Sensor based on on-chip optical diffraction neural network computing architecture

Abstract

The present application relates to the technical field of intelligent sensor chips, and in particular to a sensor based on an on-chip optical diffraction neural network computing architecture, including an input end and an output end; and a metasurface sensing structure arranged between the input end and the output end. A sensing medium is filled in the metasurface unit of the metasurface sensing structure corresponding to the diffraction neural network to form a detector array, and the detector array is used to obtain the diffraction calculation relationship between the incident light input at the input end and the output light output at the output end, and the type and / or concentration of the substance to be tested is determined according to the change law of the metasurface refractive index after the sensing medium absorbs the substance to be tested. Thus, the intelligent sensing calculation is completed by using the controlled propagation of light, and the sensing unit is set on the path of the controlled propagation of light, and the identification and sensing of the substance to be tested is completed by changing the propagation result of the light.

Description

Technical Field

[0001] The present application relates to the field of smart chip sensing technology, and in particular to a sensor based on an on-chip optical diffraction neural network computing architecture. Background Art

[0002] Common electrical sensing architectures include the use of sensitive electrodes or sensitive field effect transistors. The performance drift and noise of electrical sensing architectures affect the reliability of their performance. Optical sensing has no electrical interference and low background noise. It is widely used in gas / liquid sensing technologies, including spectral analysis technology (ultraviolet-visible-infrared), fluorescence analysis technology, chemiluminescence technology, infrared Fourier transform spectroscopy technology, Raman spectroscopy technology, etc. However, these technologies require sample pretreatment, and the equipment is large, costly, and energy-intensive, and are not suitable for real-time in-situ detection applications.

[0003] Silicon-based optoelectronic technology provides a highly integrated solution. Currently, silicon waveguide material modification technology can be used to realize silicon optical sensing functions, including Mach-Zehnder interferometer refractive index control technology and micro-ring resonant cavity resonance modulation technology. However, these technologies often require narrow linewidth lasers and high-resolution / wavelength-resolvable detector systems, which are costly, have low yields, and often only have one substance to be tested. Summary of the invention

[0004] The present application provides a sensor based on an on-chip optical diffraction neural network computing architecture, which uses the controlled propagation of light to complete intelligent sensing calculations, sets a sensing unit on the path of controlled light propagation, and completes the identification and sensing of the substance to be measured by changing the propagation results of light.

[0005] The first aspect of the present application provides a sensor based on an on-chip optical diffraction neural network computing architecture, comprising: an input end and an output end; a metasurface sensing structure arranged between the input end and the output end, wherein a sensing medium is filled in a metasurface unit of the metasurface sensing structure corresponding to the diffraction neural network to form a detector array, and the detector array is used to obtain a diffraction calculation relationship between the incident light input by the input end and the output light output by the output end, and the type and / or concentration of the substance to be tested is determined according to the change law of the metasurface refractive index after the sensing medium absorbs the substance to be tested.

[0006] Optionally, when the substance to be tested is absent, the refractive index of the sensing medium does not change, and the incident light is mapped to the detector array through weighted summation using preset weights of the diffraction neural network to obtain a reference output light; when the substance to be tested is present, the refractive index of the sensing medium changes, and the incident light is mapped to the detector array through weighted summation using the changed weights of the diffraction neural network to obtain the output light.

[0007] Optionally, the supersurface sensing structure is a silicon-based supersurface structure.

[0008] Optionally, the supersurface sensing structure is an organic supersurface structure.

[0009] Optionally, the plurality of sensing units of the detector array are arranged on a light-controlled propagation path.

[0010] Optionally, the type is gas or liquid.

[0011] The second aspect of the present application provides a sensor detection method based on an on-chip optical diffraction neural network computing architecture. The sensor based on the on-chip optical diffraction neural network computing architecture is used to place the sensor in the environment of the substance to be tested, and the incident light is passed into the sensor. The detection method includes the following steps: obtaining the diffraction calculation relationship between the incident light and the output light through the detector array; determining the type and / or concentration of the substance to be tested based on the change law of the metasurface refractive index after the sensing medium absorbs the substance to be tested and the diffraction calculation relationship.

[0012] Optionally, when the substance to be tested is absent, the refractive index of the sensing medium does not change, and the incident light is mapped to the detector array through weighted summation using preset weights of the diffraction neural network to obtain a reference output light; when the substance to be tested is present, the refractive index of the sensing medium changes, and the incident light is mapped to the detector array through weighted summation using the changed weights of the diffraction neural network to obtain the output light.

[0013] The third aspect of the present application provides a chip, including: the above-mentioned sensor based on the on-chip optical diffraction neural network computing architecture.

[0014] The sensor, detection method and chip based on the on-chip optical diffraction neural network computing architecture of the embodiment of the present application use the controlled propagation of light to complete intelligent sensing calculations, set up sensing units on the path of controlled light propagation, and complete target recognition and sensing by changing the propagation results of light. It has the advantages of small size, low power consumption and high speed.

