Improved Particle Swarm Optimization With Less Manual Intervention for Photonic Inverse Design

Inverse design algorithms for nanophotonic devices have been extensively investigated, leading to the realization of multifunctional devices. Some of these algorithms can only handle structures with continuous refractive index distribution. Others either require extensive manual intervention to tune the associated parameters, or tend to produce locally optimal solutions. To mitigate these problems, we propose an improved particle swarm optimization (PSO) algorithm with similar complexity to traditional PSO and less manual intervention, capable of obtaining the globally optimal solution. We demonstrate the feasibility of the proposed algorithm by designing a TE0 to TE1 mode converter and a compact single-mode waveguide bend with 800 and 60 structure parameters, respectively. The fabricated silicon waveguide bend with an effective radius of$1.75~\mu \text{m}$shows a loss of 0.02 dB per 90°.