Design of an arbitrary ratio optical power splitter based on a discrete differential multiobjective evolutionary algorithm

Traditional photonic integrated devices are designed to predict their optical response by transforming the structure and parameters, and it is often difficult to obtain devices with excellent performance in all aspects. The nanophotonic computing design method based on the optimization algorithm has revolutionized the traditional photonic integrated device design technology. Here, we report a discrete differential evolution algorithm that simulates a natural selection process to achieve an ultracompact arbitrary power ratio splitter. The footprint of the designed splitter is only ${2.5}\;\unicode{x00B5} {\rm m} \times {2}.{5}\;\unicode{x00B5} {\rm m}$2.5µm×2.5µm, the simulated total transmission efficiency is above 90%, the power ratio error is less than 3%, and it can work normally over the C-band. Our algorithm can provide new ideas for the application of genetic algorithms to the automatic optimization of photonic integrated devices.