Photonic integrated circuits and their manufacturing methods, photonic computing systems, photonic computing chips

Abstract

This invention relates to the field of semiconductors, and provides a photonic integrated circuit and its manufacturing method, a photonic computing system, and a photonic computing chip. The photonic integrated circuit may include multiple directional couplers, including a first directional coupler and a second directional coupler, which have different beam splitting ratios. The first and second directional couplers may each have a region of metal-semiconductor compound material, and the distribution of the metal-semiconductor compound material region contributes to the beam splitting ratio.

Description

Technical Field

[0001] This invention relates to the field of semiconductors, and more specifically, to photonic integrated circuits and their manufacturing methods, photonic computing systems, and photonic computing chips. Background Technology

[0002] In some applications, directional couplers are integrated into photonic integrated circuits. A directional coupler is a photonic device that distributes the energy of light. It typically consists of two waveguide arms spaced close together in the coupling region, allowing the optical energy (optical power) on one waveguide arm to be coupled to the other. Directional couplers are generally manufactured with a specific splitting ratio. To obtain directional couplers with different splitting ratios, the length and spacing of the waveguide arms in the coupling region, or other geometric features of the directional coupler can be changed.

[0003] In photonic computing and optical interconnect technologies, photonic devices or photonic integrated circuits are used as the hardware foundation. The stability, energy consumption, manufacturing complexity, and cost of these devices or photonic integrated circuits are factors that need to be considered. Summary of the Invention

[0004] This invention provides a photonic integrated circuit and its manufacturing method, a photonic computing system, and a photonic computing chip.

[0005] In one exemplary embodiment, a photonic integrated circuit is provided, comprising: a first directional coupler including a first arm and a second arm, wherein the first arm and the second arm of the first directional coupler respectively include a first waveguide coupling section and a second waveguide coupling section, and the light output intensities of the first arm and the second arm of the first directional coupler are different, and the ratio of the light output intensities of the first arm and the second arm is a first beam splitting ratio, wherein at least one of the first arm and the second arm of the first directional coupler includes a metal semiconductor compound material, and the distribution of the metal semiconductor compound material of the first directional coupler contributes to the allocation of the first beam splitting ratio; and a second directional coupler including a first arm and a second arm, wherein the first arm and the second arm of the second directional coupler respectively include a first waveguide coupling section and a second waveguide coupling section, wherein the light output intensities of the first arm and the second arm of the second directional coupler are different, and the ratio of the light output intensities of the first arm and the second arm is a second beam splitting ratio, wherein at least one of the first arm and the second arm of the second directional coupler includes a metal semiconductor compound material, and the distribution of the metal semiconductor compound material of the second directional coupler contributes to the allocation of the second beam splitting ratio; wherein the first beam splitting ratio and the second beam splitting ratio are different.

[0006] In some embodiments, the first arm of the first directional coupler includes a first metal-semiconductor compound layer, such that at least one of the first arm and the second arm of the first directional coupler includes a metal-semiconductor compound material; the first arm of the second directional coupler includes a second metal-semiconductor compound layer, such that at least one of the first arm and the second arm of the second directional coupler includes a metal-semiconductor compound material.

[0007] In some embodiments, the first directional coupler includes a waveguide body layer, and the first metal-semiconductor compound layer is disposed above and / or on the sidewall of the waveguide body layer in the first waveguide coupling segment of the first directional coupler.

[0008] In some embodiments, the first metal-semiconductor compound layer is obtained by heat treatment of a metal material and a semiconductor material.

[0009] In some embodiments, the photonic integrated circuit is formed based on a semiconductor-on-insulator layer.

[0010] In some embodiments, the photonic integrated circuit is formed based on silicon-on-insulator.

[0011] In some embodiments, the metal semiconductor compound material includes at least one semiconductor element selected from silicon or germanium.

[0012] In some embodiments, the metal semiconductor compound material includes at least one metallic element selected from nickel, cobalt, titanium, tungsten, tantalum, platinum, palladium, and erbium.

[0013] In one exemplary embodiment, a method for manufacturing a photonic integrated circuit is provided. The photonic integrated circuit includes a plurality of directional couplers, each of the plurality of directional couplers including a first arm and a second arm. The beam splitting ratio of the directional coupler is the ratio of the light intensity emitted by the first arm to the light intensity emitted by the second arm. The first arm and the second arm respectively include a first waveguide coupling segment and a second waveguide coupling segment. The manufacturing method includes: forming a waveguide body layer for the first arm and a waveguide body layer for the second arm of each of the plurality of directional couplers; forming a metal semiconductor compound region according to the beam splitting ratio, such that at least one of the first arm and the second arm includes a metal semiconductor compound material, the distribution of the metal semiconductor compound material contributing to the beam splitting ratio distribution.

[0014] In some embodiments, the plurality of directional couplers includes a first directional coupler and a second directional coupler, wherein the first directional coupler and the second directional coupler have different beam splitting ratios.

[0015] In some embodiments, forming the metal-semiconductor compound region includes forming a first metal-semiconductor compound layer over and / or on the sidewall of the waveguide body layer in the first waveguide coupling segment of the first directional coupler, and forming a second metal-semiconductor compound layer over and / or on the sidewall of the waveguide body layer in the first waveguide coupling segment of the second directional coupler.

[0016] In some embodiments, forming the metal-semiconductor compound region includes: forming an initial waveguide layer, the initial waveguide layer comprising a semiconductor material; forming a metal material layer; and causing a portion of the semiconductor material of the initial waveguide layer to react with the metal material in the metal material layer to form the metal-semiconductor compound region; wherein the waveguide body layer includes the initial waveguide layer that does not react with the metal material layer.

[0017] In some embodiments, forming the metal-semiconductor compound region includes: forming a semiconductor material layer on the waveguide host layer; forming a metal material layer; and causing the metal material layer to react with the semiconductor material layer to form the metal-semiconductor compound region.

[0018] In some embodiments, the metal semiconductor compound material includes at least one semiconductor element selected from silicon or germanium.

[0019] In some embodiments, the metal semiconductor compound material includes at least one metallic element selected from nickel, cobalt, titanium, tungsten, tantalum, platinum, palladium, and erbium.

[0020] In some embodiments, the metal semiconductor compound material includes at least one semiconductor element selected from silicon or germanium.

[0021] In some embodiments, the method for manufacturing a photonic integrated circuit includes a metal semiconductor compound material comprising at least one metallic element selected from nickel, cobalt, titanium, tungsten, tantalum, platinum, palladium, and erbium.

[0022] In one exemplary embodiment, a photonic computing system is provided, comprising a photonic integrated circuit. The photonic computing system includes: a photonic computing unit configured to receive a first plurality of optical signals, the first plurality of optical signals corresponding to a first group of multiple values; the photonic computing unit includes a plurality of weighting modules, the plurality of weighting modules corresponding to a plurality of preset values, each weighting module corresponding to a preset value, each weighting module including: an optical input unit configured to receive one optical signal from the first plurality of optical signals, and at least one directional coupler, wherein each directional coupler is designed to achieve a preset splitting ratio, such that the weighting module corresponds to one preset value and can be used to perform a multiplication operation between the value corresponding to the one optical signal input by the optical input unit and the preset value corresponding to the weighting module; wherein the plurality of weighting modules includes a first weighting module and a second weighting module, the at least one directional coupler in the first weighting module includes the first directional coupler, and the at least one directional coupler in the second weighting module includes the second directional coupler.

[0023] In some implementations, the at least one directional coupler is configured during the manufacturing stage to have a spectral ratio associated with the preset value corresponding to the weighting module, based on the preset value corresponding to the weighting module.

[0024] In some implementations, the photonic computing unit includes one or more summing modules, and for the outputs of two or more weighting modules, a corresponding summing module is configured to generate an electrical signal representing the sum of the multiplication results corresponding to the two or more weighting modules.

[0025] In some embodiments, a photonic computing chip is provided, including a photonic integrated circuit.

[0026] In some implementations, the first splitting ratio of the first directional coupler is associated with a first value, and the second splitting ratio of the second directional coupler is associated with a second value, wherein the first value and the second value are used for calculation.

