On-chip PROCESSOR BASED ON MMI MICRORING RESONATORS for IMAGE EDGE DETECTION in All-optical Domain

Abstract

We propose a new optical processor that implement an optical vector matrix multiplication (OVMM) or dot product circuit using multimode interference (MMI) based microring resonator structures. Our structure can be integrated into only one single chip without the need of wavelength division multiplexers (WDM). The control of weight and input signals is based on the graphene material for high speed computing. The proposed architecture provides high bandwidth, controllable accuracy of input signals and weights, high speed, compactness and capacity of working with negative values. The whole device is designed and simulated in Si3N4 platform which can provide a low loss and works directly with visible wavelength range of images. The new architecture can be applied to optical neural networks with high order of connections of neurons for artificial neural networks. We demonstrate the edge detection using Robert, Prewitt and Sobel operators. The speed of processing is up to 28GHz and MSE difference with the conventional method is of order 0.1.