Artificial intelligence aids in the creation of innovative battery design

As more devices rely on batteries for power, the search for innovative, safer, and more cost-effective materials intensifies. Historically, this quest involved extensive trial-and-error experimentation in the laboratory. However, recent research suggests that artificial intelligence (AI) could accelerate this process, potentially unlocking new battery materials tailored to specific requirements.

Initiated by a team of 11 researchers based in Washington, including individuals from Microsoft in Redmond and the Energy Department’s Pacific Northwest National Laboratory (PNNL) in Richland, the project commenced with a vast selection of potential battery materials, totaling over 32 million candidates. Leveraging AI technology, the team efficiently narrowed down this extensive list to 23 promising options, ultimately choosing one material to construct a functional battery.

While previous efforts have utilized AI to predict material behavior, achieving the actual creation of a new material has been rare. This study stands out for its comprehensive approach, as highlighted by materials scientist Shyue Ping Ong from the University of California, San Diego.

Evaluating the Electrolyte

In powering batteries, the crucial role of electrolytes involves converting chemical energy into electrical energy. Both standard AA batteries and lithium-ion (Li-ion) batteries found in electronic devices operate on this principle, allowing electrically charged ions to flow through the electrolyte medium.

Traditionally, batteries utilize liquid electrolytes, which can present safety risks such as leaks and fires. As a result, the focus has shifted towards designing solid electrolytes. The Washington research team specifically aimed to develop a solid electrolyte material for battery applications.

Distinction Between Batteries and Capacitors

The initial stage of selecting potential candidate materials involved analyzing the structures of known electrolytes and incorporating different elements into their compositions through a computer-based approach, resulting in 32 million potential electrolyte recipes.

Thanks to AI capabilities, particularly machine learning, the team efficiently sifted through this vast list in just 80 hours, contrasting with the decades it would have taken via traditional methods, as explained by lead chemist Nathan Baker from Microsoft’s research team. The machine learning algorithm offered rapid predictions based on established material patterns.