Caveats of mean first-passage time methods applied to the crystallization transition: effects of non-Markovianity

TL;DR

This paper demonstrates that mean first-passage time methods can underestimate reaction rates in crystallization due to non-Markovian dynamics caused by poor reaction coordinate choices.

Contribution

It identifies the impact of non-Markovianity on reaction rate estimation and highlights the importance of selecting appropriate reaction coordinates.

Findings

Mean first-passage time methods may underestimate reaction rates.

Non-Markovian dynamics are caused by poorly chosen reaction coordinates.

Recurrence times reveal non-Markovian behavior.

Abstract

Using the crystallization transition in a Lennard-Jones fluid as example, we show that mean first-passage time based methods may underestimate the reaction rates. We trace the reason of this deficiency back to the non-Markovian character of the dynamics caused by the projection to a poorly chosen reaction coordinate. The non-Markovianity of the dynamics becomes apparent in the behavior of the recurrence times.