[0015] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0017] Figure 1A structural block diagram of a sensor based on an on-chip optical diffraction neural network computing architecture provided according to an embodiment of the present application;

[0018] Figure 2 A schematic diagram of the structure of a sensor based on an on-chip optical diffraction neural network computing architecture provided according to an embodiment of the present application;

[0019] Figure 3 The present invention provides a flow chart of a detection method based on an on-chip optical diffraction neural network computing architecture according to an embodiment of the present application. DETAILED DESCRIPTION

[0020] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0021] Figure 1 The present invention is a structural block diagram of a sensor based on an on-chip optical diffraction neural network computing architecture provided according to an embodiment of the present application.

[0022] like Figure 1 As shown, the sensor 10 based on the on-chip optical diffraction neural network computing architecture includes an input end 100 and an output end 200; and a metasurface sensing structure 300 arranged between the input end and the output end.

[0023] Among them, a sensing medium is filled in the supersurface unit of the supersurface sensing structure 300 corresponding to the diffraction neural network to form a detector array, and the detector array is used to obtain the diffraction calculation relationship between the incident light input at the input end and the output light output at the output end, and the type and / or concentration of the substance to be tested is determined according to the change law of the supersurface refractive index after the sensing medium absorbs the substance to be tested.

[0024] In the embodiments of the present application, the supersurface sensing structure may be a silicon-based supersurface structure or an organic supersurface structure.

[0025] In the embodiment of the present application, the waveguide area is a neural node of the diffraction neural network. Thus, the optical diffraction neural network architecture on a silicon substrate is adopted to intelligently analyze the sensor calculation results, with low energy consumption and high speed. Among them, the optical diffraction neural network can be designed by an artificial intelligence algorithm, which can make the sensor in a resonant sensitive state with high sensitivity.

[0026] In the embodiments of the present application, the type of substance to be detected may be gas or liquid. By setting the neuron device in the sensor as a gas / liquid sensitive material, it is possible to simultaneously monitor a variety of different gases / liquids.

[0027] It can be understood that the sensor based on the on-chip optical diffraction neural network computing architecture of the embodiment of the present application mainly includes a silicon-based metasurface structure, in which light propagates in a restricted manner. By absorbing the substance to be measured (gas / liquid) to change the refractive index of the metasurface, different restricted propagation modes can be obtained, thereby completing sensing.

[0028] Optionally, the light source of the incident light may be a CW laser, and the detector array may be a waveguide detector. A plurality of sensing units of the detector array are arranged on the light-controlled propagation path.

[0029] Furthermore, the sensitive materials of the sensor can be distributed in the neuron nodes in the network, and the sensitive materials can also be distributed in other waveguide areas to assist in completing the limited propagation function of light. In addition, in addition to silicon, the optical sensing / computing area can also be replaced by functional materials such as organic matter.

[0030] In an embodiment of the present application, when there is no substance to be measured, the refractive index of the sensing medium does not change, and the incident light is mapped to the detector array through the weighted summation effect of the preset weights of the diffraction neural network to obtain a reference output light; when there is a substance to be measured, the refractive index of the sensing medium changes, and the incident light is mapped to the detector array through the weighted summation effect of the changed weights of the diffraction neural network to obtain the output light.

[0031] Specifically, combined Figure 2 As shown, the incident light passes through the diffraction area from the input port along the light propagation direction and is output from the output port. 1 , I 2 …, I r ) is mapped to the detector array through the weighted summation of the neural network to obtain the output light (O 1 , O 2 …, O l ). When there is a substance to be tested, the weights of the neural network nodes change, and the weighted summation result becomes (O' 1 , O' 2 …, O' l ). Therefore, the embodiments of the present application can highly sensitively infer changes in information such as the type and concentration of the substance to be detected by detecting changes in output light information.

[0032] It can be understood that the operational relationship of the sensor of the embodiment of the present application conforms to certain mathematical laws and can be regarded as the multiplication of vectors and matrices, including input vectors, output vectors and intermediate matrices. When the metasurface structure forms a detector array structure that conforms to the matrix relationship, when there is no substance to be measured, the refractive index of the sensing medium does not change, the incident light propagates according to the known ray propagation law, and the output mode of the sensor output light is the standard mode and is known. When the sensor exists in the substance to be measured, the substance to be measured changes the refractive index of the sensing medium, thereby changing the ray propagation law of the incident light, and the output mode of the output light changes. At the same time, for different substances to be measured, the refractive index of the sensing medium is changed differently. Therefore, referring to the standard output mode, the embodiment of the present application can use the detector array to obtain the diffraction calculation relationship between the incident light and the output light, and the law of change of the refractive index of the metasurface after the sensing medium absorbs the substance to be measured, and determine what kind of substance the substance to be measured belongs to, such as refractive index A corresponds to gas, refractive index B corresponds to liquid, etc., and the concentration of the corresponding substance can be further obtained according to the refractive index. For this, those skilled in the art can set it according to the actual situation without specific restrictions.

