Biasing crystallization in fused silica: An assessment of optimal metadynamics parameters

Abstract

Metadynamics (MetaD) is a useful technique to study rare events such as crystallization. It has been only recently applied to study nucleation and crystallization in glass-forming liquids such as silicates, but the optimal set of parameters to drive crystallization and obtain converged free energy surfaces is still unexplored. In this work, we systematically investigated the effects of the simulation conditions to efficiently study the thermodynamics and mechanism of crystallization in highly viscous systems. As a prototype system, we used fused silica, which easily crystallizes to β-cristobalite through MetaD simulations, owing to its simple microstructure. We investigated the influence of the height, width, and bias factor used to define the biasing Gaussian potential, as well as the effects of the temperature and system size on the results. Among these parameters, the bias factor and temperature seem to be most effective in sampling the free energy landscape of melt to crystal transition and reaching convergence more quickly. We also demonstrate that the temperature rescaling from T > Tm is a reliable approach to recover free energy surfaces below Tm, provided that the temperature gap is below 600 K and the configurational space has been properly sampled. Finally, albeit a complete crystallization is hard to achieve with large simulation boxes, these can be reliably and effectively exploited to study the first stages of nucleation.