Programmable optical computing chip

Abstract

The invention relates to a programmable optical computing chip, comprising: an input layer comprising an on-chip light source module, a wavelength division multiplexer and an optical modulator group; the input end of the linear layer is connected with the output end of the input layer so as to perform matrix-vector multiplication operation on the plurality of input light splitting signals input by the input layer to obtain a plurality of output light signals; the output layer is connected with the output end of the linear layer, and the output layer detects the optical signal intensity of a plurality of output optical signals and determines a matrix-vector multiplication operation result according to the optical signal intensity of each output optical signal; and the control module is respectively connected with the input layer, the linear layer and the output layer. The method at least can realize large-scale parallel data operation, has polarization insensitivity, can process a plurality of signals in different polarization states at the same time, and improves data throughput and calculation efficiency.

Description

Technical Field

[0001] The present disclosure relates to the field of optical computing technology, and in particular to a programmable optical computing chip. Background Art

[0002] In recent years, integrated photonics has made significant breakthroughs in many fields, especially in data communication, sensing, computing and quantum information processing. By integrating optical components (such as light sources, modulators, waveguides, detectors, etc.) onto a single chip, integrated photonics has achieved higher system integration and lower power consumption. Among them, optical computing chips are hardware devices that use optical principles for information processing and computing, and have broad application prospects in high-performance matrix-vector multiplication calculations.

[0003] With the continuous expansion of large-scale data sets and cloud-based artificial intelligence services, the demand for specialized hardware is growing, especially in accelerating matrix-vector multiplication operations, where the demand for multiplication and addition is more prominent. Although a variety of customized silicon computing hardware (such as FPGA, ASIC and GPU) has been developed to improve computing throughput and efficiency, these hardware still rely on traditional electronic components and are subject to physical limitations in speed and energy, such as Joule heating, electromagnetic crosstalk and capacitance effects. Integrated photonic technology takes advantage of the modularity and scalability of integrated circuits, while enabling large-scale parallel data transmission through wavelength division multiplexing and extremely high data modulation speeds through the high bandwidth of on-chip optical modulators and photodetectors. These advantages have given optical networks an important position in data center interconnection technology and are expected to challenge the dominance of traditional electronic technology in the field of information processing. At the same time, the parallel processing advantages of optical computing chips enable them to perform well in image processing, pattern recognition and computer vision tasks.

[0004] Incoherent optical accelerators are based on the intensity addition of optical signals. Although they have advantages in noise resistance and operational stability, they still face challenges in scalability. Existing optical computing technology still has shortcomings in handling large-scale parallel computing and improving system efficiency, which limits its application in high-performance computing and complex task processing. For example, existing optical computing chips usually rely on single-wavelength optical signals to implement computing functions or use wavelength division multiplexing technology to increase the number of channels for parallel data processing, but their parallel processing capabilities are still limited. In particular, when complex computing functions need to be implemented, larger chips (such as interferometer units with a larger footprint) are often required to accommodate more optical components. This not only increases the physical size of the system, but also leads to higher power consumption. For another example, existing optical computing chips are very sensitive to the polarization state of the input light, and optical signals with different polarization states may lead to inconsistent calculation results or reduced efficiency. Therefore, in practical applications, the polarization state of the input light must be strictly controlled and managed, which increases the complexity and cost of the system. Summary of the invention

[0005] Based on this, it is necessary to provide a programmable optical computing chip to address the above technical issues, which can at least realize large-scale parallel data operations and has polarization insensitivity, and can simultaneously process multiple signals of different polarization states to improve data throughput and computing efficiency. By controlling the input layer, output layer, and linear layer through the control module, a large number of parallel matrix-vector multiplication operations can be performed at the optical data rate, which can achieve efficient and stable computing performance.

[0006] At the same time, the use of optical computing chips can avoid inconsistent calculation results or reduced efficiency due to unstable polarization states of optical signals, and optical computing chips have extremely high data modulation speeds, which are only limited by the bandwidth of on-chip optical modulators and photodetectors.

