Hot electrons at metal-organic interface: Time-resolved two-photon photoemission study of phenol on Ag„111… Sunmin Ryu, Jinyoung Chang, Hyuksang Kwon, and Seong Keun Kima兲 School of Chemistry, Seoul National University, Seoul 151-747, Korea

共Received 18 October 2005; accepted 14 November 2005; published 23 June 2006兲 We used time-resolved two-photon photoemission 共2PPE兲 spectroscopy to investigate the excitation mechanism and dynamical behavior of the anionic molecular resonance 共MR兲 state of phenol weakly interacting with Ag共111兲. The photoexcited MR state of phenol was found at 3.1 eV above the Fermi level at 1 ML 共monolayer兲 coverage, and the binding energy of this state remained rather constant at 0.74± 0.05 eV for all coverages. The polarization angle dependence of the 2PPE signal clearly showed that the MR state is populated by an indirect excitation process involving scattering of photoexcited hot electrons rather than direct electronic transition from a bulk band. The lifetime of the MR state was found to increase from 33 to 60 fs upon increasing the coverage from 1 to 9 ML, implying that the MR state becomes further decoupled from the bulk at a higher coverage. These results constitute the first time-resolved 2PPE study that clearly demonstrates the hot-electron-mediated mechanism operating for molecules that are potentially active photochemically but weakly interacting with a metal surface. © 2006 American Vacuum Society. 关DOI: 10.1116/1.2167076兴

I. INTRODUCTION The way an electron traverses a metal-dielectric interface has attracted much interest from many different scientific disciplines.1 Most of the recent attention to the interfacial electron transfer or transport are paid by those who wish to improve yet primitive molecular electronic devices with metal-molecule-metal junctions in nanometer scale.2 Improved understanding of charge-injection phenomena in organic light-emitting diodes1,3,4 or dye-sensitized solar cells5 has been another motivation. Furthermore, electron transfer is intimately related to surface photochemistry of adsorbed molecules with significant electronic-nuclear coupling.6,7 The photoinduced electron transfer, in particular, occurring in surface photochemistry or dye-sensitized solar cells, has become a highly appropriate subject for a dynamical study of electronic behavior at the interfaces with the arrival of ultrafast lasers. Interfacial electronic structures and electron dynamics relevant to many of the above processes can be directly probed by two-photon photoemission 共2PPE兲 spectroscopy.1,8–11 In 2PPE, a normally unoccupied state can be populated by the first photon and then the decay of this state is probed by the second photon as in conventional photoelectron spectroscopy. Time-resolved 2PPE experiments using ultrafast lasers can thus provide detailed dynamical information about the transient state, the identification of which is essential to a comprehensive understanding of interfacial electron transfer.1,10 From this mechanistic viewpoint, it is highly desirable to investigate the photoexcitation mechanism of a normally unoccupied state originating from an adsorbed molecule, and to learn how the strength of substrate-molecule interaction afa兲

Electronic mail: [email protected]

1454

J. Vac. Sci. Technol. A 24„4…, Jul/Aug 2006

fects the excitation mechanism.1,10 For example, Wolf et al.12 found for CO chemisorbed on Cu共111兲 that its unoccupied 2␲*-derived state can be populated in a direct electronic transition by an electric field either perpendicular or parallel to the surface, which was attributed to significant hybridization of the CO 2␲* molecular orbital with the Cu substrate. Even for a weakly bound molecule such as C6F6 or naphthalene on Cu共111兲,13,14 their respective anionic resonances were found to undergo a direct electronic transition rather than an indirect or hot-electron-mediated6,7 excitation. These observations appear quite puzzling, in view of the fact that many photochemical reactions of adsorbates on metals occur through an indirect excitation mechanism since photogeneration of charge carriers is, in general, highly efficient.7,15 In this article we report an extension, by femtosecond time-resolved 2PPE, of our recent investigation16 on the excitation mechanism and the dynamical behavior of the anionic resonance found in phenol/ Ag共111兲. From the polarization angle dependence of the 2PPE intensity for the anionic state, we could unambiguously show that a substratemediated excitation mechanism is exclusively operative in this weakly interacting system. II. EXPERIMENT Since our time-resolved 2PPE setup has been described in detail elsewhere,16,17 only a brief overview is given here. All experiments were performed in an ultrahigh vacuum 共UHV兲 chamber equipped with standard surface science tools.18 As a light source, we used a visible laser light from an optical parametric amplifier, which was fed by a commercial Ti:sapphire femtosecond laser with a regenerative amplifier. Some of the visible beam 共probe pulse, h␯2兲 was frequencydoubled with a beta-barium borate 共BBO兲 crystal to give an ultraviolet 共UV兲 light 共pump pulse, h␯1兲 over 290–380 nm.

