Classification of local chemical environments from x-ray absorption spectra using supervised machine learning

Abstract

X-ray absorption near-edge structure (XANES) spectroscopy is a robust and element-specific tool for probing the atomic structure of materials. Traditional spectroscopy methods work in the forward direction by simulating XANES spectra from atomic models. Here, the authors present the opposite: a computational method predicting local structural geometry from XANES spectra in which the so-called inverse problem is solved using supervised machine learning. The robustness and fidelity of the method are demonstrated by an average of 86% classification accuracy in the K-edge XANES spectra of hundreds of materials across eight $3d$ transition metal families.