THE VALENCE ELECTRON EXCITATIONS AND THE OPTICAL PROPERTIES OF ADSORBED ATOMS AND MOLECULES ON METAL SURFACES E. Burstein, A. Brotman, P. Apell
To cite this version:
E. Burstein, A. Brotman, P. Apell. THE VALENCE ELECTRON EXCITATIONS AND THE OP- TICAL PROPERTIES OF ADSORBED ATOMS AND MOLECULES ON METAL SURFACES. Journal de Physique Colloques, 1983, 44 (C10), pp.C10-429-C10-439. 10.1051/jphyscol:19831086. jpa-00223544
HAL Id: jpa-00223544 https://hal.archives-ouvertes.fr/jpa-00223544 Submitted on 1 Jan 1983
HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. JOURNAL DE PHYSIQUE Colloque C10, supplkrnent au n012, Tome 44, d6cembre 1983 page Ci0-429
THE VALENCE ELECTRON EXCITATIONS AND THE OPTICAL PROPERTIES OF ADSORBED ATOMS AND MOLECULES ON METAL SURFACES*
E. Burstein, A. Brotman and P. Apellf
Physics Department and Laboratory for Research on the Structure of Matter, University of Pennsy Zvania, PhiZadeZphia, PA 19204, U.S. A.
~6sum6- Nous presentons une revue sur les excitations des electrons de valence qui jouent un r81e dans les propridtds optiques, par exemple diffusion Raman, luminescence, etc., des atomes et des mole'cules adsorb& Les propridt6s optiques elles-m&mes fournissent un moyen pour dtudier la structure ilectronique des complexes adsorbat-substrat.
Abstract - We present an overview of the valence electron excitations that play a role in the optical p~opertieseg., Raman scattering, lumi- nescence, etc., of adsorbed atoms and molecules. The optical properties can themselves be used as surface-sensitive spectroscopic probes of the electronic structure of the adsorbate-substrate complexes.
INTRODUCTORY REMARKS
The lack of information about the electronic structure and, specifically about the energies, wavefunctions and widths of virtual bound states and bonding and antibonding states of the adsorbed atoms and molecules, has been a major barrier to efforts to elucidate the key mechanisms, other than surface roughness enhanced EM fields, that play a role in their optical properties, eg., Raman scattering, second harmonic generation, etc. There has been a tendency in the past to view the metal substrate and the adsorbed molecules as separate entities, albeit perturbed by each other's presence, and to attribute the enhanced optical phenomena to enhancement by the metal substrate. Thus, the enhanced Raman scattering by the adsorbed molecules on Ag has been termed llsurface enhanced Raman scattering". It is now clear, from a variety of experimental evidence, that the Raman scattering cross-sections of the adsorbed molecules is due, in part, to contributions from "intermoleculartt (e.g., charge transfer) electronic excitations of the adsorbed atoms and molecules, that are absent in the free atoms and molecules /1,2,3/. The appropriate point of view is that the optical properties of the adsorbed atoms and molecules are those of adsorbate-substrate complexes whose "intermolec- ular" and "intram~lecular'~excitations interact with the electronic excitations of the underlying metal.
The enhancement of the macroscopic and local EM fields at an A-S complex by surface roughness, image dipoles, etc., does correspond to an enhancement by the metal substrate. On the other hand, the enhanced optical response arising from "intermolecularlt excitations is not a surface enhancement. It is simply the manifestation of the formation of an A-S complex. We note in this connection that, to observe the Raman scattering (RS) by a monolayer, or submonolayer, of adsorbed molecules on a metal substrate in the absence of any "surface enhance- ment" of the incident and scattered EM fields, it is advantageous to carry out the RS measurements under resonance enhanced conditions /4,5/ (eg., to use excita- tion wavelengths at which intramolecular or intermolecular resonances occur) and, for this purpose, to extend the excitation wavelengths from the visible into the ultraviolet and infrared,
Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphyscol:19831086 JOURNAL DE PHYSIQUE
BONDING, ELECTRONIC STRUCTURE AND VALENCE ELECTRON EXCITATIONS
To adequately understand the optical properties of adsorbed atoms and molecules on metal substrates, one needs information about the electronic structure and, in particular, information about the electronic excitations, (eg., energies, widths, oscillator strengths, etc.) of the A-S complexes that are formed. The determination of the electronic structure of molecules adsorbed on metal surfaces (and, thereby, the elucidation of the bonding between adsorbates and substrates) has been a major objective of surface physicists and chemists, and there has been considerable experimental and theoretical progress in attaining this objective for a number of adsorbed molecules and metal substrates /6/.
