Evaluation of various DFT protocols for computing¹H and ¹³C chemical shifts to distinguish stereoisomers: Diastereomeric 2-, 3-, and 4-methylcyclohexanols as a test set

¹H and ¹³C NMR chemical shifts were measured for a set of six isomers - the cis and trans 2-, 3-, and 4-methylcyclohexanols. ¹H and ¹³C NMR chemical shifts were computed at the B3LYP, WP04, WC04, and PBE1 density functional levels for the same compounds, taking into account the Boltzmann distribution among conformational isomers (chair-chair forms and hydroxyl rotamers). The experimental versus computed chemical shift values for proton and carbon were compared and evaluated (using linear correlation (r²), total absolute error (|Δδ| _T), and mean unsigned error (MUE) criteria) with respect to the relative ability of each method to distinguish between cis and trans stereoisomers for each of the three constitutional isomers. For C shift data, results from the B3LYP and PBE1 density functionals were not sufficiently accurate to distinguish all three pairs of stereoisomers, while results using the WC04 functional did do so. For ¹H shift data, each of the WP04, B3LYP, and PBE1 methods was sufficiently accurate to make the proper stereochemical distinction for each of the three pairs. Applying a linear correction to the computed data improved both the absolute accuracy and the degree of discrimination for most of the methods. The nature of the cavity definition used for continuum solvation had little effect. Overall, use of proton chemical shift data was more discriminating than use of carbon data.