Oral
02 Nov 2023
In recent years, the quest for advanced material design, crucial for low-power, high-speed next-generation electronic devices and high-efficiency electric vehicle motors, has intensified. Microscopic image data, direct evidence of device operation, forms a critical source of information for this design process. However, the interpretation of such data, especially for nanoscale magnetic bodies exhibiting complex nanostructures and interactions, has been a challenging task. Until now, image data interpretation has been largely qualitative and subjective, leading to a significant issue in functional design: an insufficient understanding of the underlying mechanisms, hindering device performance improvement. To address these challenges, we developed the "Extended Landau Free Energy Model," a unique blend of topology, information science, and free energy, enabling automated interpretation of image data (Figure 1) [1-6]. Traditionally, the Landau Free Energy Model, a tool to explain magnetization reversal phenomena based on magnetization and magnetic fields, was only applicable to single crystals, making it inadequate for real materials with nanostructures. We overcame this limitation by incorporating topology and data science into the model, creating a framework capable of analyzing real materials. Our approach employed Persistent Homology, a concept of topology, to extract complex magnetic domain structures as features. We then applied interpretable machine learning to draw a new energy landscape in the information space, constructing the Extended Landau Free Energy Model. Characterized by the use of physics-based features for analyzing the magnetization reversal process, our model establishes a relationship between magnetic domain structure changes and energy barriers through simple variable transformation and differentiation. It facilitates a bidirectional connection between micro magnetic domain structures and macro magnetization reversal phenomena across hierarchies (Figure 2), further allowing for the quantitative analysis of underlying physical interactions. Applying our model to the analysis of the information recording process of nanomagnetic bodies revealed the dominance of the demagnetization field effect. It also succeeded in visualizing the spatial concentration of energy barriers impeding the change in magnetic domain structure. This visualization indicates an improved understanding of mechanisms previously challenging to analyze visually and the ability to highlight subtle changes in microscopic images. The model potentially transforms overlooked microscopic data into a "goldmine" of information. Moreover, based on our model, we successfully proposed nanostructures with lower energy consumption, suggesting potential applications in the reverse design of devices. Applicable to various materials driven by complex mechanisms, our model promises wide-ranging contributions to the fabrication of diverse products, including quantum information technologies, electric vehicle motors, and autonomous distributed systems.References: 1) S. Kunii, K. Masuzawa, A. L. Fogiatto, C. Mitsumata & M. Kotsugi, Sci. Rep. 12, 19892 (2022) 2) A. L. Foggiatto, S. Kunii, C. Mitsumata & M. Kotsugi, Communications Physics 5, 277 (2022) 3) S. Kunii, A. L. Foggiatto, C. Mitsumata & M. Kotsugi, Sci. Tech. and Adv. Mater. Methods, 2 445-459 (2022) 4) C. Mitsumata, M. Kotsugi, J. Magn. Soc. Jpn, 46, (2022) pp. 90-96 5) K. Masuzawa, S. Kunii, A. Foggiatto, C. Mitsumata, M. Kotsugi, T. Magn. Soc. Japan. 6 (2022), 1-9. 6) T. Nishio, M. Yamamoto, T. Ohkochi, D. Nanasawa, A. Foggiatto, M. Kotsugi, Sci. Tech. and Adv. Mater. Methods, 2, pp 345-354, (2022)
