- Email: [email protected]
Surface enhanced Raman scattering as a probe of adsorbate–substrate charge-transfer excitations
Surface Science 427–428 (1999) 115–125
www.elsevier.nl/locate/susc
Surface enhanced Raman scattering as a probe of adsorbate–substrate charge-transfer excitations Patanjali Kambhampati, Alan Campion * Department of Chemistry and Biochemistry, The University of Texas at Austin, Austin, TX 78712, USA
Abstract The chemical mechanism of surface enhanced Raman scattering (SERS) is used as a probe of adsorbate–substrate charge-transfer excitations on atomically smooth, single crystal copper surfaces in ultra-high vacuum. It is demonstrated that SERS and charge-transfer excitations are supported on atomically smooth surfaces in contradiction to earlier consensus. The combination of SERS with electron energy loss spectroscopy (EELS ) provides a great deal of insight into the nature of these excitations. It is found that these excitations are sensitive probes of local electronic structure at metal surfaces. The spatial extent, transition moment, and excited state potential energy surfaces of these excitations are probed using SERS, EELS and molecular spectroscopy simulations. © 1999 Elsevier Science B.V. All rights reserved. Keywords: Alkali metals; Cesium; Charge transfer; Copper; Electron energy loss spectroscopy; Excited states; Fermi energy; Image potential states; Localization; Metal surfaces; Pyromellitic dianhydride; Surface enhanced Raman scattering; Surface states; Work function
1. Introduction Much is now known about the properties of the ground electronic states of molecules adsorbed on solid surfaces, due, in large part, to the tremendous power of modern surface science experimental and theoretical approaches. We know considerably less about the properties of the excited electronic states of adsorbed species and the manner in which they couple to the substrate, however, due to the relative paucity of techniques available to probe such states. Absorption spectroscopy, whether by reflectance, ellipsometry or inelastic electron scattering, generally yields only broad, featureless spectra because excited state lifetimes are dramatically shortened by rapid non* Corresponding author. Fax: +1 512-4718696. E-mail address: [email protected] (A. Campion)
radiative energy transfer to the surface [1]. The importance of electronically excited states in both physical and chemical processes at surfaces can hardly be overstated, though. They participate in DIET (desorption induced by electronic transitions) [2], photochemistry [3,4] and photoelectrochemistry [5] at surfaces, and photon-assisted etching of electronic materials [6 ]. Surface photochemistry is also of great importance in the surface science of heterogeneous catalysis where it can be used to effect the selective synthesis of adsorbed fragments which cannot be made by other means [3,4]. While it is true that some understanding of the nature of electronic excited states has come from the comprehensive body of DIET work, these states are, by definition, dissociative and these experiments have generally been interpreted using only one-dimensional models. There is clearly an
0039-6028/99/$ – see front matter © 1999 Elsevier Science B.V. All rights reserved. PII: S0 0 39 - 6 0 28 ( 99 ) 0 02 4 5 -9
116
P. Kambhampati, A. Campion / Surface Science 427–428 (1999) 115–125
important need to achieve, for adsorbed molecules, a comprehensive understanding of the full normal coordinate dependence of the excited state potential surface which we have for a number of molecules in the gas phase and in solution. In addition, we seek information about excited state lifetimes, couplings to other states and the orientations of transition dipole moments for adsorbed species. We describe here experiments which use surface enhanced Raman scattering (SERS) and electron energy loss spectroscopy ( EELS ) along with molecular spectroscopy simulations to achieve such an understanding for a general class of excited states, those which arise from adsorbate–substrate charge-transfer excitations. These excitations are significant in that they are model systems with which to study excited state interactions between molecules and metal surfaces, and for their importance in the chemical mechanism of surface enhanced Raman scattering. When adsorbed on some metal surfaces such as silver, copper and gold, molecules can exhibit enormously enhanced Raman scattering which has become known as surface enhanced Raman scattering [7–9]. The SERS effect has historically been associated with substrate roughness on two characteristic length scales. Surface roughness on the 10 to 100 nm length scale supports the electromagnetic resonances which are the dominant mechanism of enhancement. These electromagnetic resonances can increase the scattered intensity by ca. 104 to 107. A second mechanism, often thought to require atomic scale roughness, is referred to as the chemical mechanism. This second mechanism involves the creation of new electronic excited states which result from adsorbate–substrate chemical interactions. It is estimated that the chemical mechanism can enhance the scattering crosssection by a factor of ca. 10 to 102. Estimates have been made of the magnitude of the chemical mechanism in the presence of the electromagnetic mechanism in silver island films in ultra-high vacuum ( UHV ) [10]. These two mechanisms operate simultaneously (multiplicatively), making it difficult to isolate the role and magnitude of each mechanism. The chemical mechanism of SERS is a process analogous to molecular resonance Raman scatter-
ing except that resonances involved are not intramolecular in origin [9–12]. They are chargetransfer resonances in which there is photoinduced charge hopping between the adsorbate and the substrate. The general consensus had been that atomic scale roughness is required to couple electronically the adsorbate states to the substrate states in such a manner as to produce chargetransfer excitations and SERS [7,8]. It has proven to be very difficult to study the chemical enhancement mechanism selectively for two reasons. First, the mechanism contributes a small fraction to the total enhancement. Second, almost any experimental parameter which can be varied to probe a system will have an influence via both mechanisms, making the separation of effects difficult, if not impossible. For this reason, there has been considerable effort to understand the effect in detail [9–22]. The use of single crystal substrates can provide a great deal of insight into the problem. We can minimize the inhomogeneous broadening present in the roughened electrode experiments that were used to assign the nature of the intermediate state in SERS [7,8,12]. These ordered substrates also facilitate a more detailed understanding of the coupling between the adsorbed molecule and the metal surface such as the dynamics, spatial and electronic structure of these excited states. It is extremely important to understand the chemical mechanism both for its relevance to analytical applications and to achieve a fundamental understanding of the electronic coupling between an adsorbed molecule and a metal surface.
2. Experiment The experimental methods used have been described in detail previously [23–29] and will be only briefly discussed here. The surfaces were polished by standard metallographic techniques and were mounted in UHV chambers with base pressures ca. 1×10−10 Torr, equipped with the standard surface analysis tools. A Leybold–Heraeus ELS-22 high resolution electron energy loss (HREEL) spectrometer was used for the electronic EELS studies. Primary beam energies between 10 and 12 eV were used
P. Kambhampati, A. Campion / Surface Science 427–428 (1999) 115–125
and the instrumental resolution (convolution of both the monochromator and analyzer stages) was ca. 50 meV ( FWHM ). We have used EELS to measure the electronic spectra reported herein, and have verified the dipolar scattering mechanism by angular dependence studies [30]; the use of EELS to obtain electronic spectra for both adsorbed molecules [11,30,31] and molecules in the gas phase [32,33] is well documented. Raman spectra were obtained using ~150 mW of 647 nm light for Cu(100) and ~300 mW of 647, 660, 693, 725 nm light for Cu(111), all produced by a Coherent Innova-200 Ar+ laser pumping Kiton Red or Pyridine-2 (Exciton) in a Coherent Cr-599 dye laser. The sample temperature, as monitored by a chromel–alumel thermocouple, did not change upon illumination. The configuration and detailed descriptions of the optics and detection system have been discussed in more detail previously [23,27,34]. On our apparatus, the background count rates are typically ca. 10 counts s−1 W−1 for non-luminescent substrates such as Ag excited with 488 nm light and ca. 50 counts s−1 W−1 for luminescent substrates such as Cu excited with 647 nm light. An unenhanced Raman experiment on a molecule such as benzene typically produces ca. 1 counts s−1 W−1 (peak) or 30 counts s−1 W−1 (integrated area), whereas the PMDA/Cu system typically produces ca. 100 counts s−1 W−1 (peak) or 3000 counts s−1
117
W−1 (integrated area). All spectra were acquired in 1000 s scans with the luminescent background subtracted. For the Raman excitation profile of the mode reported here, required corrections were made for the filter and monochromator throughput and detector efficiency.
