Design of high-order series-coupled microring resonator wavelength filter with differential evolution method

Abstract

A design method for higher-order series-coupled microring resonator (MRR) wavelength filters is proposed and discussed. The differential evolution method is one of the machine learning methods, and it is a type of evolutionary algorithm that can be applied to a variety of optimization problems, including non-linear, partially impossible, and multi-modal problems. We design an evaluation function that satisfies multiple requirements by weighting each evaluation item, and optimize the design parameters using the differential evolution method. The weighting values of the evaluation function are adjusted by supervised learning Support Vector Machine to produce a more accurate evaluation function. The designs of high-order MRR filters with target parameters, such as a 3 dB passband, a free spectral range, ripples, and crosstalk, are successfully demonstrated, which shows that the differential evolution method is one of the most effective methods for designing high-order MRR filters.