Coverage dependence of the sticking probability of ethylene on Ag(410)

Transcription

1 Surface Science 587 (2005) Coverage dependence of the sticking probability of ethylene on Ag(410) L. Savio, L. Vattuone, M. Rocca * IMEM-C.N.R, sezione di Genova, INFM, and Dipartimento di Fisica, via Dodecaneso 33, I Genova, Italy Available online 1 June 2005 Abstract The interaction of ethylene with Ag(410), a vicinal surface of Ag(100) characterised by a high density of open steps, is investigated. Energy and angle resolved measurements of the sticking coefficient are performed with the retarded reflector method of King and Wells using a supersonic molecular beam. We find that open steps remove the translational barrier for adsorption into the p-bonded state present for Ag(1 0 0). Steering removes the angle dependence of the initial sticking probability. The coverage dependence of the adsorption probability indicates that a precursor mechanism is present and that adsorbate assisted adsorption is important at hyperthermal energies. With increasing coverage steering becomes ineffective and rotations inhibit adsorption. Admolecules with second nearest neighbours are destabilised and these sites can be populated only by gas-phase molecules impinging along trajectories for which the step heights are exposed. Ó 2005 Elsevier B.V. All rights reserved. Keywords: Defects; Ag; Ethylene; King and Wells 1. Introduction In the study of catalytic processes in the heterogeneous phase the importance of defects in overcoming the structure gap between industrial reactors and laboratory, ultra high vacuum conditions has been debated since the early days of surface science [1 3]. In spite of that, the role of well * Corresponding author. address: rocca@fisica.unige.it (M. Rocca). defined defects was investigated thoroughly only recently. For Ru(0 001) it was shown, both by theory and experiment, that atomic steps are the only active sites for NO and N 2 bond dissociation [4 6]; for N 2, steps were indeed found to be 10 9 times more reactive than terraces [6]. Similarly, it was demonstrated by scanning tunnelling microscopy that oxygen dissociation on Pt(111) occurs at the upper side of closed packed steps [7]. The structure gap was attacked in spectroscopic investigations by artificially introducing defects at /$ - see front matter Ó 2005 Elsevier B.V. All rights reserved. doi: /j.susc

2 L. Savio et al. / Surface Science 587 (2005) low Miller index surfaces by ion bombardment. Such a technique was successfully applied to the study of the NH 3 dissociation rate on Ru(0 001), which increases for the damaged surface [8], and of the oxygen interaction with Ag(100) [9]. In this latter case dissociation occurs in the presence of defects also at a crystal temperature T = 105 K, at which only admolecules are stable on the flat surface. When indications about the nature of the active site are available, an alternative approach to the problem consists in the investigation of particular high Miller index surfaces characterised by a high density of the defect of interest. Studies in this direction have been performed, e.g., for O 2 /Pt(533), showing that (100) steps are responsible for the conversion from the physisorbed into the chemisorbed precursor on the, otherwise inert, Pt(111) planes [10]. For O 2 / Ag(4 10) [11] and O 2 /Ag(2 10) [12] we found that the reactivity of terrace sites is affected by the limited terrace width and that dissociation occurs at the open steps. The interest in ethylene adsorption on metal surfaces is mainly due to the role played by this molecule as a test species for the investigation of the chemistry of hydrocarbons, a field which attracts much attention because of the many technological applications connected to it. The C 2 H 4 /Ag system is particularly intriguing since silver powders are the only catalyst for the ethylene epoxidation reaction. Although experiments performed in ultra high vacuum failed so far in the production of ethylene epoxide, they lead to a quite complete characterisation of the interaction of C 2 H 4 with the different high symmetry Ag faces [13]. In spite of that, the ethylene reactivity at defect sites was relatively poorly studied. Microcalorimetric measurements showed that C 2 H 4 initially adsorbs on Pt(2 1 1) as quad-r-acetylene and, at higher coverage, as ethylidyne (C CH 3 ); on the contrary on the more corrugated Pt(3 1 1) surface only ethylylidyne (C CH 2 ) forms [14]. Our previous investigation on Ag(1 0 0) showed that at T = 105 K ethylene adsorbs in different states, depending on the translational and vibrational energy of the gas-phase molecules and leading to a metastable physisorbed state [15], to a translationally activated