Capturing the Role of Temperature and Entropy in Crystal Polymorphs Through Molecular Simulations

The presence of multiple stable crystal structures in solid organic materials can limit their commercial viability when two or more observable polymorphs exhibit markedly different physical properties. Unintended restructuring events have hindered pharmaceutical solid form development in numerous therapeutic candidates and have led to costly market recalls. Conversely, intentional synthesis of a metastable form has led to significantly more favorable performance in numerous materials.

Current computational methods for predicting polymorphic behavior evaluate candidate crystal structures based on the minimized lattice energy. However, these static lattice energy-based approaches generate far more lattice energy minima than there are experimentally observed structures. Thermal motion of the crystals under working conditions has the potential to explain why many of these lattice minima are not observed experimentally. Lattice minima that are identified from a static crystal structure prediction can ultimately be unstable at experimental conditions through either temperature-mediated stability reranking, kinetic interconversion of multiple minima, or inaccuracies of the energy function in producing the real crystal ensemble.

In this work, we explore the role of thermal motion in eliminating candidate crystal structures using fully atomistic molecular dynamics simulations. Enthalpically favorable structures with low entropy will become unfavorable at high temperatures through temperature-mediated stability reranking. Our simulations correctly identify the high temperature solid form as having a larger entropy than the low temperature form in twelve small molecule organic systems with known temperature-mediated transformations. The estimated entropy differences in the classical point-charge potential are significantly closer to experimental measurements than estimated enthalpy differences. This result suggests that entropy difference estimates are less sensitive to the complexity of the simulation potential than the corresponding enthalpy estimates. We additionally find that a cheaper harmonic approximation provides a sufficient estimate of entropic contributions in small rigid molecules. However, entropies with the harmonic approximation diverge from molecular dynamics-derived entropies in systems with multiple rotatable degrees of freedom or dynamically disordered crystalline structures.

We additionally probe the sensitivity of the stability estimates to the energy function by directly computing the added effects of a more accurate polarizable Hamiltonian on polymorph free energies using a novel Hamiltonian reweighting approach. We show that the change in free energy to the more complex potential comes predominately from enthalpic rather than entropic contributions in the system examined.

Finally, we demonstrate the utility of molecular dynamics in identifying lattice minima interconversion and order-disorder transitions in organic solids. A rapid conversion of multiple lattice minima into a single ambient temperature ensemble is presented in six of the systems examined. These kinetics events can significantly reduce the number of plausible candidate structures in future crystal structure prediction studies.