Highly Efficient Ultrafast Energy Transfer into Molecules at Surface

6149

2007, 111, 6149-6153 Published on Web 04/05/2007

Highly Efficient Ultrafast Energy Transfer into Molecules at Surface Step Sites Ellen H. G. Backus,† Mattias Forsblom,‡ Mats Persson,‡ and Mischa Bonn*,†,§ Leiden Institute of Chemistry, Leiden UniVersity, P.O. Box 9502, 2300 RA Leiden, The Netherlands, Surface Science Research Centre/Department of Chemistry, The UniVersity of LiVerpool, LiVerpool L69 3BX, United Kingdom, and FOM Institute for Atomic and Molecular Physics, Kruislaan 407, 1098 SJ Amsterdam, The Netherlands ReceiVed: January 19, 2007; In Final Form: March 8, 2007

Ultrafast laser-induced desorption of NO and CO molecules on a platinum surface reveals that the rate of energy transfer between the laser-heated electrons and the adsorbate depends critically on the precise adsorbate location. An analysis based on a simple electronic friction model suggests that this femtosecond energy transfer occurs ∼3 times faster to adsorbates at step sites compared to adsorbates located on atomically flat surface regions. Density functional theory calculations of vibrational damping rates by electron-hole pair excitations corroborate the increased rate of energy transfer at step sites and show that it is caused primarily by an increased local density of adsorbate-induced states around the Fermi level.

One of the key goals in surface science is to provide a fundamental, atomic-level understanding of catalytic surface reactivity. This requires insight into molecular length and time scales and relating the reactivity to the local electronic and geometric structure. Real-time studies of surface reactivity have been enabled through the application of ultrafast spectroscopic techniques to molecules on well-defined single-crystal surfaces in ultrahigh vacuum (UHV).1-5 By exciting the surface with an ultrashort laser pulse and monitoring, as a function of time, the subsequent excitation and reaction dynamics, these studies have provided important information on, for example, the rates and pathways of energy flow into reaction coordinates. One key question that remains to be explored, however, is how the chemical dynamics are affected by surface defects, despite the well-established fact that defects, such as steps and missing atoms, play an essential role in heterogeneous catalysis.6-15 The ultrafast dynamics of reactions occurring at such sites can provide insight into important, and technologically relevant, elementary processes on these sites. Here, we present the first femtosecond study of reactivity on a surface with well-defined defects. We elucidate the efficiency of energy flow between substrate and adsorbate for step and terrace sites by studying the desorption of nitrogen monoxide (NO) and carbon monoxide (CO) from the Pt(533) surface (see the inset of Figure 1a), consisting of atomically flat, four-atomwide (111) terraces divided by monatomic (100) steps. The dynamics occurring after substrate excitation are monitored by detecting ultrafast desorption events.16,17 Our results reveal marked variations in the coupling of the laser-heated electrons to the adsorbate with respect to the adsorption site. Although the different adsorption sites are separated by only a few angstroms, an analysis based on a simple electron friction model * Corresponding author. E-mail: [email protected]. † Leiden University. ‡ The University of Liverpool. § FOM Institute for Atomic and Molecular Physics.

10.1021/jp070470g CCC: $37.00

Figure 1. (a) Time-of-flight spectrum for NO desorbing from a fully covered Pt(533) surface (gray) and from a surface on which NO is adsorbed only on the step sites (black, multiplied by 10). The macroscopic normal of the crystal was facing the mass spectrometer. A Maxwell-Boltzmann fit (dashed black line) reveals a high translational energy (∼3500 K, fluence: 100 J/m2) of desorbed NO for both the steps and terraces. The inset shows schematically NO bound atop on the step and terrace sites of the stepped Pt(533) surface, consisting of four-atom-wide (111) terraces separated by monatomic (100) steps. (b) Time-of-flight spectrum for NO desorbing from a surface of which only the step sites are covered with NO for two different collection angles shown in the inset: along the step local normal (40°, gray) and its symmetric opposite (-40°, black). The relatively large yield along the step local normal (40°) indicates that NO adsorbed on the step sites preferentially desorbs along the local step normal. All spectra are averages of the first 150 shots.

