Rare Event Sampling using Smooth Basin Classification

TL;DR

This paper introduces a universal, data-efficient method called smooth basin classification that uses graph neural networks to define reaction coordinates, improving the simulation of materials and molecules with large energy barriers.

Contribution

The method leverages internal feature representations and transfer learning to outperform traditional reaction coordinates based on human intuition.

Findings

Outperforms human-designed reaction coordinates on benchmarks

Utilizes symmetry and transfer learning for high data efficiency

Effective on challenging chemical and physical transformations

Abstract

The efficiency of atomic simulations of materials and molecules can rapidly deteriorate when large free energy barriers exist between local minima. We propose smooth basin classification, a universal method to define reaction coordinates based on the internal feature representation of a graph neural network. We achieve high data efficiency by exploiting their built-in symmetry and adopting a transfer learning strategy. We benchmark our approach on challenging chemical and physical transformations, and show that it matches and even outperforms reaction coordinates defined based on human intuition.