Design method and device of polarization beam splitting metasurface and beam splitter

Abstract

The application provides a design method and device of a polarized light beam splitting metasurface and a light beam splitter, and belongs to the optical technical field.The method comprises the following steps: obtaining a far-field intensity distribution image; performing feature extraction and reconstruction on the far-field intensity distribution image to obtain a right circular polarization phase distribution map and a left circular polarization phase distribution map of the far-field intensity distribution image; inputting the right circular polarization phase distribution map and the left circular polarization phase distribution map into a preset angular spectrum propagation model to recover a metasurface phase distribution corresponding to the far-field intensity distribution image based on the angular spectrum propagation model; and performing polarized light beam splitting metasurface design based on the metasurface phase distribution.The technical scheme can improve the design efficiency of the polarized light beam splitting metasurface.

Description

Technical Field

[0001] This application relates to the field of optical technology, and in particular to a design method, apparatus and beam splitter for a polarizing beam splitting metasurface. Background Technology

[0002] The basic building blocks of advanced photonic systems can manipulate light according to its polarization state. Although components such as bulk optical elements can also achieve the function of manipulating light according to its polarization state, the increasing demand for miniaturization, integration and multifunctionality in modern photonic devices has led to a shift in research on optical components towards developing more compact and multifunctional alternatives, such as metasurface polarization beam splitters.

[0003] In related technologies, metasurfaces, as a two-dimensional subwavelength optical nanoresonator array, provide a new platform for manipulating light at the nanoscale. However, current metasurface designs still suffer from low design efficiency in the process of miniaturization and integration of optical components. Summary of the Invention

[0004] The main objective of this application is to propose a design method, apparatus, and beam splitter for a polarized beam splitting metasurface, aiming to improve the design efficiency of polarized beam splitting metasurfaces.

[0005] To achieve the above objectives, a first aspect of this application proposes a method for designing a polarizing beam splitting metasurface, the method comprising:

[0006] Obtain the far-field intensity distribution image;

[0007] Feature extraction and reconstruction are performed on the far-field intensity distribution image to obtain a right-handed circular polarization phase distribution map and a left-handed circular polarization phase distribution map of the far-field intensity distribution image;

[0008] The right-hand circular polarization phase distribution map and the left-hand circular polarization phase distribution map are input into a preset angular spectrum propagation model to recover the metasurface phase distribution corresponding to the far-field intensity distribution image based on the angular spectrum propagation model.

[0009] The design of a polarization beam splitting metasurface is based on the phase distribution of the metasurface.

[0010] In some embodiments, the step of inputting the right-hand circularly polarized phase distribution map and the left-hand circularly polarized phase distribution map into a preset angular spectrum propagation model, so as to recover the metasurface phase distribution corresponding to the far-field intensity distribution image based on the angular spectrum propagation model, includes:

[0011] The right-hand circular polarization phase distribution map and the left-hand circular polarization phase distribution map are used as model input data to input into a preset angular spectrum propagation model;

[0012] Based on the angular spectrum propagation model, far-field intensity prediction calculation is performed using the model input data to obtain the predicted far-field intensity.

[0013] Based on a preset intensity-phase gradient relationship, the mean square error between the predicted far-field intensity and the input far-field intensity is optimized in reverse to obtain the metasurface phase distribution corresponding to the far-field intensity distribution image; the input far-field intensity is the far-field intensity of the far-field intensity distribution image.

[0014] In some embodiments, the method further includes:

[0015] The preset angular spectrum propagation model is transformed into a differentiable neural network layer, and the angular spectrum propagation model is set as a partially embedded training loop architecture to construct a deep learning framework driven by physical information.

[0016] The angular spectrum propagation model is the physical information driving layer in the deep learning framework; the deep learning framework is used to train a neural network with the far-field intensity distribution image as input data, and outputs the metasurface phase distribution corresponding to the far-field intensity distribution image recovered based on the angular spectrum propagation model.

[0017] In some embodiments, the step of extracting and reconstructing features from the far-field intensity distribution image to obtain a right-handed circularly polarized phase distribution map and a left-handed circularly polarized phase distribution map of the far-field intensity distribution image includes:

[0018] The far-field intensity distribution image is input into a preset convolutional neural network architecture; the convolutional neural network architecture includes a shared encoder and a dual decoder branch;

[0019] Based on the shared encoder and the dual decoder branch, feature extraction and reconstruction are performed on the far-field intensity distribution image to obtain the right-hand circular polarization phase distribution map and the left-hand circular polarization phase distribution map of the far-field intensity distribution image.

[0020] In some embodiments, the step of extracting and reconstructing features from the far-field intensity distribution image based on the shared encoder and the dual decoder branch to obtain a right-handed circularly polarized phase distribution map and a left-handed circularly polarized phase distribution map of the far-field intensity distribution image includes:

[0021] Based on the shared encoder, hierarchical features are obtained by feature extraction from the far-field intensity distribution image;

[0022] Based on the first branch of the dual decoder branch, the right-hand circular polarization phase distribution is reconstructed using the hierarchical features as input to obtain the right-hand circular polarization phase distribution map of the far-field intensity distribution image; and based on the second branch of the dual decoder branch, the left-hand circular polarization phase distribution is reconstructed using the hierarchical features as input to obtain the left-hand circular polarization phase distribution map of the far-field intensity distribution image.

[0023] In some embodiments, the design of the polarization beam splitting metasurface based on the metasurface phase distribution includes:

[0024] The phase distribution of the metasurface is converted into nanopillar structure parameters, and a nanopillar superlattice is designed based on the nanopillar structure parameters to construct a polarization beam splitting metasurface.

[0025] To achieve the above objectives, a second aspect of this application provides a design apparatus for a polarizing beam splitting metasurface, the apparatus comprising:

[0026] The acquisition module is used to acquire far-field intensity distribution images;

[0027] The image processing module is used to extract and reconstruct features from the far-field intensity distribution image to obtain a right-handed circular polarization phase distribution map and a left-handed circular polarization phase distribution map of the far-field intensity distribution image.

