AmericanChemicalSociety.com

Transition metal-catalyzed enantioselective hydrogenation of functionalized alkenes has become widely accepted as an important method for the synthesis of chiral, non-racemic organic compounds not only in basic research laboratories, but also in large-scale industrial manufacturing processes. Examples of industrial applications include the production of α- and β-amino acids, L-dopa, α-tocopherol, and Lotrafiban. Although such applications have become commonplace, the selection of chiral ligands and catalysts to optimize the enantiomeric excess (e.e.) for any given hydrogenation substrate is largely accomplished by trial-and-error screening, which can be time-consuming and costly in light of the large variety of chiral, often expensive catalyst systems that are available. Also, in the course of developing new chiral catalysts, there is generally no assurance in advance of superior performance compared to previously known catalysts.

Our specific aim is to develop computational protocols for prediction of the enantiomeric excess (e.e.) of the product resulting from any combination of either a known or previously unknown catalyst with a desired substrate. With such a computational tool, virtual in silico screening of thousands of potential combinations could be conducted, and only those for which high e.e.s were predicted would be selected for experimental validation. It would make no difference whether a given catalyst was previously known or not. Considerable effort would be saved by not choosing to prepare a new catalyst unless high e.e.’s were predicted. By coupling this protocol with structural diversity algorithms, entire virtual libraries of new catalysts could be created and screened for e.e. before any of the catalysts are prepared. In order to be feasible, a fast computational tool is required to handle the large numbers of catalyst/substrate combinations envisioned in this strategy. High-level quantum mechanics-based methods can be very accurate in the prediction of transition state structures and energies as required for e.e. prediction but are very CPU intensive and do not allow for conformational sampling. Molecular mechanics force field calculations are very fast but are normally limited to the study of ground state structures, which do not lend themselves to reliable prediction of relative reaction rates for the formation of two enantiomers. To accomplish our goals, we have adopted the Q2MM method developed by Professor Per-Ola Norrby of Gothenburg University in Sweden. This method allows for the automated generation of force fields for a transition state of a reaction, once the mechanism of the reaction is established and the transition state structure has been computed quantum mechanically. When applied to the enantioselectivity determining step of a mechanism, this method uses the force fields for the rapid calculation of diastereomeric transition state energies, which in turn allows the prediction of relative rates of formation of two enantiomers and therefore the e.e. for any given chiral catalyst/substrate combination.  Once optimized, this procedure permits the virtual screening of hundreds of combinations in just a few hours with a multiple node computer cluster. The implementation of the Q2MM method is done in close collaboration with Professor Norrby.

The first substrate class to be investigated in this program was the enamide-based dehydroamino ester system, which has served admirably in the enantioselective synthesis of α-amino acids. It was chosen for our initial studies due to the wide range of published data sets that were available for the use of a very large number of different chiral catalysts in these reactions. As we have reported in recent publications, we were successful in the development of the required force fields, which in turn proved highly successful in the prediction of e.e.’s for test set catalyst/substrate combinations. An R² correlation factor of 95% was seen between predicted and experimental values for over 70 examples. Most recently we have succeeded in conducting the first of our virtual library screenings for all possible combinations of six substrates with 111 ligands (unpublished). In addition to providing accurate predictions of e.e., we have found that another useful outcome of this approach is to provide a tool for probing the fine details of reaction mechanisms. The small number of outliers that we have seen in some of our computational data have suggested subtle changes of mechanism for some catalyst/substrate combinations. These cases could then be probed in more detail with high-level electronic calculations to elucidate the mechanistic variations.

With the success of our initial enamide studies, we have begun to extend this effort to α,β-unsaturated amides, esters, and acids. Although enantioselective hydrogenation of unsaturated amides have many potential applications in synthesis, there have been relatively few reports of their use compared to other substrates. We have now performed the first computational study of the hydrogenation of this class of substrates at the B3LYP LACVP** level of theory.  The overall geometries of the transition structures and intermediates involved in the reaction pathway closely resemble the ones determined previously for the enamide substrates. The main differences in the structures and energetics of the reaction can be traced to the smaller ring size of a chelate, which leads to a weaker coordination of the olefin to the metal, and the absence of a second acceptor that was provided in the enamide/ester substrates. Together, these differences lead to higher relative free energies of the species involved in the pathway. These results have important consequences for the overall reaction because an isomerization between competing pathways now becomes energetically competitive. This possibility has not been described for the enamide reaction, presumably because it involves a dissociation of the olefin from the metal, which would render it energetically inaccessible in that case. As a result, the overall mechanism of the acrylamide reaction is found to proceed through formation of the octahedral metal dihydride complex through an approach of molecular hydrogen parallel to the P-Rh-C, followed by an isomerization to dihydride complexes in which the hydrogens are aligned along the P-Rh-O axis. Irreversible hydride transfer then completes the reaction. With this level of understanding of the reaction mechanism, we can next move on to the same type of Q2MM study and e.e. prediction that was conducted with the enamide system.