Testing molecules for solar cells is like trying a new sauce in a restaurant where the chef changes every five minutes. Disturbances (humidity, substrate cleanliness, deposition method) hide the true effect. To extract real benefit from noisy data, scientists created a 'smart sieve'—an algorithm that sifts through the results and leaves only reliable improvements.
Training the model on 240 meticulous experiments revealed that what matters are not fancy structures, but just two parameters — the ability to cling to the surface via hydrogen bonds and the charge difference across the molecule. With these simple rules, the program combed through 121 million organic compounds from the PubChem database and selected five candidates. Quantum calculations confirmed: they stick firmly, readily give up electrons, and align energy levels. This 'smart sifting' closed the loop from data to verification and will speed up the development of solar cells, as well as LEDs — a technology where Nobel laureate Shuji Nakamura made a breakthrough. Surprisingly, the secret to improvement fit into just two rules, hidden among 121 million molecules.
In a decade, the efficiency of perovskite solar cells soared from 3% to 25%, almost matching silicon — thanks to the constant hunt for new materials.