Professor Duho Kim and his research team at the Department of Mechanical Engineering presents an innovative solution to improve the capacity of next-generation sodium-ion batteries

As electric vehicles become the main focus of sustainable, eco-friendly next-generation transportation, the search for better and cheaper rechargeable batteries is also intensifying. While the current mainstay is lithium-ion batteries, research on sodium-ion batteries is actively underway to make rechargeable batteries more economical by replacing lithium, the expensive rare-earth material, with sodium, one of the most abundant elements on Earth. But sodium-ion batteries have not yet reached the point of commercial viability due to the capacity erosion after charging and issues of performance instability.

Professor Duho Kim's research team at the Department of Mechanical Engineering proposed a new method to improve the performance of sodium-ion batteries using machine learning. This research was well-received for its excellence and published in the March 15, 2024, issue of the renowned international academic journal Advanced Energy Materials (IF=29.698) under the title, “Data-Driven Decoupling Structural Feature Correlation for Harnessing Anionic Capacity in Na Layered Oxide.”

Professor Kim's research team overcame the limitations of sodium-ion batteries with a new approach that combines machine learning and computational science. Student Jongbeom Kim (3rd year master’s, Mechanical Engineering), the first author of the study, explained the research and said, “While computational science can save the cost and time of experiments by running virtual predictive models, there are limitations in processing large amounts of data. We incorporated machine learning to sidestep the problem with good results, so we will continue to look into the use of machine learning to augment computational science.”

Professor Kim explained, “We identified the main characteristics of sodium-ion batteries at each scale through a systematic data-driven analysis approach and found a way to increase the possibility of commercialization of high energy density sodium-ion batteries.” The research team diagnosed structural problems based on data on the reaction mechanism of the sodium-ion battery anode model.

As a result of the formation energy analysis, it was discovered that the formation energy of the sodium-ion battery should be linear and have a high slope. The team also used Pearson correlation coefficient and machine learning to find key factors that affect formation energy. Afterwards, the ideal charging and discharging mechanism was analyzed based on the main features proposed by the research team.

Although it presented a new convergent approach combining computational science and machine learning, the actual implementation process was particularly challenging in the sense that the team, with background primarily in mechanical engineering, had to learn machine learning to design the research model properly. Student Kim had to attend lectures and classes in other departments to improve his proficiency in machine learning while working on the research full-time. He said, “I read many research articles to analyze cases that applied machine learning while going through a lot of trial and error in my own experiments.”

All the hard work led to spectacular results. In the submission review, the article received high praise for systematically classifying batteries and the effective application of machine learning. Student Kim said, “I enjoyed the process of reading papers and finding answers, so I went on to graduate school after working as an undergraduate research student. I was able to feel that I was making progress thanks to Professor Kim’s detailed guidance.” On his future goals, Student Kim said, “I would like to look into whether the methodology I have developed in this paper could be applied to other metals.”