A compact fluid refractive index measurement system and a fluid refractive index measurement method

Abstract

This invention relates to a fluid refractive index measurement device and method, specifically a compact fluid refractive index measurement system and method. It addresses the technical problem that existing optical refractive index sensors have low sensing sensitivity and overall performance evaluation indicators, making it difficult to meet the high-precision measurement requirements of fluid refractive index. The compact fluid refractive index measurement system provided by this invention is based on a double-layer metasurface. It utilizes a high-Q metasurface to convert the refractive index into spectral information. By leveraging its narrow linewidth and near-field enhancement characteristics, it improves the interaction intensity between the measurement beam and the fluid under test, thereby increasing measurement accuracy. Combined with a miniaturized spectrometer, the spectral information is converted into intuitive light intensity information. Through analysis and processing of this light intensity information, the refractive index of the fluid under test can be quickly obtained, simplifying the measurement process and improving measurement efficiency. The fluid refractive index measurement method provided by this invention introduces an error term in the spectral inversion, enabling accurate inversion of spectral information.

Description

Technical Field

[0001] This invention relates to a fluid refractive index measurement device and method, specifically to a compact fluid refractive index measurement system and a fluid refractive index measurement method. Background Technology

[0002] In the field of modern detection technology, refractive index sensing technology, with its unique detection mechanism, has become a key tool for revealing the microscopic properties and dynamic processes of matter. Refractive index sensing technology can accurately capture changes in the flow rate of a sample, enabling real-time tracking of even minute fluctuations in refractive index, whether for the precise delivery of cell suspensions in microfluidic chips or the monitoring of liquid flow rates in industrial pipelines. Furthermore, this technology has demonstrated significant advantages in observing molecular binding processes and monitoring interactions between cells.

[0003] As the core platform for refractive index sensing technology, optical sensors, with their inherent high sensitivity, non-contact measurement, and high resolution, have further promoted the application expansion of this technology. During measurement, optical sensors do not require direct contact with the sample, effectively avoiding sample contamination and sensor wear, making them particularly suitable for measuring bioactive or highly corrosive samples. Their high sensitivity allows them to capture minute changes in refractive index, even achieving single-molecule level measurements, meeting the needs of scientific research and clinical practice for measuring trace substances. High resolution ensures the accuracy and reliability of the measurement results, providing a solid foundation for subsequent data analysis and conclusion derivation.

[0004] In the development of optical refractive index sensing technology, spectral sensing chips have always played a core role. In recent years, researchers have proposed various implementation schemes for optical refractive index biosensors based on these chips. Among them, surface plasmon resonance sensors, due to their surface electromagnetic field enhancement effect, have been widely used in biomolecular measurements. They convert changes in refractive index into changes in the intensity or phase of optical signals through the interaction between plasmon resonances on the surface of a metal thin film and incident light. However, due to the inherent optical absorption loss of metal materials, their energy loss is relatively large, making it difficult to further improve the measurement sensitivity. Photonic crystal sensors utilize the photonic bandgap characteristics of photonic crystals to achieve measurement through the bandgap shift caused by changes in refractive index. However, their structural design is complex, requiring extremely high precision in fabrication processes, and their stability is poor over a wide wavelength measurement range. Whispering-gallery mode sensors rely on the whispering-gallery mode formed by total internal reflection of light on the surface of a microcavity for measurement, and have high potential for quality factor. However, the large optical absorption and surface scattering losses of the microcavity material significantly reduce the quality factor in practical applications.

[0005] Overall, due to the inherent defects of different technical principles and the high optical absorption loss of the sensor materials themselves, the quality factors of most mainstream optical refractive index sensors are currently in the tens to hundreds range. This directly leads to low sensing sensitivity and overall performance evaluation indicators, making it difficult to meet the requirements of high precision and high stability in actual measurement scenarios, thus restricting their application in a wider range of fields. Summary of the Invention

[0006] The purpose of this invention is to solve the technical problem that the sensing sensitivity and overall performance evaluation index of existing optical refractive index sensors are too low, making it difficult to meet the requirements of high-precision measurement of fluid refractive index, and to provide a compact fluid refractive index measurement system and fluid refractive index measurement method.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0008] A compact fluid refractive index measurement system is characterized by comprising a light source, a high-Q metasurface, a miniaturized spectrometer and a detector arranged sequentially, and a data processing module electrically connected to the detector;

[0009] The high-Q metasurface is encapsulated in a microfluidic cavity, which has an inlet and an outlet for injecting and discharging the fluid to be tested, respectively. The microfluidic cavity is used to bring the fluid to be tested into contact with the high-Q metasurface.

