AI speeds up discovery of 'new' materials as lithium-ion alternatives

Researchers at the New Jersey Institute of Technology (NJIT) have leveraged artificial intelligence to accelerate the discovery of new battery materials that could serve as safer, cheaper, and more sustainable alternatives to lithium-ion technology. Using generative AI models, specifically a Crystal Diffusion Variational Autoencoder (CDVAE) combined with a fine-tuned large language model (LLM), the team rapidly explored thousands of potential porous crystal structures. These structures are designed to facilitate the movement of multivalent ions—such as magnesium, calcium, aluminum, and zinc—that carry multiple positive charges, offering higher energy density than lithium ions. The AI-driven approach overcame the traditional bottleneck of experimentally testing millions of material combinations, enabling the identification of five novel porous transition metal oxide materials with large channels ideal for fast and safe ion transport. The researchers validated the AI-generated materials through quantum mechanical simulations and thermodynamic stability assessments, confirming their practical synthesizability and promising performance for energy storage applications. This breakthrough not only advances the development of