[0007] In order to achieve the above-mentioned purpose and other purposes, in a first aspect, the present disclosure provides a programmable optical computing chip, comprising: an input layer, the input layer comprising an on-chip light source module, a wavelength division multiplexer and an optical modulator group, the on-chip light source module inputs a multi-wavelength light source into the wavelength division multiplexer to decompose the input light source into multiple split light signals, the multiple split light signals are input into the optical modulator group, the optical modulator group is used to perform intensity encoding on each split light signal to obtain multiple input split light signals; a linear layer, the input end of the linear layer is connected to the output end of the input layer to perform matrix-vector multiplication operation on the multiple input split light signals input into the input layer to obtain multiple output light signals; an output layer, connected to the output end of the linear layer, the output layer detects the light signal intensity of the multiple output light signals, and determines the matrix-vector multiplication operation result according to the light signal intensity of each output light signal; a control module, the control module is respectively connected to the input layer, the linear layer and the output layer, the control module is used to control the split light encoding mode of the input layer and the matrix-vector multiplication operation mode of the linear layer, and is used to receive the matrix-vector multiplication operation result output by the output layer.

[0008] In the programmable optical computing chip in the above-mentioned embodiment, the input layer adopts an on-chip light source module and a wavelength division multiplexer, so that the input layer can be applied to different polarized light sources and can realize parallel computing based on different polarized lights. The input layer also adopts an optical modulation group to perform intensity encoding on optical signals of different wavelengths. The linear layer realizes the matrix-vector multiplication operation of the input split light signal through the control of the control unit. The output layer outputs the vector result after the matrix multiplication operation by detecting the intensity of the optical signal of different output ports. Therefore, the embodiment of the present disclosure can at least realize large-scale parallel data operation, and has polarization insensitivity, and can process multiple signals of different polarization states at the same time, thereby improving data throughput and computing efficiency. The control module controls the input layer, the output layer and the linear layer to realize a large number of parallel matrix-vector multiplication operations at the optical data rate, which can achieve efficient and stable computing performance.

[0009] At the same time, the use of optical computing chips can avoid inconsistent calculation results or reduced efficiency due to unstable polarization states of optical signals, and optical computing chips have extremely high data modulation speeds, which are only limited by the bandwidth of the on-chip light source and photodetector.

[0010] In one embodiment, the on-chip light source module includes: a substrate; an insulating layer located on the substrate; and a waveguide layer located in the insulating layer.

[0011] In the programmable optical computing chip in the above embodiment, an on-chip light source is used to implement a polarization-insensitive waveguide design.

[0012] In one embodiment, the thickness of the waveguide layer is 0.6 μm-1.0 μm, the thickness of the insulating layer is 6.6 μm-7 μm, and the width of the waveguide layer is 0.6 μm-1.0 μm.

[0013] In the programmable optical computing chip in the above embodiment, by limiting the size of the on-chip light source, the refractive indexes of the TE mode and the TM mode of the on-chip light source are made similar, thereby further realizing the polarization-insensitive waveguide design of the optical computing chip.

[0014] In one embodiment, the wavelength division multiplexer includes a microcavity unit and four splitting units, the microcavity unit is formed between the on-chip light source module and the wavelength division multiplexer, the microcavity unit includes an input end and four output ends, the input end is used to connect to the on-chip light source, each output end is connected to a splitting unit, multiple first splitting signals are transmitted to multiple splitting units, each splitting unit decomposes a first splitting signal into multiple splitting signals, and the multiple splitting signals are transmitted to the optical modulator group.

[0015] In the programmable optical computing chip in the above embodiment, the separation of optical signals of different wavelengths can be achieved through the microcavity and the optical splitting unit, and the optical signal can be split efficiently.

[0016] In one embodiment, the optical modulator group includes multiple sub-optical modulator groups, each sub-optical modulator group includes multiple optical modulation lines, each optical modulation line includes an optical modulator and two first polarization demultiplexers; wherein, in each optical modulation line, an input end of a first polarization demultiplexer receives a split light signal, the other end of the first polarization demultiplexer is connected to the input end of the optical modulator, the output end of the optical modulator is connected to the input end of another first polarization demultiplexer, and the output end of the other first polarization demultiplexer outputs a first intermediate split light signal.

[0017] In the programmable optical computing chip in the above embodiment, by combining the optical modulator and the polarization demultiplexer, ultra-high-speed transmission can be achieved while maintaining the stability and reliability of the optical computing chip.

