A dual-wavelength velocity measurement device, a velocity measurement method and a manufacturing method of an on-chip integrated optical waveguide and graphene

Abstract

This invention discloses an on-chip integrated optical waveguide and graphene dual-wavelength velocimetry device, velocimetry method, and manufacturing method. The wavelength velocimetry device includes a silicon wafer substrate, a lower cladding layer disposed on the silicon wafer substrate, a waveguide core layer disposed on the lower cladding layer, and an upper cladding layer disposed on the waveguide core layer. The waveguide core layer includes an input / output waveguide, an optical beamsplitter, and an interferometer arm. A metal reflective layer is disposed at the end of the interferometer arm, which includes a sensing arm and a reference arm. The sensing arm and the reference arm are symmetrically arranged, and the surface of the sensing arm is covered with graphene. The output end of the input / output waveguide is connected to the input end of the optical beamsplitter, and the two output ends of the optical beamsplitter are respectively connected to the reference arm and the sensing arm of the interferometer arm.

Description

Technical Field

[0001] This invention relates to the field of optical sensing and micro / nano electromechanical systems technology, specifically to a dual-wavelength velocities device, velocities measurement method, and manufacturing method that integrates on-chip optical waveguides and graphene. Background Technology

[0002] In modern precision measurement, intelligent equipment, and micro / nano systems, the ability to sense motion speed in real time with high accuracy is crucial. Traditional speed measurement methods mainly fall into two categories: mechanical contact methods (such as tachogenerators and turbine flow meters) and non-contact optical methods (such as laser Doppler velocimetry and particle image velocimetry, PIV). Mechanical methods are limited by issues such as friction and wear, response hysteresis, and short lifespan, making them unsuitable for miniaturization or clean environments. While classic optical velocimetry technology has the advantage of being non-contact, it typically relies on discrete optical components (such as lenses, mirrors, and interferometers), resulting in large size, complex alignment, and susceptibility to vibration and environmental disturbances. These limitations make it difficult to meet the urgent needs of current fields such as micro-aircraft, wearable devices, and semiconductor manufacturing equipment for miniaturized, low-power, and highly integrated sensors.

[0003] In recent years, with the rapid development of silicon-based photonics and microelectromechanical systems (MEMS) technology, on-chip integrated optical sensors have become a research hotspot. Among them, waveguide sensors based on the Mach-Zehnder (MZI) or Michelson interferometry principle are widely used for the detection of physical quantities such as refractive index, stress, temperature and acceleration due to their high sensitivity, resistance to electromagnetic interference and CMOS process compatibility.

[0004] However, existing on-chip sensors still face significant challenges in velocity measurement, particularly in detecting relative motion velocities induced by airflow. Most on-chip interferometric velocimetry schemes rely on direct modulation of optical path difference by mechanical displacement (e.g., the movement of a MEMS cantilever beam), requiring complex driving or sensing structures and exhibiting extremely weak response to non-contact wind fields. Regarding materials applications, while graphene possesses ultra-high carrier mobility, broadband light absorption, and high sensitivity to surface stress, temperature, and charge transfer, its optical response under wind-induced cooling or aerodynamic stress modulation has not been effectively coupled into interferometric phase detection systems. Furthermore, existing on-chip sensors generally operate in a single-wavelength mode, and applications combining on-chip interferometric velocimetry with two-dimensional materials have not yet been reported.

[0005] Therefore, there is an urgent need for a novel on-chip velocimetry solution that integrates graphene-sensitive materials, dual-wavelength synergistic excitation, and interferometric phase detection to simultaneously achieve high sensitivity, strong anti-interference capability, and system miniaturization. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a dual-wavelength velocimetry device that integrates optical waveguide and graphene on a chip. The dual-wavelength velocimetry device can effectively improve the velocimetry accuracy and significantly enhance the anti-interference capability of the system.

[0007] The second objective of this invention is to provide a dual-wavelength velocimetry method integrating on-chip optical waveguides and graphene.

[0008] The third objective of this invention is to provide a method for manufacturing a dual-wavelength velocimetry device that integrates an on-chip optical waveguide and graphene.