0734-2101/2006/24„4…/1454/6/$23.00

©2006 American Vacuum Society

1454

1455

Ryu et al.: Hot electrons at metal-organic interface

1455

FIG. 2. Energy of the MR state 共EMR兲 and the vacuum level 共Evac兲, referenced to the Fermi level 共EF兲, as a function of coverage.

FIG. 1. 2PPE spectra of phenol/ Ag共111兲 obtained in normal emission at various coverages 共⌰兲. h␯2 = 1.90 eV for ⌰ 艋 1 ML and h␯2 = 1.75 eV for ⌰ ⬎ 1 ML. The spectra are shown as a function of the final state energy E − EF = Ekin + ⌽, where EF , Ekin, and ⌽ are the Fermi level, the electron kinetic energy, and the work function, respectively. “SS,” “MR,” and “IBT” stand for the surface state, the molecular resonance state, and the interband transition, respectively.

The time delay between the UV and visible pulses was controlled with a precision translator. Both beams were collinearly combined by a dichroic mirror and incident on the Ag共111兲 surface at 66° from the surface normal. The polarizations of the pump and probe beams were independently controlled with half wave plates. The kinetic energy of photoelectrons was measured by an electron time-of-flight 共TOF兲 spectrometer. The Ag共111兲 surface was cleaned by standard cycles of Ar-ion sputtering and annealing at 700 K. Phenol 共Merck, ⬎99.8%兲 was purified by several freezepump-thaw cycles and introduced onto the crystal surface held at 163 K through a pinhole doser. All 2PPE experiments were carried out with the crystal maintained at 90 K. III. RESULTS AND DISCUSSION A. Interfacial electronic structures

In a recent study,16 we investigated the electronic structure of phenol adsorbed on Ag共111兲 by 2PPE spectroscopy using a nanosecond laser and discovered an anionic molecular resonance 共MR兲 state with a large effective mass. Based on its negligible electronic dispersion and energetics consistent with the threshold energy18 for photodissociation, the MR state was identified as the intermediate state for photodissociation of phenol. In the present work, phenol/ Ag共111兲 has been reexamined with femtosecond time-resolved 2PPE spectroscopy in order to address dynamical aspects of the MR state and corroborate our earlier proposition for the excitation mechanism of the MR state. JVST A - Vacuum, Surfaces, and Films