We present here an overview of the bonding of atoms and simple molecules at metal surfaces and of the electronic structure and electronic excitations of the A-S complexes that are formed. In doing so, we will limit ourselves to isolated, e.g., low coverage, adsorbates on smooth metal surfaces in ultra high vacuum. Our primary objective is to clarify the nature of the valence electron excitations that play a role in optical phenomena exhibited by adsorbed atoms and molecules on metal surfaces /7/.
The metal substrates fall qualitatively into several groups. i) Metals, such as Al, Na and K, which do not have d-bands and are free-electron-like. The covalent bonding of adsorbates by the s,p electrons of these metals is relative- ly weak. ii) Noble metals (eg., Ag, Cu and Au), whose d-bands lie well below the Fermi level, for which the chemisorption of adsorbates via the d electrons is only moderately strong. iii) Transition metals (eg., Ni, Pd, Pt) whose d-bands overlap the Fermi level and, thereby, contribute to a strong covalent bonding of adsorbates.
The adsorbates to be discussed also fall into several groups. i) simple aromatic molecules, eg., benzene, pyridine, pyrazine, etc. ii) Diatomic molecules, eg., CO and N2. iii) Atomic adsorbates, eg., hydrogen, halogen, alkali and rare gas atoms.
Aromatic Adsorbates
The aromatic molecules, eg., benzene, pyridine, etc., are chemisorbed on metal surfaces via covalent bonding of their n electrons with the d electrons of the metal substrate, and generally lie flat on surface. The covalent bonding of benzene is only moderately strong, even in the case of transition metal sub- strates /8/.
In the case of pyridine, the plane of the adsorbed molecules may be inclined, or perpendicular, to the surface due to covalent bonding via the lone pair of electrons of the N /9/. As shown by Demuth and co-workers /10,11/ who have determined the electronic excitations of pyridine, pyrazine and benzene adsorbed on Ag(ll1) using electron energy loss measurements, the intramolecular excitations of the adsorbed molecules are only moderately shifted and broadened relative to those of the free molecules. Thus, there is only a moderate mixing of the adsorbate and substrate orbitals. The electron energy loss spectra for pyridine and pyrazine also exhibit a very broad feature with an onset at -2 to 2.5 eV, which Demuth et a1 identify as corresponding to A-S charge-transfer excitations.
The energy levels of free pyridine molecules and of pyridine molecules adsorbed on pg, based on the electronic excitations energies reported by Demuth et a1 and on the valence electron binding energies obtained from photoemission measure- ments /12/, are shown schematically in Fig. la. The affinity level of free pyridine lies above the vacuum level (E ) i.e., the negative pyridine ion is unstable. When adsorbed on a free-electron-like metal substrate, the affinity level is lowered by the coulomb interaction of the negative pyridine ion with its image charge in the metal substrate. The affinity level of the adsorbed pyridine lies at an energy -2 eV below EV and corresponds to a "virtual bound state". An electron in the affinity level of the adsorbed pyridine is unstable with respect to the metal substrate and will "hopff into an empty level of the metal. We note that the excited states of the adsorbed (neutral) pyridine molecules lie above the Fermi level (E ) and also correspond to ffvirtual bound states". The fact, that the observed F;intramolecularlpelectronic excitations of adsorbed pyridine are not appreciably broadened, indicates that the contribution to the broadening from the decreased lifetime of the tfunstableflexcited states is not large. There are two possible charge-transfer excitations. One involves the transition of an electron in the metal below E to the affinity level of adsorbed pyridine /13/ with an onset energy ECT = EA* -F~Fequal to - 2 eV. The other involves the transition of an electron from the ground state of adsorbed pyridine to an empty level above EF in the substrate (a process which corresponds to photoemission from the adsorbate into the metal substrate) with an onset energy ECT = EF - EI* equal to - 3 eV.
FREE ADSORBED FREE ADSORBED co CO PYRlDlNE PYRlDlNE (b) (a 1 Fig. 1. Schematic energy level diagram for (a) pyridine adsorbed on Ag and (b) CO adsorbed on Ni. Diatomic Adsorbates
CO is strongly chemisorbed by transition metals with the C end attached to the metal /11,12/. CO is also chemisorbed by the noble metals with the axis of the molecule normal to the substrate. The covalent bonding is, however, appreciably weaker than on the transition metals /13/. CO is only weakly adsorbed by A1 /14/. The molecule lies flat on the A1 surface and is presumed to be physisorbed.