[0009] The technical solution of the present invention to solve the above-mentioned technical problems is:

[0010] A dual-wavelength velocimetry device integrating an on-chip optical waveguide and graphene includes a silicon substrate, a lower cladding layer disposed on the silicon substrate, a waveguide core layer disposed on the lower cladding layer, and an upper cladding layer covering the waveguide core layer and the graphene surface. The waveguide core layer includes an input / output waveguide, an optical beamsplitter, and an interferometer arm. A metal reflective layer is disposed at the end of the interferometer arm, which includes a sensing arm and a reference arm. The sensing arm and the reference arm are symmetrically arranged, and the surface of the sensing arm is covered with graphene. The output end of the input / output waveguide is connected to the input end of the optical beamsplitter, and the two output ends of the optical beamsplitter are respectively connected to the reference arm and the sensing arm of the interferometer arm.

[0011] Preferably, the input / output waveguide is used to couple a 980nm light source and a 1550nm light source, wherein the 980nm light source is used to photothermally excite the graphene on the surface of the sensing arm, placing the graphene in a preset thermal bias state to enhance its sensitivity to wind speed; the 1550nm light source is used to generate an interference light intensity signal to achieve speed measurement; the optical beam splitter is used to distribute the beams from the two light sources to the reference arm and the sensing arm respectively; the metal reflective layer is used to reflect the beams to form a dual-path reflective Michelson interferometer; and the graphene is used to sense the motion of the moving object being measured. The relative wind speed, in turn, generates temperature and stress changes to modulate the phase of the light beam. The temperature change changes through the thermo-optic effect, and the stress change changes through the photoelastic effect, together changing the refractive index of the graphene, thereby modulating the phase of the light beam transmitted in the sensing arm, causing a phase difference between the light beams of the reference arm and the sensing arm. The light beam from the 1550nm light source is distributed by the optical beam splitter and transmitted along the reference arm and the sensing arm respectively. After being reflected by the metal reflective layer at the end of the reference arm and the sensing arm, the two reflected light beams merge again, and interference occurs due to the phase difference, forming an interference light intensity signal.

[0012] Preferably, the reference arm of the interferometer arm is composed of a first curved section and a first straight section; one end of the first curved section is connected to the optical beam splitter, and the other end is connected to one end of the first straight section; the other end of the first straight section is provided with a metal reflective layer.

[0013] Preferably, the sensing arm of the interferometer arm is composed of a second curved section and a second straight section; one end of the second curved section is connected to the optical beam splitter, and the other end is connected to one end of the second straight section; the other end of the second straight section is provided with a metal reflective layer.

[0014] Preferably, the first straight line segment and the second straight line segment are arranged in parallel.

[0015] A dual-wavelength velocimetry method integrating on-chip optical waveguides and graphene includes the following steps:

[0016] S1: The dual-wavelength speed measuring device is fixedly installed on the surface of the moving object to be measured. The installation direction needs to ensure that the graphene on the surface of the sensing arm is perpendicular to the direction of movement of the moving object to maximize the perception of relative wind speed.

[0017] S2: A 980nm light source and a 1550nm light source are respectively coupled into the input / output waveguides of the dual-wavelength velocimetry device. The 980nm light source is used to photothermally excite the graphene on the surface of the sensing arm, placing the graphene in a preset thermal bias state to enhance its sensitivity to wind speed. When the moving object generates relative wind speed, the graphene on the surface of the sensing arm senses the wind speed and generates temperature and stress changes, thereby modulating the phase of the beam transmitted in the sensing arm. The 1550nm light source is used to form an interference light intensity signal to achieve speed measurement. The beam of the 1550nm light source is distributed by an optical beam splitter and transmitted along the reference arm and the sensing arm respectively. After being reflected by the metal reflective layer at the end of the reference arm and the sensing arm, a dual-path reflection Michelson interferometer is formed. The 980nm beam is distributed by an optical beam splitter and transmitted only along the sensing arm.

[0018] S3: A photodetector is used to synchronously acquire the interference output light intensity signal corresponding to a wavelength of 1550nm, forming an interference light intensity time series. After preprocessing the acquired interference light intensity time series, the preprocessed interference light intensity time series is input into a pre-trained Mamba model. The Mamba model, based on the state space equation, models the long-range time dependence of the interference light intensity signal, extracts the phase change and frequency domain features in the interference output light intensity signal, and mines the interference dynamic features. The phase change and frequency domain features in the interference dynamic features are positively correlated with the velocity of the moving object under test. At the same time, the state transition relationship of the state space equation is used to recursively update the target velocity and position state, and output the velocity estimate of the moving object under test.

