Steady-state molecular dynamics simulation of vapour to liquid nucleation with McDonald's daemon

TL;DR

This paper introduces a steady-state molecular dynamics simulation method using McDonald's daemon to study vapour to liquid nucleation, overcoming limitations of traditional methods by sampling steady-state properties and kinetics.

Contribution

It presents a novel application of grand canonical MD with McDonald's daemon for steady-state nucleation simulation, providing more accurate nucleation rates and free energy estimates.

Findings

Classical nucleation theory overestimates free energy of cluster formation.

Nucleation rates deviate significantly from classical predictions at high supersaturations.

Method successfully samples steady-state nucleation properties in Lennard-Jones fluids.

Abstract

The most interesting step of condensation is the cluster formation up to the critical size. In a closed system, this is an instationary process, as the vapour is depleted by the emerging liquid phase. This imposes a limitation on direct molecular dynamics (MD) simulation of nucleation by affecting the properties of the vapour to a significant extent so that the nucleation rate varies over simulation time. Grand canonical MD with McDonald's daemon is discussed in the present contribution and applied for sampling both nucleation kinetics and steady-state properties of a supersaturated vapour. The idea behind that approach is to simulate the production of clusters up to a given size for a specified supersaturation. In that way, nucleation is studied by a steady-state simulation. A series of simulations is conducted for the truncated and shifted Lennard-Jones fluid which accurately describes the fluid phase coexistence of noble gases and methane. The classical nucleation theory is found to overestimate the free energy of cluster formation and to deviate by two orders of magnitude from the nucleation rate below the triple point at high supersaturations.