[0027] In some embodiments, a photonic computing unit is included, configured to receive a first plurality of optical signals, the first plurality of optical signals corresponding to a first group of multiple values; the photonic computing unit includes a plurality of weighting modules, the plurality of weighting modules corresponding to a plurality of preset values, each weighting module corresponding to a preset value, each weighting module including: an optical input unit configured to receive one optical signal from the first plurality of optical signals, and at least one directional coupler, wherein each directional coupler is designed to achieve a preset splitting ratio, such that the weighting module corresponds to one preset value, and can be used to perform a multiplication operation between the value corresponding to the one optical signal input by the optical input unit and the preset value corresponding to the weighting module; wherein the plurality of weighting modules includes a first weighting module and a second weighting module, the at least one directional coupler in the first weighting module includes the first directional coupler, and the at least one directional coupler in the second weighting module includes the second directional coupler.

[0028] In some implementations, the at least one directional coupler is configured during the manufacturing stage to have a spectral ratio associated with the preset value corresponding to the weighting module, based on the preset value corresponding to the weighting module.

[0029] In some implementations, the first plurality of optical signals are used to represent one or more elements in a vector, and the plurality of weight modules are used to represent multiple elements in a matrix.

[0030] According to one aspect of the present invention, when it is necessary to change the beam splitting ratio of a directional coupler, this can be achieved by designing a suitable region of the metal-semiconductor compound layer without altering the initial waveguide layer (or waveguide body layer) of the directional coupler and the associated photonic devices, or without requiring significant redesign or extensive redesign or layout of the entire photonic integrated circuit. Exemplarily, the present invention can be applied to directional couplers in photonic integrated circuits with different required beam splitting ratios. Exemplarily, the formation region of the corresponding metal-semiconductor compound layer can be set using a mask, thereby allowing the beam splitting ratio of the directional coupler to be changed simply by adjusting the corresponding mask.

[0031] Various aspects, features, advantages, etc., of the embodiments of the present invention will be specifically described below in conjunction with the accompanying drawings. These aspects, features, advantages, etc., will become clearer from the following detailed description in conjunction with the accompanying drawings. Attached Figure Description

[0032] Figure 1A A top view of a directional coupler in a photonic integrated circuit is shown;

[0033] Figure 1BIt shows Figure 1A A cross-sectional view of a photonic integrated circuit at a specific location in the waveguide coupling region;

[0034] Figures 2A-2C This illustrates some states in the fabrication of photonic integrated circuits;

[0035] Figure 3-5 The distribution of the metal semiconductor compound material is shown;

[0036] Figure 6A A photonic computing system is shown;

[0037] Figure 6B A specific example of a photonic computing system is shown;

[0038] Figure 7 This demonstrates the use of a directional coupler as a weight module;

[0039] Figure 8 A photonic computing system is shown;

[0040] Figure 9 The weighting module is shown to include two directional couplers;

[0041] Figure 10 The weighting module is shown to include an MMI (Multimode Interference Coupler);

[0042] Figure 11 The weighting module is shown to consist of two MMIs. Detailed Implementation

[0043] To facilitate understanding of the various aspects, features, and advantages of the technical solution of this invention, the invention will be described in detail below with reference to the accompanying drawings. It should be understood that the various embodiments described below are for illustrative purposes only and are not intended to limit the scope of protection of this invention.

[0044] In photonic integrated circuits, directional couplers are devices that distribute light energy. In some scenarios, photonic integrated circuits include directional couplers, which are typically manufactured according to a specific beam splitting ratio. The inventors have discovered that when a new beam splitting ratio is required, manufacturing is often tailored to the new requirements. This often necessitates altering the length and spacing of the waveguide arms in the coupling region of the directional coupler, or other geometric features of the directional coupler. During photonic device integration or interconnection, the associated structures, devices, or the overall layout of the photonic integrated circuit will change accordingly, significantly increasing design and manufacturing costs and complexity. For example, in semiconductor processes, manufacturing devices or integrated products of multiple devices typically requires multiple photomasks. Modifying the design of multiple photomasks may be necessary to complete the manufacturing of the relevant directional couplers. Furthermore, changes in the geometry of the directional coupler may necessitate a rearrangement of other devices or structures.

[0045] Photonic integrated circuits can be used in photonic computing and optical interconnect technologies. In some existing photonic computing applications, modulators are required to perform multiplication calculations. Modulators require external electrical signals to operate, and large-scale integration requires numerous electrical interfaces. Modulators may also suffer from problems such as low modulation accuracy and high noise. In addition, when implementing photonic computing, some photonic device interconnect networks have high requirements for the coherence of light throughout the network.

[0046] In one exemplary embodiment, a photonic integrated circuit and a method for manufacturing the same are provided.

[0047] Figure 1A A top view of a directional coupler in a photonic integrated circuit is shown. The directional coupler has a first arm 105 and a second arm 106. The directional coupler includes an optical coupling region where light energy can be redistributed. Within the coupling region, the first arm 105 includes a first waveguide coupling section 105a, and the second arm 106 includes a second waveguide coupling section 106a. Exemplarily, the first arm 105 also includes a first optical input section 105p and a first optical output section 105q connected to the first waveguide coupling section 105a, and the second arm 106 includes a second optical input section 106p and a second optical output section 106q connected to the second waveguide coupling section 106a. Each part of the first and second arms has a waveguide body layer. Figure 1B It shows Figure 1AThe image shows a cross-sectional view of a photonic integrated circuit at a specific location (dashed line) within the waveguide coupling region. It illustrates the waveguide body layer 107 and metal-semiconductor compound layer 109 of the first arm 105, and the waveguide body layer 108 of the second arm. The metal-semiconductor compound layer 109 is disposed above the waveguide body layer 107, and its distribution area can be configured as needed. In some embodiments, the waveguide body layers of the first optical input section 105p and the first optical output section 105q may not have a metal-semiconductor compound layer. The photonic integrated circuit can be fabricated based on a semiconductor layer on an insulating layer, such as a silicon substrate on an insulator. Figure 1B The diagram shows a back substrate 101 and an insulating layer 102 in a silicon substrate on an insulator, wherein the waveguide body layer can be formed based on etching and other process steps.

[0048] A photonic integrated circuit may include multiple directional couplers, such as a first directional coupler and a second directional coupler. The first and second directional couplers each have a first and a second beam splitting ratio, respectively, and the beam splitting ratios of the first and second directional couplers are different. The beam splitting ratio of the directional coupler is the ratio of the light intensity emitted by the first arm to the light intensity emitted by the second arm. As shown in Figure 1B, exemplarily, the light intensity of the first and second arms of the first directional coupler is different. Furthermore, a suitable metal-semiconductor compound layer may be provided on the second directional coupler as needed, so that the light intensity of the first and second arms of the second directional coupler is different.

[0049] like Figure 2A As shown, specifically, photonic integrated circuits can be fabricated based on a semiconductor layer on an insulator, such as silicon-on-insulator (SOI) or silicon-germanium-on-insulator (S-SiGeOI). Alternatively, other substrates can be provided for fabricating photonic integrated circuits. Substrate materials can be silicon, germanium, silicon carbide, gallium arsenide, gallium phosphide, and can be compound semiconductors, alloy semiconductors, or combinations of the above materials. The substrate can be a wafer, such as an SOI wafer. Taking a silicon-on-insulator (SOI) substrate as an example, it includes providing an SOI substrate, which comprises a back substrate, an insulating layer, and a top silicon layer. Exemplarily, the insulating layer is a buried oxide layer. Figure 2A The diagram shows a back substrate 101, an insulating layer 102, and a top silicon layer 103. The back substrate can be a wafer, such as a silicon wafer.

[0050] The steps of forming a photonic integrated circuit may include forming a photonic device layer based on the SOI substrate. This photonic device layer may include various types of photonic devices, such as waveguides, grating couplers, optical modulators, directional couplers, multi-mode inferometers (MMIs), photodetectors, and optical beam splitters. Photonic devices can be formed through steps such as photolithography, etching, and patterning based on the top silicon layer, as well as deposition and doping steps. During the formation of photonic devices, different types of semiconductor materials and metallic materials may be deposited.

