A systematic benchmarking of <sup>31</sup>P and <sup>19</sup>F <scp>NMR</scp> chemical shift predictions using different <scp>DFT</scp>/<scp>GIAO</scp> methods and applying linear regression to improve the prediction accuracy
Peng Gao, Jie Zhang, Hongming Chen
Abstract
Abstract A systematic benchmark study of phosphorus and fluorine nuclear magnetic resonance chemical shift predictions using six different density functional theory (DFT)/the gauge‐including atomic orbital (GIAO) methods was conducted. Two databases were compiled: one consists of 35 phosphorus‐containing molecules, which cover the most common intramolecular bonding environments of trivalent and pentavalent phosphorus atoms; the other is composed of 46 fluorine‐containing molecules. The characteristics of each DFT/GIAO method with different solvent models were demonstrated in detail. The application of linear regression between the calculated isotropic shielding constants and experimental chemical shifts was applicable to improving the prediction accuracy. The best methods with the solvation model based on density (SMD) and conductor‐like polarizable continuum model (CPCM) implicit solvent models for 31 P chemical shifts predictions are able to yield a root‐mean‐square deviation of 5.58 and 5.42 ppm, respectively; for 19 F, the corresponding lowest prediction errors with these two applied solvent models are 4.43 and 4.12 ppm, respectively. The developed scaling factors fitted from linear regression are applicable to enhancing the chance of successful structural elucidations of phosphorus or fluorine‐containing compounds as an efficient complement to 13 C, 1 H, 11 B, and 15 N chemical shift predictions.