suggests that the rate of energy transfer between the substrate electrons and the adsorbate is larger by a factor of 3 at the steps than on the terraces. Density functional theory (DFT) calcula© 2007 American Chemical Society

6150 J. Phys. Chem. C, Vol. 111, No. 17, 2007 tions of vibrational damping rates by electron-hole pair excitations agree with these results in that they show that vibrational damping rates are also enhanced at the step sites relative to terrace sites and that this enhancement is caused primarily by an increased density of electron states around the Fermi level at the step site compared to a terrace site. The experiments were performed in a UHV chamber18 with a base pressure of 2 × 10-11 mbar in combination with a Ti: sapphire femtosecond laser system (Quantronix, Titan) producing 800 nm pulses of 3 mJ (1 kHz repetition rate) with a pulse duration of 130 fs (full width at half-maximum (fwhm)). For the desorption measurements, 2 mJ of the laser was used to irradiate the sample at a 35° angle of incidence with a fluence, defined throughout this paper as absorbed fluence, below the damage threshold (thermal desorption spectra before and after irradiation are identical). Experiments were typically singlepulse, enabled through the combination of a mechanical chopper and shutter. Desorbed molecules were detected as a function of the flight time by a differentially pumped mass spectrometer (Balzers QS422) amplified with a fast amplifier and counted with a multichannel scaler after a 70 mm flight path along the surface normal. The desorption yield was obtained by integrating such a time-of-flight (TOF) spectrum (see Figure 1). All experiments were performed at ∼100 K, at a background pressure lower than 2 × 10-10 mbar. To obtain saturation coverage, the Pt(533) single crystal was dosed with NO (15N18O, unless otherwise stated) or CO from background gas with the crystal at ∼100 K. To selectively cover the step sites with NO or CO, we simply dosed at elevated temperatures (400 K), making use of the higher binding energy for NO and CO on the step sites (NO: Ebind,terrace ) 1.3 eV; NO: Ebind,step ) 2 eV; CO: Ebind,terrace ) 1.5 eV, Ebind,step ) 2 eV18,19). Thermal desorption spectroscopy (TDS) confirmed that in this manner only the step sites were covered (see the insets in Figure 2). One key quality of femtosecond laser-driven surface chemistry is the temporal separation of reactions driven by electrons and phonons, respectively.20 The laser pulse first heats the electrons near the metal surface, which reach very high temperatures (thousands of K) because of their low heat capacity, after which electron-phonon equilibration (to hundreds of K) occurs on longer time scales. The very high translational temperatures of both NO and CO (exceeding 3000 K at 100 J/m2 fluence) observed in the TOF (Figure 1a for NO) indicate that desorption was induced by laser-heated electrons. The signals were much weaker (by a factor of 60 [NO] and 30 [CO]) when only the steps were covered than for the fully covered surface. These large reductions of the signals are remarkable in view of the number of adsorbed molecules on steps being lower by only factors of 4 (NO) and 3 (CO). Part of this large difference can be accounted for by the observation that molecules on steps do not desorb along directions around the macroscopic [533] surface normal. This is illustrated in Figure 1b, which shows TOF spectra for NO on the step sites, in the two geometries indicated in the figure (with the crystal rotated along the axis parallel to the monatomic steps). As the crystal angle was varied symmetrically around the incident laser beam, the laser fluence remained unchanged, allowing for a direct comparison of yields. Clearly, NO desorption from steps is biased to the local [100] surface normal, apparently along the tilted molecular axis.18 All other experimental data described here were measured along the macroscopic normal of the crystal. Because the signal was 30-60 times higher at saturation coverage than for the selective occupation of the steps, we could use the fully covered surface to obtain information on the

Letters

Figure 2. (a and b) Depletion of the surface after multipulse excitation for NO and CO, respectively, on terrace and step sites (normalized to the first shot), illustrating the faster depletion of terraces sites. Solid lines are exponential fits. The insets show TDS spectra of a fully saturated surface and of one in which only the step sites are covered for NO (four peaks: three from terrace sites and one at high temperature from steps18) and CO (one terrace and one step peak), respectively. (c) Two-pulse (40 J/m2 each) correlation traces of NO desorbing from terrace (gray points) and step sites (multiplied by 60, black points), both indicating a time response of 0.8 ps (fwhm). The lines (multiplied by 5 for step sites) are obtained with the friction model (see the text) with an electron-coupling time of 0.075 ps for the terraces and 0.025 ps for the steps.

