MIT teaches AI to 'see' crystals in 3D — and its battery picks match the experts'

Finding better materials — for batteries, solar panels, chips — usually means years of trial and error. MIT researchers led by Ju Li and Elsa Olivetti, with first author Yang Jeong Park, have built an AI model that studies materials the way a person might: it looks at a crystal's 3D structure and reads plain-language descriptions of it at the same time. Published in Nature Communications, the model was asked for materials suited to solid-state lithium-ion battery electrolytes — the safer, longer-lasting kind of battery — and its top picks matched candidates scientists had independently identified as promising. The team compares the leap to what AlphaFold did for biology: a faster way to search the material world.

SDG 9 · Industry, innovation and infrastructure — Mechanism: tech innovation