3. Charge-transfer excitations and SERS Motivated by an interest in studying adsorption and polymerization reactions of molecular dielectrics on technologically important surfaces, our laboratory conducted surface Raman experiments of pyromellitic dianhydride adsorption on Cu(111) [23]. The chemical interaction of PMDA with copper surfaces is now well understood on a structural level [23–25,35,36 ]. PMDA dissociatively chemisorbs by splitting out CO, producing a surface carboxylate. The chemisorbed carboxylate plane is tilted away from the surface normal with its oxygen atoms inequivalently bound to the surface, whereas the plane containing the aromatic portion of the molecule is oriented perpendicular to the plane of the carboxylate (Fig. 1). It has been suggested that this geometry, which tilts the long axis towards the surface, allows an unsatisfied valence on the aromatic portion of the molecule to interact with the surface copper atoms. In conducting the Raman experiments, we
Fig. 1. The chemical structure of pyromellitic dianhydride and the geometry of the adsorbed surface carboxylate. The black atoms are carbon, the gray are oxygen and the white are hydrogen.
118
P. Kambhampati, A. Campion / Surface Science 427–428 (1999) 115–125
found that the count rates were significantly greater than those of all other physisorbed and chemisorbed molecules that we had studied [23,26– 28]. The surface Raman spectra of monolayer coverage PMDA adsorbed on Cu(111) are shown in Fig. 2 at two excitation wavelengths. It is readily apparent that the total intensity is strongly dependent upon the excitation wavelength, in contrast to normal, unenhanced surface Raman scattering [34]. That the total intensity at 647 nm excitation is on the order of 100 to 200 times more intense than most physisorbed or chemisorbed molecules is the clearest evidence of an operative enhancement mechanism. Consistent with a resonance type of effect, the relative intensities are also dependent upon excitation wavelength. It is surprising, however, to encounter a resonance when exciting a small aromatic molecule with visible light. PMDA is a colorless compound with its lowest energy transitions in the ultraviolet. In order to identify the resonance and thus the origin of the enhancement, we performed electron energy loss spectroscopy ( EELS) experiments on the Cu(111) and Cu(100) surfaces [26–28] (Fig. 3). Both surfaces show an intense, narrow loss feature at 1.9 eV which matches the laser wavelength of 647 nm. This feature saturates at monolayer coverage whereas the three higher energy peaks increase in intensity at all coverage. Therefore, the low energy 1.9 eV peak is interfacial in character and responsible for the operative
Fig. 2. Raman spectra of one monolayer PMDA adsorbed on the Cu(111) surface. p-Polarized excitation at 725 nm (a) and 647 nm (b) was used.
Fig. 3. Electron energy loss (EEL) spectra of one monolayer of PMDA adsorbed on the Cu(111) and Cu(100) surfaces.
enhancement mechanism, whereas the higher energy peaks are simply the intramolecular excitations of the adsorbed carboxylate at monolayer coverage and the physisorbed parent for subsequent multilayers.
4. Surface structure and laser polarization dependence of SERS We initially hypothesized that the low energy interfacial excitation was a charge-transfer excitation by analogy to the previous experiments in the field [7–12]. We recently confirmed this hypothesis by experiments in which we were able to vary the Fermi energy of the substrate independently of the energies of the adsorbate molecular orbitals [29]. The idea of charge-transfer has been invoked to explain the potential dependence of the enhancement of molecules adsorbed on roughened electrodes. In those experiments the chargetransfer linewidths were in the range of 500 to 1000 meV. Charge-transfer had never been observed, however, on atomically smooth surfaces, making the nature of the interfacial excitation we observed unclear. In adsorbate–substrate chargetransfer excitations, it was generally assumed that there was photogenerated charge hopping between the adsorbate and the manifold of states available in the bulk band structure [7,8,11,12,37–40]. The
P. Kambhampati, A. Campion / Surface Science 427–428 (1999) 115–125
119
bulk density of states was considered a source of the large linewidth. Since the interfacial excitations we observed were so narrow, we performed experiments on two low index faces of copper to pursue analogies to surface states [41–45] and image potential states [45–49] which are also quite narrow and with properties that depend upon crystal face. The EELS data show, that while superficially similar, the excitations are different on both surfaces. Fig. 4 more clearly shows the differences between the two surfaces. While the energies and widths are nearly the same, the vibronic shoulder is more pronounced and appears at higher energy on the Cu(111) surface. The Raman spectra are superficially similar but show subtle and meaningful differences. Fig. 5 shows that the Raman spectra on the two surfaces are different when excited with p-polarized light at 647 nm. The total intensity on Cu(111) is ca. twice that on Cu(100). Additionally, the relative intensities are dependent on the crystal face. We see that, by analogy to surface and image potential states, the interfacial Fig. 5. Raman spectra of monolayer PMDA adsorbed on Cu(111) and Cu(100) at 647 nm using both p-polarized and s-polarized excitation.
Fig. 4. EEL spectra of monolayer PMDA adsorbed on Cu(111) and Cu(100) in the region of the interfacial charge-transfer excitation. The Cu(111) system has a more intense and higher energy vibronic shoulder.
excitation is sensitive to crystal face and to the local electronic structure at the surface. To further understand the role of the interfacial excitation in the enhancement process, we performed simulations [27,28] which utilize Heller and coworkers’ wavepacket picture of molecular spectroscopy [50–53]. In these simulations, the standard Kramers–Heisenberg–Dirac dispersion equation for the molecular polarizability tensor is recast in the time domain. Franck–Condon factors are evaluated by computing the full or half Fourier transforms of the appropriate autocorrelation functions of the nuclear wavepackets as they evolve on the excited state potential energy surface. The transitions and relevant potentials are shown schematically in Fig. 6. The initial vibrational state is promoted to the excited state where it is propagated by the excited state Hamiltonian as a vibrational wavepacket. A downward transition may occur to the final vibrational state on the ground electronic surface producing a resonance Raman
120
P. Kambhampati, A. Campion / Surface Science 427–428 (1999) 115–125
Fig. 6. Schematic of the wavepacket picture of resonance Raman scattering. The increased displacement and curvature of the potential energy surface for the Cu(111) system is schematically represented along an arbitrary normal coordinate.
effect, or back to the initial vibrational state resulting in the electronic absorption spectrum. Vibronic activity in the EEL spectrum and peak intensity in the SER spectrum is governed by the displacement of the excited state potential energy surface (PES) along the various normal coordinate. Since the Cu(100) surface spectrum shows less vibronic activity and has a lower energy vibronic shoulder, we conclude that the interfacial excited state PES for PMDA/Cu(100) is shallower and less displaced than its counterpart on Cu(111). That the relative intensities in the SER spectrum are different on the two surfaces may also be explained by this picture. The multidimensional potential energy surface for the interfacial excitation is displaced to different extents in the full normal coordinate space of the adsorbate [28]. Fig. 7 demonstrates that the simulations successfully reproduce both the EEL spectra and the Raman excitation profile (REP) for the symmetric surface carboxylate mode [27,28]. Furthermore, the Raman enhancement is clearly produced by this interfacial excitation. The EEL spectrum provides a picture of the excited state distortions
Fig. 7. Experimental and simulated Raman excitation profile (top) and electron energy loss spectrum (bottom) of monolayer PMDA adsorbed on Cu(111). The Raman enhancement tracks the interfacial charge-transfer excitation for the symmetric surface carboxylate mode [27,28].
averaged over all coordinates of the PES, whereas the SER spectra provide cuts along each normal coordinate of the excited state PES. The simulations predict [28] that, within the Condon approximation (constant transition dipole moment), the intensity of scattering with p-polarized excitation relative to that with s-polarized excitation should be the same for all totally symmetric modes. The spectra of PMDA adsorbed on the two surfaces using s-polarized excitation shows that this prediction is not accurate. The s-polarized spectra (Fig. 5) are less intense than the p-polarized, as expected, but the relative intensities are different between surfaces and also between polarizations for a given surface. The total intensities of the p-polarized excitation relative to the s-polarized excitation may be used to compute the orientation of the transition dipole moment. Fig. 8 shows the results of this calculation which indicates that, as the transition moment is oriented closer to the surface, the p-polarized
P. Kambhampati, A. Campion / Surface Science 427–428 (1999) 115–125
Fig. 8. The effect of the transition dipole moment orientation on the polarization dependence of Raman intensities as a function of angle from the surface normal, in the range corresponding to our experimental system. The units are arbitrary for I and I but are absolute for I / I . p s p s
scattering becomes diminished, the s-polarized enhanced, and the ratio of p- to s- intensities becomes smaller. Since the p- to s- intensity ratio is smaller for PMDA/Cu(100), we conclude that the transition dipole moment of the interfacial excitation is oriented closer to the surface for this system. The origin of the dependence upon laser polarization for the relative peak intensities was found to be due to non-Condon effects which create asymmetry in the laboratory frame polarizability tensor [28]. Fig. 9 shows the intensity of Raman scattering using p-polarized relative to s-polarized excitation as a function of non-Condon effects perpendicular and parallel to the surface. The dimensionless parameter, D, represents the fraction of non-Condon effects relative to the Franck– Condon term. This figure shows that, as the perpendicular and parallel non-Condon effects change from mode to mode, the p- to s- intensity ratios of each mode show variations as observed experimentally.