p-bonded state [16,17], and to a more strongly bound, but still undissociated state, accessed only by vibrationally excited molecules [16]. Recent local density approximation (LDA) calculations [18] do not find, however, stable chemisorption for C 2 H 4 /Ag(1 00), suggesting that the observed states are related to adsorption at defects, as in the case of C 2 H 4 /Al [19]. The unexpected barrier for ethylene adsorption into the p- bonded state on Ag(100) could thus be linked to the energy required for the creation of the relevant defect. Indeed LDA calculations find that ethylene chemisorption occurs only at isolated adatoms [20]. Here we report on a complete investigation on the ethylene interaction with Ag(4 1 0). Angle and energy resolved sticking measurements are performed by means of a supersonic molecular beam. We find that open steps remove the translational barrier to adsorption into the p-bonded state. Angle and temperature dependence of the sticking probability indicate that a precursor mechanism is operative. In the low coverage limit clear evidence is present for strong steering effects in the chemisorption process and for an adsorbate-assisted adsorption mechanism for ethylene physisorption. Contrary to ethylene/ag(1 0 0) [15] rotations play little or no role in the adsorption process. At 110 K the admolecules are destabilised with increasing coverage, becoming unstable above half occupancy of the step sites ( 1 ML, 1 ML = atoms cm 2 ). Access into the high coverage sites depends on the angle at which the gas-phase molecules hit the surface and is inhibited when the step heights are not exposed. The angular dependence of sticking changes with increasing coverage, indicating that for more weakly bound molecules steering becomes ineffective and rotations inhibit adsorption. Steering phenomena in the initial sticking probability were presented in [21]. 2. Experimental Experiments were performed in the ultra high vacuum (UHV) system described in detail in [22], which combines a high resolution electron energy loss spectrometer (HREELS) and a supersonic

3 112 L. Savio et al. / Surface Science 587 (2005) <140> <001> θ=31 1 st molecular beam. A quadrupole mass spectrometer (QMS) not in line of sight with the beam allows us to measure the sticking coefficient using the retarded reflector method of King and Wells (KW in the following [23]), while the final adsorption state of ethylene molecules was monitored by vibrational spectroscopy in preliminary experiments not reported here [24]. The supersonic molecular beam is collimated to a spot diameter of 2 mm at the crystal for KW measurements. The translational energy of the impinging molecules,, is measured by time of flight and is varied from 0.10 ev to 3 ev both by seeding C 2 H 4 in helium (4% of C 2 H 4 ) and by heating the ceramic nozzle up to T N = 870 K. The energy resolution is ±10%. The flux of the pure beam is measured with a spinning rotor gauge and reads, for the present experiments, 94 ML/s when the nozzle is kept at room temperature. The flux for the seeded beam is obtained by comparing the partial pressure rise recorded by QMS in the main chamber with that due to the pure beam and reads 47 ML/s. For isothermal desorption our experimental setup allows detecting molecules leaving the crystal with lifetimes between 0.3 and 2 s. The lower limit is due to the time constant of the vacuum system, while molecules with large lifetimes yield a flux too small to be detected with the present signal to noise ratio. The Ag(410) crystal is a disk of 7 mm diameter, oriented within 1 of the (410) plane. Before each experiment the surface was prepared by sputtering with Ne + ions and annealing to T = 700 K, until a sharp low energy electron diffraction (LEED) pattern is observed. Such a procedure led to a clean spectrum in preliminary HREELS experiments [24]. The (4 1 0) geometry is shown schematically in Fig. 1: the surface is formed by three-atomrow wide (1 0 0)-like terraces and monoatomic (1 1 0)-like steps. The step edge consists therefore of a zig-zag chain of atoms. This picture is a representation of the ideal surface, which could be quite unlike the real Ag(410) face morphology [25]. In fact the (410) surface is expected to be more defective than a low Miller index plane because the open steps are prone to roughen at relatively low temperatures, forming (1 1 1) nanofacets. No significant thermally induced surface disorder is observed at T = 110 K, at which temperature most of the present experiments were performed, as demonstrated in [11]. The angle of incidence of the impinging molecules with respect to the normal to the surface is designed by h i. In all experiments the scattering plane is defined by the normal to the surface and the h 1; 4; 0i direction (across the steps). Since the surface forms an angle of 14 with the (1 0 0) terraces, h i = h t = 14 and h i = h s = +31 correspond to molecules impinging normally to the (100) terraces and to the step heights, respectively. 