Coordinated through the SuperC international project launched in 2023, researchers at Aalto University in Finland trained a machine-learning model to search through untested chemical structures. The algorithm flagged two previously unknown compounds: YRu₃B₂ and LuRu₃B₂. Physicists at Rice University in Texas then synthesized the materials in their laboratory, testing the physical samples and confirming their predicted superconducting properties.

The mechanics of zero-loss power

Superconductors can carry electricity without losing power to resistance. When cooled past a specific temperature threshold, electrons form bonded pairs that pass through the metal lattice without scattering. Standard copper wiring loses power as heat, but superconducting currents flow without energy loss.

Eliminating electrical resistance could change power distribution, shrink industrial motors, and keep data centers cool. Finding practical materials, however, remains hard. Most superconductors only work near absolute zero, requiring liquid helium cooling that makes widespread use impractical.

Basket weaves and flattened electron bands

The two newly discovered compounds combine yttrium or lutetium with ruthenium and boron. Both share an atomic structure called a kagome lattice, named for a Japanese basket-weaving pattern of repeating triangles.

This triangular geometry restricts how electrons move through the material. The layout forces electrons into flat energy bands where their motion slows, allowing quantum forces to dominate. Those strong electron interactions help form the bound pairs required for superconductivity.

Replacing guesswork with predictive screening

Historically, scientists cataloged more than 7,000 superconducting materials, but they predicted only a small handful before making them in a lab. Most discoveries emerged from accidental observations or intuition-guided trial and error. Running detailed quantum physics calculations on thousands of unverified compounds was simply too slow and computationally expensive.

The SuperC team reversed that traditional workflow using a multi-tiered digital pipeline. Their machine-learning algorithm screened large databases of untested chemical combinations, discarding unpromising structures in seconds. Heavy quantum calculations then evaluated only the top candidates, allowing laboratory scientists to spend physical resources strictly on materials with a high probability of success.

What lies ahead for practical technology

While this demonstration proves the predictive model works, neither compound offers an immediate fix for electrical grids. Both YRu₃B₂ and LuRu₃B₂ still require extreme cold to reach their superconducting state, operating well below room temperature.

The true value of the study lies in the discovery method itself. By proving that machine learning can flag complex quantum properties in advance, researchers can now apply the same algorithm to screen candidate materials for room-temperature superconductors, higher-density batteries, and novel industrial chemical catalysts.

Sources: Aalto University; Rice University; SuperC Consortium (August 26, 2026).