Figure - available from: Advanced Photonics Research
This content is subject to copyright. Terms and conditions apply.
a) The perspective view of the three‐layer unit cells containing 64 × 64 configurations of metal and void square blocks with twofold symmetry. b) Top layer, c) middle layer, and d) bottom layer.
Source publication
Electromagnetic (EM) metasurfaces can present a versatile platform for realization of multiple diverse EM functionalities with incident wave frequency, polarization, or propagation direction through appropriate choice of unit cells structures. However, the inverse design of multifunctional metasurfaces relies on massive full‐wave EM numerical simul...
Citations
... Deep learningbased algorithms can generate ultra-fast solutions for a given target response once the computationally expensive training using the sample dataset is over. There have been several reports of utilizing neural network architectures for inverse designing multifunctional metasurfaces [31,32,33,34]. However, they have mostly been limited to Generative Neural Networks, namely vari-ant autoencoder (VAE) and generative adversarial network (GAN). ...
In this article, we report, for the first time, broadband multifunctional metasurfaces with more than four distinct functionalities. The constituent meta-atoms combine two different phase change materials, $\mathrm{VO_2}$ and $\mathrm{Sb_2S_3}$ in a multi-stage configuration. FDTD simulations demonstrate a broadband reflection amplitude switching between the four states in visible range due to the enhanced cavity length modulation effect from the cascaded Fabry-Perot cavities, overcoming the inherent small optical contrast between the phase change material (PCM) states. This, along with the reflection phase control between the four states, allows us to incorporate both amplitude and phase-dependent properties in the same metasurface - achromatic deflection, wavelength beam splitting, achromatic focusing, and broadband absorption, overcoming the limitations of previous functionality switching mechanisms for the visible band. We have used a Tandem Neural network-based inverse design scheme to ensure the stringent requirements of different states are realized. We have used two forward networks for predicting the reflection amplitude and phase for a meta-atom within the pre-defined design space. The excellent prediction capability of these surrogate models is utilized to train the reverse network. The inverse design network, trained with a labeled data set, is capable of producing the optimized meta-units given the desired figure-of-merits in terms of reflection amplitude and phase for the four states. The optical characteristics of two inverse-designed metasurfaces have been evaluated as test cases for two different sets of design parameters in the four states. Both structures demonstrate the four desired broadband functionalities while closely matching the design requirements, suggesting their potential in visible-range portable medical imaging devices.
... To tackle the challenge of generating meta-surfaces with multiple functionalities, two inverse design models have been introduced. Kiani et al. utilized conditional GANs in the microwave regime to design multilayer metal-dielectric meta-surfaces with three different functions and full-space coverage [42]. In a similar vein, An et al. employed a combination of conditional and Wasserstein GANs for the inverse design of all-dielectric meta-surfaces in photonics [43]. ...
This paper outlines a new approach to designing tunable electromagnetic (EM) graphene-based meta-surfaces using convolutional neural networks (CNNs). EM meta-surfaces have previously been used to manipulate EM waves by adjusting the local phase of subwavelength elements within the wavelength scale, resulting in a variety of intriguing devices. However, the majority of these devices have only been capable of performing a single function, making it difficult to achieve multiple functionalities in a single design. Graphene, as an active material, offers unique properties, such as tunability, making it an excellent candidate for achieving tunable meta-surfaces. The proposed procedure involves using two CNNs to design the passive structure of the graphene meta-surfaces and predict the chemical potentials required for tunable responses. The CNNs are trained using transfer learning, which significantly reduced the time required to collect the training dataset. The proposed inverse design methodology demonstrates excellent performance in designing reconfigurable EM meta-surfaces, which can be tuned to produce multiple functions, making it highly valuable for various applications. The results indicate that the proposed approach is efficient and accurate and provides a promising method for designing reconfigurable intelligent surfaces for future wireless communication systems.
... In the conditional GAN, information about specific boundary conditions is provided to the discriminator and generator to form conditional images rather than random images. Reproduced under the terms of the Creative Commons CC-BY license [125]. Copyright 2022, the authors, Advanced Photonics Research published by Wiley-VCH GmbH. ...
... Liu et al. [71] demonstrated a GAN for achieving customer-defined optical spectra with high fidelity using nanophotonic metasurfaces. Kianin et al. [125] employed a conditional GAN, which uses auxiliary information as inputs of the network, for vortex beam generation and wave manipulation of inversedesigned metasurfaces. Mall et al. [82] demonstrated a cyclical DNN framework by applying a pseudo-GA to a conditional GAN and a simulation neural network to train a generative model efficiently. ...
Evolving nanotechnologies and further understanding of nanophotonics have recently enabled the control of electromagnetic waves using metasurfaces. Since metasurfaces can provide diverse optical characteristics depending on their geometries, the forward design of metasurfaces conventionally has been employed through an understanding of the physical effects of each geometrical parameter. In contrast, the inverse
design approach optimizes the metasurface geometry using computational algorithms. This review discusses recent studies on constructing generative models for the inverse design of nanophotonic metasurfaces. The generative model for inverse design is constructed mainly with three components: an evaluator, a generator, and a criterion. The evaluator, which can be implemented by physical simulators or deep neural networks, determines whether the input metasurface geometry satisfies the target optical characteristics. The generator suggests new possible design candidates that may have optical properties close to the target. The criterion, which includes algorithms based on mathematical optimization and artificial intelligence, manages the operation of a generative model while satisfying the convergence of optimal solutions. Inverse design takes advantage of larger design space for customized applications along with the possibility of investigating new physics, and hence it is
expected to improve metasurfaces further with the emerging computational algorithms.
