5 Center for Basic Research on Materials, National Institute for Materials Science (NIMS)Fig. 1 Relationship between dimensional reduction results and valence electrons.Fig. 1 Temporal changes in the standard deviation of a peak component.85Cobalt-based Heusler alloys exhibit functional properties such as high spin polarization and ferromagnetism, rendering them promising candidates for spintronic applications. These properties are derived from the Fermi surface, yet extracting information from it remains a significant challenge. In this study, we utilized machine learning to analyze the Fermi surfaces in the kx-ky direction for four compositions: Co2MnGaxGe1-x, Co2FeGaxGe1-x, Co2MnAlxSi1-x, and Co2FeAlxSi1-x. Our findings reveal a correlation between the shape changes of the Fermi surface and the valence electron counts and magnetic moments in each composition. The results indicate that simple machine learning can extract -In, Poster Award NomineeP5-09Fermi Surface Analysis of Multi-component Co-based Heusler Alloys Using Machine Learning Soichi Takase1, Daichi Ishikawa1, Kentaro Fuku1,Yoshio Miura2,3,Yasuhiko Igarashi4, Yuma Iwasaki5, Yuya Sakuraba2, Koichiro Yaji5, Alexandre Lira Foggiatto1, Varadwaj Arpita1, Naoka Nagamura5,and Masato Kotsugi11 Department of Material Science and Technology, Tokyo University of Science2 Research Center for Magnetic and Spintronics Materials, National Institute for Materials Science (NIMS)3 Electrical Engineering and Electronics, Kyoto Institute of Technology4 Institute of System and Information Engineering, University of Tsukubaparameters that correlate with physical quantities from the shape of Fermi surfaces.P5-10Real-time in-situ Machine Learning Analysis of RHEED Images for MBE Film-growthToshiro Osawa1, 2, Asako Yoshinari1, 2, Yasunobu Ando3, Tarojiro Matsumura4, Masato Kotsugi1,and Naoka Nagamura1,21 Tokyo University of Science, 2 National Institute for Materials Science (NIMS), 3 Institute of Science Tokyo,4 National Institute of Advanced Industrial Science and Technology (AIST)Molecular beam epitaxy (MBE) allows precise film-growth and reflection high-energy electron diffraction (RHEED) is used to evaluate structure and thickness. However, quantitative RHEED analysis requires expertise. We have developed machine learning approaches for automatic detection of surface superstructure changes and extraction of structural information[1]. In this study, we extend this method to a real-time structure evaluation during MBE film growth by integrating hardware and software. We applied this system to optimize the synthesis conditions for the target surface superstructures in MBE deposition of indium on a clean Si(111)”7×7” surface. During deposition, four steps were repeated every second: (1) RHEED image capture, (2) luminance histogram generation, (3) peak fitting using EMPeaks[2,3], and (4) plotting the maximum likelihood estimation results, including the standard deviation in each peak component. We succeeded in observing temporal changes in the standard deviation of a peak component in luminance histograms during deposition(Fig. 1). We found that the extreme values of the curve were the optimal conditions for the synthesis of -In, and 4×1-In surface superstructures.[1] A. Yoshinari et.al., Sci. Technol. Adv. Mater.: Methods 2, 162 (2022). [2] T. Matsumura et al., Sci. Technol. Adv. Mater., 20, 734 (2019). [3] EMPeaks (PyPI): https://pypi.org/project/EMPeaks/
元のページ ../index.html#85