AI cuts screening time for new materials, MIT says

Industries currently expend significant computational resources screening out unstable material designs generated by artificial intelligence, often yielding only a small fraction of usable options. Now, MIT researchers have developed a framework called “crystal generator with valence-constrained design,” or CrysVCD, to improve material stability early in the design process. “You can plug this into any model, not only existing diffusion models but also future models, where generating enough stable materials is difficult, and it can improve stability,” says Mingda Li, associate professor of nuclear science and engineering.…