The Design Of Compact Operators Based On The Single-layer Metasurface

In today’s environment of exploding information growth in the Internet,communication,and other fields,a huge amount of data needs to be processed and computed in real-time,which challenges the processing speed and power consumption performance of traditional computing solutions.Recently,metamaterial-based optical computing solutions offer the possibility to break through the limitations of traditional electronic and optical computing systems in terms of processing speed and device integration.Such an approach can process data at the speed of light with low power consumption,while having compact volume and strong parallel data processing capabilities.Based on the theory of phase-tunable singlelayer metasurfaces and optical diffractive neural networks,this paper designs compact logic and trigonometric operators operating at the microwave frequency for the two core problems in computing systems: logic and trigonometric operations,respectively.The proposed strategy can run at the speed of light and has powerful parallel processing capability,which greatly improves the speed of data processing of the computing system.In addition,the design of the single-layer artificial electromagnetic structure allows operators to be better integrated with other optical devices and systems and also simplifies the processing process.The above advantages offer a possible route for the realization of ultra-high-speed photonic computing and signal processing systems.Firstly,the deep learning algorithm is carefully analyzed and studied in this paper,and a fully connected diffractive neural network with a single hidden layer is proposed to map the working process of optical logic operators by combining the scattering characteristics of electromagnetic waves in space.In this paper,ten incident light signals representing all the operations of the three basic logic operations(AND,OR,and NOT)independently are designed as the inputs to the diffractive neural network.Two-dimensional images characterizing "0" and "1" with different output logical information are also designed as the target outputs of the neural network.After building a diffractive neural network that can map the optical logic operation process,iterative training is performed and good training results are obtained.Then,a singlelayer metasurface composed of P-B(Pancharatnam-Berry)phase elements is designed based on the phase bias parameters obtained from the training,and the corresponding system simulation and physical experiment are performed.The simulated and measured results are in good agreement with the calculated results,and both of them are also consistent with the theoretical values.Finally,in order to verify the practical value and scalability of the proposed optical operation strategy,the realization of the optical trigonometric function operation system is also explored and studied based on the design process of the optical diffractive neural network mentioned before.In this paper,a multimode composite input and output strategy are designed for optical trigonometric operations,which can realize four basic trigonometric operations(sin,cos,tan,cot)in one period with only a single hidden layer.The designed optical trigonometric operator has been rigorously verified by theoretical calculations,simulation experiments,and physical testing sessions,further proving the practicality and reliability of the single-layer metasurface-based optical computing strategy designed in this paper.