[0033] The sensor based on the on-chip optical diffraction neural network computing architecture according to the embodiment of the present application has the advantages of high integration of existing silicon-based photoelectric sensors. Unlike the Mach-Zehnder interferometer refractive index control technology and the micro-ring resonant cavity resonance modulation technology, it adopts a neural network intelligent algorithm, does not require a narrow linewidth laser and a high-resolution / wavelength-resolvable detector system, and can monitor multiple gases / liquids at the same time.

[0034] Next, the detection method based on the on-chip optical diffraction neural network computing architecture proposed in accordance with an embodiment of the present application is described with reference to the accompanying drawings.

[0035] Figure 3 The present invention provides a flow chart of a detection method based on an on-chip optical diffraction neural network computing architecture according to an embodiment of the present application.

[0036] like Figure 3 As shown, the sensor is placed in the environment of the material to be tested, and the incident light is passed into the sensor. The detection method based on the on-chip optical diffraction neural network computing architecture includes:

[0037] Step S101, obtaining the diffraction calculation relationship between the incident light and the output light through the detector array.

[0038] Specifically, incident light is emitted to the input port through a light source, the incident light is propagated through the diffraction region, and optical information of output light at the output port is acquired.

[0039] Step S102, determining the type and / or concentration of the substance to be measured according to the change law of the metasurface refractive index after the sensing medium absorbs the substance to be measured and the diffraction calculation relationship.

[0040] Optionally, when there is no substance to be measured, the refractive index of the sensing medium does not change, and the incident light is mapped to the detector array through the weighted summation effect of the preset weights of the diffraction neural network to obtain a reference output light; when there is a substance to be measured, the refractive index of the sensing medium changes, and the incident light is mapped to the detector array through the weighted summation effect of the changed weights of the diffraction neural network to obtain the output light.

[0041] It should be noted that the aforementioned explanation of the sensor embodiment based on the on-chip optical diffraction neural network computing architecture is also applicable to the detection method based on the on-chip optical diffraction neural network computing architecture of this embodiment, and will not be repeated here.

[0042] According to the detection method based on the on-chip optical diffraction neural network computing architecture proposed in the embodiment of the present application, the optical information of the output light outputted from the output end is obtained; the metasurface refractive index changed by the absorption of the substance to be tested is calculated according to the optical information of the incident light inputted from the input end and the output light outputted from the output end; the type and / or concentration of the substance to be tested is determined according to the metasurface refractive index. The calculation method has low energy consumption and high speed, and can monitor a variety of different gases / liquids at the same time.

[0043] This embodiment also provides a chip, including the sensor based on the on-chip optical diffraction neural network computing architecture of the above embodiment.

[0044] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or N embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0045] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0046] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or N executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

Claims

1. A sensor based on an on-chip optical diffraction neural network computing architecture, characterized in that: include: Input and output terminals; A metasurface sensing structure is arranged between the input end and the output end, and a sensing medium is filled in the metasurface unit of the metasurface sensing structure corresponding to the diffraction neural network to form a detector array, and the diffraction calculation relationship between the incident light input by the input end and the output light output by the output end is obtained by using the detector array, and the type and / or concentration of the substance to be tested is determined according to the change law of the metasurface refractive index after the sensing medium absorbs the substance to be tested; When the substance to be tested is absent, the refractive index of the sensing medium does not change, and the incident light is mapped onto the detector array through a weighted summation effect performed by the preset weights of the diffraction neural network to obtain a reference output light; When the substance to be detected is present, the refractive index of the sensing medium changes, and the incident light is mapped onto the detector array through weighted summation through the changed weights of the diffraction neural network to obtain the output light.

2. The sensor according to claim 1, characterized in that The supersurface sensing structure is a silicon-based supersurface structure.

3. The sensor according to claim 1, characterized in that The supersurface sensing structure is an organic supersurface structure.

4. The sensor according to claim 1, characterized in that The plurality of sensing units of the detector array are arranged on a light-controlled propagation path.

5. The sensor according to claim 1, characterized in that The species is gas or liquid.

6. A sensor detection method based on an on-chip optical diffraction neural network computing architecture, characterized in that: Using the sensor based on the on-chip optical diffraction neural network computing architecture as described in any one of claims 1 to 5, wherein the sensor is placed in the environment of the substance to be detected, and the incident light is passed into the sensor, and the detection method comprises the following steps: Acquiring a diffraction calculation relationship between the incident light and the output light through the detector array; Determine the type and / or concentration of the substance to be measured according to the change law of the metasurface refractive index after the sensing medium absorbs the substance to be measured and the diffraction calculation relationship; When the substance to be tested is absent, the refractive index of the sensing medium does not change, and the incident light is mapped onto the detector array through a weighted summation effect performed by the preset weights of the diffraction neural network to obtain a reference output light; When the substance to be detected is present, the refractive index of the sensing medium changes, and the incident light is mapped onto the detector array through weighted summation through the changed weights of the diffraction neural network to obtain the output light.

7. A chip, characterized in that: A sensor based on an on-chip optical diffraction neural network computing architecture comprising any one of claims 1-5.

Images