Solid electrolytes are materials that can replace liquid electrolytes in batteries and make them safer and longer-lasting. However, a material that performs perfectly in the laboratory often fails in a real battery. This is precisely the problem being discussed by researchers led by Associate Professor Eric Jianfeng Cheng and Professor Hao Li from the Advanced Institute for Materials Research at Tohoku University (WPI-AIMR). Their review article has been published in the journal Advanced Materials.
For a long time, the main criterion was considered to be the speed at which ions - charged particles - move through a material. But, as the authors note, this alone is not enough. A practical solid electrolyte must also block electrons, maintain chemical stability, withstand mechanical loads, be manufacturable and work well with electrodes.
Instead of using artificial intelligence (AI) only to predict the properties of individual materials, the team proposes a connected research system. It combines continuously updated databases on how materials were created and tested, AI simulations that track ion movement over long timescales under more realistic conditions, and large language models (LLMs) - algorithms capable of extracting structured knowledge from the texts, figures and tables of scientific papers. Automated experiments then quickly test promising candidates, while humans manage the entire process.
The ultimate goal is a "closed loop": AI proposes candidates, physical models test them, experiments confirm the results, and new data are returned to the AI for the next round. "Finding a material with high ionic conductivity is only the first step," Cheng says. "The real task is to build a system in which data, simulations, experiments and researchers constantly learn from each other, so that the materials discovered are not only fast but also reliable and safe."
The authors emphasize that the approach is intended to support scientists, not replace them. Humans are still needed to ask the right questions, set safety limits, understand the causes of material behavior and interpret unexpected results. According to the researchers, this approach can move the search for solid electrolytes from scattered trial and error to a more reproducible and continuously improving process. In the future, competition may be not over who is first to find the fastest material, but over who builds the most reliable discovery system. Over time, this could accelerate the emergence of safer and longer-lasting batteries for electric vehicles, portable electronics and large-scale energy storage systems.