Machine Learning Revolutionizes Battery Design

In a recent study in the Journal of Materials Chemistry A, a research team led by Professor Shinichi Komaba, Ms. Saaya Sekine, and Dr. Tomooki Hosaka from Tokyo University of Science (TUS), in collaboration with Chalmers University of Technology and Professor Masanobu Nakayama from Nagoya Institute of Technology, employed machine learning to enhance the search for promising compositions.

Energy storage is critical to many rapidly expanding sustainable technologies, such as electric vehicles and renewable energy sources. Although lithium-ion batteries (LIBs) currently dominate the market, lithium is a limited and expensive resource, presenting economic and supply stability challenges. Consequently, researchers worldwide are exploring new types of batteries made from more abundant materials.

Sodium-ion (Na-ion) batteries, which use sodium ions as energy carriers, offer a promising alternative to LIBs due to the abundance of sodium, increased safety, and potential cost savings. In particular, sodium-containing transition-metal layered oxides (NaMeO₂) demonstrate remarkable energy density and capacity, making them effective materials for the positive electrode of Na-ion batteries.

However, identifying the optimal composition for multi-element layered oxides that incorporate various transition metals is both challenging and time-consuming, given the vast number of possible combinations. Even minor adjustments in the proportion and selection of transition metals can lead to significant changes in crystal morphology and battery performance.

The research team aimed to automate the screening of elemental compositions in various NaMeO₂ O₃-based materials. To achieve this, they first compiled a database of 100 samples from O₃-type sodium half-cells with 68 different compositions, collected by Komaba's group over an 11-year period.

The database included the composition of NaMeO₂ samples, with Me being a transition metal like Mn, Ti, Zn, Ni, Zn, Fe, and Sn, among others, as well as the upper and lower voltage limits of charge-discharge tests, initial discharge capacity, average discharge voltage, and capacity retention after 20 cycles.

Shinichi Komaba, Professor, Tokyo University of Science

To conduct an effective search, the researchers trained a model using this database, incorporating several machine learning algorithms alongside Bayesian optimization. The goal of this model was to predict the optimal ratio of elements needed to achieve a desired balance between properties such as operating voltage, capacity retention (lifetime), and energy density, as well as to understand how these properties relate to the composition of NaMeO₂ layered oxides.

After analyzing the results, the group determined that the model indicated the best composition for attaining the highest energy density—one of the most critical properties of electrode materials—would be Na[Mn₀.36Ni₀.44Ti₀.15Fe₀.05]O₂. They then created samples with this composition and assembled standard coin cells to perform charge-discharge tests, confirming the model’s predictive accuracy.

The model’s effectiveness and potential for investigating novel battery materials were demonstrated by the measured values, which largely aligned with the predicted outcomes.

Komaba added, “The approach established in our study offers an efficient method to identify promising compositions from a wide range of potential candidates. Moreover, this methodology is extendable to more complex material systems, such as quinary transition metal oxides.”

A developing trend in materials science is the application of machine learning to identify promising research directions, as it can significantly reduce the time and number of experiments required to screen new materials. The approach suggested in this study has the potential to accelerate the development of next-generation batteries, which could fundamentally transform energy storage technologies.

This transformation would impact not only consumer electronics, such as laptops and smartphones, but also the production of renewable energy and hybrid or electric vehicles. Moreover, effective machine learning applications in battery research can serve as a model for material development in other fields, potentially driving innovation across the broader materials science community.

Komaba concluded, “The number of experiments can be reduced by using machine learning, which brings us one step closer to speeding up and lowering the cost of materials development. Furthermore, as the performance of electrode materials for Na-ion batteries continues to improve, it is expected that high-capacity and long-life batteries will become available at lower cost in the future.”

Advantages and Challenges in Solid-State Sodium Battery Production

Video Credit: Tokyo University of Science

Hopefully, sodium-ion batteries that are economically feasible will soon be available!

This study is funded by JST-CREST, DX-GEM, and JST-GteX.

Journal Reference:

Sekine, S. et. al. (2024) Na[Mn_0.36Ni_0.44Ti_0.15Fe_0.05]O₂ predicted via machine learning for high energy Na-ion batteries. Journal of Materials Chemistry A. doi.org/10.1039/D4TA04809A