Fast design of plasmonic metasurfaces enabled by deep learning

Abstract

Metasurfaces is an emerging field that enables the manipulation of light by an ultra-thin structure composed of sub-wavelength antennae and fulfills an important requirement for miniaturized optical elements. Finding a new design for a metasurface or optimizing an existing design for a desired functionality is a computationally expensive and time consuming process as it is based on an iterative process of trial and error. We propose a deep learning (DL) architecture dubbed bidirectional autoencoder for nanophotonic metasurface design via a template search methodology. In contrast with the earlier approaches based on DL, our methodology addresses optimization in the space of multiple metasurface topologies instead of just one, in order to tackle the one to many mapping problem of inverse design. We demonstrate the creation of a Geometry and Parameter Space Library (GPSL) of metasurface designs with their corresponding optical response using our DL model. This GPSL acts as a universal design and response space for the optimization. As an example application, we use our methodology to design a multi-band gap-plasmon based half-wave plate metasurface. Through this example, we demonstrate the power of our technique in addressing the non-uniqueness problem of common inverse design. Our network converges aptly to multiple metasurface topologies for the desired optical response with a low mean absolute error between desired optical response and the optical response of topologies searched. Our proposed technique would enable fast and accurate design and optimization of various kinds of metasurfaces with different functionalities.

Metamaterials Design Method based on Deep learning Database

“…By using the above methods, CST electromagnetic simulation software has been fully put into practical application. It can quickly and easily analyse the properties of metamaterials through modelling and simulation and uses internal adaptive grids to ensure the convergence of results and the credibility of numerical calculations [5][6][7][8][9][10][11][12][13][14].…”

Inverse deep learning methods and benchmarks for artificial electromagnetic material design

“…Five of these models are included because they have been employed to solve inverse AEM problems in prior publications: the Tandem (TD), Neural-Adjoint (NA), Genetic Algorithm (GA), Variational Auto-Encoder (VAE), and the Mixture Density Network (MDN). We provide the associated AEM references for each model in 18 Ma et al, 19 Gao et al, 35 Hou et al So et al 36 Long et al, 37 He et al, 38 Xu et al 39 Ashalley et al, 40 Mall et al, 41 Pilozzi et al 42 Phan et al, 43 Singh et al, 44 Malkiel et al 45 Genetic Algorithm (GA) ✓ ✓ Zhang et al a , 46 Johnson et al 47 Forestiere et al, 48 a Indicates that we adopted the implementation presented in the published work with minimal change in our benchmark. More details about the implementation can be found in the ESI.…”

Deep Learning the Electromagnetic Properties of Metamaterials—A Comprehensive Review

“…MSE) to minimize. One‐to‐manyness has been widely recognized as a problem in the literature, [ 14,98,123,125,151,172–177 ] and has motivated the development of several different approaches to AEM inverse design. As we will discuss later in this section, the underlying strategies of most recent DNN‐based inverse models seek to either i) make networks non‐deterministic and able to identify multiple solutions, or ii) modify the loss criteria to allow the DNN to better search the solution space.…”

“…In 2018 Liu et al., first used a tandem model to predict the thicknesses of a multilayer dielectric nanostructure, [ 123 ] with similar works appearing shortly thereafter using more refined networks. [ 107,112,145 ] Since these initial reports, tandem (bidirectional) DNN models have been used extensively in many different inverse design problems in AEM, including in chiral metamaterials, [ 112,144 ] core–shell nanoparticles, [ 148 ] plasmonic nanostructures, [ 106,145,183 ] metasurfaces, [ 111,173 ] topological photonics, [ 107,174,184 ] and dielectric resonators. [ 125,177 ]…”