[0019] S4: Construct a speed correction model. During actual speed measurement, perform error compensation and calibration on the speed estimate based on the speed correction model to obtain the speed prediction value.

[0020] Preferably, in step S3, the specific process of the preprocessing is as follows: the time window is divided, the amplitude is normalized and differentially processed on the time series of the collected 1550nm interference output light intensity to form a feature sequence containing information on the change of interference light intensity.

[0021] Preferably, in step S3, the training process of the Mamba model is as follows: based on a standard wind tunnel platform, time series samples of interference light intensity corresponding to different known speeds are obtained. The time series samples of interference light intensity are divided into a training set and a test set. The state transition matrix, Kalman filter parameters and feature extraction weights inside the Mamba model are calibrated using the training set. The accuracy of the Mamba model is verified and optimized using the test set until the prediction error of the Mamba model meets a preset threshold. When the model test error is lower than the preset threshold, the training of the Mamba model is determined to be complete.

[0022] Preferably, in step S4, the speed correction model is constructed as follows: the true wind speed values ​​under different wind speeds and different airflow directions are obtained in advance through a standard wind tunnel platform, the speed estimates of the Mamba model under the corresponding working conditions are collected simultaneously, and the mapping correction relationship between the speed estimates and the true speed values ​​is established by using the least squares method, thereby obtaining the speed correction model.

[0023] Preferably, the method further includes step S5, which involves: weighting and fusing the motion velocity value obtained by physical inversion of the interference light intensity signal with the velocity prediction value output by the Mamba model; and dynamically adjusting the weighting coefficients of the motion velocity value and the velocity prediction value based on the real-time quality of the interference light intensity signal to obtain the final velocity value.

[0024] A method for manufacturing a dual-wavelength velocimetry device integrating an on-chip optical waveguide and graphene includes the following steps:

[0025] Step 1: Select a 100-oriented silicon wafer as the silicon wafer substrate, and use plasma cleaning process to activate the surface of the silicon wafer to remove surface impurities and oxide layer. The cleaning time is controlled at 5-10 minutes, and the plasma power is set to 100-200W.

[0026] Step 2: Apply EpoCladd material uniformly to the pretreated silicon wafer surface using a spin coating process. The coating speed is 3000-5000 r / min, and the coating thickness is controlled at 1-2 μm. After coating, use a 365nm ultraviolet lamp to irradiate for 10-20s for curing to form the lower cladding layer.

[0027] Step 3: Apply SU-8 material as the waveguide core layer to the lower cladding surface using a spin coating process. The coating speed is 2000-4000 r / min, and the coating thickness is controlled at 3-5 μm. After coating, transfer the pre-designed mask patterns of the input / output waveguides, beam splitters, reference arms, and sensing arms (excluding the metal reflective layer) onto the SU-8 thin film using a photolithography machine. The exposure parameters of the photolithography machine are set as follows: exposure power 100-150 mW, exposure time 20-30 s, to ensure that the mask patterns are clearly transferred onto the SU-8 thin film.

[0028] Step 4: The waveguide height of the SU-8 thin film is finely adjusted using reactive ion etching. After etching, the ends of the reference arm and the sensor arm are naturally exposed as vertical surfaces, serving as the deposition substrate for the metal reflective layer.

[0029] Step 5: Deposit a metal film on the vertical surface of the end of the reference arm and the end of the sensor arm using evaporation or sputtering processes to form a metal reflective layer; the metal film is made of gold, silver or aluminum, and the film thickness is controlled between 50-100nm to ensure that the beam reflection efficiency is not less than 95%;

[0030] Step 6: Graphene grown by chemical vapor deposition is transferred to the surface of the waveguide core layer of the sensing arm using a wet transfer process. Then, EpoCladd material, the same as the lower cladding, is spin-coated onto the waveguide core layer and the graphene surface as an upper cladding to restrict light transmission. The spin coating speed of the upper cladding is 3000-5000 r / min, and the coating thickness is controlled at 1-2 μm. After coating, it is cured by irradiation with a 365nm wavelength, 100-150mW ultraviolet lamp for 10-20 seconds. Specifically, the graphene transfer involves transferring the CVD-grown graphene film to the surface of the waveguide core layer of the sensing arm using polymethyl methacrylate (PMMA). After transfer, PMMA is removed by dissolving in acetone, and annealing is performed in a nitrogen or argon inert atmosphere at a temperature of 150-200℃ for 10-15 minutes to ensure tight adhesion between the graphene and the waveguide core layer.