[0028] The intensity phase mapping module is used to input the right-hand circular polarization phase distribution map and the left-hand circular polarization phase distribution map into a preset angular spectrum propagation model, so as to recover the metasurface phase distribution corresponding to the far-field intensity distribution image based on the angular spectrum propagation model;

[0029] A metasurface design module is used to design a polarization beam splitting metasurface based on the phase distribution of the metasurface.

[0030] To achieve the above objectives, a third aspect of the present application proposes a metasurface beam splitter, wherein the metasurface of the metasurface beam splitter is constructed based on the design method of the polarization beam splitting metasurface described in the first aspect.

[0031] To achieve the above objectives, a fourth aspect of the present application provides an electronic device, the electronic device including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method described in the first aspect.

[0032] To achieve the above objectives, a fifth aspect of the present application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described in the first aspect.

[0033] To achieve the above objectives, a sixth aspect of the present application provides a computer program product storing a computer program that, when executed by a processor, implements the method described in the first aspect.

[0034] The present application provides a method, apparatus, metasurface beam splitter, electronic device, computer-readable storage medium, and computer program product for designing polarized beam splitting metasurfaces. This involves acquiring a far-field intensity distribution image; extracting and reconstructing features from the far-field intensity distribution image to obtain a right-hand circularly polarized phase distribution map and a left-hand circularly polarized phase distribution map; inputting the right-hand circularly polarized phase distribution map and the left-hand circularly polarized phase distribution map into a preset angular spectrum propagation model to recover the metasurface phase distribution corresponding to the far-field intensity distribution image based on the angular spectrum propagation model; and designing a polarized beam splitting metasurface based on the metasurface phase distribution.

[0035] Thus, in this embodiment of the application, after obtaining the right-hand circular polarization phase distribution map and the left-hand circular polarization phase distribution map of the far-field intensity distribution image, the metasurface phase distribution corresponding to the far-field intensity distribution image is determined based on the right-hand circular polarization phase distribution map and the left-hand circular polarization phase distribution map using the angular spectrum propagation model. This not only enables accurate mapping from far-field intensity to phase distribution, but also effectively reduces the computational resources required for metasurface design, thereby improving the design efficiency of polarized beam splitting metasurfaces. Attached Figure Description

[0036] Figure 1 A flowchart illustrating the steps of the design method for a polarizing beam splitting metasurface provided in this application in some embodiments;

[0037] Figure 2 for Figure 1 A detailed flowchart of step S102;

[0038] Figure 3 for Figure 2 A detailed flowchart of step S202;

[0039] Figure 4 The design method for polarization beam splitting metasurfaces provided in the embodiments of this application includes, in some embodiments, a schematic diagram of a deep learning-based convolutional neural network U-Net architecture.

[0040] Figure 5 for Figure 1 A detailed flowchart of step S103;

[0041] Figure 6The diagram shows a deep learning-based physical information feedback framework involved in some embodiments of the design method for the polarization beam splitting metasurface proposed in this application.

[0042] Figure 7 This is a schematic diagram illustrating the construction of a metasurface through complete Jones matrix decoupling achieved by geometric parameter engineering and Pancharatnam-Berry (PB) phase modulation, as described in an embodiment of this application.

[0043] Figure 8 A schematic diagram of the design device for a polarizing beam splitting metasurface provided in the embodiments of this application;

[0044] Figure 9 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0046] It should be noted that although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first," "second," etc., in the specification, claims, and the aforementioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0047] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0048] First, the overall concept of the embodiments of this application will be explained.

[0049] Elements capable of manipulating light based on its polarization state are fundamental building blocks of advanced photonic systems. While traditional bulk optical elements can achieve these functions, research has shifted towards more compact and multifunctional alternatives as modern photonic devices demand miniaturization, integration, and versatility. Metasurfaces, as two-dimensional subwavelength optical nanoresonator arrays, offer a new platform for manipulating light at the nanoscale. However, designing metasurface polarization beamsplitters that simultaneously achieve high efficiency, large extinction ratios, and arbitrary beam directionality remains challenging.

[0050] This application proposes a design method, device, and beam splitter for a polarization beam splitting metasurface. Based on a physical information-driven deep learning method, a metasurface beam splitter with arbitrary polarization multiplexing is designed. Complete Jones matrix decoupling can be achieved through silicon nanopillar superlattice design, thereby enabling independent control of orthogonal polarization states and achieving high-efficiency, high extinction ratio, and arbitrary beam direction polarization beam splitting function.

[0051] The design method for polarization beam splitting metasurfaces proposed in this application is based on a physics-driven deep learning framework to design metasurface beam splitters for arbitrary polarization. Through three core technology components—an improved U-Net architecture, an angular harmonic propagation model, and silicon nanopillar superlattice design—it effectively solves the problems of low efficiency, poor extinction ratio, and difficulty in directional control that exist in the miniaturization and integration of traditional optical elements.

[0052] The core technical means of the design method of polarization beam splitting metasurface provided in this application include: (1) U-Net architecture optimized for optical characteristics, which achieves efficient feature extraction and reconstruction through encoder-decoder structure and skip connections; (2) embedding physical information into the angular harmonic propagation model of the deep learning framework to achieve accurate mapping from far-field intensity to phase distribution; (3) silicon nanopillar superlattice design based on geometric parameter engineering to achieve precise control of Pancharatnam-Berry phase (a phase concept in quantum mechanics that describes the phase rotation of light in an optical device).

[0053] Through the above technical means, the embodiments of this application can achieve the following technical effects: (1) Design efficiency is significantly improved and the computing resource requirements are reduced by more than 50%; (2) A high extinction ratio of 34.11dB is achieved, which is about 15% higher than related technologies; (3) A high efficiency of 63.91% is achieved; (4) It supports precise control of arbitrary polarization states and its application scope is extended to multiple application fields such as quantum optics and biological imaging.

[0054] Next, based on the overall concept of the embodiments of this application described above, specific embodiments of the design method, apparatus, metasurface beam splitter, electronic device, computer-readable storage medium, and computer program product of the polarization beam splitter provided in the embodiments of this application are proposed. First, the various specific embodiments of the design method of the polarization beam splitter metasurface in the embodiments of this application are described in detail.

