Development of predictive models of π -facial selectivity; a critical study of nucleophilic addition to sterically unbiased ketones

Abstract

Quantum chemical calculations at B3LYP/6-31G^∗ and semiempirical levels have been performed on a series of sterically unbiased ketones, where facial differentiation during nucleophilic additions is electronically induced through distal functional groups. The face selectivity data for fifty-four substrates representing nine different skeleta were computed and compared with the available experimental data on thirty-eight of them. The predictive abilities of various computational methods such as, charge model, hydride model, LiH transition state model, Cieplak hyperconjugation effect estimated by NBO analysis and the cation complexation model have been evaluated. A comparison of the computed and experimental face-selectivity data indicates that the hydride model and the LiH transition state model at the semiempirical levels are the best choices to predict diastereoselectivity. Unexpectedly, the performance of charge, hydride and LiH transition state models are inferior at the B3LYP level compared to the semiempirical methods in predicting the facial selectivities. On the other hand, the Cieplak type hyperconjugation evaluated using the NBO analysis, and cation complexation model are less reliable despite the fact that these two involve higher (B3LYP/6-31G^∗) level calculations. The inadequate performance of the charge model, NBO and the cation complexation models were traced to their emphasis on only one or two factors which are responsible for stereodifferentiation and undermining of the other subtle aspects involving a combination of orbital and electrostatic effects. On the other hand, the hydride and LiH transition state models, at semiempirical levels, provide reliable results to model the face-selectivities.