Symmetry-Constrained Monte Carlo for the Crystal Structure Prediction of Small Organic Molecules

The prediction of crystal structures from the bare knowledge of molecular connectivity remains a formidable (and mostly unsolved) challenge. Traditional brute force algorithms clash with the complexity of the structural landscape of even the simplest chemicals, where lattice energies of different polymorphs often lie far apart just by a few kJ mol–1. In this proof-of-concept contribution, we test a novel crystal structure prediction approach, based on the symmetry-constrained Monte Carlo (SC-MC) method. SC-MC leverages full group symmetry to propagate random changes of the asymmetric unit across the whole crystal. Thus, it is effective in quickly exploring a vast region of the polymorph landscape. In all test cases, the algorithm retrieves the experimental structure within the top-ranked 4–6 guesses. However, owing to the inherent limitations of classical force field methods, the absolute energy turns out to be a poor descriptor to rank predictions onto an absolute scale. By submitting the most promising predictions to the scrutiny of Molecular Dynamics, we highlight those structures that, despite bearing the highest cohesive energy, could be prone to mechanical instabilities because of the shape of their crystal field potential around the minimum. MD-based reranking allows discrimination of the most promising predictions amidst the background noise of apparently similar but incorrect structures. Strengths and limitations of the SC-MC procedure are critically discussed.