Employees at New Jersey Institute of Technology (NJIT) used artificial intelligence to discover five new materials that could potentially form the basis for environmentally friendly and higher-capacity batteries. They used magnesium, calcium, aluminum, and zinc instead of lithium.
These batteries are called multivalent batteries. They are potentially safer and can store more energy due to their multiply charged ions. The main challenge with multivalent batteries is that large, heavy ions have difficulty penetrating the battery materials. Previously, searching for options was done manually, which took years.
Now, the researchers took a dual AI approach. A CDVAE autoencoder analyzed the crystal structure, and a large language model selected the most stable and promising compounds. As a result, the algorithm discovered five new porous transition metal oxides.
These materials are characterized by high permeability to multiply charged ions and could form the basis of future commercial batteries. Their properties were confirmed through quantum mechanical modeling and stability calculations. The use of AI has made it possible to identify and test promising compounds without decades of laboratory experiments.
