Google’s DeepMind, with the help of artificial intelligence, has calculated the composition of 2.2 million inorganic materials that should be stable. This greatly increases the number of possible materials for all kinds of products.
The researchers at DeepMind say in their Nature article that ‘graph neural networks’ have collected compositions for new materials based on 48,000 existing materials. As a result, there is now a collection of 421,000 materials, which is 381,000 more than there were before. The other materials compete for stability and therefore become unavailable to other scientists.
The aim is to find suitable structures among these materials that could improve batteries and solar panels, for example. Testing new materials can be done in an automated laboratory as A-Laballowing practical test results of promising materials to be obtained quickly.