[0010] The high-Q metasurface adopts a continuous domain bound state and includes N periodically arranged superpixels. The high-Q metasurface is used to convert the refractive index information of the fluid under test into spectral information, where N is an integer and N≥4;

[0011] The miniaturized spectrometer includes M periodically arranged spectral units that match the size of the superpixel, used to convert spectral information into light intensity information;

[0012] The detector includes M detection units, each corresponding to one of the M spectral units, for detecting light intensity information;

[0013] The data processing module includes a neural network, which is used to perform spectral inversion based on light intensity information to obtain a spectral curve, thereby deriving the refractive index of the fluid under test.

[0014] Furthermore, the superpixel comprises 2n amorphous silicon square blocks arranged in an array, whose overall structure does not overlap after being rotated 180°, wherein n≥1;

[0015] The spectral unit includes 2m different amorphous silicon nanostructures, which are arranged in an array, wherein m ≥ 1.

[0016] Furthermore, the superpixel includes four amorphous silicon square blocks arranged in an array. The side lengths of the four amorphous silicon square blocks are 0.26μm, 0.26μm, 0.31μm and 0.31μm, respectively, and the height of each block is 0.2μm. The center-to-center spacing of the four amorphous silicon square blocks is 0.4μm.

[0017] Furthermore, the spectral unit includes four different amorphous silicon nanostructures. All four amorphous silicon nanostructures are cylindrical structures with the same operating wavelength, ranging from 1000 nm to 2000 nm, and a height of 1200 nm.

[0018] The four types of amorphous silicon nanostructures have different structural periods and duty cycles, with structural periods ranging from 1000nm to 1800nm ​​and duty cycles ranging from 0.2 to 0.8.

[0019] Furthermore, the arrangement period of the N superpixels is 0.8 μm.

[0020] Furthermore, the high Q-value metasurface also includes a metasurface substrate, on which N superpixels are periodically arranged;

[0021] The miniaturized spectrometer also includes a spectral substrate, on which M spectral units are periodically arranged;

[0022] The metasurface substrate is a quartz substrate or a silicon dioxide substrate, and the spectral substrate is a quartz substrate or a silicon dioxide substrate.

[0023] Furthermore, the neural network is a decoupled neural network or a multilayer fully connected network.

[0024] This invention also provides a method for measuring the refractive index of a fluid, employing the aforementioned compact fluid refractive index measurement system, characterized by the following steps:

[0025] Step 1: Train the neural network of the data processing module;

[0026] Step 2: Inject the fluid to be tested into the microfluidic cavity to bring it into contact with the high-Q metasurface; then turn on the light source to emit a measurement beam that is incident on the high-Q metasurface;

[0027] Step 3: After the measurement beam passes through the fluid under test and the high Q metasurface, the fluid under test causes the high Q metasurface to resonate with a spectral shift, forming transmitted light carrying the refractive index information of the fluid under test, which is then incident on the miniaturized spectrometer.

[0028] Step 4: The miniaturized spectrometer detects the spectral information of the transmitted light and converts it into light intensity information, which is then detected by the detector.

[0029] Step 5: The detector sends the detected light intensity information to the data processing module. The neural network trained in Step 1 performs spectral inversion based on the light intensity information to obtain the spectral curve. Then, the corresponding refractive index is determined based on the spectral curve to complete the refractive index measurement of the fluid under test.

[0030] Furthermore, in step 5, the spectral curve is obtained using the following formula:

[0031] I=Tf+δ I

[0032] Where I represents light intensity information; T represents the transmission coefficient of the miniaturized spectrometer; f represents the spectral curve; δ I The error term is determined through statistics on actual noise sources.

