Automatic Identification of X-ray Absorption Fine Structure Spectra via Machine Learning

Abstract

X-ray absorption fine structure (XAFS) spectroscopy is a characterization method that can be used to uncover information about the material electronic structure and structural information. XAFS analysis is generally performed by comparing available spectra and relies on experience and knowledge. This work utilizes supervised machine learning with descriptors derived from XAFS and physical quantities of elements to establish an automated and rapid classification method for oxidation states. Two classification methods are explored: a general classification of whether a material is an oxide and a valence number classification. As a result, descriptors and a machine learning model to identify the oxidation states are unveiled where oxidation states are predicted with high accuracy. These results show that the oxide and valence classifications of the target materials can be made with high accuracy from XAFS spectral information according to highly dimensional and complex patterns.