Large-Scale Parametrized Metasurface Design Using Adjoint Optimization

Abstract

Optical metasurfaces are planar arrangements of subwavelength meta-atoms that implement a wide range of transformations on incident light. The design of efficient metasurfaces requires that the responses of and interactions among meta-atoms are accurately modeled. Conventionally, each meta-atom’s response is approximated by that of a meta-atom located in a periodic array. Although this approximation is accurate for metastructures with slowly varying meta-atoms, it does not accurately model the complex interactions among meta-atoms in more rapidly varying metasurfaces. Optimization-based design techniques that rely on full-wave simulations mitigate this problem but thus far have been mostly applied to topology optimization of small metasurfaces. Here, we describe an adjoint-optimization-based design technique that uses parametrized meta-atoms. Our technique has a lower computational cost than topology optimization approaches, enabling the design of large-scale metasurfaces that can be readily fabricated. As proof of concept, we present the design and experimental demonstration of high numerical aperture metalenses with significantly higher efficiencies than their conventionally designed counterparts.

Intelligent designs in nanophotonics: from optimization towards inverse creation

“…More TO-designed meta-lens can be checked in the references [17,81,82]. In addition, the general adjoint-based optimizations [83,84] have been reported by Mansouree et al [74] for designing a double-wavelength nano-post meta-lens as shown in Fig. 6 (d).…”

“…Here the inter-post and inter-layer coupling were comprehensively taken into consideration. The systemically designed dual-λ metalens showed the near-diffraction-limited focusing ability and the intensity distributions Moreover, there is increasing attention on the DL-assisted meta-lens design [84]. A specific example can be found from the Pestourie's report [85], where they presented an active-learning algorithm to reduce at least one order of magnitude of the training time for the surrogate model.…”

Multiobjective Statistical Learning Optimization of RGB Metalens

“…Other computational frameworks relying on gradient based architecture have been proposed in Refs. [87][88][89] but they often do not reach the global designs.…”

“…It is worth mentioning that due to the simulation cost, optimization of full-metalenses have mostly been considered for 2D metalenses (cylindrical lens) , 84,85,[89][90][91][92][93] considering different optimization architectures and various numerical apertures. In the present work, we focus only on 3D spherical designs where the focusing is maintained along the three spatial dimensions at the visible regime.…”

Biomimicry-Gradient-Based Algorithm as Applied to Photonic Devices Design: Inverse Design of Flat Plasmonic Metalenses

“…This FOM variation is then associated with the fitness of the design parameters. In the topology optimization method based on homogenisation [17,18] or shape optimization [19], the gradient of the FOM, is used to update simultaneously the design parameters vector. This strategy consisting of simultaneously pushing the design vector elements at each iteration, enforces the local nature of this kind of optimizer.…”