Machine learning leads to green energy breakthrough

Kyushu University joined Osaka University and the Fine Ceramics Center to utilise machine learning to accelerate the discovery of vital materials for advancing green energy technologies

They have found the strongest two materials tailored for use in solid oxide fuel cells, making a significant step forward in sustainable green energy solutions.

What are solid oxide fuel cells?

Solid oxide fuel cells are devices designed to generate energy using eco-friendly fuels like hydrogen, emitting zero carbon dioxide. The discovery holds promise for enhancing the efficiency of hydrogen fuel cells and lays the groundwork for the accelerated discoveries of groundbreaking materials across various sectors.

Addressing the urgent need for eco-friendly energy solutions due to climate concerns, Professor Yoshihiro Yamazaki from Kyushu University’s Department of Materials Science and Technology outlined the potential of creating a hydrogen society.

He emphasised the importance of optimising hydrogen production, storage, and transportation while concurrently boosting the power-generating efficiency of hydrogen fuel cells.

The key to the electric current generation in solid oxide fuel cells lies in their ability to conduct hydrogen ions through a solid electrolyte material efficiently. Traditionally, researchers focused on oxides with specific crystal structures, known as perovskite structures, for electrolyte materials. However, the team at Kyushu University sought to broaden the search to non-perovskite oxides, aiming to discover materials that could conduct protons more efficiently.

Trial and error

Traditional trial and error methods in discovering proton-conducting materials faced limitations, particularly in finding optimal combinations of base materials and dopants.

Dopants, or additional substances added to the base material, enhance proton conductivity. The search became difficult with numerous base and dopant candidates, each with different atomic and electronic properties.

Facing and overcoming the challenges

In response to these challenges, the researchers used machine learning to analyse and predict potential combinations by calculating the properties of different oxides and dopants.

This data-driven approach allowed them to identify factors influencing proton conductivity and advance in the discovery process. The researchers successfully synthesised two materials with unique crystal structures using these insights.

Both materials exhibited proton conductivity in a single experiment. One of the materials boasted a selenite crystal structure, representing this arrangement’s first known proton conductor. The other, with a eulytite structure, showed a high-speed proton conduction path distinct from conventional perovskite structures.

Professor Yamazaki envisions the framework’s potential to broaden the search space for proton-conducting oxides, accelerating advancements in solid oxide fuel cells. He concluded by stating that with minor modifications, this groundbreaking framework could be adapted to various fields of materials science.