2022 Volume 65 Issue 3 Pages 121-127
This is a research report about fully automated analysis of X-ray photoelectron spectroscopy (XPS). We developed a fully automated method to perform XPS spectral analysis based on the information criteria. Our method searches a large number of initial fitting models by changing the degree of smoothing, and obtains a series of fitting results. The goodness of those optimized models is ranked using information criteria. We found that, using the Akaike information criterion, a complicated model tended to be selected, with a larger number of peaks than expected from the spectral shape. On the other hand, using the Bayesian information criterion (BIC), a simple model with reasonably good agreement and a moderate number of peaks was selected. The model selected by the BIC was close to the result of peak fitting performed by XPS analysis experts. We also present the difference in modeling between Gaussian noise and Poisson noise.