Mechanistic and Machine Learning Insights into Borrowing Hydrogen Reactions Catalyzed by Transition Metal Complexes with N-Heterocyclic Ligands

Abstract

Pyrazole-based transition metal complexes have attracted increasing attention in borrowing hydrogen (BH) reactions, particularly when employing the metal–ligand cooperation (MLC) strategy to achieve high activity and selectivity. However, a systematic understanding of their mechanistic selectivity and the factors governing catalytic performance remains lacking. Herein, the BH reaction of alcohols catalyzed by such complexes was investigated using density functional theory (DFT) calculations in combination with machine learning (ML) methods. Three possible pathways—N₂-site-assisted, O-site-assisted, and N₁-site-assisted—were proposed, among which the N₂-site-assisted route was identified as the most favorable. Both the dehydrogenation and hydrogenation steps proceed via an outer-sphere concerted transfer mechanism. Distortion/interaction analysis revealed that the ligand-assisted distortion energy plays a decisive role in determining the activation barrier. Furthermore, an ML model with high predictive accuracy (R² = 0.96) was established to correlate catalytic performance with electronic and steric descriptors. Feature importance analysis identified the HOMO energy level, dipole moment, and molecular volume as key factors, reflecting the roles of electron-donating ability, transition-state polarization, and steric effects, respectively. This study not only deepens the mechanistic understanding of MLC-enabled BH reactions catalyzed by pyrazole-based transition metal complexes but also provides a predictive framework for the rational design of efficient and tunable catalysts.