Broadband transmission-to-polarization metasurface unit and design method

By designing a square broadband transmission to polarized metasurface unit and a deep learning-based design method, the problem of low design efficiency of broadband metasurface unit in the prior art is solved, and a metasurface unit with high bandwidth and high transmission amplitude is realized. Suitable for a variety of application scenarios.

CN118943743AInactive Publication Date: 2024-11-12NANJING UNIV OF INFORMATION SCI & TECH

Patent Information

Application Number
CN202410926396.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-11
Publication Date
2024-11-12
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

In the prior art, when designing broadband metasurface units, it is difficult to achieve efficient polarization conversion and flexible bandwidth regulation, with a long design cycle and low efficiency.

Method used

A square broadband transmission to polarized metasurface unit is designed, including a dielectric substrate, upper and lower metal layers and metal branches, and forward prediction and reverse design through a deep learning framework, and structural parameters are optimized to achieve high bandwidth and high transmission amplitude.

Benefits of technology

The relative bandwidth is achieved up to 20%, the transmission amplitude is higher than 0.9, the design efficiency and performance are significantly improved, and the bandwidth is easy to regulate, and it is suitable for different application scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118943743A_ABST
    Figure CN118943743A_ABST
Patent Text Reader

Abstract

The invention discloses a broadband transmission-to-polarization metasurface unit and a design method, the metasurface unit is set to be square, the metasurface unit comprises a dielectric substrate, the dielectric substrate is provided with an upper metal layer and a lower metal layer, and each of the upper metal layer and the lower metal layer comprises an incomplete rectangular metal patch and a plurality of metal branches; according to the design method, forward prediction and reverse design are carried out on a transmission-to-polarization metasurface unit based on deep learning, a forward prediction neural network and a reverse design network are trained based on a deep learning framework, different relative bandwidths of different frequency points serve as input, and the network automatically generates structure parameters of the corresponding metasurface unit; and the generated structure parameters are respectively put into a forward prediction network and an electromagnetic simulation tool to obtain corresponding predicted and simulated electromagnetic response spectrums. According to the method, flexible bandwidth regulation and control are realized, the design efficiency and performance are improved, and the requirements on the bandwidth in different application scenes are met.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of electromagnetic metasurfaces, and in particular to a broadband transmission polarization conversion metasurface unit and a design method thereof. Background Art

[0002] Metasurface is a new type of artificial two-dimensional subwavelength electromagnetic material. By adjusting the parameters, geometry and arrangement of the dielectric structure on its surface, it can flexibly control the amplitude, phase and polarization of electromagnetic waves in space, thereby realizing a variety of optical functions. Compared with traditional electromagnetic devices, it has the characteristics of diverse functions, ultra-thinness and easy integration. Polarization is one of the basic properties of electromagnetic waves. It refers to the law of change of the electric field intensity vector of electromagnetic waves with time. The polarization of electromagnetic waves can be divided into linear polarization, circular polarization and elliptical polarization. The polarization characteristics of electromagnetic waves are widely used in satellite communications, radar reception anti-interference, aerospace and medical fields.

[0003] Artificial electromagnetic metasurfaces, with their excellent electromagnetic control performance and characteristics, provide a new method for building broadband, efficient and ultra-thin polarization converters. At the same time, the polarization control technology based on this metasurface has great application potential and broad development prospects in many cutting-edge fields such as radar cross-section reduction (RCS), wavefront fine control, and invisible cloaks. According to different functional mechanisms, metasurface polarization conversion devices are divided into two categories: reflection type and transmission type. Reflection type polarization converters need to ensure that the material is lossless or the loss is relatively low. At a certain frequency, it can be approximately considered that the electromagnetic wave has achieved total reflection, and the conversion efficiency is close to 100%. Transmission type polarization converters allow electromagnetic waves to undergo polarization conversion when passing through the metasurface without being completely reflected. Their design needs to take into account factors such as material loss and phase difference to achieve efficient polarization conversion.

[0004] With the continuous development of wireless communication technology, the performance requirements for metasurface units are increasing. In particular, in the design of broadband metasurface units, it is necessary to achieve the ability to flexibly adjust the bandwidth. Traditional design methods often rely on manual experience and trial and error, with a long design cycle and difficulty in achieving ideal results. Summary of the invention

[0005] Purpose of the invention: The purpose of the present invention is to provide a broadband transmission-polarization metasurface unit and a design method to solve the problems raised in the background technology.