[0051] For example, the photonic integrated circuit includes a directional coupler, which includes an optical coupling region. The energy of light can be redistributed within the optical coupling region. The directional coupler includes a first arm and a second arm. Within the coupling region of the directional coupler, the first arm and the second arm respectively include a first waveguide coupling section and a second waveguide coupling section. The first waveguide coupling section and the second waveguide coupling section can be connected to their respective optical input section and optical output section. For example, the optical input section and the optical output section can be retained according to the input and output requirements of light.

[0052] Figure 2B The etching is shown in the figure. Figure 1A In the top silicon layer 103, an initial waveguide layer 104 is formed for the first and second arms of a directional coupler. When multiple directional couplers exist in a photonic integrated circuit, an initial waveguide layer 104 for the first and second arms of each can be formed. In some embodiments, the initial waveguide layer 104 can be asymmetric, or it can be a curved waveguide layer.

[0053] Figure 2C The image shows a metal-semiconductor compound layer 109 formed in one arm 105 of a directional coupler. Figure 2C The diagram also shows a directional coupler with a first arm 105 having a waveguide body layer 107 and a second arm 106 having a waveguide body layer 108. The waveguide body layer 107 of the first arm 105 and the waveguide body layer 108 of the second arm can be directly formed from the initial waveguide layer or formed based on the initial waveguide layer. Figure 2CIn this configuration, a metal-semiconductor compound layer 109 is located above the waveguide host layer 107 of the first arm 105. The metal-semiconductor compound layer can be formed by reacting a metal material with a semiconductor material. As an example, the metal-semiconductor compound layer can be formed by heat-treating a semiconductor material layer and a metal material layer. The metal material layer can include at least one of nickel, cobalt, titanium, tungsten, tantalum, platinum, palladium, and erbium, and the semiconductor material layer can include at least one of silicon or germanium. Exemplarily, the formed metal-semiconductor compound layer can be, for example, a metal silicide layer. The formation of the metal-semiconductor compound layer can be achieved using conventional CMOS processes, enabling large-scale, low-cost photonic integration. The distribution area of ​​the metal-semiconductor compound material is not limited to the waveguide host layer within the waveguide coupling section of the first / second arm, but can extend to other regions. It should be noted that the metal-semiconductor compound material in some other regions may contribute little or no to the beam splitting.

[0054] In some embodiments, a metal compound semiconductor layer can be formed by reacting a metallic material with a semiconductor material in an initial waveguide layer. Exemplarily, this can be achieved through processes such as etching. Figure 2B The initial waveguide layer 104 shown is shown. Figure 2B The initial waveguide layer 104 may contain a semiconductor material such as silicon; for example, a nickel material layer may be formed on the initial waveguide layer 104. Figure 2B (Not shown in the image), and then heat treatment is performed to react a portion of the silicon material in the initial waveguide layer with the nickel material, thereby forming the metal-semiconductor compound layer 109 as shown in Figure 2C. The nickel material can be made to react substantially completely by controlling the reaction conditions; of course, residual nickel metal material may remain around the metal-semiconductor compound layer. The unreacted initial waveguide layer constitutes the waveguide body layer 107, and the first arm 105 of the directional coupler includes the waveguide body layer 107 and the metal-semiconductor compound layer 109.

[0055] like Figure 3 As shown, in some embodiments, the initial waveguide layer can be directly used as the waveguide host layer 107 of the first arm 105 and the waveguide host layer 108 of the second arm 106. Then, a semiconductor material and a metal material are formed on the waveguide host layer, and a metal-semiconductor compound layer is formed through the reaction of the semiconductor material and the metal material. For example, the waveguide host layer 107 of the first arm is a silicon waveguide host layer, with a germanium material formed on top of it. Then, a metal material layer is formed on top of the germanium material layer, and a metal germanide layer is formed through the reaction of the germanium material and the metal layer, constituting the metal-semiconductor compound layer 109.

[0056] In some embodiments, a metal-semiconductor compound layer may also be disposed on the sidewalls of the waveguide body layer. Optionally, the metal-semiconductor compound layer may also be disposed in the region surrounding the waveguide body layer, or the metal-semiconductor compound layer may extend from above and / or the sidewalls of the waveguide body layer to the region surrounding the waveguide body layer. Figure 4 For example, the metal-semiconductor compound layer 109 may be disposed above and on the sidewall of the waveguide body layer 107 of the first waveguide coupling segment of the first arm 105, and the metal-semiconductor compound layer 109 may also extend to the surrounding area of ​​the waveguide body layer 107.

[0057] like Figure 5 Both the first arm 105 and the second arm 106 can be provided with metal-semiconductor compound materials, and their distribution areas can be set as needed. Metal-semiconductor compound layers are respectively provided in the first arm 105 and the second arm 106. For example, the metal-semiconductor compound material in the first arm 105 has a distribution length L0 and a width W0, while the distribution of the metal-semiconductor compound material in the second arm 106 can be different from that in the first arm 105. For example, as shown... Figure 5 The width of the metal-semiconductor compound layer in the second arm 106 can be different from the width of the second arm waveguide body layer (or the initial waveguide layer), and the first arm can also be similarly configured. In some embodiments, a metal-semiconductor compound material can be formed in at least one of the first and second arms as needed. The metal-semiconductor compound material substantially affects the propagation of light waves in the corresponding first and second arms, thereby changing the beam splitting characteristics of the directional coupler. That is, the distribution of the metal-semiconductor compound material contributes to the beam splitting ratio of the directional coupler.

[0058] In various embodiments of the present invention, the beam splitting ratio of the directional coupler is defined as the ratio of the light intensity emitted by the first arm to the light intensity emitted by the second arm of the directional coupler, so that the beam splitting ratio can be set between 100%:0% and 0%:100%, for example 30%:70%, 40%:60%, 80%:20%. The metal-semiconductor compound layer constitutes the beam splitting configuration layer, thereby allowing the beam splitting configuration layer to be adjusted to achieve the desired beam splitting ratio.

[0059] In some embodiments, a photonic integrated circuit may include multiple directional couplers, each of which includes a first arm and a second arm. For example, the number of directional couplers is four, namely, directional couplers 1 to 4. Each of these directional couplers includes a metal-semiconductor compound material to achieve a desired beam splitting ratio. Of course, in addition to the aforementioned multiple directional couplers (directional couplers 1 to 4), the photonic integrated circuit may also include other directional couplers, such as directional couplers 5 and 6, whose beam splitting ratio does not require adjustment using the metal-semiconductor compound material. For example, directional couplers 1 to 4 may form part of a photonic computing unit or a photonic computing chip. Depending on functional requirements, directional couplers 1 to 6 may also be part of a photonic computing unit or a photonic computing chip. When it is said that a photonic integrated circuit "includes" multiple directional couplers, the word "includes" means that in addition to the aforementioned "multiple directional couplers," the photonic integrated circuit may also have other directional couplers.

[0060] In one exemplary embodiment, the plurality of directional couplers includes a first directional coupler and a second directional coupler, wherein the first directional coupler and the second directional coupler have different beam splitting ratios. Not limited thereto, further exemplary embodiments may also include a third directional coupler and a fourth directional coupler. The coupling segments of the waveguide arms of each of the first to fourth directional couplers may have the same length, and the beam splitting ratio of each directional coupler can be set by providing a suitable metal-semiconductor compound layer.

[0061] According to various embodiments of the present invention, when it is necessary to change the beam splitting ratio of the directional coupler, this can be achieved by designing a suitable region of the metal-semiconductor compound layer, without changing the initial waveguide layer (or waveguide body layer) of the directional coupler and the associated photonic devices, or without requiring extensive redesign, nor excessive redesign or layout of the entire photonic integrated circuit. For example, the present invention can be applied to directional couplers in photonic integrated circuits with different required beam splitting ratios. For example, the formation region of the corresponding metal-semiconductor compound layer can be set using a mask, thereby changing the beam splitting ratio of the directional coupler simply by adjusting the corresponding mask.