[0018] In one embodiment, the optical modulator further includes a plurality of wavelength multiplexer groups and a plurality of cross waveguides, wherein each wavelength multiplexer group includes a plurality of wavelength multiplexers, each wavelength multiplexer is interconnected and receives a plurality of first intermediate optical splitting signals, an output end of each wavelength multiplexer group outputs an input optical splitting signal, and each cross waveguide is formed at an intersection of a transmission path of a first intermediate signal.

[0019] In the programmable optical computing chip in the above embodiment, by combining multiple wavelength multiplexer groups and multiple cross waveguides, a highly integrated optical computing chip design can be achieved to meet the requirements of high-speed and large-capacity optical communication systems.

[0020] In one embodiment, each wavelength multiplexer includes a plurality of microring resonators and a plurality of first phase modulators, and each first phase modulator is used to modulate the phase of a microring resonator.

[0021] In the programmable optical computing chip in the above embodiment, the phase of light is changed by multiple first phase modulators to adjust the resonance condition of the microring resonator, thereby achieving modulation and control of the optical signal.

[0022] In one embodiment, the linear layer includes a plurality of second phase modulators, each of which is formed on a plurality of matrix-vector multiplication operation circuits in the linear layer, and is used to control and modulate the optical signals on the plurality of matrix-vector multiplication operation circuits according to a control signal of the control layer.

[0023] In the programmable optical computing chip in the above embodiment, the phase information of the optical signal is adjusted in the linear layer by multiple second phase modulators, thereby providing computing power for the optical computing chip.

[0024] In one embodiment, the output layer includes multiple wavelength demultiplexers and a photodetector array, wherein each wavelength demultiplexer is connected to an output optical signal, and decomposes an output optical signal into multiple sub-output optical signals according to different wavelengths of the output optical signal, and the photodetector array detects the optical signal intensities of the multiple sub-output optical signals to determine the result of the matrix-vector multiplication operation.

[0025] In the programmable optical computing chip in the above embodiment, the vector result after the matrix multiplication operation is determined by detecting the intensity of the output optical signal.

[0026] In one embodiment, the photodetector array includes a plurality of second polarization demultiplexers, each of which receives a sub-output optical signal and is used to perform calculations on the sub-output optical signals with respect to different polarizations.

[0027] In the programmable optical computing chip in the above embodiment, multiple second polarization demultiplexers are used to separate optical signals of different polarization states, so that the photodetector array can process multiple signals of a single polarization state, which can greatly improve the signal processing efficiency and sensitivity of the photodetector array. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0029] Figure 1 is a schematic diagram of a programmable optical computing chip provided in an embodiment;

[0030] Figure 2 is a schematic diagram of another programmable optical computing chip provided in an embodiment;

[0031] Figure 3 is a schematic diagram of a polarization-insensitive structure of an on-chip light source module provided in an embodiment;

[0032] Figure 4 A schematic diagram of a polarization-independent broadband wavelength (de)multiplexer design provided in one embodiment;

[0033] Figure 5A schematic diagram of a transmission spectrum of a polarization-independent broadband wavelength demultiplexer based on a silicon nitride waveguide layer provided in one embodiment;

[0034] Figure 6 A schematic diagram of a coupling coefficient of a microring resonator based on a silicon nitride waveguide layer provided in an embodiment;

[0035] Figure 7 A schematic diagram of a polarization-independent broadband cross-waveguide based on silicon nitride provided in an embodiment;

[0036] Figure 8 is a transmission spectrum of a polarization-independent broadband Mach-Zehnder modulator based on silicon nitride provided in one embodiment;

[0037] Fig. 9 A polarization-independent broadband directional coupler based on silicon nitride is provided in one embodiment;

[0038] Fig.10 Matrix-vector multiplication calculation results based on a silicon nitride waveguide 4×4 MZI array provided in one embodiment.

[0039] Description of reference numerals:

[0040] 101, input layer; 102, linear layer; 103, output layer; 104, control module; 201, on-chip light source; 202, microcavity; 203, wavelength division multiplexer; 204, light splitting unit; 205, optical modulator; 206, first polarization demultiplexer; 207, wavelength multiplexer; 208, cross waveguide; 209, second phase modulator; 210, wavelength demultiplexer; 211, second polarization demultiplexer; 212, photodetector array; 301, substrate; 302, insulating layer; 303, waveguide layer. DETAILED DESCRIPTION

[0041] In order to facilitate understanding of the present disclosure, the present disclosure will be described more fully below with reference to the relevant drawings. The preferred embodiments of the present disclosure are shown in the drawings. However, the present disclosure can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present disclosure more thorough and comprehensive.

