We address the crystal structure prediction problem by combining the evolutionary algorithm USPEX (used to predict sets of low-energy crystal structures) and the synthon approach (extracting preferable supramolecular synthons from the Cambridge Structural Database, CSD, and then indicating the most stable structure as well as the most likely metastable polymorphs among the predicted low-energy structures with synthons). This approach is applied to three compounds: N-chlorophthalimide (ncp), N-bromophthalimide (nbp), and N-iodophthalimide (nip). Database search has indicated the importance of halogen bonding N–X⋯OC synthons (X = Cl, Br, I) as a touchstone. The predicted structures of ncp of rank 1 and nbp of rank 3 match the experimental ones. Our results imply that the synthon approach can successfully suggest the right structures amongst a handful of lowest-energy predicted structures. Two candidate crystal structures were suggested using the synthon approach for nip.

Keywords