5. Lateral spatial localization of charge-transfer excitations Thus far we have referred to the source of the Raman enhancement as an interfacial excitation
121
Fig. 9. The role of non-Condon effects on the polarization dependence of Raman scattering.
because we have not proven its charge-transfer character. The earliest experimental evidence for the importance of charge transfer excitations in the chemical SERS mechanism was provided by electrochemical experiments in which the Raman excitation profiles were shown to depend upon the electrode potential [7,8,12]. These experiments were interpreted by assuming that the energies of the molecular states were unaffected by the applied potential which served only to shift the occupancies of the metal states by uniformly (spatially) raising or lowering the work function (Fermi energy) of the substrate. The direction of charge transfer was determined by the relationship between the change in electrode potential and the resulting spectral shift. Transitions from filled metal orbitals to empty adsorbate orbitals red shift as the Fermi energy is raised; transitions in the opposite direction blue shift ( Fig. 10). The analogous experiment can be conducted in UHV using co-dosed electropositive or electronegative elements to shift the work function of the substrate [29,54]. The EEL spectra of PMDA adsorbed on Cu(100) with various co-doses of Cs are shown in Fig. 11. In the presence of Cs, the PMDA/Cu interfacial excitation observed at 1.9 eV is split into a doublet at 1.9 and 2.5 eV. The energies of the peaks are independent of Cs coverage but the
122
P. Kambhampati, A. Campion / Surface Science 427–428 (1999) 115–125
Fig. 10. An energy level diagram of the states involved in charge transfer. The primed values represent the raised Fermi energy ( lowered work function) which results from the adsorption of Cs.
intensity of the 2.5 eV peak relative to that of the 1.9 eV peak increases with increasing Cs coverage. That a new peak appears which is blue shifted relative to the original peak suggests that the interfacial excitation is in fact charge-transfer-like in character with the direction of transfer being
Fig. 11. The effect of Cs adsorption on the interfacial chargetransfer excitation. A two-dimensional energy level diagram.
molecule to metal. But the observation of a doublet rather than a shifting of the charge-transfer peak is inconsistent with the initial assumption of a spatially uniform raising of the Fermi energy. Both the observation of the doublet and the independence of its peak energies on Cs coverage support the idea that the Cs-induced perturbation is spatially localized laterally and that the chargetransfer coupling is also laterally localized. Recent jellium calculations predict that the length scale of localization is on the order of a few angstroms [54–59]. Thus, at these Cs coverages, the adatoms are essentially non-interacting. Increasing Cs coverage simply provides more sites for interaction, which is manifested in the increased intensity of the Cs related peak, but not to change the energies involved noticeably The length scale of lateral localization can be estimated by lineshape analysis. The peak widths are estimated by fitting each peak with a single Gaussian. As shown in the figure, the width of the original charge transfer loss is ca. 150 meV ( FWHM ), whereas that which results from Cs perturbation is ca. 280 meV. That the second charge transfer peak is fairly symmetric allows a test of the jellium predictions. If the length scale of the lateral interactions were much greater than ˚ (approximately the ‘size’ of a PMDA about 5 A molecule) then molecules located in the second or third coordination circle would still experience some influence, albeit exponentially attenuated. In this situation the EELS lineshape would be asymmetric with a sloping tail on the low energy side of the 2.5 eV charge transfer peak (Fig. 12). Since only an increase in linewidth is observed rather than lineshape asymmetry, we conclude that it is primarily the first ring of PMDA molecules that ‘feels’ the effect of the Cs, but that interactions with molecules in the second coordination ring manifest themselves in the greater linewidths observed. While the length scale of Fermi localization has been examined recently, the length scale of chargetransfer localization is not self-evident. The electron originates in a three-dimensionally spatially localized molecular orbital. It then hops to the metal surface which may be considered twoor three-dimensionally delocalized depending upon
P. Kambhampati, A. Campion / Surface Science 427–428 (1999) 115–125
123
Fig. 12. A two-dimensional energy level diagram. A charge-transfer excitation in the vicinity of Cs produces a blue shifted excitation. Further away from Cs, the unperturbed value is reached. The insets schematically represent how the degree of delocalization of the Cs perturbation and the charge-transfer excitations would manifest themselves in the EELS lineshape.
whether the coupling to the bulk bands is indirect or direct. The metal surface is generally treated as a two-dimensionally delocalized electron gas. Thus the spatial characteristics of the charge-transfer excitation which is inherently composite in nature, is uncertain. Is the excitation localized or delocalized? These experiments suggest that the answer is somewhere in between.
6. Conclusions and prospects The concept of photoinduced charge transfer is not new. Metal–ligand and metal–metal charge transfer complexes have long been studied using resonance Raman spectroscopy in the frequency domain to elucidate the structure of the excited state and pump/probe spectroscopy in the time domain to elucidate the dynamics [60]. Our studies are conceptually similar, with the difference being the replacement of a metal atom with an atomically smooth single crystal metal surface. The electronic coupling of these highly dissimilar systems has been shown to produce very rich and complex interactions. The large degree of dissimilarity of the constituent components and the strength of
coupling makes these charge-transfer excitations model systems in which to study electronic interactions between an adsorbed molecule and a metal surface. These experiments have provided the first experimental evidence for the spatial localization of an electronically excited state of a molecule strongly interacting with a metal surface. Our results support the emerging picture that the effects of alkali adsorption on metal surfaces are local in nature [54–59]. Analysis of our experimental EELS lineshape produces a picture which suggests that the ˚ , which length scale of the interaction is about 5 A is consistent with the results of recent jellium calculations. These results are also consistent with recent STM work demonstrating lateral localization of ground state electronic structure [61–63]. We believe these experiments are the first evidence of the two-dimensional spatial localization of excited state coupling between an adsorbed molecule and an atomically smooth metal surface. Recent experiments in our laboratory have demonstrated the existence of charge-transfer excitations of similar molecules on a copper surface [64]. The differences between these molecules suggest variations in lifetimes within an order of magnitude
124
P. Kambhampati, A. Campion / Surface Science 427–428 (1999) 115–125
and a variable degree of spatial localization normal to the metal surface of these charge-transfer excitations. We found that SERS on atomically smooth surfaces is quite sensitive to surface structure, incident laser polarization, and scattering mode. Specifically, we have shown that the shape and curvature of the excited state potential energy surface depend upon surface structure. We have also shown that the orientation of the transition dipole moment which connects the ground state to this excited state, as well as the nuclear coordinate dependence of the transition dipole moment, are sensitive to surface structure. Non-Condon effects may be probed by polarization resolved Raman experiments. The differences in the Raman and EEL spectra obtained for the two crystal faces clearly imply that a microscopic understanding of the local electronic structure of the adsorbate– substrate complex is required for a complete understanding of charge-transfer excitations and the chemical mechanism of SERS. We believe that these results are novel and important for surface studies. For some time resonance Raman spectroscopy has been used to understand excited state structure and dynamics in the gas phase and solution. It now appears that this technique can be profitably applied to molecules adsorbed on single crystal surfaces. Very little is known about the excited state interactions between an adsorbed molecule and a metal surface, despite their importance in the areas described in the Introduction. We believe the methodology described in this paper will be valuable in the general effort to understand the manner in which the electronic states of the adsorbate couple to those of the substrate. An accurate picture of the excited state potential energy surfaces of adsorbed molecules is important from both a fundamental physical point of view, as well as for providing an increased understanding of surface chemical reactions.
