Scholarlink

<p>The performance of orthopedic implants depends on biocompatibility, mechanical compatibility, and long-term reliability. Magnesium–praseodymium (Mg-Pr) alloys, with biodegradability and bone-like mechanical response, are promising candidates. Their low alloying content and pronounced local chemical complexity, however, challenge conventional empirical potentials, limiting atomistic simulation accuracy. Here, a deep potential (DP) model within an “AI for Science” framework is developed to investigate crystallography, mechanics, energetics, and corrosion kinetics of Mg-Pr alloys. The DP model outperforms empirical and mainstream machine-learning potentials, achieving root mean square errors of 3.61 × 10<sup>−3</sup>&#xa0;eV/atom for energies and 1.84 × 10<sup>−3</sup>&#xa0;eV/Å for forces. The model reproduces density functional theory (DFT)-level accuracy in lattice constants, phonon spectra, and elastic moduli, and is able to predict vacancy and surface energies with a relative error of less than 1%. Coupling the DP model with anodic dissolution theory, corrosion polarization is accurately reproduced, consistent with DFT and experiments. This work provides an efficient atomic-scale framework for understanding structure–property–corrosion relationships in Mg-Pr alloys and a reliable tool for designing biodegradable magnesium implants.</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

Atomic-Scale Modeling of Mechanical Properties and Corrosion Behavior in Mg-Pr Alloys Using a Deep Potential Approach

  • 期刊论文

  • 发表日期：

    2026年7月23日

查看全文
  • Ruotong Gao,
  • Dayong Cui,
  • Zhijie He,
  • Xiangjie Fu,
  • Jinlong Chen,
  • Xiaohua Yu,
  • Hongying Hou,
  • Ju Rong

摘要

The performance of orthopedic implants depends on biocompatibility, mechanical compatibility, and long-term reliability. Magnesium–praseodymium (Mg-Pr) alloys, with biodegradability and bone-like mechanical response, are promising candidates. Their low alloying content and pronounced local chemical complexity, however, challenge conventional empirical potentials, limiting atomistic simulation accuracy. Here, a deep potential (DP) model within an “AI for Science” framework is developed to investigate crystallography, mechanics, energetics, and corrosion kinetics of Mg-Pr alloys. The DP model outperforms empirical and mainstream machine-learning potentials, achieving root mean square errors of 3.61 × 10−3 eV/atom for energies and 1.84 × 10−3 eV/Å for forces. The model reproduces density functional theory (DFT)-level accuracy in lattice constants, phonon spectra, and elastic moduli, and is able to predict vacancy and surface energies with a relative error of less than 1%. Coupling the DP model with anodic dissolution theory, corrosion polarization is accurately reproduced, consistent with DFT and experiments. This work provides an efficient atomic-scale framework for understanding structure–property–corrosion relationships in Mg-Pr alloys and a reliable tool for designing biodegradable magnesium implants.