In a landmark achievement that could reshape the global race toward one of materials science’s greatest prizes, an international team of researchers has used machine learning to guide the discovery and experimental confirmation of two entirely new superconductors. The work, published in Physical Review Research on June 17, 2026, marks the first successful experimental validation of materials predicted through a machine-learning-accelerated workflow — a major step forward in the quest for room-temperature superconductors.
The newly discovered compounds, YRu₃B₂ and LuRu₃B₂, both crystallize in a kagome lattice structure and exhibit bulk superconductivity at very low temperatures (Tc = 0.81 K and 0.95 K, respectively). While these critical temperatures are far from room temperature, the real breakthrough lies in the method: an intelligent combination of machine learning pre-screening and targeted quantum calculations that dramatically narrows an otherwise impossibly vast search space of potential materials.
The Discovery: New Kagome Superconductors Confirmed
The two materials belong to the CeCo₃B₂-type structure family, featuring planar kagome networks formed by ruthenium (Ru) atoms. Kagome lattices — named after a traditional Japanese basket-weaving pattern — are known for hosting exotic electronic states, including flat bands that can enhance electron correlations and potentially boost superconducting properties.
Researchers at Rice University, led by Professor Emilia Morosan, synthesized high-quality samples of both compounds. Comprehensive experimental measurements — including magnetization, specific heat, and electrical transport — confirmed true bulk superconductivity with nearly 100% superconducting volume fractions in both materials. This rules out impurity phases or surface effects as the source of the observed behavior.
Compared to the isostructural compound LaRu₃Si₂, the new yttrium and lutetium variants show a more dispersive Ru d_{x²-y²} quasiflat band (leading to lower density of states at the Fermi level) and an overall hardening of the phonon spectrum. These changes reduce the electron-phonon coupling strength λ, consistent with their modestly lower Tc values. Superfluid weight calculations indicate that conventional phonon-mediated pairing dominates, with quantum geometric contributions playing a secondary role due to the dispersive nature of the bands near the Fermi energy.
How Machine Learning Supercharged the Search
For decades, the discovery of new superconductors has been largely serendipitous. Scientists have identified more than 7,000 superconducting materials, yet only about 20 have been successfully predicted theoretically before synthesis — largely because accurate calculations of superconducting properties (such as the Eliashberg spectral function) are extremely computationally expensive.
The new approach, developed within the SuperC consortium, changes that equation. Machine learning models first rapidly screen enormous numbers of possible elemental combinations and crystal structures. Only the most promising candidates — those predicted to have favorable electronic and phononic properties for superconductivity — advance to expensive first-principles density functional theory (DFT) calculations and more sophisticated modeling.
This two-stage pipeline allowed the team to efficiently explore chemical space that would otherwise be intractable. The method successfully identified YRu₃B₂ and LuRu₃B₂ as high-priority targets, which were then synthesized and verified experimentally. The researchers emphasize that the workflow is generalizable and scalable: with further refinement, it could evaluate billions of material candidates.
Professor Päivi Törmä of Aalto University, who leads the SuperC consortium, highlighted the transformative potential: “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.”
Why Room-Temperature Superconductors Are the Ultimate Prize
Superconductivity — the ability of certain materials to conduct electricity with zero resistance and expel magnetic fields (the Meissner effect) — was discovered in 1911. For over a century, practical applications have been limited by the need for extreme cooling. Conventional superconductors require liquid helium temperatures (around 4 K), while even the record-holding hydrogen-rich compounds under extreme pressure only reach superconductivity well below room temperature.
A true room-temperature superconductor operating at ambient pressure would be revolutionary:
- Energy infrastructure: Lossless power transmission could slash global electricity losses (currently ~8–10% in many grids).
- Computing and data centers: Replacing copper interconnects with superconducting lines could dramatically reduce energy consumption and heat generation in the ICT sector, which already accounts for a growing share of global electricity use.
- Quantum technologies and magnets: Enable more powerful, efficient magnets for fusion reactors, maglev trains, MRI machines, and particle accelerators.
- Climate impact: Lower energy demand directly supports decarbonization goals.
The SuperC consortium, coordinated by Aalto University and involving leading groups across Europe and the United States (including Rice University and Princeton), was explicitly formed in 2023 with the ambitious target of discovering a room-temperature superconductor by 2033. The project integrates advanced quantum geometry theory, machine learning, and experimental synthesis in a coordinated global effort.
Challenges Remain — But Momentum Is Building
The two new kagome superconductors are conventional, phonon-mediated materials with relatively low Tc. They do not yet point directly to a room-temperature solution. However, they validate the entire discovery pipeline and provide valuable new data points for refining machine learning models.
Key remaining challenges include:
- Improving the accuracy of ML predictions for higher-Tc candidates.
- Better understanding and engineering of mechanisms beyond conventional electron-phonon coupling (such as quantum geometry and flat-band physics).
- Developing scalable synthesis routes for complex new materials.
- Closing the loop between theory, computation, and experiment more rapidly.
This work demonstrates that the integration of machine learning screening, first-principles theory, and targeted experimental synthesis is not only feasible but highly effective for accelerating materials discovery.
Looking Ahead
The research will be featured in Aalto University’s Designs for a Cooler Planet exhibition running from September 1 to October 30, 2026, in Greater Helsinki, Finland — underscoring the connection between fundamental materials breakthroughs and sustainable technology development.
As the SuperC consortium and parallel efforts worldwide continue to refine these AI-guided methods, the timeline for discovering a practical room-temperature superconductor is shortening. What once seemed like a distant dream is now being approached with systematic, data-driven speed.
The discovery of YRu₃B₂ and LuRu₃B₂ is more than just two new entries in the periodic table of superconductors — it is proof that artificial intelligence, when thoughtfully combined with deep physical insight, can fundamentally change how humanity hunts for the materials that could power a cleaner, more efficient future.
Reference: Mustaf, R. A., et al. “Machine-learning-guided discovery of kagome superconductors YRu₃B₂ and LuRu₃B₂.” Physical Review Research (2026). DOI: 10.1103/lpqj-7hyg.
The hunt has accelerated. The next breakthroughs may arrive sooner than we think.
