On-Chip Photonic Platform Based on Hetero-Integrated Perovskite/Si₃N₄

Integrated photonic chips present considerable promise for applications in optical communications, computing, lidar, sensing, and imaging, providing exceptional data throughput alongside low power consumption. A primary goal is to develop a monolithic on-chip photonic system that consolidates light sources, processors, and photodetectors onto a single chip. Nevertheless, achieving this integration poses challenges due to constraints in materials engineering, chip-integration methods, and design strategies. In this work, we propose and experimentally demonstrate a near-infrared monolithic on-chip photonic system built upon a perovskite/Si₃N₄ photonic platform. We develop a nano-hetero-integration technology that enables the integration of efficient light-emitting diodes, high-performance processors, and sensitive photodetectors. Our implementation of photonic neural networks addresses tasks such as quantum simulations and image classification. Specifically, our network effectively predicts the topological invariant in a two-dimensional disordered SSH model and simulates nonlinear topological models with an average fidelity of 87%. For computer vision applications, we attain over 85% accuracy in edge detection and 56% accuracy on the CIFAR-10 dataset using a scaled-up architecture. This work tackles the challenge of integrating diverse nanophotonic components on a single chip, presenting a promising approach for multifunctional photonic information processing.