[0042] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present disclosure belongs. The terms used herein in the specification of the present disclosure are only for the purpose of describing specific embodiments and are not intended to limit the present disclosure.

[0043] It should be understood that when an element or layer is referred to as being "on, adjacent to, connected to or coupled to other elements or layers, it may be directly on, adjacent to, connected to or coupled to other elements or layers, or there may be intervening elements or layers. On the contrary, when an element is referred to as being "directly on, directly adjacent to, directly connected to or directly coupled to other elements or layers, there are no intervening elements or layers. It should be understood that although the terms first, second, third, etc. may be used to describe various elements, components, regions, layers, doping types and / or parts, these elements, components, regions, layers, doping types and / or parts should not be limited by these terms. These terms are merely used to distinguish one element, component, region, layer, doping type or part from another element, component, region, layer, doping type or part. Therefore, without departing from the teachings of the present disclosure, the first element, component, region, layer, doping type or portion discussed below may be represented as a second element, component, region, layer or portion; for example, the first doping type may be referred to as the second doping type, and similarly, the second doping type may be referred to as the first doping type; the first doping type and the second doping type are different doping types, for example, the first doping type may be P-type and the second doping type may be N-type, or the first doping type may be N-type and the second doping type may be P-type.

[0044] Spatially relative terms such as "under," "beneath," "below," "under," "above," "above," and the like may be used herein to describe the relationship of an element or feature shown in the figures to other elements or features. It should be understood that, in addition to the orientations shown in the figures, spatially relative terms also include different orientations of the device in use and operation. For example, if the device in the accompanying drawings is flipped, an element or feature described as "under other elements" or "under it" or "under it" will be oriented as being "above" the other elements or features. Thus, the exemplary terms "under" and "under" may include both upper and lower orientations. In addition, the device may also include additional orientations (e.g., rotated 90 degrees or other orientations), and the spatial descriptors used herein are interpreted accordingly.

[0045] When used herein, the singular forms "a", "an" and " / the" may also include the plural forms, unless the context clearly indicates otherwise. It should also be understood that when the terms "consisting of" and / or "comprising" are used in this specification, the presence of features, integers, steps, operations, elements and / or components can be determined, but the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups is not excluded. At the same time, when used herein, the term "and / or" includes any and all combinations of the relevant listed items.

[0046] Embodiments of the invention are described herein with reference to cross-sectional views which are schematic representations of ideal embodiments (and intermediate structures) of the present disclosure, such that variations in the shapes shown due to, for example, manufacturing techniques and / or tolerances are anticipated. Thus, embodiments of the present disclosure should not be limited to the particular shapes of the regions shown herein, but rather include deviations in shapes due to, for example, manufacturing techniques. For example, an implanted region shown as a rectangle typically has rounded or curved features and / or an implant concentration gradient at its edges rather than a binary change from an implanted region to a non-implanted region. Similarly, a buried region formed by implantation may result in some implantation in the region between the buried region and the surface through which the implantation is performed. Thus, the regions shown in the figures are schematic in nature, their shapes do not represent the actual shape of the region of the device, and do not limit the scope of the present disclosure.

[0047] Please refer to Figure 1-Figure 10 An embodiment of the present disclosure provides a programmable optical computing chip, including: an input layer 101, an output layer 103, a linear layer 102 and a control module 104.

[0048] Among them, the input layer 101 includes an on-chip light source 201 module, a wavelength division multiplexer 203 and an optical modulator 205 group. The on-chip light source 201 module inputs a multi-wavelength light source into the wavelength division multiplexer 203 to decompose the input light source into multiple split light signals. The multiple split light signals are input into the optical modulator 205 group. The optical modulator 205 group is used to perform intensity encoding on each split light signal to obtain multiple input split light signals.

[0049] A linear layer 102, wherein the input end of the linear layer 102 is connected to the output end of the input layer 101, so as to perform a matrix-vector multiplication operation on a plurality of input optical splitting signals inputted by the input layer 101, and obtain a plurality of output optical signals; an output layer 103, connected to the output end of the linear layer 102, wherein the output layer 103 detects the optical signal strength of the plurality of output optical signals, and determines the matrix-vector multiplication operation result according to the optical signal strength of each output optical signal; a control module 104, wherein the control module 104 is respectively connected to the input layer 101, the linear layer 102 and the output layer 103, wherein the control module 104 is used to control the optical splitting coding mode of the input layer 101 and the matrix-vector multiplication operation mode of the linear layer 102, and is used to receive the matrix-vector multiplication operation result outputted by the output layer 103.