[0062] In some implementations, photonic integrated circuits can be used in photonic computing. The photonic computing unit may include the directional couplers listed in the embodiments herein. For example, in a weighting module (multiplication module) of photonic computing, the directional coupler is designed to achieve a preset splitting ratio, such that one weighting module corresponds to one preset value, and can be used to perform a multiplication operation between a value corresponding to one optical signal input to the optical input section of the weighting module and a preset value corresponding to one weighting module. For example, the photonic computing unit may include multiple weighting modules to represent different preset values, and appropriate splitting ratios can be created for each directional coupler in the multiple weighting modules to achieve the preset values.

[0063] In an exemplary embodiment, a photonic computing system is proposed, comprising: a photonic computing unit configured to receive a first plurality of optical signals, the first plurality of optical signals corresponding to a first group of multiple numerical values; the photonic computing unit includes a plurality of weighting modules, the plurality of weighting modules corresponding to a plurality of preset numerical values, each weighting module corresponding to a preset numerical value. Each weighting module includes: an optical input unit (optical signal input unit) configured to receive one optical signal from the first plurality of optical signals; and at least one directional coupler, each directional coupler being designed to achieve a preset splitting ratio, and / or at least one micro-mixing element (MMI), each MMI being designed to achieve a preset splitting ratio, such that the weighting module corresponds to one preset numerical value and can be used to perform a multiplication operation between one numerical value corresponding to one optical signal input by the optical input unit and one preset numerical value corresponding to one weighting module. The weighting module may include a directional coupler and an MMI, and the structures of the directional coupler and MMI can be designed to achieve the preset splitting ratio, such that they can be associated with the preset numerical value corresponding to the weighting module. The directional coupler and MMI are passive devices, requiring no external electrical signal to adjust the splitting ratio. Their manufacturing process is also simple. Their splitting ratio is designed to be related to a preset value corresponding to the weighting module, making them suitable for weighting modules with preset fixed values. Using these splitting devices to achieve passive beam splitting offers advantages in photonic computing, such as stable beam splitting, anti-interference, low nonlinearity, and no static power consumption. When light is input to the weighting module, the passive devices such as the directional coupler and MMI are used as splitting devices to represent the corresponding weighting module values. These splitting devices achieve passive beam splitting, eliminating the need for electrical modulation for beam splitting calculations and avoiding excessive electrical interfaces. In some embodiments, the photonic computing system may also include an optical encoding unit and an optical replication module.

[0064] Figure 6AA photonic computing system 600 is shown, comprising an optical encoding unit 610, a replication unit 620, and a photonic computing unit 630. Exemplarily, the optical encoding unit may include multiple modulators, which can encode light according to numerical values. Figure 6B Modulators 601a and 601b in the optical encoding unit are shown. The optical encoding unit is optically connected to a replication unit, which may include one or more replication modules. Exemplary replication modules include beam splitters that can split the light input to the replication module into two or more beams, i.e., perform a "replication" operation to generate optical copies. These optical copies can serve as a first plurality of optical signals. Figure 6B The diagram illustrates replication modules 602a and 602b within a replication unit. The photonic computing system also includes a photonic computing unit configured to receive a first plurality of optical signals, each corresponding to a first set of multiple numerical values. The photonic computing unit includes multiple weighting modules, exemplarily such as... Figure 6B The weighting modules 603a, 603b, 603c, 603d, 603e, and 603f in the diagram represent multiple preset values, with each weighting module corresponding to one preset value. Each weighting module includes: an optical input unit configured to receive one optical signal from a plurality of optical signals; at least one directional coupler; and / or at least one MMI (Mechanical Management Interface). Figure 6A and Figure 6B (Not shown), the directional coupler and the MMI are designed to achieve a preset splitting ratio so that the weighting module corresponds to a preset value and is used to perform a multiplication operation between a value corresponding to one optical signal and a preset value corresponding to one weighting module.

[0065] The output signal of the replication module does not necessarily have the same amplitude as the input signal. For example, if an optical beam splitter (optical power splitter) is used to uniformly separate the input signal power between the two output signals, then each of the two output signals will have power equal to or less than 50% of the input signal power.

[0066] In some cases, the replication module or optical encoding unit can be omitted, and the optical signal carrying the numerical value can be directly input into the photonic computing unit through some optical ports; in addition, the optical signal output by the optical encoding unit can also be directly input into the photonic computing unit.

[0067] An optical encoding unit can encode light so that the optical signal represents a numerical value. The encoded optical signal can be input into a photonic computing unit as a data source for computation. For example, the optical encoding unit can encode n numerical values ​​to generate n corresponding optical signals. Exemplarily, the photonic computing unit includes multiple weighting modules, and the n optical signals can be input into their respective n weighting modules for computation. The optical encoding unit may include, for example, multiple modulators that modulate the light based on the electrical signals representing the numerical values, thereby encoding and generating the initial optical signal. In a photonic computing system, in addition to the aforementioned photonic computing unit, other photonic computing units may be included, such as a second photonic computing unit, a third photonic computing unit, etc. The computation modules and functions implemented in different photonic computing units may differ.

[0068] In addition to the aforementioned weighting modules (which correspond one-to-one with multiple preset values), the photonic computing unit may also include other structures or photonic computing modules capable of performing photonic computing, such as a photonic computing module that uses an MZI (Mach-Zehnder interferometer) for calculations.

[0069] Figure 7 A weighting module 200 is shown, having two input sections 201a and 201b and two output sections 202a and 202b. This example illustrates a schematic diagram using a directional coupler as a weighting module. The directional coupler includes two input sections and two output sections. In this embodiment, any one input section of the directional coupler can serve as an input section of the weighting module, or communicate with an input section of the weighting module, to receive optical signals. One input section of the weighting module, such as the first input section 201a, receives an optical signal, which is assigned a numerical value, such as x. The optical signal can have different intensities, thus corresponding to different numerical values. The weighting module itself is designed to correspond to a preset value (preset weight), such as y. Specifically, the directional coupler is designed to achieve a preset splitting ratio. In some embodiments, the weighting module only requires one input section; therefore, the weighting module can have only one input section, retaining only one input section of the directional coupler. In some embodiments, the weighting module can also have only one output section.

[0070] The directional coupler in the weighting module can be any of the directional couplers described in the embodiments of this paper. A metal compound semiconductor material is incorporated into the directional coupler to adjust the beam splitting ratio as needed. Photonic computing typically involves multiple different values. The directional couplers in these embodiments are suitable for different beam splitting ratios, thereby creating photonic computing structures corresponding to different values.

[0071] For example, the corresponding x and y multiplication result can be obtained by receiving the output of at least one output unit of a directional coupler. In this example, one weighting module can be implemented using one directional coupler. This weighting module is configured to perform a multiplication operation between a value represented by one input optical signal and a value represented by one weighting module. The directional coupler is a passive optical element, eliminating the need for electrical signal adjustment of light distribution like in modulated MZI, and its manufacturing process is simple. Its splitting ratio is designed to be related to a preset value corresponding to the weighting module, making it suitable for weighting modules with fixed values. For example, a computing unit (photonic computing unit) or photonic processor that requires multiplication calculations can include multiple weighting units. Depending on the needs, a photonic computing system can include photonic computing units or photonic processors to implement functions such as multiplication calculations.

[0072] In some embodiments, the coupling region of the directional coupler is designed to achieve a preset splitting ratio. Different desired splitting ratios are achieved by designing the length of the coupling region structure, waveguide width, coupling gap width, materials, etc. (the directional coupler can be set to correspond to the desired preset value), thereby configuring the weighting module to correspond to a specific value. Based on the preset value corresponding to a weighting module, each directional coupler in the weighting module is configured during the manufacturing stage to have a splitting ratio associated with the weighting module.

