Cascaded Micro‐Ring Resonators for Low‐Crosstalk High‐Density Photonic Convolutional Computing

Abstract Photonic neural networks (PNNs) based on micro‐ring resonators (MRRs) have attracted significant attention for their compactness and low power consumption. However, there remains substantial room for improvement in spectral density and network performance. Here, a novel PNN architecture is introduced based on double‐stage serially coupled ring resonators (DCRRs), incorporating specially designed optoelectronic signal modulation and detection circuits, achieving a PNN with high spectral density, robustness, and accuracy. The DCRR achieves an extinction ratio of 55 dB and a narrow bandwidth of 0.17 nm. Through systematic innovation, it addresses the challenge of representing negative numbers in optoelectronic neural networks caused by the non‐negativity of light intensity, enabling positive and negative weighting operations using a single photodiode‐based architecture. Experimental validation in digital and cell edge extraction and classification tasks demonstrates high accuracies above 95%. Compared to single‐ring computing architectures with the same parameters, this method significantly reduces inter‐channel crosstalk and spectral spacing, leading to a sixfold increase in spectral density and achieving a compute density of 2.48 TOPS/mm 2 . Furthermore, utilizing DCRR‐based nonlinear activation results in faster convergence speed and higher recognition accuracy. The method provides the technical foundation for achieving high‐density, high‐precision photonic computing.