[0050] In the programmable optical computing chip in the above embodiment, the input layer 101 uses an on-chip light source 201 module and a wavelength division multiplexer 203, so that the input layer 101 can be applied to different polarized light sources and can realize parallel computing based on different polarized lights. The input layer 101 also uses an optical modulation group to perform intensity encoding on optical signals of different wavelengths. The linear layer 102 realizes the matrix-vector multiplication operation of the input split light signal through the control of the control unit. The output layer 103 outputs the vector result after the matrix multiplication operation by detecting the intensity of the optical signals at different output ports. Therefore, the embodiment of the present disclosure can at least realize large-scale parallel data operation, and has polarization insensitivity, and can process multiple signals of different polarization states at the same time, thereby improving data throughput and computing efficiency. The control module 104 controls the input layer 101, the output layer 103 and the linear layer 102 to realize a large number of parallel matrix-vector multiplication operations at the optical data rate, which can achieve efficient and stable computing performance.

[0051] At the same time, the use of optical computing chips can avoid inconsistent computing results or reduced efficiency due to unstable polarization states of optical signals, and optical computing chips have extremely high data modulation speeds, which are only limited by the bandwidth of the on-chip light source 201 and the photodetector.

[0052] For details, please refer to Figure 2-Figure 3 The on-chip light source 201 module includes: a substrate 301; an insulating layer 302 located on the substrate 301; and a waveguide layer 303 located in the insulating layer 302.

[0053] In the programmable optical computing chip in the above embodiment, the on-chip light source 201 is used to implement a polarization-insensitive waveguide design.

[0054] The thickness of the waveguide layer 303 is 0.6 μm-1.0 μm, the thickness of the insulating layer 302 is 6.6 μm-7 μm, and the width of the waveguide layer 303 is 0.6 μm-1.0 μm.

[0055] As an example, the thickness of the waveguide layer 303 may be 0.6 μm, 0.8 μm, 1.0 μm, etc.

[0056] As an example, the thickness of the insulating layer 302 is 6.6 μm, 6.8 μm, 7 μm, and so on.

[0057] As an example, the waveguide layer 303 has a width of 0.6 μm, 0.8 μm, 1.0 μm, and so on.

[0058] In the programmable optical computing chip in the above embodiment, by limiting the size of the on-chip light source 201, the refractive indexes of the TE mode and the TM mode of the on-chip light source 201 are similar, further realizing the polarization-insensitive waveguide design of the optical computing chip.

[0059] For example, please continue to refer to Figure 3 , Figure 3 A schematic diagram of a possible polarization-insensitive structure of an on-chip light source 201 module is shown. Figure 3 An optional size of the insulating layer 302 and the waveguide layer 303 is shown in FIG. Figure 3 As shown in Figure a, the thickness of the insulating layer 302 can be set to 6.8 μm, and the thickness and width of the waveguide layer 303 can be set to 0.8 μm. Figure 3 Taking the size of the on-chip light source 201 module in Figure a as an example, Figure 3 Figures b and c in the figure show that Figure 3 The simulation results of TE mode and TM mode under the size of Figure a are as follows: Figure 3 As shown in Figures b and c in FIG. 1 , the refractive indexes of the TE mode and the TM mode of the on-chip light source 201 module are very close, specifically 1.729.

[0060] Among them, TE wave (i.e. s wave): has a magnetic field component but no electric field component in the propagation direction, which is called transverse electric wave. TM wave (i.e. p wave): has an electric field component but no magnetic field component in the propagation direction, which is called transverse magnetic wave.

[0061] Wherein, the substrate 301 can be made of semiconductor material, insulating material, conductor material or any combination thereof. The substrate 301 can be a single-layer structure or a multi-layer structure. For example, the substrate 301 can be a silicon substrate 301, a silicon germanium substrate 301, a silicon germanium carbon substrate 301, a silicon carbide substrate 301, a gallium arsenide substrate 301, an indium arsenide substrate 301, an indium phosphide substrate 301 or other III / V semiconductor substrates 301 or II / VI semiconductor substrates 301. Alternatively, for example, the substrate 301 can be a layered substrate 301 including silicon / silicon carbide, silicon on insulator or silicon germanium on insulator. Therefore, the type of substrate 301 should not limit the scope of protection of the present disclosure.

