Optical processing chip and optical computing system

Abstract

The application provides an optical processing chip and an optical computing system, and relates to the field of optoelectronic technology. The optical coupling structure comprises an optical coupling structure, a micro-nano superstructure and an optical signal receiving structure. The first surface of the optical coupling structure has a plurality of light output ports for outputting a plurality of first optical signals. The micro-nano superstructure is arranged on the first surface and comprises a plurality of super units, and the orthographic projections of the super units on the first surface correspond to the plurality of light output ports one by one. Each super unit comprises a plurality of nano pillars arranged at intervals, which are used for processing the plurality of first optical signals into second optical signals, and the first optical signals and the second optical signals are different. The optical signal receiving structure is connected with the micro-nano superstructure, and the optical signal receiving structure is configured to receive the second optical signals and obtain the information carried by the second optical signals. Parallel computing is realized by using the plurality of super units, so as to improve the computing efficiency of the optical processing chip and enhance the integration of the optical processing chip.

Description

Technical Field

[0001] This application relates to the field of optoelectronic technology, and in particular to an optical processing chip and an optical computing system. Background Technology

[0002] In the field of optoelectronics, existing optical processing chips often rely on a combination of traditional optical and electronic components, such as lenses and prisms, when performing calculations. When using these optical components to process complex optical signals, problems such as energy loss and signal distortion can easily occur, affecting the accuracy and speed of the optical processing chip's calculations. Furthermore, the overall size of the optical and electronic components used in optical processing chips is relatively large, occupying a large amount of chip space, resulting in low integration of the optical processing chip. Consequently, optical processing chips cannot meet the requirements of three-dimensional stacked optical computing architecture for high-speed computing, miniaturization, and high integration. Summary of the Invention

[0003] This application proposes an optical processing chip and an optical computing system, which aims to utilize multiple meta-units to achieve parallel computing, thereby improving the computing efficiency of the optical processing chip and enhancing its integration.

[0004] In a first aspect, a light processing chip is provided. The light coupling structure includes: a light coupling structure, a micro / nano metastructure, and a light signal receiving structure. The light coupling structure has multiple light output ports on a first surface, which are used to output multiple first light signals. The micro / nano metastructure includes multiple meta-units. The micro / nano metastructure is disposed on the first surface, and the orthographic projections of the multiple meta-units on the first surface correspond one-to-one with the multiple light output ports. Each meta-unit includes multiple spaced-apart nanopillars. The meta-units are used to process the multiple first light signals into a second light signal, which is different from the first light signal. The light signal receiving structure is connected to the micro / nano metastructure and is configured to receive the second light signal and obtain the information carried by the second light signal.

[0005] In this embodiment, multiple meta-units in a micro-nano metastructure are used to perform parallel modulation on multiple first optical signals output from multiple optical output ports, i.e., the first optical signals of the beam split, to achieve parallel computation. This can improve the computational efficiency of the optical processing chip. Furthermore, the multiple meta-units are composed of multiple nanopillars, and their overall size is small, which makes the area occupied by the meta-units on the optical processing chip smaller. This can improve the integration of the optical processing chip, thereby enabling the optical processing chip to meet the requirements of high-speed computing, miniaturization, and high integration of a three-dimensional stacked optical computing architecture.

[0006] In some embodiments, at least some of the nanopillars in the metaunit have different dimensions in a first direction; and / or, at least some of the nanopillars in the metaunit have different cross-sectional areas. The cross-sectional areas are perpendicular to the first direction.

[0007] In some embodiments, each metaunit includes a first component and a second component, which are arranged along a first direction perpendicular to the first surface. Each of the first and second components includes a substrate and a plurality of nanopillars, with at least a portion of the nanopillars of the first component engaging with at least a portion of the nanopillars of the second component. The substrates of the first and second components are disposed opposite to each other along the first direction.

[0008] In some embodiments, each metaunit includes at least one component, and each component includes a first component and a second component. When the metaunit includes multiple components, the multiple components are arranged sequentially along a first direction.

[0009] In some embodiments, the number of components in different superunits varies.

[0010] In some embodiments, the optical coupling structure includes a substrate, a waveguide layer, and a grating layer stacked sequentially. The waveguide layer includes multiple waveguides, the ends of which form multiple optical output ports. A first surface is perpendicular to the substrate. The grating layer includes multiple spaced-apart nanopillars, the extension direction of which is perpendicular to the substrate.

[0011] In some embodiments, the material of the multiple nanopillars of the metaunit is the same as that of the waveguide layer.

[0012] In some embodiments, the optical processing chip further includes a photodetector structure. The micro / nano metastructure and the optical signal receiving structure are connected via the photodetector structure. The photodetector is configured to receive a second optical signal and convert it into a corresponding electrical signal. The optical signal receiving structure is configured to receive the electrical signal and obtain the information carried by the electrical signal.

[0013] Secondly, an optical computing system is provided, comprising an optical processing chip and a memory as described in the above embodiments. The memory stores multiple neural network weights. The memory is connected to a micro / nano metastructure in the optical processing chip, which processes multiple first optical signals into second optical signals based on the multiple neural network weights.

[0014] The optical computing system described above has the same structure and beneficial technical effects as the optical processing chip provided in some of the above embodiments, and will not be described again here. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in this application, the accompanying drawings used in some embodiments of this application will be briefly described below. Obviously, the drawings described below are only drawings of some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings. In addition, the drawings described below can be regarded as schematic diagrams and are not actual dimensions of the products or actual processes of the methods involved in the embodiments of this application.

[0016] Figure 1 This is a schematic diagram of a light processing chip structure provided in an embodiment of this application; Figure 2 This is a front view of an optical processing chip provided in an embodiment of this application; Figure 3 A top view of a metacellular unit provided in an embodiment of this application; Figure 4 This is a schematic diagram of another optical processing chip structure provided in an embodiment of this application; Figure 5 A top view of another metastructure provided in an embodiment of this application; Figure 6 A schematic diagram of the overall structure of another optical processing chip provided in this application; Figure 7 A schematic diagram of an optical coupling structure provided in this application; Figure 8 This is a schematic diagram of the structure of an optical computing system according to an embodiment of this application. Detailed Implementation

[0017] The technical solutions in some embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments provided in this application are within the scope of protection of this application.

