John D. Roberts Lecture

Catalysts make reactions occur faster, but their control over reaction rates enables catalytic reactions to occur with selectivities that are different from those of related uncatalyzed reactions. Our group has discovered reactions that occur at one functional group that is less reactive than others and that occur at typically unreactive C–H bonds and the least reactive of these C–H bonds. In each of these processes, we have sought to reveal the mechanisms of the reactions and origins of selectivity. This lecture will include a series of mechanistic puzzles, the solutions of which have led to catalysts with increased reactivity and, in one case, an ability to create macromolecules with unknown combinations of properties. Recognizing the challenge of explaining selectivity from small differences in energy and complex, multistep pathways, we have developed approaches using machine learning to predict the site selectivity of catalytic reactions by encoding the three-dimensional structures of these catalysts in a fashion that allows predictions of the selectivity of reactions with a large diversity of catalysts.