High computational density nanophotonic media for machine learning inference

作者：Zhenyu Zhao, Yichen Pan, Jinlong Xiang, Yujia Zhang, An He, Yaotian Zhao, Youlve Chen, Yu He, Xinyuan Fang, Yikai Su, Miṅ Gu, Xuhan Guo · 发表于：Nature Communications · 年份：2025 · DOI：10.1038/s41467-025-65213-0 · 被引用次数：3 · 研究领域：Neural Networks and Reservoir Computing、Photonic and Optical Devices、Metamaterials and Metasurfaces Applications

Efficient machine learning inference is essential for the rapid adoption of artificial intelligence (AI) across various domains. On-chip optical computing has emerged as a transformative solution due to its ultra-low power consumption, yet improving computational density remains challenging because of the difficulty of miniaturizing interference-based components. Here, we demonstrate fabrication-constrained scattering optical computing within nanophotonic media, enabled by fabrication-aware inverse design. This yields an ultra-compact optical neural architecture occupying 64 µm²—a three-order reduction compared to conventional optical neural networks. Our prototype achieves 86.7% accuracy on the Iris dataset, closely matching simulations. To further validate scalability, we train a larger 64-input design for optical character recognition using 8×8 handwritten digits, reaching 92.8% test accuracy. These results highlight the potential of nanophotonic media to perform large-scale tasks in ultra-small footprints, paving the way for dense, energy-efficient optical processors for next-generation AI. Advancing AI increasingly requires compact, energy-efficient hardware for fast computation. Here, authors create ultra-compact optical intelligent structures in nanophotonic media, performing inference on Iris and handwritten digit tasks while dramatically reducing size versus conventional designs.