(272h) Anatomy of a Complex Crystallization Pathway

Crystallization, the process of molecules or particles assembling into an ordered structure from the disordered phase, is integral to a wide range of applications, such as material fabrication and drug production. At the core of this process is the nucleation step, during which, according to the Classical Nucleation Theory, the basic units of the crystals assemble into small compact clusters that are usually spherical. Upon reaching the critical size, the cluster survives thermal fluctuations and continues to grow.

However, this classical depiction of nucleation has been called into question repeatedly. In polymorphic systems, where the basic units can assemble into multiple structures, the competition between polymorphs complicates nucleation and crystallization even more.

Another practical issue with polymorphism is that different structures often lead to different physicochemical properties, e.g., dissolution rate and solubility. Therefore, it is crucial for us to gain a fundamental understanding of the nucleation process in polymorphic systems such that we can engineer the pathways towards the target crystal with desired properties.

Due to the small time and length scales on which nucleation happens, computational tools are well-suited for such studies. In this work, we utilize molecular simulations to model a system of point particles that interact via an oscillatory pair potential. We find that the system can assemble into four competing polymorphs. Two of them are simple crystals: BCC and FCC, and the other two are highly complex: -Brass and -Manganese. -Brass has 52 particles in its unit cell, divided among four Wyckoff sites, and -Manganese has 20 particles in the unit cell with two Wyckoff sites.

Using local structure metrics, we demonstrate how we can use simple machine learning tools strategically to dissect the complex pathways down to particle level, with which we identified four types of structural transitions, and we show how local environments of the particles direct the pathways towards different polymorphs.

To explore the thermodynamic origin of the competition, we computed the Gibbs free energies of the four crystals. We find that the FCC crystal has the lowest free energy. Yet it is the most rarely seen polymorph in our self-assembly simulations, which indicates that kinetic factors play an important role in the assembly pathways. We also calculated the free energy surface, which, in combination with our local environment analysis, elucidates the mechanisms of the four types of transitions.

With our results, we hope to contribute to the theoretical framework of nucleation in polymorphic systems that involve complex crystal structures and provide insights on how to engineer the pathways towards desired polymorphs with high quality and yield.