[0006] Technical solution: The broadband transmission polarization metasurface unit described in the present invention is arranged in a square shape, including a dielectric substrate, the dielectric substrate is provided with an upper metal layer and a lower metal layer, the upper metal layer and the lower metal layer each include an incomplete rectangular metal patch and a plurality of metal branches; wherein the incomplete rectangular metal patch is obtained by cutting the corners of a pair of complete rectangular metal patches and simultaneously digging out four rectangular grooves inside the patches, and the lower metal layer is a translational mirror image of the upper metal layer; the upper and lower metal layers are connected by eight metal through holes, and the end of each metal through hole is connected to the incomplete rectangular metal patch through a metal branch.

[0007] Furthermore, the side length of the metasurface unit is approximately equal to seven twentieths of the wavelength of the central operating frequency, and the thickness of the metasurface unit is less than one tenth of the wavelength of the central operating frequency.

[0008] Furthermore, there are 7 metal branches.

[0009] The first one is F4B plate.

[0010] Furthermore, when a linearly polarized wave with an arbitrary incident azimuth angle φ is irradiated onto the metasurface unit, the transmitted linearly polarized wave is rotated 90° and remains stable as the incident azimuth angle of the electromagnetic wave changes.

[0011] Furthermore, the maximum transmission amplitude of the metasurface unit is higher than 0.95, and the maximum relative bandwidth is 20%.

[0012] The present invention discloses a method for designing the bandwidth of a polarizer based on deep learning, which performs forward prediction and reverse design on a transmission polarization conversion metasurface unit based on deep learning.

[0013] Further, the following steps are included: (S1) Collecting data sets: determining the structural parameters of the broadband metasurface unit, dividing the structural parameters into ranges and performing random sampling according to uniform distribution, using electromagnetic wave numerical simulation tools to obtain the electromagnetic response spectrum corresponding to each group of structural parameters, and collecting the simulated structural parameters and the corresponding electromagnetic response spectrum and organizing them into a data set; (S2) Based on a deep learning framework, training a forward prediction neural network, wherein the input of the neural network is the structural parameters of the metasurface unit, and the output is the electromagnetic response spectrum of the corresponding metasurface unit; (S3) Based on the deep learning framework, the reverse design network is trained, and the different relative bandwidths of different frequency points are taken as input. The network automatically generates the structural parameters of the corresponding metasurface units; the generated structural parameters are respectively put into the forward prediction network and the electromagnetic simulation tool to obtain the corresponding predicted and simulated electromagnetic response spectra.

[0014] Furthermore, the electromagnetic response spectrum is a linear S21 parameter curve.

[0015] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: the metasurface can achieve a higher bandwidth, the maximum bandwidth can reach 20%, and the bandwidth is easy to control. Within the working bandwidth, the transmission amplitude is higher than 0.9, and the maximum value can reach 0.98. In addition, the unit also has the characteristics of small size and thin thickness; the bandwidth design method of the polarizer based on deep learning, based on the deep learning framework, the broadband transmission polarization metasurface unit is forward predicted and reversely designed, and the error between the designed relative bandwidth and the target relative bandwidth is controlled within 2%, realizing flexible bandwidth control, improving design efficiency and performance, and meeting the bandwidth requirements in different application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 Schematic diagram of the upper metal layer structure of the super surface unit of the present invention; Figure 2 Schematic diagram of the structure of the lower metal layer of the super surface unit of the present invention; Figure 3 is a graph of the transmission amplitude coefficient of the metasurface unit of the present invention when the incident azimuth angle increases from 0 degrees to 90 degrees; Figure 4 A network framework diagram of the method for adjusting the bandwidth of a metasurface according to the present invention; Figure 5 The training network loss diagram of the metasurface bandwidth adjustable method of the present invention; Figure 6 It is a prediction result diagram of the forward prediction network of the present invention; Figure 7 This is a result diagram of the inverse design network of the metasurface bandwidth adjustable method of the present invention. DETAILED DESCRIPTION

[0017] The technical solution of the present invention is further described below in conjunction with the accompanying drawings. Example

[0018] like Figure 1-2 As shown: a broadband transmission polarization conversion metasurface unit, including a dielectric substrate 1; the material of the dielectric substrate is a common material such as F4B plate; the metasurface unit is set to a square, the side length of the metasurface unit is approximately equal to seven twentieths of the wavelength of the center working frequency, and the thickness of the metasurface unit is slightly less than one tenth of the wavelength of the center working frequency.