Acknowledgements We thank the National Science Foundation and the Welch Foundation for financial support.
References [1] D.H. Waldeck, A.P. Alivisatos, C.B. Harris, Surf. Sci. 158 (1985) 103. [2] Ph. Avouris, R.E. Walkup, Annu. Rev. Phys. Chem. 40 (1989) 173. [3] X.L. Zhou, X.-Y. Zhu, J.M. White, Surf. Sci. Rep. 13 (1991) 73. [4] X.-Y. Zhu, Annu. Rev. Phys. Chem. 45 (1994) 113. [5] N. Serpone, E. Pelizetti (Eds.), in: Photocatalysis: Fundamentals and Applications, Wiley, New York, 1989. [6 ] I.P. Herman, Chem. Rev. 89 (1989) 1323. [7] M. Moskovits, Rev. Mod. Phys. 57 (1985) 783. [8] B. Pettinger, in: J. Lipkowski, P.N. Ross (Eds.), Adsorption of Molecules at Metal Electrodes, VCH, New York, 1992, p. 285. [9] A. Campion, P. Kambhampati, Chem. Soc. Rev. 27 (1998) 241. [10] I. Mrozek, A. Otto, J. Electron Spectrosc. Relat. Phenom. 54 (1990) 895. [11] Ph. Avouris, J.E. Demuth, J. Chem. Phys. 75 (1981) 4783. [12] J.R. Lombardi, R.L. Birke, T. Lu, J. Xu, J. Chem. Phys. 84 (1986) 4174. [13] J.D. Jiang, E. Burstein, H. Kobayashi, Phys. Rev. Lett. 57 (1986) 1793. [14] H. Yamada, H. Nagata, K. Toba, Y. Nakao, Surf. Sci. 182 (1987) 269. [15] H. Park, S.B. Lee, K. Kim, M.S. Kim, J. Phys. Chem. 94 (1990) 7576. [16 ] A. Rosenberg, D.P. DiLella, Chem. Phys. Lett. 223 (1994) 76. [17] M. Osawa, N. Matsuda, K. Yoshii, I. Uchida, J. Phys. Chem. 98 (1994) 12702. [18] J.C. Rubim, P. Corio, M.C.C. Ribeiro, M. Matz, J. Phys. Chem. 99 (1995) 15765. [19] J.F. Arenas, I. Lopez Tocon, J.C. Otero, J.I. Marcos, J. Phys. Chem. 100 (1996) 9254. [20] J.F. Arenas, M.S. Wooley, J.C. Otero, J.I. Marcos, J. Phys. Chem. 100 (1996) 3199. [21] A.G. Brolo, D.E. Irish, J. Lipkowski, J. Phys. Chem. B 101 (1997) 3906. [22] A. Kudelski, J. Bukowski, Chem. Phys. Lett. 253 (1996) 246. [23] J.E. Ivanecky III, C.M. Child, A. Campion, Surf. Sci. 325 (1995) L428. [24] C.M. Child, J.E. Fieberg, A. Campion, Surf. Sci. 372 (1997) L254. [25] S.S. Perry, A. Campion, Surf. Sci. 234 (1990) L275. [26 ] A. Campion, J.E. Ivanecky III, C.M. Child, M. Foster, J. Am. Chem. Soc. 117 (1995) 11807. [27] P. Kambhampati, C.M. Child, A. Campion, J. Chem. Soc., Faraday Trans. 92 (1996) 4775. [28] P. Kambhampati, C.M. Child, M.C. Foster, A. Campion, J. Chem. Phys. 108 (1998) 5013. [29] P. Kambhampati, M.C. Foster, A. Campion, J. Chem. Phys. 110 (1999) 551.
P. Kambhampati, A. Campion / Surface Science 427–428 (1999) 115–125 [30] H. Ibach, D.L. Mills, Electron Energy Loss Spectroscopy and Surface Vibrations, Academic Press, New York, 1982. [31] H.-J. Freund, R.P. Messmer, W. Spiess, H. Behner, G. Wedler, C.M. Kao, Phys. Rev. B 33 (1986) 5228. [32] J.F. Ying, K.T. Leung, J. Chem. Phys. 100 (1994) 1011. [33] M. Allan, N.J. Mason, J.A. Davies, J. Chem. Phys. 105 (1996) 5665. [34] A. Campion, in: J.T. Yates Jr., T.E. Madey ( Eds.), Vibrational Spectroscopy of Molecules on Surfaces, Plenum, New York, 1987, p. 345. [35] B.G. Frederick, M.R. Ashton, N.V. Richardson, T.S. Jones, Surf. Sci. 292 (1993) 33. [36 ] M.R. Ashton, T.S. Jones, N.V. Richardson, R.G. Mack, W.N. Unertl, J. Electron Spectrosc. Relat. Phenom. 54 (1990) 1133. [37] B.N.J. Persson, Chem. Phys. Lett. 82 (1981) 561. [38] K. Arya, R. Zeyher, Phys. Rev. B 24 (1981) 1852. [39] F.J. Adrian, J. Chem. Phys. 77 (1982) 5302. [40] H. Ueba, Surf. Sci. 131 (1983) 347. [41] S.D. Kevan, N.G. Stoffel, N.V. Smith, Phys. Rev. B 31 (1985) 3348. [42] P. Heimann, J. Hermanson, H. Miogsa, H. Neddermeyer, Phys. Rev. B 20 (1979) 3059. [43] P. Heimann, H. Neddermeyer, H.F. Roloff, J. Phys. C 10 (1977) L17. [44] P.O. Gartland, B.J. Slagsvold, Phys. Rev. B 12 (1975) 4047. [45] S.D. Kevan, W. Eberhardt, in: S.D. Kevan ( Ed.), Angle Resolved Photoemission, Elsevier, New York, 1992, p. 99. [46 ] R.W. Schoenlein, J.G. Fujimoto, G.L. Eesley, T.W. Capehart, Phys. Rev. B 43 (1991) 4688. [47] M. Wolf, Surf. Sci. 377 (1997) 343.
125
[48] N.-H. Ge, C.M. Wong, R.L. Lingle Jr., J.D. McNeill, K.J. Gaffney, C.B. Harris, Science 279 (1998) 202. [49] W. Steinmann, Th. Fauster, in: H.L. Dai, W. Ho ( Eds.), Laser Spectroscopy and Photochemistry on Metal Surfaces, World Scientific, Singapore, 1995, p. 184. [50] S.Y. Lee, E.J. Heller, J. Chem. Phys. 71 (1979) 4777. [51] E.J. Heller, Acc. Chem. Res. 14 (1981) 368. [52] E.J. Heller, R.L. Sundberg, D. Tannor, J. Phys. Chem. 86 (1982) 1822. [53] A.B. Myers, R.A. Mathies, T.G. Spiro (Ed.), Biological Applications of Raman Spectroscopy 2 , Wiley, New York, 1987, p. 1. [54] H.P. Bonzel, A.M. Bradshaw, G. Ertl ( Eds.), in: Physics and Chemistry of Alkali Metal Adsorption, Elsevier, New York, 1989. [55] P. Nordlander, N.D. Lang, Phys. Rev. B 44 (1991) 13681. [56 ] N.D. Lang, S. Holloway, J.K. Norskov, Surf. Sci. 150 (1985) 24. [57] H. Ishida, Phys. Rev. B 38 (1988) 8006. [58] C.C. Hsu, H. Bu, J.W. Rabalais, P. Nordlander, Phys. Rev. Lett. 69 (1992) 188. [59] G.A. Kimmel, D.M. Goodstein, Z.H. Levine, B.H. Cooper, Phys. Rev. B 43 (1991) 9403. [60] P.F. Barbara, T.J. Meyer, M.A. Ratner, J. Phys. Chem. 100 (1996) 13148. [61] Ph. Avouris, I.W. Lyo, Science 264 (1994) 942. [62] D.M. Eigler, P.S. Weiss, E.K. Scheizer, N.D. Lang, Phys. Rev. Lett. 66 (1991) 1189. [63] M.F. Crommie, C.P. Lutz, D.M. Eigler, Science 262 (1993) 218. [64] P. Kambhampati, C.M. Child, M.C. Foster, A. Campion, submitted to J. Chem. Phys.
