Interpretable Catalysis Models Using Machine Learning with Spectroscopic Descriptors

The complexity and dynamics of catalytic systems make it challenging to study the catalysts reactions. Fortunately, advance machine learning (ML) has made descriptor-based catalyst screening rational design a mainstream research approach. Herein, spectroscopic descriptors reported in recent years are highlighted field catalysis. Both vibrational spectra X-ray absorption have demonstrated strong ability predict structures properties. Through several cases, interpretable ML models based on discussed reveal physical knowledge mechanism exhibit superiority transfer tasks imperfect data scenarios. Finally, this Viewpoint, we illustrate challenges with provide perspectives.