Abstract
Several computer methods of differentiating Auger electron spectra are analysed in terms of their relative signal-to-noise ratios, their distortion of the peak shape and their output intensity as a function of the differentiation amplitude, in order to maximise the measurement precision and detectability in surface analysis. These methods, a direct difference of two channels, the quadratic and cubic smoothing methods of Savitzky and Golay (1964), a variant of the latter and multiple smoothings are each compared in detail with the traditional analogue sinusoidal modulation method. The relative distortions of the techniques are established using the parameters of the Ag M45NN Auger electron doublet proposed by Seah and Anthony (1983). The overall comparisons show that the more complex methods are not necessarily of significantly greater benefit.