[0055] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0056] Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0057] It should be noted that in all specific embodiments of this application, whenever processing is required based on user information, user behavior data, user historical data, and user location information—data related to user identity or characteristics—user permission or consent will be obtained first. Furthermore, the collection, use, and processing of this data will comply with relevant laws, regulations, and standards. In addition, when embodiments of this application require access to sensitive personal information of users, separate permission or consent from the user will be obtained through pop-ups or redirects to confirmation pages. Only after obtaining the user's separate permission or consent will the necessary user-related data for the proper functioning of these embodiments be acquired.

[0058] Furthermore, the design method for polarized beam splitting metasurfaces provided in this application can be applied to terminals, servers, or software running on either a terminal or a server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, or other terminal device; the server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application that implements the design method for polarized beam splitting metasurfaces, but is not limited to the above forms.

[0059] Alternatively, embodiments of this application can also be used in numerous general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer computer devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In a distributed computing environment, program modules can reside in local and remote computer storage media, including storage devices.

[0060] For ease of understanding and explanation, the following description will use the design method of the polarized beam splitting metasurface provided in the embodiments of this application for terminal devices as an example. The implementation of any of the above-described subject matter using the design method of the polarized beam splitting metasurface provided in the embodiments of this application can be carried out by referring to the process of the design method of the polarized beam splitting metasurface for terminal devices described below.

[0061] Please refer to Figure 1 , Figure 1 The schematic diagram shows the steps of the design method for a polarizing beam splitting metasurface provided in this application in some embodiments. It should be understood that, although... Figure 1 The figure shows the execution order of some method steps, but based on different design needs in practical applications, the design method for polarized beam splitting metasurfaces provided in this application embodiment can of course adopt a different execution order of method steps than that shown in the figure. That is, Figure 1 The order of the method steps shown does not constitute a limitation on the execution logic order of the design method for the polarization beam splitting metasurface provided in the embodiments of this application. Any other method based on this method is not limited to this method. Figure 1 Reasonable variations in the order of steps shown should be included within the protection scope of the design method for polarized beam splitting metasurfaces provided in the embodiments of this application.

[0062] like Figure 1 As shown, in some embodiments, the design method of the polarization beam splitting metasurface provided in this application may include, but is not limited to, steps S101 to S104.

[0063] Step S101: Obtain the far-field intensity distribution image.

[0064] During the design process of polarized beam splitting metasurface, the terminal device first acquires a far-field intensity distribution image as input data for the entire polarized beam splitting metasurface design process.

[0065] In some embodiments, the terminal device may acquire a target far-field intensity distribution image, including information such as the beam splitting point location, beam splitting intensity ratio, and polarization state, as input data.

[0066] Step S102: Perform feature extraction and reconstruction on the far-field intensity distribution image to obtain the right-hand circular polarization phase distribution map and the left-hand circular polarization phase distribution map of the far-field intensity distribution image.

[0067] After acquiring the far-field intensity distribution image as input data, the terminal device uses a preset U-Net architecture to extract and reconstruct features from the far-field intensity distribution image based on the encoder-decoder structure and skip connections, thereby obtaining the right-hand circular polarization phase distribution map and the left-hand circular polarization phase distribution map of the far-field intensity distribution image.

[0068] It should be noted that the U-Net architecture features a shared encoder and dual decoder branches, enabling simultaneous optimization of the phase distribution of both right-hand circularly polarized (RCP) and left-hand circularly polarized (LCP) light. Specifically, the phase distribution of RCP is a right-hand circularly polarized phase distribution map, while the phase distribution of LCP is a left-hand circularly polarized phase distribution map.

[0069] Step S103: Input the right-hand circular polarization phase distribution map and the left-hand circular polarization phase distribution map into a preset angular spectrum propagation model to recover the metasurface phase distribution corresponding to the far-field intensity distribution image based on the angular spectrum propagation model.

[0070] After obtaining the right-hand circularly polarized phase distribution map and the left-hand circularly polarized phase distribution map of the far-field intensity distribution image, the terminal device further inputs the right-hand circularly polarized phase distribution map and the left-hand circularly polarized phase distribution map into a preset angular spectrum propagation model. Based on the angular spectrum propagation model, the right-hand circularly polarized phase distribution map and the left-hand circularly polarized phase distribution map are used to perform an accurate mapping from far-field intensity to phase distribution, so as to recover the metasurface phase distribution corresponding to the far-field intensity distribution image.

[0071] Step S104: Design a polarization beam splitting metasurface based on the phase distribution of the metasurface.

[0072] After the terminal device obtains the metasurface phase distribution corresponding to the far-field intensity distribution image, it can design a polarization beam splitter metasurface based on the metasurface phase distribution, thereby obtaining the designed metasurface beam splitter.

[0073] In this embodiment, a far-field intensity distribution image is acquired via a terminal device as input data for the entire polarization beam splitting metasurface design process. Then, using a pre-defined U-Net architecture, feature extraction and reconstruction are performed on the far-field intensity distribution image based on an encoder-decoder structure and skip connections, resulting in right-hand circularly polarized phase distribution maps and left-hand circularly polarized phase distribution maps. Next, these right-hand and left-hand circularly polarized phase distribution maps are input into a pre-defined angular spectrum propagation model. Based on this model, a precise mapping from far-field intensity to phase distribution is performed using the right-hand and left-hand circularly polarized phase distribution maps to recover the metasurface phase distribution corresponding to the far-field intensity distribution image. Finally, a polarization beam splitting metasurface is designed based on this metasurface phase distribution, resulting in the designed metasurface beam splitter.

[0074] Thus, in this embodiment of the application, after obtaining the right-hand circular polarization phase distribution map and the left-hand circular polarization phase distribution map of the far-field intensity distribution image, the metasurface phase distribution corresponding to the far-field intensity distribution image is determined based on the right-hand circular polarization phase distribution map and the left-hand circular polarization phase distribution map using the angular spectrum propagation model. This not only enables accurate mapping from far-field intensity to phase distribution, but also effectively reduces the computational resources required for metasurface design, thereby improving the design efficiency of polarized beam splitting metasurfaces.

[0075] Please refer to Figure 2 , Figure 2 for Figure 1 A detailed flowchart of step S102.

