Researchers from the SuperC consortium have used machine learning to discover two new superconductors and to build a screening method that could dramatically accelerate the search for a room-temperature superconductor. Their work, published in Physical Review Research in July 2026, shows how AI can rapidly narrow the vast space of potential materials, targeting only the most promising candidates for detailed quantum calculations.
Superconductors carry electricity without resistance, but today they require cooling to near absolute zero. They already power quantum computers, MRI machines, fusion reactors, and maglev trains. A material that superconducts at everyday temperatures would reshape energy grids and computing. "Superconductive materials that can operate at room temperature would forever change the way we consume energy," said Päivi Törmä, professor at Aalto University and head of the SuperC consortium. "If such a material could replace regular conductors in applications like computers and data centers, global energy consumption could be slashed and the heat footprint of the ICT sector vastly reduced."
Machine Learning and Quantum Geometry Join Forces
The new superconductors YRu3B2 and LuRu3B2 get their zero-resistance properties from electrons that form flat bands within a kagome lattice, a hexagonal pattern inspired by Japanese basket weaving. To find them, the team first used machine learning to pre-screen an enormous number of elemental combinations. A specialized algorithm selected the most promising ones, which were then analyzed with quantum calculations that predicted superconductivity. Collaborators at Rice University, led by Professor Emilia Morosan, synthesized the materials and experimentally verified that both are indeed superconductors. The proof-of-concept study was published in Physical Review Research.
From Serendipity to Systematic Search
Despite decades of work, scientists have identified roughly 7,000 superconductors, most found by chance. Traditional theoretical screening is so computationally heavy that only about 20 of those were predicted before their discovery. "Over the decades researchers have recognized over 7,000 superconductors, but mostly serendipitously," said Törmä. "Our method uses machine-learning-based pre-screening followed by targeted calculations on the promising candidates. This approach will greatly speed up superconductor discovery in the future. With machine learning, we may be able to push the number of materials we can process into the billions. This will take us a critical step closer to finding a room-temperature superconductor."
The SuperC consortium aims to discover a room-temperature superconductor by 2033. As AI proves its value in materials research, many scientists are building these capabilities through dedicated resources such as the AI Learning Path for Research Scientists.
Why this matters for research scientists
For researchers in condensed matter physics and materials science, this study demonstrates a practical workflow that combines machine learning with quantum-level calculations to drastically reduce the computational cost of discovering new functional materials. The approach is transferable to other classes of compounds, offering a model for accelerating AI-driven discovery across many labs.
Your membership also unlocks:
