Google DeepMind's AI Predicts 380,000 Stable Crystal Structures for Future Technologies

Collaboration with Lawrence Berkeley National Laboratory successfully synthesizes 41 of the predicted materials, paving the way for advancements in energy storage, solar cells, and superconductor chips.

Google DeepMind's AI model, GNoME, has predicted 2.2 million potential crystal structures, with 380,000 believed to be stable enough for use in future technologies.
The AI model was trained on data from 69,000 known crystals and can predict potential properties of new materials, aiding scientists in assessing their stability.
GNoME has generated 52,000 compounds with a similar structure to graphene, and 528 lithium ion conductors, as well as 15 lithium transition-metal oxides that could potentially be used in superconductors and rechargeable batteries.
In a collaboration with the Lawrence Berkeley National Laboratory, 58 of GNoME's predicted materials were selected for synthesis by a robotic arm, with 41 successfully produced.
The breakthrough could accelerate the discovery of new materials for applications such as energy storage, solar cells, and superconductor chips.

One America News Network

Nature