3. Data presentation θ=-14 2 nd 3 rd 4 th 14 Side view Top view Fig. 1. Geometry of the Ag(410) surface. The scattering plane is aligned along the h 1,4,0i direction (across the monoatomic steps). Positive h i values correspond to molecules impinging against the step heights. The primitive unit cell is also drawn. Fig. 2 shows QMS traces of the normalised C 2 H 4 partial pressure during KW experiments performed on Ag(410) at T = 110 K, with h i = h s and = 0.10 ev and = 0.36 ev. At time t = t 0 the molecular beam enters the UHV chamber and a sudden increase in the partial pressure takes place. The beam does not strike the Ag sample since it is still intercepted by an inert flag placed in front of it. Since the background pressure remains very low ( mbar), no significant adsorption occurs in these conditions. At t = t 1 the flag is re-

4 L. Savio et al. / Surface Science 587 (2005) C 2 H 4 norm. partial press S S Smax 0 t 0 t =0.10 ev moved and the beam hits the sample. The drop of the C 2 H 4 partial pressure is due to the pumping action of the surface and gives a direct measurement of the total sticking probability, S. The missing area in the normalised partial pressure curve multiplied by the flux gives the surface coverage, H. S(t) can thus be converted by integration into S(H). Its initial value, denoted with S 0, is particularly important since it corresponds to the interaction with the bare surface. As it is apparent in Fig. 2, S 0 is smaller at larger, as expected for non-kinetic energy activated adsorption. S(t) is initially nearly constant for = 0.10 ev, while for = 0.36 ev it increases with exposure up to a maximum value S max. In both cases it drops rapidly to zero when a maximum coverage is achieved. This condition is dynamical since, when at t = t 2 the exposure is stopped, the QMS signal increases because of desorption from a metastable state, the population of which is no longer counterbalanced by adsorption. The area of the desorbed flux is smaller than the one of the uptake curve, indicating that the major part of the adsorbed molecules t 2 =0.36 ev 60 t 3 Time (sec) % C 2 H 4 T=110 K θ i =31 o Fig. 2. QMS traces of the normalised C 2 H 4 partial pressure during two uptake experiments performed on the Ag(410) surface at T = 110 K. The beam impinges normally onto the steps heights (h i =31 ) with = 0.10 ev (upper trace) and = 0.36 ev (lower trace). The upper trace is shifted in the normalised partial pressure coordinate by 1.2 units. S denotes the sticking probability, S 0 and S max correspond to the initial and maximum S value. Times t 0, t 1, t 2 and t 3 are respectively, the moments when the beam enters the chamber, when it first strikes the sample, when it is intercepted and when it hits the sample for the second time. are stable at T = 110 K, i.e. that they are chemisorbed. When exposing the crystal to the C 2 H 4 beam for a second time (t 3 ), the shape of the QMS trace is radically different: S(t) reaches immediately its maximum value and drops eventually rapidly for both energies. Further exposures yield QMS signals identical to the second one, proving that the additional admolecules are in the metastable state. The active sites for chemisorption have thus been saturated already during the first exposure. In Fig. 3 we report H(t) calculated from integration of the lower trace of Fig. 2. It increases initially up to 0.13 ML and drops to 0.11 ML when the beam is stopped and desorption off the metastable state sets in. The latter value, corresponding to the saturation of the stable species, is close to one half occupancy of the step sites; this is indicative that admolecules repel each other and inhibit occupation of the nearest neighbours. No ordered superstructures are however observed by low energy electron diffraction. In Fig. 4 we show S(H) att = 110 K for two beam energies and for different h i.at = 0.10 ev (left panel) two regimes can be identified: S(H) is nearly constant up to a critical coverage, then it drops linearly to values below the KW sensitivity. When the step heights are illuminated by the beam (h i > 59, all curves except dotted one) the critical coverage reads 0.10 ML, while the maximum Θ (ML) Exposure (ML) t 1 40 t 2 t Time (sec) Pure C 2 H 4 =0.36 ev T=110 K θ i =31 o Fig. 3. Total coverage, H, vs. exposure time calculated for the high energy QMS trace of Fig. 2. The exposure (in ML of Ag(410)) during the first shot is reported on the upper axis.

