Image sensor based on metasurface color routing and design method thereof

Abstract

The application discloses an image sensor based on a super-structured surface color routing. The image sensor comprises an upper super-structured surface layer and a lower pixel layer, wherein the pixel layer is composed of a plurality of pixel units arranged in cycles, each pixel unit comprises four pixels, namely R, G, B and N pixels; the super-structured surface is composed of super-structured units, and the cycle of the super-structured units is consistent with the cycle of the pixel units; the super-structured surface is used for separating incident light in the visible to near-infrared range into four colors and focusing the four colors on the four pixels respectively. The application can route incident light in the 400-1100 nanometer wave band to four image sensor pixels arranged in a 2*2 array, correspond to blue, green, red and near-infrared channels respectively, and directly realize color imaging.

Description

Technical Field

[0001] This invention relates to the fields of micro-nano optics and imaging technology, specifically to a visible-to-near-infrared image sensor based on metasurface color routing, used to efficiently split and route light in the visible-to-near-infrared band to different regions. The invention also relates to a method for designing metasurfaces. Background Technology

[0002] Color imaging technology plays a crucial role in modern society. Currently, CMOS image sensors equipped with Bayer color filter arrays are the mainstream in the market. However, Bayer color filters filter colors by absorbing light in specific wavelengths, resulting in the waste of most incident light. Theoretically, the utilization rate of blue and red light is no more than 25%, and that of green light is no more than 50%, while completely filtering out near-infrared light. This causes serious waste of light energy, especially given the continuous shrinking of pixel sizes and the urgent need for low-light imaging, becoming a bottleneck restricting the improvement of imaging performance. The spectral response range of silicon-based CMOS image sensor chips originally covered 400nm to 1100nm, but to avoid color crosstalk caused by near-infrared light in the 700nm-1100nm band, an infrared cutoff filter must be added in front of the color filter. This further sacrifices light energy utilization and completely abandons valuable near-infrared spectral information.

[0003] To overcome these limitations, the industry has made several attempts. First, it has developed novel color filters based on nanophotonic structures (such as plasmon resonance and photonic crystals), but these are essentially absorption filters and fail to fundamentally solve the energy loss problem. Second, it has adopted novel pixel arrangements such as RYYB and RGBW, using algorithms for color restoration, but the spatial multiplexing filtering mode remains unchanged, and the theoretical ceiling for light energy utilization efficiency still exists. Third, it has employed stacked sensors like those used in the Foveon X3, but their complex multi-layer photoelectric conversion structure and large pixel size limit their application in video imaging requiring high resolution and high frame rates. The natural environment is rich in near-infrared light, whose intensity is often higher than visible light, and silicon-based sensors can detect this band. Effectively utilizing near-infrared light energy is a feasible and efficient path to improve low-light imaging performance.

[0004] In recent years, the concept of color routing has been proposed. It adopts a physical mechanism of "splitting light" rather than "filtering light" to route light of different wavelengths to designated pixels, theoretically achieving a light energy utilization efficiency of nearly 100%. Although color routing based on three-dimensional bulk metamaterials can achieve near-perfect routing function, its complex three-dimensional structure at the nanoscale far exceeds the limits of current micro-nano fabrication technology and is even more difficult to be compatible with standard CMOS processes.

[0005] Metasurfaces, as planar optical elements composed of subwavelength antennas, offer a new technological approach to color routing due to their powerful electromagnetic wave manipulation capabilities and compact form. While forward-designed multifocal metalenses have achieved Bayer-type routing, meeting 0-2π phase coverage typically requires nanostructures with high aspect ratios or complex cross-sectional shapes, posing significant manufacturing challenges. Reverse-designed metasurfaces have reduced fabrication difficulty to some extent and have achieved Bayer-type beam splitting in the visible light range. However, these approaches still fail to effectively solve the problem of efficient, low-crosstalk integrated routing of visible and near-infrared light, limiting their application in broadband imaging.

[0006] Therefore, there is an urgent need in the field for a new type of color router device that can simultaneously cover the visible and near-infrared bands, has high light energy utilization efficiency, low crosstalk, is easy to manufacture, and can be integrated with CMOS technology. Summary of the Invention

[0007] The purpose of this invention is to overcome the technical defects of existing Bayer color filters, such as low light energy utilization, inability to effectively utilize near-infrared light, and difficulty of existing metasurface routing devices to achieve high efficiency, low crosstalk, and polarization-insensitive beam splitting in a wide band (visible-near-infrared). This invention provides a visible-to-near-infrared image sensor based on metasurface color routing and its design method.

