Expedited Screening of Active and Regioselective Catalysts for the Hydroformylation Reaction

The discovery of new homogeneous catalysts that preferentially form one product over another in regio- or enantioselective chemical reactions has traditionally been the province of experimental chemists. Today, computational-based approaches have carved an increasingly important role, which, for computational catalytic designs, often rely on highly inefficient combinatorial-based screening methods. To increase the pace of discovery, tools capable of rapidly assessing large numbers of prospective species and identify those possessing desirable properties, such as activity and selectivity, are vital. Here, through the examination of the hydroformylation of 2-methylpropene, we demonstrate how a new tool built upon molecular volcano plots can be used to quickly predict the activity of molecular catalysts as well as estimate the intrinsic ability of each species to form one regioisomer over the other with striking accuracy. Following training and validation, these regioselective molecular volcanoes are employed to predict catalysts that preferentially form the branched product (2,2-dimethylpropanal) in violation of Keulemans' 70-year-old law. Eighteen species (out of a total of 68 predicted) were computationally predicted to have regiomeric excess (r.e.) values >90. Overall, these tools can be used to quickly screen the activity and selectivity of potential catalysis based on two easily computed descriptor variables.