CO adsorbed on Ni is probably better understood than any other adsorbate on a metal substrate. The highest occupied lone pair orbital (50 ) and the lowest unoccupied orbital (21r ) are appreciably admixed with the d/s orbitals of Ni. The admixture of adsorbed molecule and metal substrate orbitals leads to bonding and antibonding states whose wavefunctions are largely localized in the vicinity of the molecule. The one-electron energy levels for free CO, and for CO adsorbed on Ni /I?/, are shown schematically in Fig. Ib. The occupied 50 -M d/s bonding state (50 ) is predominantly CO in character, and the corresponding unoccupied antibonding state (Md/s) is predominantly metal in character. On the other hand, the occupied 2m -Md/s bonding state (~d/s)is predominantly metal in charac- ter, and the corresponding unoccupied antibonding state (2% ) is predominantly CO in character. We note also that the orbital of the CO affinity level, which in the free molecule lies -2 eV above EV, is also admixed with the metal d/s orbitals. On the basis of recent inverse photoemission data /15/, the affinity level of the adsorbed CO molecule (i.e., the energy level of the A-S complex C10-432 JOURNAL DE PHYSIQUE
with an additional electron) lies "4 eV above EF.
The valence electron excitations of the adsorbed CO molecule include intermolecular (bonding-antibonding) transitions of the A-S complex, charge-transfer transitions between the electron levels of the metal substrate and the A-S complex, as well as intramolecular transitions between levels of the A-S complex which are pre- dominantly CO in charact~r. In this connection we note that an electronic transi- tion from the occupied 50 bonding state to the unoccupied Md/s antibonding state involves a transfer of charge from the CO to the metal and, thus, corresponds also to a "charge-transfer" excitation in which the CO is the donor and the metal is the acceptor. The corre_sponding transition from the occupied Md/s bonding state to the unoccupied 2~ antibonding state also corresponds to a charge-transfer excitation from the metal as the donor to the CO as the acceptor.
The covalent bonding of the homonuclear molecule N by Ni is moderately strong /15,19/. The adsorbed N molecule is orienged with its axis normal to the 2 substrate, and the N -Ni complex that is formed lacks a center of inversion. The N-N stretching vibragion mode is accordingly infrared active and, in fact, has a sizeable effective charge. The effective charge of the N-Ni vibration mode is com- parable in magnitude to that of the C-Ni vibration mode of CO adsorbed on Ni. The valence electron excitations of the adsorbed N2 molecule are similar in type to those of the adsorbed CO molecule.
Atomic Adsorbates
Theoretical investigations of atomic adsorbates have been carried out by Lang and Williams /20/, by Lundqvist and coworkers /21/ and by Flynn and coworkers /22/. Experimental investigations have been carried out by Flynn's group (e.g., optical absorption in the ultraviolet by halogen, alkali and rare gas atoms adsorbed on various metal substrates including the alkali metals and Mg) /19/, and by Walden and Lindgren /22/and Anderson and Jostell /21/(e.g., photoemission and electron energy loss measurements for alkali atoms adsorbed on noble metals and transition metals).
The charge state of the monovalent adsorbates (e.g., hydrogen, halogen, and alkali atoms) on a given metal substrate is determined by the relative magnitudes of the ionization and affinity energies of the adsorbed atom E * and EA*, respec- tively, and the work function 0 of the metal substrate /21/. BI* corresponds to the difference in energy of the adsorbed atom and that of the adsorbed ionized- atom with an electron at infinity. In the absence of any sizeable covalent bonding of the adsorbate by the substrate, which is generally the case when the substrate is a free-electron like metal, EI* will be smaller than EI, the ionization energy of the free atom, because of the attractive coulomb interaction of the positive adsorbate ion with its image charge in the metal. Similarly E * corresponds to the difference in energy of the adsorbed atom with an electron a! infinity and that of the adsorbed atom with the electron attached (e.g., that of the negative adsorbate ion). Here too, in the absence of any sizeable covalent bonding, EA* will be larger than EA, the affinity energy of the free atom, because of the attractive coulomb interaction between the negative adsorbate ion and its image charge in the metal.
