X-ray absorption spectroscopy (XAS) is a premier element-specific experimental technique for materials characterization in catalysis. Specifically, X-ray absorption near edge structure (XANES) carries rich local structural and chemical information around X-ray absorbing species, which makes it a powerful tool to probe catalytic processes under <i>in situ</i> and <i>operando</i> conditions. However, the correlation between XANES spectral features and the underlying local structural motifs and electronic properties is obscure, which often requires prior knowledge of fingerprints of known crystals or modeling from first-principles theory to uncover. Here we show how data science can be combined with theory to decipher the structure-spectrum relationship from XANES, with only a small amount of experimental data typically available for training. Such an approach can be applied in multiple contexts. We demonstrated how to combine machine learning algorithms and XANES modeling to predict the 3D structures of metal nanoparticles, how to determine the local structure characteristics in amorphous materials from a XANES database, and a proof of concept on local chemical environment classification using XANES database and machine learning methods.

Unraveling local atomic structures from X-ray absorption spectroscopy using theory and machine learning

2019 Spring National Meeting

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.