www.elsevier.nl/locate/susc
Surface enhanced Raman scattering as a probe of adsorbate–substrate charge-transfer excitations Patanjali Kambhampati, Alan Campion * Department of Chemistry and Biochemistry, The University of Texas at Austin, Austin, TX 78712, USA
Abstract The chemical mechanism of surface enhanced Raman scattering (SERS) is used as a probe of adsorbate–substrate charge-transfer excitations on atomically smooth, single crystal copper surfaces in ultra-high vacuum. It is demonstrated that SERS and charge-transfer excitations are supported on atomically smooth surfaces in contradiction to earlier consensus. The combination of SERS with electron energy loss spectroscopy (EELS ) provides a great deal of insight into the nature of these excitations. It is found that these excitations are sensitive probes of local electronic structure at metal surfaces. The spatial extent, transition moment, and excited state potential energy surfaces of these excitations are probed using SERS, EELS and molecular spectroscopy simulations. © 1999 Elsevier Science B.V. All rights reserved. Keywords: Alkali metals; Cesium; Charge transfer; Copper; Electron energy loss spectroscopy; Excited states; Fermi energy; Image potential states; Localization; Metal surfaces; Pyromellitic dianhydride; Surface enhanced Raman scattering; Surface states; Work function
1. Introduction Much is now known about the properties of the ground electronic states of molecules adsorbed on solid surfaces, due, in large part, to the tremendous power of modern surface science experimental and theoretical approaches. We know considerably less about the properties of the excited electronic states of adsorbed species and the manner in which they couple to the substrate, however, due to the relative paucity of techniques available to probe such states. Absorption spectroscopy, whether by reflectance, ellipsometry or inelastic electron scattering, generally yields only broad, featureless spectra because excited state lifetimes are dramatically shortened by rapid non* Corresponding author. Fax: +1 512-4718696. E-mail address: [email protected] (A. Campion)
radiative energy transfer to the surface [1]. The importance of electronically excited states in both physical and chemical processes at surfaces can hardly be overstated, though. They participate in DIET (desorption induced by electronic transitions) [2], photochemistry [3,4] and photoelectrochemistry [5] at surfaces, and photon-assisted etching of electronic materials [6 ]. Surface photochemistry is also of great importance in the surface science of heterogeneous catalysis where it can be used to effect the selective synthesis of adsorbed fragments which cannot be made by other means [3,4]. While it is true that some understanding of the nature of electronic excited states has come from the comprehensive body of DIET work, these states are, by definition, dissociative and these experiments have generally been interpreted using only one-dimensional models. There is clearly an
0039-6028/99/$ – see front matter © 1999 Elsevier Science B.V. All rights reserved. PII: S0 0 39 - 6 0 28 ( 99 ) 0 02 4 5 -9
116
P. Kambhampati, A. Campion / Surface Science 427–428 (1999) 115–125
important need to achieve, for adsorbed molecules, a comprehensive understanding of the full normal coordinate dependence of the excited state potential surface which we have for a number of molecules in the gas phase and in solution. In addition, we seek information about excited state lifetimes, couplings to other states and the orientations of transition dipole moments for adsorbed species. We describe here experiments which use surface enhanced Raman scattering (SERS) and electron energy loss spectroscopy ( EELS ) along with molecular spectroscopy simulations to achieve such an understanding for a general class of excited states, those which arise from adsorbate–substrate charge-transfer excitations. These excitations are significant in that they are model systems with which to study excited state interactions between molecules and metal surfaces, and for their importance in the chemical mechanism of surface enhanced Raman scattering. When adsorbed on some metal surfaces such as silver, copper and gold, molecules can exhibit enormously enhanced Raman scattering which has become known as surface enhanced Raman scattering [7–9]. The SERS effect has historically been associated with substrate roughness on two characteristic length scales. Surface roughness on the 10 to 100 nm length scale supports the electromagnetic resonances which are the dominant mechanism of enhancement. These electromagnetic resonances can increase the scattered intensity by ca. 104 to 107. A second mechanism, often thought to require atomic scale roughness, is referred to as the chemical mechanism. This second mechanism involves the creation of new electronic excited states which result from adsorbate–substrate chemical interactions. It is estimated that the chemical mechanism can enhance the scattering crosssection by a factor of ca. 10 to 102. Estimates have been made of the magnitude of the chemical mechanism in the presence of the electromagnetic mechanism in silver island films in ultra-high vacuum ( UHV ) [10]. These two mechanisms operate simultaneously (multiplicatively), making it difficult to isolate the role and magnitude of each mechanism. The chemical mechanism of SERS is a process analogous to molecular resonance Raman scatter-
ing except that resonances involved are not intramolecular in origin [9–12]. They are chargetransfer resonances in which there is photoinduced charge hopping between the adsorbate and the substrate. The general consensus had been that atomic scale roughness is required to couple electronically the adsorbate states to the substrate states in such a manner as to produce chargetransfer excitations and SERS [7,8]. It has proven to be very difficult to study the chemical enhancement mechanism selectively for two reasons. First, the mechanism contributes a small fraction to the total enhancement. Second, almost any experimental parameter which can be varied to probe a system will have an influence via both mechanisms, making the separation of effects difficult, if not impossible. For this reason, there has been considerable effort to understand the effect in detail [9–22]. The use of single crystal substrates can provide a great deal of insight into the problem. We can minimize the inhomogeneous broadening present in the roughened electrode experiments that were used to assign the nature of the intermediate state in SERS [7,8,12]. These ordered substrates also facilitate a more detailed understanding of the coupling between the adsorbed molecule and the metal surface such as the dynamics, spatial and electronic structure of these excited states. It is extremely important to understand the chemical mechanism both for its relevance to analytical applications and to achieve a fundamental understanding of the electronic coupling between an adsorbed molecule and a metal surface.
2. Experiment The experimental methods used have been described in detail previously [23–29] and will be only briefly discussed here. The surfaces were polished by standard metallographic techniques and were mounted in UHV chambers with base pressures ca. 1×10−10 Torr, equipped with the standard surface analysis tools. A Leybold–Heraeus ELS-22 high resolution electron energy loss (HREEL) spectrometer was used for the electronic EELS studies. Primary beam energies between 10 and 12 eV were used
P. Kambhampati, A. Campion / Surface Science 427–428 (1999) 115–125
and the instrumental resolution (convolution of both the monochromator and analyzer stages) was ca. 50 meV ( FWHM ). We have used EELS to measure the electronic spectra reported herein, and have verified the dipolar scattering mechanism by angular dependence studies [30]; the use of EELS to obtain electronic spectra for both adsorbed molecules [11,30,31] and molecules in the gas phase [32,33] is well documented. Raman spectra were obtained using ~150 mW of 647 nm light for Cu(100) and ~300 mW of 647, 660, 693, 725 nm light for Cu(111), all produced by a Coherent Innova-200 Ar+ laser pumping Kiton Red or Pyridine-2 (Exciton) in a Coherent Cr-599 dye laser. The sample temperature, as monitored by a chromel–alumel thermocouple, did not change upon illumination. The configuration and detailed descriptions of the optics and detection system have been discussed in more detail previously [23,27,34]. On our apparatus, the background count rates are typically ca. 10 counts s−1 W−1 for non-luminescent substrates such as Ag excited with 488 nm light and ca. 50 counts s−1 W−1 for luminescent substrates such as Cu excited with 647 nm light. An unenhanced Raman experiment on a molecule such as benzene typically produces ca. 1 counts s−1 W−1 (peak) or 30 counts s−1 W−1 (integrated area), whereas the PMDA/Cu system typically produces ca. 100 counts s−1 W−1 (peak) or 3000 counts s−1
117
W−1 (integrated area). All spectra were acquired in 1000 s scans with the luminescent background subtracted. For the Raman excitation profile of the mode reported here, required corrections were made for the filter and monochromator throughput and detector efficiency.
