Scholarlink

<p>In modern computational materials, machine learning has shown the capability to predict interatomic potentials, thereby supporting and accelerating conventional molecular dynamics (MD) simulations. However, existing models typically sacrifice either accuracy or efficiency. Moreover, efficient models are highly demanded for offering simulating systems on a considerably larger scale at reduced computational costs. Here, we introduce an efficient equivariant graph neural network (E<sup>2</sup>GNN) that can enable accurate and efficient interatomic potential and force predictions for molecules and crystals. Rather than relying on higher-order representations, E<sup>2</sup>GNN employs a scalar-vector dual representation to encode equivariant features. By learning geometric symmetry information, our model remains efficient while ensuring prediction accuracy and robustness through the equivariance. Our results show that E<sup>2</sup>GNN consistently outperforms the prediction performance of the representative baselines and achieves significant efficiency across diverse datasets, which include catalysts, molecules, and organic isomers. Furthermore, we conduct MD simulations using the E<sup>2</sup>GNN force field across solid, liquid, and gas systems. It is found that E<sup>2</sup>GNN can achieve the accuracy of ab initio MD across all examined systems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Efficient equivariant model for machine learning interatomic potentials

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年2月26日

查看全文
  • Ziduo Yang,
  • Xian Wang,
  • Yifan Li,
  • Qiujie Lv,
  • Calvin Yu-Chian Chen,
  • Lei Shen

摘要

In modern computational materials, machine learning has shown the capability to predict interatomic potentials, thereby supporting and accelerating conventional molecular dynamics (MD) simulations. However, existing models typically sacrifice either accuracy or efficiency. Moreover, efficient models are highly demanded for offering simulating systems on a considerably larger scale at reduced computational costs. Here, we introduce an efficient equivariant graph neural network (E2GNN) that can enable accurate and efficient interatomic potential and force predictions for molecules and crystals. Rather than relying on higher-order representations, E2GNN employs a scalar-vector dual representation to encode equivariant features. By learning geometric symmetry information, our model remains efficient while ensuring prediction accuracy and robustness through the equivariance. Our results show that E2GNN consistently outperforms the prediction performance of the representative baselines and achieves significant efficiency across diverse datasets, which include catalysts, molecules, and organic isomers. Furthermore, we conduct MD simulations using the E2GNN force field across solid, liquid, and gas systems. It is found that E2GNN can achieve the accuracy of ab initio MD across all examined systems.