Research On Topology Optimization Design For Integrated Photonic Multiplexer/Demultiplexer

Photonics has received widespread attention as a feasible alternative to electronics for entering the post-Moore era.Various multiplexing technologies in the field of photonics have been widely developed and applied to solve the problem of data expansion.The topology optimization method,which utilizes the continuous characteristics of design parameters to solve gradient information and accelerate the design process,has been broadly applied in the field of photonics design.This work aims to efficiently design ultra-compact multidimensional integrated(de)multiplexers using topology optimization.The main research contents are as follows:(1)A dynamic update rate adjustment strategy was introduced into the topology optimization design process in photonics,drawing inspiration from the dynamic learning rate adjustment schemes used in machine learning.The natural exponential decay method is adopted to ensure a large update rate in the early stages,which not only improves convergence speed but also avoids falling into a local optimal solution to a certain extent.As the number of iterations increases,the update rate gradually decreases.In the later stages,the smaller update rate helps achieve better convergence and avoids oscillations near the optimal solution.(2)A universally applicable simulation model for mode(de)multiplexing devices is established,and the corresponding mathematical model is derived.Based on the above mathematical model,two demultiplexers are designed.The two mode(de)multiplexers have ultra-compact footprint and achieve excellent performance with only a few iterations of optimization.The above mode(de)multiplexers demonstrate that the design method has both high design efficiency and excellent design effectiveness.(3)In photonics,to address the problem of multi-objective imbalance in the multiobjective optimization,a linear weighting coefficient is adjusted using an update method with equal optimization result steps.The method updates the linear weighting coefficient each time the optimization result meets certain conditions,ensuring that the linear weighting coefficient changes gradually and prevents the optimization result from repeatedly jumping laterally.Initially,the weights between multiple objectives are consistent,and so are the linear weighting coefficients.As the optimization process proceeds,the imbalance between multiple objectives emerges.To prevent the gap from widening further,the linear weighting coefficient is dynamically adjusted to balance the objectives.This dynamic adjustment of the weighting coefficient can prevent the optimization process from getting stuck in a local optimum due to the imbalance of the optimization objectives.(4)A universally applicable simulation model for the mode-wavelength multidimensional(de)multiplexer is established,and the corresponding mathematical model is derived.The electromagnetic structure solution process for multi-objective problems is optimized,and parallel algorithms are used to solve multi-objective electromagnetic problems simultaneously.To a large extent,it further improves design efficiency.Based on the above mathematical model and optimized design process,a dual-mode-dualwavelength(de)multiplexer was designed.The footprint is only 4.8 μm ×4.8 μm,with transmission loss and crosstalk less than 0.3 d B and-16 d B,respectively.The above device shows that the design method is suitable for the design of multidimensional(de)multiplexers,and has the advantages of high efficiency and effectiveness.