Scholarlink

<p>The development of deep potential (DP) energy has enabled high-accuracy modeling of material energy potential functions, leading to precise characterization of material properties. Recent discoveries on transition metal magnesium (Mg) highlight the need for in-depth molecular dynamics (MD) studies of its structural and physical characteristics. However, the practical use of MD is often hindered by the accuracy-efficiency trade-off. This study proposes using DP methods to address the lack of suitable potential functions for Mg, successfully generating accurate interatomic potentials for classical MD simulations. We validate the constructed model by comparing key material properties predicted by DP with those obtained from density functional theory (DFT). Through deep potential molecular dynamics simulations, we reveal the structural evolution of Mg at different temperatures, demonstrating a phase transition from α-hexagonal close-packed to β-body-centered cubic at 912&#xa0;K, followed by melting at 948&#xa0;K. A comparison of phonon dispersion and lattice thermal conductivity shows that the DP potential matches first-principles calculations in computational accuracy. The findings not only expand the application of deep learning in material simulation but also provide a powerful tool for accelerating the discovery and design of future materials.</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

Machine Learning-Driven Molecular Dynamics Investigation of the Physical and Heat Transfer Properties of Magnesium

  • 期刊论文

  • 发表日期：

    2025年6月23日

查看全文
  • Hanqing Li,
  • Jinlong Chen,
  • Cheng Cheng,
  • Li Ma,
  • Ju Rong,
  • Yudong Sui,
  • Zhijie He,
  • Jueyi Ye,
  • Xiangjie Fu,
  • Xiaohua Yu,
  • Jing Feng

摘要

The development of deep potential (DP) energy has enabled high-accuracy modeling of material energy potential functions, leading to precise characterization of material properties. Recent discoveries on transition metal magnesium (Mg) highlight the need for in-depth molecular dynamics (MD) studies of its structural and physical characteristics. However, the practical use of MD is often hindered by the accuracy-efficiency trade-off. This study proposes using DP methods to address the lack of suitable potential functions for Mg, successfully generating accurate interatomic potentials for classical MD simulations. We validate the constructed model by comparing key material properties predicted by DP with those obtained from density functional theory (DFT). Through deep potential molecular dynamics simulations, we reveal the structural evolution of Mg at different temperatures, demonstrating a phase transition from α-hexagonal close-packed to β-body-centered cubic at 912 K, followed by melting at 948 K. A comparison of phonon dispersion and lattice thermal conductivity shows that the DP potential matches first-principles calculations in computational accuracy. The findings not only expand the application of deep learning in material simulation but also provide a powerful tool for accelerating the discovery and design of future materials.