From Semiempirical Descriptors to Catalyst Discovery: Interpretable Machine Learning for Ethylene Polymerization

Abstract The discovery of high-performance ethylene polymerization catalysts remains challenging because catalytic behavior is governed by a complex interplay between ligand structure, conformational flexibility, and reaction conditions. Herein, we present a data-driven framework for catalyst discovery that integrates semiempirical descriptor engineering, interpretable machine learning, and DFT analysis. Using an automated workflow based on SMILES-derived structures, conformer sampling, and xTB/RDKit descriptors, chemically meaningful representations were constructed for a diverse set of Co- and Fe-based bis(imino)pyridyl catalysts. The resulting structure-condition models accurately predict catalytic activity, achieving an R2 of 0.86 with the optimized XGBoost model. Interpretable analysis establishes a local angular descriptor associated with the N-aryl substituent geometry as an important structural factor governing catalytic activity, which is further supported by DFT calculations. Guided by these insights, systematic ligand design establishes practical rules favoring reduced proximal steric congestion and electronically tunable ortho substituents, enabling the identification of promising Fe/Co catalyst candidates. This framework provides a general strategy for accelerating the rational discovery and design of homogeneous polymerization catalysts.