Russian scientists from Skoltech and other research institutes have developed an innovative artificial intelligence system capable of significantly accelerating the process of creating solid-state batteries. The neural network, based on graph neural networks, allows identifying energy-intensive and fire-safe electrolytes, as well as selecting optimal protective coatings for electrodes.
Solid-state batteries are a promising alternative to traditional lithium-ion batteries, as they can increase the range of electric vehicles by about 50% and improve fire safety. However, their widespread use is limited due to the shortcomings of existing solid electrolytes, which either conduct ions poorly or quickly degrade when interacting with electrode materials.
The development of a new AI system can significantly speed up the process of searching and testing new materials. Scientists have prepared an extensive database, including information on the structure of various materials and the results of accurate quantum chemical calculations. This allows the neural network to quickly find solutions that previously required lengthy and laborious calculations.
As a test, the system was applied to the promising solid-state electrolyte Li10GeP2S12, consisting of lithium, germanium, phosphorus, and sulfur. The results showed that the neural network was able to identify several new protective materials that can significantly improve the durability and safety of solid-state batteries.
Read more on the topic:
Scientists trained a neural network to sing the songs of sperm whales
Named errors that prevent the use of neural networks to the maximum