[0031] Compared with the prior art, the present invention has the following advantages:

[0032] 1. The dual-wavelength velocimetry device of this invention deeply integrates the graphene wind-induced response mechanism with an on-chip integrated Michelson interferometer, enabling highly sensitive non-contact detection of object motion speed. By employing a 980nm / 1550nm dual-wavelength collaborative working mode, the 980nm light excites the graphene to generate a thermal bias amplification of the airflow response amplitude, which, combined with the phase interference light intensity signal generated by the 1550nm light, significantly enhances the device's ability to sense weak airflow disturbances. Simultaneously, the introduction of the Mamba time-series modeling algorithm overcomes the limitations of traditional demodulation schemes, allowing the dual-wavelength velocimetry device of this invention to maintain ultra-high measurement accuracy and stability even under complex airflow conditions.

[0033] 2. The dual-wavelength velocity measurement device of the present invention adopts an on-chip integrated design, is compatible with standard semiconductor processes, and is easy to integrate with intelligent systems. It has the advantages of high sensitivity, strong anti-interference, miniaturization and easy integration, and can be widely used in scenarios such as airspeed sensing of micro UAVs, airflow monitoring in clean rooms, and motion control of precision instruments. Attached Figure Description

[0034] Figure 1 This is a schematic diagram of the dual-wavelength velocimetry device integrating on-chip optical waveguide and graphene of the present invention.

[0035] In the diagram: 1-1: Input / output waveguide; 1-2: Optical beam splitter; 1-3: Sensor; 1-4: Reference arm; 1-5: Metal reflective layer; 1-6: Graphene. Detailed Implementation

[0036] The present invention will be further described in detail below with reference to the embodiments and accompanying drawings, but the embodiments of the present invention are not limited thereto.

[0037] Example 1

[0038] See Figure 1 The on-chip integrated optical waveguide and graphene dual-wavelength velocimetry device of the present invention includes a silicon wafer substrate, a lower cladding layer disposed on the silicon wafer substrate, a waveguide core layer disposed on the lower cladding layer, and an upper cladding layer covering the waveguide core layer and the graphene surface. The waveguide core layer includes an input / output waveguide, an optical beamsplitter, and an interferometer arm. A metal reflective layer is disposed at the end of the interferometer arm, which includes a sensing arm and a reference arm. The sensing arm and the reference arm are symmetrically arranged, and the surface of the sensing arm is covered with graphene. The output end of the input / output waveguide is connected to the input end of the optical beamsplitter, and the two output ends of the optical beamsplitter are respectively connected to the reference arm and the sensing arm of the interferometer arm.

[0039] In this embodiment, the input / output waveguide is used to couple a 980nm light source and a 1550nm light source. The 980nm light source is used to photothermally excite the graphene on the surface of the sensing arm, placing the graphene in a preset thermal bias state to enhance its sensitivity to wind speed. The 1550nm light source is used to generate an interference light intensity signal to achieve speed measurement. The optical beam splitter is used to distribute the beams from the two light sources to the reference arm and the sensing arm, respectively. The metal reflective layer is used to reflect the beams to form a dual-path reflective Michelson interferometer. The graphene is used to sense the motion of the moving object being measured. The resulting relative wind speed, in turn, causes temperature and stress changes that modulate the phase of the light beam. The temperature change alters the refractive index of the graphene through the thermo-optic effect, while the stress change alters it through the photoelastic effect. This, in turn, modulates the phase of the light beam transmitted in the sensing arm, creating a phase difference between the light beams of the reference arm and the sensing arm. The light beam from the 1550nm light source is distributed by the optical beam splitter and then transmitted along the reference arm and the sensing arm, respectively. After being reflected by the metal reflective layers at the ends of the reference arm and the sensing arm, the two reflected light beams converge again. Due to the phase difference, interference occurs, forming an interference light intensity signal.

[0040] See Figure 1 The reference arm of the interferometer arm consists of a first curved section and a first straight section; one end of the first curved section is connected to the optical beam splitter, and the other end is connected to one end of the first straight section; a metal reflective layer is provided at the other end of the first straight section; wherein, the bending angle and bending length of the first curved section and the length of the first straight section can be flexibly selected according to the actual situation, for example, the bending angle of the first curved section is 90°-180°, the bending length is 1000-1500μm; and the length of the first straight section is 20000-25000μm.