[0018] Unless the context otherwise requires, throughout the specification and claims, the term "comprising" is interpreted as open and encompassing, that is, "including, but not limited to".

[0019] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of embodiments of this application, unless otherwise stated, "a plurality of" means two or more.

[0020] In describing some embodiments, the term "connection" and its derivative expressions may be used. The term "connection" should be interpreted broadly; for example, "connection" can be a fixed connection, a detachable connection, or an integral part; it can be a direct connection or an indirect connection through an intermediate medium. For example, in describing some embodiments, the term "connection" may be used to indicate that two or more components have direct physical or electrical contact with each other.

[0021] In addition, the use of “based on” implies openness and inclusivity, because processes, steps, calculations or other actions “based on” one or more of the stated conditions or values ​​may in practice be based on additional conditions or values ​​beyond those stated.

[0022] It should be understood that when a layer or element is referred to as being on another layer or substrate, it can mean that the layer or element is directly on the other layer or substrate, or that there is an intermediate layer between the layer or element and the other layer or substrate.

[0023] This document describes exemplary embodiments with reference to cross-sectional views, which are intended as idealized exemplary drawings. In the drawings, the thickness of the layers and the area of ​​the regions are enlarged for clarity. Therefore, variations in shape relative to the drawings are contemplated due to, for example, manufacturing techniques and / or tolerances. Thus, exemplary embodiments should not be construed as limited to the shapes of the regions shown herein, but rather include shape deviations caused, for example, by manufacturing processes. For example, etched areas shown as rectangular would typically have curved features. Therefore, the regions shown in the drawings are schematic in nature, and their shapes are not intended to show the actual shapes of the areas of the device, nor are they intended to limit the scope of the exemplary embodiments.

[0024] In the field of optoelectronics, existing optical processing chips often rely on a combination of traditional optical and electronic components, such as lenses and prisms, when performing calculations. When using these optical components to process complex optical signals, problems such as energy loss and signal distortion can easily occur, affecting the accuracy and speed of the optical processing chip's calculations. Furthermore, the overall size of the optical and electronic components used in optical processing chips is relatively large. For example, the overall size of components used for optical computing (such as Mach-Zehnder modulators) is on the order of hundreds of micrometers or even millimeters, occupying a large amount of chip space. This results in low integration of the optical processing chip, which means that the optical processing chip cannot meet the requirements of high-speed computing, miniaturization, and high integration of a three-dimensional stacked optical computing architecture.

[0025] To address the aforementioned problems, embodiments of this application provide an optical processing chip. Figure 1 This is a schematic diagram of the structure of an optical processing chip provided in an embodiment of this application. Figure 2 This is a front view of an optical processing chip provided in an embodiment of this application. Figure 3This is a top view of a metacellular unit provided in an embodiment of this application. Figure 4 This is a schematic diagram of another optical processing chip provided in an embodiment of this application.

[0026] See Figure 1 and Figure 2 The aforementioned optical processing chip 100 includes an optical coupling structure 1, a micro / nano metastructure 2, and an optical signal receiving structure 3.

[0027] Among them, optical coupling structure 1 refers to a device that transmits optical signals from one optical path to another, and is used to realize the directional transmission and distribution of optical signals.

[0028] For example, the optical coupling structure 1 described above can be a vertical grating coupler, which can efficiently couple optical signals in free space vertically into the waveguide. In this process, the optical signal that was originally incident vertically or had a small angular deviation will have its propagation direction changed after passing through the coupler, and will be converted to propagation in the horizontal direction. That is, after the optical signal passes through the coupler, its propagation direction changes, causing the optical signal to propagate in the horizontal direction or generate a horizontal component.

[0029] For example, see Figure 1 After the optical signal enters the optical coupling structure 1 along the direction -Z, it is deflected after being coupled by evanescent waves, causing the optical signal to continue to propagate along the direction X, or there is a horizontal optical signal component that continues to propagate along the direction X.

[0030] See Figure 1 The first surface of the aforementioned optical coupling structure 1 has multiple optical output ports 11.

[0031] The first surface is one side surface of the optical coupling structure 1 in the X direction, for example... Figure 1 In the middle, along the direction X, the right surface of the optical coupling structure 1.

[0032] The aforementioned multiple optical output ports 11 are used to output multiple first optical signals so that these first optical signals can be processed in parallel subsequently.

[0033] For example, the aforementioned plurality of optical output ports 11 can be output ports of a waveguide beam splitter structure, used to output the transmitted optical signal of each beam, for example, Figure 1 In the middle, the Y-shaped waveguide beam splitter structure has two optical output ports 11.

[0034] The first optical signal here is the optical signal coupled into the waveguide through the optical coupling structure 1, and the output optical signal after being split by the beam splitting waveguide structure. The sum of the intensities of the multiple first optical signals is slightly less than the intensity of the optical signal before beam splitting.

[0035] For example, the optical processing chip 100 may include multiple optical coupling structures 1, which can be arranged in an array to enable optical signals in space to be coupled into the waveguide efficiently and orderly, thereby achieving accurate reception and directional transmission of optical signals. This provides stable and high-quality optical input for subsequent complex optical signal processing operations (such as optical beam splitting, optical modulation, optical filtering, etc.) within the chip, ensuring the stable performance and efficient operation of the overall optical processing chip.

[0036] See Figure 1 The aforementioned micro / nano metastructure 2 is disposed on the first surface of the optical coupling structure 1 to ensure that the first optical signal is further processed (e.g., optical calculation) along the propagation direction of the first optical signal using the micro / nano metastructure 2, thereby reducing the number of bends of the first optical signal during internal transmission and thus reducing optical transmission loss while reducing optical path connections.

