 and theory36,59 have shown that for CO/Pd(111), the strength of the molecule—surface bond is a strong function of coverage due to repulsive, surface-mediated lateral interactions and the occupation of increasingly less tightly binding sites with increasing coverage. The activation energy for desorption from a clean Pd(111) surface varies dramatically, from ∼1.5 eV (extrapolated to zero coverage) to ∼0.2 eV at saturation.57 In addition, DFT calculations37 have indicated that the activation energy for CO oxidation is a strong function of the site at which
Efficient Subpicosecond Photoinduced Surface Chemistry the CO is bound: the activation energy for the CO+O reaction is ∼0.6 eV less when CO starts at the top site as compared to the hollow site, where it is more tightly bound. Thus, we propose the following tentative explanation for our observations. For the desorption channel the decrease in coupling strength with increasing coverage results in a ∼2-fold increase in the relaxation time of the molecule-surface complex (Table 1). At the same time, the large decrease in the activation energy is sufficient to more than compensate for the weaker coupling, and the desorption efficiency increases markedly (Figure 5b). For the oxidation channel, the large decrease in coupling strength is also accompanied by a large decrease in activation energy, but the data suggest that the expected increase in photooxidation yield is offset by the strong increase in the desorption probability. Thus, despite the large decrease in the oxidation activation energy for CO starting from the top site compared to CO starting from the hcp site, the preference for desorption apparently outweighs any increase in the photooxidation yield and the probability of oxidation remains essentially the same. Experiments or simulations with the capability to more directly address adsorbate excitation are required to confirm this interpretation. Finally, the 2PC measurements indicate that excitation of both O and CO is involved in initiating the ultrafast photooxidation. The measured relaxation times in Table 1 show that at both coverages studied, relaxation of the CO desorption coordinate is significantly slower than that of the reaction coordinate. The relatively quick cooling of the reaction coordinate indicates that the CO remains “hot” for a period of time after the oxidation has shut down. This indicates that strong electronic coupling to the oxygen is responsible for the fast dynamics of the oxidation, i.e., it is the rapid cooling of the O excitation that effectively drains energy out of the reaction coordinate. On the other hand, the observed coverage dependence of the dynamics associated with both oxidation and desorption is correlated with a decrease of the CO-Pd coupling strength; previous work with the lower-coverage (3×3)R30° structure suggests that this coupling is also primarily electronic.16 We believe that this indicates that the excitation of the CO is also important to driving the photooxidation—when the CO–Pd coupling is weaker (at higher coverages, where top sites are occupied) the CO remains hot longer, resulting in a slower relaxation of the excitation in the reaction coordinate. Thus excitation of both the O and CO plays a significant role in driving the reaction. This concept is consistent with indications from DFT that (i) the CO is more mobile than the O and (ii) reaching the transition state for oxidation requires both translation of the O and frustrated rotation of the CO.37-39 Conclusions In summary, we have observed subpicosecond-pulse-induced CO2 production from (CO+O)/Pd(111) surfaces containing 0.25 ML O and varying amounts of CO. The oxidized fraction of the desorbed molecules ranges from ∼15% to 40%, which is significantly higher than the ∼3% observed from Ru(001), a difference consistent with the lower activation energy for O desorption from Pd(111). Despite the presence of an efficient thermal oxidation channel on Pd(111), an electron-mediated process is responsible for the ultrafast photooxidation. The reaction takes place in a mixed phase in contradistinction to the thermal chemistry between segregated phases, and we attribute the coverage-dependence of the dynamics primarily to coverage- and site-dependent CO-Pd interactions. The strong coverage dependence of the dynamics is consistent with a