... To tackle the challenge of generating metasurfaces with multiple functionalities, two inverse design models have been introduced. Kiani et al. utilized conditional GANs in the microwave regime to design multi-layer metal-dielectric metasurfaces with three different functions and full-space coverage [45]. In a similar vein, An et al. employed a combination of conditional and Wasserstein GANs for the inverse design of all-dielectric metasurfaces in photonics [46]. ...
This paper outlines a new approach to designing tunable electromagnetic (EM) graphene-based metasurfaces using convolutional neural networks (CNNs). EM metasurfaces have previously been used to manipulate EM waves by adjusting the local phase of subwavelength elements within the wavelength scale, resulting in a variety of intriguing devices. However, the majority of these devices have only been capable of performing a single function, making it difficult to achieve multiple functionalities in a single design. Graphene, as an active material, offers unique properties, such as tunability, making it an excellent candidate for achieving tunable metasurfaces. The proposed procedure involves using two CNNs to design the passive structure of the graphene metasurfaces and predict the chemical potentials required for tunable responses. The CNNs are trained using transfer learning, which significantly reduced the time required to collect the training dataset. The proposed inverse design methodology demonstrates excellent performance in designing reconfigurable EM metasurfaces, which can be tuned to produce multiple functions, making it highly valuable for various applications. The results indicate that the proposed approach is efficient and accurate and provides a promising method for designing reconfigurable intelligent surfaces for future wireless communication systems.
... The prediction model is trained with cycle loss to make sure that ( (y)) ≈ y while keep the forward model fixed. • Conditional Generative Adversarial Networks (CGAN) [15,24,30,47]: Generative adversarial networks (GAN) have a generator network and a discriminator network. The discriminator network discriminates between real data and generated data. ...
Aiding humans with scientific designs is one of the most exciting of artificial intelligence (AI) and machine learning (ML), due to their potential for the discovery of new drugs, design of new materials and chemical compounds, etc. However, scientific design typically requires complex domain knowledge that is not familiar to AI researchers. Further, scientific studies involve professional skills to perform experiments and evaluations. These obstacles prevent AI researchers from developing specialized methods for scientific designs. To take a step towards easy-to-understand and reproducible research of scientific design, we propose a benchmark for the inverse design of nanophotonic devices, which can be verified computationally and accurately. Specifically, we implemented three different nanophotonic design problems, namely a radiative cooler, a selective emitter for thermophotovoltaics, and structural color filters, all of which are different in design parameter spaces, complexity, and design targets. The benchmark environments are implemented with an open-source simulator. We further implemented 10 different inverse design algorithms and compared them in a reproducible and fair framework. The results revealed the strengths and weaknesses of existing methods, which shed light on several future directions for developing more efficient inverse design algorithms. Our benchmark can also serve as the starting point for more challenging scientific design problems. The code of IDToolkit is available at https://github.com/ThyrixYang/IDToolkit.
... Deep learning (DL) based on deep neural networks (DNN) has been applied to many fields, such as computer vision (CV) [12] and natural language processing (NLP) [13], which have achieved significant advancements. DL has also been widely used in the field of optics, including metasurface design [14,15], quantum optics [16], fiber optics [17], and multi-layer thin films design [18]. There is also a small body of research on combining DL with optical design. ...
In this paper, we propose a method to automatically generate design starting points for free-form three-mirror imaging systems with different folding configurations using deep neural networks. For a given range of system parameters, a large number of datasets are automatically generated using the double seed extended curve algorithm and coded optimization. Deep neural networks are then trained using a supervised learning approach and can be used to generate good design starting points directly. The feasibility of the method is verified by designing a free-form three-mirror system with three different folding configurations. This method can significantly reduce the design time and effort for free-form imaging systems, and can be extended to complex optical systems with more optical surfaces.
The interplay between two paradigms, artificial intelligence (AI) and optical metasurfaces, nowadays appears obvious and unavoidable. AI is permeating literally all facets of human activity, from science and arts to everyday life. On the other hand, optical metasurfaces offer diverse and sophisticated multifunctionalities, many of which appeared impossible only a short time ago. The use of AI for optimization is a general approach that has become ubiquitous. However, here we are witnessing a two-way process—AI is improving metasurfaces but some metasurfaces are also improving AI. AI helps design, analyze and utilize metasurfaces, while metasurfaces ensure the creation of all-optical AI chips. This ensures positive feedback where each of the two enhances the other one: this may well be a revolution in the making. A vast number of publications already cover either the first or the second direction; only a modest number includes both. This is an attempt to make a reader-friendly critical overview of this emerging synergy. It first succinctly reviews the research trends, stressing the most recent findings. Then, it considers possible future developments and challenges. The author hopes that this broad interdisciplinary overview will be useful both to dedicated experts and a general scholarly audience.
Multifunctional metasurfaces have demonstrated extensive potential in various fields. During the design of metasurfaces, optimization of components for polarization, amplitude distribution, phase distribution, and other factors is necessary. This process typically demands expert involvement and is time‐consuming. In this paper, a metasurface inverse design method is introduced that combines a high‐precision ultra‐wideband spectrum forward prediction utilizing a neural network and a genetic algorithm. A neural network is constructed and trained to accurately predict the amplitude and phase of a 16 × 16 discrete grid structure with high degrees of freedom in the frequency range of 0.5–2 THz. Leveraging the neural network's ultra‐fast spectrum prediction capabilities (producing approximately 1000 spectra per second), the average optimization time for a single component is reduced to 1.5 min. Finally, the effectiveness of this inverse design method is validated through the design and simulation of multifunctional reflective deflection metasurfaces with two sets of 3‐bit frequency multiplexing and polarization multiplexing. The proposed metasurface inverse design method offers a new approach for the rapid design of components in complex application scenarios and holds significant reference value for metasurface designers.
