Surface science studies of fullerenes have been used in many unique ways to advance our knowledge of fullerenes and doped fullerenes and their interaction with substrates. The interpretation of surface science studies frequently depends on measurements using complementary surface science techniques. Photoemission and inverse photoemission studies (§17.1) have provided unique information on the density of states and, in conjunction with optical absorption and luminescence spectra (see §13.2 and §13.3), have clarified fundamental aspects of the electronic structure of fullerenes in the solid state (see §12.7). Electron energy loss spectra (EELS) provide complementary information to optical studies regarding interband transitions, the frequency and wave-vector dependence of the dielectric function e((a,q), as well as information on fundamental excitations such as plas-mons and phonons (§17.2). Auger spectroscopy is very important for identifying the species that is desorbed from a surface and for monitoring the quantity that is desorbed (see §17.3). Auger spectroscopy studies have also yielded unique information about the on-site hole-hole Coulomb interaction (§17.3), while scanning tunneling microscopy and spectroscopy studies have also yielded valuable information about the crystalline structure, the surface reconstruction and the electronic structure of fullerenes (§17.4). Temperature-programmed desorption has proved especially useful for the identification of sublimation energies and the relative bonding of fullerenes to each other and to the substrate species (§17.5). Of importance to many potential applications of fullerenes (§20.2.6, §20.2.7, and §20.3.2) is the nature of the interface between fullerenes and various semiconductors and metallic substrates, and this topic is discussed in §17.9.
Photoemission and inverse photoemission spectroscopies (PES and IPES) have been extensively applied [17.1-4] to study the density of states within a few eV of the Fermi level for both undoped and doped fullerenes, and many calculations of the density of states have been compared to these measurements [17.4-6]. This work has been extensively reviewed [17.2,4,7].
In the standard photoemission experiment, an incident photon of energy hbj excites an electron from an occupied state to the vacuum level and the electron kinetic energy Ee is measured to determine hw — Ee, which is plotted with reference to the Fermi level. Comparison of the photoemission intensity vs. (ha> - Ee) taken at a variety of incident photon energies is used to relate the PES spectra to a density of occupied states (see upper inset to Fig. 17.1(a)). Correspondingly, for the inverse photoemission experiment, an incident electron enters a bound excited state, emitting a photon, whose energy is measured, so that the spectrum provided by IPES measurements yields the density of unoccupied states. In the PES and IPES spectra of fullerenes, there are few complications due to excitonic and vibronic interactions, which dominate studies of the optical absorption and luminescence spectra near the fundamental absorption edge (see §13.1.3).
Ultraviolet (UV) photoemission and inverse photoemission experiments are highly sensitive to the first two or three nanometers of material near the surface and because of surface reconstruction effects, photoemission measurements of the density of states of semiconductors do not usually provide an accurate determination of the density of states in bulk compounds. The low density of electron charge found on the surface of fullerene solids indicates a low density of surface states. Thus, surface reconstruction effects may not be important for fullerite surfaces, thereby making PES and IPES especially useful tools for studying fullerenes in the solid state. Nevertheless, the large unit cell dimensions of fullerite crystals imply a sensitivity of the PES and IPES techniques to only a few unit cells at the surface of the crystalline solid, thereby emphasizing the surface properties relative to bulk properties.
With regard to doped fullerenes, surface states, screening effects, and surface reconstruction effects may be more important, so that PES and IPES measurements may not provide as accurate a measurement of the density of states for MxCfi0 compounds as for the case of undoped fullerenes. Furthermore upon addition of an alkali metal such as K, spatially separated phases such as K3C60 and K^C«, are formed, and these phases may be inho-
mogeneously distributed throughout the sample. In particular, the surface layer may have different stability regimes for these phases relative to the bulk. Such inhomogeneities may present significant difficulties with the interpretation of PES and IPES spectra in the near-surface layers (20-30 A thick) of doped fullerenes.
Typical photoemission spectra (E < E,. , where EF is the Fermi energy) and inverse photoemission [17.3,8] spectra (E > EF) are shown in the lowest trace in Fig. 17.1(a) for C60, where the intensity maxima in the spectra correspond to peaks in the density of states, and the energy difference between the thresholds for the photoemission and inverse photoemission spectra yield important information on the HOMO-LUMO gap in solid C60 [17.9,10], Also shown in Fig. 17.1(a) are PES and IPES spectra for K,C60 or potassium-doped C60, and these spectra are further discussed in §17.1.3.