5 114 L. Savio et al. / Surface Science 587 (2005) Fig. 4. Coverage dependence of the sticking probability for several h i at low (left panel) and high (right panel) energy. coverage is between 0.14 and 0.19 ML depending on the different population of the metastable species [27]. For h i > 59 (dotted curve), on the contrary, the critical coverage reads only 5 ML, indicating a lower reactivity of the surface if step heights are in shadow. The Kisliuk-like behaviour [26] of S(H) indicates that the adsorption process is precursor mediated. For = 0.36 ev (right panel) a further regime is present in the initial stage of the adsorption process. Indeed S increases initially with H up to a coverage of 6 ML; then, if step heights are exposed, S(H) stays constant until a critical coverage between 9 and 0.12 ML, above which it drops rapidly below the KW sensitivity. Also at this energy the behaviour is different if step heights are in shadow: for h i = 59 the plateau is missing and the drop is observed already beyond 5 ML. The maximum coverage remains angle dependent also after subtracting the metastable coverage desorbed at the end of the dose (between 2 and 3 ML for the present experimental conditions [27]). We conclude therefore that, when the step height is little or not exposed, adsorption is limited to H < 6 ML, while when the step height is illuminated an additional, angle dependent, coverage is present, corresponding to the area below the flat part of the S(H)-curve. In Fig. 5 S(H) is reported for = 0.36 ev, parametric in surface temperature T, for h i = 40 (left panel) and h i = h s (right panel). The former angle corresponds to the largest value of the sticking probability and to molecules hitting at 19 grazing incidence on the step heights. For both angles it is apparent that: (1) the saturation coverage decreases continuously with increasing surface temperature; (2) the width of the flat part of S(H) decreases with T and is nearly absent at 123 K, at which temperature the maximum coverage becomes angle independent and (3) no increase of S(H) with H is observed for T = 135 K. In Fig. 6, S 0 and S max are plotted as a function of and for normal incidence. Two data points are present for = 0.36 ev, corresponding, respectively, to a seeded beam at T N = 300 K (open symbol) and to a pure beam at T N = 870 K (filled symbol). The data show that: (1) S 0 and S max decrease with increasing, as expected for non-activated adsorption because of the larger amount of energy to be dissipated in order to trap the molecule into the adsorption well; (2) S max is significantly higher than S 0 at high translational energy while it is indistinguishable from it at the lowest ; (3) at = 0.36 ev the difference between pure and seeded beam is small for S 0 but large for S max, indicating that internal energy influences the adsorption process only at non-zero coverage. In Fig. 7 we show the dependence of S 0 (upper panels) and S max (lower panels) on T for h i = 40 (left panels) and h i =31 (right panels). Different

6 L. Savio et al. / Surface Science 587 (2005) % C 2 H 4 =0.36 ev θ i = K 117 K 134 K 3 % C 2 H 4 =0.36 ev θ i = K 123 K 136 K 157 K S(Θ) Θ (ML) Θ (ML) Fig. 5. Coverage dependence of S at h i = 40 (left panel) and h i =31 (right panel), for = 0.36 ev and parametric in surface temperature, T. C 2 H 4 /Ag(410) θ i =0, T=110 K Pure Seeded C 2 H 4 /Ag(410) θi =0, T=110 K Pure Seeded S 0 S max (ev) (ev) Fig. 6. Behaviour of S 0 (left panel) and S max (right panel) as a function of translational energy for T = 110 K and normal incidence on the Ag(410) surface. Filled and empty symbols denote that the exposure was performed with pure and seeded C 2 H 4 beams, respectively. symbols correspond to different beam energies; for = 0.10 ev S 0 and S max coincide. The initial sticking probability is constant up to 140 K and decreases thereafter, becoming negligible above 180 K. S max, on the contrary, starts decreasing already at low T and, as mentioned above, becomes indistinguishable from S 0 above 140 K. In Fig. 8 we report the angle dependence of S at T = 110 K and parametric in H. The data points are obtained from the analysis of S(H) curves similar to those reported in Fig. 4 but recorded for different h i values. For H < 6 ML, S(h i ) shows a smooth angle dependence when step heights are exposed and drops abruptly when step heights are shadowed. For larger H dosing at = 0.36 ev with pure-hot or seeded-cold beams is no longer equivalent as rotations play now a role in inhibiting adsorption. For pure beams S develops initially a