[0073] exist Figure 7 In the embodiments shown, in some embodiments, the preset value of the weighting module can be a preset value of 0 to 1, and is defined as the ratio of the output light of one of the two output sections of the directional coupler (e.g., the output section located at the top in Figure 7) to the input light of one input section. The splitting ratio of the directional coupler is designed according to this rule. When the preset value is 1, the power ratio of the output light to the input light is 1; when the preset value is 0.3, the power ratio of the output light to the input light is 0.3. If the optical signals of the two output sections of the directional coupler are converted into electrical signals respectively, and differential processing is performed before outputting as electrical signals, negative preset values ​​can also be defined. Examples of differential processing can be found in Chinese Patent Publication CN113159305A, paragraphs 595 to 621, or US Patent Publication US20200250532A1, paragraphs 566 to 635.

[0074] When multiple weight modules exist, the preset values ​​of the weight modules are defined according to the same rules, and the corresponding beam splitting ratios of the directional couplers are designed. The range of the preset values ​​(weights) represented by the weight modules is not limited to 0 to 1. For example, a photonic computing unit includes four weight modules, and the preset values ​​(preset weights) they need to represent are b1=3, b2=4, b3=7, and b4=8, respectively. Each weight module can be implemented by one directional coupler. According to the rules, the ratio of the optical power of one output section to the optical power of the input section of the directional coupler can be manufactured as 0.3, 0.4, 0.7, and 0.8, respectively, which correspond to beam splitting ratios of 30%:70%, 40%:60%, 70%:30%, and 80%:20%. The output light information is obtained from the corresponding output section of the directional coupler, and according to the correspondence between the output light and the preset values, the result of the multiplication calculation can be obtained according to the corresponding rules. Thus, the weight module can realize the calculation of the value represented by the input light and the preset values. Although the weighting modules are designed according to preset values ​​and represent multiplication of these preset values, for multiple weighting modules, they can also represent the calculation of the input light multiplied by the product of each corresponding weighting module and then multiplied by the same scaling factor. For example, if we want to use weighting modules to represent b1'=30, b2'=40, b3'=70, b4'=80, we can still complete the multiplication calculation using weighting modules manufactured with preset values ​​b1=3, b2=4, b3=7, b4=8. In this case, we only need to redefine the correspondence between the preset values ​​and the splitting ratio in the mathematical rules. This does not require changing the structure of the weighting modules, and the multiplication calculation implemented by the weighting modules is equivalent. Therefore, multiplication operations with preset values ​​implemented by weighting modules can be considered to include the above-mentioned equivalent multiplication operations.

[0075] In some implementations, the first plurality of optical signals can be used to represent one or more elements of a vector. The plurality of weight modules can be used to correspond to multiple elements in a matrix, thereby performing at least a portion of the matrix-vector multiplication operation; of course, the complete vector-matrix operation can also be performed in the manner described above. Since the first plurality of optical signals and the weight modules can all represent elements in a vector or matrix, these operations represent a set of general linear operations capable of performing various computations, including but not limited to: vector-vector dot product, vector-vector element-wise multiplication, vector-scalar element-wise multiplication, or matrix-matrix element-wise multiplication. Some examples described herein illustrate techniques and configurations for vector-matrix multiplication, but the corresponding techniques and configurations can be used for any of these types of computations.

[0076] In some implementations, the corresponding optical signal parameters can be obtained by detecting the light output of the directional coupler, thus obtaining the multiplication result. For example, the weighting module may include a photoelectric conversion element, which may include, for example, a photodetector, such as a photodiode, phototransistor, or photoresistor, to convert the optical signal into an electrical signal, representing the corresponding multiplication result. For example, the optical signals from the first and second outputs of the directional coupler are first converted into current signals, respectively. For example, the electrical signal can be used as the output of the weighting module, for example, by converting the output current signal through a photodiode.

[0077] In some embodiments, the optical signals from the first and second outputs of the directional coupler are first converted into current signals, and then the current signals are differentially processed. This differential processing can employ, for example, a differential operational amplifier, with the differentially processed electrical signal representing the result of the multiplication calculation. For example, the weighting module includes a first photoelectric conversion element and a second photoelectric conversion element. The first and second outputs of the directional coupler are respectively coupled to the first and second photoelectric conversion elements. The outputs of the first and second photoelectric conversion elements serve as the two outputs of the weighting module, meaning the weighting module can have an electrical signal output. The first and second outputs of the weighting module can be a first electrical signal output and a second electrical signal output. In some embodiments, the first and second outputs of the directional coupler can serve as the two outputs of the weighting module, meaning the weighting module outputs an optical signal, and the two outputs of the weighting module are respectively connected to the photoelectric conversion elements, converted into electrical signals, and then differentially processed.

[0078] Figure 8 An example of a photonic computing system is shown. In this embodiment, the optical signal to be computed can be transmitted through multiple input ports (such as...). Figure 8 (as shown), or in other ways, multiple weight modules can be input via an optical transmission channel, which can be, for example, an optical waveguide, and can constitute an optical waveguide unit.

[0079] Reference Figure 8 Examples of the optoelectronic computing system 1800 include a set of optical ports 1802A, 1802B, etc., that provide optical signals. For example, in some embodiments, optical port 1802A may include an optical input coupler that provides an optical signal coupled to optical path 1803. In other embodiments, optical port 1802A may receive an optical signal coupled to optical path 1803. Some embodiments may include a combination of a port that couples an optical signal to the system 1800 and a source that generates the optical signal within the system 1800. The optical signal may include any light wave (e.g., an electromagnetic wave whose spectrum includes wavelengths in the range of approximately 100 nm to approximately 1 mm) that has been or is being modulated using any of various forms of modulation to utilize information. Optical path 1803 may be defined, for example, based on the guidance mode of an optical waveguide (e.g., a waveguide embedded in a photonic integrated circuit (PIC) or an optical fiber), or based on a predetermined free-space path between optical port 1802A and another module of system 1800.

[0080] In some embodiments, the optoelectronic computing system 1800 is configured to perform computations on an array of input values ​​encoded on corresponding optical signals provided via optical ports 1802A, 1802B, etc. For example, for various machine learning applications based on neural networks, the computation can implement vector-matrix multiplication (or vector-by-matrix multiplication), where an input vector is multiplied by a matrix to produce an output vector as a result. The optical signal can represent elements of the vector, possibly including only a selected subset of the vector's elements. For example, for some neural network models, the size of the matrix used in the computation can be larger than the size of the matrix that can be loaded into the hardware system (e.g., the engine or coprocessor of a larger system) performing the vector-matrix multiplication portion of the computation. Therefore, performing a portion of the computation may involve dividing the matrix and vector into smaller segments that can be provided separately to the hardware system.

[0081] Figure 8 The modules shown can be part of a larger system that performs vector-matrix multiplication on relatively large matrices (or submatrices), such as 64×64 element matrices. However, for illustrative purposes, the modules will be described in the context of an example computation that performs vector-matrix multiplication using 2×2 element matrices. The modules referenced in this example will include two replication modules 1804A and 1804B, four weighting modules 1806A, 1806B, 1806C, and 1806D, and two summing modules, with only one summing module 1808... Figure 8 The above is displayed. These modules will make the input vector... Multiply by matrix To generate the output vector For this vector matrix multiplication Output vector Each of the two elements can be represented by a different equation, as shown below.

[0082] y A =M A x A +M B x B

[0083] y B =M C x A +M D x B

[0084] These equations can be broken down into separate steps that can be performed in System 1800 using a set of basic operations: copying, multiplication, and summation. In these equations, each element of the input vector appears twice, so there are two copying operations. There are also four multiplication operations and two summation operations. For systems that implement vector-matrix multiplication using larger matrices, the number of operations performed will be greater, and the relative number of instances per operation will differ depending on the matrix shape (i.e., the number of columns differs from the number of rows).

