(544a) Enhanced Sampling Molecular Dynamics Simulations of Polymorphism in Molecular Crystals

Polymorphism affects the properties of molecular solids and is crucial in processing pharmaceuticals, pigments, agrochemicals, and energetic materials. Molecular dynamics (MD) simulations can provide important information about these materials' polymorphic transitions and crystallization processes by estimating the underlying free energy surface (FES). However, constructing the FES from MD simulations is difficult due to the long timescales (>1µs) associated with crystal phase transitions. Enhanced sampling methods, such as well-tempered Metadynamic (WT-MetaD), bias the system to increase the occurrence of phase transitions during an MD simulation. MetaD requires defining reaction coordinates (collective variables, CVs) that sample the slowest degree of freedom of the process. In conformational polymorphism, the constituent molecules crystallize in diverse forms, which differ by the molecular conformation and not just by the molecular packing or lattice parameters. Thus, finding good CVs to accelerate transitions between the different polymorphs is challenging. Data-driven approaches can be used to catalyze CV discovery.

Herein, we used MD simulations and dimensionality reduction algorithms to describe conformational polymorphic transitions of CL-20, a complex energetic material with three anhydrous conformational polymorphs (β, γ, and ϵ) at ambient pressure.¹ We demonstrate that a local molecular CV obtained from short gas phase simulations of isolated conformers can drive polymorphic phase transitions. We successfully converge the FES of the system and obtain the correct qualitative thermodynamic stability order for the three polymorphs. Moreover, we identified several defect and intermediate forms, which are local minima in FES. We next investigate polymorphism in a phosphorescent organic material by combing descriptors of molecular orientation and global molecular packing. The approach allows for identifying polymorphic transitions into new forms. These findings demonstrate the benefits of combining data-driven approaches with biased MD simulations to describe polymorphism in complex systems and find new strategies to obtain desired molecular polymorphs.

[1] Collective Variables for Conformational Polymorphism in Molecular Crystals, Oren Elishav, Roy Podgaetsky, Olga Meikler, and Barak Hirshberg, The Journal of Physical Chemistry Letters 2023 14 (4), 971-976, DOI: 10.1021/acs.jpclett.2c03491