60High brilliance synchrotron radiation (SR) X-rays realize multi-dimensional, multi-scale, and extremely high-resolution analysis. Advanced SR spectromicroscopy potentially produce huge number of datasets. Efficient interpretation of spectral big data beyond manual peak assignment is an urgent issue. Here we propose a high-throughput and low-computational-cost spectral peak fitting analysis method assisted by machine learning. We have developed “spectrum-adapted” expectation-maximization (EM) algorithm for maximum likelihood estimation of parameters at each peak component[1]. We adopted this technique for peak shift detection in spectral imaging data of X-ray photoemission spectroscopy (XPS) taken by a SR soft X-ray operando scanning photoelectron microscopy system. Drastic acceleration of peak fitting analysis was achieved in comparison to the conventional approach. Moreover, we succeeded to clarify the local carrier doping in the atomic layer device microstructures, which had been overlooked by conventional analyses[2,3]. Our analysis methods can be applied to several kinds of spectral data and peak structures, such as X-ray absorption spectra, Raman spectra, fluorescence spectra, luminescence histograms, and so on. In coordination chemistry, the only methods for structure estimation have been single crystal X-ray structure analysis or DFT calculations. Recently, X-ray absorption fine structure (XAFS) has been attracting attention as a method for estimating coordination environments. However, the use of XAFS (especially XANES) in coordination chemistry is not common technique, and the most of its use are limited to “fingerprint matching” such as comparison with reference samples. In this study, we measured the XAFS spectra of Ni complexes, and aimed to estimate the coordination environment conveniently by spectral clustering using machine learning. As a result of our experiments, we succeeded in cluster partitioning by the number of coordination and the chemical environment of the coordinating atoms. And the clusters to which they belonged were changed following the ligand exchange reaction by adding ligands. These results suggest that the combination of XANES and machine learning will make it possible to estimate the coordination environment more easily. Poster Award NomineeP2-07Machine-Learning Based Analysis for Synchrotron X-ray Spectral Imaging Naoka Nagamura1,2,3, Tarojiro Matsumura4, Yasunobu Ando5, Kenji Nagata1, and Shotaro Akaho4 1 National Institute for Materials Science (NIMS), 2 Tokyo University of Science, 3 Tohoku University,4 National Institute of Advanced Industrial Science and Technology, 5 Institute of Science Tokyo [1] T. Matsumura et al., Sci. Tech. Adv. Mat., 20, 733 (2019).[2] T. Matsumura et al., Sci. Tech. Adv. Mat Methods., 4, 2373046 (2024).[3] M. Okada et al., APL Materials, 9, 121115 (2021). P2-08XANES spectral analysis of Ni complexes using machine learning Kentaro Fuku1, Takefumi Yoshida2, Tetsu Sato3, Hiroaki Iguchi4, Shinya Takaishi5, Ryota Sakamoto5,Hitoshi Abe6,7,81 Tokyo Univ. of Science, 2 Wakayama Univ., 3 IMRAM, Tohoku Univ., 4 Nagoya University, 5 Tohoku University, 6 KEK-IMSS, 7 SOKENDAI, 8 Ibaraki Univ.
元のページ ../index.html#60