[0033] Furthermore, step 1 specifically includes:

[0034] Step 1.1: Collect the resonance spectral curves of fluids with different known refractive indices on high Q-value metasurfaces and use them as tag data;

[0035] Step 1.2: Use a miniaturized spectrometer to obtain the light intensity information corresponding to the resonance spectrum curves of fluids with different known refractive indices on high Q-value metasurfaces, and use it as input data;

[0036] Step 1.3: Divide the input data and corresponding label data into training set and test set, and then input the training set into the neural network;

[0037] Step 1.4: The neural network performs spectral inversion on a set of input data in the training set to obtain the predicted spectral curve. Then, it calculates the mean square error between the predicted spectral curve and the label data corresponding to the input data to obtain the loss function.

[0038] Step 1.5: Optimize the parameters of the neural network through backpropagation based on the loss function, then return to step 1.4 and train the neural network using the next set of input data in the training set until the preset training rounds are reached.

[0039] Step 1.6: Input the test set into the neural network to obtain the test spectrum curve, and calculate the error of the test spectrum curve. If the error of the test spectrum curve meets the preset accuracy requirements, the training of the neural network is completed; otherwise, adjust the framework and / or parameters of the neural network, and then input the training set into the neural network and return to step 1.4.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] 1. The present invention provides a compact fluid refractive index measurement system based on a double-layer metasurface. Utilizing a high-Q metasurface with a continuous-domain bound state, refractive index information is converted into spectral information. The narrow linewidth and near-field enhancement characteristics increase the interaction intensity between the measurement beam and the fluid under test, thereby enhancing spectral response sensitivity and improving measurement accuracy. Furthermore, a miniaturized spectrometer is used to convert spectral information into intuitive light intensity information. Through analysis and processing of this light intensity information, the refractive index of the fluid under test can be quickly obtained, simplifying the measurement process and improving measurement efficiency.

[0042] 2. The present invention provides a compact fluid refractive index measurement system, which adopts a miniaturized spectrometer, significantly reducing the system size and realizing the compact design of the spectrometer. It can be integrated into portable measurement devices to meet the needs of on-site measurement, bedside measurement and other scenarios.

[0043] 3. The present invention provides a compact fluid refractive index measurement system, which uses a microfluidic cavity to provide a stable measurement environment for the fluid to be measured, ensuring the controllability and accuracy of the measurement process;

[0044] 4. The present invention provides a compact fluid refractive index measurement system in which the overall structure of the superpixels in the high Q-value metasurface does not overlap after rotating 180°, which can realize the leakage excitation of the quasi-BIC (Bound State un Continuum) resonance mode, thereby improving the interaction intensity between the measurement beam and the fluid under test, enhancing the resolution of small spectral shifts, and further enhancing the spectral response sensitivity.

[0045] 5. The present invention provides a compact fluid refractive index measurement system. The parameter design of the amorphous silicon square block in the high Q metasurface can enable the high Q metasurface to form a narrow linewidth resonance peak in its working band, thereby improving the spectral response sensitivity to refractive index changes.

[0046] 6. The present invention provides a compact fluid refractive index measurement system. The parameter design of the amorphous silicon nanostructure in the miniaturized spectrometer can ensure efficient modulation and broadband tuning of the light field within the working wavelength range of the miniaturized spectrometer.

[0047] 7. The fluid refractive index measurement method provided by this invention introduces an error term in the spectral inversion process. The distribution characteristics of the error term match the actual noise source, providing more realistic light intensity information for spectral inversion based on the detected light intensity information, realizing accurate inversion of spectral information, thereby improving the accuracy of refractive index measurement. Attached Figure Description

[0048] Figure 1 This is a system structure diagram of an embodiment of the present invention (microfluidic cavity not shown).