J. Phys. Chem. C, Vol. 112, No. 40, 2008 15807 decrease in the CO-Pd coupling strength with increasing CO coverage, and a decrease in the activation energies for desorption and reaction as the CO moves from occupying only hcp sites to occupying both hcp and top sites. In addition, the timeresolved measurements indicate that excitation of both O and CO play a significant role in driving the oxidation. As the same transition state(s) can apply to both segregated and mixed phases, our studies, with well-defined initial conditions, point toward a means for investigating fundamental interactions in the catalysis of CO oxidation. In particular, the relative efficiency of the CO2 channel in photodesorption suggests the possibility of time-domain measurements of energy transfer, molecular motion and bimolecular reaction dynamics by timeresolved pump-probe surface spectroscopy. Acknowledgment. We thank M. G. White and SUNY Stony Brook for the extended loan of the Ti:S amplifier used in these experiments. We also thank Ping Liu and Yixiong Yang for useful discussions. This research was supported by the U.S. Department of Energy, Office of Basic Energy Sciences, Division of Chemical Sciences, under Contract No. DE-AC0298CH10886. References and Notes (1) Backus, E. H. G.; Eichler, A.; Kleyn, A. W.; Bonn, M. Science 2005, 310, 1790. (2) Gu¨dde, J.; Ho¨fer, U. J. Phys.: Condens. Matter 2006, 18, S1409. (3) Kubota, J.; Yoda, E.; Ishizawa, N.; Wada, A.; Domen, K.; Kano, S. S. J. Phys. Chem. B 2003, 107, 10329. (4) Domen, K.; Bandara, A.; Kubota, J.; Onda, K.; Wada, A.; Kano, S. S.; Hirose, C. Surf. Sci. 1999, 427–428, 349. (5) Bauer, M.; Lei, C.; Read, K.; Tobey, R.; Gland, J.; Murnane, M. M.; Kapteyn, H. C. Phys. ReV. Lett. 2001, 87, 025501. (6) Petek, H.; Nagano, H.; Weida, M. J.; Ogawa, S. J. Phys. Chem. B 2001, 105, 6767. (7) Kubota, J.; Wada, A.; Domen, K. J. Phys. Chem. B 2005, 109, 20973. (8) Budde, F.; Heinz, T. F.; Loy, M. M. T.; Misewich, J. A.; de Rougemont, F.; Zacharias, H. Phys. ReV. Lett. 1991, 66, 3024. (9) Prybyla, J. A.; Tom, H. W. K.; Aumiller, G. D. Phys. ReV. Lett. 1992, 68, 503. (10) Kao, F.-J.; Busch, D. G.; Cohen, D.; Gomes da Costa, D.; Ho, W. Phys. ReV. Lett. 1993, 71, 2094. (11) Struck, L. M.; Richter, L. J.; Buntin, S. A.; Cavanagh, R. R.; Stephenson, J. C. Phys. ReV. Lett. 1996, 77, 4576. (12) Funk, S.; Bonn, M.; Denzler, D. N.; Hess, C.; Wolf, M.; Ertl, G. J. Chem. Phys. 2000, 112, 9888. (13) Bonn, M.; Hess, C.; Funk, S.; Miners, J. H.; Persson, B. N. J.; Wolf, M.; Ertl, G. Phys. ReV. Lett. 2000, 84, 4653. (14) Cai, L.; Xiao, X. D.; Loy, M. M. T. J. Chem. Phys. 2001, 115, 9490. (15) Szymanski, P.; Harris, A. L.; Camillone, N., III J. Chem. Phys. 2007, 126, 214709. (16) Szymanski, P.; Harris, A. L.; Camillone, N., III J. Phys. Chem. A 2007, 111, 12524. (17) Busch, D. G.; Ho, W. Phys. ReV. Lett. 1996, 77, 1338. (18) Deliwala, S.; Finlay, R. J.; Goldman, J. R.; Her, T. H.; Mieher, W. D.; Mazur, E. Chem. Phys. Lett. 1995, 242, 617. (19) Bonn, M.; Funk, S.; Hess, Ch.; Denzler, D. N.; Stampfl, C.; Scheffler, M.; Wolf, M.; Ertl, G. Science 1999, 285, 1042. (20) Arnolds, H. Surf. Sci. 2004, 548, 151. (21) Denzler, D. N.; Frischkorn, C.; Hess, C.; Wolf, M.; Ertl, G. Phys. ReV. Lett. 2003, 91, 226102. (22) Lane, I. M.; King, D. A.; Liu, Z. P.; Arnolds, H. Phys. ReV. Lett. 2006, 97, 186105. (23) Misewich, J. A.; Heinz, T. F.; Newns, D. M. Phys. ReV. Lett. 1992, 68, 3737. (24) Brandbyge, M.; Hedegård, P.; Heinz, T. F.; Misewich, J. A.; Newns, D. M. Phys. ReV. B 1995, 52, 6042. (25) Gergen, B.; Nienhaus, H.; Weinberg, W. H.; McFarland, E. W. Science 2001, 294, 2521. (26) Ji, X. Z.; Somorjai, G. A. J. Phys. Chem. B 2005, 109, 22530. (27) Tully, J. C. Annu. ReV. Phys. Chem. 2000, 51, 153. (28) Morin, M.; Levinos, N. J.; Harris, A. L. J. Chem. Phys. 1992, 96, 3950. (29) Zhang, V. L.; Arnolds, H.; King, D. A. Surf. Sci. 2005, 587, 102.
