Download or read book New Multi-Objective Inverse-Design Techniques for Functionalized Metasurfaces written by Ronald Jenkins and published by . This book was released on 2022 with total page 0 pages. Available in PDF, EPUB and Kindle. Book excerpt: Metasurfaces and meta-devices have the potential to offer substantial improvements to many properties of interest (e.g., size, weight, and power) for many kinds of electromagnetic devices by offering radio frequency and optics engineers direct control over the boundary conditions between two media. As the study and application of metasurfaces has advanced from using primarily canonical structures to more freeform device parameterizations, the need for sophisticated inverse-design methods has grown. In this dissertation, the possibilities of multi-objective and evolutionary optimization techniques are explored in the domain of functionalized metasurface design, with applications ranging from nanofabrication robustness and foundry design rules to topology optimization convergence characteristics. The design and realization of a varactor-tunable frequency selective surface will first explore the variety of applications which functionalized surfaces offer. In the optical regime, the high performance of freeform devices becomes limited by fabrication uncertainty and structural constraints. Therefore, two methods are proposed and validated for how to design within this more challenging context. First, a method of direct lithographic mask design rule coercion called minimum feature size enforcement (MFSE) is articulated in the language of morphological operators, and the substantial effects these constraints have on metasurface performance upper-bounds at optical wavelengths are studied using the covariance matrix adaptation evolutionary strategy. Previous demonstrations of this kind of enforcement were limited to the domain of topology optimization, whereas the proposed method can be applied within a broader range of inverse-design techniques. Second, deep learning is used in conjunction with multi-objective optimization to make direct optimization of robust meta-devices feasible. It is shown that despite the use of RCWA which is a highly efficient solver for the specifically considered case of dielectric supercells, deep learning-augmentation offers more than a 10x speedup over an equivalent tradeoff study conducted using only full-wave solvers. In a final application, multi-objective methods are used to modify the prevailing formulation of adjoint topology optimization for strictly planar metasurfaces in order to improve its convergence characteristics. By introducing the multigradient, topology optimization can be made to converge more than 2x faster for the design problem considered, with implications extending to a large class of similar topology optimization design problems. These results demonstrate how multi-objective inverse-design methods can advance the SOA in metasurface design, and in turn has the potential to have substantial impact in device design from radio to optical regimes.