[0041] See Figure 1 The sensing arm of the interferometer arm consists of a second curved section and a second straight section; one end of the second curved section is connected to the optical beam splitter, and the other end is connected to one end of the second straight section; a metal reflective layer is provided at the other end of the second straight section; wherein, the bending angle and bending length of the second curved section and the length of the second straight section can be flexibly selected according to the actual situation, for example, the bending angle of the second curved section is 90°-180° and the length is 1000-1500μm; the length of the second straight section is 20000-25000μm.

[0042] See Figure 1The first and second straight segments are arranged in parallel, with the distance between them controlled at 50-100μm. The parallel arrangement of the first and second straight segments can ensure that the optical path of the two reflected beams is consistent, guaranteeing the stability of the interference effect, and at the same time facilitating the graphene layer to uniformly sense the relative wind speed.

[0043] See Figure 1 The dual-wavelength velocimetry method of on-chip integrated optical waveguide and graphene of the present invention includes the following steps:

[0044] S1: The dual-wavelength speed measuring device is fixedly installed on the surface of the moving object to be measured. The installation direction needs to ensure that the graphene on the surface of the sensing arm is perpendicular to the direction of movement of the moving object to maximize the perception of relative wind speed.

[0045] S2: A 980nm light source and a 1550nm light source are respectively coupled into the input / output waveguides of the dual-wavelength velocimetry device. The 980nm light source is used to photothermally excite the graphene on the surface of the sensing arm, placing the graphene in a preset thermal bias state to enhance its sensitivity to wind speed. When the moving object generates relative wind speed, the graphene on the surface of the sensing arm senses the wind speed and generates temperature and stress changes, thereby modulating the phase of the beam transmitted in the sensing arm. The 1550nm light source is used to form an interference light intensity signal to achieve speed measurement. The beam of the 1550nm light source is distributed by an optical beam splitter and transmitted along the reference arm and the sensing arm respectively. After being reflected by the metal reflective layer at the end of the reference arm and the sensing arm, a dual-path reflection Michelson interferometer is formed. The 980nm beam is distributed by an optical beam splitter and transmitted only along the sensing arm.

[0046] S3: A photodetector is used to synchronously acquire the interference output light intensity signal corresponding to a wavelength of 1550nm, forming an interference light intensity time series. After preprocessing the acquired interference light intensity time series, the preprocessed interference light intensity time series is input into a pre-trained Mamba model. The Mamba model, based on the state space equation, models the long-range time dependence of the interference light intensity signal, extracts the phase change and frequency domain features in the interference output light intensity signal, and mines the interference dynamic features. The phase change and frequency domain features in the interference dynamic features are positively correlated with the velocity of the moving object under test. At the same time, the state transition relationship of the state space equation is used to recursively update the target velocity and position state, and output the velocity estimate of the moving object under test.

[0047] In this embodiment, the specific preprocessing process is as follows: the time sequence of the acquired 1550nm interference output light intensity is divided into time windows, normalized in amplitude, and differentially processed to form a feature sequence containing information on the change in interference light intensity; the time window is divided into a sliding window with a window size of 10-100ms; the differential processing uses first-order difference to eliminate DC offset of the signal; the Mamba model is modeled through a state-space model, with the target velocity and position as the core state variables, to mine the long-range time dependence in the feature sequence, suppress phase disturbances caused by airflow turbulence, environmental noise, and thermal drift, and improve the accuracy of velocity estimation; the amplitude normalization adopts the min-max normalization method.

[0048] Furthermore, the Mamba model can efficiently model long-term time series, capturing the implicit states highly correlated with velocity changes in the interference light intensity signal with low computational complexity. This enables enhanced extraction of effective phase information and suppression of random noise, significantly improving velocity measurement accuracy and anti-interference capabilities. The training process of the Mamba model is as follows: Interference light intensity time series samples corresponding to different known velocities are obtained based on a standard wind tunnel platform. These samples are divided into training and testing sets. The state transition matrix, Kalman filter parameters, and feature extraction weights within the Mamba model are calibrated using the training set. The accuracy of the Mamba model is verified and optimized using the testing set until the prediction error of the Mamba model meets a preset threshold. The preset threshold is a velocity prediction error ≤ 5%. When the model's test error is lower than this preset threshold, the Mamba model training is considered complete. Specifically, the state-space equation of the Mamba model is:

[0049] ;

[0050] In the formula: The hidden state vector at time K is used to characterize the dynamic temporal characteristics of the interference light intensity signal; The input is the time-series data of the interference light intensity; The output velocity is the estimated value of the model; A, B, C, and D are learnable parameter matrices, which are obtained through offline training and optimization using wind tunnel calibration samples. For system process noise, This is light intensity observation noise, used to characterize environmental disturbances and acquisition errors in actual detection scenarios.

