Non-Markovian dynamics in ice nucleation

TL;DR

This study investigates whether the size of the crystalline nucleus in ice nucleation follows Markovian dynamics, finding evidence of history dependence and proposing improved reaction coordinates using neural networks and symbolic regression.

Contribution

It demonstrates non-Markovian behavior in ice nucleation and introduces a machine learning approach to identify better reaction coordinates.

Findings

Nucleus size exhibits non-Markovian dynamics with history dependence.

Structural descriptors reveal differences between early and late recurrences.

Neural network and symbolic regression improve reaction coordinate approximation.

Abstract

In simulation studies of crystallisation, the size of the largest crystalline nucleus is often used as a reaction coordinate to monitor the progress of the nucleation process. Here, we investigate, for the case of homogeneous ice nucleation, whether the nucleus size exhibits Markovian dynamics, as assumed in classical nucleation theory. Using 300 independent nucleation trajectories generated by molecular dynamics, we evaluate the mean recurrence time required to reach selected values of the largest nucleus size. Early recurrences consistently take longer than later ones, revealing a clear history dependence and thus non-Markovian dynamics. To identify the slow modes underlying this behaviour, we analyse several structural descriptors of the nucleus, observing subtle but systematic differences between nuclei at early and late recurrences. By training a neural network on 2,700 short trajectories to learn the committor, we identify relevant collective variables. Based on these features, symbolic regression provides a compact approximation of the committor, i.e., an improved reaction coordinate, which we subsequently test for Markovianity.