Quantitative Analysis of X-Ray Spectral Data for a Mixture of Compounds Using Machine-Learning Algorithms

Based on machine-learning algorithms, a method is developed for determining the structural parameters of the components of a mixture from X-ray absorption spectra. For each component, a database of spectra is constructed for all possible deformations of its structure. The machine-learning method implemented in the PyFitIt software package allows quick calculation of the spectrum for deformations of structures from the considered family and optimization of the structural parameters of the mixture by fitting the theoretical spectrum to the experimental one. The capabilities of the method are examined by analyzing changes in the structural characteristics and concentrations of the components of the mixture for the bis-dioxolene complex of cobalt with functionalized iminopyridine ligands during its valence-tautomeric interconversion depending on temperature.