Signature line: michael matz
Newswise — Argonne researchers outline a comprehensive technology roadmap for the use of large-scale language models in battery research.
Scientists predict that batteries will play a central role in making the U.S. energy system safer and more cost-effective. However, achieving this will require solving many technical challenges, including designing high-performance batteries and understanding how batteries degrade, which is no easy task.
Can artificial intelligence (AI) help overcome these challenges? A team of researchers at the U.S. Department of Energy’s (DOE) Argonne National Laboratory has outlined an ambitious technology roadmap to accelerate battery breakthroughs using an AI tool known as large-scale language models (LLMs).
The Institute is also advancing AI for research through DOE’s Genesis Mission, a historic national effort to transform American science and innovation through the power of AI and strengthen the nation’s technological leadership and global competitiveness.
“At Argonne, we offer a rare combination of leading battery researchers and data scientists working under the same roof,” said Khalil Amin, Argonne Distinguished Fellow and leader of Argonne’s battery technology development group. “Together, they conducted a comprehensive review of emerging LLM applications in the battery field. This review presents short- and long-term goals to leverage the huge potential of LLM to revolutionize battery research.”
“Soon, most battery scientists will also be AI experts, and LLMs will serve as their wise research assistants. Scientists will spend significantly less time reading papers, scrutinizing data, and conducting experiments. This will allow them to spend more time developing ideas and strategic research plans.” — Wenhua Zuo, Argonne Postdoctoral Fellow
LLM Agent and Autonomous Driving Institute
LLM is an extremely versatile machine learning tool. It can be trained on large and diverse datasets to perform a wide range of tasks. For example, it can answer questions, write and summarize articles, and translate languages in response to human prompts. LLMs like OpenAI’s ChatGPT and Google’s Gemini are already transforming the way industries work.
Although the use of LLM in academia, industry, and research institutions is increasing, its potential is largely untapped in battery research. In our review of Argonne, we discuss potential LLM applications and explore ways to expand its use and improve its effectiveness.
Some application examples: LLM can text mine hundreds to millions of battery research papers to extract important insights, identify knowledge gaps, and suggest new research directions. Battery performance datasets can be analyzed to identify failure mechanisms and mitigation strategies. It can also monitor and optimize battery operation in the field and provide personalized training to early battery researchers.
The Argonne review clearly articulates a vision for coordinating activities between multiple LLM agents. These advanced AI systems use LLM to analyze information, make decisions, and use tools. Agents specializing in different subject areas of the battery perform different research tasks while working together to achieve a common goal.
“LLM can be integrated with existing battery research tools, such as simulation software and material property databases,” said Guiliang Xu, corresponding author of the review and a chemist at Argonne. “This will help scientists build an AI-powered self-driving laboratory that accelerates the research process through automation.”
Self-driving research labs could have a major impact on the discovery of new battery materials. Traditionally, this type of research has been conducted through a process of trial and error. Scientists manually test one material or synthesis parameter at a time.
Self-driving laboratories can speed up this process by continuously running iterative experimental cycles. It will review the literature, screen databases of material properties, and propose promising new battery chemistries. It then instructs the robotic equipment to manufacture and characterize the material. Next, analyze the experimental data and use the results to refine your hypotheses, methods, and experiments.
The benefits go beyond speed and efficiency. Automating research reduces errors and increases reproducibility of experiments.
Successful implementation of these AI systems will require extensive collaboration between battery researchers and LLM experts.
“Battery researchers can inform LLM experts about their most important research questions, and LLM experts can inform battery researchers about the most appropriate models and techniques to address those questions,” Xu said.
Knowledge base, data sharing, adaptability
This paper points out the technical challenges that need to be addressed before the full potential of LLM can be realized. When selecting an existing LLM or developing a new one, researchers should carefully consider its computational efficiency and adaptability to a particular battery research task. For critical applications such as predicting battery failure, it is important to use LLMs that can explain the step-by-step reasoning behind the conclusions. Protocols need to be developed for effective collaboration between LLM agents.
Training an LLM requires a high-quality knowledge base. It is built from existing published literature and various battery datasets. Significant work is required to standardize dataset formats, and consortia are required to share datasets across industry and research communities.
“Traditionally, researchers only publish data about successful results,” said Huihuo Zheng, one of the review’s authors and a computer scientist at the Argonne Leadership Computing Facility (ALCF), a DOE Office of Science user facility. “For LLM to function optimally, it must also be trained on failure data, such as battery materials with poor experimental performance. Industry and academia must implement new methods to make this data easily accessible.”
Battery and data scientists should regularly evaluate the functionality, performance, and usability of LLMs and make improvements as needed. This may include activities such as testing the LLM’s data interpretation abilities and retraining with newly published literature.
If these challenges are resolved, what will the future of battery research look like?
“Soon, most battery scientists will also be AI experts, and LLMs will serve as their smart research assistants,” said Wenhua Zuo, one of the review authors and an Argonne postdoctoral fellow. “Scientists will spend significantly less time reading papers, sifting through data, and conducting experiments. This will allow them to spend more time developing ideas and strategic research plans.”
This review was published in the August 20, 2025 issue of Joule.
Other Argonne contributors to this review include Tanjin He, Venkatram Vishwanath, Maria Chan, and Rick Stevens.
This review was supported by DOE’s Office of Transportation Technology through the Advanced Battery Materials Research Program, which includes the Earth-Abundant Low-Cost Na Ion Storage Consortium. It uses ALCF resources and is based on research supported by the DOE Office of Science’s Advanced Scientific Computing Research Program.
This research also received funding in part from the Energy Storage Research Alliance, an energy innovation hub funded by the DOE Office of Science Basic Energy Sciences (BES). Research conducted at the Center for Nanoscale Materials, a DOE Office of Science user facility, also received support from BES.
michael matz He is a freelance science and technology writer with over 30 years of experience covering topics such as energy, environment, physics, AI, quantum information, and materials. He has contributed to a wide range of publications, translating complex ideas into compelling stories.
About Argonne’s Center for Nanoscale Materials
The Nanoscale Materials Center is one of DOE’s five nanoscale science research centers and is the nation’s primary user facility for interdisciplinary research at the nanoscale supported by the DOE Office of Science. Together, the NSRC constitutes a set of complementary facilities that provide researchers with state-of-the-art capabilities to manufacture, process, characterize, and model nanoscale materials and constitute the National Nanotechnology Initiative’s largest infrastructure investment. NSRC is located at DOE’s Argonne, Brookhaven, Lawrence Berkeley, Oak Ridge, Sandia, and Los Alamos national laboratories. For more information about the DOE NSRC, please visit https://sci .osti .gov/U ser – F acilities / U ser – F acilities – at – a – Glance.
Argonne Leadership Computing Facility It provides supercomputing capabilities to the scientific and engineering community to advance fundamental discovery and understanding in a wide range of fields. Supported by the U.S. Department of Energy (DOE) Office of Science’s Advanced Scientific Computing Research (ASCR) program, ALCF is one of two DOE Leadership Computing Facilities in the nation dedicated to open science.
Argonne National Laboratory We seek solutions to pressing national problems in science and technology by conducting cutting-edge basic and applied research in virtually every scientific field. Argonne is managed by UChicago Argonne, LLC of the U.S. Department of Energy’s Office of Science.
U.S. Department of Energy Office of Science is the largest supporter of basic research in the physical sciences in the United States and is working to address some of the most pressing challenges of our time. For more information, please visit https:// ener gy .gov/s science.
