Appropriate basis selection based on Bayesian inference for analyzing measured data reflecting photoelectron wave interference

Abstract: In this study, we applied Bayesian inference for extended X-ray absorption fine structure (EXAFS) to select an appropriate basis from among Fourier, wavelet and advanced Fourier bases, and we extracted a radial distribution function (RDF) and physical parameters from only EXAFS signals using physical prior knowledge, which is to be realized in general in condensed systems. To evaluate our method, the well-known EXAFS spectrum of copper was used for the EXAFS data analysis. We found that the advanced Fourier basis is selected as an appropriate basis for the regression of the EXAFS signal in a quantitative way and that the estimation of the Debye-Waller factor can be robustly realized only by using the advanced Fourier basis. Bayesian inference based on minimal restrictions allows us to not only eliminate some unphysical results but also select an appropriate basis. Generally, FEFF analysis is used for estimating physical parameters such as Deby-Waller and extracting RDF. Bayesian inference enables us to simultaneously select an appropriate basis and optimized physical parameters without FEFF analysis, which results in extracting RDF from only EXAFS signals. These advantages lead to the general usage of Bayesian inference for EXAFS data analysis.