[0049] Figure 2 This is a schematic diagram of the microfluidic cavity structure in an embodiment of the present invention;

[0050] Figure 3 This is a schematic diagram of the structure of the high-Q metasurface in an embodiment of the present invention;

[0051] Figure 4 This is a schematic diagram of the superpixel structure of the high Q-value metasurface in an embodiment of the present invention;

[0052] Figure 5 This is a schematic diagram of the miniaturized spectrometer in an embodiment of the present invention;

[0053] Figure 6 This is a schematic diagram of the amorphous silicon nanostructure of the miniaturized spectrometer in an embodiment of the present invention;

[0054] Figure 7 The images show transmission spectrum curves of high Q-value metasurfaces in contact with fluids of different refractive indices in embodiments of the present invention, wherein the fluid refractive indices n in (a), (b), (c), and (d) are 1.0402, 1.46231, 1.67337, and 1.94772, respectively.

[0055] Figure 8 The images show the transmission spectrum curves of four amorphous silicon nanostructures in the miniaturized spectrometer of this invention, where (a), (b), (c), and (d) are the transmission spectrum curves of four different amorphous silicon nanostructures, respectively.

[0056] Figure 9 This is a schematic diagram illustrating the working principle of an embodiment of the present invention.

[0057] The annotations in the attached figures are explained as follows:

[0058] 1-Light source, 2-High Q metasurface, 3-Miniaturized spectrometer, 4-Detector, 5-Data processing module, 6-Microfluidic cavity, 7-Inlet, 8-Outlet, 9-Superpixel, 10-Spectral unit. Detailed Implementation

[0059] The present invention provides a compact fluid refractive index measurement system and method in further detail below with reference to the accompanying drawings and specific embodiments. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the invention and are not intended to limit the scope of protection of the invention.

[0060] A compact fluid refractive index measurement system, such as Figure 1As shown, the system includes a light source 1, a high-Q metasurface 2, a miniaturized spectrometer 3, and a detector 4 arranged sequentially, as well as a data processing module 5 electrically connected to the detector 4. The light source 1 is a broadband light source, capable of generating a measurement beam encompassing multiple wavelengths within its spectral range. The high-Q metasurface 2 is encapsulated within a microfluidic cavity 6, as shown... Figure 2 As shown, the microfluidic cavity 6 has an inlet 7 and an outlet 8 for injecting and discharging the fluid to be tested, respectively. The microfluidic cavity 6 is used to bring the fluid to be tested into contact with the high-Q metasurface 2. In this embodiment, the microfluidic cavity 6 is composed of polydimethylsiloxane, and the fluid to be tested is injected into it through the inlet 7 using a syringe. Polydimethylsiloxane has good biocompatibility and sealing properties, which can prevent fluid leakage. At the same time, its low optical absorption characteristics do not affect the transmission of infrared light, ensuring the stability of the optical signal during the measurement process.

[0061] like Figure 3 As shown, the high-Q metasurface 2 employs a continuous-domain bound state, comprising a metasurface substrate and N periodically arranged superpixels 9 on the metasurface substrate. The high-Q metasurface 2 is used to convert the refractive index information of the fluid under test into spectral information, where N is an integer and N≥4. Figure 4 As shown, superpixel 9 comprises four amorphous silicon square blocks arranged in an array. Their overall structure does not overlap after being rotated 180°, allowing for leakage excitation of the quasi-BIC resonance mode and enhancing the sensitivity of the spectral response. In this embodiment, the side lengths of the four amorphous silicon square blocks are 0.26 μm, 0.26 μm, 0.31 μm, and 0.31 μm, respectively, and their thickness is 0.2 μm. The center-to-center spacing between the four amorphous silicon square blocks is 0.4 μm. Two amorphous silicon square blocks with a side length of 0.26 μm and two amorphous silicon square blocks with a side length of 0.31 μm are arranged in an upper and lower structure. The arrangement period of the N superpixels 9 in the high-Q metasurface 2 is 0.8 μm, where the arrangement period refers to the distance between the centers of two superpixels 9.