[0085] In this example, the copy operation is performed using copy modules 1804A and 1804B. Input vector x A and x B The elements are represented by values ​​encoded on optical signals from optical ports 1802A and 1802B, respectively. Each of these values ​​is used in two equations, thus each value is copied to provide two copies to different corresponding weighting modules. For example, as described in more detail below, values ​​can be encoded in a specific time slot using light waves modulated with power from a set of multiple power levels, or light waves with a set of multiple duty cycles. The values ​​are copied by optical signals on which the encoded values ​​are replicated. The encoded elements have the representation element x. A The optical signal with the value is copied by the copy module 1804A and encoded with a representation element x. B The optical signal of value is copied by the copying module 1804B. Each copying module can be implemented, for example, using an optical power splitter, such as a waveguide optical splitter that couples the guided mode in the input waveguide to each of the two output waveguides on a Y-shaped splitter that progressively (e.g., adiabatically) separates the power, or the optical power splitter can be, for example, a free-space beam splitter that uses a dielectric interface or thin film with one or more layers to transmit and reflect the two output beams from the input beam, respectively.

[0086] In this application, it is said that the encoded representation element x is copied by means of the copy module 1804A. A When referring to the value of an optical signal, it means generating a representation element x based on the input signal. AThe multiple signal copies produced by the replication module 1804A do not necessarily have the same amplitude as the input signal. For example, if the replication module 1804A uniformly distributes the input signal power between the two output signals, each of the two output signals will have power equal to or less than 50% of the input signal power. The two output signals are copies of each other, and the amplitude of each output signal of the replication module 1804A differs from the amplitude of the input signal. Moreover, in some embodiments having a set of multiple replication modules for replicating a given optical signal or a subset of optical signals, each individual replication module does not necessarily uniformly distribute power among its produced copies, but the set of replication modules can be collectively configured to provide copies with substantially equal power to the input of a downstream module (e.g., a downstream weighting module).

[0087] In this embodiment, the multiplication operation is performed by four weighting modules 1806A, 1806B, 1806C, and 1806D. For each copy of an optical signal, one weighting module multiplies the copy of the optical signal by the matrix element value.

[0088] Encoded with vector elements x A Optical signals can be encoded using different forms of amplitude modulation. The amplitude of the optical signal can correspond to a specific instantaneous power level P of the physical light wave within a specific time slot. A Or it could correspond to a specific energy E of a physical light wave in a specific time slot. A (The power integrated over time produces the total energy). For example, the power of a laser source can be modulated to have a specific power level from a predetermined set of multiple power levels. In some embodiments, it may be useful to operate the electronics near an optimized operating point, thus using an optimized "on" power level instead of varying the power across many possible power levels, where the signal is modulated to be "on" and "off" (at zero power) for specific portions of a time slot. The time portion of the power at the "on" level corresponds to a specific energy level. Any of these specific values ​​of power or energy can be mapped to an element x. A A specific value (using a linear or nonlinear mapping relationship). After the signal is in the electrical domain, the actual integration over time that produces a specific total energy level can occur downstream of system 1800, as described in more detail below.

[0089] Furthermore, the term "amplitude" can refer to the amplitude of a signal, expressed as instantaneous or integrated power in a light wave, or equivalently, the "electromagnetic field amplitude" of the light wave. This is because the electromagnetic field amplitude and the signal amplitude have a well-defined relationship (e.g., generating instantaneous power by integrating the electromagnetic field strength (proportional to the square of the electromagnetic field amplitude) over the transverse dimension of a guided mode or free-space beam). This leads to a relationship between modulation values, because by a specific value... A modulator that modulates the amplitude of an electromagnetic field can also be considered as modulating the power-based signal amplitude by a corresponding value M (because optical power is proportional to the square of the electromagnetic field amplitude).

[0090] In this embodiment, the summation operation is performed by two summation modules, wherein summation module 1808 (e.g., ...) Figure 8 (As shown) is used to calculate the output vector element y B The summation is performed within the equation. The corresponding summation module (not shown) is used to calculate the elements of the output vector y. A The summation is performed in the equation. The summation module 1808 generates an electrical signal representing the sum of the results from the two weighting modules 1806C and 1806D. In this example, the electrical signal is the current i. sum The current i is in the form of a sum of power in the output optical signal generated by the weighting modules 1806C and 1806D, respectively. In some embodiments, this current i is generated... sum The summation operation is performed in the photoelectric domain, and in other embodiments in the electrical domain. Alternatively, some embodiments may use photoelectric domain summation for some summation modules and electrical domain summation for other summation modules.

[0091] In embodiments where summation is performed in the electrical domain, summation module 1808 may be implemented using: (1) two or more input conductors, each carrying an input current whose magnitude represents the result of one of the weighting modules, and (2) at least one output conductor carrying a current as the sum of the input currents. This occurs, for example, if the conductors are wires that meet at a node. For example (not bound by theory), this relationship can be understood based on Kirchhoff's current law, which states that the current flowing into a node is equal to the current flowing out of the node. In these embodiments, signals 1810A and 1810B provided to summation module 1808 are input currents that may be generated by a photodetector, which is part of a weighting module that generates a corresponding photocurrent whose magnitude is proportional to the power in the received optical signal. Summation module 1808 then provides an output current i sumThe quantitative value of the sum can then be expressed using either the instantaneous value or the integrated value of the output current.

[0092] In embodiments where summation is performed in the optoelectronic domain, the summation module 1808 can be implemented using a photodetector (e.g., a photodiode) that receives optical signals generated by different corresponding weighting modules. For these embodiments, signals 1810A and 1810B provided to the summation module 1808 are input optical signals, each comprising a light wave whose power represents the result of one of the weighting modules. The output current i in this embodiment... sum The photocurrent is generated by a photodetector. Since the wavelengths of the light waves are different (e.g., different enough that no significant constructive or destructive interference occurs between them), the photocurrent will be proportional to the sum of the powers of the received optical signals. The photocurrent is also essentially equal to the sum of the individual currents, which will result in individual detected optical powers detected by separate equivalent photodetectors. The wavelengths of the light waves are different, but close enough that the photodetectors have essentially the same response (e.g., wavelengths within the essentially flat detection bandwidth of the photodetector). As described above, using a summation in the electrical domain using current summation allows for a simpler system architecture by avoiding the need for multiple wavelengths.

[0093] In some implementations, a weighting module may include multiple directional couplers. Examples include... Figure 9 The weighting module 400 includes two directional couplers and a beam splitter 401. The beam splitter 401 can perform 50%-50% beam splitting (e.g., a 3dB beam splitter), but the splitting ratio is not limited to this. As shown in the figure, the input of the 3dB beam splitter serves as the input of the weighting module 400, or the input of the weighting module 400 is optically connected to the input of the beam splitter to input a light signal representing a numerical value. The light signal is split into two equal paths by the 3dB beam splitter, which then enter the two directional couplers respectively. Based on the preset value corresponding to the weighting module, the splitting ratio of the first directional coupler and the splitting ratio of the second directional coupler are designed. The splitting ratio of the directional coupler is related to the preset value of the weighting module and can be physically modified by designing the length of the coupling region structure, waveguide width, coupling gap width, materials, etc. Figure 9In this design, one or more outputs of two directional couplers can be selected as corresponding outputs of the weighting module. Of course, a weighting module can also include other numbers of directional couplers, such as three or four, but is not limited to these. Furthermore, the connection method of the directional couplers can be changed to achieve suitable beam splitting; for example, the output of one directional coupler can be used as the input of another.

[0094] exist Figure 9 The paper also shows that for unwanted waveguide terminals, optical absorption structures can be used to absorb the residual light in the waveguide, such as dummy photodetectors 402, which convert unwanted optical signals into electrical signals, but the electrical signals no longer play any other role.

[0095] In some implementations, such as Figure 9 As shown, the weighting module also includes a photoelectric conversion element 405, such as a photodetector. Each optical output section of the two directional couplers is connected to the photoelectric conversion element. The weighting module includes four electrical output ports: a first electrical output port 406a and a second electrical output port 406b corresponding to the first directional coupler 403, and a first electrical output port 406c and a second electrical output port 406d corresponding to the second directional coupler 404. The currents of the two first electrical output ports are first summed, and the currents of the two second electrical output ports are summed. The summation results are then differentially processed, and the resulting electrical signal can represent the result of a multiplication operation between an input optical signal and the weighting module. Optionally, in some embodiments, among the multiple electrical output ports in the weighting module, the currents of multiple first ports are summed, and the currents of multiple corresponding second ports are summed. The resulting differential electrical signal represents the result of a multiplication operation between an input optical signal and the weighting module.

