Researchers at Ames National Laboratory are combining fundamental physics with artificial intelligence to discover permanent magnets that don't require rare earth elements, addressing a critical U.S. supply chain vulnerability. For over two decades, scientists have pursued rare-earth-free alternatives through trial-and-error experimentation. Now, a new physics-informed AI approach is accelerating discovery by predicting material behavior before anything is synthesized in the lab.

Why Does America Need Rare-Earth-Free Magnets?

Rare earth elements are essential to today's high-performance magnets, giving them exceptional strength and resistance to demagnetization. These properties are critical for energy generation and defense applications. However, rare earths are expensive and often sourced from unstable international suppliers outside the United States. Reducing or eliminating dependence on these elements would lower costs and strengthen domestic production capacity.

The challenge has been significant. Traditional approaches relied on researchers synthesizing materials in the lab, testing them, and building knowledge one data point at a time. This process is slow, expensive, and limits the scope of what scientists can explore.

How Does Physics-Informed AI Speed Up Materials Discovery?

Ames Lab scientist Prashant Singh has outlined a faster, more systematic approach that combines physics-based modeling, high-throughput simulations, and reasoning-based AI tools to guide discovery before materials are made. The key innovation is embedding physics knowledge directly into computational models, allowing researchers to understand how a material's atomic structure and electronic behavior determine critical magnetic properties.

This approach focuses on predicting several essential characteristics:

• Magnetization Strength: How powerful the magnet will be under various conditions
• Energy Storage Capacity: How much magnetic energy the material can hold and release
• Resistance to Demagnetization: How well the magnet maintains its strength over time and temperature
• High-Temperature Performance: How the material behaves when exposed to heat, critical for real-world applications

By predicting these properties computationally, researchers can identify the most promising material candidates and dramatically reduce the need for costly experimental iteration.

"Ames Lab's strength comes from its deep expertise and a long history of data in the magnet space that no other institution has. That's what makes our role critical. In any material design problem, you need to know how combining two elements will change their performance before you ever run an experiment," said Prashant Singh, Ames Lab Scientist.

Prashant Singh, Ames Lab Scientist

What Makes This AI Approach Different From Generic Machine Learning?

A critical distinction sets this work apart from standard AI applications. Rather than training models on generalized datasets, Ames Lab researchers are training AI on experimentally measured and scientifically calculated material properties. This ensures predictions remain grounded in real-world behavior, not just statistical patterns.

Singh emphasized the importance of understanding physics when designing materials. "If you just use the data to train your models, you are going to get only the predictions within the range of information you have," he explained. "But once you understand the physics of what controls specific properties, then you and your agentic tools or AI frameworks can search arbitrary material space." This distinction allows researchers to explore materials that have never been tested before, expanding the frontier of discovery.

Singh

The team is also developing new AI-assisted tools that allow researchers to interact with models more directly. These emerging capabilities, including a tool called DuctGPT, enable scientists to pose design questions, refine requirements, and explore potential materials more efficiently.

How Can AI Account for Real-World Manufacturing Constraints?

One often-overlooked advantage of AI-driven discovery is its ability to factor in supply chain realities. Material availability, costs, and market conditions shift constantly, and traditional research methods rarely account for these practical constraints. By incorporating supply chain data into the discovery process, Ames Lab researchers can identify materials that are not only high-performing but also practical to produce and scale.

"Supply chain conditions shift by the hour, material costs fluctuate, availability changes daily, and the market never stands still. By factoring those conditions into the discovery process, we can better target materials that are not only high-performing but also practical to produce and scale," noted Prashant Singh.

Prashant Singh, Ames Lab Scientist

This holistic approach addresses the full critical materials pipeline, from accelerating discovery to ensuring industrial viability. The result could reduce America's reliance on outside sources for critical materials and build a more resilient domestic supply chain.

What Competitive Advantages Does Ames Lab Bring to This Challenge?

Ames National Laboratory is uniquely positioned to lead this effort. The institution has seven decades of work in critical materials, strong foundations in theory and simulation, and proprietary magnetic material data exclusive to Ames. No other laboratory has accumulated this depth of knowledge in the magnet space.

By combining these long-standing strengths with emerging AI capabilities, researchers hope to dramatically expand the pace and scope of magnetic materials innovation. This work is part of the U.S. Department of Energy's Genesis Mission, which unites DOE National Labs, industry, and academia to harness AI for scientific breakthroughs in energy, discovery science, and national security.

The implications extend beyond magnets. Similar physics-informed AI approaches are being explored globally for materials discovery. At Nanyang Technological University in Singapore, researchers are developing multimodal AI systems that bridge computational, synthesis, and characterization data to build richer representations of materials. These techniques are being applied to identify compounds essential for next-generation clean energy technologies.

Meanwhile, South Korea's Ministry of Science and ICT is investing 22.5 billion Korean won over four years in AI-driven research across six fields, including materials and chemistry. A team at Seoul National University is developing AI models that predict the complex properties of next-generation polymer and electronic materials, or design new materials optimized for target performance, thereby reducing development time and cost.

The convergence of physics-informed AI, decades of accumulated materials data, and global investment in AI-driven discovery signals a fundamental shift in how scientists approach materials innovation. Rather than waiting for serendipitous breakthroughs, researchers can now systematically explore vast material spaces, constrained by physics and practical manufacturing realities. For critical applications like magnets, this acceleration could reshape supply chains and accelerate the transition to clean energy technologies.