[0062] The high-Q metasurface 2, with its continuous-domain bound states, achieves a strong beam confinement effect, resulting in a narrow-linewidth optical response. This narrow-linewidth characteristic makes it more sensitive to changes in refractive index; even minute fluctuations in refractive index can induce significant spectral changes. Simultaneously, the strong confinement effect also generates a significant near-field enhancement effect, forming a high-intensity electromagnetic field region near the high-Q metasurface 2. When the analyte comes into contact with the high-Q metasurface 2, the refractive index change is amplified by the near-field enhancement effect, further improving measurement sensitivity and enabling it to meet the high-precision requirements of single-molecule measurements and ultra-low concentration biomarker measurements.

[0063] like Figure 5As shown, the miniaturized spectrometer 3 is an infrared broadband spectrometer, including a spectral substrate and M periodically arranged spectral units 10 on the spectral substrate, each matching the size of the superpixel 9. Here, M is an integer, and M≥4. The miniaturized spectrometer 3 is used to convert spectral information into light intensity information. The size matching between the superpixel 9 and the spectral units 10 ensures minimal loss during optical signal transmission, improving the accuracy of light intensity information acquisition. The spectral units 10 include four different amorphous silicon nanostructures, arranged in an array. All four amorphous silicon nanostructures are as follows... Figure 6 The cylindrical structures shown all operate at the same wavelength, ranging from 1000 nm to 2000 nm, with a height of 1200 nm. The four amorphous silicon nanostructures have different structural periods and duty cycles, with structural periods ranging from 1000 nm to 1800 nm and duty cycles ranging from 0.2 to 0.8. By adjusting the structural period and duty cycle, each amorphous silicon nanostructure exhibits a linearly independent spectral response curve, thereby achieving efficient modulation and differentiation of the spectrum. The spectral substrate is either a quartz substrate or a silicon dioxide substrate. Similarly, the metasurface substrate is either a quartz substrate or a silicon dioxide substrate. The refractive index of the quartz substrate is n = 1.46, and the thickness of the silicon dioxide substrate is 2 μm, ensuring structural stability while avoiding additional interference to the optical field.

[0064] The number of superpixels 9 in the high-Q metasurface 2 and the number of spectral units 10 in the miniaturized spectrometer 3 are not forced to be equal. This is because the structures of the high-Q metasurface 2 and the miniaturized spectrometer 3 are both periodically arranged, which means that they are both repetitive arrangements of unit structures. In essence, the response of each superpixel 9 and spectral unit 10 is the same, that is, the modulation of light is essentially a similar result.

[0065] Detector 4 includes M detection units, each corresponding to one of the M spectral units 10, used to detect light intensity information. Data processing module 5 includes a neural network used for spectral inversion based on light intensity information to obtain spectral curves, and then deriving the refractive index of the fluid under test based on the refractive indices corresponding to different spectral curves. In this embodiment, the neural network uses a decoupled neural network or a multilayer fully connected network. During training, the resonance spectral curves corresponding to fluids with different refractive indices are used as label data, and the light intensity information corresponding to the resonance spectral curves is used as input data. Through training, it can acquire the ability to obtain spectral curves through spectral inversion based on light intensity information, thereby deriving the refractive index. The training samples of the neural network cover the target measurement range of refractive index and include resonance spectral curves of fluids at different temperatures and concentrations, ensuring that it maintains high spectral inversion accuracy even in complex real-world scenarios.

[0066] This embodiment utilizes a bilayer metasurface comprised of a high-Q metasurface 2 (continuous-domain bound state) and a miniaturized spectrometer 3. These two components work together to cover a wavelength range of 1 μm-2 μm, suitable for measuring the refractive index of fluids in common biological and chemical sample ranges. The high-Q metasurface 2, with its narrow linewidth, exhibits a significant near-field enhancement effect, thereby increasing the interaction intensity between the measurement beam and the analyte. Stronger interaction results in higher measurement sensitivity. The core functional unit of the miniaturized spectrometer 3 is composed of amorphous silicon nanostructures. By selecting various nanostructures with linearly independent spectral response curves, the miniaturized spectrometer 3 achieves efficient modulation and broadband tuning of the light field within the 1 μm-2 μm wavelength range. Based on the narrow linewidth characteristic of the high-Q metasurface 2, this embodiment maps the fluid refractive index into spectral information, which is then decoded by the miniaturized spectrometer 3 to obtain light intensity information, thus enabling precise measurement of the fluid refractive index.

