Surface Science 0 Northllolland
100 (1980) Publishing
SURFACE SCIENCE LETTERS DEEXCITATION
MECHANISMS IN METASTABLE
H. CONRAD, G. ERTL, J. KiiPPERS and W. SESSELMAN Institut fiir Physikalische Chemie, Universitci’tMiinchen, Sophienstrasse Germany
II, D-8000 Mtinchen 2,
and H. HABERLAND Fakuitd Received
fiir Physik, Universitit Freiburg,
29 April 1980
Electron emission caused by impact of metastable He atoms on surfaces can either proceed by a two-stage resonance ionization+ Auger neutralisation (RI + AN) or by a one-stage Auger deexcitation (AD or Penning ionization) mechanism. The RI + AN mechanism will dominate with clean transition metal surfaces and may be suppressed for example by lowering the local work function as demonstrated for K and Cs adlayers.
Electron emission caused by impact of metastable He (2’S or 23S) atoms (He*) on clean or adsorbate covered metal surfaces has recently been used to obtain information on their electronic properties [l-4]. Although there is general agreement concerning the extreme surface sensitivity of this technique there exists still controversial discussion on the interpretation of the obtained spectra. The present contribution is intended to cast new light on the mechanisms of He*/surface interactions and to specify the conditions under which well-defined analysis of the data may be achieved. If an excited He atom (E* = 20.6 eV for 2lS, 19.6 eV for 23S) interacts with a gaseous species.M, electron emission occurs through an Auger deexcitation (AD) or Penning ionization process AD:
whereby the kinetic energy distribution of the electrons is governed by the ionization potentials of M [5,6]. Theoretical models have been developed in order to describe this process on a first-principles basis [7,8]. The mechanism of its operation at a metal surface is illustrated schematically by fig. la. An alternative mechanism of deexcitation and electron emission from a solid surface is represented by fig. 1b: Here the excited He* electron tunnels into an empty level of the solid above the Fermi level EF (resonance ionisation, RI), the remaining L46 1
T % @-i&l
1 I _
20 18 16 1.~32 10 8
Fig. f . PotenhI energy diagrams for a He* atom in front of a metal surface @strating the two possible mechanisms of deexcitation: (a) Auger deexcitation (= Penning ionization), (b) fesonance ionization i- Auger neutraiisation. Fig. 2. Kinetic energy distributions of electrons emitted from K-overfayers: (a) KIPd(f 111, Be* (2lS) excitation,(b) K/Ni(lOOf, We+excitation (see ref. ).
He* ion is subsequen~y neutrabsed and the energy of this process is rebased via the kinetic energy of an Auger electron (Auger neutraiisation, AN): RI:
He* + M -+ He’ + M-,
This latter AN mechanism is obviously identical to that involved in ion neutralisation spectroscopy (INS) as developed by Hagstrum [9,10], where He’ ions are striking the surface. Because only one electron from the target is involved in the AD process, measured electron distributions reflect directly the density of valence states of the target, if the AD mechanism (1) is operating. Otherwise they are determined by the se~~~on~olut~on of the density of states since the AN step (2b) is an effective twoelectron process [9,10]. The crucial point is ob~ously whether a large fraction of the He* atoms deexcited at the surface is ionized by resonance tunneling (2a) or not. The following results will demonstrate that this step may be suppressed by the presence of proper adsorbate layers, but that on the other hand it will dominate with clean transition metal surfaces which provide empty states above EF. The experimental arrangement used in the present work has been briefly outlined previously 131 and will be described in full detail elsewhere. Metastable He atoms were produced by electron impact of a nozzle beam [ 111, Since electron emission may also be caused by photons and fast ground-state He atoms  it
H. Conrad et al. / Deexcitation
