We developed open-source software Open Data Analysis Tool for Science and Engineering (ODAT-SE) [1], by the PASUMS project in FY2024 as a major upgrade of 2DMAT [2, 3]. ODAT-SE solves an inverse problem, when one gives a direct or forward model, a physical or statistical model representing the system under investigation. Currently, ODAT-SE offers five analysis methods: (i) grid search, (ii) Nelder-Mead optimization, (iii) Bayesian optimization, (iv) replica exchange Monte Carlo method, and (v) population-annealing Monte Carlo (PAMC) method.

The code structure of ODAT-SE is drawn schematically in Fig. 1. The original architecture in 2DMAT was tightly coupled with specific experimental techniques, limiting its flexibility and reusability across other scientific fields. In ODAT-SE, the architecture explicitly separates direct problems from the optimization or search algorithms. This modular approach enables researchers to apply ODAT-SE to diverse fields; users can easily add their own direct problem solvers or search algorithms tailored to their research needs.

Hereafter, we focus on the PAMC method, a massively parallel Bayesian inference. In general, the Bayesian inference gives the posterior probability distribution $P(X''\mid ''Y)$, as histogram, where $X$ is the target quantity (vector), the quantity that we would like to know, and $Y$ is the experimentally observed quantity (vector). The PAMC method is suitable to supercomputers. Since the PAMC method is a global search algorithm, one can find the global solution and local solutions in the data space of $X$. The PAMC method was used to determine the surface structure of the 3×3-Si phase on the Al (111) surface by total reflection high-energy positron diffraction experiment (https://www2.kek.jp/imss/spf/eng/, Fig. 2(a)) and core-level photoemission spectroscopy [4]. The analysis finds the global solution as a flat surface structure shown in Fig. 2(b) and local solutions, which indicates the crucial importance of global search algorithm.

In future outlook, ODAT-SE will be developed further and used both in plasma and material science, for example, through our project launched recently in the Moonshot R&D Program [5]. Notably, a recent study [6] using 2DMAT have successfully demonstrated efficient fitting of high-dimensional experimental parameters. With the modular architecture of ODAT-SE, similar complex analyses can now be performed more easily, enabling broader applicability across various scientific fields.

References

• [1] https://www.pasums.issp.u-tokyo.ac.jp/odat-se/
• [2] Y. Motoyama et al., Comp. Phys. Commun. 280, 108465 (2022).
• [3] https://www.pasums.issp.u-tokyo.ac.jp/2dmat/
• [4] Y. Sato et al., Phys. Rev. Materials 9, 014002 (2025).
• [5] https://www.jst.go.jp/moonshot/en/program/goal10/A3_hoshi.html
• [6] S. Liu et al., Phys. Rev. Lett. 135, 056502 (2025).

Authors

• T. Hoshi^a,b, A. Nakano^a,b, T. Aoyama, Y. Motoyama, and K. Yoshimi
• ^aNational Institute for Fusion Science
• ^bThe Graduate University for Advanced Studies, SOKENDAI