Rational design of Pd-catalysts for C-H activation and Ru-catalysts for C=O hydrogenation: from operando X-ray absorption spectroscopy identification of metal complexes to multi-technique machine learning-based characterization

 X-ray absorption spectroscopy is a powerful technique to study the active compounds in catalytic cycles, since it is element specific and can thus selectively probe the catalytic species in complex mixtures. Its major drawbacks include the complexity of the data analysis when simultaneously studying multiple catalytic species and the lack of the required reference compounds for reactive intermediates. Here, we will develop a new and powerful analytical method by first applying statistical methods to extract the spectra of all present catalytic species and then using artificial intelligence (AI) to identify all unknown intermediates for which no references are available. The AI algorithm will be trained with a large library of calculated spectra of hypothetical intermediates and will be much stronger than the current analytical algorithms, because it will simultaneously analyze FTIR, XANES and EXAFS spectra. The AI algorithm will be applied on two highly topical reactions, the Ru-catalyzed hydrogenative deoxygenation of bio-based polyols to olefins and the Pd-catalyzed arene alkenylation by C-H activation. The newly developed AI algorithm will allow us to identify all species in the catalytic cycles and get a profound knowledge of the reaction mechanism and its side reactions. Based on these insights, we will be able to design new catalysts in a much more structured and effective way than the current intuitive optimization, finally reaching a much higher performance.