Data Preprocessing in Peak Shape Analysis of Auger Electron Spectra

 Factor analysis is an established method of peak shape analysis in Auger electron spectrometry. The influence of different commonly used data preprocessing tools onto the results of factor analysis is demonstrated on AES depth profiles of multilayers and implantation profiles. For the analysis of Auger electron spectra it has been traditional to differentiate spectra by Savitzky and Golay’s method to remove background and to elucidate changes in peak shape. For phosphorus implanted in titanium it is shown that background removal works not ideal so that inelastic losses of the Ti(LMM) Auger peak can affect the result of factor analysis for the P(LVV) peak located at ca. 250 eV lower in kinetic energy. The contribution of such losses to the background can be corrected by shifting the spectra so that the high energy side above the peak equals zero. Numerical differentiation can introduce correlated error into the data set. To diminish edge effects the reduction of filter width at the edges and cutting off the outermost data points is recommended. The precision of spectrum reproduction is considered as a crucial test for the number of principal components. The reliability factor is investigated as a measure for the goodness of spectrum reproduction.