Inverse Design of Electromagnetic Metasurfaces Utilizing Infinite and Separate Latent Space Yielded by a Machine Learning-Based Generative Model

Aug 6, 2026

Upload by Finn

An inverse design framework for electromagnetic metasurfaces (EMMS) is proposed using a neural network generative model that produces infinite, continuous latent representations to comprehensively span the EMMS property space. The EMMS inverse mapping problem is inherently one-to-many, because identical electromagnetic properties can be realized by multiple scatterer geometries. Unlike prior solutions requiring elaborate configurations or dataset preprocessing modules, the approach builds an end-to-end network for multimodal data and integrates a classification scheme, validating performance via comparisons with PSSFSS simulations.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages13

Reading Time33 min

AccessFree to download

Abstract

I. Introduction

Electromagnetic metasurfaces and applications

Motivation: limits of conventional iterative design

ML-based inverse design and generative models

Latent spaces in VAE and GAN approaches

JOURNAL OF ELECTROMAGNETIC ENGINEERING AND SCIENCE, VOL. 24, NO. 2, 178~190, MAR.2024 [https://doi.org/10.26866/jees.2024.2.r.218](https://doi.org/10.26866/jees.2024.2.r.218)[ ](https://doi.org/10.26866/jees.2024.2.r.218)ISSN 2671-7263 (Online) ∙ ISSN 2671-7255 (Print) Inverse Design of Electromagnetic Metasurfaces Utilizing Infinite and Separate Latent Space Yielded by a Machine Learning-Based Generative Model Jong-Hoon Kim1 ∙ Ic-Pyo Hong2,* | Abstract | | --- | | This study proposes an inverse design framework for metasurfaces based on a neural network capable of generating infinite and continuous latent representations to fully span the electromagnetic metasurfaces (EMMS) property space. The inverse design of EMMS inherently poses the one-to-many mapping problem, since one set of electromagnetic properties can be provided by many different shapes of scatterers. Previous studies have addressed this issue by introducing machine learning-based generative models and regularization strategies. However, most of these approaches require highly complex operating configurations or external modules for preprocessing datasets. In contrast, this study aimed to construct a more streamlined and end-to-end solver by building a network to process multimodal datasets and then incorporating a classification scheme into the network. The validity of the idea was confirmed by comparing the accuracy of the results predicted by the proposed approach and the outcomes simulated using PSSFSS.
KeyWords: EMMS Property, Generative Model, Inverse Design, Latent Spaces, Metasurfaces. | I. INTRODUCTION Electromagnetic metasurfaces (EMMS) are composed of thin two-dimensional arrays of subwavelength-sized metallic unit cells that can manipulate the phase/magnitude response and polarization of electromagnetic waves at specific frequencies [1–4]. Owing to these capabilities, EMMS are used as spatial filters, polarizers, or microwave absorbers in broad applications that require the manipulation of electromagnetic waves or optical frequencies, such as communication systems, energy harvesting, and cloaking [5–10]. These unique and extraordinary functionalities are largely determined by the physical structures and material properties of the metasurfaces, which are judiciously selected or designed. More importantly, the potentially limitless pattern shapes of scatterers can significantly broaden the applicability of EMMS to extremely vast areas. In this context, the accurate geometric design of meta-atoms plays a crucial role in the successful construction of metasurfaces that implement targeted electromagnetic (EM) properties. However, conventional design processes [11, 12] that map unknown design parameters to specific EM properties through numerous iterations of fullwave simulations are painfully time consuming. Furthermore, the process relies heavily on a limited number of designers, with expertise in electromagnetism and intuitive insights derived Manuscript received September 20, 2023 ; Revised November 20, 2023 ; Accepted December 08, 2023. (ID No. 20230920-175J) 1Smart Natural Space Research Center, Kongju National University, Cheonan, Korea. 2Department of Information and Communication Engineering, Kongju National University, Cheonan, Korea. *Corresponding Author: Ic-Pyo Hong (e-mail: [iphong@kongju.ac.kr](iphong@kongju.ac.kr)) This is an Open-Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License ([http://creativecommons.org/licenses/by-nc/4.0](http://creativecommons.org/licenses/by-nc/4.0)) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited. ⓒ Copyright The Korean Institute of Electromagnetic Engineering and Science. KIM and HONG: INVERSE DESIGN OF ELECTROMAGNETIC METASURFACES UTILIZING INFINITE AND SEPARATE LATENT SPACE YIELDED… from extensive experience, who can efficiently narrow down t

FAQs

What problem does the study address in electromagnetic metasurface inverse design?

It addresses the one-to-many inverse mapping where many different scatterer shapes can produce the same electromagnetic properties, making direct inverse prediction difficult.

How does the proposed method handle the latent space challenge?

It uses a neural generative model that yields infinite and continuous latent representations, aiming to fully cover the EMMS property space.

Why is the proposed approach described as more streamlined than prior work?

Most prior approaches need complex operating setups or external preprocessing modules, while this work targets a more end-to-end solver by processing multimodal data within the network and adding an internal classification scheme.

Aug 6, 2026

Upload by Finn