In tQe case of atoms adsorbed on free-electron-like metal substrates, the magnitude of U = EI* - EA* is appreciably smaller than U = EI - EA. (In the case og Li adsorbed on metal substrate with a high electron denslty, such as Al, U may,actually be negative, e.g. E * < EA*). When the work function 0 is larger than EA*, but smaller than E *, the adsorbed atom will be neutral. When 0 is E smaller than A *, which is t$e case for halogens adsorbed on Al, the affinity level will lie below EF and be occupied, e.g. the adsorbate will be a negative ion. When 0 is larger than EI*, which is the case for alkali atoms adsorbed on Al, the atoms will be ionized, e.g., positively charged.
When the covalent bonding of the adsorbate by the metal substrate is strong, which is often the case for transition metal substrates, the increase in the valence electron binding energy due to covalent bonding offsets the decrease due to the screening of the coulomb interaction by the metal substrate. Moreover, the affinity energy EA* of the A-S complex does not have a simple relation to that of the free adsorbate and is generally much smaller than EI* and smaller than the 0 of the substrate. The adsorbed atoms on a transition metal substrate will accordingly be largely neutral.
Among the atomic adsorbates, H has been the most extensively studied theoretical- ly and experimentally /21,25,26/. The free H- ion which has a He configuration is stable with EA = +0.75 eV. When H is adsorbed on a metal substrate, its a_ffinity level is lowered appreciably due to the coulomb interaction of the H ion with its image charge in the metal. Because of the small size of the H- ion, and the small 0 of the alkali metals, the affinity level of H adsorbed on an alkali metal substrate lies below EF. Thus, H adsorbs on the alkali metals as an H- ion and the bonding is predominantly ionic. (H also adsorbs on Ag as the H- ion). On the other hand H is chemisorbed by transition metals (e.g., Ni, Pd and Pt) at low temperature via strong covalent bonding of the 1s electron with the d electrons of the transition metal substrate. (On warming to 300 K the adsorbed H atoms undergo an irreversible transition to a sub-surface state). Photoemission data indicate that a bonding H level is split off from the substrate d bands. In the case of H adsorbed on Ni at low coverage, the split-off level lies -8 eV below EF.
Alkali and halogen ions (e.g. ~i+and Cl') have rare gas configurations and are bonded to the metal substrate predominantly by the coulomb interaction of the ion with its image charge in the metal. The s electrons of the neutral alkali atoms and the p electrons of neutral halogen atoms form covalent bonds with the s,p,d electrons of the metal substrate. The covalent bonds are, in general, relatively weak when the alkali and halogen atoms are adsorbed on a free-electron-like metal substrate. In the case of transition metal substrates, however, the covalent bonding between the s and p electrons of the alkali and halogen atoms with the metal d electrons can be relatively strong.
Rare gas atoms are physisorbed by metals via the van er Waals interaction. As pointed out by Flynn et a1 /22/, when one of the pg valence electrons of adsorbed rare gas atoms is excited into the next s level, the excited rar gas atom has an alkali atom configurati n. (In the case of Xe which has a 5p con- figuration, the excited Xe has a 5p 86s' Cs configuration). As a consequence, the interaction of the excited rare gas atom with the metal substrate will be similar to that of the corresponding alkali atom (e.g. the s electron forms a covalent bond with the electrons of the substrate).
Experimental data on the charge-states of adsorbates is obtained from the depen- dence of the work function of the metal substrate on adsorbate coverage, optical absorption and electron energy loss measurements of the electronic excitations, infrared data on the vibration frequency of the A-S bond, and data on Ex* derived from photoemission measurements. The general trends in the observed charge- states of halogen and alkali adsorbates on different metal substrates are as follows:
Halogen atoms which have affinity energies greater than, or close to, the work function of a free-electron-like metal substrate (e.g. C1 and Br for which EA = 3.8 and 3.4 eV, respectively, adsorbed on Cs, Mg and A1 for which 0 = 1.8, 3.8 and 4.2 eV, respectively) are adsorbed as negative ions, i.e., the affinity energy of the adsorbed halogens EA* fs greater than the 0 of the substrate. In general the image charge coulomb interaction, and, therefore, the magnitude of E * - EA, for a given halogen, increases with increase in the electron density of tke metal. The bonding is also ionic in the case of halogen atoms adsorbed at low coverage on Ag and Cu. On the other hand, halogen atoms adsorbed on transition metal substrates are adsorbed as "neutral" atoms, in part, because of the sizeable covalent bonding which tends to decrease EA* and, in part, because JOURNAL DE PHYSIQUE of the large values of 0 of the metal substrates.
