Magnetic materials could now be developed faster than ever before, thanks to computer modelling techniques used to build two new types of magnets, atom-by-atom.

By using software to predict atom energy, stability, and other interactions inside a computer model, the researchers whittled down 236,115 potentially promising compounds to a shortlist of just 14 very quickly.

That's a huge improvement over the traditional trial-and-error methods currently used by scientists, according to the team from Duke University, and could lead to the rapid discovery of new magnets for all kinds of purposes, from medical devices to car engines.

"Predicting magnets is a heck of a job and their discovery is very rare," says one of the researchers, Stefano Curtarolo from the Centre for Materials Genomics at Duke. "Even with our screening process, it took years of work to synthesise our predictions."