[0037] The aforementioned micro / nano metastructure 2 includes multiple metaunits 21.

[0038] The superstructure unit 21 is used to process the first optical signal to obtain the second optical signal. That is, the first optical signal can be modulated by the superstructure unit 21 to perform complex optical calculations on the first optical signal to obtain the second optical signal. For example, the first optical signal can be superimposed, subtracted, subjected to matrix operations, and subjected to Fourier transforms (first order and second order).

[0039] The second optical signal here is different from the first optical signal, that is, the phase, intensity, polarization, wavelength and other parameters of the second optical signal are different from those of the first optical signal. In addition, the second optical signal can also carry information after complex calculations on the first optical signal, such as Fourier spectrum information, linear combination information and feature mapping information.

[0040] The orthographic projections of the aforementioned multiple meta-units 21 on the first surface correspond one-to-one to cover multiple optical output ports 11, that is, one optical output port 11 is connected to one corresponding meta-unit 21, so as to utilize multiple meta-units 21 to process multiple first optical signals in parallel, thereby realizing parallel operation on the coupled optical signals and improving the computational efficiency of the optical processing chip 100.

[0041] For example, multiple meta-units 21 in the micro-nano metastructure 2 can be arranged sequentially or in an array along the first direction. In this case, one optical output port 11 is connected to one meta-unit 21. In this way, by changing the arrangement of the meta-units 21, the micro-nano metastructure 2 can meet different optical signal tasks, thereby improving the applicability of the optical processing chip 100.

[0042] For example, multiple supercells 21 can be used to process multiple first optical signals respectively, and each supercell 21 can process each first optical signal in a different way, so as to flexibly select the appropriate operation mode according to the characteristics of different first optical signals and the required computing objectives, thereby further improving the computing performance of the optical processing chip 100.

[0043] For example, one superunit 21 can be used to perform superposition operation on the first optical signal, and another superunit 21 can be used to perform matrix multiplication operation on the first optical signal.

[0044] For example, multiple meta-units 21 can be used to process the same first optical signal sequentially to obtain a second optical signal, for example, Figure 4 In this process, two meta-units 21 are used to process the same first optical signal in sequence.

[0045] Multiple meta-units 21 can be connected by waveguides.

[0046] For example, when multiple superunits 21 are used to process the same first optical signal sequentially, the operations performed by the multiple superunits 21 are different, for example, Figure 4 In the process of using two superunits 21 to process the same first optical signal in sequence, the first superunit 21 performs a superposition operation on the first optical signal, and the second superunit 21 performs a matrix multiplication operation on the first optical signal after the superposition operation.

[0047] The multiple superstructure units 21 here can be arranged sequentially along the first direction.

[0048] In this way, by using multiple meta-units 21 to process the same first optical signal multiple times, not only can the flexibility and accuracy of optical computing be improved, but also some complex optical tasks that are difficult to achieve with a single meta-unit 21 (such as advanced image processing, optical pattern recognition, etc.) can be accomplished, thereby improving the applicability and competitiveness of the optical processing chip 100.

[0049] See Figure 1 and Figure 3 Each metaunit 21 includes a plurality of nanopillars N spaced apart.

[0050] Multiple nanopillars N are used to enable the metaunit 21 to modulate the first optical signal to complete optical computation. When the first optical signal is transmitted to the multiple nanopillars N, diffraction occurs. At this time, each nanopillar N can be regarded as a new diffraction source. The wavelets generated by these diffraction sources interfere with each other in space to form a diffracted light field. The diffracted light field carries the modulated optical signal information. Its light intensity and phase distribution will change according to the arrangement and size parameters of the nanopillars N and the initial characteristics of the first optical signal. Subsequently, the diffracted light field is collected and analyzed by other components to extract the effective signal after optical computation processing and complete the calculation of the first optical signal based on the principles of diffraction and interference. For example, multiple nanopillars N are used to modulate the light field of the first optical signal to complete matrix operations.

[0051] For example, see Figure 3 In the Z direction, the cross-sectional shape of the nanopillar N can be rectangular, trapezoidal, or other shapes, which are limited in this application.

[0052] For example, see Figure 3 Multiple nanopillars N can be arranged periodically. In this case, the interval d between two adjacent nanopillars N1 or two adjacent nanopillars N2 in each metaunit 2 can be the same or different in the direction Y, so as to adapt to different computing task requirements. This application does not impose any restrictions on this.

[0053] Among them, nanopillars N1 and N2 extend in opposite directions. Nanopillar N1 can extend in the positive direction of X, while nanopillar N2 can extend in the negative direction of X.

[0054] For example, the multiple nanopillars N in each metaunit 21 can be arranged in an array along the Y direction to obtain a row of nanopillars, that is, along the Z direction, a layer of nanopillars can be obtained to ensure that the first optical signal can be modulated by the layer of nanopillars N to realize optical computing.

[0055] For example, the multiple nanopillars N in each metaunit 21 can be arrayed along both the Y and Z directions, so that multiple nanopillars N are obtained in the Z direction. In this way, multiple nanopillars N can be used to perform complex calculations and further improve the accuracy of calculations on the first optical signal.

[0056] For example, the dimensions of the multiple nanopillars N in the meta-unit 21 in the X direction and the area of ​​the orthographic projection on the first surface can reach the hundred-nanometer level, thereby enabling the meta-unit 21 to occupy less area of ​​the light processing chip 100, thereby improving the integration of the light processing chip 100.

[0057] For example, each meta-unit 21 may also include a substrate, and the substrate may be disposed on the first surface. In this case, multiple nanopillars N may be disposed on the substrate to provide sufficient support for the multiple nanopillars N, making the structure of the meta-unit 21 more stable. This effectively reduces the displacement or deformation of the nanopillars N caused by external interference (such as temperature changes, mechanical vibrations, etc.) during optical signal processing, thereby improving the optical performance stability and reliability of the optical processing chip 100 and ensuring that the chip can maintain accurate optical signal processing capabilities under different working environments.