7 116 L. Savio et al. / Surface Science 587 (2005) θ i =-40 o θ i = 31 o 0.10 ev 0.36 ev 0.36 ev S 0 C 2H4 /Ag(410) S max T (K) T (K) Fig. 7. Comparison between S 0 (T) (upper panels) and S max (T) (lower panels) for h i = 40 (left panels) and h i =31 (right panels). Full symbols: pure beam; open symbols: seeded beam. minimum close to normal incidence, while the curve corresponding to the seeded beam is flatter. At 0.12 ML coverage S drops also at large positive h i, indicating that access to the higher coverage sites is inhibited also for grazing incidence on the (1 0 0) nanofacets. Pure-hot and seeded-cold beams behave now very differently since the latter shows a maximum at normal incidence. 4. Discussion The data presented in the previous section provide an almost complete description of the dynamics of the C 2 H 4 Ag(410) interaction. HREELS inspection, reported elsewhere [24], indicates that only one chemisorbed species is present after the 20 min needed to record the spectrum. This finding is consistent with the rapid desorption of the less strongly bound state. The stable state has a main vibrational loss at 125 mev, corresponding to the C 2 H 4 wag motion, and desorbs below 155 K as expected for p-bonded molecules. The vibrational spectrum is the same for all h i and. The wag frequency is 4 mev higher than for C 2 H 4 /Ag(100) confirming that the adsorption site is not identical. The KW traces in Fig. 2 indicate that at T = 110 K a stable and a metastable species are present at the surface. Since S 0 is constant up to 140 K (see Fig. 7) we conclude that in the initial stages of the adsorption process the molecules end up in the most stable state. From their desorption temperature we estimate the adsorption en-

8 L. Savio et al. / Surface Science 587 (2005) A C 2 H 4 /Ag(410) S 0 S(9 ML) T=110 K C 2 H 4 -Ag(410) =0.10 ev (100%) =0.36 ev (100%) =0.36 ev (3%) (100) (110) S(0.10 ML) S(0.12 ML) θ i ( o ) θ i ( o ) Fig. 8. Angular dependence of S at different H, T = 110 K and different translational and internal energies. Full symbols: pure beam; open symbols: seeded beam. The rapid decrease of S(H) at the highest coverage causes a larger scattering of the data points which might reflect slight differences of the incident ethylene flux. ergy to be ev/molecule (assuming a prefactor of Hz). For the species metastable at 110 K we deduce, on the contrary, a binding energy of 5 ev/molecule from the measured lifetime (0.7 s, see Fig. 2) and assuming the same prefactor. Such value is in agreement with our previous experiments for C 2 H 4 Ag(001) [15,17] and is compatible with physisorption. We must note, however, that for C 2 H 4 Ag(410) the distinction between physisorbed and weakly chemisorbed molecules is not straightforward, at least with the available techniques. In fact similar desorption traces are observed also for experiments performed at T > 110 K (not shown) and assigned to weakly chemisorbed molecules having at least one second nearest neighbour. These findings are in accord

9 118 L. Savio et al. / Surface Science 587 (2005) with the quite broad thermal desorption spectra. More conservatively we should thus conclude that different ethylene moieties with different binding energies can form. Inspection of the coverage dependence of S vs temperature and angle (see Figs. 4 and 5) shows that in presence of pre-adsorbed ethylene, the sticking probability is characterised: (1) at low, by a Kisliuk-like behaviour [26]; (2) at high, by an increase of the sticking probability with coverage at low T and a Kisliuk-like behaviour at higher T; (3) at all and low T, by a strong angle dependence of the coverage at which S drops below the sensitivity of the KW experiment; such coverage decreases with T and is angle independent at T = 135 K. The Kisliuk-like