The power that allows electric cars to travel long distances and extend the lifespan of smartphones comes from battery materials. Among these, the core material that directly determines the performance and lifespan of a battery is the cathode material. What if artificial intelligence could replace the numerous experiments required to develop battery materials? A KAIST research team has developed an artificial intelligence (AI) framework that provides both the particle size and reliability of predictions for cathode materials even in situations where experimental data is insufficient, opening the possibility of expansion to next-generation energy technologies such as solid-state batteries.

KAIST announced on January 26 that a research team led by Professor Hong Seung-beom of the Department of Materials Science and Engineering, in collaboration with Professor Cho Eun-ae’s team, has developed a machine learning framework that accurately predicts the particle size of battery cathode materials even when experimental data is incomplete and provides reliable results.

The cathode material inside the battery is the core material that allows lithium-ion batteries to store and use energy. Currently, the most widely used cathode materials in electric vehicle batteries are NCM-based metal oxides with a mixture of nickel (Ni), cobalt (Co), and manganese (Mn), which have a significant impact on battery life, charging speed, range, and safety.

The KAIST research team focused on the fact that the size of the very small primary particles that make up these cathode materials is a key factor in determining battery performance. This is because particles that are too large will reduce performance, while particles that are too small may cause problems with stability. Therefore, the research team developed an AI-based technology that can accurately predict and control particle size.

Previously, determining particle size required many repeated experiments with varying sintering temperatures, times, and material compositions. However, in actual research settings, it is difficult to measure all conditions without exception, and experimental data is often missing, which limits the ability to accurately analyze the relationship between process conditions and particle size.

To solve this problem, the research team designed an AI framework that imputed missing data and presented predictive results with confidence. This framework is characterized by a combination of a technology that imputes missing experimental data by considering chemical properties (MatImpute) and a probabilistic machine learning model that calculates prediction uncertainty (NGBoost).

This AI model not only predicts particle size, but also provides information about how reliable the predictions are. This is an important criterion for deciding under what conditions the material will actually be synthesized.

As a result of learning from expanded experimental data, the AI ​​model showed a high prediction accuracy of approximately 86.6%. Our analysis reveals that the particle size of the cathode material is more influenced by process conditions such as baking temperature and time than by material composition, which is in good agreement with existing experimental understanding.

In order to verify the reliability of the AI ​​predictions, the research team conducted experiments by creating four new cathode material samples synthesized under manufacturing conditions not included in existing data, while keeping the metal composition of NCM811 the same (Ni 80%/Co 10%/Mn 10%). As a result, it was found that the particle diameter predicted by AI almost matched the actual microscopic measurement results, with most errors being less than 0.13 micrometers (μm), which is much smaller than the thickness of a human hair. In particular, the actual experimental results were within the range of prediction uncertainty presented by the AI, confirming that not only the predicted values ​​but also their reliability were reasonable.

This work is important because it opens the way for battery research to first find conditions with a high probability of success without having to perform every experiment. This is expected to accelerate the development of battery materials and significantly reduce wasteful experiments and costs.

Professor Hong Seung-beom said, “It is important for AI to present not only predicted values, but also how reliable the results are,” and “This will be of practical use in designing next-generation battery materials more quickly and efficiently.”

Benediktus Madika, a doctoral student in the Department of Materials Science and Engineering, participated as the first author in this study, which was published in Advanced Science, an internationally authoritative academic journal in the fields of materials science and chemical engineering, on October 8, 2025.

* Paper title: Uncertainty-Quantified Primary Particle Size Prediction in Li-Rich NCM Materials via Machine Learning and Chemistry-Aware Imputation, DOI: https://doi.org/10.1002/advs.202515694

Meanwhile, this study was conducted by a research team of researchers Benediktus Madika, Chaeyul Kang, JooSung Shim, Taemin Park, and Jung Hyeon Moon, as well as Professor EunAe Cho and Professor Seungbum Hon, and was supported by the Ministry of Science, Information and Communications Technology (MSIT), National Research Foundation of Korea (NRF), and Future Fusion Technology Pioneer (Strategic). RS-2023-00247245).