Researchers from New Jersey Institute of Technology (NJIT) used artificial intelligence (AI) to look at multivalent-ion batteries as an alternative to lithium-ion batteries.
These batteries use elements like magnesium, calcium, aluminium and zinc that are more abundant in comparison to lithium.
Lithium-ion batteries tend to use ions that carry a single positive charge, but the researchers said that multivalent-ion batteries would use elements that can carry two or three charges.
It would mean that these batteries could store more energy – being ideal for energy storage solutions.
The research paper, published in the Cell Reports Physical Science journal, used generative AI to tackle a key challenge with these batteries.
The main challenge is that the increased size and greater electrical charge would make it difficult to be put efficiently into battery materials.
Professor Dibakar Datta, who led the NJIT team, said: “One of the biggest hurdles wasn’t a lack of promising battery chemistries — it was the sheer impossibility of testing millions of material combinations. 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.”
It used a dual-AI approach via a large language model (LLM) and a crystal diffusion variational auto-encoder (CDVAE). These AI tools were used to quickly explore thousands of new crystal structures that could be used.
The CDVAE model was trained on datasets of known crystal structures, which the researchers claim allowed it to propose completely novel materials with diverse structural possibilities.
The LLM was used to focus in on materials closest to thermodynamic stability, which is crucial for practical synthesis for potential multivalent-ion batteries.
The NJIT team uncovered five new porous transition metal oxide structures that Professor Datta said show promise.
“These materials have large, open channels ideal for moving these bulky multivalent ions quickly and safely, a critical breakthrough for next-generation batteries,” he said.
The researchers validated the AI-generated structures using quantum mechanical simulations and stability tests, which confirmed the materials could be synthesised experimentally.
Image: The open, sponge‑like network inside a porous transition‑metal oxide, which could be used in a multivalent-ion battery. Credit: NJIT.