[0062] The material of the insulating layer 302 may be silicon oxide, and the material of the waveguide layer 303 may be silicon nitride.

[0063] For details, please continue to refer to Figure 1 and Figure 2 The wavelength division multiplexer 203 includes a microcavity 202 unit and four light splitting units 204 .

[0064] The microcavity 202 unit is formed between the on-chip light source 201 module and the wavelength division multiplexer 203. The microcavity 202 unit includes an input end and four output ends. The input end is used to connect to the on-chip light source 201. Each output end is connected to a splitter unit 204 to transmit multiple first splitter signals to multiple splitter units 204. Each splitter unit 204 decomposes a first splitter signal into multiple splitter signals, and transmits the multiple splitter signals to the optical modulator 205 group.

[0065] Please continue reading Figure 1 and Figure 2 The optical modulator 205 group includes multiple sub-optical modulator 205 groups, each sub-optical modulator 205 group includes multiple optical modulation lines, and each optical modulation line includes an optical modulator 205 and two first polarization demultiplexers 206.

[0066] In each optical modulation line, an input end of a first polarization demultiplexer 206 receives a split light signal, the other end of the first polarization demultiplexer 206 is connected to the input end of the optical modulator 205, the output end of the optical modulator 205 is connected to the input end of another first polarization demultiplexer 206, and the output end of the other first polarization demultiplexer 206 outputs a first intermediate split light signal.

[0067] For details, please continue to refer to Figure 1 and Figure 2 The optical modulator 205 further includes a plurality of wavelength multiplexers 207 and a plurality of cross waveguides 208.

[0068] Each wavelength multiplexer 207 group includes multiple wavelength multiplexers 207, each wavelength multiplexer 207 is interconnected and receives multiple first intermediate optical splitting signals, the output end of each wavelength multiplexer 207 group outputs an input optical splitting signal, and each cross waveguide 208 is formed at the intersection of the transmission path of a first intermediate signal.

[0069] For details, please continue to refer to Figure 1 , Figure 2 and Figure 4 Each wavelength multiplexer 207 includes a plurality of microring resonators and a plurality of first phase modulators, and each first phase modulator is used to modulate the phase of a microring resonator.

[0070] The circumference of each microring resonator is 71.70 microns to 85.31 microns.

[0071] As an example, the circumference of each microring resonator can be 71.70 microns, 75.10 microns, 81.60 microns, 85.31 microns, and so on.

[0072] Preferably, the circumferences of the microring resonators 1-1, 1-2, 2-1, 2-2, 3-1, 3-2, 4-1, and 4-2 may be 71.70 microns, 85.15 microns, 71.75 microns, 85.20 microns, 71.80 microns, 85.26 microns, 71.84 microns, and 85.31 microns, respectively.

[0073] See also Figure 6 , Figure 6 Schematic diagram of the coupling coefficient of the microring resonator based on silicon nitride waveguide, where the TE and TM modes show a high degree of consistency. Since the refractive index and group refractive index of the waveguide are also relatively consistent, the broadband wavelength (de)multiplexer design has polarization-independent properties.

[0074] See also Figure 7 , Figure 7 Schematic diagram of a polarization-independent broadband cross-waveguide 208 based on silicon nitride. Figure 7 As shown, the design and transmission spectrum of the polarization-independent broadband cross-waveguide 208 based on silicon nitride are shown, and the device exhibits polarization-insensitive properties over a wide wavelength range.

[0075] For details, please continue to refer to Figure 1-Figure 2 The linear layer 102 includes a plurality of second phase modulators 209, each of which is formed on a plurality of matrix-vector multiplication operation lines in the linear layer 102, and is used to control and modulate the optical signals on the plurality of matrix-vector multiplication operation lines according to the control signal of the control layer.

[0076] Specifically, the output layer 103 includes a plurality of wavelength demultiplexers 210 and a photodetector array 212 .

[0077] Each wavelength demultiplexer 210 is connected to an output optical signal, and decomposes an output optical signal into multiple sub-output optical signals according to different wavelengths of the output optical signal. The photoelectric detection array detects the optical signal intensities of the multiple sub-output optical signals to determine the matrix-vector multiplication result.

