Acceleration of crystal structure relaxation with Deep Reinforcement Learning

TL;DR

This paper presents a Deep Reinforcement Learning approach to accelerate crystal structure relaxation, outperforming classical methods and capable of generalizing from small to complex atomic systems.

Contribution

Introduces a novel DRL model for crystal structure relaxation, compares neural network architectures, and benchmarks against classical algorithms to demonstrate improved efficiency.

Findings

DRL models outperform classical optimization algorithms in structure relaxation.

Model generalizes learned interaction patterns from small to complex systems.

Parameter tuning significantly influences model performance.

Abstract

We introduce a Deep Reinforcement Learning (DRL) model for the structure relaxation of crystal materials and compare different types of neural network architectures and reinforcement learning algorithms for this purpose. Experiments are conducted on Al-Fe structures, with potential energy surfaces generated using EAM potentials. We examine the influence of parameter settings on model performance and benchmark the best-performing models against classical optimization algorithms. Additionally, the model's capacity to generalize learned interaction patterns from smaller atomic systems to more complex systems is assessed. The results demonstrate the potential of DRL models to enhance the efficiency of structure relaxation compared to classical optimizers.