Phenol forms three distinct adsorption layers on Ag共111兲: a weakly chemisorbed layer, an intermediate layer, and an infinitely growing multilayer.19 The chemisorption layer desorbs molecularly with laterally repulsive interaction as exhibited in thermal desorption spectra. Figure 1 shows 2PPE spectra obtained at zero time delay at various coverages 共⌰兲 from 0 to 7.6 ML 共monolayer兲. On bare Ag共111兲, one finds a prominent peak at 5.7 eV 共designated SS兲 originating from nonresonant two-photon ionization of the occupied surface state located 50 meV below the Fermi level.20 At ⌰ = 0.21 ML, another band 共designated MR for molecular resonance state兲 shows up and moves towards lower energy with increasing coverage, in parallel with the shift in the vacuum level as identified by the edge of the secondary-electron band at the low-energy onset of the photoelectron spectra. The decrease in work function implied by the latter allows onephoton photoemission by the UV pump, generating the sharp peak at 3.9 eV at coverages of 0.78 and 1.0 ML. We note that all the spectral features remain virtually the same at coverages higher than 1.0 ML, where the photon energy was reduced to avoid one-photon photoemission. Although most of these features have already been found with nanosecond 2PPE, the present femtosecond 2PPE study revealed a new peak at 4.8 eV 共designated IBT for interband transition兲, which survives adsorption and remains at the same energy up to 1 ML. By comparing the 2PPE result at various photon energies with calculated electronic bands of Ag bulk,21 it was suggested that this band originates from nonresonant twophoton excitation between sp-derived bulk bands.22 The energy of the MR state 共EMR兲 can be found by subtracting the probe photon energy from the final state energy as shown in Fig. 2. Also shown in Fig. 2 is the vacuum level energy 共Evac兲 from the Fermi level 共i.e., work function兲 determined from the onset of the secondary-electron bands in Fig. 1. Upon increasing the coverage to 1 ML, both Evac and EMR decrease sharply but their difference 共Evac − EMR兲, the binding energy of the MR state, remains virtually constant at 0.74± 0.05 eV. The constant binding energy of the MR state suggests that the global work function determined by 2PPE

1456

Ryu et al.: Hot electrons at metal-organic interface

FIG. 3. 共a兲 Polarization-dependent 2PPE spectra obtained in normal emission for phenol/ Ag共111兲 at 2 ML. The solid line is for p-polarized pump and p-polarized probe, while the dashed line is for s-polarized pump and p-polarized probe, and the dotted line is for p-polarized pump and s-polarized probe. 共b兲 Integrated intensity of the MR peak 共IMR; circles兲 as a function of the polarization angle between the pump and the probe maintained p-polarized. 0° and 90° correspond to p- and s-polarized pumps, respectively. The solid line indicates the normalized absorptance of a silver surface at the photon energy h␯1 employed, while the dashed line represents the square of the electric field normal to the surface, 兩E⬜兩2.

in Fig. 2 is nearly equal to the local work function that dictates the energetics of the interfacial electronic structure.3,8 This implies that the adsorbates in a submonolayer coverage are rather evenly dispersed on the surface probably due to a lateral repulsion19 instead of forming close-packed islands. B. Excitation mechanism of the MR state

In our previous article,16 we suggested that the photoexcitation of the MR state is due to the scattering of photoexcited hot electrons from the substrate on the basis of a preliminary experiment on polarization angle dependence. In the present work, we employed femtosecond pulses that turned out to give a higher signal-to-noise ratio because of the reduced photoreaction of the adsorbate. The enhanced experimental sensitivity allowed us to unambiguously identify the excitation mechanism as discussed below. Figure 3共a兲 shows the 2PPE spectrum at ⌰ = 2 ML with different combinations of polarization for the pump and probe pulses. The solid line represents the pump and probe pulses both p polarized. When we change the probe from pto s-polarized light with the pump fixed p polarized, the intensity of the MR band decreases drastically to less than a few percent, as shown by the dotted line. Since the photoJ. Vac. Sci. Technol. A, Vol. 24, No. 4, Jul/Aug 2006