[0078] As an example, see Figure 5 , Figure 5 The transmission spectrum of the polarization-independent broadband wavelength demultiplexer 210 based on silicon nitride waveguide obtained by simulation is shown in FIG. Figure 5 As shown in , the polarization-independent broadband wavelength demultiplexer 210 based on silicon nitride waveguide realizes four-channel wavelength division multiplexing with a spacing of 100 GHz and the sideband noise is less than 25 dB.

[0079] Specifically, the photodetector array 212 includes a plurality of second polarization demultiplexers 211 , each of which receives a sub-output optical signal and is used to perform calculations on the sub-output optical signals with respect to different polarizations.

[0080] As an example, see Figure 8 , Figure 8 Figure a shows a polarization-independent broadband directional coupler design based on silicon nitride. Figure 8 Figures b and c show the Figure 8 Figure a shows the design and simulation results of a polarization-independent broadband directional coupler made of silicon nitride. The optical computing chip exhibits polarization-insensitive properties over a wide wavelength range and can achieve 50:50 splitting of TE and TM modes.

[0081] As an example, see Fig. 9 , Fig. 9 The transmission spectrum of a polarization-independent broadband Mach-Zehnder modulator based on silicon nitride is shown. The optical computing chip is implemented based on the two above-mentioned 2×2 directional couplers and waveguides. The optical computing chip realizes optical interference in a wide wavelength range, with an extinction ratio greater than 15dB and has polarization-insensitive characteristics.

[0082] As an example, see Fig.10 , Fig.10 The results of matrix-vector multiplication calculation based on silicon nitride waveguide 4×4 MZI array are shown. Fig.10 The matrix-vector multiplication calculation results of the silicon nitride waveguide 4×4 MZI array are shown in Figure 1 when the input optical powers of the four input ports are 0.25mW, 0.25mW, 0.25mW, and 0.25mW. Fig.10 Figure a in Fig.10 Figure a in Figure 1 shows the output optical powers (TE and TM modes) of the four ports when the transmission matrix of the 4×4 MZI array is [1,1,1,1;0,0,0,0;0,0,0,0,0;0,0,0,0;], which is consistent with the calculation results of matrix multiplication. Fig.10 Figure b in the figure, Fig.10 Figure b shows the output optical power (TE and TM modes) of the four ports when the transmission matrix of the 4×4 MZI array is [0,0,0,0;0,0,0,0,0;0,0,0,0;1,1,1,1;], which is consistent with the calculation result of matrix multiplication.

[0083] In this way, the optical computing chip proposed in this application can integrate WDM (Wavelength Division Multiplexing) and MDM (Mode Division Multiplexing) technologies, and the system can achieve high-dimensional parallel computing in a smaller physical space, while reducing the need for a large number of optical and electronic components, thereby reducing power consumption.

[0084] It should be noted that WDM (Wavelength Division Multiplexing) and MDM (Mode Division Multiplexing) technologies use the spatial freedom of light waves to increase the capacity of optical communications. As a multiplexing method that can overcome the transmission capacity limitations of single-mode optical fiber communication systems, it has received widespread attention. The long-distance transmission of MDM signals requires optical amplifiers to supplement energy. Few-mode erbium-doped fiber amplifiers (FM-EDFA) can simultaneously amplify multiple wavelengths and multiple optical fiber guided modes, which can greatly reduce the cost of MDM systems.

[0085] By combining WDM and MDM technologies, this application enables data processing in multiple dimensions simultaneously, significantly improving throughput and computing efficiency. The high precision and programmability of the optical frequency comb light source further enhance the performance of the system. At the same time, by introducing polarization-insensitive devices, this application enables the system to maintain compatibility with different polarization states when processing multi-channel optical signals, ensuring efficient and stable computing performance.

[0086] The programmable optical computing chip of the present application allows for dynamic adjustment of computing paths and resource allocation according to task requirements, enabling the system to adapt to a variety of application scenarios, thereby improving resource utilization.

[0087] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features of the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0088] The above embodiments only express several implementation methods of the present disclosure, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the patent application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present disclosure, and these all belong to the protection scope of the present disclosure. Therefore, the protection scope of the patent of the present disclosure shall be subject to the attached claims.