[0008] To achieve the above objectives, the present invention adopts the following technical solution: an image sensor based on metasurface color routing, comprising an upper metasurface layer and a lower pixel layer, wherein the pixel layer is composed of multiple pixel units arranged periodically and closely, each pixel unit containing 4 pixels, namely R, G, B, and N pixels; the metasurface layer is composed of metasurface units, the period of which is consistent with the period of the pixel units; the metasurface layer is used to separate the incident light in the visible to near-infrared range into four colors and focus them onto the 4 pixels respectively.

[0009] Furthermore, the metasurface layer is composed of nanopillars and a substrate.

[0010] Furthermore, each metaunit is divided into n*n cells, and the nanopillars are arranged within the cells.

[0011] Furthermore, the nanopillars are not completely arranged within the cell.

[0012] Furthermore, the arrangement of nanopillars corresponding to each pixel in the pixel unit is not exactly the same.

[0013] Furthermore, the side length of the nanopillar is equal to the side length of the cell.

[0014] Furthermore, the side length of the cell is 2P / n, where P is the side length of a single pixel; the height of the nanopillar is 600-1200nm.

[0015] Furthermore, the wavelength of R pixels is 600-700nm, the wavelength of G pixels is 500-600nm, the wavelength of B pixels is 400-500nm, and the wavelength of N pixels is 700-1100nm.

[0016] This invention also provides a design method for the above-mentioned image sensor based on metasurface color routing, comprising the following steps: Step 1, the metasurface layer is composed of nanopillars and a substrate, each metasurface unit is divided into n*n cells, each cell is assigned to place a nanopillar or air, the placement position is represented by 1 and 0 respectively; the arrangement of the nanopillars of the metasurface layer is mapped to a matrix mat containing only 0 and 1; the geometric parameters of the metasurface layer include matrix mat, nanopillar height and substrate thickness; Step 2, the metasurface layer is optimized by a genetic algorithm, the optimization process includes: first, the geometric parameters of the metasurface layer are transformed into a single-row matrix as a population; then, the light field distribution of the metasurface layer with different geometric parameters is calculated based on the finite difference method in the time domain; then, the fitness function is calculated and the population is selected, crossover and mutated; Step 3, Step 2 is repeated until the fitness function converges, and the optimized metasurface layer parameters are output.

[0017] The visible to near-infrared image sensor based on metasurface color routing designed in this invention has the following main advantages:

[0018] (1) The present invention uses the principle of beam splitting to replace the filter array and introduces a near-infrared imaging channel. The overall light energy utilization efficiency is much higher than that of traditional Bayer color filters, which is beneficial for use in low light environments.

[0019] (2) By optimizing the design of the metasurface, the present invention makes the structure insensitive to the polarization of incident light, thus possessing polarization insensitive characteristics and improving its practicality;

[0020] (3) The metasurface structure of the present invention is compact and can be fabricated using standard CMOS processes (such as PECVD, electron beam lithography, ICP-RIE). It is easy to achieve on-chip integration on sensor chips and is suitable for scenarios that require simultaneous acquisition of color and near-infrared information, such as low-light imaging, autonomous driving, and face recognition. Attached Figure Description

[0021] Figure 1 This is a schematic diagram illustrating the principle of visible light to near-infrared color routing in this invention.

[0022] Figure 2 These are schematic diagrams (side view and top view) of an embodiment of the present invention.

[0023] Figure 3 This is a flowchart of the design method according to an embodiment of the present invention.

[0024] Figure 4 The image shows the pixel-plane light field diagrams of incident light with different polarizations and wavelengths for the metasurface color routing of this invention.

[0025] Figure 5 This is a structural diagram of the metasurface color routing of the present invention.

[0026] Figure 6 for Figure 5 A table of specific parameters for metasurface structures. Detailed Implementation

[0027] This embodiment provides a visible to near-infrared image sensor based on metasurface color routing, comprising an upper metasurface layer and a lower pixel layer. The pixel layer consists of M basic pixel units arranged periodically and closely, where M is a positive integer. Each pixel unit contains four pixels: R, G, B, and N pixels, with colors representing blue (400-500nm), green (500-600nm), red (600-700nm), and near-infrared (700-1100nm), respectively. The metasurface layer is composed of metasurface units whose period matches that of the pixel units. By modulating light of different wavelengths through the metasurface layer, the broadband incident light from the visible to near-infrared spectrum is separated into four colors and focused onto the four pixels respectively.