1456

emission is of ␴ symmetry, i.e., allowed only by the electric field normal to the surface 共E⬜兲, the MR state essentially belongs to a point group of high symmetry, at least Cn. In the electric dipole approximation, this suggests that the MR state is totally symmetric, since the final state observed in normal emission along 关111兴 should be totally symmetric23 and the electric field normal to the surface also spans a totally symmetric species for the relevant point group. The little peak intensity for the s-polarized probe should be attributed to polarization impurity and/or imperfection of the surface flatness, since the SS peak involving a transition dipole moment of ␴ symmetry had a similar intensity ratio for p- and s-polarized probes. When the pump was changed from p- to s-polarized light with the probe fixed p polarized, however, the decrease in the MR peak intensity was just 65% 关shown by the dashed line in Fig. 3共a兲兴, indicating a significant contribution of the electric field parallel to the surface. In Fig. 3共b兲, the MR peak intensity 共denoted IMR兲 is shown by circles as a function of the polarization angle 共⍀兲 of the pump. ⍀ was defined with respect to the plane of incidence with 0° and 90° corresponding to the p- and s-polarized lights, respectively. The solid line in Fig. 3共b兲 represents the normalized absorptance 共A⍀ / A p兲 of a silver surface as a function of the polarization angle at the pump photon energy h␯1 : A⍀ / A p = 共A p cos2 ⍀ + As sin2 ⍀兲 / A p, where A p = 共1 − 兩r p兩2兲 , As = 共1 − 兩rs兩2兲, and r p and rs represent the Fresnel reflection coefficients for the pand s-polarized lights, respectively.12 The complex refractive index used to calculate r p and rs is 0.209+ 1.44i at h␯ = 3.49 eV 共Ref. 24兲 and the incidence angle was 66°. The calculated absorptance fits the experimental data very nicely, indicating that the population of the MR state is directly proportional to the amount of photons absorbed by the substrate. This remarkable agreement is a strong piece of evidence for the indirect excitation mechanism,12 which has been frequently invoked in surface photochemistry under the name of the hot-electron-mediated or substrate-mediated mechanism.6,7 Alternatively, one may think that the excitation to the MR state is due to a direct electronic transition from an occupied bulk state. In such a case, the ⍀ dependence of IMR can be expressed in the dipole approximation as follows:12 IMR ⬀ 兩␮ · E⍀兩2, where ␮ and E⍀ represent the transition dipole moment from the initial bulk state to the MR state and the electric field at the surface, respectively. In view of the electronic band structure of Ag,25 the occupied sp-derived band that disperses along the ⌫-L direction in the bulk Brillouin zone and is totally symmetric should be the most probable state within reach of the pump photon energy below the MR state. Since the MR state was found to be also totally symmetric, only the electric field normal to the surface would give a nonzero photoelectron intensity. The normalized 兩E⬜兩2 represented by the dashed line in Fig. 3共b兲 is, however, not consistent with the experimental data. An unoccupied state originating from a weakly adsorbed molecule such as the MR state of phenol on Ag共111兲 is likely to undergo an indirect excitation from the bulk since a direct

1457

Ryu et al.: Hot electrons at metal-organic interface

1457

FIG. 4. Time-resolved 2PPE spectra of 2 ML phenol obtained at various time delays 共⌬t兲 between the pump and probe pulses.

transition requires a strong electronic coupling or spatial overlap with the corresponding wave function of the bulk. However, the anionic resonances observed for C6F6 and naphthalene weakly interacting with Cu共111兲 共Refs. 13 and 14兲 were found to undergo a direct excitation of ␴ symmetry. These rather counterintuitive results remain largely unresolved, although they can be partly explained by the argument that the observed anionic state is mixed with the image potential 共IP兲 state that undergoes a direct excitation of ␴ symmetry.1 Application of such an argument to the MR state of phenol would not go against the observed result, since the indirect mechanism is also operative for the IP states of Ag共111兲 and CCl4 / Ag共111兲.17 Although we cannot satisfactorily answer why the excitation mechanism for the unoccupied interfacial state is different between Ag共111兲 and Cu共111兲, one argument can be made. Since the evanescent tail of the wave function of the respective IP state into the bulk band gap governs the electronic coupling with the bulk,8,26 the excitation mechanism should depend on the magnitude of the coupling in the first approximation.1 While there are a few contradicting theoretical estimates,8,26,27 the IP state of Ag共111兲 should have a smaller probability density penetrating into the bulk than that of Cu共111兲 since the lifetime of the IP state, which is roughly inversely proportional to the magnitude of the penetration,8 is twice larger for Ag共111兲 共Refs. 17 and 28兲 than for Cu共111兲.29 C. Dynamical behavior of the MR state

