Momentum-Assisted Adjoint Method for Highly Efficient Inverse Design of Large-Scale Digital Nanophotonic Devices

Gradient-descent-based digitized adjoint method offers a way to realize the high-efficiency inverse design of digital nanophotonic devices with diverse functions. However, vanishing gradient problem encountered in high-dimension may lead significant inefficiencies, making it difficult integrate novel functions on single chip. Here, we propose highly efficient for large-scale design, called adaptive gradient-descent momentum. It uses first- and second-order momentum, instead gradient, update device pattern during optimization. To demonstrate efficiency proposed method, coarse wavelength division multiplexer three-mode power divider dimensions 800 1360, respectively, which are approximately 2-4 times that conventional devices. The simulation results show that, compared descent momentum-assisted has about 4-6 higher obtains better optimization performance, provides powerful tool