desorption dynamics from the (111) terraces of the stepped Pt surface. By only covering the step sites, the same information about the (100) steps was obtained. The different angular distribution for step and terrace NO accounts for a factor of ∼3 in the difference in desorption probability of steps and terraces (interpolating the data in Figure 1b to 0°21). Combined with the lower amount of NO on the steps (4×), we can conclude that the desorption probability for NO on steps is lower by a factor 5 ( 2 compared to terrace sites. This value was corroborated by an independent measurement of the relative desorption probability through the decay of the depletion curves (Figure 2a) resulting in a factor of 3 ( 1. For CO, the depletion curves depicted in Figure 2b indicate an even smaller difference in desorption probability from steps and terraces, namely, a factor of 2 ( 1 lower desorption probability for the step sites.22 The factor of 2-3 difference, obtained directly from independent measurements in Figure 1 and 2, between the desorption probability from terrace and step sites for both NO and CO, is much smaller than what is expected from the large difference in binding energy between step and terrace sites. Under equilibrium conditions at T ) 3000 K (the kinetic temperature of desorbing molecules), the difference would be a factor of ∼20 for both CO and NO using a simple Arrhenius expression. The relatively strong enhancement of desorption from step sites compared to terrace sites suggests that the energy transfer rate from the laser-heated electrons in the Pt substrate to the adsorbate is much more efficient for the step sites. The rate of energy transfer from the surface into the adsorbate can be quantified by analyzing a two-pulse correlation (2PC) spectrum using a phenomenological two-temperature and one-

Letters

Figure 3. (a) Dependence of the first shot yield (FSY) for NO desorbing from Pt(533) for saturation coverage on the absorbed yield weighted fluence (). Every point is an average of four measurements of typically 10 sequential laser shots each on a fresh spot on the crystal. The FSY is obtained from the amplitude of an exponential fit. The spatial profile of the laser beam is explicitly taken into account by calculating the yield weighted fluence in the manner described in refs 23 and 38. This procedure yields a power law exponent of 4 ( 0.5 (gray dotted line). (b) Two-pulse (20 J/m2 each) correlation of NO desorbing from Pt(533) for saturation coverage, indicating a time response of 0.8 ps (fwhm). Every point is an 8× averaged first shot yield obtained from an exponential fit through the depletion curve. (c) 14N16O and 15N18O laser-induced desorption yield (4× average) as a function of the fraction of 15N18O on the surface. The surface fraction is obtained from TDS of the residual NO (>90%) after the laser experiments. In this mixed-isotopomers experiment, 15N18O and 14N16O are first mixed in the desired ratio in the vacuum chamber with the crystal at 500 K (where adsorption does not occur). The crystal was subsequently cooled down, resulting in a homogeneous distribution of both isotopomers over the crystal as confirmed by TDS.39 (d) Translational temperature as a function of the delay between two laser pulses (see panel b). The lines in all panels are obtained within a friction model calculation (see the text) using an electron-coupling time of 0.075 ps. Typical error bars are depicted in each panel.

dimensional electronic friction model in the low-friction limit: 23,24 For the 2PC measurement, the excitation pulse was split into two equal parts, which excited the sample with a variable time delay. The yield as a function of pulse-pulse delay is determined by the rate of energy transfer, in addition to independently determined variables such as laser characteristics and binding energy. Figure 2c shows the 2PC spectrum for NO on the step and terrace sites. These spectra have a comparable time-scale for decay, but the baseline is appreciably lower for the step sites compared to the terrace sites. To obtain accurate values for the electron-coupling times, we also measured the fluence dependence (Figure 3a), the isotope effect (Figure 3c), and the translational temperature (Figure 3d) as a function of the delay between the two pump pulses in a 2PC experiment for NO on the terrace sites (Figure 3b). The complete data set for NO desorption from terrace sites can be described well with +0.025 the friction model using a coupling time of 0.075-0.045 ps.25 The simple model reproduces the data remarkably well: in addition to the fluence dependence and the 2PC trace, the kinetic temperatures are reproduced, as well as the ∼15% higher desorption probability for 14N16O compared to 15N18O. The lines in the figure are the results of the model calculation. Because the most detailed information about the coupling time comes from the 2PC spectrum, for NO on the step sites only this data has been measured. On the basis of this 2PC spectrum and the lower desorption probability, the friction model gives a 3 times smaller electron-coupling time (0.025 ps) for the step sites. Although here the absolute error is also significant, the

