Hybrid light-matter quasiparticles promise faster, cooler AI chips.
Engineers at the University of Pennsylvania coupled photons with excitons inside a 2-D perovskite microcavity, forming polaritons whose group velocity reaches 0.8 c. Logic gates built from these polaritons execute matrix-vector products in 180 femtoseconds while dissipating 4 attojoules per operation.
You begin viewing computation as a wave phenomenon rather than an electron shuffle. The design loop now includes optical path lengths and cavity tuning instead of transistor sizing alone.
Penn’s Quantum Photonics Group fabricated a 64-by-64 polariton array that performed MNIST inference at 1.2 picojoules per image, two orders of magnitude below the 120 picojoules of a comparable electronic ASIC.
Step 1: Download the open-source polariton-sim package at https://github.com/pennqpg/polariton-sim. Step 2: Define a cavity geometry and run polariton.propagate(input_vector, steps=50). Step 3: Compare the simulated energy per MAC against your current electronic baseline to quantify the projected gain.