| Metasurfaces are artificial electromagnetic structure composed of subwavelength-scale elements,which are usually in periodic or non-periodic two-dimensional arrangements.Metasurfaces have structural advantages such as low profile,low loss,easy integration and unique electromagnetic properties.The special electromagnetic properties have led to perfect lenses,cloaking and illusion devices,super-resolution electromagnetic imaging,as well as other novel research results.The concept of digital coding metasurfaces further enhances the electromagnetic control capability of metamaterials by means of digitalization.The software-defined metamaterial proposed in this paper is based on digital coding metasurfaces,combined with intelligent algorithms such as deep learning,and gradually builds a series of mapping relationships from the physical structure of the unit to the electromagnetic response of the element,and then to the pattern of the metasurface array.Using the software-defined metasurface,the user can flexibly switch the electromagnetic function of the metasurface according to the complex environment.The beam deflection function ensures the uniqueness of the direction of electromagnetic information transmission.The orbital angular momentum technology provides secure encryption protection for the transmission process of holographic influences.Low RCS diffuse scattering function provides technical support for the electromagnetic stealth of the device,making software-defined metamaterials of great engineering significance in the fields of information encryption and information security.The main work and ideas of this thesis are summarized as follows:Aiming at the element custom design of metasurfaces,a metasurface element custom design system based on deep neural network is proposed,which includes forward deep neural network and reverse search intelligent algorithm.The forward-network takes the parameters of the element structure and the material of the dielectric layer as input,and predicts the reflection phase-frequency characteristics of the structural element.The error between the predicted value and the simulation result is only 0.0561° per frequency point in X-band.The reverse search intelligent algorithm is based on the forward network,and combines the depth-first algorithm with the prediction results of the forward network to realize the structure parameters of a group of elements defined by the user from 1-bit coding to 4-bit coding.Aiming at the custom design of electromagnetic functions of digital coding metasurfaces,an array custom design system based on an intelligent coding algorithm is proposed.Firstly,an intelligent coding algorithm based on the generalized Snell’s law and metasurface convolution coding theorem is designed,which combines the analysis of element amplitude to enhance the forecast capability of the coding algorithm.Then,based on the intelligent coding algorithm,a custom array design system of digital metasurface is proposed,which has a high degree of freedom parameter selection,including beam deflection under the incident plane wave and spherical wave.Afterwards,a reflective coding metasurface unit operating at9.2GHz is proposed,which consists of a seven-layer structure,indicating that the metal layer mainly consists of a rectangular metal patch,two small metal patches and two diodes.The turn-on and turn-off of 2-bit reflection phase control is realized.Finally,a software custom design system for coding metasurfaces is proposed,which realizes the common electromagnetic functions of single-beam deflection,double-beam deflection,multi-beam deflection,orbital angular momentum encoding,and low RCS scattering. |
