Comparison of various density functional methods for distinguishing stereoisomers based on computed (1)H or (13)C NMR chemical shifts using diastereomeric penam beta-lactams as a test set.

Full (1)H and (13)C NMR chemical shift assignments were made for two sets of penam beta-lactams: namely, the diastereomeric (2S, 5S, 6S)-, (2S, 5R, 6R)-, (2S, 5S, 6R)-, and (2S, 5R, 6S)-methyl 6-(1,3-dioxoisoindolin-2-yl)-3,3-dimethyl-7-oxo-4-thia-1-aza-bicyclo[3.2.0]heptane-2-carboxylates (1-4) and (2S, 5R, 6R)-, (2S, 5S, 6R)-, and (2S, 5R, 6S)-6-(1,3-dioxoisoindolin-2-yl)-3,3-dimethyl-7-oxo-4-thia-1-aza-bicyclo[3.2.0]heptane-2-carboxylic acids (6-8). Each penam was then modeled as a family of conformers obtained from Monte Carlo searches using the AMBER* force field followed by IEFPCM/B3LYP/6-31G(d) geometry optimization of each conformer using chloroform solvation. (1)H and (13)C chemical shifts for each conformer were computed at the WP04, WC04, B3LYP, and PBE1 density functional levels as Boltzmann averages of IEFPCM/B3LYP/6-311 + G(2d,p) energies over each family. Comparisons between experimental and theoretical chemical shift data were made using the total absolute error (|Deltadelta| (T)) criterion. For the (1)H shift data, all methods were sufficiently accurate to identify the proper stereoisomers. Computed (13)C shifts were not always successful in identifying the correct stereoisomer, regardless of which DFT method was used. The relative ability of each theoretical approach to discriminate among stereoisomers on the basis of proton shifts was also evaluated.