[0058] For example, each meta-unit 21 may also include only a plurality of nanopillars N, so as to ensure the computing power of the light processing chip 100 while reducing the area occupied by each meta-unit 21 in the light processing chip 100, thereby further improving the integration of the light processing chip 100.

[0059] For example, see Figure 1 and Figure 2 The size of the micro-nano metastructure 2 in the X direction is related to the number of meta-units 21 required for the actual task. That is, after analyzing the actual task requirements, it is determined that more meta-units 21 are needed. In this case, the size of the micro-nano metastructure 2 in the X direction is made larger so as to ensure that the optical processing chip 100 meets the task requirements while making its structure compact and highly integrated.

[0060] See Figure 1 and Figure 2 The aforementioned optical signal receiving structure 3 is connected to the micro / nano metastructure 2.

[0061] The optical signal receiving structure 3 is used to receive the second optical signal and extract the information carried by the second optical signal to provide accurate data for the subsequent processing of the optical processing chip 100, such as extracting information such as the phase, intensity, polarization, wavelength and calculation results of the second optical signal.

[0062] For example, the optical processing chip 100 described above can adopt a standardized packaging form, such as a butterfly package or a plug-in package, to ensure that the optical processing chip 100 can be easily connected to external optical fibers or other optical components.

[0063] For example, when packaging the light processing chip 100, the packaging material can be a transparent optical material, such as quartz or sapphire, to reduce the loss of light signals during the packaging process.

[0064] It is understood that, in the embodiments of this application, multiple meta-units 21 in the micro-nano metastructure 2 are used to perform parallel modulation on multiple first optical signals output from multiple optical output ports 11, i.e., multiple first optical signals after beam splitting, in order to achieve parallel operation, which can improve the computing efficiency of the optical processing chip 100. Furthermore, the multiple meta-units 21 are composed of multiple nanopillars N, and their overall size is small, which makes the area occupied by the meta-units 21 in the optical processing chip 100 smaller, thereby improving the integration of the optical processing chip 100. In turn, the optical processing chip 100 can meet the requirements of high-speed computing, miniaturization and high integration of the three-dimensional stacked optical computing architecture.

[0065] In some embodiments, see Figure 3 In the metaunit 21, at least some of the nanopillars N may have different dimensions in the first direction, and / or, at least some of the nanopillars N may have different cross-sectional areas.

[0066] Wherein, the first direction is the transmission direction of the first optical signal in the waveguide, and the cross-sectional area is perpendicular to the first direction, that is, the cross-sectional area of ​​the nanopillar N can be the area of ​​the orthogonal projection of the nanopillar N in the first direction.

[0067] The aforementioned multiple nanopillars N can form an apodized structure, which can precisely control the propagation path and intensity distribution of the first optical signal, making its optical field smoother. By rationally designing the arrangement, size parameters, and spacing of the nanopillars N, the sidelobe effect of the first optical signal during transmission can be effectively suppressed, the scattering loss of the first optical signal can be reduced, and the crosstalk of the signal can be reduced, thereby improving the computational speed and accuracy of the metaunit 21, and thus improving the computational performance of the optical processing chip 100.

[0068] For example, the dimensions of multiple nanopillars N in the first direction can be designed using mathematical functions, custom rules, etc., to form the desired apodized structure. For instance, the dimensions of multiple nanopillars N in the first direction can be designed using Gaussian functions, cosine functions, etc., to form Gaussian apodized structures, cosine apodized structures, etc.

[0069] For example, see Figure 3 At least two nanopillars N1 in the nanopillar N can have different dimensions in the direction X (i.e., the first direction) in order to adjust the coupling efficiency and phase delay of the nanopillar N1, thereby assigning different weights to the optical signals passing through each nanopillar N1. In this case, a nanopillar N1 can be regarded as a weighting unit, and its combination with nanopillar N2 can realize the weighted calculation of optical signals and the function of modulating optical signals.

[0070] The phase delay can be calculated using the formula Φ=L1×β=L1×2πn / λ, where L1 represents the size of the nanopillar N1 in the X direction, β represents the phase constant, λ represents the wavelength of the first optical signal, and n represents the effective refractive index of the nanopillar N1.

[0071] The effective refractive index of the aforementioned nanopillar N1 is used to reflect the degree to which the propagation speed of the first optical signal in the nanopillar N1 is slowed down relative to its propagation speed in a vacuum.

[0072] Figure 5 Top view of another superstructure 21 provided for implementation of this application For example, see Figure 3 and Figure 5 Along the Y direction, the diffraction and interference of the first optical signal can be precisely controlled by adjusting the interval d between two adjacent nanopillars N1 or N2 in the multiple nanopillars N, thereby realizing optical computation of the first optical signal.

[0073] Diffraction can be described by the following diffraction formula: where θ d λ represents the diffraction angle of the first optical signal, λ represents the wavelength of the first optical signal, and d represents the spacing between two adjacent nanopillars N1 or N2 in the direction Y.

[0074] For example, see Figure 3 and Figure 5 Along the Y direction, the spacing d between two adjacent nanopillars N1 or N2 in the plurality of nanopillars N can be the same or different to enrich the structure of the metaunit 21, enabling it to perform a variety of operations, thereby improving the applicability of the optical processing chip 100.

[0075] In this way, the dimensions and cross-sectional areas of some of the nanopillars N in the meta-unit 21 are different in the first direction, which can form an apodized structure. This improves the computing speed and accuracy of the meta-unit 21, while reducing the loss of the first optical signal. As a result, the optical processing chip 100 can meet the requirements of the three-dimensional stacked optical computing architecture for optical computing efficiency and performance.

[0076] In some embodiments, see Figure 3 Each superstructure 21 includes a first component 211 and a second component 212, and the first component 211 and the second component 212 are arranged along a first direction.

[0077] The first direction is perpendicular to the first surface of the optical coupling structure 1.

