Digital Pareto-front mapping of homogeneous catalytic reactions

Abstract

We report a digital framework for accelerated exploration and optimization of transition metal-based homogeneous catalytic reactions through autonomous experimentation and Bayesian optimization (BO). Specifically, we utilize a machine learning model constructed with deep neural networks for a rhodium-catalyzed hydroformylation reaction to investigate the role of BO hyperparameters, including the acquisition function and sampling size, on the efficiency of reaction Pareto-front mapping.