Catenaa, Saturday, February 14, 2026- Researchers at the University of New Hampshire have used artificial intelligence to identify 25 previously unknown compounds that remain stable at high temperatures

The research also identified tens of thousands of magnetic materials, a development that could reduce reliance on rare earth elements.

The team said its AI-driven approach produced a searchable database of 67,573 magnetic materials after scanning decades of published scientific literature.

The newly identified compounds meet a core requirement for industrial use, retaining magnetic properties under heat stress common in power systems, transportation and electronics.

The database, known as the Northeast Materials Database, is designed to speed up materials discovery by allowing researchers to filter candidates by composition, magnetic behavior and thermal stability.

Permanent magnets are essential to devices ranging from electric vehicles and wind turbines to medical imaging systems, yet most high-performance magnets still depend on rare earth elements that are costly and largely imported.

According to the researchers, traditional laboratory testing cannot realistically cover the millions of possible element combinations that may form useful magnets.

The AI system was trained to read and extract experimental data from scientific papers, then combine those findings with computational models to assess whether materials exhibit magnetism and how much heat they can withstand.

The study was published in Nature Communications in October.

The research team said the method could shorten discovery timelines from years to weeks and help lower costs tied to clean energy technologies and advanced manufacturing.

Beyond magnet research, the scientists said the underlying AI tools could be adapted for broader academic use, including converting older scientific records into modern digital formats. The work was supported by the US Department of Energy.