AI‑Designed Photonic Chip Components Shrink 500‑Fold, Paving Way for Ultra‑Compact Optics

AI‑Designed Photonic Chip Components

Scientists at the Max Planck Institute have demonstrated that AI‑generated designs can shrink three core photonic‑chip elements up to ↑ 500× smaller than conventional layouts. The breakthrough, published in Nature Communications on May 28, leverages an inverse‑design algorithm that starts with target optical functions—splitting wavelengths, sorting spatial modes, and acting as mirrors—and iterates until a nanostructure emerges that no human would sketch.

“Inverse design lets us define what we want light to do, and the optimization finds a structure that does it,” said Toby Bi, lead author.

By fabricating the components in 400‑800 nm thick silicon nitride, the team achieved reflectivity of ↑ 98.5% and low loss, with mirrors only 11 µm long bouncing light over a hundred times. The wavelength splitter fits within a 5 µm footprint, comparable to a single bacterium, while the mode sorter is only marginally larger. Photonic chips route photons instead of electrons, delivering data at the speed of light, higher bandwidth, and reduced heat—a key advantage for data‑center interconnects, AI accelerators, lidar, and emerging quantum processors. These ultra‑compact components free valuable on‑chip real estate, allowing engineers to pack more functionality into a single die. The next milestone is integrating the three devices into a full optical circuit, a step that could accelerate commercial adoption. Industry observers note that AI‑driven semiconductor design has already cut development cycles from weeks to hours, with firms like Google deploying “superhuman” layouts in production chips (Reuters). As the photonic‑chip market expands, the ability to generate fabrication‑tolerant, sub‑micron structures on demand may become a decisive competitive edge (Bloomberg).