主催: The Japan Society of Vacuum and Surface Science
会議名: 2023年日本表面真空学会学術講演会
開催地: 名古屋
開催日: 2023/10/31 - 2023/11/02
To promote materials development based on data-driven science, it is essential to develop technologies for extracting features from high-dimensional measurement data such as images and spectra. There is a need for a system that automatically converts high-dimensional measurement data into features and stores the table data under AI-ready status. A cloud system, Research Data Express (RDE) [1], is being developed to register experimental and computational data quickly. The RDE has workflows to automatically extract features from high-dimensional measurement data.
As one of the feature extraction tools, we have developed an automatic spectral decomposition tool using reference data associated with physical states such as electron bound state and crystalline structure, etc. The developed tool can be used as part of the workflow functionality of the RDE. This spectral decomposition has high interpretability because it uses reference spectral data. The figure shows four concepts in the design of this spectral decomposition tool. Four concepts are presented as follows:
・[Reference] A spectral decomposition tool using a basis (reference data) linked to a physical state.
・[Integration] Multiple spectra can be integrated and shared parameters can be controlled.
・[Selection] Reference spectra from candidates based on data can be selected automatically.
・[Customization] Analysis models (peak, BG, noise models) can be customized. This spectral decomposition tool enables the automatic analysis of spectral data using reference data. In this presentation, examples of its use will be presented on the subjects of X-ray diffraction data and X-ray photoelectron spectroscopy data.
[1] Research Data Express (RDE), https://dice.nims.go.jp/services/RDE/