J. Phys. Chem. C, Vol. 111, No. 17, 2007 6151 relative value (compared to terrace NO) can be determined reliably. A similar analysis of the CO data reveals an electroncoupling time of 0.3 and 0.1 ps for CO adsorbed on the terraces and the steps, respectively. Thus, for both CO and NO the electron coupling time is 3 times faster at the step sites than at the terrace sites.26 The flow of energy from laser-heated substrate electrons into the adsorbate results in the excitation of adsorbate vibrational modes. For desorption of diatomics from various surfaces, it has been suggested that excitation of the frustrated rotation is responsible for the desorption process.27-29 Our results are consistent with those observations: We cannot, strictly speaking, dismiss (a contribution from) the Pt-CO stretch vibration, but the frustrated translational mode can be excluded based on the independently determined electron-coupling times found to be 2.5 and 4 ps for terrace- and step-adsorbed molecules, respectively.19 To rationalize the faster electron-coupling times of adsorbates at the step sites than at the terrace sites, we have carried out DFT calculations of damping rates of the frustrated rotational modes for step- and terrace-adsorbed molecules due to electronhole excitations. Note that a full calculation of the laser-induced desorption is a daunting task and requires at least a treatment of the mode-mode coupling and calculations of electronic friction away from the equilibrium configuration and at high electronic temperatures.30 Thus, the vibrational lifetimes, as obtained from the inverse of the calculated damping rate, cannot be compared directly to the electron coupling times extracted from the experiments using a one-dimensional electronic friction model but should provide a useful measure of the relative efficiency of the energy transfer pathways resulting in desorption.31 The first step in the DFT calculations of the vibrational damping rates is calculations of the geometric and electronic structures of the adsorbed molecules using a supercell geometry. This was carried out in the same manner as reported previously.18,19 In the second step, the damping rates were obtained from the calculated Kohn-Sham states, ψυ, and energies, υ, using a Fermi Golden-rule like formula32,33

γ)

2π p M

|〈ψυ′|υ′|ψυ〉|2 δ(F - υ′) δ(F - υ) ∑ υ,υ′

(1)

where F is the Fermi energy and M is the reduced mass of the vibrational mode with coordinate Q. The electron-vibration coupling υ′(r) ) (∂υ(r;Q)/∂Q)|Q)0(r) is obtained from the Kohn-Sham potential υ(r;Q) as a function of Q. For details about applying eq 1 to a supercell with slab geometry, see ref 33. The adsorbate vibrational modes and energies for a rigid substrate lattice (Table 1) were obtained through a diagonalization of the dynamical matrix for the adsorbate. We constructed the dynamical matrix from the calculated forces, obtained by finite displacements of the adsorbate atoms in three mutually orthogonal directions. The calculated vibrational lifetimes τ ) 1/γ, as shown in Table 1, agree qualitatively with the experimentally observed trend that the electron-coupling times are lower for step-adsorbed molecules than for terrace-adsorbed molecules. However, we note that the calculated magnitude of τ and the electron-coupling times differ quantitatively. As mentioned above, there may be several reasons for this discrepancy that calls for further studies. To explain the observed behavior of γ, we have related it to the adsorbate-induced electronic structure. We have developed an approximation to γ, involving the projected density of states

6152 J. Phys. Chem. C, Vol. 111, No. 17, 2007

Letters

TABLE 1: Calculated Vibrational Energies pΩ and Damping Rates for the Frustrated Rotational (FR) Modes of NO and CO Adsorbed at a Step and Terrace Site on a Pt(533) Surface a adsorbate

pΩ γ γlocal γ5σ-2π* γ4σ-2π* τ ) 1/γ mode (cm-1) (ps-1) (ps-1) (ps-1) (ps-1) (ps)