Figure 4 shows the time-resolved 2PPE spectra of 2 ML phenol obtained at various time delays 共⌬t兲 between the pump and probe pulses. Both the secondary-electron band and the MR band decrease together in intensity upon increasing the time delay, but the decay of the MR band is significantly slower than that of the secondary-electron band, imJVST A - Vacuum, Surfaces, and Films

FIG. 5. 共a兲 Integrated intensity of the MR peak 共IMR; circles兲 at 2 ML coverage as a function of the time delay. The dashed line is the crosscorrelation curve between the pump and the probe, while the circles are the experimental data, and the solid line is a theoretical fit to the data. 共b兲 Lifetime of the MR state as a function of the phenol coverage.

plying that the MR state has a considerable lifetime. On the other hand, there is no indication of relaxation for the MR state in the energetic sense since the peak position of the MR band remains nearly constant within 10 meV throughout the entire range of time delay. In Fig. 5共a兲, the MR band intensity IMR at 2 ML coverage is plotted as a function of the time delay. We used the SS peak of clean Ag共111兲 originating from nonresonant two-photon ionization of the occupied surface state to represent the pump-probe cross correlation 关shown by the dashed line in Fig. 5共a兲兴, of which the full width at half maximum was ⬃133 fs. In order to estimate the temporal change of population for the MR state, we used a simple kinetic model that involves a single-exponential decay.30 When deconvoluted against the pump-probe cross correlation, the decay curve 关the solid line fitting the experimental data in Fig. 5共a兲兴 gives a time constant of 33± 5 fs. This time constant can be considered as the effective lifetime of the MR state before it decays back to the metal. Figure 5共b兲 shows the lifetime of the MR state determined in the same manner for various coverages. At low coverages of 1 or 2 ML, the lifetime of the MR state is essentially equal to that of the IP state for bare Ag共111兲 共Refs. 17 and 28兲 within the uncertainty denoted by the error bar in Fig. 5共b兲. As the coverage increases, the lifetime increases, reaching 60± 5 fs at 9 ML.

1458

Ryu et al.: Hot electrons at metal-organic interface

Dynamical behavior of an unoccupied interfacial state such as the IP state,9 the anionic resonance state of a weakly interacting molecule,13 the antibonding state of a chemisorbed molecule,31 and the solvated electron state32 has been actively investigated with the time-resolved 2PPE method by several groups. While the observed dynamics of the IP state on metal surfaces with or without simple dielectric molecules has been explained by simple model calculations with some success,9,33 a rigorous analysis appears to be quite out of reach because of the intrinsic complexity of the problem. Therefore, the results will be only qualitatively discussed in this article, taking also into consideration the electronicnuclear coupling1 that is responsible for the O–H bond scission.16 Although many of the unoccupied states mentioned above are often highly excited to several eVs above the Fermi level, most of them survive for tens of femtoseconds rather than a few femto seconds34 as expected from the Fermi liquid theory. Their relatively long lifetime is due to the fact that their wave function is largely located outside of, and thus decoupled from, the bulk.8 In particular, some lowindex planes such as our own Ag共111兲 surface provide a far smaller coupling due to the bulk band gap over the pertinent energy range. In this weak coupling limit, the lifetime of the MR state should be inversely proportional to its electronic coupling to the bulk,8 since the excited electron should return to the bulk despite the weak coupling. In this sense, the increased lifetime at higher coverages suggests that the wave function of the MR state becomes less coupled to the metal surface. We note that the wave function of the MR state is laterally localized as indicated by its large effective mass16 and also confined vertically within the adsorbate layer since the adsorbate exerts an attractive force13 because of the drastically increased electron affinity.16 At a higher coverage, the wave function of the MR state is thus displaced farther away from the surface, which results in a smaller coupling with the bulk. The long lifetimes observed at all coverages of this study are also consistent with the insignificant degree of wave-function penetration into the bulk, which has already been inferred from the predominance of the indirect excitation. Since the MR state is believed to be the intermediate state for the photodissociation of phenol, the bond scission process as well as the back transfer of electron to the bulk may play some role in the decay of the MR state. The essentially time-independent energy of the MR state in Fig. 4, however, implies that the population of the excited molecule that eventually undergoes the O–H bond scission is negligibly small. Even in the case that the bond scission predominates the decay process of the MR state, a similar coverage dependence of lifetime is expected. Since only the first phenol layer is subject to photodissociation as we found in our previous photochemical study,18 the wave function displaced towards the upper layer at a higher coverage will be less likely to induce a chemical reaction in the first monolayer, thereby leading to an extended lifetime. J. Vac. Sci. Technol. A, Vol. 24, No. 4, Jul/Aug 2006

