Quantum mechanical investigations of organocatalysis: mechanisms, reactivities, and selectivities.

Organocatalysis has captured the imagination of a significant group of synthetic chemists. Much of the mechanistic understanding of these reactions has come from computational investigations or studies involving both experimental and complementary computational explorations. As much as any other area of chemistry, organocatalysis has advanced because of both empirical discoveries and computational insights. Quantum mechanical calculations, particularly with density functional theory (DFT), can now be applied to real chemical systems that are studied by experimentalists; this review describes the quantum mechanical studies of organocatalysis. The dramatic growth of computational investigations on organocatalysis in the last decade reflects the great attention focused on this area of chemistry since the discoveries of List, Lerner, and Barbas of the proline-catalyzed intermolecular aldol reaction, and by MacMillan in the area of catalysis by chiral amino-acid derived amines. The number of reports on the successful applications of organocatalysts and related mechanistic investigations for understanding the origins of catalysis and selectivities keep growing at a breathtaking pace. Literature coverage in this review is until October 2009, except for very recent discoveries that alter significantly the conclusions based on older literature. 1.1 Computational methods for organocatalysis Over the last two decades, DFT has become a method of choice for the cost-effective treatment of large chemical systems with high accuracy.1 Most of the studies reported in this review were carried out using the B3LYP functional with the 6-31G(d) basis set, which is a standard in quantum mechanical calculations. Nevertheless, DFT is experiencing continuing developments of new functionals and further improvements. The availability of many new functionals and, in particular, the rapidly evolving performance issues of B3LYP have stimulated extra efforts on benchmarking DFT methods for the prediction of key classes of organic reactions.2 The well-documented deficiencies of B3LYP include the failure to adequately describe medium-range correlation and photobranching effects,3,4 delocalization errors causing significant deviations in π→σ transformations,2b,5 and incorrect description of non-bonding and long-range interactions,6 which are likely to be key factors in determining stereoselectivities. Benchmark results also show that newer functionals considerably improve some of the underlying issues.2–7 Recent advances, especially in the treatment of dispersion effects, now offer more reliable models of the reaction profiles and stereoselectivities. Most benchmarks focus on energetics rather than stereoselectivities. Systematic benchmarking for stereoselectivities requires more sophisticated techniques and averaging over conformations. To date, such benchmarking based upon stereoselectivity is available for only three reactions,8 and even there only various basis sets with B3LYP, as well as comparisons of results predicted using enthalpies and free energies. It is not possible to assign error bars for stereoselectivities for the majority of reports discussed in this review. Because stereoisomeric transition structures are very similar species, their relative energies are likely to be calculated accurately, as shown by the good agreement between calculated and experimental values. More recently Harvey (Harvey, 2010, faraday discussions) has studied two typical organic reactions of polar species (Wittig and Morita-Baylis-Hillman reactions) at different levels of theory.2i He showed that many standard computational methods, involving B3LYP, are qualitatively useful, but the energetics may be misleading for larger reactive partners; the quantitative prediction of rate constants remains difficult. These studies suggest that although B3LYP provides valuable qualitative insight into the reaction mechanisms and selectivities, the energetics may require testing with higher accuracy methods for complex organic systems. On the other hand, Simon and Goodman found B3LYP to be “only slightly less accurate” than newer methods, and recommended its use for organic reaction mechanisms.9