Xiaoyan Song*, Guojing Xu, Hao Lu, Chongyu Han, Mengyao Shan
Faculty of Materials and Manufacturing, Key Laboratory of Advanced Functional Materials,
Education Ministry of China, Beijing University of Technology, Beijing, 100124, China
EXTENDED ABSTRACT: Among various Sm-Co compounds, SmCo7 shows very important potential as a high-performance high-temperature permanent magnet due to its excellent comprehensive magnetic properties and low temperature coefficient of coercivity. However, due to the TbCu7-type crystal structure, SmCo7 is a metastable phase (1 :7H) and cannot exist stably at room temperature. The phase stability of 1 :7H greatly limits the application of SmCo7 based alloys. Though previous studies have improved the phase stability of 1 :7H by reducing the grain size of the alloy and adding appropriate elements, it is very difficult to obtain the law of stability of a specific phase in the Sm-Co based alloys and the law of phase transformation through experimental studies. With the development of data-driven materials design, it has become possible to study the stabilities of performance-dominant phases by data-driven methods based on data mining and machine learning. The methods are applicable to a wide range of materials composition and more influencing factors. In this field, we established a machine learning model for the phase stability of Sm-Co-based alloys, and proposed two key characteristics affecting the structural stability of the 1 :7H phase, i.e., the melting point of the doped element and the difference in the electronegativity between the element and Co. The high throughput predictions of the virtual samples of the SmCo7-xMx (M stands for doping elements) alloys were carried out. The characteristics of the phase constitution and the phase stability of the alloy system under the conditions of the doping elements and the grain size of the alloy were given for the first time. The model and method established in this study provide important quantitative guidance and scientific basis for the composition design and grain structure control for the alloy systems containing metastable phases.
Keywords: metastable phase; phase stability; machine learning; data-driven multi-phase alloy
Prof. Xiaoyan Song, from Beijing University of Technology, is winner of China National Science Fund for Distinguished Young Scholars, Talents of Outstanding Contribution of Beijing, and Humboldt Fellow. She was awarded the "Provincial Science and Technology Progress Award First Prize" once and the "Municipal Natural Science and Technology Invention Award Second Prize" for three times. She has published over 300 SCI papers and has over 70 patents authorized. She has been engaged as the Associate Editor of Int. J. Refract. Met. Hard Mater. since 2013, and served as committee members of international and domestic academic societies such as PTC, ISS and CMRS.