As the clean transition leads to widespread use of electric vehicles and energy storage in power grids, and an ever-increasing reliance on variable renewable energy sources such as wind and solar, the risk of battery fires is also increasing. To improve battery performance while limiting this risk, next-generation batteries will likely rely on novel solid-state electrolytes, but research is hampered by the large number of material options and associated parameters.
But machine learning is coming to the rescue: A group of materials scientists has developed a new dynamic database of hundreds of solid-state electrolytes and is applying artificial intelligence techniques to it to already guide their research for the better.
A paper describing their approach has been published in the journal Nanomaterials Science September 10, 2023.
Organic solvents are commonly used as electrolytes (usually liquid or gel-like substances that facilitate the movement of charged particles or ions between positive and negative electrodes) in many rechargeable batteries. This type of solvent is highly conductive, allowing ions to move between the electrodes efficiently, but various safety and performance concerns have led battery researchers to long search for alternative electrolyte materials.
In particular, organic solvents are flammable and can cause thermal runaway reactions, resulting in fires and explosions. Organic solvents are also prone to chemical decomposition, which over time can lead to the evolution of gases and electrolyte decomposition, reducing the performance and lifespan of the battery. Additionally, batteries may only operate over a limited voltage range.
One alternative is all-solid-state batteries (ASSBs), which replace traditional liquid or gel organic solvents with solid electrolytes, eliminating issues of leakage and explosion. These solid electrolytes not only improve safety, but also potentially increase energy density and reduce charging times.
However, the road to finding a solid electrolyte (SSE) with high ionic conductivity (the ability of ions to move through the battery and generate electric current) is fraught with challenges, mainly due to their complex structures and their structure-performance relationships. So far, only SSEs with slow ion transport have been identified. Without high-performance SSEs, the development of ASSBs has been greatly hindered.
“To make matters worse, there are so many SSEs to choose from,” says Hao Li, a materials scientist at the Advanced Institute for Materials Research at Tohoku University and corresponding author of the paper. “There are hundreds of possibilities, and it's really challenging for researchers to wade through so many options while keeping track of the different parameters for optimal performance.”
To explore the relationships between the various variables, the team developed an experimental dynamic database, the Dynamic Database of Solid Electrolytes (DDSE), that contains more than 600 candidate solid electrolyte materials spanning a wide range of operating temperatures and covering a range of cations and anions (positive and negative ions).
A dynamic database is a type of database that is designed to be easily updated and frequently modified, allowing changes and additions to be made in real time to the data contained in the database. This type of database is often used in situations where information is constantly evolving, in which case the DDSE is continually updated with new experimental data. The database is updated weekly and contains over 1,000 documents as of January 2024.
Researchers then applied machine learning to DDSE to overcome the limitations of human analysis and the enormous computational cost of theoretical calculations. Without machine learning, researchers struggled to computationally unravel the complexities of SSE's large atomic systems and associated chemical reactions.
By leveraging machine learning, researchers can make better predictions about new solid electrolyte materials with minimal time waste and at a much lower computational cost (and expense) compared to the traditional trial-and-error approach in SSE design.
In doing so, researchers have begun to unravel the complex relationships between several different variables, including ion transport, composition, activation energy (the amount of energy needed to start a chemical reaction), and electrical conductivity, allowing for the development of a new set of guidelines for the design of SSEs. Researchers have already identified trends in the development and performance of SSEs across different classes of materials, as well as performance bottlenecks for each class of SSE.
DDSE was designed with a user-friendly interface so that battery and materials scientists outside the original team could update and use it themselves.
About the World Premier International Research Center Initiative (WPI)
The World Premier International Research Center Initiative was launched by the Ministry of Education, Culture, Sports, Science and Technology in 2007 to foster world-class research environments and centers of excellence. These research centers, run by more than 10 research institutions across the country, are given a high degree of autonomy to pursue innovative operations and research. The program is run by the Japan Society for the Promotion of Science (JSPS).
For the latest research news from each center, visit the WPI News Portal: https://www.eurekalert.org/newsportal/WPI
Main WPI program site: www.jsps.go.jp/english/e-toplevel
Advanced Institute for Materials Research (AIMR)
Tohoku University
Establishing a world-leading materials science research center
AIMR brings together outstanding researchers in the fields of physics, chemistry, materials science, engineering, and mathematics, and has created a world-class research environment. As a world-leading research center in materials science, the AIMR aims to push the boundaries of research and contribute to society.
Disclaimer: Neither AAAS nor EurekAlert! are responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for any use of information provided through the EurekAlert system.