[0096] In some implementations, the weighting module has an optical signal output. The optical signal output of the weighting module can be further optically processed or photoelectrically converted. For example, its output is connected to an optical nonlinear module to process the output optical signal of the weighting module, thereby achieving other mathematical transformations, such as nonlinear transformations.

[0097] In some implementations, the weighting module may include a photoelectric conversion element for converting optical signals into electrical signals. An exemplary photoelectric conversion element includes a photodiode, in which case the weighting module has an electrical signal output.

[0098] In an exemplary embodiment, the weighting module can be used for matrix calculations, including, for example, vector-matrix multiplication, matrix-matrix multiplication, which can typically be decomposed into element-wise multiplication and summation operations. Exemplarily, in one embodiment, one or more elements of the matrix can be implemented using the weighting module, which can be implemented using the weighting modules described in various embodiments of the present invention. When performing matrix calculations, such as vector-matrix operations, it is necessary to multiply one element (one weight) of the matrix with another input value. A light signal is mapped to one input value, and one matrix element is mapped to one weighting module, thereby completing the multiplication calculation between the value corresponding to one element and one input value. For example, for an 8×8 matrix, multiplying it by an 8-bit vector requires each element in the vector to be repeated 8 times, performing multiplication with each of the 8 elements in the matrix. To achieve this, the optical signal (optical signal power) representing an element 'a' in the vector can be distributed, for example, equally divided into 8 parts. That is, an element in the vector can be represented by 8 optical signal copies. A replication module can be used to generate these optical signal copies. The replication module may include, for example, a splitter or a combination of splitters. Multiple replication modules can constitute a replication unit. Taking the complete calculation of the vector and matrix as an example, an 8×8 matrix requires 64 weight modules. The 8-bit elements of the vector require a total of 64 optical signals to represent (because each element in the vector needs to be repeated 8 times). During calculation, these 64 optical signals are input into 64 weight modules respectively for multiplication. Of course, in matrix multiplication operations, only weight modules with preset values ​​can be used to perform partial multiplication operations, and an appropriate number of these weight modules with preset values ​​can be configured according to requirements. Each weighting module may include different types of devices.

[0099] In some implementations, the photonic computing unit may include multiple replication modules, meaning the replication modules may be part of the photonic computing unit. In other implementations, the multiple replication modules may also be independent of the photonic computing unit, and may be part of a replication unit, meaning the replication unit includes multiple replication modules and is optically connected to the photonic computing unit to transmit optical copies to the photonic computing unit.

[0100] like Figure 10-11 The directional coupler described above can be replaced by an MMI (Multimode Interferometer Coupler). For example, a weighting module may include one or more MMIs, where the MMIs can be designed to have different light distribution ratios. This can be achieved, for instance, by designing the coupling region (interference region) of the MMI, where the MMI can be set to be asymmetric to achieve arbitrary light distribution ratios. Figure 10The diagram shows a weighting module 500 including an MMI. The input and output sections of the weighting module 500 can correspond to the input and output sections of the MMI. Therefore, the weighting module has one optical signal input section and two optical signal output sections. Figure 11 In the weighting module 500, there are a first MMI 503 and a second MMI 504, a beam splitter 501, and four photodetectors 505. The weighting module 500 has one optical input section and four electrical signal output sections. Inside the weighting module, the two optical output sections of each MMI are connected to the photodetectors. Finally, the electrical signal is output from the weighting module through the electrical signal output section.

[0101] In some implementations, a weighting module may include at least one of a directional coupler and a micro-interface (MMI), and the number of these can be one or more. The structures of the directional couplers and MMIs can be designed to achieve a preset splitting ratio, enabling them to be associated with a preset value corresponding to the weighting module. The directional couplers and MMIs are passive devices, requiring no external electrical signal to adjust the splitting ratio, and their manufacturing process is simple. Their splitting ratio is designed to be related to a preset value corresponding to the weighting module, making them suitable for weighting modules with preset fixed values. Using these splitting devices to achieve passive beam splitting offers advantages in photonic computing, such as stable beam splitting, anti-interference, low nonlinearity, and no static power consumption. When light is input into the weighting module, the values ​​of the corresponding weighting module are represented by passive devices such as directional couplers and MMIs. These splitting devices achieve passive beam splitting, eliminating the need for electrical modulation for beam splitting calculations and avoiding excessive electrical interfaces.

[0102] A weighting module may have one or more outputs, which may include electrical signal outputs or optical signal outputs as needed.

[0103] In some implementations, the output of the weighting module can output an optical signal; that is, the weighting module includes an optical signal output section. One or more outputs of a weighting module can be, for example, the output of one or more directional couplers, the output of an MMI (Multi-Instrument Machine), or a directional coupler. The output of the MMI can be coupled to the output of the weighting module via a waveguide or other optical element. By reading the optical signal information from one or more outputs of the weighting module and performing further processing, the result of the multiplication calculation can be obtained.

[0104] In some embodiments, the output of the weighting module can output an electrical signal; that is, the weighting module includes an electrical signal output section. For example, the weighting module may include a photoelectric conversion element that can detect the light from the optical output section of the directional coupler and the MMI and convert it into an electrical signal to represent the calculation result of the weighting module. The photoelectric conversion element may include, for example, a photodetector, such as a photodiode, phototransistor, or photoresistor, which converts the optical signal into an electrical signal to represent the corresponding multiplication calculation result. For example, the electrical signal is the output of the weighting module, such as a current signal output through a photodiode. A weighting module may have only one electrical signal output section or multiple electrical signal output sections. The electrical signal from the electrical signal output section of the weighting module can be processed to obtain the multiplication calculation result. For example, the light output from at least one optical output section of the directional coupler and the MMI is photoelectrically converted and output as a current signal, and signal processing is performed in the electronic domain to read the corresponding multiplication calculation result. For example, a differential operational amplifier may be used for signal processing in the electronic domain.

[0105] In some implementations, a summation module sums the currents of the first electrical signal outputs of each of the multiple weighting modules, then sums the currents of the second electrical signal outputs of each of the multiple weighting modules, and then performs differential processing. The result output after differential processing represents the sum of the products obtained by the multiple weighting modules. For example, each of the four weighting modules includes a first electrical signal output and a second electrical signal output. The summation module receives the electrical signals (e.g., photocurrent signals) from the first electrical signal outputs of the four weighting modules, sums them, and the result is current S1. Then, it sums the currents of the second electrical signal outputs, and the result is recorded as current S2. Then, it performs differential processing on currents S1 and S2 to obtain the sum of the results corresponding to the multiple weighting modules.

[0106] In one exemplary embodiment, a photonic computing chip is proposed. The photonic computing chip may include a photonic integrated circuit. The photonic devices in the photonic integrated circuit can be manufactured using the manufacturing methods of the relevant embodiments of the present invention, and may include, for example, the directional couplers in various embodiments. The photonic integrated circuit may include the photonic computing unit mentioned herein, and exemplaryly, may also include an optical encoding unit, a replication unit, and a summing module. In some embodiments, the photonic computing chip can be used in a photonic computing system. The photonic computing system may include the photonic computing chip, or it may include electronic chips. The electronic chips can implement storage, computing, or control functions, and may be, for example, memory chips, computing chips, control chips, etc.

[0107] Those skilled in the art should understand that the above-disclosed embodiments are merely implementations of the present invention and should not be construed as limiting the scope of the patent protection claimed in this invention. Equivalent variations made according to the embodiments of the present invention are still within the scope of the claims of the present invention.