15808 J. Phys. Chem. C, Vol. 112, No. 40, 2008 (30) Lane, I. M.; King, D. A.; Arnolds, H. J. Chem. Phys. 2007, 126, 024707. (31) Cai, L.; Xiao, X.; Loy, M. M. T. Surf. Sci. 2000, 464, L727. (32) Fournier, F.; Zheng, W.; Carrez, S.; Dubost, H.; Bourguignon, B. J. Chem. Phys. 2004, 121, 4839. (33) Kao, F.-J.; Busch, D. G.; Gomes da Costa, D.; Ho, W. Phys. ReV. Lett. 1993, 70, 4098. (34) Me´ndez, J.; Kim, S. H.; Cerda´, J.; Wintterlin, J.; Ertl, G. Phys. ReV. B 2005, 71, 085409. (35) Kim, S. H.; Me´ndez, J.; Wintterlin, J.; Ertl, G. Phys. ReV. B 2005, 72, 155414. (36) Seitsonen, A. P.; Kim, Y. D.; Schwegmann, S.; Over, H. Surf. Sci. 2000, 468, 176. (37) Eichler, A. Surf. Sci. 2002, 498, 314. (38) Salo, P.; Honkala, K.; Alatalo, M.; Laasonen, K. Surf. Sci. 2002, 516, 247. (39) Zhang, C. J.; Hu, P. J. Am. Chem. Soc. 2001, 123, 1166. (40) Matsushima, T.; Asada, H. J. Chem. Phys. 1986, 85, 1658. (41) Engel, T.; Ertl, G. J. Chem. Phys. 1978, 69, 1267. (42) Stampfl, C.; Scheffler, M. Phys. ReV. Lett. 1997, 78, 1500. (43) Kostov, K. L.; Rauscher, H.; Menzel, D. Surf. Sci. 1992, 278, 62. (44) Lee, H.-I.; White, J. M. J. Catal. 1980, 63, 261. (45) Conrad, H.; Ertl, G.; Ku¨ppers, J.; Latta, E. E. Surf. Sci. 1977, 65, 245. (46) Szymanski, P.; Harris, A. L.; Camillone, N., III Surf. Sci. 2007, 601, 3335. (47) Frischkorn, C.; Wolf, M. Chem. ReV. 2006, 106, 4207.
Szymanski et al. (48) Eichler, A.; Mittendorfer, F.; Hafner, J. Phys. ReV. B 2000, 62, 4744. (49) Todorova, M.; Reuter, K.; Scheffler, M. J. Phys. Chem. B 2004, 108, 14477. (50) Kuhn, W. K.; Szanyi, J.; Goodman, D. W. Surf. Sci. 1992, 274, L611. (51) Giessel, T.; Schaff, O.; Hirschmugl, C. J.; Fernandez, V.; Schindler, K.-M.; Theobald, A.; Bao, S.; Lindsay, R.; Berndt, W.; Bradshaw, A. M.; Baddeley, C.; Lee, A. F.; Lambert, R. M.; Woodruff, D. P. Surf. Sci. 1998, 406, 90. (52) Ohtani, H.; Van Hove, M. A.; Somorjai, G. A. Surf. Sci. 1987, 187, 372. (53) Conrad, H.; Ertl, G.; Ku¨ppers, J. Surf. Sci. 1978, 76, 323. (54) For electron-mediated desorption and reaction, the rate typically decreases by an order of magnitude on a timescale comparable to the fwhm of the 2PC. (55) Madey, T. E.; Engelhardt, H. A.; Menzel, D. Surf. Sci. 1975, 48, 304. (56) Anisimov, S. I.; Kapeliovich, B. L.; Perel’man, T. L. SoV. Phys. JETP 1974, 39, 375. (57) Guo, X.; Yates, J. T., Jr J. Chem. Phys. 1989, 90, 6761. (58) Rose, M. K.; Mitsui, T.; Dunphy, J.; Borg, A.; Ogletree, D. F.; Salmeron, M.; Sautet, P. Surf. Sci. 2002, 512, 48. (59) Glassey, W. V. Surf. Sci. 2006, 600, 173.
JP8044737