게시판 최근논문

[Editor's Pick] Current Optics and Photonics Vol. 8 no. 6 (2024 December)_1st

사무국 hit 96 date 2024-12-26

A Tutorial on Inverse Design Methods for Metasurfaces

Jin-Young Jeong† , Sabiha Latif† , and Sunae So*

 

Current Optics and Photonics  Vol. 8 No. 6 (2024 December), pp. 531-544
DOI: https://doi.org/10.3807/COPP.2024.8.6.531 

 

 

  Fig. 1  Schematic illustration of an overview of inverse design metasurfaces using machine learning and optimization methods.

 

 

Keywords: Inverse design, Machine learning, Metasurface, Optimization algorithm
OCIS codes: (150.1135) Algorithms; (240.0240) Optics at surfaces


Abstract
This paper provides a tutorial on inverse design approaches for metasurfaces with a systematic analysis of the fundamental methodologies and underlying principles for achieving targeted optical properties. Traditionally, metasurfaces have been designed with extensive trial-and-error methods using analytical modeling and numerical simulations. However, as metasurface complexity grows, these conventional techniques become increasingly inefficient in exploring the vast design space. Recently, machine learning and optimization algorithms have emerged as powerful tools for overcoming these challenges and enabling more efficient and accurate inverse design. We begin by introducing the fundamentals of optical simulations used for forward modeling of metasurfaces and their relevance to inverse design. Next, we explore recent advancements in applying machine learning techniques such as neural networks, Markov decision processes, and Monte Carlo simulations, as well as optimization algorithms, including automatic differentiation, the adjoint method, genetic algorithms, and particle swarm optimizations, and show their potential to revolutionize the metasurface design process. Finally, we conclude with a summary of key findings and insights from this review.

 

첨부파일