[0019] The dielectric substrate 1 is provided with an upper metal layer, and the dielectric substrate 1 is provided with a lower metal layer. The upper metal and the lower metal layer both include an incomplete rectangular metal patch 2 and a first metal branch 3, a second metal branch 4, a third metal branch 5, a fourth metal branch 6, a fifth metal branch 7, a sixth metal branch 8, a seventh metal branch 9, and an eighth metal branch 10. The incomplete rectangular metal patch 2 is obtained by cutting the corners of the complete rectangular metal patch and digging out four rectangular grooves inside it at the same time. The lower metal is a translation mirror image of the upper metal. The upper and lower metal layers are connected by eight metal through holes, and the end of each metal through hole is connected to the incomplete rectangular metal patch 2 through a metal branch. Among them, the incomplete rectangular metal patch 2 is obtained by cutting the corners of the complete rectangular metal patch and digging out four rectangular grooves inside it. This operation has the effect of widening the bandwidth. The upper and lower metal layers of the dielectric substrate are connected by eight metal through holes, and the ends of the metal through holes are connected to the incomplete rectangular metal patch 2 through metal branches. This method can effectively transmit electromagnetic waves. When linearly polarized waves with incident azimuth angles of 0°, 30°, 60°, and 90° are irradiated onto the unit, linearly polarized waves with incident azimuth angles of 90°, 120°, 150°, and 180° will be transmitted within the working bandwidth.

[0020] like Figure 3 As shown in the figure, when linear polarized waves with incident azimuth angles of 0°, 30°, 60°, and 90° are irradiated onto the unit, the working bandwidth is 11.6-14.3 GHz, and the relative bandwidth reaches 20%. Within the working bandwidth, the transmission amplitude of the unit is higher than 0.9, with the highest value reaching 0.98, and remains stable as the incident azimuth angle of the electromagnetic wave changes. The central working frequency and working bandwidth of the metasurface unit are adjustable. Example

[0021] A method for designing the bandwidth of a polarization converter based on deep learning. The transmission polarization conversion metasurface unit is forward predicted and reversely designed based on a deep learning framework. The structural block diagram of the neural network is shown in Figure 4 , the method comprises the following steps: Step 1: Prepare a data set: Determine the structural parameters based on the above-mentioned broadband metasurface unit, divide the range of each structural parameter and perform random sampling, use electromagnetic wave numerical simulation software to perform simulation, and obtain the electromagnetic response spectrum of the metasurface unit under different structural parameters. Collect each set of simulated structural parameters and the corresponding electromagnetic response spectrum, organize them into the final data set, and finally divide the data set into training set, validation set and test set according to appropriate proportions.

[0022] The metasurface unit used in Example 2 of the present invention is the broadband transmission polarization metasurface unit of Example 1, so the electromagnetic response spectrum is the S21 parameter curve. The frequency range selected here is 10-16 GHz, which is divided into 601 points. The four structural parameters that have the greatest impact on the spectrum are screened out, namely the side length of the dielectric substrate, the thickness of the dielectric substrate, the width w of the rectangular groove, and the distance r from the metal through hole to the geometric center of the model.

[0023] In addition, a total of 3200 sets of data were collected in this embodiment 2 of the present invention, and the training set, validation set and test set were divided into 8:1:1. The hyperparameters for training the forward prediction network and the reverse design network were set as follows: the initial learning rate was 0.01, the training network was iterated 500 times, the Adam optimizer was used, and the loss function was defined as the mean square error between the network prediction value and the simulation value.

[0024] Step 2: Use the collected data set and based on the deep learning network framework to train a forward prediction network from structural parameters to the corresponding electromagnetic response spectrum; after the network training is completed, the validation set is needed to test whether the network is overfitting or underfitting. If overfitting or underfitting occurs, the network parameters need to be adjusted until the network no longer overfits, or the overfitting and Loss values ​​are very small. The network contains an input layer, multiple hidden layers and an output layer. Different numbers of hidden layers need to be set according to different tasks and data characteristics. The number of hidden layers also has a great influence on the Loss. Here we compare the Loss when the number of layers is 3, 4 and 5, respectively. Figure 5 It can be seen that when the number of hidden layers is 4, the Loss value is the smallest. After the network training is completed, four sets of data need to be randomly selected from the test set to check whether the electromagnetic response spectrum predicted by the network is consistent with the electromagnetic response spectrum of the electromagnetic wave numerical simulation tool, such as Figure 6 As shown in the figure, the predicted spectra and simulated spectra of the four sets of data can fit well, and the prediction is more accurate.

[0025] Step 3: Interchange the input and output of the data set, and train the reverse design network from electromagnetic response spectrum to structural parameters based on the deep learning framework. Take the different relative bandwidths at different frequencies as input, and the network automatically generates the structural parameters of the corresponding metasurface unit. Put the generated structural parameters into the forward prediction network and electromagnetic simulation tool respectively, and obtain the corresponding predicted and simulated electromagnetic response spectra. Select the target spectra with center frequencies of 12 GHz, 13 GHz and 14 GHz and relative bandwidths of 10%, 15% and 20% respectively, as shown in Figure 3. Figure 7 As shown in the figure, the predicted curve and the simulation curve are basically consistent and can both meet our target requirements. The error between the S21 relative bandwidth obtained by electromagnetic simulation and the target relative bandwidth is controlled within 2%, which is a small error and achieves the purpose of adjustable bandwidth.

