基于多种群差分进化算法的分子动力学模拟体系中局部晶体结构识别

Local crystal structure identification in molecular dynamics simulation system based on differential evolution with multi-population

为了研究晶体形成的相变过程,提出了一种基于多种群差分进化算法的分子动力学模拟体系中局部晶体结构识别算法.首先,结合模拟体系原子/离子间连通性及其与晶体结构的几何相似度,构造出表示两者间重合程度的目标函数.然后,利用多种群差分进化算法获得目标函数最优解,进而确定模拟体系中排列方式与晶体结构相似的原子/离子.最后,利用Al、Mg、Fe、α-Fe_(2)O_(3)、Fe_(3)O_(4)、α-FeOOH、Fe(OH)_(2)七个聚集体对该算法进行验证.结果表明,所提算法不仅能够从模拟体系中识别出面心立方、密排六方、体心立方等简单晶体结构,还可识别出排布较为复杂的晶体结构,并可将聚集体直接分割为数个取向不同的晶体区域,适用于晶体成核、生长及相变过程的分子动力学模拟研究.

To study the phase transition process of crystal formation,a local crystal structure identification algorithm for molecular dynamics simulation systems based on the multi-population differential evolution algorithm was proposed.Firstly,combining with the connectivity between the atoms or ions of the simulation system and its geometric similarity to the crystal structure,an objective function representing the coincidence degree was constructed.Then,the multi-population differential evolution algorithm was used to obtain the optimal solution of the objective function,and the atoms or ions with the arrangement similar to the crystal structure were determined in the simulation system.Finally,this algorithm was verified by using 7 aggregates of Al,Mg,Fe,α-Fe_(2)O_(3),Fe_(3)O_(4),α-FeOOH,Fe(OH)_(2).The results show that the proposed algorithm can identify not only simple crystal structures such as face centered cubic,hexagonal close packed and body centered cubic structures from the simulation system,but also complex crystal structures,and even can directly segment the aggregate into several crystal regions with different orientations.It provides an effective data analysis method for molecular dynamics simulation of crystal nucleation,growth and phase transition.