AI Accelerates Superconductor Discovery, Finds Two New Materials

An international research consortium has successfully used machine learning to discover two new superconducting materials, demonstrating a method that could dramatically accelerate the decades-long search for a practical room-temperature superconductor.

The SuperC consortium identified YRu3B2 and LuRu3B2 by combining AI-driven screening with quantum physics calculations, then synthesized and experimentally verified both materials. The findings, detailed in Physical Review Research, represent a proof of concept for using artificial intelligence to narrow billions of potential material combinations down to the most promising candidates.

Why it matters

Room-temperature superconductors would eliminate the need for expensive cooling systems and could transform energy infrastructure. Replacing conventional conductors in data centers and computers alone could significantly reduce global electricity consumption and the information technology sector's thermal footprint. The discovery method itself matters as much as these specific materials—it offers a scalable approach to a problem that has largely relied on chance.

The Discovery Process

Traditionally, finding superconductors has been computationally prohibitive. According to Aalto University Professor Päivi Törmä, who leads the SuperC consortium, researchers have identified over 7,000 superconductors over the decades, but have only been able to theoretically predict about 20 of them due to computational constraints.

The new approach flips that paradigm. Machine learning algorithms first screen enormous numbers of elemental combinations, selecting candidates most likely to exhibit superconducting properties. Researchers then apply detailed quantum mechanical calculations only to these pre-screened materials, dramatically reducing computational overhead.

For YRu3B2 and LuRu3B2, the team focused on materials where electrons form flat bands within a kagome lattice—a geometric pattern inspired by Japanese basket weaving. After theoretical predictions confirmed their potential, collaborators at Rice University, led by Professor Emilia Morosan, chemically synthesized the compounds and experimentally verified their superconducting behavior.

Scaling the Search

The computational efficiency gains are substantial. Törmä notes that machine learning could enable researchers to process billions of material candidates, compared to the handful that traditional methods allow. This matters because even theoretically promising materials often prove impractical—too difficult to synthesize or impossible to manufacture at scale.

Superconductors conduct electricity with zero resistance, but existing materials only exhibit this property at temperatures near absolute zero, requiring costly cooling systems. The SuperC consortium, established in 2023, has set an ambitious goal: discovering a room-temperature superconductor by 2033.

Broader Context

Superconductors already enable technologies including quantum computers, medical neuroimaging systems, fusion reactors, and magnetic levitation trains. However, their requirement for extreme cooling limits widespread adoption. A material that superconducts at room temperature would remove this barrier entirely.

The SuperC consortium brings together leading physicists focused on using quantum physics to address climate change, representing the first coordinated global effort specifically targeting superconductor discovery at this scale.

These findings were first reported by Aalto University and detailed in Physical Review Research. The consortium receives funding from The Kavli Foundation, Klaus Tschira Stiftung, and several other foundations.