(436i) Polymorphic Transformations at Ambient Conditions Elucidated with Molecular Dynamics

A long sought-after goal in the field of computational chemistry is the ability to predict all observable crystal structures, or polymorphs, of small-molecule organic solids. Computer-based models can, in principle, rapidly input a particular molecule and search through possible crystal structures, identifying the set of stable or metastable polymorphs which are likely to be observed experimentally. Current approaches for crystal structure prediction typically rank candidate structures using lattice energies of individual configurations. Such approaches assume that entropy and lattice expansion effects caused by thermal motions at ambient conditions contribute negligibly to the relative stability of different crystal structures. However, many temperature and pressure driven solid-solid transformations are known to exist experimentally, and these cannot be predicted without including thermal motions.

In this work, we present the prediction of finite temperature solid-solid transformations using temperature and pressure dependent free energies derived from a novel molecular dynamics approach. Molecular dynamics simulations implicitly include the thermal motions present at finite temperatures and can therefore predict the emergence of a more stable crystal structure at high temperatures and pressures. We examine the five organic molecules benzene, formamide, glycine, acetic acid, and imidazole which are known experimentally to have different globally stable structures at low and high temperatures. The relative free energies for these solids in the point-charge OPLS potential correctly identify the high temperature / high pressure polymorphs as being more stable than the low temperature form.

We also compute the temperature and pressure dependent free energies in the more expensive polarizable AMOEBA potential to determine the added benefit of using a more accurate energy function. We use a novel Boltzmann reweighting approach to circumvent the additional expense of the polarizable potential which has kept the more accurate model from being used in previous MD studies. The polarizable AMOEBA potential is shown to have a higher accuracy in predicting the low temperature crystal structure in these systems. However, both OPLS and AMOEBA have equal fidelity at predicting the high temperature and high pressure structure. These preliminary results suggest that high level energy functions are more accurate at predicting low temperature structures, but have a less significant advantage at high temperatures relative to cheaper levels of theory. Finally, we show that these short ambient temperature MD simulations can allow metastable lattice minima to restructure into more stable minima.