Neuromorphic Computing with the Plasmonic Microcavity for all Types of Logic Tasks

We propose a new passive-integrated photonic reservoir computer for processing analog signals and study it through detailed numerical simulations. The computer utilizes plasmonic microcavity and silicon arrays to achieve high- dimensional mapping and mixing of the input temporal signal. It can handle all types of sequential logic tasks after simple training. Our reservoir computer has a tiny footprint (60 ^x 60 μm ² ) and a high information processing speed (50 GHz). Our research offers a way for on-chip photonic computing with high- dimensional dynamics.