3. Charge-transfer excitations and SERS Motivated by an interest in studying adsorption and polymerization reactions of molecular dielectrics on technologically important surfaces, our laboratory conducted surface Raman experiments of pyromellitic dianhydride adsorption on Cu(111) [23]. The chemical interaction of PMDA with copper surfaces is now well understood on a structural level [23–25,35,36 ]. PMDA dissociatively chemisorbs by splitting out CO, producing a surface carboxylate. The chemisorbed carboxylate plane is tilted away from the surface normal with its oxygen atoms inequivalently bound to the surface, whereas the plane containing the aromatic portion of the molecule is oriented perpendicular to the plane of the carboxylate (Fig. 1). It has been suggested that this geometry, which tilts the long axis towards the surface, allows an unsatisfied valence on the aromatic portion of the molecule to interact with the surface copper atoms. In conducting the Raman experiments, we
Fig. 1. The chemical structure of pyromellitic dianhydride and the geometry of the adsorbed surface carboxylate. The black atoms are carbon, the gray are oxygen and the white are hydrogen.
118
P. Kambhampati, A. Campion / Surface Science 427–428 (1999) 115–125
found that the count rates were significantly greater than those of all other physisorbed and chemisorbed molecules that we had studied [23,26– 28]. The surface Raman spectra of monolayer coverage PMDA adsorbed on Cu(111) are shown in Fig. 2 at two excitation wavelengths. It is readily apparent that the total intensity is strongly dependent upon the excitation wavelength, in contrast to normal, unenhanced surface Raman scattering [34]. That the total intensity at 647 nm excitation is on the order of 100 to 200 times more intense than most physisorbed or chemisorbed molecules is the clearest evidence of an operative enhancement mechanism. Consistent with a resonance type of effect, the relative intensities are also dependent upon excitation wavelength. It is surprising, however, to encounter a resonance when exciting a small aromatic molecule with visible light. PMDA is a colorless compound with its lowest energy transitions in the ultraviolet. In order to identify the resonance and thus the origin of the enhancement, we performed electron energy loss spectroscopy ( EELS) experiments on the Cu(111) and Cu(100) surfaces [26–28] (Fig. 3). Both surfaces show an intense, narrow loss feature at 1.9 eV which matches the laser wavelength of 647 nm. This feature saturates at monolayer coverage whereas the three higher energy peaks increase in intensity at all coverage. Therefore, the low energy 1.9 eV peak is interfacial in character and responsible for the operative
Fig. 2. Raman spectra of one monolayer PMDA adsorbed on the Cu(111) surface. p-Polarized excitation at 725 nm (a) and 647 nm (b) was used.
Fig. 3. Electron energy loss (EEL) spectra of one monolayer of PMDA adsorbed on the Cu(111) and Cu(100) surfaces.
enhancement mechanism, whereas the higher energy peaks are simply the intramolecular excitations of the adsorbed carboxylate at monolayer coverage and the physisorbed parent for subsequent multilayers.
4. Surface structure and laser polarization dependence of SERS We initially hypothesized that the low energy interfacial excitation was a charge-transfer excitation by analogy to the previous experiments in the field [7–12]. We recently confirmed this hypothesis by experiments in which we were able to vary the Fermi energy of the substrate independently of the energies of the adsorbate molecular orbitals [29]. The idea of charge-transfer has been invoked to explain the potential dependence of the enhancement of molecules adsorbed on roughened electrodes. In those experiments the chargetransfer linewidths were in the range of 500 to 1000 meV. Charge-transfer had never been observed, however, on atomically smooth surfaces, making the nature of the interfacial excitation we observed unclear. In adsorbate–substrate chargetransfer excitations, it was generally assumed that there was photogenerated charge hopping between the adsorbate and the manifold of states available in the bulk band structure [7,8,11,12,37–40]. The
P. Kambhampati, A. Campion / Surface Science 427–428 (1999) 115–125
119
bulk density of states was considered a source of the large linewidth. Since the interfacial excitations we observed were so narrow, we performed experiments on two low index faces of copper to pursue analogies to surface states [41–45] and image potential states [45–49] which are also quite narrow and with properties that depend upon crystal face. The EELS data show, that while superficially similar, the excitations are different on both surfaces. Fig. 4 more clearly shows the differences between the two surfaces. While the energies and widths are nearly the same, the vibronic shoulder is more pronounced and appears at higher energy on the Cu(111) surface. The Raman spectra are superficially similar but show subtle and meaningful differences. Fig. 5 shows that the Raman spectra on the two surfaces are different when excited with p-polarized light at 647 nm. The total intensity on Cu(111) is ca. twice that on Cu(100). Additionally, the relative intensities are dependent on the crystal face. We see that, by analogy to surface and image potential states, the interfacial Fig. 5. Raman spectra of monolayer PMDA adsorbed on Cu(111) and Cu(100) at 647 nm using both p-polarized and s-polarized excitation.
Fig. 4. EEL spectra of monolayer PMDA adsorbed on Cu(111) and Cu(100) in the region of the interfacial charge-transfer excitation. The Cu(111) system has a more intense and higher energy vibronic shoulder.
excitation is sensitive to crystal face and to the local electronic structure at the surface. To further understand the role of the interfacial excitation in the enhancement process, we performed simulations [27,28] which utilize Heller and coworkers’ wavepacket picture of molecular spectroscopy [50–53]. In these simulations, the standard Kramers–Heisenberg–Dirac dispersion equation for the molecular polarizability tensor is recast in the time domain. Franck–Condon factors are evaluated by computing the full or half Fourier transforms of the appropriate autocorrelation functions of the nuclear wavepackets as they evolve on the excited state potential energy surface. The transitions and relevant potentials are shown schematically in Fig. 6. The initial vibrational state is promoted to the excited state where it is propagated by the excited state Hamiltonian as a vibrational wavepacket. A downward transition may occur to the final vibrational state on the ground electronic surface producing a resonance Raman
120
P. Kambhampati, A. Campion / Surface Science 427–428 (1999) 115–125
Fig. 6. Schematic of the wavepacket picture of resonance Raman scattering. The increased displacement and curvature of the potential energy surface for the Cu(111) system is schematically represented along an arbitrary normal coordinate.
effect, or back to the initial vibrational state resulting in the electronic absorption spectrum. Vibronic activity in the EEL spectrum and peak intensity in the SER spectrum is governed by the displacement of the excited state potential energy surface (PES) along the various normal coordinate. Since the Cu(100) surface spectrum shows less vibronic activity and has a lower energy vibronic shoulder, we conclude that the interfacial excited state PES for PMDA/Cu(100) is shallower and less displaced than its counterpart on Cu(111). That the relative intensities in the SER spectrum are different on the two surfaces may also be explained by this picture. The multidimensional potential energy surface for the interfacial excitation is displaced to different extents in the full normal coordinate space of the adsorbate [28]. Fig. 7 demonstrates that the simulations successfully reproduce both the EEL spectra and the Raman excitation profile (REP) for the symmetric surface carboxylate mode [27,28]. Furthermore, the Raman enhancement is clearly produced by this interfacial excitation. The EEL spectrum provides a picture of the excited state distortions
Fig. 7. Experimental and simulated Raman excitation profile (top) and electron energy loss spectrum (bottom) of monolayer PMDA adsorbed on Cu(111). The Raman enhancement tracks the interfacial charge-transfer excitation for the symmetric surface carboxylate mode [27,28].
averaged over all coordinates of the PES, whereas the SER spectra provide cuts along each normal coordinate of the excited state PES. The simulations predict [28] that, within the Condon approximation (constant transition dipole moment), the intensity of scattering with p-polarized excitation relative to that with s-polarized excitation should be the same for all totally symmetric modes. The spectra of PMDA adsorbed on the two surfaces using s-polarized excitation shows that this prediction is not accurate. The s-polarized spectra (Fig. 5) are less intense than the p-polarized, as expected, but the relative intensities are different between surfaces and also between polarizations for a given surface. The total intensities of the p-polarized excitation relative to the s-polarized excitation may be used to compute the orientation of the transition dipole moment. Fig. 8 shows the results of this calculation which indicates that, as the transition moment is oriented closer to the surface, the p-polarized
P. Kambhampati, A. Campion / Surface Science 427–428 (1999) 115–125
Fig. 8. The effect of the transition dipole moment orientation on the polarization dependence of Raman intensities as a function of angle from the surface normal, in the range corresponding to our experimental system. The units are arbitrary for I and I but are absolute for I / I . p s p s
scattering becomes diminished, the s-polarized enhanced, and the ratio of p- to s- intensities becomes smaller. Since the p- to s- intensity ratio is smaller for PMDA/Cu(100), we conclude that the transition dipole moment of the interfacial excitation is oriented closer to the surface for this system. The origin of the dependence upon laser polarization for the relative peak intensities was found to be due to non-Condon effects which create asymmetry in the laboratory frame polarizability tensor [28]. Fig. 9 shows the intensity of Raman scattering using p-polarized relative to s-polarized excitation as a function of non-Condon effects perpendicular and parallel to the surface. The dimensionless parameter, D, represents the fraction of non-Condon effects relative to the Franck– Condon term. This figure shows that, as the perpendicular and parallel non-Condon effects change from mode to mode, the p- to s- intensity ratios of each mode show variations as observed experimentally.
