Efficient Adjoint-Based Shape Optimization Method for the Inverse Design of Microwave Components

An adjoint-based shape optimization approach for the inverse design of microwave components is proposed. The proposed approach is a pure postprocessing process that only needs the field solutions. It is independent on numerical methods, system matrix, and mesh information, thus can be used in conjunction with mainstream Maxwell solvers currently available on the market, which makes it an efficient, convenient, and user-friendly method. To be applicable to the shape optimization of complex cases, this approach can calculate shape derivatives at all points in design region with only one simulation by a rigorous analytical solution, which makes it have arbitrarily many degrees of freedom (DoFs) while the reported similar approaches can only deal with several variables. Flexible parameters' updating is achieved by the level set method. Owing to the speed of calculating shape derivatives at boundary points within adjoint method, coupled with the accurate updating of shape by gradient descent and level set method, our approach normally requires several or dozens of iterations even in cases with numerous DoFs. The developed approach is used to optimize a high-directivity directional coupler and an unequal Wilkinson power divider to verify its validity and robustness.