[0067] like Figure 7 As shown, the transmission spectrum curves of high-Q metasurface 2 under different refractive index fluid environments are displayed. It can be seen that as the refractive index n increases, the resonance peak wavelength shifts towards longer wavelengths, and the peak width becomes narrower, reflecting the high-Q characteristics. Figure 8 As shown, the transmission spectral curves of four amorphous silicon nanostructures of the miniaturized spectrometer 3 are displayed. It can be seen that each amorphous silicon nanostructure has a significant difference in transmittance for different wavelengths of light, and can achieve a linearly independent spectral response. Furthermore, the spectral characteristics can be accurately reflected through light intensity information, which verifies the effectiveness of the miniaturized spectrometer 3 in converting spectral information into light intensity information.

[0068] This embodiment constructs a compact fluid refractive index measurement system by combining a high-Q metasurface 2 in a continuous domain bound state, a miniaturized infrared broadband spectrometer 3, a microfluidic cavity 6, and a neural network. The fluid to be measured is injected into the microfluidic cavity 6, contacting the high-Q metasurface 2 within it. Changes in the refractive index of the fluid cause a shift in the resonance spectral curve of the high-Q metasurface 2; for example... Figure 9 As shown, after the measurement beam is incident on the high-Q metasurface 2, it forms transmitted light carrying the refractive index information of the fluid to be measured. Subsequently, the miniaturized spectrometer 3 receives the transmitted light modulated by the high-Q metasurface 2 and converts it into light intensity information, which is then detected by the detector 4. Finally, the neural network performs spectral inversion based on the light intensity information to obtain the corresponding spectral curve. According to the preset correspondence between the spectral curve and the refractive index, the precise measurement of the fluid's refractive index is finally achieved. This embodiment combines the high-Q metasurface 2 with the miniaturized spectrometer 3, which not only fully utilizes the high sensitivity advantage of the high-Q metasurface 2, but also leverages the compactness and efficient information processing capabilities of the miniaturized spectrometer 3 to promote the development of refractive index measurement systems towards portability and intelligence.

[0069] This embodiment also provides a method for measuring the refractive index of a fluid, using the aforementioned compact fluid refractive index measurement system, and includes the following steps:

[0070] Step 1: Train the neural network of data processing module 5. Specifically:

[0071] Step 1.1: Collect resonance spectral curves of fluids with different known refractive indices on high-Q metasurface 2 and use them as tag data;

[0072] Step 1.2: Use miniaturized spectrometer 3 to obtain the light intensity information corresponding to the resonance spectrum curves of fluids with different known refractive indices on high Q-value metasurfaces, and use it as input data;

[0073] Step 1.3: Divide the input data and corresponding label data into training set and test set, and then input the training set into the neural network;

[0074] Step 1.4: The neural network performs spectral inversion on a set of input data in the training set to obtain the predicted spectral curve. Then, it calculates the mean square error between the predicted spectral curve and the label data corresponding to the input data to obtain the loss function.

[0075] Step 1.5: Optimize the parameters of the neural network through backpropagation based on the loss function, then return to step 1.4 and train the neural network using the next set of input data in the training set until the preset training rounds are reached.

[0076] Step 1.6: Input the test set into the neural network to obtain the test spectrum curve, and calculate the error of the test spectrum curve. If the error of the test spectrum curve meets the preset accuracy requirements, the training of the neural network is completed; otherwise, adjust the framework and / or parameters of the neural network, and then input the training set into the neural network and return to step 1.4.

[0077] Step 2: Inject the fluid to be tested into the microfluidic cavity 6 so that it comes into contact with the high Q metasurface 2; then turn on the light source 1 and emit a measurement beam that is incident on the high Q metasurface 2.

[0078] Step 3: After the measurement beam passes through the fluid to be tested and the high-Q metasurface 2, the fluid to be tested causes the high-Q metasurface 2 to produce a resonant spectral shift, forming transmitted light carrying the refractive index information of the fluid to be tested, which is then incident on the miniaturized spectrometer 3.