NO (step)

FR(x) FR(y) NO (terrace) FR(x) FR(y) CO (step) FR(x) FR(y) CO (terrace) FR(x) FR(y)

514 306 105 289 381 396 388 392

0.38 0.34 0.030 0.25 0.35 0.32 0.27 0.18

0.32 0.28 0.024 0.21 0.44 0.33 0.36 0.30

0.52 0.49 0.039 0.087 0.46 0.35 0.37 0.32

0.050 0.033 0.016 0.00 0.00 0.00 0.00 0.00

2.6 2.9 33 4.0 2.9 3.1 3.7 5.6

a The rate γ has been calculated with the full Kohn-Sham wave function as basis set and γlocal using a truncated local basis set based on the valence MOs of an isolated NO or CO molecule. Also, the dominant sums of the individual contributions from intra- and interorbital terms in γlocal are given (see the text). x and y denote the coordinate axes perpendicular to the [533] (z) direction, where x is parallel to the direction of the step. For the modes FR(x) and FR(y) of step-adsorbed NO and FR(y) of terrace-adsorbed NO, there are relatively large contributions (∼50% of γ5σ-2π*) to γlocal from terms similar to the one in eq 5, which involve an additional MO.

(DOS) on the molecular orbitals (MO) of the isolated adsorbate. This approximation is based on the fact that the efficient screening of the field from the vibrating atoms by the conduction electrons limits the spatial range of υ′(r) to the vicinity of the molecule. An expansion in a truncated localized basis set, such as the valence MOs |φn〉 of the isolated CO or NO molecule, is then meaningful:34

υ′ ≈

|φn〉υ′nm〈φm| ∑ n,m

(2)

where υ′nm ) 〈φn|υ′|φm〉. Inserting eq 2 in eq 1 gives the damping rate

γlocal )

2π p M

∑

Fnm(F) υ′nkFkl(F)υ′nm

(3)

n,m,k,l

where Fnm() are the one-electron density matrix elements

Fnm() )

∑V 〈φn|ψV〉〈ψV|φm〉 δ( - V)

(4)

Note that the diagonal element Fn(F) ≡ Fnn(F) is nothing else than the DOS projected on the MO |φn〉 at the Fermi level and is influenced by the chemical bonding to the surface. The specific details of the vibrational mode and the electronic nonadiabatic coupling enter only through the electron-vibration coupling matrix elements υ′nm, which are obtained from the first-order change of the first-order moment of Fnn() with respect to vibrational amplitude. As shown in Table 1, the calculated γlocal using the valence MOs of an isolated NO or CO molecule are in near-quantitative agreement with the calculated γ. Thus, the approximate expression in eq 3 provides a means to analyze the trends of the calculated γ in terms of individual, dominant intra- and interorbital contributions to γlocal shown in Table 1. For instance, if the sum in eq 3 is dominated by a contribution from a single orbital, then the Persson-Persson result for γ is recovered.35 Such an analysis shows that the larger γ at the step relative to the terrace sites are caused primarily by the enhanced projected DOS on the 5σ orbital at the Fermi level (F5σ(F)) at the step sites relative to the terrace sites. As shown in Table 1, the dominant orbital terms to γlocal involve the 5σ orbital and

the twofold degenerated 2π* antibonding orbital and are given by

γ5σ-2π* ∝ F2π*(F) F5σ(F)υ′2π*5συ′5σ2π*

(5)