1458

IV. CONCLUSION We employed time-resolved 2PPE spectroscopy to investigate the excitation mechanism and dynamical behavior of the anionic resonance state of phenol weakly interacting with Ag共111兲. From the polarization angle dependence of the 2PPE signal, we clearly showed that the MR state is populated by an indirect excitation process involving scattering of photoexcited hot electrons rather than a direct electronic transition from a bulk band. Upon increasing the coverage from 1 to 9 ML, the lifetime of the MR state was found to increase from 33 to 60 fs, implying that the electronic coupling between the MR state and the bulk becomes reduced at a higher coverage. The weak coupling manifested in the population dynamics of the MR state is also consistent with the indirect excitation mechanism deduced from the polarization angle dependence study. A possible role of the electronic-nuclear coupling in the decay process has also been discussed. These results are significant in the sense that they reveal an indirect excitation process that is responsible for a photochemical reaction occurring at a weakly interacting molecule-metal interface. ACKNOWLEDGMENTS This work was supported by the grant 共R02-2003-00010073-0兲 from the Basic Research Program of the Korea Research Foundation and also by the National Research Laboratory program of the Ministry of Science and Technology. X. Y. Zhu, J. Phys. Chem. B 108, 8778 共2004兲. M. A. Reed, C. Zhou, C. J. Muller, T. P. Burgin, and J. M. Tour, Science 278, 252 共1997兲; M. N. Bussac, D. Michoud, and L. Zuppiroli, Phys. Rev. Lett. 81, 1678 共1998兲; A. Nitzan, Annu. Rev. Phys. Chem. 52, 681 共2001兲; A. Nitzan and M. A. Ratner, Science 300, 1384 共2003兲. 3 H. Ishii, K. Sugiyama, E. Ito, and K. Seki, Adv. Mater. 共Weinheim, Ger.兲 11, 605 共1999兲. 4 D. Cahen and A. Kahn, Adv. Mater. 共Weinheim, Ger.兲 15, 271 共2003兲. 5 A. Hagfeldt and M. Gratzel, Acc. Chem. Res. 33, 269 共2000兲. 6 X. Y. Zhu, Annu. Rev. Phys. Chem. 45, 113 共1994兲. 7 F. M. Zimmermann and W. Ho, Surf. Sci. Rep. 22, 129 共1995兲. 8 T. Fauster and W. Steinmann, in Photonic Probes of Surfaces, edited by P. Halevi 共Elsevier Science, Amsterdam, 1995兲, Vol. 2, p. 347. 9 C. B. Harris, N. H. Ge, R. L. Lingle, J. D. McNeill, and C. M. Wong, Annu. Rev. Phys. Chem. 48, 711 共1997兲. 10 X. Y. Zhu, Annu. Rev. Phys. Chem. 53, 221 共2002兲. 11 H. Petek and S. Ogawa, Prog. Surf. Sci. 56, 239 共1997兲. 12 M. Wolf, A. Hotzel, E. Knoesel, and D. Velic, Phys. Rev. B 59, 5926 共1999兲. 13 C. Gahl, K. Ishioka, Q. Zhong, A. Hotzel, and M. Wolf, Faraday Discuss. 117, 191 共2000兲. 14 H. F. Wang, G. Dutton, and X. Y. Zhu, J. Phys. Chem. B 104, 10332 共2000兲. 15 W. Ho, in Laser Spectroscopy and Photochemistry on Metal Surfaces, edited by H.-L. Dai and W. Ho 共World Scientific, Singapore, 1995兲, Vol. II, p. 1054. 16 J. Lee, S. Ryu, J. Chang, S. Kim, and S. K. Kim, J. Phys. Chem. B 109, 14481 共2005兲. 17 S. Ryu, J. Chang, and S. K. Kim, J. Chem. Phys. 123, 114710 共2005兲. 18 J. Lee, S. Ryu, J. S. Ku, and S. K. Kim, J. Chem. Phys. 115, 10518 共2001兲. 19 J. Lee, S. Ryu, and S. K. Kim, Surf. Sci. 481, 163 共2001兲. 20 K. Giesen, F. Hage, F. J. Himpsel, H. J. Riess, and W. Steinmann, Phys. Rev. B 33, 5241 共1986兲. 21 H. Eckardt, L. Fritsche, and J. Noffke, J. Phys. F: Met. Phys. 14, 97 1 2

