Understanding of nucleation process based on large-scale molecular dynamics simulations and theoretical improvements

Abstract

The nucleation process during the initial stage of a phase change is a fundamental process related to various fields. However, the theoretically predicted nucleation rate is known to be highly indeterminate. This study performed direct, large-scale molecular dynamics (MD) simulations of homogeneous nucleation processes including vapor-to-liquid nucleation and liquid-to-vapor nucleation. MD simulations are used to investigate the details of the nucleation process at the molecular level, and provide the nucleation rate, critical nuclei, Gibbs free energy for nucleation, sticking probabilities, and other information for verifying the theory from various aspects. Using systems with a large number of particles, such as nucleation from the gas phase using 8 billion Lennard-Jones or 4 million water molecules and bubble nucleation from the liquid phase with 500 million Lennard-Jones molecules, low nucleation rates were produced under the same conditions as that of experimental data. Moreover, the nucleation models exhibited high accuracy.