was checked by time-of-flight analysis that the flux of these particles was always less than 10m3 of the He* flux. The 2lS : 23S ratio was about 5 : 1 so that the presented spectra are to a good approximation characteristic for He* 2lS excitation. The resonance ionization process (2a) is governed by the overlap of the singly occupied He 2s level with an empty level above EF, while the Auger deexcitation process (1) is governed by the overlap of a filled level below E, of the target with the He 1s hole. As the “diameter” of the He 2s orbital is considerably larger than that of the 1s orbital, it becomes highly probable that resonance ionization will dominate if empty orbitals at the target near the He* 2s energy are available. If this process can be suppressed, electron emission will switch from a two-stage (RI + AN) to a one-stage process. This can be achieved in the following ways: (i) If the target IS an insulator or semiconductor where the energy of the He 2s level falls into the band gap. This situation is obviously found with condensed aromatics where the operation of the AD process has been established [ 131. (ii) If close approach of the He* atom to the solid is inhibited by the presence of a dense layer of adsorbed molecules which themselves possess no empty levels near the He* 2s energy which might assist resonance tunnelling. CO adsorbed on Pd(ll1) is an example of this type  where the spectra reflect the operation of the AD process and exhibit emission from the CO-derived 50 t In and 40 levels (similar to UPS), but no features from the metallic d band which is geometrically “shielded” by the presence of the adlayer. A theoretical treatment of the Penning ionization (AD) process at an oriented CO layer  revealed that the electronic transition occurs indeed at distances >2 A from the 0 nucleus which in turn is separated by 2.5 A from the topmost metal atoms. This “shielding” effect will be not so effective with small atomic adsorbates which in addition form no denselypacked adlayer such as 0: spectra obtained with Oad-covered Pd(lI1) indeed strongly suggest that the two-stage RI + AN mechanism is still predominating [ 151. (iii) If the work function @Jof the metal is lowered so that its Fermi level is above the He 2s level. This can, e.g., be achieved by adsorption of alkali metal atoms. Fig. 2. shows the electron energy distribution from Pd(ll1) with adsorbed K. The work function was lowered to a value of 1.6 eV so that resonance ionization of He* is inhibited. The maximum kinetic energy Ek,,,= = 19.0 eV is just equal to the difference of the He* excitation energy and the work function, 20.6-1.6 eV. The kinetic energy scale can thus be directly converted into a scale for the initial statebinding energy with respect to the Fermi level EF, E, = E* - @ - Ekin = 19.0 EG,, (eV). The spectrum then exhibits a relatively narrow peak just below EF, which is ascribed to a resonance level derived from the K 4s state, and two relatively broad maxima around E, = 7 and 10 eV. The latter are tentatively interpreted as arising from traces of oxygen, since it is extremely difficult to keep a K-adlayer free from oxygen. Similar experiments with a K-covered Ni(lOO) surface were recently performed by Hagstrum  who used He’ ions instead of He* atoms as incident particles. The resulting data are included in fig. 2 and show remarkable agreement. As dis-
H. Conrad et al. / Deexcitation
18 16 It
12 10 8
Fig. 3. He* (2l S) excited spectra from clean and Cs covered Pd(ll1) surfaces: (a) surface saturated with Cs at 300 K, (b) after brief annealing to 750 K, (c) 9.50 K, (d) clean surface. All data were recorded under identical instrumental conditions. The scale of kinetic energies refers to the vacuum level Evac of the sample, using the low-energy cut-off of electron emission as zero reference.
cussed by Hagstrum [ 161, in this case an electron tunnels from the solid to the He’ ion forming a He* (23S) atom which subsequently undergoes Auger deexcitation, so that similar kinetic energy distributions of the emitted electrons result, irrespective whether He* or He’ are the primary particles. The present findings support nicely these conclusions. (The peaks in Hagstrum’s spectrum occur at about 1 eV lower kinetic energies, which corresponds to about the difference of the excitation energies of He 2’S and 23S. Another source for this difference might arise from the fact that with He’ ions the deexcitation process occurs probably somewhat closer to the surface.) Further insight into the local character of this work function effect (which obviously determines if the RI + AN or AD mechanism is operating), is obtained from the following results with a Pd(ll1) surface covered by various amount of (oxi-
