Speciation of Ru Molecular Complexes in a Homogeneous Catalytic System: Fingerprint XANES Analysis Guided by Machine Learning

Abstract

X-ray absorption spectroscopy is a powerful tool for the characterization of local atomic structure. Commonly, bond lengths and coordination numbers are extracted from the extended energy region of the spectrum (extended X-ray absorption fine structure, EXAFS). However, for many diluted systems, such as homogeneous catalysts, with a low concentration of the active component and under in situ or operando conditions, one cannot collect sufficient EXAFS data for a quantitative analysis. Considering the case of a homogeneous ruthenium-based catalyst, where the ligand surrounding the ruthenium atoms can change from Br to CO depending on the reaction conditions, we establish here an effective machine learning approach based on the descriptor analysis of spectral features. After the training procedure, the algorithm predicts both the ligand surrounding ruthenium and the distances to Br and CO ligands. The prediction quality of the approach was verified by means of a cross-validation procedure applied to the mixture of compounds and was validated for experimental spectra of reference RuBr3 and [RuBr2(CO)3]2 complexes. This work describes a practical route to improve classical fingerprint analysis and linear combination fit by more sophisticated data science algorithms.