Scholarlink

<p>TaC is considered as a promising material for extreme environments due to its excellent high-temperature performance. Molecular dynamics simulations are essential for elucidating atomic-scale mechanisms. However, no existing interatomic potential for TaC achieves both satisfactory accuracy and computational efficiency across a wide range of applications. In this study, we develop a deep learning interatomic potential for TaC based on a neural network framework to overcome these limitations. Compared to classical semi-empirical potentials and other deep learning-based models, the present potential demonstrates superior representability and transferability across a comprehensive set of properties, such as bulk properties, surface energies, grain boundary energies, generalized stacking fault energies, point defect formation energies, and melting point. All of these results are in good agreement with density functional theory calculations and experimental data. Extensive benchmark tests further confirm that the potential achieves a favorable balance between accuracy and computational efficiency. In addition, the stress–strain curves obtained at different temperatures reveal that increasing temperature leads to a reduction in yield strength and elastic modulus, while the corresponding yield strain remains nearly constant, except at 3000&#xa0;K. Fracture dynamics and dislocation mobility analyses are performed to provide a comprehensive understanding of the tensile fracture mechanism, together with the influence of irradiation-induced vacancy clusters on dislocation mobility. The present potential is well suited for molecular dynamics simulations of TaC across a wide range of applications, particularly for studying deformation behavior in extreme environments.</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

An accurate deep learning interatomic potential for TaC and its application in tensile deformation and dislocation mobility

  • 期刊论文

  • 发表日期：

    2026年2月26日

查看全文
  • Shunming Yang,
  • Yi Kong,
  • Yuanyuan Zhu,
  • Yongqi Lv,
  • Jinglian Fan

摘要

TaC is considered as a promising material for extreme environments due to its excellent high-temperature performance. Molecular dynamics simulations are essential for elucidating atomic-scale mechanisms. However, no existing interatomic potential for TaC achieves both satisfactory accuracy and computational efficiency across a wide range of applications. In this study, we develop a deep learning interatomic potential for TaC based on a neural network framework to overcome these limitations. Compared to classical semi-empirical potentials and other deep learning-based models, the present potential demonstrates superior representability and transferability across a comprehensive set of properties, such as bulk properties, surface energies, grain boundary energies, generalized stacking fault energies, point defect formation energies, and melting point. All of these results are in good agreement with density functional theory calculations and experimental data. Extensive benchmark tests further confirm that the potential achieves a favorable balance between accuracy and computational efficiency. In addition, the stress–strain curves obtained at different temperatures reveal that increasing temperature leads to a reduction in yield strength and elastic modulus, while the corresponding yield strain remains nearly constant, except at 3000 K. Fracture dynamics and dislocation mobility analyses are performed to provide a comprehensive understanding of the tensile fracture mechanism, together with the influence of irradiation-induced vacancy clusters on dislocation mobility. The present potential is well suited for molecular dynamics simulations of TaC across a wide range of applications, particularly for studying deformation behavior in extreme environments.