In a laboratory at Clarkson University in Potsdam, Associate Professor Dhara Trivedi and collaborators at Los Alamos National Laboratory have cracked open a shortcut to discovering tomorrow's materials. Rather than spending years testing compounds one by one in the lab, they've used artificial intelligence and quantum simulations to screen over 2,000 potential perovskite combinations in a fraction of the time—materials that could power more efficient solar panels, quantum computers, advanced sensors, and light-emitting devices.
The challenge of finding new materials has long been the unglamorous bottleneck of innovation. Scientists would synthesize compounds, test their properties, watch them fail, and start over. In renewable energy and quantum technology especially, the search space is vast and unpredictable. Two-dimensional perovskites—layered crystalline materials with unique electronic and quantum properties—have shown real promise for these applications, but identifying which combinations work best requires testing thousands of possibilities. That's where AI changes the equation.
Trivedi's team created a comprehensive database of over 2,000 possible two-dimensional perovskite material combinations, then trained machine learning models to predict which candidates would have the most useful electronic properties. By combining high-throughput computational modeling with quantum-scale simulations, they could evaluate materials virtually before any atoms were assembled in a test tube. The results were published recently in npj Computational Materials, co-authored by Robert Stanton and others.
The impact of this work ripples across multiple industries. The perovskites identified through this screening could enable advances in renewable energy through more efficient solar cells. They could accelerate quantum information science by supporting better quantum computing and communication systems. For everyday technology, the materials could improve low-power electronic devices, photodetectors, LEDs, fiber-optic technologies, and advanced sensing platforms—the kinds of innovations that eventually reach consumer products and critical infrastructure.
What makes this collaboration between a university and a national laboratory particularly significant is how it demonstrates a new model for research. Rather than treating physics, computing, and artificial intelligence as separate domains, the team wove them together. Physics-based simulation provided the foundation; machine learning did the rapid screening; quantum simulations captured the subtle electronic behavior that matters most. "Artificial intelligence helps us narrow down the best possibilities much faster," Trivedi explained, capturing the elegance of the approach. "That means scientists can spend more time developing technologies and less time searching for materials that may not work."
The partnership also signals how universities and national labs can tackle problems too large or complex for either to solve alone. Clarkson brought computational expertise and fresh research perspectives; Los Alamos brought its massive computational resources and decades of materials science experience. The result was faster discovery with broader reach.
Looking ahead, Trivedi's team envisions this method becoming standard practice. Rather than being limited to perovskites, the same machine learning and simulation framework could accelerate the discovery of materials for countless other applications—whatever the next technological challenge demands. "The long-term goal is to create materials that can improve technologies people use every day," Trivedi said. In a world racing toward cleaner energy and quantum breakthroughs, that timeline just got a lot shorter.