Performance of DFT methods and origin of stereoselectivity in bipyridine N,N′-dioxide catalyzed allylation and propargylation reactions

Abstract

Enantioselectivities for the allylation and propargylation of benzaldehyde catalyzed by bipyridine N,N'-dioxides were predicted using popular DFT methods. The results reveal deficiencies of several DFT methods while also providing a new explanation for the stereoselectivity of these reactions. In particular, even though many DFT methods provide accurate predictions of experimental ee's for these reactions, these predictions sometimes stem from qualitatively incorrect transition states. Overall, B97-D/TZV(2d,2p) provides the best compromise between accurate predictions of low-lying transition states and stereoselectivities for these reactions. The origin of stereoselectivity in these reactions was also examined, and arises from electrostatic interactions within the chiral electrostatic environment of a hexacoordinate silicon intermediate; the previously published transition state model for these reactions is flawed. Ultimately, these results suggest two strategies for the design of highly stereoselective catalysts for the propargylation of aromatic aldehydes, and pave the way for the computational design of novel catalysts for these reactions.

Steps toward Rationalization of the Enantiomeric Excess of the Sakurai–Hosomi–Denmark Allylation Catalyzed by Biisoquinoline N,N’-Dioxides Using Computations

“…These results suggested that the stereoselectivity of the corresponding reactions can be modulated by changing the secondary interactions between the ligands. In a subsequent paper from the same group [14], Sepùlveda et al expanded their previous computation to assess the performance of several DFT functionals and basis set combinations for prediction of the enantiomeric excess (ee) using the C 2 -symmetric catalyst 5 (Figure 1). These results confirmed that the secondary interactions between the ligands are responsible for the relative ordering of the TSs but also showed that the DFT results are in general not easily transferrable.…”

“…All the reported structures have been characterized as transition states with one negative frequency, and Gibbs free energies have been obtained using the harmonic approximation. Since our initial results showed differences between some of the structure of less than 2 kcal mol −1 , we performed a benchmark of the electronic energy results with 33 additional exchange-correlation functionals, including the one identified by Wheeler et al as the best performer for allylation and propargylation reactions (B97-D/TZV(2p,2d); see the Supporting Information for details) [13][14][15][16]. Relevant results are presented in the next section, and detailed ones for each structure are also available in the Supporting Information.…”

“…The stereochemical aspects of these reactions have been initially rationalized using conventional stereoelectronic arguments for the transition states (TSs) of the transient hypervalent silicon intermediates [12]. Computational work from Wheeler, [13][14][15][16] has elucidated that the interaction between the ligands in these TSs play a central role for the enantioselectivity of a reaction. Specifically, a C2-symmetric bidentate Lewis base, RSiCl3, and an electrophile Significant experimental contributions in this field include the work of Denmark, Kočovský, Malkov, Nakajima, and more recently from one of us [4][5][6][7][8][9][10][11].…”

Reaction-based machine learning representations for predicting the enantioselectivity of organocatalysts

“…values were not predicted from Gibbs free energy barriers, but rather from relative energy barriers (i.e., electronic energies plus solvation free energies), since they have been found to be more reliable than those based on either relative enthalpies or free energy barriers for this reaction. 14 The symbol E a was therefore used to indicate the energy (electronic plus solvation) difference between the TS and the preceding intermediate. For each C 2symmetric catalyst (Scheme 1), 10 distinct ligand arrangements around a hexacoordinate Si centre are possible (BP1-5, (R)-and (S)-, Fig.…”

“…9 To this end, Wheeler and co-workers have investigated 76 Lewis base organocatalysts (Scheme 1) 74 and used the computational toolkit AARON 20 to build a database of 760 stereocontrolling transition states to predict their enantioselectivity in the propargylation of benzaldehyde (Scheme 2). 14,74,86 Large databases of kinetic data for asymmetric catalysis generated in silico are scarce. 63 Therefore, this library constitutes an ideal training and validation set for the development of an atomistic ML model with reaction-based representations capable of predicting the e.e.…”