[0076] In some embodiments, such as Figure 2 As shown, step S102 above, which involves extracting and reconstructing features from the far-field intensity distribution image to obtain a right-handed circular polarization phase distribution map and a left-handed circular polarization phase distribution map of the far-field intensity distribution image, may include, but is not limited to, steps S201 and S202 as shown below.

[0077] Step S201: Input the far-field intensity distribution image into a preset convolutional neural network architecture; the convolutional neural network architecture includes a shared encoder and a dual decoder branch.

[0078] During the feature extraction and reconstruction of the far-field intensity distribution image, the terminal device inputs the far-field intensity distribution image into a preset convolutional neural network architecture, thereby extracting deep features of the image through the encoder of the convolutional neural network architecture. Each encoding block contains two 3×3 convolutional layers and one max-pooling layer.

[0079] It should be noted that the default convolutional neural network architecture is the U-Net architecture described above. The U-Net architecture is a pre-modified version of the U-Net network, featuring a shared encoder and dual decoder branches, which can simultaneously optimize the phase distribution of right-hand circularly polarized (RCP) and left-hand circularly polarized (LCP) light.

[0080] Step S202: Based on the shared encoder and the dual decoder branch, feature extraction and reconstruction are performed on the far-field intensity distribution image to obtain the right-hand circular polarization phase distribution map and the left-hand circular polarization phase distribution map of the far-field intensity distribution image.

[0081] After the terminal device inputs the far-field intensity distribution image into the U-Net architecture, the U-Net architecture extracts and reconstructs the features of the far-field intensity distribution image through a shared encoder and dual decoder branches. By simultaneously optimizing the phase distribution of right-hand circularly polarized light (RCP) and left-hand circularly polarized light (LCP), the right-hand circularly polarized phase distribution map and the left-hand circularly polarized phase distribution map of the far-field intensity distribution image are obtained.

[0082] Please refer to Figure 3 , Figure 3 for Figure 2 A detailed flowchart of step S202.

[0083] In some embodiments, such as Figure 3 As shown, step S202 above may include, but is not limited to, steps S301 and S302 as shown below.

[0084] Step S301: Extract hierarchical features from the far-field intensity distribution image based on the shared encoder.

[0085] When the terminal device performs feature extraction and reconstruction of the far-field intensity distribution image based on the U-Net architecture, it first performs deep feature extraction on the far-field intensity distribution image through the shared encoder in the U-Net architecture to obtain the hierarchical features of the far-field intensity distribution image.

[0086] Step S302: Based on the first branch of the dual decoder branch, the right-hand circular polarization phase distribution is reconstructed with the hierarchical features as input to obtain the right-hand circular polarization phase distribution map of the far-field intensity distribution image; and based on the second branch of the dual decoder branch, the left-hand circular polarization phase distribution is reconstructed with the hierarchical features as input to obtain the left-hand circular polarization phase distribution map of the far-field intensity distribution image.

[0087] After extracting the hierarchical features of the far-field intensity distribution image through the shared encoder in the U-Net architecture, the terminal device further reconstructs the right-hand circular polarization phase distribution using these hierarchical features as input through the first branch of the dual decoder branch in the U-Net architecture, thus obtaining the right-hand circular polarization phase distribution map of the far-field intensity distribution image. Simultaneously, the terminal device reconstructs the left-hand circular polarization phase distribution using these hierarchical features as input through the second branch of the dual decoder branch in the U-Net architecture, thus obtaining the left-hand circular polarization phase distribution map of the far-field intensity distribution image.

[0088] For example, the U-Net architecture can consist of three main components: an encoder path dedicated to feature extraction, a bottleneck layer for feature transformation, and a dual decoder path responsible for phase reconstruction. Figure 4 As shown, the shared encoder path (yellow block) in the U-Net architecture first extracts hierarchical features from the far-field intensity distribution image, supplemented by intermediate pooling layers (gray blocks). Then, because the U-Net architecture integrates dual decoder branches (each branch consisting of an upsampling convolutional layer (brown block) and a feature-refining convolutional block (red block), it can simultaneously reconstruct the phase distributions of right-hand circular polarization (RCP, upper part) and left-hand circular polarization (LCP, bottom part), i.e., the right-hand circular polarization phase distribution map and the left-hand circular polarization phase distribution map of the far-field intensity distribution image. Furthermore, the U-Net architecture integrates skip connections (horizontal arrows) to preserve spatial details by connecting the corresponding encoder and decoder layers. The feature dimensions of each layer are represented numerically. The outputs (RCP and LCP) demonstrate the ability to recover complementary phase modes from a single target intensity image (Target, i.e., the far-field intensity distribution image), achieving polarization selection functionality through physical information optimization.

[0089] In some embodiments, the U-Net architecture employs a systematic mathematical formulation, wherein the encoding path involves successive convolution and pooling operations, and its mathematical formula can be expressed as:

[0090]

[0091] Here, el represents the feature map of the l-th layer of the encoding path, * represents the convolution operation, and σ is the ReLU activation function. This is a 2×2 max pooling operation. l，RCP and d l，LCP Let represent the feature map of the l-th layer of the dual decoder path, [,] represent the feature connections along the channel dimension, and U represents upsampling using a 2×2 kernel and a transposed convolution with a stride of 2.

[0092] Furthermore, the U-Net architecture performs feature transformation at the network bottleneck layer, connecting the encoder and decoder, as expressed mathematically below:

[0093] f bottleneck =σ(W bottleneck *e L +b bottleneck ),

[0094] Here, L represents the depth of the encoder path. The dual convolutional blocks within the decoder gradually reduce feature complexity, ensuring a smooth transition from high-level features to detailed phase maps. Features are reconstructed through the decoder; each decoding block contains an upsampling layer, two 3×3 convolutional layers, and skip connections. Thus, the U-Net architecture can output phase distribution maps of RCP and LCP polarized light.

[0095] In this embodiment, by adopting the above-described U-Net architecture, the hierarchical features of the far-field intensity distribution image can be captured efficiently. At the same time, the independent synchronous reconstruction of the right-hand circular polarization (RCP) and left-hand circular polarization (LCP) phase distributions can be achieved through a single forward propagation, resulting in the right-hand circular polarization phase distribution map and the left-hand circular polarization phase distribution map of the far-field intensity distribution image. This significantly optimizes the design process.

