Insider Brief
- NJIT researchers used public funding and generative AI to identify new materials for multivalent-ion batteries, a next-generation alternative to lithium-ion systems.
- The team applied a dual-AI approach combining a crystal-generating model and a large language model to quickly identify five porous transition metal oxides with potential to host bulkier multivalent ions like magnesium, calcium, zinc, and aluminum.
- The study, published in Cell Reports Physical Science, addresses key scalability and sustainability hurdles for lithium alternatives and points to broader uses for AI in materials discovery across energy and electronics sectors.
Researchers at the New Jersey Institute of Technology report using artificial intelligence to identify new materials that could improve next-generation batteries and reduce reliance on lithium, a costly and limited resource.
The team at NJIT applied advanced generative AI tools to accelerate the search for alternatives to lithium-ion battery materials. The research, published in Cell Reports Physical Science, aimed at solving the scalability and sustainability challenges of lithium-dependent energy storage systems.
According to the researchers, the study focused on so-called multivalent-ion batteries, a promising but underdeveloped class of batteries that use elements such as magnesium, zinc, calcium, and aluminum. These elements are more abundant than lithium and carry multiple positive charges, giving them the potential to store more energy in a smaller space. However, their larger size and greater charge make it difficult for them to move efficiently within battery materials, a key technical hurdle.
“One of the biggest hurdles wasn’t a lack of promising battery chemistries — it was the sheer impossibility of testing millions of material combinations,” noted Professor Dibakar Datta, who ead the team. “We turned to generative AI as a fast, systematic way to sift through that vast landscape and spot the few structures that could truly make multivalent batteries practical.”
To address that problem, the NJIT team said it combined two different AI models. The first, a Crystal Diffusion Variational Autoencoder (CDVAE), was trained on a large dataset of known crystal structures and used to generate entirely new porous materials. These virtual materials featured the open internal spaces needed to accommodate bulkier multivalent ions. The second tool, a large language model (LLM), helped filter the results, selecting structures that were likely to be stable in real-world conditions.
The dual-AI system allowed the researchers to search through thousands of possible materials in a fraction of the time it would take using laboratory-based methods. The result was a shortlist of five porous transition metal oxides that simulations showed could serve as hosts for multivalent ions. The researchers used quantum mechanical simulations to validate the structures’ thermodynamic stability and ion mobility.
The findings could have wide-ranging implications for battery design. Multivalent-ion batteries have long been seen as a possible successor to lithium-ion systems, which dominate markets ranging from electric vehicles to grid storage but face long-term questions over mineral availability, cost, and environmental impact. By identifying stable, scalable materials for multivalent batteries, the NJIT study points to a faster path toward commercial alternatives.
The researchers noted that the tools developed in the study can be reused for other materials discovery efforts beyond batteries, including semiconductors and clean energy catalysts. They also emphasized the potential for public-private collaborations to translate their findings into working prototypes.
“This is more than just discovering new battery materials — it’s about establishing a rapid, scalable method to explore any advanced materials, from electronics to clean energy solutions, without extensive trial and error,” Datta said.
The NJIT team plans to continue refining its models and collaborating with experimentalists to validate and scale the most promising materials.
