Neural network - assisted analysis of X-ray spectra of bimetallic nanoparticles

In X-ray absorption spectroscopy, it is problematic to analyze and interpret polyatomic systems consisting of elements that are neighbors in Periodic Table due to the overlapping edge regions and similar photoelectron scattering properties. This limits our ability to solve the local structure of interesting bimetallic nanocatalysts such as PtAu PdAg, IrPt, and RhAu. We have shown, recently, that X-ray absorption near edge structure (XANES) can be inverted to provide structural properties due the region's sensitivity to photoelectron scattering. Now, we take advantage of XANES sensitivity to electronic structure, specifically charge transfer. In this work, we demonstrate how our new Neural Network XANES (NN-XANES) method can be used to solve the structure of these difficult systems with better accuracy than existing methods. Our work suggests that NNs can yield distinct partial Pt-Au and Pt-Pt coordination numbers from Pt L3-edge XANES in PtAu, a feat impossible with EXAFS and other Z-contrast-limited techniques.