Photonic Passive Integrated Reservoir Computing Based on Silicon-Cylinder-Assisted Plasmonic Cavity
Yihang Lai, Tian Zhang, Jian Dai, Kun Xu · Journal of Lightwave Technology · 2025
Integrated reservoir computing (RC) systems have attracted growing research interest owing to their excellent temporal signal processing capabilities. However, the current integrated schemes face a trade-off between short-term memory capacity (STMC) and device footprint, resulting in limited system performance. In this article, we propose a novel passive integrated photonic reservoir computing scheme based on a silicon-cylinder-assisted plasmonic cavity (RC-SPC), achieving a large STMC (up to 22.2-bit at an input signal rate of 100 GHz) within a compact footprint (80 μm × 80 μm). Specifically, by introducing the correlation dimension (DF) to measure the dynamic complexity of SPC and then using STMC to evaluate the memory capacity of RC-SPC, the optimal parameters are selected, including the cavity size, the number of output ports and cylinders, the minimum distance of cylinders and the number of interpolation points. Furthermore, we show the STMC of our proposed RC-SPC has a large tuning range (from 5.3-bit to 22.2-bit) by adjusting the input signal rate (from 20 GHz to 100 GHz), indicating the flexible capacity of our structure to deal with the temporal sequential tasks in which the STMC exhibits gain-saturation-inhibition effect sequentially. In four benchmark tasks, including the nonlinear auto-regressive moving average (NARMA), header recognition (HR), Santa-Fe time-series prediction, and nonlinear channel equalization (NCE), the RC-SPC demonstrates comparable to or better performance, while with a smaller footprint, than previously reported typical integrated RCs. Our proposed RC-SPC scheme provides a novel approach for integrated RC system design and establishes a technical foundation for practical applications of photonic intelligent hardware systems.