(606f) Polymorphism in a Lennard-Jones Fluid Nucleating from Its Melt

Polymorph selection, which is predicting the outcome of a crystallization process in terms of the resulting crystal structure, has important applications in fine chemicals industries. For example, polymorphs of active pharmaceutical ingredients have different solubilities and dissolution rates resulting in varied bioavailability of drugs. Computational prediction of the kinetically favored polymorph can reduce experimental costs significantly and accelerate formulation development. In this work, we take an atomic-resolution look into the competition between HCP and FCC polymorphs in a Lennard-Jones (LJ) fluid nucleating from its melt using molecular dynamics (MD) simulations and 2-dimensional free energy calculation with hybrid Monte Carlo/Molecular Dynamics (HMC/MD) simulations. Structural composition of the nucleus is considered as well as nucleus size by calculating the free energy as a function of structure specific nucleus size coordinates. Unbiased MD trajectories reveal that polymorph selection happens after nucleation of an FCC-dominated cluster, also confirmed by the presence of a single free energy saddle point on the polymorph specific nucleus size space. The critical cluster transforms into either FCC or HCP phases in the growth stage. By calculating the structure specific diffusion coefficients through unbiased MD simulations, we construct the complete mesoscopic system for LJ nucleation from the melt and calculate nucleation rates. The choice of LJ fluid as a model system is advantageous in elucidating the strategy for formulating a mesoscopic model for calculating nucleation rates of multiple structures, which can be applied to the more computationally intensive problem of real systems with more complex molecules. Direct comparison of computational predictions with experimental measurements is challenging for polymorph specific nucleation rates since monitoring the structure of the critical nuclei is difficult and solvent mediated polymorphic transitions can cause errors in the measured rates. We propose a population balance model that captures the nucleation, growth, and polymorphic transformations of a population of clusters in LJ nucleation from the melt. This multi-scale, first-principles based model can potentially predict the polymorphic outcome of a crystallization process and allow a direct comparison to experiments.