29Si NMR Shielding Calculations Employing Density Functional Theory, Focussing on Hypervalent Silicon Compounds

Abstract 29Si NMR chemical shifts of 41 silicon‐containing compounds (14 tetra‐, 10 penta‐ and 17 hexacoordinate compounds) were calculated by employing DFT using the gauge including atomic orbital (GIAO) formalism and compared to experimental values. The hybrid B3LYP functional, the generalized gradient approximation functional (GGA) BPW91 and the meta‐GGA functional HCTH407 were used in conjunction with the 6‐31G(d,p) and 6‐311+G(2d) basis sets and a locally dense basis set [6‐311G+(2d) on Si and 6‐31G(d,p) on all other atoms]. The HCTH407/6‐311+G(2d) level of theory consistently performed best with a mean average error (MAE) of 5.9 ppm between calculated and experimental chemical shifts over the entire test set (4.4, 5.7 and 7.9 ppm for tetra‐, penta‐ and hexacoordinate compounds, respectively). The choice of a suitable basis set had a much larger influence on the accuracy of the results than the choice of functional. Additional calculations for some outliers with the 6‐311++G(2df,2pd) basis set yielded only small improvements, whereas the economically attractive, locally dense basis set (LDBS) yielded an MAE of 6.5 ppm with the HCTH407 functional over the entire test set after correction of systematic errors. For a subset of compounds, GIAO‐HF and MP2 ab initio calculations were performed yielding more accurate results than the DFT calculations (MAEs for HF, MP2 and HCTH407 were 3.9, 4.6 and 5.6 ppm, respectively) but required roughly two orders of magnitude more CPU time.