Efficient Parameterization Of A Surrogate Model Of Molecular Interactions In Crystals

We propose a surrogate model for lattice energy that allows the accurate prediction of the crystal structures formed by a given molecule and their relative stability ranking. The model is derived from a combination of isolated-molecule quantum mechanical calculations and a relatively small number of more expensive solid-state DFT-D computations. The surrogate model provides an effective mechanism for refining the crystal structure landscape predicted by current Crystal Structure Prediction methodologies. Applied to the agrochemical Chlorothalonil, the approach is shown to be highly accurate whilst reducing the computational costs by approximately a factor of 20 compared to refinement of all structures using solid-state DFT.