Multiparameter Machine Learning Quantifies Electronic Dominance in Pd-Catalyzed Formic Acid Dehydrogenation

Given the limited efficiency of geometric optimization in enhancing formic acid dehydrogenation (FAD), advancing Pd-based catalysts requires deeper insight into electronic structural regulation. Here, we developed a catalytic system confined within metal-nitrogen-doped carbon supports (Pd@MNC) and applied machine learning to innovatively establish a multiparameter correlation model integrating intrinsic kinetic barriers (E_ads) with diverse descriptors. Unlike traditional single-factor analyses, our findings unravel the central role of electronic structure engineering (48% relative importance) over geometric tunability (12%) in regulating catalytic kinetics, with the d-band center offset (ε_d, 30%) and Pd(II) proportion (ω_Pd(II), 18%) accounting for the electronic contribution. Validated experimentally via Co and Cr doping, this theory-based machine learning framework offers a predictive paradigm for rational catalyst design and activity trend. Ultimately, this multidimensional electronic regulation strategy elevates FAD performance while providing broad applicability for accelerating other critical Pd-catalyzed processes, such as Suzuki coupling and CO₂ reduction reactions.