[0028] like Figure 1 As shown, specifically, each pixel unit contains four pixels of the same size, with each pixel having a side length of P. Therefore, the side length of the pixel unit is 2P, where P is typically 0.56-2.8 μm. A silicon dioxide substrate is deposited on the pixel layer surface, and then silicon nitride nanopillar structures are fabricated on the substrate through etching. The substrate and nanopillars together constitute the metasurface layer. The nanopillars have a height of h (600-1500 nm), a substrate thickness of f (2-10 μm), and a square cross-section with a side length of a. The arrangement of the nanopillars in this embodiment is as follows... Figure 2 As shown, different coupling or resonance effects are generated between the individual nanopillars, resulting in specific responses to different wavelengths. By using a metasurface layer, the wavefront of incident light can be modulated according to different wavelengths, and the light can be routed to a designated sensor pixel after propagation through the substrate, thereby significantly improving light energy utilization and showing great application potential in low-light scenarios.

[0029] To achieve efficient routing across the four bands of visible and near-infrared wavelengths simultaneously, this embodiment employs a design method combining genetic algorithms and numerical simulation. Specifically, to optimize the arrangement of the nanopillars in the metasurface layer, "binary encoding" is first performed, dividing each metasurface unit into n*n cells, with a cell side length of 2P / n. Each cell is assigned either a nanopillar or air (i.e., the missing portion of the nanopillar), with placement positions represented by "1" and "0" respectively. The arrangement of the metasurface layer nanopillars is mapped to a binary matrix mat containing only "0" and "1", with a total of n*n elements. In this embodiment, the nanopillars are designed to completely fill the cells (though they can also be smaller than the cells), hence the nanopillar side length a = 2P / n. Therefore, the geometric parameters of the metasurface layer are determined by the matrix mat, the nanopillar height h, and the substrate thickness f, resulting in a total of n*n+2 parameters, where h = 600-1200 nm, f = 2-10 μm, and mat is a randomly varying matrix of 0s and 1s. Then, the metasurface layer is optimized using a genetic algorithm. The optimization process includes: first, representing the geometric parameters of the metasurface layer as a 1*(n*n+2) single-row matrix as the population; then, calculating the light field distribution for metasurface layers with different geometric parameters based on the finite-difference time-domain method; finally, calculating the fitness function and performing selection, crossover, and mutation on the population. The population is continuously iterated and optimized until the fitness function converges, and the optimized metasurface layer parameters are output. The fitness function is:

[0030] in These represent different incident light polarization angles. To represent different channels, Then in the formula This represents the channels collected on each pixel. The ratio of the energy of light to the energy of incident light. , Indicates channel The corresponding pixel area Indicates channel The ratio of light energy to incident light energy at each sampling point. These are adjustable weighting coefficients used to prevent the response of a particular channel from becoming too prominent and causing crosstalk. The incident wavelength is [wavelength]. A flowchart illustrating the design method can be found [here]. Figure 3 It should be noted that the method of this invention can be extended to other application bands. During the optimization process, since different polarization incident conditions are calculated simultaneously, the metasurface layer will have similar wavefront modulation for incident light of different polarizations, thereby producing an approximately consistent routing effect and maintaining the polarization insensitivity of the sensor.

[0031] This embodiment provides a specific example of structural parameters based on the design method described above. For example... Figure 2 As shown, a single pixel is a square with a side length P of 1000nm and a pixel area of ​​1000nm*1000nm. The period of both the pixel unit and the meta-unit is 2000nm. Figure 5 and Figure 6 The arrangement of the nanopillar array in the metasurface layer is shown. A metaunit is divided into 16*16 cells, each assigned as either a nanopillar or air, resulting in a nanopillar cross-section with a side length of 125 nm. The optimized nanopillar height is 800 nm, and the substrate thickness is 2400 nm. Figure 6 It is the optimized binary matrix mat. Figure 5 The left image is a top view of the metasurface layer nanopillars arranged according to mat, while the enlarged view on the right shows a complete metaunit structure. The nanopillars corresponding to the four pixels in this unit are arranged in completely different ways; of course, some similarities may occur in other optimized embodiments. According to the optimization results, incident light with different wavelengths and polarizations passing through the metasurface layer is focused on the pixel region, achieving a color routing effect in the 400-1100nm range, as shown in the image. Figure 4 As shown. Blue, green, red, and gray correspond to incident light with wavelengths of 450nm, 540nm, 630nm, and 900nm, respectively. Figure 4 (a)-(d) show the results at polarization of 0°, 30°, 60° and 90°, respectively.

