Application of a Convolutional Neural Network for Automated Analysis of X-ray Photoelectron Spectra of Heterogeneous Catalysts

A convolutional neural network was used to solve the problem of segmentation of XPS spectra. The developed combination of recognition using a machine learning model and a post-processing algorithm provided fast automatic analysis of XPS data. The results of determining the positions and areas of peaks were in good agreement with both the results of manual analysis and handbook values. The proposed approach was applied to analyze the XPS spectra of heterogeneous catalysts (Pd/Al2O3 and Sr2TiO4) and chemical compounds used in the preparation of catalysts (AgCl and TiO2).