How AI Identifies Magnetic Structures Behind High-Frequency Energy Loss

Tokyo University of Science

From electric vehicles to renewable energy systems and data centers, modern technologies depend on efficient power conversion. These systems depend on magnetic components such as transformers and inductors. However, magnetic materials lose energy during operation, generating heat and limiting efficiency. As electrical systems operate at increasingly high frequencies, understanding the microscopic origins of these losses is becoming crucial to improving energy efficiency.

In soft magnetic materials, magnetization reversal occurs largely through the movement of magnetic domain walls, which separate regions magnetized in different directions. Although conventional models estimate overall magnetic losses, they do not readily identify which microscopic structures contribute most to resistance during magnetization reversal. Understanding these structures could offer new directions for designing lower-loss magnetic materials.

In an effort to address these challenges, a research team led by Professor Masato Kotsugi from the Department of Material Science and Technology, Tokyo University of Science (TUS), Japan, developed a physics-interpretable artificial intelligence (AI) approach to investigate the microscopic origins of high-frequency magnetic loss. Their study, published in NPG Asia Materials on November 17, 2026, was coauthored by Kosuke Nishioka (master's course completed in 2025) from the Department of Material Science and Technology at TUS.

Prof. Kotsugi says, "Using AI to see the invisible, this was the motivation that drove us to develop reliable AI methods that explain results through physical principles rather than simply producing predictions."

To investigate magnetization reversal, the researchers performed large-scale micromagnetic simulations using the Fugaku supercomputer, calculating an idealized iron-based nanocrystalline soft magnet at frequencies of 100 kHz, 1 MHz, and 10 MHz. They applied persistent homology, a mathematical technique that quantifies shapes and connectivity in images, to characterize evolving magnetic domain-wall networks. Principal component analysis compressed the resulting image-based descriptors into a two-dimensional representation. They then linked this representation to an image-derived exchange-energy measure and constructed an energy landscape with inverse mapping to identify the structures contributing most strongly to the energy barrier.

The analysis showed that magnetic domains became increasingly subdivided as excitation frequency increased. They identified an energy barrier near the coercive region, where magnetization reverses, that increased with frequency. Most importantly, inverse mapping revealed that high-contribution features were concentrated around propagating reversed-domain tips and selected fine wall segments, rather than being distributed uniformly across the network. This finding suggests that similar walls can play different roles during reversal. The identified barrier represents an exchange-derived structural energy cost, not a direct measurement of dissipated heat.

These findings offer a new way to investigate the microscopic structures associated with energy loss in magnetic materials. They could eventually guide the development of low-loss magnetic materials for electric vehicles and power-conversion equipment used in data centers. Understanding structures linked to greater resistance could help researchers investigate how material composition and processing influence domain-wall behavior.

Prof. Kotsugi explains, "Our study could contribute to energy conservation and CO₂ emission reduction through improved power conversion efficiency and reduced heat generation."

By integrating supercomputing, topology, magnetic physics, and explainable AI, this study connects complex magnetic structures with their contributions to energy barriers. While establishing a promising framework for investigating microscopic sources of energy loss and developing more efficient power electronics.

Reference

Title of original paper: Physics-interpretable topological machine learning identifies domain-wall contributors to high-frequency loss in a nanocrystalline soft magnet

Journal: NPG Asia Materials

DOI: 10.1038/s41427-026-00683-2

About The Tokyo University of Science

Tokyo University of Science (TUS) is a well-known and respected university, and the largest science-specialized private research university in Japan, with four campuses in central Tokyo and its suburbs and in Hokkaido. Established in 1881, the university has continually contributed to Japan's development in science through inculcating the love for science in researchers, technicians, and educators.

With a mission of "Creating science and technology for the harmonious development of nature, human beings, and society," TUS has undertaken a wide range of research from basic to applied science. TUS has embraced a multidisciplinary approach to research and undertaken intensive study in some of today's most vital fields. TUS is a meritocracy where the best in science is recognized and nurtured. It is the only private university in Japan that has produced a Nobel Prize winner and the only private university in Asia to produce Nobel Prize winners within the natural sciences field.

Website: https://www.tus.ac.jp/en/mediarelations/

About Professor Masato Kotsugi from Tokyo University of Science

Dr. Masato Kotsugi is a Professor in the Department of Materials Science and Technology, Faculty of Advanced Engineering, at Tokyo University of Science (TUS) in Japan. With a PhD from Osaka University in 2001, he joined TUS in 2015. His research focuses on magnetism, nanostructures, synchrotron radiation, and materials informatics. His work combines magnetic physics, data science, and explainable artificial intelligence to understand complex material behavior and support the discovery and design of high-performance functional materials. His research aims to contribute to energy-efficient technologies and a greener society.

Website: https://www.kotsugi.jp/

Funding information

This work was supported by the Ministry of Education, Culture, Sports, Science and Technology (MEXT), Japan, through the Program for Creation of Innovative Core Technology for Power Electronics (grant number JPJ009777); the Japan Society for the Promotion of Science (JSPS) KAKENHI (grant numbers 21H04656, 22A201 and 22H05107); and the Japan Science and Technology Agency (JST) through CREST (grant number JPMJCR21O1). Chiharu Mitsumata received support from the Tsukuba Research Center for Energy Materials Science (TREMS), University of Tsukuba.

/Public Release. This material from the originating organization/author(s) might be of the point-in-time nature, and edited for clarity, style and length. Mirage.News does not take institutional positions or sides, and all views, positions, and conclusions expressed herein are solely those of the author(s).View in full here.