[0078] The first component 211 includes a substrate 2111 and a plurality of nanopillars N1, and the second component 212 includes a substrate 2121 and a plurality of nanopillars N2.

[0079] Wherein, the base 2111 of the first component 211 and the base 2121 of the second component 212 can be disposed opposite to each other along the first direction, for example, Figure 3 In the middle, along the direction X, substrate 2111 is disposed on the left side of nanopillar N, and substrate 2121 is disposed on the right side of nanopillar N.

[0080] See Figure 3 At least some of the aforementioned nanopillars N1 can be engaged with at least some of the nanopillars N2, such that a microcavity (i.e., a resonant cavity) is formed between adjacent nanopillars N1 and N2 in the Y direction, thereby making each metaunit 21 include a plurality of such microcavities.

[0081] The multiple nanopillars N1 here can be a comb-like structure, and the multiple nanopillars N2 can be a toe-like structure.

[0082] The aforementioned microcavities can cause the first optical signal entering the meta-unit 21 to undergo multiple reflections and interferences within the cavity, forming a standing wave, thereby enhancing the light intensity of the first optical signal. Furthermore, each microcavity has a specific mode field distribution. When the first optical signal enters the meta-unit 21, it will couple with the specific mode in the microcavity, thereby causing the optical signal to propagate in the direction of propagation of the microcavity mode.

[0083] For example, the microcavity described above can be used to enable the meta-unit 21 to perform weighted multiplication or participate in the operation as an activation function based on the resonant characteristics of the microcavity, thereby enabling the meta-unit 21 to have efficient optical signal processing capabilities.

[0084] In optical signal processing, the aforementioned weighted multiplication is a crucial step in simulating neural network computation. By precisely controlling the intensity and phase of the optical signal through the resonant characteristics of the microcavity, multiplication operations between different input signals and their corresponding weights can be completed quickly and accurately. When participating in the computation as an activation function, the superunit 21 can perform nonlinear transformations on the multiplication results according to preset rules, further simulating the response characteristics of biological neurons.

[0085] For example, when the microcavity participates in the computation as an activation function, the material of the multiple nanopillars N may also include a phase change material, so as to utilize the properties of the phase change material to enable the metaunit 21 to perform nonlinear calculations.

[0086] In this way, the microcavity can be used to improve the computing speed and complex computing capabilities of the metacell 21, thereby improving the computing efficiency of the optical processing chip 100 and enabling it to meet the high-speed computing requirements of the three-dimensional stacked computing architecture.

[0087] For example, the microcavity described above can also be used to enable the meta-unit 21 to process optical signals of a specific wavelength to achieve filtering, or the sensitivity of the resonant wavelength of the microcavity to changes in the external environment can be used to enable the meta-unit 21 to have optical sensing function characteristics, thereby improving the applicability of the optical processing chip 100 and enabling it to be applicable to a variety of application scenarios. For example, see Figure 3 The size of each microcavity can be adjusted by making at least some of the nanopillars N in the metaunit 21 have different cross-sectional areas, and / or by making the spacing d between two adjacent nanopillars N1 or two nanopillars N2 different, for example. Figure 3 In the process, the dimension L of the microcavity in the direction Y is adjusted, which is the resonant wavelength of each microcavity. By adjusting the resonant wavelength of the microcavity resonator, the transmission loss or phase change of the first optical signal in the microcavity can be controlled. This allows the meta-unit 21 to process optical signals of multiple wavelengths, realize parallel computing or filtering, and perform various complex operations, thereby improving the applicability of the optical processing chip 100.

[0088] For example, when filtering is achieved based on the interval d between two adjacent nanopillars N1 or two nanopillars N2 in a plurality of nanopillars N, the interval d between two adjacent nanopillars N1 or two nanopillars N2 in the plurality of nanopillars N can be adjusted so that the wavelength and incident angle of the first optical signal satisfy the Bragg condition to produce strong diffraction, thereby achieving wavelength separation. With the help of the microcavity formed by nanopillars N1 and N2, after the first optical signal that satisfies the Bragg condition passes through the microcavity, the first optical signal that matches the resonant wavelength is selected for enhancement to achieve filtering.

[0089] The diffraction angle of the first optical signal can be calculated using the formula dsin(θ)=mλ, where θ represents the diffraction angle of the first optical signal, d represents the interval between two adjacent nanopillars N1 or two nanopillars N2, m represents the diffraction order, and λ represents the wavelength of the first optical signal.

[0090] For example, when filtering is achieved by adjusting the size of the microcavity in the Y direction, the effective length L of multiple microcavities can also be calculated. et Determine the frequency and effective length L of the optical signal that can pass through the metaunit 21. et The larger the value, the lower the frequency of the optical signal that can pass through the metaunit 21, and vice versa.

[0091] Wherein, the effective length is the sum of the effective lengths of all nanopillars N and the effective lengths of the intervals d1 between all adjacent nanopillars N1 and N2; the effective length of nanopillar N is the product of the size of nanopillar N in the Y direction and its refractive index; the effective length of the interval d1 between adjacent nanopillars N1 and N2 is the product of the interval d1 and the refractive index of the medium between adjacent nanopillars N1 and N2, for example, the product of the interval d1 and the refractive index of air.

[0092] For example, see Figure 5 When multiple nanopillars N have the same dimension in the Y direction, and the spacing between two adjacent nanopillars N1 is the same, and the spacing between two adjacent nanopillars N2 is also the same, the effective length is: L et =(M1×L2)×n1+(M2-1)×d1)×n2 Where M1 represents the number of nanopillars N in a metaunit 21, L2 represents the size of nanopillar N in the Y direction, d1 represents the spacing between adjacent nanopillars N1 and N2, M2 represents the number of spacings d1, n1 represents the refractive index of nanopillar N, and n2 represents the refractive index of the medium in the spacing.