[0096] Please refer to Figure 5 , Figure 5 for Figure 1 A detailed flowchart of step S103.

[0097] In some embodiments, such as Figure 5 As shown, step S103 above: inputting the right-hand circular polarization phase distribution map and the left-hand circular polarization phase distribution map into a preset angular spectrum propagation model, so as to recover the metasurface phase distribution corresponding to the far-field intensity distribution image based on the angular spectrum propagation model, may include, but is not limited to, steps S501 to S503 as shown below.

[0098] Step S501: Input the right-hand circular polarization phase distribution map and the left-hand circular polarization phase distribution map as model input data into the preset angular spectrum propagation model.

[0099] Step S502: Based on the angular spectrum propagation model, perform far-field intensity prediction calculation using the model input data to obtain the predicted far-field intensity.

[0100] Step S503: Based on the preset intensity-phase gradient relationship, perform reverse optimization on the mean square error between the predicted far-field intensity and the input far-field intensity to obtain the metasurface phase distribution corresponding to the far-field intensity distribution image; the input far-field intensity is the far-field intensity of the far-field intensity distribution image.

[0101] After obtaining the right-hand circularly polarized phase distribution map and the left-hand circularly polarized phase distribution map of the far-field intensity distribution image, the terminal device further inputs these maps as model input data into a preset angular spectrum propagation model. The angular spectrum propagation model then uses this input data to perform far-field intensity prediction calculations, obtaining the predicted far-field intensity. Specifically, through Fourier transform, frequency domain multiplication, and inverse Fourier transform, the far-field intensity distribution corresponding to the model input data is calculated; this far-field intensity distribution is the predicted far-field intensity. Finally, using a pre-established intensity-phase gradient relationship, the mean square error between the predicted far-field intensity and the actual far-field intensity of the far-field intensity distribution image is optimized in reverse, thereby obtaining the metasurface phase distribution corresponding to the far-field intensity distribution image.

[0102] In some embodiments, the terminal device can construct an angular spectrum propagation model by transforming the traditional angular spectrum propagation formula into a differentiable neural network layer. The phase distribution predicted by the network (the right-hand circularly polarized phase distribution map and the left-hand circularly polarized phase distribution map of the far-field intensity distribution image obtained by extracting hierarchical features based on the U-Net architecture and reconstructing them) is used as input to the angular spectrum propagation layer. The angular spectrum propagation model calculates the corresponding far-field intensity distribution (predicted far-field intensity) through Fourier transform, frequency domain multiplication, and inverse Fourier transform. Furthermore, by establishing the gradient relationship between intensity and phase (intensity-phase gradient relationship), the phase distribution prediction can be directly optimized from the intensity error. Finally, by combining the mean square error and physical constraints (such as energy conservation and phase continuity), the prediction results can be ensured to conform to physical laws.

[0103] In some embodiments, the design method for a polarizing beam splitting metasurface provided in this application may also include, but is not limited to, the following steps:

[0104] The preset angular spectrum propagation model is transformed into a differentiable neural network layer, and the angular spectrum propagation model is set as a partially embedded training loop architecture to construct a deep learning framework driven by physical information.

[0105] It should be noted that the angular spectrum propagation model is the physical information-driven layer in the deep learning framework; the deep learning framework is used to train the neural network with the far-field intensity distribution image as input data, and outputs the metasurface phase distribution corresponding to the far-field intensity distribution image, which is recovered based on the angular spectrum propagation model.

[0106] Terminal devices can construct a physically-driven deep learning framework by transforming a preset angular spectrum propagation model into a differentiable neural network layer and setting the angular spectrum propagation model as a partially embedded training loop architecture. Thus, when recovering the metasurface phase distribution corresponding to the far-field intensity distribution image through the angular spectrum propagation model, the right-hand circularly polarized phase distribution map and the left-hand circularly polarized phase distribution map of the far-field intensity distribution image, obtained by extracting hierarchical features from the U-Net architecture and reconstructing them, can be used as input to the angular spectrum propagation layer. This allows for Fourier transform, frequency domain multiplication, and inverse Fourier transform through the angular spectrum propagation model to calculate the corresponding far-field intensity distribution and establish the gradient relationship between intensity and phase. This enables the deep learning architecture to directly optimize phase distribution prediction from the intensity error.

[0107] For example, deep learning architectures such as Figure 6 As shown, the angular spectral propagation model is embedded as part of the deep learning architecture in the training loop architecture. Figure 6 As indicated by the pink arrows and orange dashed boxes in the image, this deep learning architecture starts with an intensity input (far-field intensity distribution image), processes the predicted phase information through the U-Net architecture, and then inputs the obtained phase (right-hand circular polarization phase distribution map and left-hand circular polarization phase distribution map of the far-field intensity distribution image) into the physical model (angular spectrum propagation model) to predict the intensity. The mean squared error loss between the predicted intensity and the input intensity is backpropagated through an adaptive learning rate (Adam) optimizer to achieve end-to-end training.

[0108] also, Figure 6 The blue dashed boxes in the diagram indicate the training loop, the pink arrows represent the backpropagation path, and the orange dashed lines highlight the physical information feedback mechanism. The output of the deep learning architecture can be converted into a complex optical field using a formula, which is then propagated through the angular spectrum to generate the predicted far-field intensity, thereby determining the final phase distribution.

[0109] In this embodiment, after the terminal device reconstructs the RCP and LCP phase distributions to obtain the right-hand circularly polarized phase distribution maps and the left-hand circularly polarized phase distribution maps of the far-field intensity distribution image, an angular spectrum propagation model is introduced as a physical information driving layer to connect the phase distribution with the far-field intensity pattern. Since the angular spectrum propagation model can accurately describe the relationship between the metasurface phase distribution and the far-field intensity pattern, this rigorous electromagnetic framework provides a solution for paraxial optics systems that maintains physical accuracy while achieving significantly higher computational efficiency than full-wave simulation. Furthermore, by constructing a deep learning framework, the angular spectrum propagation model is embedded during neural network training to directly recover the phase distribution from the target far-field intensity distribution, thereby eliminating the need for a large amount of pre-generated datasets.

