The potential of dispersion-corrected density functional theory calculations for distinguishing between salts and cocrystals

Validation of a method for distinguishing between salts and cocrystals based on dispersion-corrected density functional theory is presented. Existing related works (van de Streek & Neumann, 2010; LeBlanc et al., 2018) indicate that this approach is problematic and leads to incorrect results in multiple situations. The method suggested here is based on the geometry optimization of an artificially constructed wrong structure (hydrogen atom placed in salt position near the potential acceptor for cocrystals and vice versa cocrystal position with hydrogen atom placed near the potential donor of the salts). The verification of the method is based on comparison of the results with an experimentally confirmed correct hydrogen position. Calculations were performed on 173 selected structures of salts and 96 cocrystals with ΔpKa in the critical 〈−1, 4〉 range. The range was chosen to test the method on the most problematic structures. When the artificial wrong model did not converge to the correct one (salt to cocrystal and vice versa), it was tested whether the correct model converged to the correct one in addition. The results confirmed that the most widely used functional (PBE) tends to generate false salt results. All salts converged to the salt from cocrystal initial models. Sixteen cocrystals showed local energy minima for both the salt and cocrystal states. Eighteen cocrystals always converged to salt. Rules were identified under which the results can be considered reliable: when a cocrystal starting model converges to cocrystal, the structure is certainly cocrystal. When both the cocrystal and salt models converge to salt for a long hydrogen-bond (longer than 2.613 Å) the structure is most likely salt. For short hydrogen bonds it is not possible to distinguished reliably between salt and cocrystal using the dispersion-corrected PBE functional. Additional calculations were performed with more advanced functionals for 18 problematic structures detected in the screening as well as for four more mentioned in the literature. The results show that the rSCAN functional (Bartók & Yates, 2019) improves the agreement with the experiment. Further improvement was observed by using hybrid functionals (PBE0, PBE50), which were tested on structures that gave incorrect results with rSCAN. The described method for distinguishing salts from cocrystals can be useful for enhancing the information given by structure solutions from powder, the verification of structure solutions from single crystals and studies related to crystal structure prediction.