Researchers Map Hidden Energy Losses in Electric Motor Magnetic Fields
Scientists at Tokyo University of Science have identified how magnetic domain structures in motors waste energy through heat, using artificial intelligence to decode patterns invisible to conventional analysis.
The work addresses a concrete problem in electric vehicle design: iron loss, which occurs when magnetic fields reverse direction inside motor cores. This process generates heat and reduces efficiency. Temperature makes the problem worse by partially demagnetizing the soft magnetic materials that form motor cores.
The culprit lies in how magnetic domains-tiny magnetic regions within materials-arrange themselves. Some materials contain intricate "maze domains," structures with zig-zag patterns that shift abruptly with temperature changes. Understanding these shifts has been difficult because multiple factors interact: microscopic structure, thermal effects, and energy stability.
A Physics-Based AI Framework
Professor Masato Kotsugi and Dr. Ken Masuzawa developed a model called the entropy-feature-eXtended Ginzburg-Landau (eX-GL) model to study how maze domains behave in rare-earth iron garnet materials. The team worked with collaborators from three other Japanese universities.
"Conventional simulations oversimplify real materials, while experiments reveal complexity without a clear way to quantify cause and effect," Kotsugi said. "Our physics-based explainable artificial intelligence framework addresses these limitations and is designed to mechanistically explain temperature-dependent magnetization reversal."
The approach combined three steps. First, researchers captured microscopic images of magnetic domains at different temperatures. They then used persistent homology-a mathematical technique for identifying structural patterns in data-to detect uneven characteristics in those images. Machine learning pattern recognition followed, producing a digital map of free energy as magnetic structures evolved.
Four Energy Barriers Identified
The analysis revealed a dominant feature the team called PC1, which tracked how magnetization reversed. By connecting PC1 to physical properties, the researchers visualized four major energy barriers controlling how magnetization changes.
They found that maze domains grow more complex as domain walls lengthen. This complexity stems from interactions between entropy and exchange forces-the magnetic forces holding atoms in place. The results clarified the physical mechanisms behind maze-domain reversal.
The findings appeared in Scientific Reports.
Broader Applications Beyond Motors
Kotsugi noted the model's potential extends beyond electric motors. "Since free energy is a universal thermodynamic metric, our model can be extended to other systems with similar characteristics," he said.
The work introduces a strategy for investigating complex energy patterns in magnetic systems and related materials. For researchers working on materials science and AI for science and research, the approach demonstrates how machine learning can automate interpretation of complex physical processes that resist conventional methods.
The research was funded by the Japan Society for the Promotion of Science and the Japan Science and Technology Agency.
Your membership also unlocks:
