Quantitative AES and XPS: convergence between theory and experimental databases

Abstract

Abstract Experimental spectral databases have been recorded for AES and XPS using fully calibrated instruments. These instruments have been calibrated so that the spectra have the true shape and peak area intensities may be integrated to give absolute yields for AES and relative yields for XPS. Removal of all the backgrounds requires care but may be completed by using information from both databases. The resultant yields may be compared with theory. The correlations for AES are the more complex and involve the total intensities for all transitions originating in each shell. The correlations are excellent using significant changes to the traditional approach. These involve the use of the Casnati et al. ionisation cross section and the restriction of the number of electrons for use in the inelastic mean free path calculations to electrons of 14 eV or less binding energy in the s, p or d sub-shells. The average ratio of experiment to theory is 1.04 with a standard uncertainty of the mean of 4%. Results for XPS are excellent using Scofield’s ionisation cross section together with the above rules for the inelastic mean free path calculations. Improvements for certain elements are still needed for removing the inelastically scattered Auger and photoelectrons in both databases. To assist analysts in using such databases a simpler measure of Auger electron intensity is developed involving differential spectra broadened with a Gaussian function of 15–20 eV width. The peak-to-peak intensities from these broadened spectra are reasonably closely related to the peak area of the direct spectra except in a few exceptional cases. The unbroadened differential spectra show strong contributions from the spectrometer resolution and changes in the chemical state which are avoided by the spectral broadening. To simplify calculations for the analyst when studying homogeneous materials by AES and XPS, the relative sensitivity factors are re-defined to be for an average matrix instead of the pure element. This leads to a matrix-less equation for calculating compositions from the spectra.