5. Lateral spatial localization of charge-transfer excitations Thus far we have referred to the source of the Raman enhancement as an interfacial excitation
121
Fig. 9. The role of non-Condon effects on the polarization dependence of Raman scattering.
because we have not proven its charge-transfer character. The earliest experimental evidence for the importance of charge transfer excitations in the chemical SERS mechanism was provided by electrochemical experiments in which the Raman excitation profiles were shown to depend upon the electrode potential [7,8,12]. These experiments were interpreted by assuming that the energies of the molecular states were unaffected by the applied potential which served only to shift the occupancies of the metal states by uniformly (spatially) raising or lowering the work function (Fermi energy) of the substrate. The direction of charge transfer was determined by the relationship between the change in electrode potential and the resulting spectral shift. Transitions from filled metal orbitals to empty adsorbate orbitals red shift as the Fermi energy is raised; transitions in the opposite direction blue shift ( Fig. 10). The analogous experiment can be conducted in UHV using co-dosed electropositive or electronegative elements to shift the work function of the substrate [29,54]. The EEL spectra of PMDA adsorbed on Cu(100) with various co-doses of Cs are shown in Fig. 11. In the presence of Cs, the PMDA/Cu interfacial excitation observed at 1.9 eV is split into a doublet at 1.9 and 2.5 eV. The energies of the peaks are independent of Cs coverage but the
122
P. Kambhampati, A. Campion / Surface Science 427–428 (1999) 115–125
Fig. 10. An energy level diagram of the states involved in charge transfer. The primed values represent the raised Fermi energy ( lowered work function) which results from the adsorption of Cs.
intensity of the 2.5 eV peak relative to that of the 1.9 eV peak increases with increasing Cs coverage. That a new peak appears which is blue shifted relative to the original peak suggests that the interfacial excitation is in fact charge-transfer-like in character with the direction of transfer being
Fig. 11. The effect of Cs adsorption on the interfacial chargetransfer excitation. A two-dimensional energy level diagram.
molecule to metal. But the observation of a doublet rather than a shifting of the charge-transfer peak is inconsistent with the initial assumption of a spatially uniform raising of the Fermi energy. Both the observation of the doublet and the independence of its peak energies on Cs coverage support the idea that the Cs-induced perturbation is spatially localized laterally and that the chargetransfer coupling is also laterally localized. Recent jellium calculations predict that the length scale of localization is on the order of a few angstroms [54–59]. Thus, at these Cs coverages, the adatoms are essentially non-interacting. Increasing Cs coverage simply provides more sites for interaction, which is manifested in the increased intensity of the Cs related peak, but not to change the energies involved noticeably The length scale of lateral localization can be estimated by lineshape analysis. The peak widths are estimated by fitting each peak with a single Gaussian. As shown in the figure, the width of the original charge transfer loss is ca. 150 meV ( FWHM ), whereas that which results from Cs perturbation is ca. 280 meV. That the second charge transfer peak is fairly symmetric allows a test of the jellium predictions. If the length scale of the lateral interactions were much greater than ˚ (approximately the ‘size’ of a PMDA about 5 A molecule) then molecules located in the second or third coordination circle would still experience some influence, albeit exponentially attenuated. In this situation the EELS lineshape would be asymmetric with a sloping tail on the low energy side of the 2.5 eV charge transfer peak (Fig. 12). Since only an increase in linewidth is observed rather than lineshape asymmetry, we conclude that it is primarily the first ring of PMDA molecules that ‘feels’ the effect of the Cs, but that interactions with molecules in the second coordination ring manifest themselves in the greater linewidths observed. While the length scale of Fermi localization has been examined recently, the length scale of chargetransfer localization is not self-evident. The electron originates in a three-dimensionally spatially localized molecular orbital. It then hops to the metal surface which may be considered twoor three-dimensionally delocalized depending upon
P. Kambhampati, A. Campion / Surface Science 427–428 (1999) 115–125
123
Fig. 12. A two-dimensional energy level diagram. A charge-transfer excitation in the vicinity of Cs produces a blue shifted excitation. Further away from Cs, the unperturbed value is reached. The insets schematically represent how the degree of delocalization of the Cs perturbation and the charge-transfer excitations would manifest themselves in the EELS lineshape.
whether the coupling to the bulk bands is indirect or direct. The metal surface is generally treated as a two-dimensionally delocalized electron gas. Thus the spatial characteristics of the charge-transfer excitation which is inherently composite in nature, is uncertain. Is the excitation localized or delocalized? These experiments suggest that the answer is somewhere in between.
6. Conclusions and prospects The concept of photoinduced charge transfer is not new. Metal–ligand and metal–metal charge transfer complexes have long been studied using resonance Raman spectroscopy in the frequency domain to elucidate the structure of the excited state and pump/probe spectroscopy in the time domain to elucidate the dynamics [60]. Our studies are conceptually similar, with the difference being the replacement of a metal atom with an atomically smooth single crystal metal surface. The electronic coupling of these highly dissimilar systems has been shown to produce very rich and complex interactions. The large degree of dissimilarity of the constituent components and the strength of
coupling makes these charge-transfer excitations model systems in which to study electronic interactions between an adsorbed molecule and a metal surface. These experiments have provided the first experimental evidence for the spatial localization of an electronically excited state of a molecule strongly interacting with a metal surface. Our results support the emerging picture that the effects of alkali adsorption on metal surfaces are local in nature [54–59]. Analysis of our experimental EELS lineshape produces a picture which suggests that the ˚ , which length scale of the interaction is about 5 A is consistent with the results of recent jellium calculations. These results are also consistent with recent STM work demonstrating lateral localization of ground state electronic structure [61–63]. We believe these experiments are the first evidence of the two-dimensional spatial localization of excited state coupling between an adsorbed molecule and an atomically smooth metal surface. Recent experiments in our laboratory have demonstrated the existence of charge-transfer excitations of similar molecules on a copper surface [64]. The differences between these molecules suggest variations in lifetimes within an order of magnitude
124
P. Kambhampati, A. Campion / Surface Science 427–428 (1999) 115–125
and a variable degree of spatial localization normal to the metal surface of these charge-transfer excitations. We found that SERS on atomically smooth surfaces is quite sensitive to surface structure, incident laser polarization, and scattering mode. Specifically, we have shown that the shape and curvature of the excited state potential energy surface depend upon surface structure. We have also shown that the orientation of the transition dipole moment which connects the ground state to this excited state, as well as the nuclear coordinate dependence of the transition dipole moment, are sensitive to surface structure. Non-Condon effects may be probed by polarization resolved Raman experiments. The differences in the Raman and EEL spectra obtained for the two crystal faces clearly imply that a microscopic understanding of the local electronic structure of the adsorbate– substrate complex is required for a complete understanding of charge-transfer excitations and the chemical mechanism of SERS. We believe that these results are novel and important for surface studies. For some time resonance Raman spectroscopy has been used to understand excited state structure and dynamics in the gas phase and solution. It now appears that this technique can be profitably applied to molecules adsorbed on single crystal surfaces. Very little is known about the excited state interactions between an adsorbed molecule and a metal surface, despite their importance in the areas described in the Introduction. We believe the methodology described in this paper will be valuable in the general effort to understand the manner in which the electronic states of the adsorbate couple to those of the substrate. An accurate picture of the excited state potential energy surfaces of adsorbed molecules is important from both a fundamental physical point of view, as well as for providing an increased understanding of surface chemical reactions.
