平面团簇稳定结构的蒙特卡罗树搜索

发布时间：2018-07-05 15:33

本文选题：团簇 + 结构识别　；参考：《物理学报》2017年16期

【摘要】：以平面团簇为例提出了一种结合结构识别和蒙特卡罗树技术搜索稳定结构的新方法.体系原子之间的相互作用由两类模型势能函数来描述:Lennard-Jones二体势函数与基于Lennard-Jones势的三体势函数.考虑可能的三角晶格碎片作为候选结构,引入编号策略对结构进行快速识别,并运用蒙特卡罗树搜索研究稳定结构随着原子数增大的演化过程;对于能量较低的候选结构,进一步采取局域优化来获得对应体系的稳定结构.计算表明,Lennard-Jones二体势函数对应的三角晶格团簇更稳定;在特定的参数下,三体势函数对应的六角晶格团簇更稳定.结合结构识别和蒙特卡罗树搜索可以对候选结构空间进行高效扫描,在较短时间内更容易搜索到稳定的团簇结构,并可以与第一原理计算结合实现材料的结构预测.
[Abstract]:Taking planar clusters as an example, a new method for searching stable structures by combining structure recognition and Monte Carlo tree technique is proposed. The interaction between atoms in the system is described by two kinds of potential energy functions of model: Lennard-Jones two-body potential function and three-body potential function based on Lennard-Jones potential. Considering the possible triangular lattice fragments as candidate structures, the numbering strategy is introduced to identify the structures quickly, and Monte Carlo tree search is used to study the evolution process of stable structures with the increase of atomic number. Further local optimization is adopted to obtain the stable structure of the corresponding system. The results show that the triangular lattice cluster corresponding to Lennard-Jones two-body potential function is more stable, and the hexagonal lattice cluster corresponding to the three-body potential function is more stable under certain parameters. Combined with structure recognition and Monte Carlo tree search, the candidate structure space can be scanned efficiently, and the stable cluster structure can be easily found in a short time, and the structure prediction of materials can be realized by combining with first principle calculation.
【作者单位】：华南理工大学物理与光电学院;
【基金】：国家自然科学基金(批准号:11474100) 中央高校基本科研业务费(批准号:2017MS119)资助的课题~~
【分类号】：O562

【相似文献】