Accelerating transition state search and ligand screening for organometallic catalysis with reactive machine learning potential

Organometallic catalysis lies at the heart of numerous industrial processes that produce bulk and fine chemicals. The search for transition states and screening for organic ligands are vital in designing highly active organometallic catalysts with efficient reaction kinetics. However, identifying accurate transition states necessitates computationally intensive quantum chemistry calculations. In this work, a reactive machine learning potential (RMLP) model is developed to expedite transition state optimizations and ligand screenings for organometallic catalysis based on an automatic transition state database construction method and a higher-order equivariant message passing neural network. In a case study involving the ethylene hydrogenation reaction catalyzed by organometallic catalysts, RMLP rapidly predicts potential energy surfaces along intrinsic reaction coordinate paths, achieving speeds nearly three orders of magnitude faster than those of rigorous quantum chemistry calculations. Meanwhile, it maintains comparable accuracy with a root mean square deviation of 0.221 Å for transition state geometries and a mean absolute error of 0.135 kcal/mol for reaction barriers on the external test set, significantly outperforming semi-empirical quantum chemistry methods. Our RMLP model offers an effective alternative to both rigorous and semi-empirical quantum chemistry approaches for rapid and precise transition state optimizations, facilitating high-throughput and large-scale screenings of advanced organometallic catalyst ligands.