Researchers at Tokyo University of Science have developed an AI-powered physics model that could help improve understanding of magnetic energy loss in electric motor materials, a factor that affects motor efficiency in electric vehicles (EVs) and other electrified systems.
Electric motors experience iron loss, also known as magnetic hysteresis loss, when magnetic fields inside the motor repeatedly reverse direction. This process can waste energy as heat within the motor core, which is typically made from soft magnetic materials. Iron loss accounts for roughly 31% of a motor’s energy loss, most of it dissipated as heat.
New computational model identifies origin of complex magnetization reversal in soft magnets.
Because EV motors often operate under demanding thermal conditions, thermal effects can also contribute to partial demagnetization, further complicating how energy loss occurs inside motor materials.
A key factor behind these effects is the behavior of magnetic domains, which are microscopic magnetic regions inside materials. The arrangement and structure of these domains can affect how magnetic materials respond to heat and how much energy they lose during operation. In some soft magnetic materials, these domains form complex zig-zag structures known as maze domains, which can change abruptly as temperatures rise or fall.
“Conventional simulations oversimplify real materials, while experiments reveal complexity without a clear way to quantify cause and effect,” said Prof. Masato Kotsugi of Tokyo University of Science. “Our physics-based explainable artificial intelligence framework addresses these limitations and is designed to mechanistically explain temperature-dependent magnetization reversal process.”
The research team, led by Kotsugi and Dr. Ken Masuzawa from the Department of Material Science and Technology at Tokyo University of Science, worked with collaborators from the University of Tsukuba, Okayama University, and Kyoto University. Their study, “Explainable analysis of the complex maze magnetic domain structure through extension of the Landau free energy model by adding an entropy feature,” was published in Scientific Reports.
The model, called the entropy-feature-eXtended Ginzburg-Landau, or eX-GL, model, was developed to analyze the energy landscape of maze domains in a rare-earth iron garnet sample. The approach extends the Landau free energy model by adding an entropy feature, allowing the researchers to connect complex magnetic-domain images with the physical mechanisms driving magnetization reversal.
To study the process, the researchers captured microscopic images of magnetic domains at different temperatures. The eX-GL model first used persistent homology, a mathematical method for identifying topological features in data, to extract structural features from the domain images. Machine-learning-based pattern recognition was then used to identify the most important features from that data and create a digital free-energy landscape showing how magnetic microstructures evolve with energy changes.
According to the researchers, the model identified a dominant feature, called PC1, that captured the magnetization reversal process. By correlating PC1 with physical parameters, the team visualized four key energy barriers that play important roles in magnetization reversal dynamics.
The analysis also showed how exchange interactions, demagnetizing effects, and entropy contribute to changes in the maze-domain structures. The researchers found that maze domains become more complex as domain-wall length increases, a process driven by the coupling of entropy and exchange interactions.
For EV motor development, the work could provide a new way to study the microscopic behavior behind magnetic energy loss. By improving how researchers interpret complex magnetic-domain patterns, the approach may support future development of motor-core materials with lower losses and better thermal performance.
“Our eX-GL approach effectively automates the interpretation of complex magnetization reversal process and enables identification of hidden mechanisms, difficult to discern using conventional methods,” said Kotsugi. “In addition, since free energy is a universal thermodynamic metric, our model can be extended to other systems with similar characteristics.”
The research was conducted by Tokyo University of Science with collaborators from the University of Tsukuba, Okayama University, and Kyoto University. The study was supported by the Japan Society for the Promotion of Science, JST-CREST, and the Tsukuba Research Center for Energy Materials Science.