Incorporating reaction kinetics into Bayesian inference － Advanced analysis of dynamics for practical applications of functional materials －
- Release Date
- 12 Sep, 2023
- BL02B2 (Powder Diffraction)
11 September 2023
JASRI
Kumamoto University
Japan Science and Technology Agency (JST)
|
The Japan Synchrotron Radiation Research Institute (JASRI) and Kumamoto University have collaborated to develop an analytical method by incorporating models of the time evolution of reactions, such as reaction kinetics, into machine-learning-based Bayesian estimation. By applying this method to the dynamic processes (dynamics) of gas adsorption in porous materials, they demonstrated that it is possible to (1) estimate the start time of dynamics, (2) select the most plausible reaction model on the basis of data, and (3) correctly evaluate the estimation accuracy, which have been considered impossible by the conventional method. Porous materials having many small holes are attracting considerable attention as functional materials for solving environmental and energy problems through gas separation and storage. Understanding the dynamics of materials exhibiting their functions is important in material development for practical applications. With the analytical method they developed, much more information about dynamics can be derived from experimental data than with the conventional method. A better understanding of dynamics will be achieved by utilizing the information, which will accelerate material development. The application of the analytical method they developed is not limited to gas storage. It can also be applied to the analyses of various dynamics including chemical reactions. This analytical method will be used in a wide range of areas in materials science. This achievement was obtained by the collaborative research group led by Yuichi Yokoyama (Research Scientist) with Shogo Kawaguchi (Senior Scientist) of JASRI and Masaichiro Mizumaki (Professor) of Kumamoto University. The results of their study were published in the online version of Scientific Reports, a scientific journal owned by Springer Nature, on 12 September 2023. 【Publication】 |
Fig. 1: Time evolution of fraction transformed that is determined by Bayesian inference and conventional method. The results showed a time lag between the gas-shot time ts and the adsorption start time to, demonstrating the importance of estimating the adsorption start time by Bayesian inference.
Fig. 2: Posterior probability distributions of adsorption start time and reaction rate. The probability of the adsorption start time coinciding with ts, the conventional indicator, was 0%. The adsorption start time was found to be 7.4461 ± 0.0287 [s]. The reaction rate was found to be 0.6192 ± 0.0235 [1/s].
Contact: |
spring8.or.jp- Previous Article
- Discovering Hidden Order in Disordered Crystals（ Press Release ）
- Current article
- Incorporating reaction kinetics into Bayesian inference － Advanced analysis of dynamics for practical applications of functional materials －