behaviour observed at low indicates that a precursor is operative. The initial increase of S(H) at higher is due to adsorbate assisted adsorption, a mechanism already reported in the past for other systems [28]. When the incident molecule interacts with an ethylene pre-covered surface the energy transfer is more efficient than when it interacts with the bare substrate due to the reduced mass mismatch. The effect is dramatic at ev, where S max overcomes S 0 by more than a factor 2 (see Fig. 6); moreover it occurs also when dosing with the step heights in shadow, thus indicating that the chemisorption site is always exposed to the beam. This result is in accord with the one obtained from ab initio calculations, which predict it at the upper terrace side of the step edge [20]. The plateau in the S(H) curves of Figs. 4 and 5 indicates that, in a certain coverage range and/or at T > 110 K, adsorbate assisted adsorption is counterbalanced by some other effect. At = 0.36 ev and low T, S(H) becomes flat above 6 ML. This is also the maximum coverage reached at T = 135 K and roughly corresponds to 1/4 occupancy of the sites at the step edge. On the other hand, the maximum stable coverage reached at low T is close to 1/8 of ML, i.e. one molecule every second Ag atoms along the step length. We conclude therefore that admolecules with second nearest neighbours are less strongly bound, as it is the case e.g. for NO/Ni(100) [29], and that total occupancy of step sites is forbidden by molecule molecule repulsion. The angle dependence of S(H) reported in Fig. 4 shows that high coverages are reached more easily for molecules impinging at h i = 40. One possibility is that exposing the lower part of the step height is important. A similar conclusion was reached for the propane interaction with stepped Pt surfaces [30]. We further note that impinging at h i = 40 (i.e. at incidence angles of 26 on (1 0 0) and 71 on (1 1 0) nanofacets) or at h = +60 (corresponding to incidence angles of 74 on (1 0 0) and 29 on (1 1 0) nanofacets) is not equivalent. The physical process is thus asymmetric with respect to the two nanofacets with sticking being higher for molecules hitting grazing on the shorter terrace. We propose this counterintuitive behaviour to be connected to the location of the chemisorption site on the upper terrace side of the step edge: trapped molecules moving against the descending side of the step will thus have a larger probability of hitting against previously chemisorbed ones and being stopped. The different adsorption paths at low and at high coverage are confirmed by the different behaviour of S 0 and S max vs. (see Fig. 6) and by the high energy S(h i ) curves reported in Fig. 8. The data show that S depends on internal energy only at non-zero coverage. As demonstrated in [15] what matters are rotations. For Ag(410) they suppress the sticking probability only in presence of pre-adsorbed ethylene and most efficiently for close to normal incidence. This behaviour is linked to the broad range over which the adsorption energy varies for ethylene/ag(4 1 0). At low coverage steering is strong enough to suppress the rotational effect and the angle dependence of the sticking probability because of the attraction towards the chemisorption site at the step edge felt by the impinging molecules, as shown in [21]. With increasing coverage the adsorption energy decreases and is more uniform over the surface, so that steering is less efficient: rotations resume their role in inhibiting physisorption and the sticking probability becomes again angle dependent. At grazing incidence the probability of being scattered back into the gas-phase is large already for cold molecules so that rotations have no influence at the scale of KW experiments.

