A new AI tool developed by MIT researchers is helping generative models create materials with exotic quantum properties, potentially accelerating breakthroughs in quantum computing and advanced technologies. While AI models from companies like Google, Microsoft, and Meta have already generated millions of new materials, they typically focus on stability and are less effective at designing materials with rare quantum behaviors such as superconductivity or unique magnetic states. To address this gap, the MIT team created SCIGEN—short for Structural Constraint Integration in GENerative model—a system that guides generative AI models to follow specific geometric rules during material design. These constraints ensure the AI produces materials with structures known to support quantum phenomena, such as Kagome and Archimedean lattices, which are critical for next-generation materials. The approach shifts focus from generating vast numbers of stable but unremarkable materials to creating a smaller pool of highly promising candidates with real-world impact. “We don’t need 10 million new materials to change the world. We just need one really good material,” says Mingda Li, the senior author and MIT’s Class of 1947 Career Development Professor. The researchers tested SCIGEN with a popular diffusion model called DiffCSP, using it to generate over 10 million materials with Archimedean lattices—2D patterns made of repeating polygons known to produce quantum spin liquids and flat bands, which mimic rare earth elements without requiring them. After filtering for stability, the team used supercomputers at Oak Ridge National Laboratory to simulate 26,000 candidates. The simulations revealed magnetism in 41% of the structures. From this group, the team successfully synthesized two previously unknown compounds—TiPdBi and TiPbSb—confirming that the AI’s predictions closely matched real-world behavior. These materials exhibit exotic magnetic properties linked to quantum spin liquids, a key goal in quantum computing research. The work, published in Nature Materials, brings together experts from MIT, Emory University, Michigan State University, Oak Ridge National Laboratory, and Princeton University. The team believes SCIGEN can dramatically speed up the search for materials that meet strict geometric requirements, which are essential but not sufficient for quantum phenomena. “Experimental progress has been painfully slow,” says Weiwei Xie. “By generating thousands of candidate materials with the right lattice structures, we give experimentalists far more opportunities to find the next breakthrough.” Experts outside the team, like Drexel University’s Steve May, praise the method as a major step forward, noting it could unlock new materials for electronics, magnetism, and optics. The researchers emphasize that AI predictions must still be validated through lab experiments. Future improvements to SCIGEN could include chemical composition rules and functional constraints, allowing even more precise design. “This tool changes the game,” says first author Ryotaro Okabe. “It sacrifices some stability to open the door to truly impactful materials.” The work was supported by the U.S. Department of Energy, the National Science Foundation, and Oak Ridge National Laboratory.