Claims

1. A photonic integrated circuit, comprising: A first directional coupler includes a first arm and a second arm. The first arm and the second arm of the first directional coupler respectively include a first waveguide coupling section and a second waveguide coupling section. The light output intensities of the first arm and the second arm of the first directional coupler are different, and the ratio of the light output intensities of the first arm and the second arm is a first beam splitting ratio. At least one of the first arm and the second arm of the first directional coupler includes a metal semiconductor compound material. The distribution of the metal semiconductor compound material of the first directional coupler contributes to the allocation of the first beam splitting ratio. A second directional coupler includes a first arm and a second arm. The first arm and the second arm of the second directional coupler respectively include a first waveguide coupling section and a second waveguide coupling section. The light output intensities of the first arm and the second arm of the second directional coupler are different, and the ratio of the light output intensities of the first arm and the second arm is a second beam splitting ratio. At least one of the first arm and the second arm of the second directional coupler includes a metal semiconductor compound material. The distribution of the metal semiconductor compound material of the second directional coupler contributes to the allocation of the second beam splitting ratio. The first spectral ratio is different from the second spectral ratio.

2. The photonic integrated circuit as described in claim 1, wherein, The first arm of the first directional coupler includes a first metal-semiconductor compound layer, such that at least one of the first arm and the second arm of the first directional coupler includes a metal-semiconductor compound material; The first arm of the second directional coupler includes a second metal semiconductor compound layer, such that at least one of the first arm and the second arm of the second directional coupler includes a metal semiconductor compound material.

3. The photonic integrated circuit as claimed in claim 2, wherein the first directional coupler includes a waveguide body layer, and the first metal semiconductor compound layer is disposed above and / or on the sidewall of the waveguide body layer in the first waveguide coupling segment of the first directional coupler.

4. The photonic integrated circuit as described in claim 3, wherein, The first metal-semiconductor compound layer is obtained by heat treatment of metal and semiconductor materials.

5. The photonic integrated circuit as described in any one of claims 3 or 4, wherein, The photonic integrated circuit is formed based on a semiconductor-on-insulator layer.

6. The photonic integrated circuit as described in claim 5, wherein, The photonic integrated circuit is formed based on silicon-on-insulator.

7. The photonic integrated circuit as described in claim 5, wherein, The metal semiconductor compound material includes at least one semiconductor element selected from silicon or germanium.

8. The photonic integrated circuit as described in claim 7, wherein, The metal semiconductor compound material includes at least one metallic element selected from nickel, cobalt, titanium, tungsten, tantalum, platinum, palladium, and erbium.

9. A method for manufacturing a photonic integrated circuit, the photonic integrated circuit comprising a plurality of directional couplers, each of the plurality of directional couplers comprising a first arm and a second arm, wherein the beam splitting ratio of the directional coupler is the ratio of the light intensity emitted by the first arm to the light intensity emitted by the second arm, wherein... The first arm and the second arm respectively include a first waveguide coupling section and a second waveguide coupling section; The manufacturing method includes: The waveguide body layer of the first arm and the waveguide body layer of the second arm are formed in each of the plurality of directional couplers; A metal-semiconductor compound region is formed according to the spectral splitting ratio, such that at least one of the first arm and the second arm includes a metal-semiconductor compound material, the distribution of which contributes to the spectral splitting ratio distribution.

10. The method for manufacturing a photonic integrated circuit as described in claim 9, wherein the plurality of directional couplers includes a first directional coupler and a second directional coupler, and the first directional coupler and the second directional coupler have different beam splitting ratios.

11. The method for manufacturing a photonic integrated circuit as claimed in claim 10, wherein forming the metal semiconductor compound region includes forming a first metal semiconductor compound layer above and / or on the sidewall of the waveguide body layer in the first waveguide coupling segment of the first directional coupler, and forming a second metal semiconductor compound layer above and / or on the sidewall of the waveguide body layer in the first waveguide coupling segment of the second directional coupler.

12. The method for manufacturing a photonic integrated circuit as claimed in claim 10 or 11, wherein forming the metal-semiconductor compound region comprises: An initial waveguide layer is formed, wherein the initial waveguide layer comprises a semiconductor material; Forming a layer of metallic material; This causes a portion of the semiconductor material in the initial waveguide layer to react with the metal material in the metal material layer to form the metal-semiconductor compound region; The waveguide body layer includes an initial waveguide layer that does not react with the metal material layer.

13. The method for manufacturing a photonic integrated circuit as described in claim 10 or 11, wherein forming the metal-semiconductor compound region comprises: A semiconductor material layer is formed on the waveguide body layer; Forming a layer of metallic material; This causes the metal material layer to react with the semiconductor material layer, forming the metal-semiconductor compound region.

14. The method for manufacturing a photonic integrated circuit as described in claim 12, wherein the metal semiconductor compound material comprises at least one semiconductor element selected from silicon or germanium.

15. The method for manufacturing a photonic integrated circuit as described in claim 14, wherein the metal semiconductor compound material comprises at least one metallic element selected from nickel, cobalt, titanium, tungsten, tantalum, platinum, palladium, and erbium.

16. The method for manufacturing a photonic integrated circuit as described in claim 13, wherein the metal semiconductor compound material comprises at least one semiconductor element selected from silicon or germanium.

17. The method for manufacturing a photonic integrated circuit as described in claim 16, wherein the metal semiconductor compound material comprises at least one metallic element selected from nickel, cobalt, titanium, tungsten, tantalum, platinum, palladium, and erbium.

18. A photonic computing system comprising a photonic integrated circuit as described in any one of claims 1-8, wherein the photonic computing system comprises: A photonic computing unit is configured to receive a first plurality of optical signals, the first plurality of optical signals corresponding to a first group of multiple numerical values; The photonic computing unit includes multiple weight modules, each weight module corresponding to a multiple preset value. Each weight module corresponds to one preset value, and each weight module includes: An optical input unit, configured to receive one optical signal from the plurality of optical signals, and At least one directional coupler, wherein each directional coupler is designed to achieve a preset splitting ratio such that the weighting module corresponds to a preset value and can be used to perform a multiplication operation between a value corresponding to a light signal input from the light input unit and a preset value corresponding to a weighting module. The plurality of weight modules include a first weight module and a second weight module, wherein at least one directional coupler in the first weight module includes the first directional coupler, and at least one directional coupler in the second weight module includes the second directional coupler.

19. The photonic computing system of claim 18, wherein, According to the preset value corresponding to the weight module, the at least one directional coupler is configured during the manufacturing stage to have a spectral ratio associated with the preset value of the weight module.

20. The photonic computing system of claim 19, wherein the photonic computing unit comprises one or more summing modules, and for the outputs of two or more weighting modules, a corresponding summing module is configured to generate an electrical signal representing the sum of the multiplication results corresponding to the two or more weighting modules.

21. A photonic computing chip, characterized in that, Including the photonic integrated circuit as described in any one of claims 1-8.

22. The photonic computing chip as described in claim 21, wherein, The first directional coupler has a first splitting ratio associated with a first value, and the second directional coupler has a second splitting ratio associated with a second value. The first value and the second value are used for calculation.

23. The photonic computing chip as described in claim 21, characterized in that, Includes a photonic computing unit configured to receive a first plurality of optical signals, the first plurality of optical signals corresponding to a first group of multiple values; The photonic computing unit includes multiple weight modules, each weight module corresponding to a multiple preset value. Each weight module corresponds to one preset value, and each weight module includes: An optical input unit, configured to receive one optical signal from the plurality of optical signals, and At least one directional coupler, wherein each directional coupler is designed to achieve a preset splitting ratio such that the weighting module corresponds to a preset value and can be used to perform a multiplication operation between a value corresponding to a light signal input from the light input unit and a preset value corresponding to a weighting module. The plurality of weight modules include a first weight module and a second weight module, wherein at least one directional coupler in the first weight module includes the first directional coupler, and at least one directional coupler in the second weight module includes the second directional coupler.

24. The photonic computing chip as described in claim 23, wherein, According to the preset value corresponding to the weight module, the at least one directional coupler is configured during the manufacturing stage to have a spectral ratio associated with the preset value of the weight module.

25. The photonic computing chip as described in claim 23 or 24, characterized in that, The first plurality of optical signals are used to represent one or more elements in a vector, and the plurality of weight modules are used to represent multiple elements in a matrix.

Images