An extensive comparison of full-QM (B3LYP) and QM/MM (B3LYP:UFF) levels of theory has been made for two enantioselective catalytic systems, namely, Pybox-Ru and Box-Cu complexes, in the cyclopropanation of alkenes (ethylene and styrene) with methyl diazoacetate. The geometries of the key reaction intermediates and transition structures calculated at the QM/MM level are generally in satisfactory agreement with full-QM calculated geometries. More importantly, the relative energies calculated at the QM/MM level are in good agreement with those calculated at the full-QM level in all cases. Furthermore, the QM/MM energies are often in better agreement with the stereoselectivity experimentally observed, and this suggests that QM/MM calculations can be superior to full-QM calculations when subtle differences in inter- and intramolecular interactions are important in determining the selectivity, as is the case in enantioselective catalysis. The predictive value of the model presented is validated by the explanation of the unusual enantioselectivity behavior exhibited by a new bis-oxazoline ligand, the stereogenic centers of which are quaternary carbon atoms.

We present a new methodology to predict the enantioselectivity of asymmetric catalysis based on quantitative quadrant-diagram representations of the catalysts and quantitative structure-selectivity relationship (QSSR) modelling. To account for quadrant occupation, we used two types of molecular steric descriptors: the Taft-Charton steric parameter (ν(Charton)) and the distance-weighted volume (V(W) ). By assigning the value of the steric descriptors to each of the positions of the quadrant diagram, we generated the independent variables to build the multidimensional QSSR models. The methodology was applied to predict the enantioselectivity in the cyclopropanation of styrene catalysed by copper complexes. The dataset comprised 30 chiral ligands belonging to four different oxazoline-based ligand families: bis- (Box), azabis- (AzaBox), quinolinyl- (Quinox) and pyridyl-oxazoline (Pyox). In the first-order approximation, we generated QSSR models with good predictive ability (r(2) =0.89 and q(2) =0.88). The derived stereochemical model indicated that placing very large groups at two diagonal quadrants and leaving free the other two might be enough to obtain an enantioselective catalyst. Fitting the data to a higher-order polynomial, which included crossterms between the descriptors of the quadrants, resulted in an improvement of the predicting ability of the QSSR model (r(2) =0.96 and q(2) =0.93). This suggests that the relationship between the steric hindrance and the enantioselectivity is non-linear, and that bulky substituents in diagonal quadrants operate synergistically. We believe that the quantitative quadrant-diagram-based QSSR modelling is a further conceptual tool that can be used to predict the selectivity of chiral catalysts and other aspects of catalytic performance.

No abstract