[0093] For example, see Figure 5 When multiple nanopillars N have different dimensions in the Y direction, or the spacing between two adjacent nanopillars N1 is different, and the spacing between two adjacent nanopillars N2 is also different, the effective length of each nanopillar N and the effective length of each spacing d1 can be calculated separately, and then added together to obtain the effective length L of the multiple microcavities. et .

[0094] For example, see Figure 3 and Figure 5 The size of the nanopillar N in the X direction, the cross-sectional area of ​​the nanopillar N, and the spacing between two adjacent nanopillars N1 or N2 can be adjusted simultaneously to adjust the coupling efficiency and phase delay of the nanopillar N1, as well as the size of the microcavity formed by the nanopillars N1 and N2 in the Y direction. This allows the metaunit 21 to perform more complex operations and realize the functional integration of the metaunit 21.

[0095] In this way, at least a portion of the nanopillars N1 of the first component 211 and at least a portion of the nanopillars N2 of the second component 212 are engaged to form a microcavity. This allows the meta-unit 21 to selectively weight the input optical signal with its resonant wavelength, supporting wavelength multiplexing, enabling parallel computing, and participating in computation as an activation function. This gives the meta-unit 21 high-efficiency computing capabilities, thereby enabling the optical processing chip 100 to be applied in various scenarios requiring high efficiency and complexity.

[0096] Figure 6 This is a schematic diagram of the overall structure of another optical processing chip provided in this application.

[0097] In some embodiments, see Figure 1 and Figure 6 Each superunit 21 may include at least one component.

[0098] Each component includes a first component 211 and a second component 212.

[0099] For example, see Figure 1 Each superunit 21 may include a component to ensure that each superunit 21 can process or calculate optical signals while reducing the area occupied by each superunit 21 in the optical processing chip 100, thereby increasing the integration of the optical processing chip 100.

[0100] See Figure 6 Each meta-component 21 may include multiple components, and these components can be arranged sequentially along a first direction, i.e., from left to right along direction X, so that each meta-unit 21 can make more precise adjustments to each beam of the first optical signal, thereby making the calculation results of the first optical signal more accurate and improving the calculation accuracy of the optical processing chip 100. For example, Figure 6 In this context, each superunit 21 comprises three components.

[0101] For example, in multiple meta-units 21, the number of components included in different meta-units 21 can be the same, so as to simplify the structural design of the meta-units 21, thereby reducing the fabrication difficulty of the meta-units 21 while ensuring the computational speed and accuracy of the meta-units 21.

[0102] Alternatively, for example, the number of components included in different superunits 21 may be different, so that different superunits 21 can perform different operations, thereby completing different calculation tasks such as superposition, subtraction, and multiplication, and obtaining a second optical signal carrying different information, thereby improving the complex computing power and applicability of the optical processing chip 100.

[0103] In this way, the computing speed and accuracy of the superunit 21 can be adjusted by adjusting the number of components in each superunit 21. This allows the optical processing chip 100 to dynamically adjust the component configuration of its internal superunit 21 according to different application scenarios and requirements, thereby enabling it to be accurately applied to various optical computing, signal processing, and optical communication tasks with different requirements for computing performance.

[0104] Figure 7 This is a schematic diagram of an optical coupling structure provided in this application.

[0105] In some embodiments, see Figure 7 The optical coupling structure 1 in the above embodiment includes a substrate 14, a waveguide layer 13 and a grating layer 12 stacked sequentially.

[0106] For example, the light processing chip 100 can share a substrate 14 with the light coupling structure 1, so as to simplify the fabrication process, reduce the fabrication cost, and improve the integration of the light processing chip 100.

[0107] The aforementioned waveguide layer 13 includes multiple waveguides, which can form a waveguide beam splitting structure to split the optical signal coupled to the waveguide through the optical coupling structure 1 to obtain multiple first optical signals, so that they can be processed in parallel by multiple meta-units 21 to achieve parallel computing, thereby improving the computing efficiency of the optical processing chip 100.

[0108] See Figure 1 , Figure 2 and Figure 4 The ends of the aforementioned multiple waveguides form multiple optical output ports 11, that is, the output ports of the waveguide beam splitting structure are optical output ports 11, so as to output multiple first optical signals.

[0109] See Figure 7 The aforementioned grating layer 12 includes a plurality of spaced nanopillars 121, and the extension direction of these nanopillars 121 is perpendicular to the substrate 14, i.e., extending along the Z direction.

[0110] For example, the optical properties of light, such as phase, polarization, and amplitude, can be flexibly controlled by precisely adjusting parameters such as the shape, size, spacing, and rotation angle of the nanopillars 121, thereby optimizing the optical coupling efficiency of the optical coupling structure 1 and ensuring that the optical signal is coupled into the waveguide in the best possible state.

[0111] For example, once the shape, size, spacing and rotation angle of the multiple nanopillars 121 in the grating layer 12 are determined, they can be adapted to process optical signals within a preset wavelength range. That is, multiple wavelengths of optical signals within the wavelength range can be independently coupled and controlled. When the shape, size, spacing and rotation angle of the multiple nanopillars 121 in the grating layer 12 are changed, the preset wavelength range adapted to by the optical coupling structure 1 will also change.

[0112] For example, when each nanopillar 121 is rectangular in shape, with a size of approximately 580nm × 210nm and a height of 550nm, and the period of the nanopillar 121 is 650nm, the optical coupling structure can be adapted to process optical signals with a preset wavelength range of 1310nm to 1550nm. Taking an optical signal with a wavelength of 1550nm as an example, its transmittance is 0.94 and its conversion efficiency is 0.88.

[0113] In this way, the optical coupling structure 1 can be made insensitive to wavelength and polarization, effectively coupling optical signals in a wide wavelength range, reducing polarization loss, thereby improving the functionality and efficiency of the optical processing chip 100. In addition, the optical coupling structure 1 also has a large aperture receiving capability and a large effective photosensitive area, which can receive a wide beam of light, reducing the requirements for optical path alignment accuracy, further improving the stability and reliability of the optical processing chip 100. Furthermore, the optical coupling structure 1 utilizes nanopillars 121 to achieve efficient optical coupling, making it occupy a smaller area of ​​the optical processing chip 100, which can further improve the integration of the optical processing chip 100.

