Neural network assisted analysis of bimetallic nanocatalysts using X-ray absorption near edge structure spectroscopy

X-ray absorption spectroscopy is a common method for probing the local structure of nanocatalysts. One portion of the X-ray absorption spectrum, the X-ray absorption near edge structure (XANES) is a useful alternative to the commonly used extended X-ray absorption fine structure (EXAFS) for probing three-dimensional geometry around each type of atomic species, especially in those cases when the EXAFS data quality is limited by harsh reaction conditions and low metal loading. A methodology for quantitative determination of bimetallic architectures from their XANES spectra is currently lacking. We have developed a method, based on the artificial neural network, trained on ab initio site-specific XANES calculations, that enables accurate and rapid reconstruction of the structural descriptors (partial coordination numbers) from the experimental XANES data. We demonstrate the utility of this method on the example of a series of PdAu bimetallic nanoalloys. By validating the neural network-yielded metal-metal coordination numbers based on the XANES analysis by previous EXAFS characterization, we obtained new results for in situ restructuring of dilute (2.6 at% Pd in Au) PdAu nanoparticles, driven by their gas and temperature treatments. Trained neural networks are used to extract the first partial coordination numbers from XANES spectra. In bimetallic nanoparticles, the four local structure descriptors provide rich information on structural motifs.