Could our smartphones and electric cars one day do without rare earths, which are widely used in the design of batteries and electric motors?
A team from the University of New Hampshire offers an encouraging answer by harnessing artificial intelligence to examine a vast array of magnetic materials that have been little studied so far.
To achieve this, the researchers trained an artificial intelligence system to read and interpret decades of scientific literature. The tool is capable of extracting experimental details to determine whether a material has magnetic properties and what temperature it can withstand. All this information is then structured within an open database, named the Northeast Materials Database.
This strategy thus allows for the rapid identification of promising candidates, without the need for systematic laboratory testing.
Currently, the most efficient permanent magnets rely heavily on rare earths. These elements are primarily imported, leading to difficulties related to their price and availability. Although many magnetic compounds are known, none have so far succeeded in replacing rare-earth-based magnets in consumer applications.
The database already contains more than sixty-seven thousand magnetic materials, among which twenty-five original compounds retain their properties at high temperatures. According to Suman Itani, a physics doctoral student and lead author of the study, this acceleration in material identification could help reduce dependence on rare earths, lower the price of electric vehicles, and strengthen the industrial base.
Professor Jiadong Zang, co-author of the work, indicates that this method addresses one of the main challenges in materials science: identifying sustainable alternatives to permanent magnets. The team is convinced of the potential of its database and artificial intelligence technologies to make this goal achievable.