Semi-empirical refinements of crystal structures using ¹⁷O quadrupolar-coupling tensors

We demonstrate a modification of Grimme's two-parameter empirical dispersion force field (referred to as the PW91-D2* method), in which the damping function has been optimized to yield geometries that result in predictions of the principal values of ¹⁷O quadrupolar-coupling tensors that are systematically in close agreement with experiment. The predictions of ¹⁷O quadrupolar-coupling tensors using PW91-D2*-refined structures yield a root-mean-square deviation (RMSD) (0.28 MHz) for twenty-two crystalline systems that is smaller than the RMSD for predictions based on X-ray diffraction structures (0.58 MHz) or on structures refined with PW91 (0.53 MHz). In addition, ¹³C, ¹⁵N, and ¹⁷O chemical-shift tensors and ³⁵Cl quadrupolar-coupling tensors determined with PW91-D2*-refined structures are compared to the experiment. Errors in the prediction of chemical-shift tensors and quadrupolar-coupling tensors are, in these cases, substantially lowered, as compared to predictions based on PW91-refined structures. With this PW91-D2*-based method, analysis of 42 ¹⁷O chemical-shift-tensor principal components gives a RMSD of only 18.3 ppm, whereas calculations on unrefined X-ray structures give a RMSD of 39.6 ppm and calculations of PW91-refined structures give an RMSD of 24.3 ppm. A similar analysis of ³⁵Cl quadrupolar-coupling tensor principal components gives a RMSD of 1.45 MHz for the unrefined X-ray structures, 1.62 MHz for PW91-refined structures, and 0.59 MHz for the PW91-D2*-refined structures.