Alkali atoms whose EI values range from 3.9 eV (e.g., Cs) to 5.4 eV (e.g. Li) are adsorbed as neutral atoms on alkali metal substrates (e.g., low electron density metals with small 0). On the other hand, they are adsorbed as positive ions on Mg and A1 which have greater 0 values and greater electron densities. Cs atoms which have the smallest E are adsorbed as ions on a Cu substrate. I On the other hand, the alkali atoms are adsorbed as ttneutrallt atoms on transition metal substrates, in large part, bepause the contribution of covalent bonding to the binding energy offsets the decrease in the ionization energy due to the coulomb interaction of the ionized atoms with its image charge.
The atomic adsorbates exhibit three types of electronic excitations: i) Atomic- type excitations of the adsorbed atoms which are the counterparts of the intra- molecular excitations of adsorbed molecules. ii) A-S charge-transfer excitations. iii) Electron transitions between bonding and antibonding levels of the adsorbed atom-metal substrate complexes.
Atomic-type electronic transitions are observed for rar gas a sorbates, e.g. for atoms that are weakly adsorbed. The ultraviolet (p8+ p5srf spectra of rare gas atoms adsorbed on a variety of metal substrates (e.g. alkali metals, Mg, Al, Au and Ti) have been investigated in some detail by Flynn and coworkers /20/. They observed two types of spectra are observed. In one type, which includes nearly all of the rare gases adsorbed on Cs, K, Mg and Al, the spectra exhibit strong atom-like excitations. In the second type, which includes Xe adsorbed on Al, Au and Ti, and Kr adsorbed on Al, the atom-like excitations do not appear. They suggested that the absence of the atomic-type exci a ions in the second type of spectra was due to the fact that the excited (pff s ) states for the adsorbed Xe and Kr lie above E F of the metal substrate and, were therefore, tlunstabletr to charge-transfer of the electrons from the excited level to the substrate. Lang, et a1 /27/, on the other hand find, on the basis of density functional calculations of the electronic structure of rare gas adsorbates, that the excited states of the rare gas adsorbates, for both types of spectra, lie above EF of the metal substrate, and that the instability of the excited states toward charge- transfer to the substrate was not a valid explanation for the differences in the two types of spectra. We have already noted earlier that the intramolecular excitations of pyridine adsorbed on Ag are not appreciably broadened even though the excited states of the adsorbed pyridine are ttvirtual bound statestt. Demuth et a1 /28/have recently obtained electron energy loss data on the electronic excitations of Ar and Xe on A1 and on the noble metals, which, in fact, show that the excitations do have an atomic-type character. Their data indicate, moreover, that the excited state lifetimes were 4 times longer than that pre- dicted by Lang et a1 /27/ on the basis of their density functional calculations. More recently, Eberhardt and Zangwill /29/ have measured the 4d Rydberg state excitations of Xe adsorbed on Au, and find that they exhibit a broad, but well defined structure whose integrated oscillator strength is roughly the same as that for free Xe atoms, i.e. the optical excitations are atomic-like in character. An explanation for the apparent discrepancy, between the data obtained by Flynn's group and those obtained by the other groups, is still lacking. It may well reside in differences in the preparation and character of the adsorbed rare gas samples.
Charge-transfer excitations are of particular interest in the case of hydrogen, halogen and alkali atoms which are adsorbed as ions (e.g., H-, C1-, ~a+)/30/. In the case of adsorbed C1- ions the charge-transfer excitations involve the transition of an electron from the adsorbate affinity level (which lies below EF) to an empty level in the metal substrate (Fig. 2a). The charge-transfer excitations have an onset at an energy hw = EF - EA*.and extend to an energy beyond .tfw = EV - EA* where they coalesce with photoemrssion excitations into the vacuum.
In the case of adsorbed ~a+ions whose energy level lies above EF, the charge- transfer excitations involve the transition of an electron from an occupied level in the metal substrate to the "virtual bound state" of the adsorbed ~a+ (Fig. 2b). The charge-transfer excitations have an onset at an energyfw = E - E and extend to an energy ilw = EIt EC. I * F -
Fig. 2. Charge-transfer transitions of atomic adsorbates on a free-electron-like metal (a) ~1-(b) ~a+(c) xeO.