Claims

1. A programmable optical computing chip, characterized in that: include: An input layer, the input layer includes an on-chip light source module, a wavelength division multiplexer and an optical modulator group, the on-chip light source module inputs a multi-wavelength light source into the wavelength division multiplexer to decompose the input light source into a plurality of split light signals, the plurality of split light signals are input into the optical modulator group, the optical modulator group is used to perform intensity encoding on each split light signal to obtain a plurality of input split light signals; A linear layer, wherein an input end of the linear layer is connected to an output end of the input layer to perform a matrix-vector multiplication operation on a plurality of input split optical signals inputted by the input layer to obtain a plurality of output optical signals; an output layer connected to an output end of the linear layer, wherein the output layer detects optical signal strengths of a plurality of output optical signals and determines a matrix-vector multiplication result according to the optical signal strength of each output optical signal; A control module, wherein the control module is connected to the input layer, the linear layer and the output layer respectively, and the control module is used to control the spectral coding mode of the input layer and the matrix-vector multiplication operation mode of the linear layer, and is used to receive the matrix-vector multiplication operation result output by the output layer.

2. The programmable optical computing chip according to claim 1, characterized in that: The on-chip light source comprises: substrate; an insulating layer, located on the substrate; The waveguide layer is located in the insulating layer.

3. The programmable optical computing chip according to claim 2, characterized in that: The thickness of the waveguide layer is 0.6 μm-1.0 μm, the thickness of the insulating layer is 6.6 μm-7 μm, and the width of the waveguide layer is 0.6 μm-1.0 μm.

4. The programmable optical computing chip according to claim 1, characterized in that: The wavelength division multiplexer comprises a microcavity unit and four light splitting units. The microcavity unit is formed between the on-chip light source module and the wavelength division multiplexer. The microcavity unit includes an input end and four output ends. The input end is used to connect to the on-chip light source. Each output end is connected to a splitter unit. A plurality of first splitter signals are transmitted to the plurality of splitter units. Each splitter unit decomposes a first splitter signal into a plurality of splitter signals, and the plurality of splitter signals are transmitted to the optical modulator group.

5. The programmable optical computing chip according to claim 4, characterized in that: The optical modulator group includes a plurality of sub-optical modulator groups, each sub-optical modulator group includes a plurality of optical modulation circuits, each optical modulation circuit includes an optical modulator and two first polarization demultiplexers; Among them, in each optical modulation line, the input end of a first polarization demultiplexer receives a split light signal, the other end of the first polarization demultiplexer is connected to the input end of the optical modulator, the output end of the optical modulator is connected to the input end of another first polarization demultiplexer, and the output end of the other first polarization demultiplexer outputs a first intermediate split light signal.

6. The programmable optical computing chip according to claim 5, characterized in that: The optical modulator further comprises a plurality of wavelength multiplexer groups and a plurality of crossing waveguides, Each wavelength multiplexer group includes multiple wavelength multiplexers, each wavelength multiplexer is interconnected and receives multiple first intermediate optical splitting signals, the output end of each wavelength multiplexer group outputs an input optical splitting signal, and each cross waveguide is formed at the intersection of the transmission path of a first intermediate signal.

7. The programmable optical computing chip according to claim 6, characterized in that: The microcavity unit includes a plurality of microring resonators and a plurality of first phase modulators, and each first phase modulator is used to modulate the phase of a microring resonator.

8. The programmable optical computing chip according to claim 1, characterized in that: The linear layer includes a plurality of second phase modulators, each of which is formed on a plurality of matrix-vector multiplication operation circuits in the linear layer and is used to control and modulate optical signals on the plurality of matrix-vector multiplication operation circuits according to control signals of the control layer.

9. The programmable optical computing chip according to claim 1, characterized in that: The output layer includes a plurality of wavelength demultiplexers and a photodetector array, Wherein, each wavelength demultiplexer is connected to an output optical signal, and decomposes an output optical signal into multiple sub-output optical signals according to different wavelengths of the output optical signal. The photoelectric detection array detects the optical signal intensity of the multiple sub-output optical signals to determine the matrix-vector multiplication operation result.

10. The photodetector array comprises a plurality of polarization demultiplexers, characterized in that: The photodetector array includes a plurality of second polarization demultiplexers, each of which receives a sub-output optical signal and is used to perform calculations on the sub-output optical signals with respect to different polarizations.

Images