Fertilizer production depends on ammonia, which is synthesized from nitrogen and hydrogen. The catalyst here is like a versatile frying pan: it must work at every stage of the multi-step reaction. A conventional iron catalyst requires high temperatures and pressures, wasting a lot of energy.

Researchers trained machine learning on millions of structures, and AI generated 15 million candidates. Computer modeling filtered out the inefficient ones, leaving 279 materials. Among them are familiar iron with ruthenium and surprising ones, like those with vanadium. Amazingly, bacteria in nature already use vanadium-containing enzymes for nitrogen fixation — evolution beat us by millions of years.

The new catalysts will enable fertilizer production without methane and water consumption, with a smaller carbon footprint. The behavior of atoms can be calculated using the laws of quantum mechanics, discovered by Erwin Schrödinger, Wolfgang Pauli, and Enrico Fermi, where entropy, a measure of disorder, plays a key role.