论文标题
通过机器学习的分子动力学模拟，MO中晶体缺陷形成的计算研究
Computational study of crystal defects formation in Mo by machine learned molecular dynamics simulations
论文作者
论文摘要
在这项工作中，我们研究了由于中子轰炸在室温下的原子原子原子范围为0.5-10 keV时，晶体钼材料样品的损害。我们根据高斯近似电势（GAP）框架进行机器学习的分子动力学（MD）模拟，具有先前开发的原子质潜力。我们利用最近开发的软件工作流程来指纹和可视化损伤晶体结构中的缺陷，通过计算碰撞级联反应期间和之后的点缺陷来分析受损的MO样品。作为基准，我们报告了形成Frenkel对（自相关原子和单个空缺）和PKA能量函数的总数（自相关原子和单个空缺）的结果。提出了与使用嵌入式原子方法（EAM）势获得的结果的比较，以讨论机器学习的MD模拟的优点和限制。 Frenkel对的形成遵循与$ e^{0.54} _ \ mathrm {pKa} $相关的sublinear缩放定律，以与GAP MD结果和$ e^{0.667} _ \ Mathrm {pka} $进行$ e^{0.667} _ \ e^{0.667} $。尽管两种方法的平均数量总缺陷都相似，但我们注意到，MD电位模型的复杂点缺陷的不同原子几何形状模型，其中人群的形成对间隙电位更有利。最后，报告并讨论了间隙MD模拟的离子束混合结果。
In this work, we study the damage in crystalline molybdenum material samples due to neutron bombardment in a primary knock-on atom range of 0.5-10 keV at room temperature. We perform machine learned molecular dynamics (MD) simulations with a previously developed interatomic potential based on the Gaussian Approximation Potential (GAP) framework. We utilize a recently developed software workflow for fingerprinting and visualizing defects in damage crystal structures to analyze the damaged Mo samples by computing the formation of point defects during and after a collision cascade. As a benchmark, we report results for the total number of Frenkel pairs (a self-interstitial atom and a single vacancy) formed and atom displacement as a function of the PKA energy. A comparison to results obtained by using an Embedded Atom Method (EAM) potential is presented to discuss the advantages and limits of the machine learned MD simulations. The formation of Frenkel pairs follows a sublinear scaling law related to the PKA energy with $E^{0.54}_\mathrm{PKA}$ to the GAP MD results and $E^{0.667}_\mathrm{PKA}$ for the EAM simulations. Although the average number total defects is similar for both methods, we notice that MD potentials model different atomic geometries for the complex point defects, where the formation of crowdions is more favorable for the GAP potential. Finally, ion beam mixing results for GAP MD simulations are reported and discussed.