[0110] In some embodiments, step S104 above, designing a polarization beam splitting metasurface based on the metasurface phase distribution, may include, but is not limited to, the following steps:

[0111] The phase distribution of the metasurface is converted into nanopillar structure parameters, and a nanopillar superlattice is designed based on the nanopillar structure parameters to construct a polarization beam splitting metasurface.

[0112] After obtaining the metasurface phase distribution corresponding to the far-field intensity distribution image, the terminal device designs a polarization beam splitting metasurface based on this metasurface phase distribution. This is achieved by converting the final determined metasurface phase distribution into nanopillar structure parameters, and then designing a silicon nanopillar superlattice based on geometric parameter engineering.

[0113] For example, the terminal device converts the phase distribution map (the metasurface phase distribution corresponding to the far-field intensity distribution image) into specific nanopillar structure parameters according to the formula shown below.

[0114]

[0115] Here, φRCP and φLCP represent the phase distributions of right-handed circularly polarized light (RCP) and left-handed circularly polarized light (LCP), respectively, S represents the activation function Sigmod, and Wfinal,RCP and Wfinal,LCP are weighting coefficients related to the polarization state. Based on this, the phase modulation of RCP and LCP can be converted into structural parameters of the nanopillar (such as diameter and height).

[0116] Next, specific embodiments of the metasurface beam splitter provided in this application will be described.

[0117] The metasurface beam splitter provided in this application embodiment is constructed based on the design method of the polarization beam splitting metasurface provided in the above-described embodiment. That is, after the terminal device constructs the polarization beam splitting metasurface by executing a specific embodiment of the design method of the polarization beam splitting metasurface provided in the above-described embodiment, a metasurface beam splitter can be further designed based on the polarization beam splitting metasurface.

[0118] In some embodiments, the structure of the metasurface beam splitter designed according to the present application may include: (1) a substrate layer: composed of an amorphous silicon (α-Si) silica substrate with a thickness of 500 nm; (2) a nanopillar array layer: composed of α-Si nanopillars with a diameter range of 100-400 nm, a height of 700 nm, and a period of 750 nm; (3) a control layer: including an input light source control unit, a polarization state detection unit, and an output beam analysis unit.

[0119] In some embodiments, the polarization beam splitting function of the metasurface beam splitter can be completely decoupled from the Jones matrix through geometric parameter engineering and Pancharatnam-Berry (PB) phase modulation, enabling independent control of the orthogonal polarization state. The mathematical relationship between the PB phase and the RCP / LCP phase is expressed as follows:

[0120]

[0121] Wherein, Φtotal RCP / LCP is the total phase of RCP or LCP, which is determined by both the geometric phase and the PB phase; ΦgeoRCP / LCP(L,W) is the geometric phase, which is related to the length (L) and width (W) of the nanopillar; and θ(x,y) is the rotation angle of the nanopillar, which is a function of the position (x,y).

[0122] Based on this, the terminal device can determine the corresponding rotation angle of the nanopillars to construct the metasurface. The mathematical expression of the rotation angle is shown below:

[0123]

[0124] Wherein, ΦtargetRCP / LCP(x,y) is the target phase distribution, that is, the phase values ​​that the RCP and LCP are expected to reach at position (x,y).

[0125] For example, please refer to Figure 7 , Figure 7 (a) in the figure is a schematic diagram of the unit structure in the metasurface, showing the rectangular silicon nanopillars (α-Si) on the silicon dioxide substrate (SiO2), and the key geometric parameters are marked in the figure. Figure 7 Figure (b) shows the phase response diagrams under right-handed circularly polarized light (RCP) and left-handed circularly polarized light (LCP), demonstrating the complete 2π phase coverage in parameter space as the length and width of the nanopillars vary. Furthermore, Figure 7 (c) in the figure is a schematic diagram of the supercell configuration, where the left side is a supercell composed of 2×2 nanopillars and the right side is a top view of the entire metasurface array containing 32×32 supercells. Figure 7(d) in the diagram is a top view of the 2×2 supercell configuration, showing the spatial distribution of the nanopillars.

[0126] It should be noted that the implementation of the polarization beam splitting function of the metasurface beam splitter, in conjunction with geometric parameter engineering and PB phase modulation, involves the following mechanisms: phase mapping, parameter optimization process, joint encoding of rotation angle and geometric parameters, and complete decoupling of the Jones matrix. Among these:

[0127] 1. Phase Mapping: By introducing a PB contribution through rotation angle, complete Jones matrix decoupling is achieved. This correlation ensures that the rotation component of phase modulation can correctly compensate for the geometric phase component, thereby simultaneously achieving the overall phase distribution required for both polarization states.

[0128] 2. Parameter optimization process: For the optimization process of each supercell position, the target phase is extracted from the neural network output, the required rotation angle is calculated based on the phase difference to determine the adjusted geometric phase requirement, and the optimal nanopillar size that minimizes the phase error of the two polarization states is found.

[0129] 3. Joint encoding of rotation angle and geometric parameters: The rotation angle θ is integrated with the length L and width W of the nanopillar into a parameter vector [L,W,θ], and the position of each unit is optimized to simultaneously meet the phase requirements of RCP and LCP.

[0130] 4. Complete decoupling of the Jones matrix: By precisely controlling the geometric parameters and rotation angle of the nanopillars, complete control of the transmission / reflection coefficients can be achieved, thereby independently regulating the two orthogonal polarization states and completing the polarization beam splitting function.

[0131] In this embodiment, the specific parameters (such as diameter, height, period, etc.) of the nanopillars mapped in the metasurface beam splitter are determined through optimization, thereby achieving the best polarization beam splitting effect. This structural design allows the device to achieve high performance indicators that are difficult to reach with traditional optical components while maintaining micro / nano dimensions.

[0132] Next, a simulation verification example of the finite-difference time-domain (FDTD) method for the metasurface beam splitter provided in the embodiments of this application is presented.

