A supercooled study of nucleation and symmetries

Nucleation is the process by which a metastable phase decays into a stable phase. It is widely observed in nature, and is responsible for many phenomena such as the formation clouds and domains in crystalline solids. The classical theory of nucleation predicts that the objects that initiate the decay from the metastable to the stable phase are compact droplets whose interior has the structure of the stable phase. For quenches deep into the metastable phase, however, the droplets may be ramified, with a structure very different from the stable phase. This difference has profound implications for material properties, especially because predicting the onset of structure early enough is useful for manipulating and controlling nucleation processes. I used molecular dynamics to simulate nucleation in Lennard-Jonesium, a model system for liquid-solid transformations. The system is quenched from a high temperature, where the liquid is stable, to a temperature where the liquid is metastable, and is allowed to nucleate via fluctuation-driven clusters referred to as critical droplets. I determined the occurrence of critical droplets by the intervention method, but found a non-monotonic variation in droplet survival rates near the saddle point. I determined the structure of the critical droplet and found evidence for a core consisting of mostly solid-like particles with hcp symmetry and a previously unknown planar structure around it. Using perturbative techniques, I showed that the planar particles have a significant influence on the nucleation and growth of critical droplets. I also introduced a novel method of learning symmetries to predict the structure and appearance of precursors to the critical nucleus. My results give added evidence for the presence of spinodal nucleation at deep quenches