[0114] In some embodiments, the constituent materials of the plurality of nanopillars N in each metaunit 21 may be the same as the constituent materials of the waveguide layer 13 in the optical coupling structure 1.

[0115] For example, the constituent materials of the multiple nanopillars N in each metaunit 21 and the constituent material of the waveguide layer 13 can both be high-resistivity silicon, so as to effectively reduce the loss in the optical signal transmission process by utilizing the high resistivity characteristics of high-resistivity silicon. At the same time, high-resistivity silicon has a moderate refractive index in the visible and near-infrared bands, which can form a good optical match with the surrounding medium, thereby achieving efficient optical coupling and optical modulation.

[0116] In this way, the constituent materials of the multiple nanopillars N in each metaunit 21 can be the same as the constituent materials of the waveguide layer 13 in the optical coupling structure 1. This can reduce the difficulty and cost of fabrication, and also provide a solid foundation for the synergistic work of the metaunit 21 and the waveguide layer 13, thereby improving the performance and stability of the entire optical processing chip 100.

[0117] In some embodiments, see Figure 1 , Figure 2 and Figure 4 The light processing chip 100 also includes a photodetector structure 4.

[0118] Among them, the micro-nano metastructure 2 and the optical signal receiving structure 3 can be connected through the photoelectric detection structure 4.

[0119] The aforementioned photoelectric detection structure 4 is used to receive the second optical signal and convert the second optical signal into a corresponding electrical signal so that other structures can use the electrical signal to obtain the information carried by the second optical signal.

[0120] The optical signal receiving structure 3 is configured to receive electrical signals and obtain the information carried by the electrical signals.

[0121] For example, see Figure 6The aforementioned photoelectric detection structure 4 can also replace the optical signal receiving structure 3. In this case, the photoelectric detection structure 4 can convert the optical signal into an electrical signal and further process the electrical signal to achieve detection, such as optical detection. This simplifies the structure of the optical processing chip 100 and improves its integration.

[0122] For example, see Figure 6 The aforementioned photoelectric detection structure 4 adopts an array structure, which consists of multiple high-sensitivity photoelectric detection units 41, and the number of rows and columns of the array is determined according to the number of output channels of the actual micro-nano metastructure 2.

[0123] For example, for a micro-nano metastructure 2 with 16 output channels, a 2-row × 8-column photodetector array can be designed, with each photodetector unit 41 having a size of approximately 200 micrometers × 200 micrometers.

[0124] For example, the photoelectric detection structure 4 can be made of semiconductor materials, such as indium gallium arsenide (InGaAs) and germanium (Ge). These materials have high sensitivity and response speed to optical signals with wavelength range of 1200nm-1700nm, so that the photoelectric detection structure 4 can efficiently capture and convert optical signals into electrical signals in this specific wavelength range, and accurately realize the detection and analysis of optical information in the corresponding wavelength band.

[0125] For example, the photodetector structure 4 described above can adopt a P-type semiconductor-intrinsic semiconductor-N-type semiconductor (PIN) structure or an avalanche photodiode (APD) structure to improve its detection efficiency and gain.

[0126] For example, a focusing structure or waveguide structure can also be set between the photodetector structure 4 and the micro / nano metastructure 2 to focus the optical signal onto the sensitive area of ​​the photodetector structure 4 and improve the coupling efficiency of the optical signal.

[0127] The electrical connection of the aforementioned photoelectric detection structure 4 can be achieved using flip-chip bonding technology to connect the detector chip to the signal processing circuit inside the chip, ensuring that the electrical signal can be transmitted quickly and stably.

[0128] Since the photoelectric detection structure 4 is made of photoelectric detection material with high sensitivity and high response speed, and can be closely matched with the micro-nano metastructure 2, it can quickly and accurately convert the calculated optical signal into an electrical signal, improve the sensitivity and accuracy of optical detection, enable the optical processing chip 100 to output electrical signals efficiently and accurately, and thus broaden the application range of the optical processing chip 100, making it adaptable to optical signal input of different intensities.

[0129] For example, the aforementioned optical processing chip also includes a cross-group amplifier circuit 5 and an output circuit 6.

[0130] Among them, the cross-group amplifier circuit 5 can be set after the optical signal receiving structure 3, and the output circuit 6 can be set after the cross-group amplifier circuit 5.

[0131] The aforementioned cross-group amplifier circuit 5 can adopt a two-stage amplifier structure. The first stage amplifier adopts a low-noise amplifier (LNA) to initially amplify the weak electrical signal output by the photodetector structure 4.

[0132] The low-noise amplifier here has a gain of approximately 20dB and a bandwidth of 100MHz-3GHz.

[0133] The second-stage amplifier uses a variable gain amplifier (VGA) to further amplify the signal. The gain can be adjusted according to the actual signal strength to achieve adaptive signal amplification. The gain adjustment range of the variable gain amplifier is 0-30dB.

[0134] The output circuit 6 is used to shape and match the amplified electrical signal to ensure that the signal finally output by the optical processing chip 100 meets the interface requirements of other components or systems in the three-dimensional stacked optical computing architecture.

[0135] The output circuit here can adopt a differential output structure to effectively suppress common-mode noise and improve the signal transmission quality.

[0136] The aforementioned output circuit 6 also has a signal level conversion function, which can convert the signal level inside the optical processing chip 100 into a level standard compatible with external systems, such as CMOS level or TTL level. Furthermore, the output impedance of the output circuit 6 can be matched to 50Ω to achieve good matching with external transmission lines and reduce signal reflection and loss.

[0137] The following is an example of using the optical processing chip 100 in the above embodiment to perform matrix multiplication calculations to complete the detection: The central design wavelength of the optical coupling structure 1 is 1550nm. The cross-sectional size of each nanopillar 121 in the grating layer 12 is 580nm×210nm, the size in the Z direction is 550nm, and the period is 650nm. Through precise design, the light transmittance in free space is 94%, and the optical coupling efficiency reaches 88%.

