Bayesian estimation analysis of X-ray photoelectron spectra: Application to Si 2p spectrum analysis of oxidized silicon surfaces

X-ray Photoelectron Spectroscopy (XPS) is known as a powerful experimental technique that provides information on the chemical states of material surfaces, and curve-fitting analyses based on methods such as leastsquares are widely used for spectrum analysis. However, a major issue with these analyses is that the number of spectral components and other analytical conditions often include qualitative or arbitrary settings. In response to this, we applied Bayesian estimation to the spectral data of the oxidized state of silicon (Si) surfaces, which we previously reported. Bayesian estimation discussed in this paper is based on stochastic modelling without incorporating physical properties such as Si oxidation. By applying this to the time-dependent oxidation spectra of Si surfaces, we reconfirmed that latent information on number of peaks, peak positions, intensity changes and spin-orbit splitting conditions can be extracted through Bayesian posterior probabilities.