1459

Ryu et al.: Hot electrons at metal-organic interface

共1984兲. T. Miller, W. E. McMahon, and T. C. Chiang, Phys. Rev. Lett. 77, 1167 共1996兲; S. Pawlik, R. Burgermeister, M. Bauer, and M. Aeschlimann, Surf. Sci. 404, 556 共1998兲. 23 S. Hufner, Photoelectron Spectroscopy, 2nd ed. 共Springer, Berlin, 2003兲. 24 Handbook of Optical Constants of Solids, edited by D. P. Palik 共Academic, San Diego, 1998兲. 25 C. N. Berglund and W. E. Spicer, Phys. Rev. 136, A1044 共1964兲. 26 E. V. Chulkov, V. M. Silkin, and P. M. Echenique, Surf. Sci. 437, 330 共1999兲. 27 P. de Andres, P. M. Echenique, and F. Flores, Phys. Rev. B 35, 4529 共1987兲. 28 J. D. McNeill, R. L. Lingle, N. H. Ge, C. M. Wong, R. E. Jordan, and C. B. Harris, Phys. Rev. Lett. 79, 4645 共1997兲. 22

JVST A - Vacuum, Surfaces, and Films

1459 M. Wolf, E. Knoesel, and T. Hertel, Phys. Rev. B 54, R5295 共1996兲. R. L. Lingle, N. H. Ge, R. E. Jordan, J. D. McNeill, and C. B. Harris, Chem. Phys. 205, 191 共1996兲. 31 H. Petek, M. J. Weida, H. Nagano, and S. Ogawa, Science 288, 1402 共2000兲. 32 N. H. Ge, C. M. Wong, R. L. Lingle, J. D. McNeill, K. J. Gaffney, and C. B. Harris, Science 279, 202 共1998兲; A. D. Miller, I. Bezel, K. J. Gaffney, S. Garrett-Roe, S. H. Liu, P. Szymanski, and C. B. Harris, ibid. 297, 1163 共2002兲; C. Gahl, U. Bovensiepen, C. Frischkorn, and M. Wolf, Phys. Rev. Lett. 89, 107402 共2002兲. 33 A. Hotzel, G. Moos, K. Ishioka, M. Wolf, and G. Ertl, Appl. Phys. B: Lasers Opt. 68, 615 共1999兲. 34 M. Wolf, Surf. Sci. 377–379, 343 共1997兲. 29 30