Physics-Trained AI Discovers Viable Rare-Earth-Free Permanent Magnets

Scientists at the Ames National Laboratory have successfully utilized a novel artificial intelligence workflow to discover a new class of permanent magnets that require zero rare-earth elements. The breakthrough, announced as part of the U.S. Department of Energy's Genesis Mission, represents a major leap in materials science that could reshape global manufacturing. By leveraging an AI model that understands the fundamental physics of electron behavior, researchers were able to bypass decades of trial-and-error laboratory work. The resulting material—a specialized composite of manganese and bismuth—demonstrates the ability to retain its magnetic field under extreme conditions, matching the performance profile required for commercial and military applications without using a single ounce of rare-earth metals.[1][2]

Permanent magnets are the invisible, indispensable workhorses of the modern industrialized world. They are essential components in everything from the electric motors powering consumer electric vehicles to critical defense applications, including advanced radar systems, nuclear submarines, and the F-35 Lightning II fighter jet. To function effectively in these high-stress environments, the magnets must maintain their magnetization even when subjected to extreme heat, radiation, and mechanical friction. Historically, only specialized alloys containing rare-earth elements like neodymium and dysprosium could meet these rigorous performance standards, forcing manufacturers to rely on a highly constrained global supply chain.[1][3][5]

While the United States does possess domestic rare-earth deposits—most notably the Mountain Pass mine in California—the domestic supply chain remains severely bottlenecked by a lack of processing infrastructure. Currently, over 95 percent of the raw minerals extracted domestically are exported to Asia for the complex, expensive, and environmentally hazardous refinement process. This near-total reliance on overseas processing has long been viewed by policymakers as a critical economic and national security vulnerability. A viable rare-earth-free alternative could simultaneously lower the production costs of green technology and secure the defense supply chain against potential geopolitical embargoes or trade disputes.[1][6]

The artificial intelligence model behind the discovery, dubbed DuctGPT, was not originally built with electric vehicle motors in mind. It was initially designed by researchers to discover highly resilient materials capable of surviving the extreme radiation and heat inside experimental fusion power plants. For fusion applications, materials must withstand unprecedented thermal stress while remaining ductile enough to be machined into workable parts without shattering. Recognizing the model's capability to solve complex material constraints, the Department of Energy pivoted the AI's focus toward the rare-earth magnet shortage under the umbrella of the Genesis Mission.[1][4]

The key advancement in DuctGPT is its architectural foundation. Unlike traditional machine learning models in materials science that simply search for hidden patterns within massive databases of known chemical compounds, DuctGPT operates as a physics-informed AI. It is trained directly on the fundamental laws of physics and the quantum behavior of electrons. Because the artificial intelligence natively understands the underlying science of how atoms interact and bond under varying conditions, it can theoretically invent entirely novel material structures from scratch rather than merely interpolating or tweaking existing compounds.[1][2][6]

Applying this physics-based reasoning to the permanent magnet problem, the AI directed researchers toward a specific composite of manganese and bismuth. While these abundant materials possess natural magnetic properties, they traditionally lose their magnetization far too easily to be useful in industrial motors. When subjected to the heat generated by a spinning motor, the magnetic alignment of the crystals breaks down. However, the AI's simulation suggested that altering the physical structure of the composite at the microscopic level could stabilize the magnetic field, prompting Ames scientists to develop a novel synthesis process based on the AI's exact parameters.[1][2][5]

To solve the degradation problem, the Ames Laboratory scientists successfully coated the individual microscopic crystals within the manganese-bismuth material with a specialized polymer. This polymer barrier acts as a microscopic shield, preventing the crystals from making direct physical contact with one another. In traditional composites, such physical contact causes a cascading loss of magnetization when the material is subjected to heat or external magnetic interference. By isolating the crystals, the polymer coating preserves the overall magnetic integrity of the material even under the extreme operating temperatures of an electric vehicle motor.[1][2]

The project's success serves as a primary validation of the Department of Energy's Genesis Mission, an initiative explicitly designed to leverage government supercomputing resources alongside academic expertise to drive breakthroughs in energy dominance and discovery science. By proving that AI can successfully design commercially viable materials that solve acute national security challenges, the DOE hopes to accelerate funding and adoption of physics-informed AI across other critical sectors, including battery chemistry and semiconductor manufacturing.[3][6]

While the laboratory synthesis of the manganese-bismuth composite is a proven scientific success, the next major hurdle is commercialization. Researchers are now working closely with industry partners and automotive manufacturers to ensure the new material can be manufactured at scale. The transition from a controlled laboratory environment to mass production requires adapting existing electric motor assembly lines to handle the new polymer-coated composite, a process that will determine how quickly these AI-designed magnets can begin replacing rare-earth components in consumer products.[5][6]