For the undoped fullerenes, the systems under investigation for the PES and IPES spectra contain C„c molecules with (nc — 1) 77-electrons (for PES spectra) and with (nc + 1) 7t electrons (for IPES spectra), where nc is the number of carbon atoms in the fullerene, as shown schematically in the inset to Fig. 17.1(a) and in Fig. 17.1(b). In contrast, the states excited in an optical absorption or EELS experiment correspond to molecules with nc 1t electrons for which an excited electron and hole are located on the same or neighboring fullerenes and are bound in a Frenkel-type exciton state, if the electron and hole are on the same molecule, or in a charge transfer state, if the electron and hole are on different molecules [see Fig. 17.1(b)]. If the binding energy of the exciton is supplied, so that the electron and hole are separated to infinity, then the nc + 1 and nc — 1 states in the IPES and PES experiments, respectively, can be realized. Using a value of 3.7 eV for the separation between the peaks for the PES and IPES spectra [17.9] and a value of 1.4 eV for the Hubbard U [17.4], we obtain a value of 2.3 eV for the intrinsic band gap A. With this value of the band gap A and a value of 1.84 eV for the singlet absorption edge, a value of 0.46 eV is obtained for the exciton binding energy as shown in Fig. 17.2. The bulk value of U for C60 may be lower than that measured near the surface because of surface screening deficiencies [17.12], The Hubbard U is the energy needed to transfer an electron from one neutral fullerene C„c to create two distinct molecules with nc + l and nc — 1 7r electrons, respectively [see Fig. 17.1(b)]. The onset energy (2.3-2.6 eV) which is equal to (U + A - W) is consistent with a bandwidth of ~1 eV for the LUMO-derived band, and an energy width W of ~ 0.5 eV [17.9],
(a) Photoemission (PES) and inverse photoemission (IPES) spectra for C«, and films [17.2]. The spectra show the effects of K incorporation into C^ films. Adding more K causes an increase of the emission intensity below EF in the PES spectra and a decrease of intensity above EF in the IPES spectra. The top spectrum shows that EF shifts into the gap when the LUMO /,„ band is filled, (b) An array of molecules exposed to a photon and an electron beam. The figure shows schematically a molecule with nc electrons (open circles), a localized Frenkel exciton (±), and an electron (—) and a hole (+) on different molecules containing nc + 1 and nc — 1 electrons, respectively [17.4],
The peaks in the photoemission spectra tend to be wide compared with other determinations of the bandwidths (~0.5 eV). The reason for the larger PES widths is not presently understood, although it is believed that molecular rotations at room temperature are in part responsible for the large PES and IPES linewidth for C60. Band calculations of the density of states, whether based on a one-electron picture [17.5,13,14] or a many-body picture [17.9,10], provide a good fit to the observed photoemission and inverse photoemission spectra for C60, at least at the level of detail
Fig. 17.2. Schematic energy levels showing the energy separation between the IPES and PES peaks which yields the HOMO-LUMO gap (A) plus the Hubbard U (1.4 eV). Using a value of 1.84 eV for the singlet optical absorption edge yields an exciton binding energy of 0.46 eV, where the zero of energy is for convenience placed at the S0 level. The triplet level (T,) at 1.55 eV [17.11] indicates an exchange energy of 0.29 eV.
that has been used in these comparisons. For example, since the density of states obtained from PES and IPES experiments is nearly an order of magnitude less than that obtained in typical calculations (see §12.7), the comparisons between theory and experiment have focused on the energy of the spectral features and on the relative intensity of these features [17-4].
To study the dispersion of the various bands, angle-resolved photoemission experiments have been carried out. The first angle-resolved photoemission study of C60 [17.15] indicated little band dispersion, suggesting that the PES linewidth might arise from vibronic side bands. Subsequent more detailed angle-resolved studies have indicated some dispersion in the HOMO-derived level (~0.4 eV) [17.16], and good agreement between experiment and theoretical calculations [17.17] has been obtained for this amount of dispersion in the C60 valence bands. These experiments and calculations show that the measured dispersion is sufficient to account for the PES linewidth, without need to identify vibronic side bands as a broadening mechanism [17.11,18-20], although vibronic side bands may indeed contribute.
Although most of the PES and IPES measurements at UV photon energies have been done on C60, a significant body of work has already been carried out on C70 [17.21,22] and on higher fullerenes C76, C82, and C84 [17.16,23-25]. The results largely show a decreasing band gap for fullerenes as nc increases, consistent with other experimental probes of the electronic structure and with theoretical expectations [17.5,6]. The photoemission spectra observed for C60, C70, and C84 (see Fig. 17.3) show some remarkable similarities, with a set of bands between 1.5-4.0 eV below the Fermi level EF, another group located about 6 eV below Er, and finally a third set of bands in the 7-9 eV range below EF [17.26]. The broadening in the density of states is also seen in Fig. 17.3 with increasing nc in C„c associ-
IPES threshold Singlet Exciton Triplet Exciton
PES threshold PES peak
Fig. 17.3. Photoemission spectra of C^, C70, and C84 fullerene thin films deposited on gold. Photoemission measurements are made at a
Was this article helpful?