MAXIMUM-LIKELIHOOD COMMON FACTOR-ANALYSIS IN AUGER-ELECTRON SPECTROSCOPY

An advanced multivariate statistical technique, maximum likelihood common factor analysis (MLCFA), is proposed as an alternative method to principal component analysis (PCA). From a critical review of both methods, it is shown that MLCFA is superior by far to PCA for decomposing overlapping peaks into their component spectra since it is not subject to the various arbitrary factors that characterize the PCA procedure and since it requires less a priori information about the system under study. Investigation of computer-simulated sets of related spectral mixtures that simulate a series of background-corrected Auger spectra in a depth profile confirms that, compared to PCA, MLCFA leads to an improved determination of the correct number of relevant components. When treating real Auger spectra, a correct interpretation of the experimental background, ion sputtering effects, etc. is necessary, as discussed previously in the literature. The present results, however, indicate that the implementation of MLCFA in the field of Auger electron spectroscopy contributes substantially to the correct characterization of the chemical composition of solid surfaces and interfaces.