[0051] S4: Construct a speed correction model. During actual speed measurement, perform error compensation and calibration on the speed estimate based on the speed correction model to obtain the final speed prediction value. .

[0052] In this embodiment, the velocity correction model is constructed as follows: the true wind speed values ​​under different motion conditions are obtained in advance through a standard wind tunnel platform, the velocity estimates of the Mamba model under the corresponding conditions are collected simultaneously, and the least squares method is used to establish the mapping correction relationship between the model estimates and the true values, thereby obtaining the velocity correction model; the different motion conditions include different wind speeds and different airflow directions, covering the actual motion range of the object under test, ensuring the universality of the correction model; the different wind speed range is 0-10 m / s, and the different airflow directions are 0°-90° to the graphene surface.

[0053] See Figure 1 The manufacturing method of the dual-wavelength velocimetry device integrating on-chip optical waveguide and graphene of the present invention includes the following steps:

[0054] Step 1: Select a 100-oriented silicon wafer as the silicon wafer substrate, and use plasma cleaning process to activate the surface of the silicon wafer to remove surface impurities and oxide layer. The cleaning time is controlled at 5-10 minutes, and the plasma power is set to 100-200W.

[0055] Step 2: Apply EpoCladd material uniformly to the pretreated silicon wafer surface using a spin coating process. The coating speed is 3000-5000 r / min, and the coating thickness is controlled at 1-2 μm. After coating, use a 365nm ultraviolet lamp to irradiate for 10-20s for curing to form the lower cladding layer.

[0056] Step 3: Apply SU-8 material as the waveguide core layer to the lower cladding surface using a spin coating process. The coating speed is 2000-4000 r / min, and the coating thickness is controlled at 3-5 μm. After coating, transfer the pre-designed mask patterns of the input / output waveguides, beam splitters, reference arms, and sensing arms (excluding the metal reflective layer) onto the SU-8 thin film using a photolithography machine. The exposure parameters of the photolithography machine are set as follows: exposure power 100-150 mW, exposure time 20-30 s, to ensure that the mask patterns are clearly transferred onto the SU-8 thin film.

[0057] Step 4: The waveguide height of the SU-8 thin film is finely adjusted using reactive ion etching. After etching, the ends of the reference arm and the sensor arm are naturally exposed as vertical surfaces, serving as the deposition substrate for the metal reflective layer.

[0058] Step 5: Deposit a metal film on the vertical surface of the end of the reference arm and the end of the sensor arm using evaporation or sputtering processes to form a metal reflective layer; the metal film is made of gold, silver or aluminum, and the film thickness is controlled between 50-100nm to ensure that the beam reflection efficiency is not less than 95%;

[0059] Step 6: Graphene grown by chemical vapor deposition is transferred to the surface of the waveguide core layer of the sensing arm using a wet transfer process. Then, EpoCladd material, the same as the lower cladding, is spin-coated onto the waveguide core layer and the graphene surface as an upper cladding to restrict light transmission. The spin coating speed of the upper cladding is 3000-5000 r / min, and the coating thickness is controlled at 1-2 μm. After coating, it is cured by irradiation with a 365nm wavelength, 100-150mW ultraviolet lamp for 10-20 seconds. Specifically, the graphene transfer involves transferring the CVD-grown graphene film to the surface of the waveguide core layer of the sensing arm using polymethyl methacrylate (PMMA). After transfer, the PMMA is dissolved and removed using acetone solvent. Annealing is then performed in a nitrogen or argon inert atmosphere at a temperature of 150-200℃ for 10-15 minutes to ensure a tight bond between the graphene and the waveguide core layer.