[0032] The visible-to-near-infrared image sensor based on metasurface color routing proposed in this invention can route incident light in the 400-1100 nm wavelength band to four image sensor pixels arranged in a 2×2 pattern, corresponding to the blue, green, red, and near-infrared channels respectively, directly achieving color imaging and bringing greater flexibility and application potential to modern optical systems. Compared with traditional solutions, the integrated advantages of this invention have disruptive value in practical applications: the device size is reduced by more than 70% compared to traditional discrete systems, allowing integration into handheld portable devices; the routing effect replaces filter devices, improving energy utilization by more than 2 times; the planar structure is compatible with CMOS technology, supporting large-scale mass production. With the advancement of micro-nano fabrication technology, this image sensor is expected to achieve technological innovation in various fields such as surveillance cameras, facial recognition, and remote sensing monitoring.

[0033] The above embodiments are merely illustrative of the principles and functions of the present invention and are not intended to limit the present invention. Any non-substantial substitutions or modifications that do not exceed the essential spirit of the present invention fall within the protection scope of the present invention.

Claims

1. An image sensor based on metasurface color routing, characterized in that, The image sensor includes an upper metasurface layer and a lower pixel layer. The pixel layer is composed of multiple pixel units arranged in a periodic and close arrangement. Each pixel unit contains four pixels, namely R, G, B, and N pixels. The metasurface layer is composed of metaunits, and its period is consistent with the period of the pixel units. The metasurface layer is used to separate the incident light in the visible and near-infrared range into four colors and focus them onto the four pixels respectively.

2. The image sensor based on metasurface color routing according to claim 1, characterized in that, The metasurface layer consists of nanopillars and a substrate.

3. The image sensor based on metasurface color routing according to claim 2, characterized in that, Each metaunit is divided into n*n cells, and the nanopillars are arranged within the cells.

4. The image sensor based on metasurface color routing according to claim 3, characterized in that, The nanopillars are not completely arranged within the cell.

5. The image sensor based on metasurface color routing according to claim 4, characterized in that, The arrangement of nanopillars corresponding to each pixel in the pixel unit is not exactly the same.

6. The image sensor based on metasurface color routing according to claim 3, characterized in that, The side length of the nanopillar is equal to the side length of the cell.

7. The image sensor based on metasurface color routing according to claim 3, characterized in that, The side length of the cell is 2P / n, where P is the side length of a single pixel; the height of the nanopillar is 600-1200nm.

8. The image sensor based on metasurface color routing according to claim 1, characterized in that, The wavelength of R pixels is 600-700nm, the wavelength of G pixels is 500-600nm, the wavelength of B pixels is 400-500nm, and the wavelength of N pixels is 700-1100nm.

9. The design method of an image sensor based on metasurface color routing as described in claim 1, characterized in that, The method includes the following steps: Step 1, the metasurface layer is composed of nanopillars and a substrate. Each metasurface unit is divided into n*n cells, and each cell is assigned to place either a nanopillar or air, with the placement position represented by 1 and 0 respectively; the arrangement of the nanopillars in the metasurface layer is mapped to a matrix mat containing only 0 and 1; the geometric parameters of the metasurface layer include the matrix mat, the height of the nanopillars, and the thickness of the substrate; Step 2, the metasurface layer is optimized using a genetic algorithm. The optimization process includes: first, transforming the geometric parameters of the metasurface layer into a single-row matrix as a population; then, calculating the light field distribution of metasurface layers with different geometric parameters based on the finite-difference time-domain method; then, calculating the fitness function and performing selection, crossover, and mutation on the population; Step 3, repeating Step 2 until the fitness function converges, and outputting the optimized metasurface layer parameters.

10. The design method according to claim 9, characterized in that, In step 2, the fitness function is: ,in Representing different incident light polarization angles, Indicates different channels, , Represents the channels collected on each pixel. The ratio of the energy of light to the energy of incident light. , Indicates channel The corresponding pixel area Indicates channel The ratio of light energy to incident light energy at each sampling point. These are adjustable weighting coefficients. λ is the incident wavelength.

Images