10 L. Savio et al. / Surface Science 587 (2005) Conclusions Our results can be summarised as follows: (1) On Ag(4 10) ethylene molecules chemisorb in the p-bonded state at the step edge. At 110 K the admolecules are stable up to coverages of 0.12 ML, corresponding to occupation of every second site along the step. At 136 K the coverage is stable only up to 6 ML, corresponding to quarter occupancy of the step edge, i.e. to the absence of second neighbours. The adsorption site is always exposed to the beam in accord with the theoretical result of Kokalji et al. [18], who find it at the upper side of the step edge. At variance with the flat Ag(100) case, no energy barrier is present for chemisorption. A presumably physisorbed phase, metastable at 110 K, is present, too, and acts as precursor to chemisorption. (2) The initial sticking probability decreases with increasing as expected for a nonkinetic energy activated system. Since the molecules end up chemisorbed, S 0 is temperature independent until desorption off the p- bonded state sets in at 140 K. S 0 is moreover nearly angle independent as long as the step heights are illuminated by the beam and drops thereafter. The loss of the memory of the angle of incidence is explained by the presence of steering, deviating the incoming trajectories towards the final adsorption site at the step edge. (3) The coverage dependence of the sticking probability indicates that initially the adsorption sites are occupied for all h i but at higher coverages, corresponding to less strongly bound molecules, they can be reached only along particular trajectories. (4) Steering is important for the initial sticking probability, removing its dependence on h i. At larger coverage the chemisorption energy decreases and steering becomes less effective. The adsorption probability shows then the minimum around normal incidence expected for a non-activated system. (5) Rotations have little or no effect for the initial adsorption process but inhibit it at significant coverage. Acknowledgement We acknowledge financial support by MIUR through PRIN 2003 and G. Rovida for providing the sample. References [1] G.A. Somorjai, Introduction to Surface Chemistry and Catalysis, Wiley, New York, [2] J.T. Yates, J. Vac. Sci. Technol. A 13 (1995) [3] G. Ertl, Adv. Catal. 45 (2000) 1. [4] B. Hammer, Phys. Rev. Lett. 83 (1999) [5] T. Zambelli, J. Wintterlin, J. Trost, G. Ertl, Science 273 (1996) [6] S. Dahl et al., Phys. Rev. Lett. 83 (1999) [7] P. Gambardella, Z. Sljivancanin, B. Hammer, M. Blanc, K. Kuhnke, K. Kern, Phys. Rev. Lett. 87 (2001) [8] H. Mortensen, L. Diekhoner, A. Baurichter, E. Jensen, A.C. Luntz, J. Chem. Phys. 113 (2000) [9] L. Vattuone, U. Burghaus, L. Savio, M. Rocca, G. Costantini, F. Buatier de Mongeot, C. Boragno, S. Rusponi, U. Valbusa, J. Chem. Phys. 115 (2001) [10] A.T. Gee, B.E. Hayden, J. Chem. Phys. 113 (2000) [11] L. Savio, L. Vattuone, M. Rocca, Phys. Rev. Lett. 87 (2001) [12] L. Vattuone, L. Savio, M. Rocca, Phys. Rev. Lett. 90 (2003) [13] R.A. van Santen, H.P.C. Kuipers, Adv. Catal. 35 (1990) 256. [14] W.A. Brown, R. Kose, D.A. King, Surf. Sci. 440 (1999) 271. [15] L. Vattuone, U. Valbusa, M. Rocca, Phys. Rev. Lett. 82 (1999) [16] L. Vattuone, L. Savio, M. Rocca, U. Valbusa, Chem. Phys. Lett. 331 (2000) 177. [17] L. Vattuone, L. Savio, M. Rocca, L. Rumiz, A. Baraldi, S. Lizzit, G. Comelli, Phys. Rev. B 66 (2002) [18] A. Kokalj, A. Dal Corso, S. de Gironcoli, S. Baroni, Surf. Sci (2002) 62. [19] E.M. King, S.J. Clark, C.F. Verdozzi, G.J. Ackland, J. Phys. Chem. B 105 (2001) 641. [20] A. Kokalj, A. Dal Corso, S. de Gironcoli, S. Baroni, J. Phys. Chem. B 106 (2002) [21] L. Savio, L. Vattuone, M. Rocca, S. Corriol, G. Darling, S. Holloway, Chem. Phys. Lett. 382 (2003) 605. [22] M. Rocca, U. Valbusa, A. Gussoni, G. Maloberti, L. Racca, Rev. Sci. Instrum. 62 (1991) 2171.

11 120 L. Savio et al. / Surface Science 587 (2005) [23] D.A. King, M.G. Wells, Surf. Sci. 29 (1972) 454. [24] L. Savio, L. Vattuone, M. Rocca, J. Electron. Spect. Rel. Phenom. 129 (2003) 157. [25] P.J. Knight, S.M. Driver, D.P. Woodruff, Surf. Sci. 376 (1997) 374. [26] P. Kisliuk, J. Phys. Chem. Solids 3 (1957) 95. [27] The population of the metastable species depends on the beam flux which has (a) slightly different absolute values for the different experiments and (b) decreases with cosh i. [28] See e.g. M. Kunat, Ch. Boas, Th. Becker, U. Burghaus, Ch. Wöll, Surf. Sci. 474 (2001) 114; A.F. Carlsson, R.J. Madix, J. Chem. Phys. 113 (2000) 838. [29] L. Vattuone, Y.Y. Yeo, D.A. King, Catal. Lett. 41 (1996) 919. [30] J.-C. Wang, Surf. Sci. 540 (2003) 326.