At the National University of Singapore, chemists have cracked a problem that has stumped researchers for years: how to turn carbon dioxide and waste nitrate into fertilizer at speeds that actually matter for industry. The breakthrough hinges on a cadmium-modified iron oxide catalyst—called Cd–Fe₂O₃—that works so efficiently that it could reshape how the world makes urea, the most widely used fertilizer on the planet.
The stakes are enormous. Conventional urea production accounts for more than two percent of global energy consumption and pumps over 200 million tons of carbon dioxide into the atmosphere each year. Scientists have long known that electrochemical production using low-carbon electricity could offer a cleaner path, converting captured CO₂ and nitrate waste into a valuable product. But the approach has remained stubbornly difficult to scale. At the high current densities needed for real manufacturing, catalysts typically botch the job—they favor unwanted side reactions like hydrogen gas formation instead of urea synthesis, tanking production efficiency and leaving the technology stuck in the laboratory.
What makes the NUS team's work distinctive is how they solved the problem. Assistant Professors Pengfei Ou and Lei Wang led a collaboration that fused artificial intelligence, quantum-level simulations, and experimental chemistry into a single design strategy. The researchers first deployed a large language model to survey the entire published literature on urea electrosynthesis, mining the data for patterns. That analysis revealed a critical bottleneck: most existing catalysts perform reasonably well at low production rates but collapse when pushed to industrially viable speeds. Techno-economic modeling showed that achieving a urea partial current density of approximately 100 milliamps per square centimeter was the bare minimum for cost-competitive manufacturing.
Armed with that concrete target, the team used density functional theory to screen materials computationally, hunting for candidates that could suppress side reactions while activating nitrate—a crucial step in urea formation. This narrowed the field to iron oxide as a promising base. But iron oxide alone wasn't quite good enough. The researchers identified a remaining bottleneck: carbon monoxide species were sticking to the iron surface and driving unwanted hydrogen evolution. Adding cadmium to modify the iron oxide's electronic structure solved the problem elegantly. Cadmium made it harder for carbon monoxide and hydrogen to adhere to the catalyst surface, freeing up active sites to promote the carbon-nitrogen bond formation that actually produces urea.
The resulting Cd–Fe₂O₃ catalyst achieved a urea partial current density of about 140 milliamps per square centimeter—above the industrial threshold—while converting more than half of the electrical charge into urea rather than wasted byproducts. The catalyst also ran stably for 100 continuous hours, a critical proof-of-concept for real-world deployment.
What resonates about this work is not just the technical achievement but the methodology itself. By combining machine learning, quantum chemistry, and experiment from the outset rather than using AI merely to explain results after the fact, the team demonstrated a template for catalyst design that could accelerate sustainable manufacturing far beyond fertilizer. Assistant Professor Ou captured this vision plainly: artificial intelligence and quantum-level simulations can identify the right design principles from the beginning, providing a powerful route for developing catalysts for sustainable chemical production at scale.