Charge-transfer excitations also occur for adsorbed rare gas atomic 6e.g. xeO) and for the alkali atoms that are adsorbed as neutral atoms (e.g. Li adsorbed on a Cs substrate) and The affinity level of these adsorbates lie below EV and above EF, and, therefore, correspond to virtual bound states. For these adsorbates, the charge-transfer excitations involve the transition of an electron from the metal into the empty affinity level, with an onset at %m = E * - EF (Fig. 2c). A
Finally we note that neutral atoms that are covalently bonded by the metal sub- strate, e.g. H chemisorbed on a transition metal substrate, exhibit intermolec- ular electronic excitations between bonding and antibonding states which are similar in nature to the intermolecular excitations that occur for molecular adsorbates.
Some comments are in order regarding the widths of the electronic excitations of A-S complexes. The energy levels of A-S complexes have sizeable widths due to the admixing of the adsorbate orbitals with the orbitals of the quasi-continuous energy levels of the substrate /18/. These widths depend on the overlap of the adsorbate and substrate orbitals and, therefore, increase with increase in the bonding strength. Effects that decrease the lifetime of the excited state also contribute to the widths of the electronic excitations. The coupling of the electronic excitations of the A-S complex with the single-particle and collective electron excitations of the metal substrate leads to a broadening, as well as shifting, of the electronic excitations. The vibronic structure of the electronic excitations can lead to an appreciable broadening of the photo- excitation (e.g. absorption) spectra when the widths of the electronic levels are greater than the energies (ha) of the vibration modes. We note also that the valence electron excitations of the A-S complex may be accompanied by sizeable shake-up, and related effects, involving electron-hole pair excitations in the metal substrate which lead to an apparent broadening of the electronic excitations.
VIBRONIC EXCITATIONS
We have thus far discussed the electronic levels and excitations of A-S complexes. We now consider the vibronic nature of the levels and, in particular, the intermolec- ular vibronic excitations which play important roles in the Raman scattering and luminescence by the A-S complexes. For simplicity, we discuss the vibronic charge-transfer excitations of an atomic adsorbate (e.g., C1-) on a free-electron- like metal substrate (e.g., Al) /31/. JOURNAL DE PHYSIQUE
Fig. 3. Dependence of the energy levels of C1- and c1° on distance from an A1 substrate.
The vibronic character of the charge transfer excitation of an atomic adsorbate arises from the dependence of the ground and excited energy levels on the adsor- bate separation from the substrate. In Fig. 3 we show the dependence of the energy levels of ~1-andc1° on their distance from the metal substrate. E represents the energy to remove an electron from a free C1- to infinity and thereby form clO. The charge-transfer excitation involves the transfer of an electron from C1- to an empty level in the metal, leaving c1° adsorbed at the surface. The energy of the adsorbed ~1-levelhas an attractive contribution from the image charge potential and, at short distances from the substrate, a repulsive contribution from the overlap of the adsorbate and substrate electrons, e.ej., the Pauli exclusion principle. On the other hand, the energy of the adsorbed C1 level has an attractive contribution from the van der Waals interactions and from weak covalent bonding of the c1° p electrons with the metal s,p electrons, and a repulsive contribution, at short distances, due to the overlap of adsorbate and substrate electrons. Since the ionic interaction of the C1- with the sub- strate is stronger than the van der Waals and covalent interactions of the c1° with the substrate, the equilibrium distance of ~1-,and the corresponding spring constant K(c~-)which determines the stretch vibration frequency of the C1-- substrate bond, is greater than the corresponding quantities for clO. The vertical energy separation E -(d) - E o(d) = ~~*(d)represents the energy to photoionize C1- at constant d. '$he variaflon of E o(d) 0 with distance from the sub- C1 - strate is also shown in Fig. 3. The magnitude of EClo(d) - 0 - E -(d) = EA*fd) - 0 represents the energy to transfer an electron from the C1- led1 to a level in the metal at EF, i.e., the tfonsettfenergy ECT(d).
