At the University of Washington, materials scientists have used artificial intelligence to simulate stacks of molybdenum ditelluride sheets—dozens of atomic layers piled on top of one another in intricate patterns—and discovered something remarkable: quantum phenomena that don't exist at smaller scales suddenly emerge when the structure reaches a certain size. It's a breakthrough that suggests the way forward for quantum materials research isn't just building better supercomputers, but training AI systems to see what humans couldn't previously compute.
Quantum materials are the foundation of an emerging technological revolution. These exotic substances, governed by quantum mechanics rather than classical physics, exhibit properties like superconductivity and unusual magnetism that could power the next generation of quantum computers and energy-efficient electronics. Yet designing them has always been a bottleneck: researchers have to predict how materials behave across vastly different scales, from individual atoms to configurations large enough to be practically useful. For decades, this meant expensive trial-and-error or grinding away on supercomputers that could only handle so much complexity.
Ting Cao, an associate professor of materials science and engineering at the University of Washington and senior author of the research, describes the shift happening in his field with striking clarity: "What is exciting is that AI and quantum computing are beginning to change not just what problems we can solve, but how we do research." His team published their findings in the Proceedings of the National Academy of Sciences on June 2, demonstrating that an AI model trained on smaller datasets can act as a fast, relatively inexpensive surrogate for a supercomputer—effectively learning to extrapolate the behavior of huge material systems without needing to simulate every atom individually.
The approach worked by stacking virtual sheets of molybdenum ditelluride on top of one another, layer after layer. With traditional computing, this would have been impractical, perhaps impossible. But the AI system could handle it, revealing complex quantum behaviors that only emerged at scale. Researchers can now take the most promising candidates and attempt to manufacture them in the laboratory to verify the simulations actually predict the real world.
In a complementary study published June 8 in Nature Communications, Cao's team demonstrated that quantum computers themselves can be part of the solution. Where AI struggles with certain quantum phenomena like entanglement, quantum computers excel naturally—they're powered by the same physics researchers want to study. The team used a quantum computer to investigate an exotic phase of matter called a Laughlin state, opening a new avenue for discovery.
Di Xiao, the chair of materials science and engineering at the University of Washington and co-author of both studies, frames this moment with unmistakable conviction: "We are at the start of a new era. Our field is fundamentally changing. Things that were literally impossible a couple of years ago are now becoming routine."
Looking ahead, Cao and his team plan to expand their datasets and develop models that can simulate a far wider range of materials. The ultimate goal is to merge their AI and quantum computing systems into a hybrid tool—using AI to guide quantum simulations while quantum computers generate new data to improve the AI models further. It's a feedback loop that promises to accelerate discovery and unlock quantum materials that would otherwise remain locked away in the realm of theory.