Solid State Communications, Vol. 79, No. 2, pp. 191-195, 1991. Printed in Great Britain. 0038-1098/91 $3.00 + .00 Pergamon Press plc THE PHOSPHORUS AUGER L2.3VV SPECTRUM OF InP(1 1 0) P.O. Nilsson Department of Physics, Chalmers University of Technology, S-412 96 G6teborg, Sweden and S.P. Svensson Martin Marietta Laboratories, Baltimore, MD 21227, USA (Received 4 March 1991 by B. Lundqvist) The L2,3VV Auger spectrum of InP(1 1 0) has been measured using synchrotron radiation excitation. In a first step a one-body theory was applied for analysis, involving convolutions of partial, local density of states functions obtained from a band calculation. The effect of hole- hole correlations in the valence band was then evaluated using the Cini-Sawatzky model. 1. INTRODUCTION AUGER electron spectroscopy has mainly been used as a routine tool for qualitative and quantitative analysis, in particular of surfaces. However, the technique can also be used for investigations of the electronic structure of solids. For studies of valence bands the CVV transitions are of special interest. In such a process the initial state consists of one core hole while the final state contains two holes in the valence band and an escaping Auger electron. The final state rule [1] says that the detected energies are those of the final state, while the transition amplitudes also depend on the initial state. Thus, as the final state contains no core hole in the CVV process (in contrast to e.g. the CCV process) we expect to test the unscreened band structure in an independent particle model. This situation should prevail when the effective Coulomb interaction Uar is much smaller than the valence band width W. In the Cini-Sawatzky model [2] the con- dition is U~J2W ~< 1. In the independent particle picture the measured spectrum is a sum of weighted convolutions of ground state density of states func- tions. As the perturbation operator in the Auger pro- cess is the Coulomb interaction the selection rules are very much relaxed compared to the dipole selection rule for optical transitions. However, combined with simple theory, information may be extracted about partial (i.e. angular momentum decomposed) density of states functions. In addition one of the most valu- able characteristics of Auger transitions is that the local valence density of states is tested. This site sen- sitivity, not present in e.g. photoemission, is of par- ticular importance for investigations of alloys and compounds since it may reveal information about the bonding of the constituents, in particular for more complicated compounds such as ternaries and quar- ternaries. Valence Auger spectra of metals have been treated in the literature in some detail [3], while spectra from semiconductors are more scarse [4]. Essentially only the Si L2,3VV spectrum has been thoroughly analyzed [5]. As there is now a very intensive research taking place on the electronic structure of surfaces and inter- faces of semiconductors it is of great importance to find out if a simple model can be applied to the analy- sis of CVV spectra for these materials. Specifically we want to determine the influence of the many-body effects on the CVV transitions and relate it to earlier studies on Si [5]. From the outset it is not clear whether a single-particle model is at all adequate. For example, even the application of the cited Ue~/W rule is not trivial for a semiconductor since several bands are involved, in some cases separated by a gap; the heteropolar gap in compound semiconductors. This means for example that one single value for Ueer may not be sufficient. 2. EXPERIMENTAL DETAILS The P L2.3VV transition in InP waschosen due to its favourable energy position relative to other spec- tral features. Only a few other, very weak, transitions from either of the constituents are expected to overlap with the peaks we are studying [6]. Furthermore, the P Lz,3VV transition appears at a sufficiently high energy so that the large background of the low energy secondary electrons does not interfere with the low energy part of the peak. These facts make it possible to obtain a very clean spectrum without complicated 191