Construction and application of materials database under data-driven research paradigm

<p indent="0mm">In recent years, with the rapid development of big data and artificial intelligence technologies, the data-driven paradigm for materials research has increasingly demonstrated significant advantages in areas such as mining structure-property relationships and screening and designing novel materials. It relies on vast amounts of data, employing high-throughput screening or artificial intelligence aided methods to shorten the materials research and development cycle while reducing scientific research costs. In this paradigm, data serves not only as the foundation but also plays a crucial role in expanding the search space for high-throughput screening and enhancing the performance of artificial intelligence models. Consequently, establishing high-quality materials databases has become an essential component of data-driven materials research. Since the launch of the Materials Genome Initiative, both theoretical and experimental databases have seen continuous growth in number and scale, characterized by automatic data generation, diversified data access, and standardized data representation. These databases strive to meet the demands for high concurrency and scalability in the context of big data while improving data reliability and usability. This review focuses on materials database construction under data-driven research paradigm, outlining the key of database construction from four aspects: Data generation, data preprocessing, data storage, and data access. In the data generation phase, methods like high-throughput computation and experimentation facilitate batch acquisition of theoretical and experimental data. When calculation or synthesis methods are reasonable, both data accuracy and efficiency of data acquisition can be achieved simultaneously. Text mining, based on natural language processing technology, can extract synthetic information, material entities, or properties from massive scientific literature, which is also an important supplementary means of data generation. Data preprocessing is vital for ensuring compliance with FAIR (findable, accessible, interoperable, and reusable) principles and enhancing data quality which can be achieved through data standardization and data filtering. During the data storage phase, it is necessary to construct a rational data model and choose between relational and non-relational databases for storing data, ensuring efficient and flexible data storage. Data retrieval can be facilitated by designing effective query strategies and offering diverse access methods. Subsequently, a brief summary is offered on representative materials databases from around the world, considering aspects such as the volume and variety of data, access methods, and distinctive features of the databases. We will also integrate selected case studies from database applications to underscore the pivotal role and advantages of databases within the data-driven research paradigm. Finally, we discuss future developments in materials databases from two perspectives: The material data itself and the construction of data infrastructure.