AI-Guided Search Finds Two New Superconductors

AI-Guided Search Turns Up Two New Superconductors — and a Faster Way to Find More

An international consortium paired machine learning with quantum physics to predict, then synthesize, two previously unknown superconductors — a screening method its leaders think could one day scan candidate materials by the billions.

Jul 8, 2026Source: ScienceDaily

One of physics' hardest search problems just got a powerful new filter. An international team — the SuperC consortium, led by Aalto University professor Päivi Törmä — has combined machine learning with quantum physics to discover two previously unknown superconductors and, more consequentially, a far faster way to hunt for others.

The discovery

The team identified two new superconductors, YRu3B2 and LuRu3B2. Both derive their properties from electrons forming so-called flat bands within a kagome lattice — a geometric arrangement where electrons move sluggishly and interact strongly, fertile ground for superconductivity.

The method

The approach uses machine-learning pre-screening to narrow a vast space of candidate materials, then runs expensive quantum calculations only on the most promising ones. Collaborators at Rice University, led by Emilia Morosan, then synthesized and experimentally verified both materials — closing the loop from AI prediction to lab confirmation.

Why it's a big deal

Of more than 7,000 known superconductors, researchers had been able to theoretically predict the viability of only about 20, because the physics calculations are so demanding. With machine learning doing the triage, the team believes the number of materials that can be screened could eventually reach the billions. The consortium's stated ambition is to find a room-temperature superconductor by 2033 — a material that would transform power grids, transport and computing.

The caveat

These are conventional superconductors, not the room-temperature holy grail — the breakthrough is the search method, not a miracle material. It fits a broader pattern of AI compressing scientific discovery, from protein structures to new compounds. (The underlying study appeared in Physical Review Research; the consortium's AI-guided pipeline drew fresh attention in early-July coverage.)

Why It Matters

The result that matters is not the two materials — it is the pipeline: machine learning that pre-screens candidates so quantum calculations run only where they will pay off, turning a field that could vet about 20 candidates into one that could eventually scan billions. It is a concrete template for how AI accelerates the physical sciences, and a measured step toward the room-temperature superconductor that would rewire energy and computing.