Abstract
A theoretical database comprising experimentally accessible and inaccessible chemical reactions could complement the existing experimental databases and contribute significantly to data-driven chemical reaction discovery. Quantum chemistry-aided retrosynthetic analysis (QCaRA) can generate a network of elementary steps called a reaction-path network and predict hundreds or more of chemical reactions along with their theoretical yields. In contrast to ordinary simulations, QCaRA traces back the reaction paths from the target product to various reactant candidates while solving the kinetic equations. In this study, we propose theoretical reaction database construction based on QCaRA. Seven reaction-path networks containing 13,190 reactions, 108,754 reaction paths, and 2,552,652 geometries have been identified and discussed as examples. In addition to well-known reactions (i.e., synthesis of fluoroglycine, Wöhler’s urea synthesis, base-catalysed aldol reaction, Lewis-acid-catalysed ene reaction, cobalt-catalysed hydroformylation, Strecker reaction, and Passerini reaction), numerous unexplored reactions with high, medium, low, near-zero, or zero yields have been identified. We anticipate that such a QCaRA-based theoretical reaction database will provide information on hitherto unexplored reactivities, especially those that are experimentally inaccessible.