6-9. Accelerating materials discovery based on generalized low-dimensional conformation-performance relationships

Xiangying Meng^1,2*, Zixi Jia³, Xiaofeng Liu⁴, Gaowu Qin^2,*

1. College of Sciences; 2. Key Laboratory for Anisotropy and Texture of Materials (Ministry of Education), School of Materials Science and Engineering; 3. Faculty of Robot Science and Engineering; 4. Information Construction and Network Security Offic, Northeastern University, Shenyang 110819, China

Abstract: The construction of conformation-performance relationships (CPRs) is the core issue of efficient material design. In this paper, we propose a generalized low-dimensional CPRs (GLD-CPRs) strategy based on data processing technology for material design. This strategy can narrow the searching space and thus accelerate materials discovery. From a pure computing perspective, the strategy covers three parts: data-generation, data-management and data-utilization. In these aspects, the critical issues of data processing technology, including correlation analysis, attributes construction, data-cleaning and classification, weight evaluation, regression model training etc, are well illustrated form the viewpoint of material science, which provides a universal solution to the efficient material oriented design. This solution has been concreted by a case study of semiconductor band gap engineering. Based on more than 20,000 items of semiconductor band gap and attributes data, a set of CPRs for band gap regulation towards arbitrary semiconductor of the same kind can be generated using GLD-CPRs strategy. By comparing with the reported studies, the reliability of the strategy is confirmed. Finally, future improvements needed by GLD-CPRs are addressed. The topic of this work highlights the construction of CPRs using material informatics method, adds new depth and width to our understanding of material gene on performance. Facing the challenge of GLD-CPRs, it is hopeful to broaden the research field of materials science.

Keywords: Machine learning; Material design; Band gap engineering; Material database; High throughput calculations

Figure 1. Implement of GLD-CPRs strategy in bandgap design.

Figure 2. Weight evaluation of bandgap genes for binary semiconductors.

Figure 3. Regression model prediction of bandgap changes with components in rutile-TiO₂.