[0138] Based on the design of the optical coupling structure 1 described above, a wide-band light beam is input into the optical processing chip 100. If we assume there is a multi-wavelength input signal, with each wavelength component carrying a data, when the light beam enters the micro-nano metastructure 2, the spacing d between adjacent nanopillars N in the first component 211 or the second component 212 of the metastructure unit 21 is designed according to the wavelength, so that the different microcavities formed by the adjacent nanopillars N1 and N2 respond to specific wavelengths (i.e., resonate). Then the output (transmission or reflection) of each microcavity is equivalent to the weighting of the input signal on the corresponding wavelength component. Combining these weighted outputs achieves matrix multiplication.

[0139] At this point, the photodetector structure 4 can use a photodiode based on semiconductor materials, which has a response speed of up to 10 GHz. It can quickly convert the second optical signal into an electrical signal. The converted electrical signal is transmitted to the cross-group amplifier 5 and the output circuit 6. The cross-group amplifier 5 adopts a two-stage amplifier structure to amplify the electrical signal to a suitable amplitude. Then, after being shaped and matched by the output circuit 6, it is output to the subsequent processing module of the three-dimensional stacked optical computing architecture to complete the entire photodetector process.

[0140] The following is an example of implementing a multi-wavelength filtering task using the optical processing chip 100 in the above embodiments: The optical coupling structure 1 on the optical processing chip 100 is used to efficiently couple optical signals of different wavelengths. By adjusting the cavity length L of the microcavity in the meta-unit 21 and the spacing d between adjacent nanopillars N in the first component 211 or the second component 212, the enhancement of light of a specific wavelength is achieved in the resonant cavity.

[0141] At this time, within the resonant cavity, the optical signal is reflected multiple times and forms interference. When the phase difference is an integer multiple of 2π, constructive interference occurs, and the light intensity is significantly enhanced, so that light of a specific wavelength forms stable constructive interference within the cavity, thereby enhancing the light intensity of that wavelength.

[0142] The frequency corresponding to this enhanced light intensity is the selected frequency, thereby enabling differentiated calculation and processing of light signals of different wavelengths.

[0143] This application also provides an optical computing system. Figure 8 This is a schematic diagram of the structure of an optical computing system according to an embodiment of this application.

[0144] See Figure 8 The optical computing system 200 includes the optical processing chip 100 and the memory 101 described in the above embodiments.

[0145] When the optical computing system is applied to an optical neural network, the aforementioned memory 101 stores multiple neural network weights, and the memory 101 is connected to the micro-nano metastructure 2 in the optical processing chip 100.

[0146] The micro-nano metastructure 2 here can be designed based on multiple neural network weights and process multiple first optical signals into second optical signals.

[0147] The optical computing system 200 described above utilizes the optical processing chip 100 in the above embodiment to process optical signals, and therefore has the same beneficial technical effects as the optical processing chip 100, which will not be elaborated here.

[0148] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A light processing chip, characterized in that, include: An optical coupling structure, wherein a first surface of the optical coupling structure has a plurality of optical output ports, the plurality of optical output ports being used to output a plurality of first optical signals; A micro / nano metastructure includes multiple meta-units disposed on a first surface, and the orthographic projections of the multiple meta-units on the first surface correspond one-to-one with the multiple light output ports; each meta-unit includes multiple nanopillars spaced apart; the meta-unit is used to process the multiple first optical signals into second optical signals, wherein the first optical signals and the second optical signals are different; An optical signal receiving structure is connected to the micro / nano metastructure, and the optical signal receiving structure is configured to receive the second optical signal and obtain the information carried by the second optical signal. Each of the meta-units includes a first component and a second component, the first component and the second component being arranged along a first direction; the first direction is perpendicular to the first surface; The first component and the second component each include a substrate and the plurality of nanopillars. At least a portion of the nanopillars of the first component are engaged with at least a portion of the nanopillars of the second component. The substrates of the first component and the substrates of the second component are disposed opposite to each other along the first direction.

2. The optical processing chip according to claim 1, characterized in that, In the metaunit, at least some of the nanopillars have different dimensions in a first direction; and / or, Among the multiple nanopillars in the metacellular unit, at least some of the nanopillars have different cross-sectional areas; the cross-sectional areas are perpendicular to the first direction.

3. The optical processing chip according to claim 1 or 2, characterized in that, Each of the superstructures includes at least one component, and each component includes the first component and the second component; when the superstructure includes multiple components, the multiple components are arranged sequentially along the first direction.

4. The optical processing chip according to claim 3, characterized in that, The number of components varies among the multiple superunits.

5. The optical processing chip according to claim 1, characterized in that, The optical coupling structure includes a substrate, a waveguide layer, and a grating layer stacked sequentially; the waveguide layer includes multiple waveguides, and the ends of the multiple waveguides form the multiple optical output ports; the first surface is perpendicular to the substrate; The grating layer includes a plurality of spaced nanopillars, the extension direction of which is perpendicular to the substrate.

6. The optical processing chip according to claim 5, characterized in that, The material of the multiple nanopillars of the metaunit is the same as that of the waveguide layer.

7. The optical processing chip according to claim 1, characterized in that, Also includes: A photoelectric detection structure, wherein the micro / nano metastructure and the optical signal receiving structure are connected through the photoelectric detection structure; The photoelectric detection structure is configured to receive the second optical signal and convert the second optical signal into a corresponding electrical signal; The optical signal receiving structure is configured to receive the electrical signal and obtain the information carried by the electrical signal.

8. An optical computing system, characterized in that, include: The light processing chip as described in any one of claims 1 to 7; The memory stores multiple neural network weights and is connected to a micro / nano metastructure in the optical processing chip. The micro / nano metastructure processes the multiple first optical signals into second optical signals based on the multiple neural network weights.

Images