标题： 点模式匹配技术用于凝聚态物质中的局部结构分析 显示英文标题

标题： Point-Pattern Matching Technique for Local Structural Analysis in Condensed Matter

摘要： 原子的局部排列是材料机械和功能性质最重要的预测因素之一。 然而，用于识别复杂材料系统中原子几何排列的算法尚不存在。 为了解决这一挑战，我们提出了一种点模式匹配算法，该算法可以检测给定原子坐标集中的“模板”结构实例。 据我们所知，这是第一个针对具有非常不同原子数的原子构型的几何比较技术，并且当对齐构型所需的最优旋转和平移未知时也是如此。 该模式匹配算法可以与适当的一组度量结合，以量化原子构型之间的相似性或差异性。 我们通过两个例子展示了点模式匹配算法的独特能力，其中自动分析局部原子结构有助于推进材料科学中结构-性能关系的理解：(a) 在界面缺陷上识别局部三维多面体单元，以及 (b) 分析金属玻璃原子结构中的准二十面体拓扑结构。 我们预计该模式匹配算法将在凝聚态系统中广泛区域的原子结构分析中具有应用价值，包括生物分子、聚合物和无序金属系统。 该算法的在线实现通过托管在GitHub上的开源代码提供（https://github.com/spatala/ppm3d）。 显示英文摘要

摘要： The local arrangement of atoms is one of the most important predictors of mechanical and functional properties of materials. However, algorithms for identifying the geometrical arrangements of atoms in complex materials systems are lacking. To address this challenge, we present a point-pattern matching algorithm that can detect instances of a `template' structure in a given set of atom coordinates. To our knowledge this is the first geometrical comparison technique for atomistic configurations with very different number of atoms, and when the optimal rotations and translations required to align the configurations are unknown. The pattern matching algorithm can be combined with an appropriate set of metrics to quantify the similarity or dissimilarity between configurations of atoms. We demonstrate the unique capabilities of the point-pattern matching algorithm using two examples where the automated analysis of local atomistic structure can help advance the understanding of structure-property relationships in material science: (a) identifying local three-dimensional polyhedral units along interfacial defects, and (b) the analysis of quasi-icosahedral topologies in the atomistic structure of metallic glasses. We anticipate that the pattern matching algorithm will be applicable in the analysis of atomistic structures in broad areas of condensed matter systems, including biological molecules, polymers and disordered metallic systems. An online implementation of the algorithm is provided via the open source code hosted on GitHub (https://github.com/spatala/ppm3d).