[0133] When verifying the performance of the metasurface beam splitter provided in this application embodiment using FDTD simulation software, parameter scanning was performed on nanopillars of different scales to obtain corresponding RCP and LCP diagrams. Based on the phase distribution obtained from the aforementioned U-net architecture, FDTD was used to map and construct the metasurface, obtaining the physical model information of the metasurface. This model file was imported into the FDTD simulation area, and simulation was performed by setting conditions such as the light source and polarization state. The results show that the metasurface beam splitter provided in this application embodiment has the characteristics of high polarization extinction ratio (PER) (34.11dB) and high transmission efficiency (63.91%).

[0134] It should be noted that in the implementation of PER, the logarithm of the ratio of the intensity of transmitted polarized light (desired polarization state) to that of cross-polarized light (undesired polarization state) was calculated, i.e.: PER = 10log 10 (I_desired / I_undesired). Furthermore, PER values ​​of 36.63dB, 28.11dB, and 37.6dB were measured at the three focal positions, averaging 34.11dB, far exceeding the 20dB threshold required for polarization-sensitive applications. This confirms that the metasurface beam splitter provided in this application achieves an extinction ratio of 34.11dB, significantly superior to related technologies.

[0135] In some embodiments, the metasurface beam splitter provided in this application can be a three-beam beam splitter, corresponding to left-hand circularly polarized light (LCP), right-hand circularly polarized light (RCP), and 45° linearly polarized light, respectively. This three-beam beam splitter successfully recovers the required phase distribution through an improved U-Net architecture and angular spectrum propagation model. Furthermore, the performance of the three-beam beam splitter is verified by FDTD simulation, and the simulation results demonstrate that the three-beam beam splitter possesses a high extinction ratio and high transmission efficiency.

[0136] In some embodiments, the metasurface beam splitter provided in this application can also be a polarization beam splitter with arbitrary beam directionality.

[0137] In some embodiments, the metasurface beam splitter provided in this application can achieve arbitrary beam directionality and polarization control, thus the metasurface beam splitter provided in this application can be applied to a variety of optical applications.

[0138] For example, specific applications of the metasurface beam splitter provided in this application embodiment may include:

[0139] a. Polarization Imaging System: Utilizing the high extinction ratio of metasurface beam splitters to acquire polarization information of biological tissues, thereby enhancing the contrast of medical images.

[0140] b. Optical communication multiplexing: The property of metasurface beam splitters to independently control polarization states enables high-bandwidth information transmission and improves communication capacity.

[0141] c. Quantum information processing: Metasurface beam splitters can provide a high-fidelity manipulation platform for polarization-coded qubits.

[0142] d. Augmented Reality (AR) display: Based on metasurface beam splitters, the direction of beams with different polarization states can be controlled simultaneously, thereby realizing multi-level information presentation.

[0143] e. Optical security encryption: Utilizing the polarization-dependent diffraction pattern of a metasurface beam splitter as a security verification mechanism.

[0144] f. Advanced Spectrometer: Combining the polarization and spatial control capabilities of metasurface beam splitters, it enables the separation and analysis of complex spectral components.

[0145] Next, please refer to Figure 8 This application also provides a design apparatus for a polarizing beam splitter metasurface. This apparatus can implement the aforementioned design method for a polarizing beam splitter metasurface. The design apparatus for a polarizing beam splitter metasurface includes:

[0146] The acquisition module is used to acquire far-field intensity distribution images;

[0147] The image processing module is used to extract and reconstruct features from the far-field intensity distribution image to obtain a right-handed circular polarization phase distribution map and a left-handed circular polarization phase distribution map of the far-field intensity distribution image.

[0148] The intensity phase mapping module is used to input the right-hand circular polarization phase distribution map and the left-hand circular polarization phase distribution map into a preset angular spectrum propagation model, so as to recover the metasurface phase distribution corresponding to the far-field intensity distribution image based on the angular spectrum propagation model;

[0149] A metasurface design module is used to design a polarization beam splitting metasurface based on the phase distribution of the metasurface.

[0150] In some embodiments, the intensity phase mapping module is further configured to input the right-hand circularly polarized phase distribution map and the left-hand circularly polarized phase distribution map as model input data into a preset angular spectrum propagation model; perform far-field intensity prediction calculation based on the angular spectrum propagation model using the model input data to obtain the predicted far-field intensity; and perform inverse optimization on the mean square error between the predicted far-field intensity and the input far-field intensity based on a preset intensity phase gradient relationship to obtain a metasurface phase distribution corresponding to the far-field intensity distribution image; the input far-field intensity is the far-field intensity of the far-field intensity distribution image.

[0151] In some embodiments, the design apparatus for the polarization beam splitting metasurface further includes:

[0152] A neural network training model is used to transform a preset angular spectrum propagation model into a differentiable neural network layer, and to set the angular spectrum propagation model as a partially embedded training loop architecture, so as to construct a deep learning framework based on physical information.

[0153] The angular spectrum propagation model is the physical information driving layer in the deep learning framework; the deep learning framework is used to train a neural network with the far-field intensity distribution image as input data, and outputs the metasurface phase distribution corresponding to the far-field intensity distribution image recovered based on the angular spectrum propagation model.

[0154] In some embodiments, the image processing module is further configured to input the far-field intensity distribution image into a preset convolutional neural network architecture; the convolutional neural network architecture includes a shared encoder and a dual decoder branch; and, based on the shared encoder and the dual decoder branch, to perform feature extraction and reconstruction on the far-field intensity distribution image to obtain a right-handed circular polarization phase distribution map and a left-handed circular polarization phase distribution map of the far-field intensity distribution image.

[0155] In some embodiments, the image processing module is further configured to extract hierarchical features from the far-field intensity distribution image based on the shared encoder; and to reconstruct a right-handed circularly polarized phase distribution based on the hierarchical features as input from the first branch of the dual decoder branch to obtain a right-handed circularly polarized phase distribution map of the far-field intensity distribution image; and to reconstruct a left-handed circularly polarized phase distribution based on the hierarchical features as input from the second branch of the dual decoder branch to obtain a left-handed circularly polarized phase distribution map of the far-field intensity distribution image.

[0156] In some embodiments, the metasurface design module is further configured to convert the metasurface phase distribution into nanopillar structure parameters, and to perform nanopillar superlattice design based on the nanopillar structure parameters to construct a polarization beam splitting metasurface.

