Researchers at the New Jersey Institute of Technology (NJIT) have achieved a significant breakthrough in the search for sustainable alternatives to lithium-ion batteries by harnessing the power of artificial intelligence. In a study published in Cell Reports Physical Science, the team led by Professor Dibakar Datta used advanced generative AI techniques to identify new porous materials ideally suited for use in multivalent-ion batteries — a potential game-changer for the future of energy storage.
Multivalent-ion batteries, which utilize abundant and cost-effective elements such as magnesium, calcium, aluminum, and zinc, present a promising alternative to lithium-based systems. These ions carry multiple positive charges (unlike lithium’s single charge), enabling the batteries to store substantially more energy. However, the bulky size and higher charge of multivalent ions have historically posed major challenges in designing suitable battery materials.
“While promising chemistries have existed for some time, the real bottleneck was the overwhelming number of possible material combinations,” explained Prof. Datta. “Manually testing them all would be virtually impossible — that’s where AI came in.”
To tackle this, the NJIT team devised a dual-AI strategy: a Crystal Diffusion Variational Autoencoder (CDVAE) and a custom-tuned Large Language Model (LLM). The CDVAE, trained on extensive databases of known crystal structures, could generate entirely new porous material blueprints, while the LLM was used to identify those structures most likely to be thermodynamically stable and practically synthesizable.
The approach led to the discovery of five novel porous transition metal oxide structures with large, open channels — ideal for efficiently transporting multivalent ions. These AI-generated materials were validated through quantum mechanical simulations and stability testing, confirming their feasibility for real-world applications.
“This is a breakthrough not just in battery research but in how we approach material discovery as a whole,” said Datta. “By using AI, we’ve created a powerful and scalable way to accelerate innovation across multiple fields, from clean energy to electronics.”
Moving forward, the team intends to collaborate with experimental labs to synthesize and further test these materials, potentially paving the way for more affordable and sustainable energy storage solutions.