Design of a transmissive metasurface antenna using deep neural networks

Abstract

This article presents design methods for a transmissive metasurface antenna composed of four layers of meta-structures based on the deep neural network (DNN). Owing to the structural complexity as well as side effects such as couplings among the adjacent meta-structures, the conventional design of metasurface unit cell strongly relies on the researcher’s intuition as well as time-consuming iterative simulations. A design method for a metasurface antenna unit cell with a size of a quarter wavelength operating at a frequency of 5.8GHz is presented. We describe two unique implementations for designing the target metasurfaces: 1) utilizing the inverse network 2) data augmentation by the forward network and a random search algorithm. With the usage of the two DNNs, the average transmittance of the unit cells is improved by about 0.024 than that of the unit cells designed by the conventional approach. This research invokes the application of DNN in designing antennas and other structures operating at radio frequency.

Tunable metasurfaces towards versatile metalenses and metaholograms: a review

“…Accordingly, we expect that alternative design methods such as inverse design and machine learning can assist researchers in designing metasurfaces that have the desired optical and electrical responses. 23,[168][169][170][171][172][173][174][175][176][177][178] Also, recently developed 3D manufacturing methods of plasmonic and dielectric metasurfaces will enhance the flexibility of possible designs. [179][180][181][182] The development of deep learning will improve human intuition and imagination, and their design will be realized with advanced 3D manufacturing methods.…”

Three-dimensional artificial chirality towards low-cost and ultra-sensitive enantioselective sensing

“…68,69 To apply the chiral nanostructure to a practical and commercial on-chip sensing system, a dynamic system, such as an opto-fluidic system that considers the mobility of particles, have been actively researched. Recently proposed advanced materials, 70–72 tunable materials, 73–76 and machine-learning design methods 77–80 empower the development of chiral metamaterials. Furthermore, artificial chiral nanostructures can be utilized in various photonic applications such as chiral metamirrors, 81 metaholograms, 82–86 metalenses, 87,88 multi-mode OAM generators, 89 beam steering, 90 and color prints.…”

Artificial Intelligence in Meta-optics

“…A rough comparison between evolutionary algorithms and deep learning algorithms is shown in Figure b and c. From the perspective of the entire optimization process, the deep learning method is more computationally efficient. Deep learning has demonstrated much efficacy in accelerating meta-optic designs, such as an auxetic metamaterial and a transmissive meta-atom …”

“…Deep learning has demonstrated much efficacy in accelerating meta-optic designs, such as an auxetic metamaterial 132 and a transmissive metaatom. 133 Inverse design assisted by deep learning could be divided into two parts according to the adopted model types. One is based on the discriminative model, and the other is based on the generative model.…”