Size and Temperature Transferability of Direct and Local Deep Neural Networks for Atomic Forces

TL;DR

This paper introduces a deep learning model for atomic forces that demonstrates comparable accuracy to state-of-the-art models, with strong transferability across system sizes and temperatures, validated on aluminum, sodium, and silicon.

Contribution

The study presents a direct, local deep learning model for atomic forces that effectively scales and transfers across different system sizes and temperatures, providing insights into physical attributes.

Findings

Model errors are comparable to state-of-the-art methods.

The model's performance scales well with system size.

It maintains accuracy across temperature variations.

Abstract

A direct and local deep learning (DL) model for atomic forces is presented. We demonstrate the model performance in bulk aluminum, sodium, and silicon; and show that its errors are comparable to those found in state-of-the-art machine learning and DL models. We then analyze the model's performance as a function of the number of neighbors included and show that one can ascertain physical attributes of the system from the analysis of the deep learning model's behavior. Finally, we test the size scaling performance of the model, and the transferability between different temperatures, and show that our model performs well in both scaling to larger systems and high-to-low temperature predictability.