[0079] Step 4: The miniaturized spectrometer 3 detects the spectral information of the transmitted light and converts it into light intensity information, which is then detected by the detector 4.

[0080] Step 5: Detector 4 sends the detected light intensity information to data processing module 5. The neural network trained in step 1 performs spectral inversion based on the light intensity information to obtain the spectral curve. Then, the corresponding refractive index is determined based on the spectral curve, completing the refractive index measurement of the fluid under test. The spectral curve is obtained using the following formula:

[0081] I=Tf+δ I

[0082] Where I represents light intensity information; T represents the transmission coefficient of the miniaturized spectrometer 3; f represents spectral information; δ I The error term is determined through statistics on actual noise sources.

[0083] The following mathematical model explains the spectral encoding process of the high-Q metasurface 2. The encoding of wavelength information ensures that the light intensity received by detector 4 is the integral of the entire spectrum. Therefore, the transmission intensity I received by the i-th detection unit of detector 4 is... i It can be represented as:

[0084]

[0085] in, , These represent the maximum and minimum wavelengths in the measured beam, respectively. Let M be the transmission coefficient of the i-th spectral unit of the miniaturized spectrometer 3 at wavelength λ, where i is an integer and 0 < i ≤ M. This represents the spectral information of the i-th spectral unit at wavelength λ.

[0086] Then we have:

[0087]

[0088]

[0089]

[0090]

[0091] Where k is the number of wavelength sampling points in the measurement beam, λ1-λ k These represent the wavelengths of the beams at the k wavelength sampling points; , These represent the transmission coefficients of the first spectral unit in the miniaturized spectrometer 3 to the first and kth wavelength sampling points at wavelength λ1, respectively. , These represent the Nth spectral unit in the miniaturized spectrometer 3 at wavelength λ. k The transmission coefficient of the beam at the first and kth wavelength sampling points.

[0092] The light intensity information can be obtained by calculating the transmission coefficient of the miniaturized spectrometer 3 and the incident spectral information, i.e., I=Tf. The transmission coefficient T of the miniaturized spectrometer 3 characterizes the modulation characteristics of each spectral unit in the miniaturized spectrometer 3 for different wavelengths of light.

[0093] However, several uncertainties exist in the actual measurement process: First, the pixel response of detector 4 is non-uniform, causing deviations in the signal response of different pixels to the same light intensity; second, the introduction of random noise such as ambient light interference and electronic noise causes the intensity signal to contain additional fluctuations. To more realistically reflect the actual measurement scenario and improve the accuracy of subsequent spectral inversion, this embodiment introduces an error term δ into the light intensity information I. I That is, I = Tf + δ I The error term δ I The amplitude is determined by statistical methods, and its distribution characteristics match the actual noise source, thus providing a more realistic input for the spectral inversion algorithm based on measurement data, and ultimately achieving accurate inversion of spectral information.

Claims

1. A compact fluid refractive index measurement system, characterized in that: It includes a light source (1), a high Q-value metasurface (2), a miniaturized spectrometer (3) and a detector (4) arranged in sequence, and a data processing module (5) electrically connected to the detector (4); The high Q-value metasurface (2) is encapsulated in a microfluidic cavity (6). The microfluidic cavity (6) is provided with an inlet (7) and an outlet (8) for injecting and discharging the fluid to be tested, respectively. The microfluidic cavity (6) is used to make the fluid to be tested contact with the high Q-value metasurface (2). The high-Q metasurface (2) adopts a continuous domain bound state and includes N periodically arranged superpixels (9). The high-Q metasurface (2) is used to convert the refractive index information of the fluid to be measured into spectral information, where N is an integer and N≥4; the arrangement period of the N superpixels (9) is 0.8μm, and each superpixel (9) includes 4 amorphous silicon square blocks arranged in an array. Their overall structure does not overlap after rotating 180°; the side lengths of the 4 amorphous silicon square blocks are 0.26μm, 0.26μm, 0.31μm and 0.31μm respectively, and the height is 0.2μm. The spacing between the centers of the 4 amorphous silicon square blocks is 0.4μm. The miniaturized spectrometer (3) includes M periodically arranged spectral units (10) that match the size of the superpixel (9) for converting spectral information into light intensity information, where M is an integer and M≥4; the spectral unit (10) includes 4 different amorphous silicon nanostructures, which are arranged in an array and are all cylindrical structures with a height of 1200nm; the 4 amorphous silicon nanostructures have the same operating wavelength, ranging from 1000nm to 2000nm, but different structural periods and duty cycles, with structural periods ranging from 1000nm to 1800nm ​​and duty cycles ranging from 0.2 to 0.8; The detector (4) includes M detection units respectively corresponding to M spectral units (10) for detecting light intensity information; The data processing module (5) includes a neural network, which is used to perform spectral inversion based on light intensity information to obtain a spectral curve, thereby deriving the refractive index of the fluid to be tested.