Here we find that the increase of these terms between the step sites and the terrace sites is governed by F5σ(F). Note that the one-electron density matrix elements only change when changing adsorption site. So the large difference in the damping rates between the two vibrational modes of NO on a terrace site can only be caused by differences between the electron-vibration coupling matrix elements. According to the simple d-band model36 for chemisorption, the enhancement of F5σ(F) for step adsorption relative to terrace adsorption has the same origin as the larger chemisorption energy for step adsorption. The center of the local d-band center shifts up in energy at step atoms compared to terrace atoms,15 resulting in a stronger covalent interaction between the metal and the molecule. In particular, the enhancement of F5σ(F) is caused by the increase of the DOS with an antibonding character between the d states and the 5σ orbital around F when the center of the d-band comes closer to F.37 It should be noted that the observed correlation between strong binding and effective coupling to the substrate electrons is not a generic phenomenon. For CO on Ru(0001), the adsorbate binding energy is comparable to NO/Pt(533), but the electronic transfer rate is negligible23sorders of magnitude different from NO/Pt(533). In conclusion, using femtosecond surface spectroscopy and DFT calculations, we have shown that the energy transfer between adsorbates and metal electrons occurs appreciably more efficient on step sites than on the atomically flat surface. Calculations of vibrational damping rates by excitation of electron-hole pairs suggest that the energy transfer at step sites is enhanced primarily by an increased local density of adsorbateinduced states around the Fermi level. The more efficient energy transfer at the step sites is in accord with the generally observed higher reactivity of these sites in surface reactions. Acknowledgment. We thank A. Eichler for providing us the optimized structures. This work is part of the research program of the “Stichting voor Fundamenteel Onderzoek der Materie (FOM)”, which is financially supported by the “Nederlandse organisatie voor Wetenschappelijk Onderzoek (NWO)”. Financial support from the EPSRC and allocations of computer resources by the Swedish National Allocations Committee (SNAC) and the University of Liverpool are also gratefully acknowledged. References and Notes (1) Petek, H.; Weida, M. J.; Nagano, H.; Ogawa, S. Science 2000, 288, 1402. (2) Bauer, M.; Lei, C.; Read, K.; Tobey, R.; Gland, J.; Murnane, M. M.; Kapteyn, H. C. Phys. ReV. Lett. 2001, 87, 025501. (3) Ste´pa´n, K.; Gu¨dde, J.; Ho¨fer, U. Phys. ReV. Lett. 2005, 94, 236103. (4) Frischkorn, C.; Wolf, M. Chem. ReV. 2006, 106, 4207. (5) Lane, I. M.; King, D. A.; Liu, Z. P.; Arnolds, H. Phys. ReV. Lett. 2006, 97, 186105. (6) Zambelli, T.; Wintterlin, J.; Trost, J.; Ertl, G. Science 1996, 273, 1688. (7) Dahl, S.; Logadottir, A.; Egeberg, R. C.; Larsen, J. H.; Chorkendorff, I.; Tornqvist, E.; Nørskov, J. K. Phys. ReV. Lett. 1999, 83, 1814. (8) Gambardella, P.; Sˇljivancanin, Z.; Hammer, B.; Blanc, M.; Kuhnke, K.; Kern, K. Phys. ReV. Lett. 2001, 87, 056103. (9) Liu, Z. P.; Hu, P. J. Am. Chem. Soc. 2003, 125, 1958. (10) Zubkov, T.; Morgan, G. A.; Yates, J. T.; Ku¨hlert, O.; Lisowski, M.; Schillinger, R.; Fick, D.; Ja¨nsch, H. J. Surf. Sci. 2003, 526, 57. (11) Honkala, K.; Hellman, A.; Remediakis, I. N.; Logadottir, A.; Carlsson, A.; Dahl, S.; Christensen, C. H.; Nørskov, J. K. Science 2005, 307, 555.

