Early-stage nucleation in metallic materials: A computational study using Monte Carlo and rare event sampling methods

Abstract

Nucleation is the initial stage of a first-order phase transition, such as vapor-liquid condensation, crystallization and precipitation, when the system evolves from the thermodynamic metastable state into a stable state. In metallic systems, the nucleation process determines the kinetics of new phase formation, and strongly influences the evolution of intermediate products during phase transition. The nucleation mechanism has drawn the academic attention for years, but there are still many questions remaining unsolved. The well-known Classical Nucleation Theory (CNT) captures the essence of nucleation, but the macroscopic description of the very small critical nucleus also calls into question in real applications. In this work, the atomistic Monte Carlo simulation technique combined with enhanced sampling methods is used to investigate the nucleation mechanism in binary metallic systems. Profound insights into microscopic aspects of the nucleation mechanism are provided and compared with the concepts from CNT, such as the cluster description, interfacial energy, nucleation energy barrier and driving force. The enhanced sampling methods provide a plausible way to bridge the gap between continuum modelling and atomic simulations, which enables the validation of classical theories or other non-classical continuum modelling by computational experiments in atomic dimension. Additionally, a novel and efficient way, the Reweighted Partial Path (RPP) method, is developed and proposed to compute free energy profiles for diffusive processes in single Transition Interface Sampling (TIS) or Forward Flux Sampling (FFS) simulations. The method employs a partial path reweighting strategy, based on the memory loss assumption for diffusive systems, to derive the equilibrium distribution of states along a chosen order parameter from TIS or FFS trajectories. No additional calculations, such as, reverse TIS or Umbrella Sampling are required.