Acknowledgements We thank the National Science Foundation and the Welch Foundation for financial support.
References [1] D.H. Waldeck, A.P. Alivisatos, C.B. Harris, Surf. Sci. 158 (1985) 103. [2] Ph. Avouris, R.E. Walkup, Annu. Rev. Phys. Chem. 40 (1989) 173. [3] X.L. Zhou, X.-Y. Zhu, J.M. White, Surf. Sci. Rep. 13 (1991) 73. [4] X.-Y. Zhu, Annu. Rev. Phys. Chem. 45 (1994) 113. [5] N. Serpone, E. Pelizetti (Eds.), in: Photocatalysis: Fundamentals and Applications, Wiley, New York, 1989. [6 ] I.P. Herman, Chem. Rev. 89 (1989) 1323. [7] M. Moskovits, Rev. Mod. Phys. 57 (1985) 783. [8] B. Pettinger, in: J. Lipkowski, P.N. Ross (Eds.), Adsorption of Molecules at Metal Electrodes, VCH, New York, 1992, p. 285. [9] A. Campion, P. Kambhampati, Chem. Soc. Rev. 27 (1998) 241. [10] I. Mrozek, A. Otto, J. Electron Spectrosc. Relat. Phenom. 54 (1990) 895. [11] Ph. Avouris, J.E. Demuth, J. Chem. Phys. 75 (1981) 4783. [12] J.R. Lombardi, R.L. Birke, T. Lu, J. Xu, J. Chem. Phys. 84 (1986) 4174. [13] J.D. Jiang, E. Burstein, H. Kobayashi, Phys. Rev. Lett. 57 (1986) 1793. [14] H. Yamada, H. Nagata, K. Toba, Y. Nakao, Surf. Sci. 182 (1987) 269. [15] H. Park, S.B. Lee, K. Kim, M.S. Kim, J. Phys. Chem. 94 (1990) 7576. [16 ] A. Rosenberg, D.P. DiLella, Chem. Phys. Lett. 223 (1994) 76. [17] M. Osawa, N. Matsuda, K. Yoshii, I. Uchida, J. Phys. Chem. 98 (1994) 12702. [18] J.C. Rubim, P. Corio, M.C.C. Ribeiro, M. Matz, J. Phys. Chem. 99 (1995) 15765. [19] J.F. Arenas, I. Lopez Tocon, J.C. Otero, J.I. Marcos, J. Phys. Chem. 100 (1996) 9254. [20] J.F. Arenas, M.S. Wooley, J.C. Otero, J.I. Marcos, J. Phys. Chem. 100 (1996) 3199. [21] A.G. Brolo, D.E. Irish, J. Lipkowski, J. Phys. Chem. B 101 (1997) 3906. [22] A. Kudelski, J. Bukowski, Chem. Phys. Lett. 253 (1996) 246. [23] J.E. Ivanecky III, C.M. Child, A. Campion, Surf. Sci. 325 (1995) L428. [24] C.M. Child, J.E. Fieberg, A. Campion, Surf. Sci. 372 (1997) L254. [25] S.S. Perry, A. Campion, Surf. Sci. 234 (1990) L275. [26 ] A. Campion, J.E. Ivanecky III, C.M. Child, M. Foster, J. Am. Chem. Soc. 117 (1995) 11807. [27] P. Kambhampati, C.M. Child, A. Campion, J. Chem. Soc., Faraday Trans. 92 (1996) 4775. [28] P. Kambhampati, C.M. Child, M.C. Foster, A. Campion, J. Chem. Phys. 108 (1998) 5013. [29] P. Kambhampati, M.C. Foster, A. Campion, J. Chem. Phys. 110 (1999) 551.
P. Kambhampati, A. Campion / Surface Science 427–428 (1999) 115–125 [30] H. Ibach, D.L. Mills, Electron Energy Loss Spectroscopy and Surface Vibrations, Academic Press, New York, 1982. [31] H.-J. Freund, R.P. Messmer, W. Spiess, H. Behner, G. Wedler, C.M. Kao, Phys. Rev. B 33 (1986) 5228. [32] J.F. Ying, K.T. Leung, J. Chem. Phys. 100 (1994) 1011. [33] M. Allan, N.J. Mason, J.A. Davies, J. Chem. Phys. 105 (1996) 5665. [34] A. Campion, in: J.T. Yates Jr., T.E. Madey ( Eds.), Vibrational Spectroscopy of Molecules on Surfaces, Plenum, New York, 1987, p. 345. [35] B.G. Frederick, M.R. Ashton, N.V. Richardson, T.S. Jones, Surf. Sci. 292 (1993) 33. [36 ] M.R. Ashton, T.S. Jones, N.V. Richardson, R.G. Mack, W.N. Unertl, J. Electron Spectrosc. Relat. Phenom. 54 (1990) 1133. [37] B.N.J. Persson, Chem. Phys. Lett. 82 (1981) 561. [38] K. Arya, R. Zeyher, Phys. Rev. B 24 (1981) 1852. [39] F.J. Adrian, J. Chem. Phys. 77 (1982) 5302. [40] H. Ueba, Surf. Sci. 131 (1983) 347. [41] S.D. Kevan, N.G. Stoffel, N.V. Smith, Phys. Rev. B 31 (1985) 3348. [42] P. Heimann, J. Hermanson, H. Miogsa, H. Neddermeyer, Phys. Rev. B 20 (1979) 3059. [43] P. Heimann, H. Neddermeyer, H.F. Roloff, J. Phys. C 10 (1977) L17. [44] P.O. Gartland, B.J. Slagsvold, Phys. Rev. B 12 (1975) 4047. [45] S.D. Kevan, W. Eberhardt, in: S.D. Kevan ( Ed.), Angle Resolved Photoemission, Elsevier, New York, 1992, p. 99. [46 ] R.W. Schoenlein, J.G. Fujimoto, G.L. Eesley, T.W. Capehart, Phys. Rev. B 43 (1991) 4688. [47] M. Wolf, Surf. Sci. 377 (1997) 343.
125
[48] N.-H. Ge, C.M. Wong, R.L. Lingle Jr., J.D. McNeill, K.J. Gaffney, C.B. Harris, Science 279 (1998) 202. [49] W. Steinmann, Th. Fauster, in: H.L. Dai, W. Ho ( Eds.), Laser Spectroscopy and Photochemistry on Metal Surfaces, World Scientific, Singapore, 1995, p. 184. [50] S.Y. Lee, E.J. Heller, J. Chem. Phys. 71 (1979) 4777. [51] E.J. Heller, Acc. Chem. Res. 14 (1981) 368. [52] E.J. Heller, R.L. Sundberg, D. Tannor, J. Phys. Chem. 86 (1982) 1822. [53] A.B. Myers, R.A. Mathies, T.G. Spiro (Ed.), Biological Applications of Raman Spectroscopy 2 , Wiley, New York, 1987, p. 1. [54] H.P. Bonzel, A.M. Bradshaw, G. Ertl ( Eds.), in: Physics and Chemistry of Alkali Metal Adsorption, Elsevier, New York, 1989. [55] P. Nordlander, N.D. Lang, Phys. Rev. B 44 (1991) 13681. [56 ] N.D. Lang, S. Holloway, J.K. Norskov, Surf. Sci. 150 (1985) 24. [57] H. Ishida, Phys. Rev. B 38 (1988) 8006. [58] C.C. Hsu, H. Bu, J.W. Rabalais, P. Nordlander, Phys. Rev. Lett. 69 (1992) 188. [59] G.A. Kimmel, D.M. Goodstein, Z.H. Levine, B.H. Cooper, Phys. Rev. B 43 (1991) 9403. [60] P.F. Barbara, T.J. Meyer, M.A. Ratner, J. Phys. Chem. 100 (1996) 13148. [61] Ph. Avouris, I.W. Lyo, Science 264 (1994) 942. [62] D.M. Eigler, P.S. Weiss, E.K. Scheizer, N.D. Lang, Phys. Rev. Lett. 66 (1991) 1189. [63] M.F. Crommie, C.P. Lutz, D.M. Eigler, Science 262 (1993) 218. [64] P. Kambhampati, C.M. Child, M.C. Foster, A. Campion, submitted to J. Chem. Phys.