基于原子间作用势的晶体结构去重算法及其应用

Crystal Structure Deduplication Algorithm Based on Interatomic Potential and Its Applications

摘要: 随着科学技术的发展，人们收集“大数据”的能力远远超过了分析它的能力。科学研究正经历从传统的实验观测、理论推演以及计算仿真向大数据研究转型，形成第四研究范式。材料数据信息学作为应用科学的重要分支之一，驱动着新材料的设计与发现。本文针对材料领域数据的冗余问题，提出一种基于原子间作用势的去重算法，可以有效地鉴别材料中的相同或相似结构，从而实现冗余数据的去重。实验结果表明，提出的算法与已有的材料结构去重算法相比，在准确性、鲁棒性以及计算效率各指标上均表现优异。进一步，本文利用开发的算法对材料领域三大知名数据库ICSD、CSD和COD数据库超过159万数据进行去重分析，有效地去除了101,643个相同和相似结构，并构建了开放共享的去冗余数据库，数据发布于：https://matgen.nscc-gz.cn/。

Abstract: With the development of science and technology, the ability of people to collect big data exceeds the ability to analyze it. Scientific research is undergoing a transformation from traditional experimental observation, theoretical deduction and simulation research to big data research, forming the fourth research paradigm. As one of the important branches of applied science, materials data informatics drives the design and discovery of new materials. Aiming at the problem of redundancy in the field of materials, this paper proposed a deduplication algorithm based on the interatomic potential, which can effectively identify the same and similar structures, thereby realizing the deduplication of redundant data. The experimental results show that this algorithm has excellent performance in accuracy, robustness and computational efficiency compared with the existing algorithms. Further, this paper used the proposed algorithm to deduplicate more than 1.59 million data in ICSD, CSD and COD databases, which effectively remove 101,643 same and similar structures, and built an open and shared de-redundancy database. The data was published at: https://matgen.nscc-gz.cn/.