Co-crystallization is a promising approach to improving both the solubility and the dissolution rate of active pharmaceutical ingredients. Crystal structure determination from powder diffraction data plays an important role in determining co-crystal structures, especially those generated by mechanochemical means. Here, two new structures of pharmaceutical interest are reported: a 1:1 co‑crystal of furosemide with urea formed by liquid-assisted grinding and a second polymorphic form of a 1:1 co‑crystal of carbamazepine with indomethacin, formed by solvent evaporation. Energy minimization using dispersion-corrected density functional theory was used in finalizing both structures. In the case of carbamazepine:indomethacin, this energy minimization step was essential in obtaining a satisfactory final Rietveld refinement.

A combined local and global optimisation approach to crystal structure determination from powder diffraction data (SDPD) is presented. Using graphics processing units (GPUs) to accelerate the underpinning calculations, the speed and power of this approach is demonstrated with the solutions of two challenging crystal structures. In both cases, the frequency with which solutions were obtained was improved by an order of magnitude relative to DASH, a well-established SDPD program. With complex crystal structures increasingly being generated in polycrystalline form, this approach is a valuable step-forward in structure determination capabilities.

An effective and efficient method for dealing with ring systems during global optimisation-based crystal structure determination from powder diffraction data.