Crystal-energy landscapes of active pharmaceutical ingredients using composite approaches

The crystal structure prediction (CSP) of organic molecular solids remains challenging, as the demand to predict more complex crystal structures increases. Low-cost (semi-)empirical methods are now more commonly being used to unburden the computational bottlenecks in intermediate stages of a CSP protocol, but remain inadequate for the energy ranking of crystal structures. With the use of composite approaches, however, these relative energies can be refinedwith higher levels of theory at the cost of a single-point energy calculation, provided that the low-level geometries are amenable to such purposes. Herein, a composite method making use of the B86bPBE-XDM density functional, and combining a low-level small-basis method using finite-support numerical orbitals with high-level plane-wave calculations, is applied to predict crystal-energy landscapes of four active pharmaceutical ingredients: 5-fluorouracil, naproxen, carbamazepine, and olanzapine. Results show that this composite method can aid in resolving realistic energy landscapes of drug-like molecules, consistently placing the experimentally isolable polymorphs as the lowest-energy structures and, in addition, providing a sound stability ordering of polymorphs, in reasonable agreement with available experimental data. This is in stark contrast to the energy rankings provided by previously reported refined anisotropic force-field data or results obtained from other small-basis set approaches, such as sHF-3c. While the B86bPBE-XDM composite method is generally more expensive than (semi-)empirical approaches, it does not rely on any parameter fitting or tuning with regards to a given system of interest, making it a suitable and more generally applicable alternative for CSP.