[17p-S2_204-1~18] 23.1 Joint Session N "Informatics"

3:15 PM - 3:30 PM JST(6:15 AM - 6:30 AM UTC)Presentation by Applicant for JSAP Young Scientists Presentation Award

[17p-S2_204-7] Data-Driven Analysis of Magnetic Domain Structures and Magnetic Free Energy Using Supervised Variational Autoencoders

〇Kanji Tsubouchi¹, Ryunosuke Nagaoka¹, Michiki Taniwaki¹, Yotaro Machida¹, Alexandre Lira Forggiatto¹, Masato Kotsugi¹ (1.Tokyo Univ. of Sci.)

Keywords:

magnetic material,spin texture,deep learning

Spin textures are a crucial source of information governing macroscopic magnetic properties. Analyzing the causal relationship between microstructure and its formation factors is essential for advancing magnetic devices. However, quantifying the spatial inhomogeneity of magnetic structures is challenging due to their complex shapes. Therefore, this study utilizes supervised variational autoencoders to map complex spatial magnetization distributions into low-dimensional latent representations. Specifically, it focuses on the energy underlying the functionality, aiming to extract structural features that encapsulate information about both structure and energy.