[0157] It should be noted that the specific implementation of the design device for polarized beam splitting metasurface provided in this application is basically the same as the specific implementation of the design method for polarized beam splitting metasurface described above, and will not be repeated here.

[0158] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described design method for a polarized beam splitting metasurface. This electronic device can be any smart terminal device, including tablet computers and personal computers.

[0159] Please see Figure 9 , Figure 9 This illustration shows the hardware structure of an electronic device in some embodiments. The electronic device may include:

[0160] The processor 901 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application.

[0161] The memory 902 can be implemented as a read-only memory (ROM), static storage device, dynamic storage device, or random access memory (RAM). The memory 902 can store the operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 902 and is called and executed by the processor 901 to execute the design method of the polarization beam splitting metasurface of the embodiments of this application.

[0162] The input / output interface 903 is used to implement information input and output;

[0163] The communication interface 904 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0164] Bus 905 transmits information between various components of the device (e.g., processor 901, memory 902, input / output interface 903, and communication interface 904);

[0165] The processor 901, memory 902, input / output interface 903, and communication interface 904 are connected to each other within the device via bus 905.

[0166] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described design method for a polarized beam splitting metasurface.

[0167] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0168] This application also provides a computer program product that stores a computer program that, when executed by a processor, implements the above-described design method for a polarized beam splitting metasurface.

[0169] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0170] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0171] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0172] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0173] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0174] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0175] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0176] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0177] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0178] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0179] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A method for designing a polarizing beam splitting metasurface, characterized in that, The method includes: Obtain the far-field intensity distribution image; Feature extraction and reconstruction are performed on the far-field intensity distribution image to obtain a right-handed circular polarization phase distribution map and a left-handed circular polarization phase distribution map of the far-field intensity distribution image; The right-hand circular polarization phase distribution map and the left-hand circular polarization phase distribution map are input into a preset angular spectrum propagation model to recover the metasurface phase distribution corresponding to the far-field intensity distribution image based on the angular spectrum propagation model. The design of a polarization beam splitting metasurface is based on the phase distribution of the metasurface.

2. The method according to claim 1, characterized in that, The step of inputting the right-hand circularly polarized phase distribution map and the left-hand circularly polarized phase distribution map into a preset angular spectrum propagation model, so as to recover the metasurface phase distribution corresponding to the far-field intensity distribution image based on the angular spectrum propagation model, includes: The right-hand circular polarization phase distribution map and the left-hand circular polarization phase distribution map are used as model input data to input into a preset angular spectrum propagation model; Based on the angular spectrum propagation model, far-field intensity prediction calculation is performed using the model input data to obtain the predicted far-field intensity. Based on a preset intensity-phase gradient relationship, the mean square error between the predicted far-field intensity and the input far-field intensity is optimized in reverse to obtain the metasurface phase distribution corresponding to the far-field intensity distribution image; the input far-field intensity is the far-field intensity of the far-field intensity distribution image.

3. The method according to claim 1, characterized in that, The method further includes: The preset angular spectrum propagation model is transformed into a differentiable neural network layer, and the angular spectrum propagation model is set as a partially embedded training loop architecture to construct a deep learning framework driven by physical information. The angular spectrum propagation model is the physical information driving layer in the deep learning framework; the deep learning framework is used to train a neural network with the far-field intensity distribution image as input data, and outputs the metasurface phase distribution corresponding to the far-field intensity distribution image recovered based on the angular spectrum propagation model.

4. The method according to claim 1, characterized in that, The step of extracting and reconstructing features from the far-field intensity distribution image to obtain a right-handed circular polarization phase distribution map and a left-handed circular polarization phase distribution map of the far-field intensity distribution image includes: The far-field intensity distribution image is input into a preset convolutional neural network architecture; the convolutional neural network architecture includes a shared encoder and a dual decoder branch; Based on the shared encoder and the dual decoder branch, feature extraction and reconstruction are performed on the far-field intensity distribution image to obtain the right-hand circular polarization phase distribution map and the left-hand circular polarization phase distribution map of the far-field intensity distribution image.

5. The method according to claim 4, characterized in that, The step of extracting and reconstructing features from the far-field intensity distribution image based on the shared encoder and the dual decoder branch to obtain a right-handed circular polarization phase distribution map and a left-handed circular polarization phase distribution map of the far-field intensity distribution image includes: Based on the shared encoder, hierarchical features are obtained by feature extraction from the far-field intensity distribution image; Based on the first branch of the dual decoder branch, the right-hand circular polarization phase distribution is reconstructed using the hierarchical features as input to obtain the right-hand circular polarization phase distribution map of the far-field intensity distribution image; and based on the second branch of the dual decoder branch, the left-hand circular polarization phase distribution is reconstructed using the hierarchical features as input to obtain the left-hand circular polarization phase distribution map of the far-field intensity distribution image.

6. The method according to any one of claims 1 to 5, characterized in that, The design of the polarization beam splitting metasurface based on the phase distribution of the metasurface includes: The phase distribution of the metasurface is converted into nanopillar structure parameters, and a nanopillar superlattice is designed based on the nanopillar structure parameters to construct a polarization beam splitting metasurface.

7. A design device for a polarizing beam splitting metasurface, characterized in that, The device includes: The acquisition module is used to acquire far-field intensity distribution images; The image processing module is used to extract and reconstruct features from the far-field intensity distribution image to obtain a right-handed circular polarization phase distribution map and a left-handed circular polarization phase distribution map of the far-field intensity distribution image. The intensity phase mapping module is used to input the right-hand circular polarization phase distribution map and the left-hand circular polarization phase distribution map into a preset angular spectrum propagation model, so as to recover the metasurface phase distribution corresponding to the far-field intensity distribution image based on the angular spectrum propagation model; A metasurface design module is used to design a polarization beam splitting metasurface based on the phase distribution of the metasurface.

8. A metasurface beam splitter, characterized in that, The metasurface of the metasurface beam splitter is constructed based on the design method of polarized beam splitting metasurface according to any one of claims 1 to 6.

9. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the design method of the polarization beam splitting metasurface according to any one of claims 1 to 6.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the design method of the polarization beam splitting metasurface as described in any one of claims 1 to 6.

Images