[0060] After the fabrication of the dual-wavelength velocimetry device is completed, the following procedures are followed to complete the packaging, integration, calibration, and practical application of the dual-wavelength velocimetry device:

[0061] In the packaging and system integration process, the fabricated dual-wavelength velocimetry device is fixed in a packaging base with an airflow channel. During packaging, it is ensured that the airflow channel precisely aligns with the graphene region on the sensor arm surface, allowing the airflow to be measured to directly blow onto the graphene surface, ensuring that the graphene can accurately sense changes in airflow velocity. Subsequently, the beams output from the 980nm DFB laser and the 1550nm DFB laser are combined by a wavelength division multiplexer (WDM) and then precisely coupled into the input end of the interferometer arm of the dual-wavelength velocimetry device via end-face coupling, realizing the dual-wavelength beam... Synchronous transmission is performed; the 980nm beam serves as a reference wavelength for intensity baseline calibration, environmental temperature drift, and noise compensation, while the 1550nm beam serves as the main sensing wavelength, highly sensitive to the refractive index and micro-deformation changes of the graphene sensing arm, and is used for velocity inversion; the interference output of the dual-wavelength velocimetry device is connected to an optical spectrum analyzer or a high-speed photodetector to acquire the interference output light intensity signal; the acquired 1550nm interference output light intensity signal is sampled at equal time intervals to form a continuous interference light intensity time series, thus completing the signal acquisition link construction.

[0062] In wind tunnel calibration and model calibration, the integrated dual-wavelength velocimetry device is set up in a standard wind tunnel platform to conduct calibration experiments, completing the training and mapping relationship establishment of the Mamba model. The Mamba model dynamically models the time-series signal of interference light intensity based on discrete state-space equations. During the calibration process, based on the principle of relative motion, the standard wind tunnel outputs multiple stable standard wind speeds within the range of 0–10 m / s, which are equivalent to the actual motion speed of the object under test. The 1550 nm interference light intensity time series under each standard wind speed is collected simultaneously. After normalization and detrending preprocessing, it is used as a training sample to input into the Mamba model to complete offline training. At the same time, the output value of the Mamba model under the corresponding wind speed is recorded. The nonlinear mapping relationship between the real wind speed and the output value of the Mamba model is established through data fitting, and the system calibration is completed.

[0063] In practical speed measurement applications, the calibrated dual-wavelength speed measuring device is installed on the surface of the target (such as the wing of a micro-aircraft, the joint of an industrial robot, or a precision motion platform). The 980nm and 1550nm dual light sources are activated, and the device enters real-time speed measurement mode: it acquires the 1550nm interference output light intensity signal in real time and generates a time series. After normalization and detrending preprocessing, the signal is input into a pre-trained and calibrated Mamba model. The Mamba model completes dynamic feature modeling through state-space equations, outputs an estimated velocity value, and then combines this with the mapping relationship established during the calibration phase for error compensation and correction. Ultimately, it achieves real-time, high-precision inversion of the target's velocity, providing accurate speed feedback for industrial robot joints, precision motion platforms, and other equipment.

[0064] Example 2

[0065] The dual-wavelength velocity measurement method in this embodiment includes the following steps:

[0066] Step S1: Couple a 980nm light source and a 1550nm light source into the input / output waveguides of the dual-wavelength velocimetry device, respectively. The 980nm light source is used to photothermally excite the graphene on the surface of the sensing arm, placing the graphene in a preset thermal bias state to enhance its sensitivity to wind speed. When the moving object generates relative wind speed, the graphene on the surface of the sensing arm senses the wind speed and generates temperature and stress changes, thereby modulating the phase of the beam transmitted in the sensing arm. The 1550nm light source is used to form an interference light intensity signal to achieve speed measurement. The beam of the 1550nm light source is distributed by an optical beam splitter and transmitted along the reference arm and the sensing arm, respectively. After being reflected by the metal reflective layer at the end of the reference arm and the sensing arm, a dual-path reflection Michelson interferometer is formed. The 980nm beam is distributed by an optical beam splitter and transmitted only along the sensing arm.

