How accurate can crystal structure prediction be for energetic molecular crystals?*

Predicting the packing and conformation of the molecules in molecular crystals (MC) generally requires the accuracy of ab initio methods [1]. Using such methods, we assess the capabilities of an evolutionary algorithm (EA) [2] for the crystal structure prediction (CSP) of well-known but challenging energetic MC. While providing the EA with the experimental conformation of the molecules quickly re-discovers the experimental packing, it is more realistic to start from a naive or neutral conformation, which reflects the limited experimental knowledge we have in computational design of MC. Doing so, and using fully flexible molecules in fully variable cells, we show that the experimental structures of β-HMX, α-RDX, ε-CL-20 and α-FOX-7 can be perfectly predicted in less than 20 generations. Nonetheless, one must be aware that some MC have naturally hindered evolutions, requiring as many attempts as space groups of interest to predict their structures, and some may require the accuracy of all-electron calculations to discriminate closely ranked polymorphs. To save computer resources, we show that a hybrid xTB/DFT-D approach could be considered to push the limits of CSP beyond 200+ atoms and for co-crystals.

[1] A.M. Reilly et al., “Report on the sixth blind test of organic crystal structure prediction methods”, Acta Crystallogr. Sect. B Struct. Sci. Cryst. Eng. Mater., 72, 4, 439-459, 2016.

[2] C.W. Glass et al., “USPEX - Evolutionary crystal structure prediction”, Comput. Phys. Commun., 175, 11, 713-720, 2006.