H. Conrad et al. / Deexcitation mechanisms
dized) Cs: Fig. 3, curve a, shows the He* 2lS spectrum from a surface saturated with Cs at 300 K. Ek, _,x = 18.2 eV (gEF) is smaller by 0.6 eV than the corresponding value obtained with UPS (hv = 2 1.2 eV) and indicates the operation of the one-stage AD mechanism. The work function is $J= 20.6 - 18.2 = 2.4 eV. A pronounced double-peak structure just below EF is ascribed to emission from states derived from the Cs 6s level. Similar features have been observed in UPS measurements with other Cs/metal systems . A peak at Ek = 11 .l eV (EB = 7.1 eV) is ascribed to an oxygen-derived level [ 181 and the two weak maxima at Ek = 5.4 and 3.6 eV (EB = 12.8 and 14.6 eV) arise from the Cs Sps,z and 5pr,z states. These latter peaks persist if the Cs coverage is lowered by continuous heating of the sample (curve b + c) and indicate that the AD mechanism is operating whenever an He* atom is deexcited at an adsorbate Cs atom. The peak below EF is, on the other hand, rapidly decreasing with decreasing coverage which might be due to a higher degree of ionicity of CS,~ at lower coverages (i.e. the 6s level shifts above EF) and/or to variations of the oxidation state. Parallel to these effects a broad increase of the background up to E, = 12 eV is observed with decreasing Cs coverage. After complete desorption this is the only structure left (curve d). Its shape is qualitatively rather similar to those obtained from clean metal surfaces by means of INS [lo]. The maximum kinetic energy of the electrons is about 12 eV. This is just the value expected for an RI + AD (= INS) mechanism if both electrons from the solid originate from the Fermi level, namely Ek, max =Ei - 2rP. Ei is the effective
electron affinity of He’ (= ionization energy of He) in front of the metal surface which is lowered by about 2 eV with respect to the free gas phase value due to image force effects . $I (x.5 .O eV) is the work function of the clean Pd(l11) surface. Indeed deconvolution of the energy distribution of fig. 3d using Hagstrum’s data reduction procedure developed for INS yielded an initial-state density of states which exhibited quite similar structure as the corresponding UPS (oneelectron) data [ 191. These results demonstrate that there is no sudden switching from the AD to the RI + AN mechanism at a coverage where the overal work function just passes through the energy of the He 2s level. Instead it has to be concluded that the incoming He* atoms “feel” the local work function near their point of impact: this will be high (+RI + AN) on bare Pd patches and low (-+AD) in the vicinity of an adsorbed Cs atom. Fruitful discussions Forschungsgemeinschaft
with G. Doyen as well as financial support by the Deutsche (SFB 128) is gratefully acknowledged.
H. Conrad et al. / Deexcitation
References [l] C. Boiziau,C. Garot, R. Nuvolone and J. Roussel, J. Phys. 39 (1978) L339.  P.D. Johnson and T.A. Delchar, Surface Sci. 77 (1978) 400.  H. Conrad, G. Ertl, J. Kiippers, S.W. Wang, K. Gerard and H. Haberland, Phys. Rev. Letters 42 (1979) 1082.  C. Boiziau, C. &rot, R. Nuvolone and J. Roussel, Surface Sci. 91 (1980) 313. [S] A. Niehaus, Ber. Bunsenges, Physik. Chem. 77 (1973) 632.  H. Hotop, Radiation Res. 59 (1974) 379.  W.H. Miller, J. Chem. Phys. 52 (1970) 3563.  A.P. Hickman, A.D. Isaacson and W.H. Miller, J. Chem. Phys. 66 (1977) 1483.  H.D. Hagstrum, Phys. Rev. 96 (1954) 336; 104 (1956) 672. [ 101 H.D. Hagstrum, in: Electron and Ion spectroscopy of Solids, Eds. L. Fiermans, J. Vennik and W. Dekeyser (Plenum, New York, 1978) p. 273. [ 111 H. Brutschy and H. Haberland, J. Phys. El0 (1977) 90.  J. Roussel and E. Labois, Surface Sci. 92x1980) 561. [ 13) T. Munataka, T. Hirooka and K. Kuchitsu, J. Electron Spectrosc. 13 (1978) 219. [ 141 S.W. Wang and G. Ertl, Surface Sci. 93 (1980) L75. [ 151 Unpublished data. [ 16) H.D. Hagstrum, Phys. Rev. Letters 43 (1979) 1050.  G. Ebbinghaus, W. Braun, A. Simon and K. Berresheim, Phys. Rev. Letters 37 (1976) 1770; S.A. Lindgren and L. Wallden, Solid State Commun. 28 (1978) 283. [ 181 The O/Cs system is characterised by the formation of a whole series of suboxides giving rise to various binding energies of the oxygen-derived levels. [ 191 H.D. Hagstrum, personal communication.