New Computational Model Decodes Maze Like Magnetic Patterns to Solve Energy Efficiency Bottlenecks in Electric Vehicle Motors

Tokyo University of Science researchers develop the eX-GL model to identify why maze-like magnetic patterns cause energy loss in high-temperature EV motors.

By: AXL Media

Published: Apr 20, 2026, 8:58 AM EDT

Source: Information for this report was sourced from EurekAlert!

The Challenge of Hysteresis in High Temperature Motors

The global transition to electric vehicles (EVs) has placed a premium on the energy conversion efficiency of electric motors. A primary obstacle to this efficiency is "iron loss," or magnetic hysteresis loss, which occurs when magnetic fields within the motor core are repeatedly reversed. These cores, typically made of soft magnetic materials, are subject to high-temperature environments that complicate their performance. Understanding the behavior of "magnetic domains"—microscopic regions with uniform magnetization—is critical to mitigating this energy dissipation, yet their complex structures have long eluded standard simulation techniques.

Decoding the Complexity of Maze Domains

In certain soft magnetic materials, these domains arrange themselves into intricate, zig-zag configurations known as maze domains. These structures exhibit abrupt and erratic changes when exposed to fluctuating temperatures, directly influencing the amount of energy lost as heat. Because these patterns are shaped by a volatile mix of metallographic structure, thermal stability, and energy exchange, conventional models often oversimplify the material, while physical experiments fail to quantify specific causes and effects. To bridge this gap, Professor Masato Kotsugi and his team introduced a new entropy-feature-eXtended Ginzburg-Landau (eX-GL) model.

An AI Framework for Explainable Physics

Published in Scientific Reports on February 11, 2026, the study details a novel "explainable artificial intelligence" framework. Unlike traditional "black box" AI, the eX-GL model is designed to mechanistically explain the temperature-dependent process of magnetization reversal. The process begins with persistent homology (PH), a mathematical tool used to extract structural features from microscopic images of rare-earth iron garnets. These features are then processed through machine learning to create a digital "free-energy landscape," essentially mapping out the energetic cost and stability of various magnetic configurations.