The electronic transitions between the ground state, (e.g., EC1-(d)) and the excited state (e.g., EClo(d)) correspond to vibronic transitions, e.g., they involve changes in the vibrational quantum number of the adsorbed C1- and clO. The optical matrix elements for such transitions depend on the overlap of the vibrational wavefunctions of the ground and excited states, which in turn depend on the difference in equilibrium distances of the adsorbed C1- and clO. Since charge-transfer electronic excitations have a continuum of energies beyond E CT' the vibronic excitations do not lead to any observable vibrational structure. The vibronic character of the charge-transfer excitations does play a key role in the Raman scattering by the vibrational modes of the A-S complex. Also, as shown by Gadzuk et a1 /32/ in their investigation of the photoemission excita- tions of Xe adsorbed on Cu(llO), the vibrational levels of the A-S complexes make a sizeable contribution to the broadening of the photoemission spectra.
RAMAN SCATTERING AND LUMINESCENCE
The point of view that the Raman scattering and luminescence by adsorbed atoms and molecules are optical phenomena of adsorbate-substrate complexes leads one immediately to the realization that the microscopic mechanisms, vibration mode selection rules, etc. for these phenomena can be markedly different from those for the free adsorbates. (This conclusion is, of course, an obvious one in the case of the Raman scattering by adsorbed atoms, which does not exist for the free atoms). The emphasis of efforts to elucidate the Raman scattering and luminescence by adsorbates on metal substrates, is, in part, placed on the elucidation of the nature of the A-S complexes. Once the nature of the A-S complexes is established, we can make use of the extensive knowledge about the Raman scattering /33/ and luminescence of molecules /34/, that already exists, to ascertain the physics underlying the Raman scattering and luminescence by the A-S complexes. It is also evident that the Raman scattering and luminescence of the adsorbed atoms and molecules can, themselves, be used as surface sensitive spectroscopic probes of the electronic excitations of A-S complexes. The measure- ment of Raman scattering excitation profiles is, in fact, a form of modulation spectroscopy which is particularly advantageous for A-S complexes, since it enables one to determine the electronic excitations that are associated with the adsorbed atoms and molecules. The measurement of the excitation profiles of the Raman scattering by vibration modes of different symmetry provides informa- tion, not only about the energies and widths of the electronic excitations that interact with the vibration modes, but also, about the character of the electronic states that are involved.
In its simplest form, the microscopic mechanism for the Raman scattering by molecules, which is also applicable to charge-transfer excitations of S-A com- plexes /2/, involves a virtual optical transition from a vibrational level in the ground electronic state to a intermediate state, that corresponds to a vibra- tional level in the excited electronic state, followed by a virtual optical transition from the intermediate state to a different vibrational level in the ground electronic state. As can be seen in the electron energy-atomic configura- tion coordinate diagram shown in Fig. 4, the equilibrium configuration-coordinates and vibrational frequencies of the ground and excited electronic states are different.
The transition polarizability (Raman scattering matrix element) for the two- level, two-step process can, under simplifying assumptions, be expressed in the following form /32/:
where
time of the excited state /35/. When the luminescence of an A-S complex is observable, measurements of the luminescence emission spectrum and the lumin- escence excitation profile can provide vital information about the vibronic excitations of the A-S complex.
le)
lo)
Fig. 4. Energy-configuration coordinate diagrams showing the vibronic transitions that are involved in (a) Raman scattering and (b) luminescence.
Acknowledgements
We wish to acknowledge valuable discussions with C. P. Flynn, J. W. Gadzuk, N. D. Lang, B. I. Lundqvist, R. P. Messmer, E. W. Plummer, P. Soven and M. Sunjic.
References
* Research supported by ONR and by the NSF MRL at the University of Pennsylvania.
+ IBM Postdoctoral Fellow. Present address, Chalmers University of Technology, Goteborg 41296 Sweden.