Letters (12) Kim, Y. K.; Morgan, G. A.; Yates, J. T. Surf. Sci. 2005, 598, 14. (13) Vang, R. T.; Honkala, K.; Dahl, S.; Vestergaard, E. K.; Schnadt, J.; Laegsgaard, E.; Clausen, B. S.; Nørskov, J. K.; Besenbacher, F. Nat. Mater. 2005, 4, 160. (14) Buatier de Mongeot, F.; Toma, A.; Molle, A.; Lizzit, S.; Petaccia, L.; Baraldi, A. Phys. ReV. Lett. 2006, 97, 056103. (15) Hammer, B. Top. Catal. 2006, 37, 3. (16) Bonn, M.; Funk, S.; Hess, C.; Denzler, D. N.; Stampfl, C.; Scheffler, M.; Wolf, M.; Ertl, G. Science 1999, 285, 1042. (17) Denzler, D. N.; Frischkorn, C.; Hess, C.; Wolf, M.; Ertl, G. Phys. ReV. Lett. 2003, 91, 226102. (18) Backus, E. H. G.; Eichler, A.; Grecea, M. L.; Kleyn, A. W.; Bonn, M. J. Chem. Phys. 2004, 121, 7946. (19) Backus, E. H. G.; Eichler, A.; Kleyn, A. W.; Bonn, M. Science 2005, 310, 1790. (20) Cavanagh, R. R.; King, D. S.; Stephenson, J. C.; Heinz, T. F. J. Phys. Chem. 1993, 97, 786. (21) Because our experimental geometry does not allow us to measure the angular distribution, we use cos 4q (based on M. Wilde et al., Surf. Sci. 1999, 27, 427-428 and references therein) to estimate the yield at 0°. (22) The relative desorption probability for CO on steps obtained in this way constitutes a lower limit because laser-induced motion of CO molecules from step to terrace sites is known to occur.19 Hence, for the step-covered surface, there is a small additional contribution from CO molecules that have not diffused back to the steps, after being driven to terrace sites by the initial laser shot. Moreover, desorption of molecules from the step sites via the terrace sites directly after excitation could also play a role. From the different angle dependence of the desorption probability for step and terrace sites (Figure 1), it can be concluded that this process alone can not explain the data, but it certainly can not be excluded. (23) Funk, S.; Bonn, M.; Denzler, D. N.; Hess, C.; Wolf, M.; Ertl, G. J. Chem. Phys. 2000, 112, 9888. (24) Brandbyge, M.; Hedegård, P.; Heinz, T. F.; Misewich, J. A.; Newns, D. M. Phys. ReV. B 1995, 52, 6042.

J. Phys. Chem. C, Vol. 111, No. 17, 2007 6153 (25) For the strong coupling observed here, the reaction rate follows the electronic transient temperature quite closely, resulting in a relative large error in the coupling time. (26) The extracted electron coupling times for NO are smaller than the inverse mode frequency of the low-frequency modes. This is unphysical because energy transfer into the low-frequency modes cannot occur faster than the motion associated with the modes. Although the absolute values for the friction coefficient obtained with this simple one-dimensional friction model may have limited meaning, the relative difference between the step and terrace coefficient clearly indicates a ∼3-fold stronger coupling of the laser-heated electrons to the adsorbate at the steps relative to the terraces. (27) Bonn, M.; Hess, C.; Funk, S.; Miners, J. H.; Persson, B. N. J.; Wolf, M.; Ertl, G. Phys. ReV. Lett. 2000, 84, 4653. (28) Fournier, F.; Zheng, W.; Carrez, S.; Dubost, H.; Bourguignon, B. J. Chem. Phys. 2004, 121, 4839. (29) Springer, C.; Head-Gordon, M.; Tully, J. C. Surf. Sci. 1994, 320, L57. (30) Tully, J. C.; Gomez, M. J. Vac. Sci. Technol., A 1993, 11, 1914. (31) Luntz, A. C.; Persson, M.; Wagner, S.; Frischkorn, C.; Wolf, M. J. Chem. Phys. 2006, 124, 244702. (32) Hellsing, B.; Persson, M. Phys. Scr. 1984, 29, 360. (33) Lorente, N.; Persson, M. Faraday Discuss. 2000, 117, 277. (34) Hassel, M.; Persson, M.; Forsblom, M.; Bird, D.; Holloway, S., in press. (35) Persson, B. N. J.; Persson, M. Solid State Commun. 1980, 36, 175. (36) Hammer, B.; Nørskov, J. K. Nature 1995, 376, 238. (37) Hammer, B.; Nielsen, O. H.; Nørskov, J. K. Catal. Lett. 1997, 46, 31. (38) Struck, L. M.; Richter, L. J.; Buntin, S. A.; Cavanagh, R. R.; Stephenson, J. C. Phys. ReV. Lett. 1996, 77, 4576. (39) Although the adsorption structure is slightly different from dosing directly at 85 K,18 the first shot yield and depletion curve are indistinguishable.