CNN-Surrogate Inverse Design of Disordered Photonic Crystals for Broadband Light Trapping in Thin-Film Silicon Photovoltaics

Ultrathin crystalline-silicon solar cells use far less silicon than conventional wafers, but a film hundreds of nanometers thick transmits much of the sunlight a wafer would absorb. Patterning the film with a deliberately disordered photonic crystal recovers much of that loss, and we show that the specific arrangement of holes is itself a design variable: layouts of identical disorder strength differ by up to 1.6\% in absorbed solar flux. Full-wave simulation is too slow to search this space exhaustively. From a dataset of 2,723 simulated layouts, we train an ensemble of convolutional neural networks to predict a layout's absorption enhancement directly from its geometry, with 0.21\% mean error at ${\sim}10^4\times$ lower cost per evaluation. Using the network to search and the full solver to verify, we obtain designs that beat an equal-budget random search in all eight disorder settings tested; the largest verified gain reaches 2.6\% above its cell's dataset average. Attribution experiments verified by the solver show that the network reads the spatial frequencies through which sunlight couples into the film. Surrogate-guided search thus makes layout-level photonic design practical, offering a route to higher-efficiency ultrathin photovoltaics without additional material. Code and the Photra-2.7k dataset: https://anonymous.4open.science/r/SEER-Photonic-Design/.