[0026] In summary, the present invention designs a broadband transmission polarization metasurface unit, the relative bandwidth can reach 20%, and the unit has the advantages of small size and thin thickness. A method for designing the bandwidth of a polarization converter based on deep learning is provided. Based on the deep learning framework, the broadband transmission polarization metasurface unit is forward predicted and reversely designed. The relative bandwidth obtained by the reverse design has an error of less than 2% with the target relative bandwidth, which realizes flexible bandwidth regulation, improves design efficiency and performance, and meets the bandwidth requirements in different application scenarios.

Claims

1. A broadband transmission polarization metasurface unit, characterized in that: The metasurface unit is arranged in a square shape, and comprises a dielectric substrate (1). The dielectric substrate (1) is provided with an upper metal layer and a lower metal layer. The upper metal layer and the lower metal layer both comprise an incomplete rectangular metal patch (2) and a plurality of metal branches (3). The incomplete rectangular metal patch (2) is obtained by cutting corners of a pair of complete rectangular metal patches and simultaneously digging out four rectangular grooves therein, and the lower metal layer is a translational mirror image of the upper metal layer. The upper and lower metal layers are connected by eight metal through holes, and the end of each metal through hole is connected to the incomplete rectangular metal patch (2) via a metal branch.

2. The broadband transmission polarization conversion metasurface unit according to claim 1, characterized in that: The side length of the metasurface unit is approximately equal to seven twentieths of the wavelength of the central operating frequency, and the thickness of the metasurface unit is less than one tenth of the wavelength of the central operating frequency.

3. The broadband transmission polarization conversion metasurface unit according to claim 1, characterized in that: There are 7 metal branches.

4. The broadband transmission polarization conversion metasurface unit according to claim 1, characterized in that: The dielectric substrate (1) is an F4B plate.

5. The broadband transmission polarization conversion metasurface unit according to claim 1, characterized in that: When a linearly polarized wave with any incident azimuth angle φ is irradiated onto the metasurface unit, the transmitted linearly polarized wave is rotated 90° and remains stable as the incident azimuth angle of the electromagnetic wave changes.

6. The broadband transmission polarization conversion metasurface unit according to claim 1, characterized in that: The maximum transmission amplitude of the metasurface unit is higher than 0.95, and the maximum relative bandwidth is 20%.

7. A method for designing polarization converter bandwidth based on deep learning, characterized in that: The method performs forward prediction and reverse design of transmission-polarization metasurface units based on deep learning.

8. The method for designing polarization converter bandwidth based on deep learning according to claim 7, characterized in that: The following steps are involved: (S1) Collecting data sets: determining the structural parameters of the broadband metasurface unit, dividing the structural parameters into ranges and performing random sampling according to uniform distribution, using electromagnetic wave numerical simulation tools to obtain the electromagnetic response spectrum corresponding to each group of structural parameters, and collecting the simulated structural parameters and the corresponding electromagnetic response spectrum and organizing them into a data set; (S2) Based on a deep learning framework, training a forward prediction neural network, wherein the input of the neural network is the structural parameters of the metasurface unit, and the output is the electromagnetic response spectrum of the corresponding metasurface unit; (S3) Based on the deep learning framework, the reverse design network is trained, and the different relative bandwidths of different frequency points are taken as input. The network automatically generates the structural parameters of the corresponding metasurface units; the generated structural parameters are respectively put into the forward prediction network and the electromagnetic simulation tool to obtain the corresponding predicted and simulated electromagnetic response spectra.

9. The method for designing polarization converter bandwidth based on deep learning according to claim 7, characterized in that: The electromagnetic response spectrum is a linear S21 parameter curve.

Citation Information

Patent Citations

  • Polarization-insensitive phase-gradient metasurface

    CN105552564A

  • Irregular polygon metasurface wave absorber with adjustable visible light and near-infrared band filtering window and design method of irregular polygon metasurface wave absorber

    CN117810706A

  • Metasurface design method based on deep learning and hologram weight partition regulation and control

    CN117933042A

  • Transmission type ultra-wideband polarization insensitive cross polarization converter

    CN117977213A

  • On-chip polarization detection and polarimetric imaging

    US20210190593A1

Cited By

  • Design method of frequency selection type reflection-transmission composite polarization converter

    CN118899673A

  • A design method for a frequency-selective reflection-transmission composite polarization converter

    CN118899673B

  • Transflective polarization conversion switchable reconfigurable intelligent metasurface device and design method thereof

    CN121189193A