TLDR: Researchers at the San Diego Supercomputer Center (SDSC), in collaboration with the New Jersey Institute of Technology (NJIT) and Rensselaer Polytechnic Institute (RPI), are leveraging generative AI and advanced simulations on the Expanse supercomputer to discover new battery materials. This initiative aims to replace scarce lithium with abundant metals like magnesium, zinc, and aluminum, identifying over 40 unique structures suitable for multivalent-ion batteries. The breakthrough promises to revolutionize energy storage for electric vehicles and electrical grids, offering more sustainable, cost-effective, and safer alternatives.
As the global demand for electric vehicles and renewable energy storage continues to surge, the critical reliance on lithium, a scarce and environmentally intensive metal, presents a significant challenge. Scientists at the San Diego Supercomputer Center (SDSC), part of the University of California San Diego, are spearheading a groundbreaking initiative to address this issue by developing next-generation batteries that utilize abundant metals such as magnesium, zinc, and aluminum.
This ambitious project employs advanced simulations on SDSC’s Expanse supercomputer, powered by generative AI models, to engineer a sustainable solution. The research, a collaborative effort with teams from the New Jersey Institute of Technology (NJIT) and Rensselaer Polytechnic Institute (RPI), focuses on identifying optimal electrode materials for multivalent-ion batteries, specifically transition metal oxides (TMOs).
Dibakar Datta, an associate professor of mechanical engineering at NJIT, highlighted the potential of TMOs, stating, “TMOs have a great deal of potential as host materials in multivalent ion batteries. They offer structural versatility, high ionic conductivity and an ability to accommodate multiple charge carriers. However, their vast compositional and structural diversity makes traditional exploration inefficient — it’s a classic ‘needle in a haystack’ problem. Thankfully, generative AI-powered by Expanse at SDSC has a knack for helping us pick through haystacks.”
The research team utilized a suite of AI models, including crystal diffusion variational autoencoders (CDVAEs), large language models (LLMs), and the atomistic graph neural network (ALIGNN). Initially, the CDVAE and LLM models generated an impressive 10,000 potential structures each. Following this, the ALIGNN model was employed to predict crucial material properties, such as thermodynamic and electronic stability, effectively narrowing down the selection. Through this AI-driven process, the researchers successfully identified over 40 unique structures that meet the necessary benchmarks for use as host materials in these advanced batteries.
Also Read:
- MIT and Kookmin University Unveil FlowER: A Generative AI Revolutionizing Chemical Reaction Prediction with Physical Constraints
- University of Pennsylvania Engineers Unveil Revolutionary AI for De Novo Antibiotic Creation
This innovative approach, supported by U.S. National Science Foundation (NSF) ACCESS allocations on Expanse, promises to deliver batteries that are not only more sustainable due to their reliance on common elements but also potentially safer, more stable, and more cost-effective than current lithium-ion technology. The implications are far-reaching, with the potential to revolutionize energy storage for everything from smartphones to large-scale electrical grids, mitigating the environmental and geopolitical constraints associated with lithium mining.