OTTO, A., J. Elec. Spect. and Related Prob. 29 (1983) 329. ADRIAN, F. J., J. Chem. Phys. 17 (1982) 53027 UEBP, H., ICHIMURA, S., and YAMADA, H., Surface Science 2 (1982) 433. YAMADA, H., App. Spect. Rev. 17(2) (1981) 227. BURSTEIN, E., BURNS, G. and DACOL, F., Solid State Commun. 3 (1983) 595. See review papers in Photoemission and the Electronic Properties of Surfaces, Edited by FEUERBACHER, B., FITTON, B., and WILLIS, R.F., (Wiley, NY 1978). Because of the "length of paper" limitations we will not attempt to cite all relevant references. It is our intention to focus primarily on work which contributes to the conceptual understanding of the subject. KELEMEN, S. R., and FISCHER, T,.E., Surf. Sci. 2 (1981) 45. DEMUTH, J.E., CHRISTMANN, K., and SANDA, P.N., Chem. Phys. Lett. 3 (1980) 201. DEMUTH, J.E., and SANDA, P.N., Phys. Rev. Lett. 3 (1981) 57. AVOURIS, P.H., and DEMUTH, J.E., J. Chem. Phys. 75 (1981) 4783. KELEMEN, S.R. and KALDOR, A., Chem. Phys. Lett. 2 (1980) 205. BURSTEIN, E., CHEN, Y.J., CHEN, C.Y., LUNDQVIST, S., and TOSATTI, E., Solid State Commun. 2 (1979) 567. FREUND, H.J., and PLUMMER, E.W., Phys. Rev. B23 (1981) 4859. STOHR, J. and JAEGER, R., Phys. Rev. B26 (1982) 4111. SPITZER, A., and LUTH, H., Surf. Sci. 2 (1981) 29. CHIANG, T.-C., KAINDL, G., and EASTMAN, D.E., Solid State Commun. 3 (1980) 25. HIMPSEL, F.J., and FAUSTER, T.H., Phys. Rev. Lett. 9 (1982) 1583. DINARDO, J., EBERHARDT, W., FREUND, H.-J., and PLUMMER, E.W. Surf. Sci. -118 (1982) 465. LANG, N.D., and WILLIAMS, A.R., Phys. Rev. B18 (1978) 616. LANG, N.D., Phys. Rey. 3 (1971) 4234. LUNDQVIST, B.I., GUNNARSSON, O., HJELMBERG, H., and NORSKOV, J.K., Surf.Sci. 89 (1979) 196. JOHANSSON, P. K., and LUNDQVIST, B.I., in Recent Develop- ments in Condensed Matter Physics, Vol. 1, Edited by DEVREESE, J.T., (Plenum Press, NY, 1981) p. 605. CUNNINGHAM, J.E., GREENLAW, D., FLYNN, C.P., and ERSKINE, J.L., Phys. Rev. El (1980) 717. CUNNINGHAM, J.E., GIBBS, D., CHIU, T.H. and FLYNN, C.P., J. Phys. C: Solid State 14 (1981) L113. GIBBS, D., CUNNINGHAM, J.E., CHIU, T.H. and FLYNN, C.P., J. Phys. C: Solid State 111 (1981) 1119. FLYNN, C.P., and CHEN, Y.C., Phys. Rev. Lett. 3 (1981) 447. FLYNN, C.P. and CUNNINGHAM, J.E., J. Phys. C: Solid State 15 (1982) 11169. GIBBS, D., PhD Thesis, University of Illinois, 1983. LINDGREN, A.A., and WALDEN, L., Solid State Commun. 5 (1978) 13 and 2f! (1980) 671; Phys. Rev. E (1981) 5967; Surf. Sci. 89 (1979) 319. ANDERSON, S., and JOSTELL, U.G., Sslid State Comrnun. lT(1973) 829 and 833; Surf. Sci. 3 (1974) 629. BLANCHET, G.B., ESTRUP, P.J. and STILES, P.J., Phys. Rev. Lett. 44 (1980) 171; BLANCHET, G.B. and STILES, P.S., Phys. Rev. B21 (1980) 3273 and 3655. EBERHARDT, W., GREUTER, F., and PLUMMER, E.W., Phys. Rev. Lett. 3 (1981) 1085. LANG, N.D., WILLIAMS, A.R., HIMPSEL, F.J., RIEHL, B. and EASTMAN, D.E., Phys. Rev. B26 (1982) 1728. DEMUTH, J.E., AVOURIS, P.H., and SCHMEISSER, S., Phys. Rev. Lett. 50 (1983) 600. EBERHARDT, W., and ZANGWILL, A., Phys. Rev. B27 (1983). APELL, P., BURSTEIN, E., and SUNJIC, M., Bull. Amer. Phys. Soc. 8 (1983) 526. SUNJIC, M., APELL, P. and BURSTEIN, E., Bull. Amer. Phys. Soc. 3 (1983) 526. GADZUK, J.W., HOLLOWAY, S., MARIANI, C., and HORN, K., Phys. Rev. Lett. 48 (1982) 1288. ALB~HT,A.C., and HUTLEY, M.C., J. Chew. Phys. 55 (1971) 4438. BIRKS, J.B., Photophysics of Aromatic Molecules (Wiley-Interscience, NY, 1972.). RITCHIE, G. and BURSTEIN, E., Phys. Rev. (1981) 4843.