2. The compact fluid refractive index measurement system according to claim 1, characterized in that: The high Q-value metasurface (2) also includes a metasurface substrate, and N superpixels (9) are periodically arranged on the metasurface substrate; The miniaturized spectrometer (3) also includes a spectral substrate, on which M spectral units (10) are periodically arranged; The metasurface substrate is a quartz substrate or a silicon dioxide substrate, and the spectral substrate is a quartz substrate or a silicon dioxide substrate.

3. The compact fluid refractive index measurement system according to claim 2, characterized in that: The neural network is a decoupled neural network or a multilayer fully connected network.

4. A method for measuring the refractive index of a fluid, employing a compact fluid refractive index measurement system as described in any one of claims 1-3, characterized in that, Includes the following steps: Step 1: Train the neural network of the data processing module (5); Step 2: Inject the fluid to be tested into the microfluidic cavity (6) so that it comes into contact with the high Q metasurface (2); then turn on the light source (1) and emit a measurement beam that is incident on the high Q metasurface (2). Step 3: After the measurement beam passes through the fluid to be measured and the high Q value metasurface (2), the fluid to be measured causes the high Q value metasurface (2) to generate a resonant spectral shift, forming transmitted light carrying the refractive index information of the fluid to be measured, which is then incident on the miniaturized spectrometer (3). Step 4: The miniaturized spectrometer (3) detects the spectral information of the transmitted light and converts it into light intensity information, which is then detected by the detector (4). Step 5: The detector (4) sends the detected light intensity information to the data processing module (5). The neural network trained in step 1 performs spectral inversion based on the light intensity information to obtain the spectral curve. Then, it determines the corresponding refractive index based on the spectral curve to complete the refractive index measurement of the fluid to be tested.

5. The fluid refractive index measurement method according to claim 4, characterized in that, In step 5, the spectral curve is obtained using the following formula: I = Tf + δ I in, I For light intensity information; T The transmittance coefficient of the miniaturized spectrometer (3); f The spectrum curve; δ I The error term is determined through statistics on actual noise sources.

6. A method for measuring the refractive index of a fluid according to claim 4 or 5, characterized in that, Step 1 is as follows: Step 1.1: Collect the resonance spectral curves of fluids with different known refractive indices on the high Q-value metasurface (2) and use them as tag data; Step 1.2: Use a miniaturized spectrometer (3) to obtain the light intensity information corresponding to the resonance spectrum curves of fluids with different known refractive indices on the high Q value metasurface (2), and use it as input data; Step 1.3: Divide the input data and corresponding label data into training set and test set, and then input the training set into the neural network; Step 1.4: The neural network performs spectral inversion on a set of input data in the training set to obtain the predicted spectral curve. Then, it calculates the mean square error between the predicted spectral curve and the label data corresponding to the input data to obtain the loss function. Step 1.5: Optimize the parameters of the neural network through backpropagation based on the loss function, then return to step 1.4 and train the neural network using the next set of input data in the training set until the preset training rounds are reached. Step 1.6: Input the test set into the neural network to obtain the test spectrum curve, and calculate the error of the test spectrum curve. If the error of the test spectrum curve meets the preset accuracy requirements, the training of the neural network is completed; otherwise, adjust the framework and / or parameters of the neural network, and then input the training set into the neural network and return to step 1.4.

Images