Automatic oxidation threshold recognition of XAFS data using supervised machine learning
Oxidation states of materials are characterized by the X-ray absorption near edge structure (XANES) region in X-ray absorption spectroscopy (XAS). However, the challenges in identifying oxide states are strong depending on the researcher’s judgment based on shift changes between measured XAS and reference spectra data. Here, automatic oxidation threshold recognition is performed using machine learning and experimental XAS spectra. In particular, a workflow from experimental data collection, data preprocessing and prediction using machine learning are proposed. 10 descriptors for recognizing the oxide state in XAS spectra are discovered. More importantly, the oxide states of unknown experimental XAS spectra are identified using a trained machine. The proposed approach thus allows for the machine learning to automatically recognize the oxidation threshold of a given XAS spectra without the presence of reference data, leading to the fast analysis of XAS spectra.