[0067] Step S2: Fix the dual-wavelength velocity measuring device on the surface of the moving object to be measured, so that the dual-wavelength velocity measuring device moves with the moving object at a speed v, generating a relative wind speed that acts on the graphene layer on the sensing arm. Using Newton's law of cooling in heat transfer... Where A is the heat transfer surface area of ​​the object, h is the convective heat transfer coefficient, m is the mass of the object, c is the specific heat capacity, and v is the velocity. Let be the temperature of the object at time t. For ambient temperature; due to ,available And thus obtain ;

[0068] Step S3: Wind-induced disturbances exert temperature and pressure effects on the graphene surface. ,in The refractive index of graphene, Thermo-optic coefficient, The photoelastic coefficient, Wind-induced stress causes changes in the optical properties of graphene, which in turn leads to changes in the effective refractive index of the waveguide in the sensing arm. The effective refractive index of the waveguide is [value missing]. , It is the overlap factor between the mode field and graphene, thereby achieving the purpose of modulating the phase of 1550nm light in the sensing arm. The phase change formula is: ,in, This is the working length of the sensor arm;

[0069] Step S4: Acquire the 1550nm interference output light intensity signal. The formula for the output light intensity is: ,in, The initial output light intensity, The initial phase is used as the reference point. A change in phase will cause a change in the output light intensity, which in turn will cause a shift in the interference spectrum. The velocity is obtained indirectly through the shift relationship of the interference spectrum. ,in, The effective refractive index of the sensor arm.

[0070] Example 3

[0071] The difference between this embodiment and Embodiments 1 and 2 is that:

[0072] In this embodiment, the speed value obtained through actual inversion in Embodiment 2 is weighted with the speed prediction value output by the Mamba model in Embodiment 1. The weighting formula is as follows:

[0073]

[0074] in, It is the final velocity value after weighting. and These are weighting coefficients, satisfying... + =1, The velocity values ​​obtained through actual inversion and These are speed predictions obtained based on the Mamba model.

[0075] In this embodiment, an appropriate weight can be selected based on the reliability of the interference light intensity signal. If the quality of the interference light intensity signal is good and the noise is low, the weight can be increased. Conversely, if the Mamba model, after being trained on a large amount of data, can provide accurate estimation results, then the impact will increase. The reliability of the interference intensity signal is comprehensively judged based on three indicators: signal-to-noise ratio, time-series fluctuation variance, and baseline drift. The judgment logic is as follows: real-time acquisition of the 1550nm interference intensity time series, quantitative calculation of the signal-to-noise ratio, data fluctuation dispersion, and baseline offset, thereby objectively reflecting the strength of interference from complex airflow disturbances, environmental temperature drift, and photoelectric acquisition noise on the interference waveform. When the interference waveform is regular, the signal-to-noise ratio is high, the signal variance is small, and the baseline drift is weak, the interference intensity signal is judged to have good quality and high reliability. If waveform distortion, increased noise, drastic data fluctuation, or severe baseline offset occurs, the reliability of the interference intensity signal is judged to be poor. Based on this judgment result, the weighting coefficient is dynamically adjusted to autonomously allocate the weight ratio of physical inversion speed and Mamba model prediction speed.

[0076] The above are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above content. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A dual-wavelength velocimetry device integrating on-chip optical waveguide and graphene, characterized in that, The device includes a silicon wafer substrate, a lower cladding layer disposed on the silicon wafer substrate, a waveguide core layer disposed on the lower cladding layer, and an upper cladding layer covering the waveguide core layer and the graphene surface. The waveguide core layer includes an input / output waveguide, an optical beamsplitter, and an interferometer arm. A metal reflective layer is disposed at the end of the interferometer arm, which includes a sensing arm and a reference arm. The sensing arm and the reference arm are symmetrically arranged, and the surface of the sensing arm is covered with graphene. The output end of the input / output waveguide is connected to the input end of the optical beamsplitter, and the two output ends of the optical beamsplitter are respectively connected to the reference arm and the sensing arm of the interferometer arm.

2. The dual-wavelength velocimetry device integrating on-chip optical waveguide and graphene according to claim 1, characterized in that, The input / output waveguides are used to couple a 980nm light source and a 1550nm light source. The 980nm light source is used to photothermally excite the graphene on the surface of the sensing arm, placing the graphene in a preset thermal bias state to enhance its sensitivity to wind speed. The 1550nm light source is used to generate an interference light intensity signal to achieve speed measurement. The optical beam splitter is used to distribute the beams from the two light sources to the reference arm and the sensing arm, respectively. The metal reflective layer is used to reflect the beams to form a dual-path reflective Michelson interferometer. The graphene is used to sense the phase generated by the motion of the object being measured. The wind speed, in turn, causes temperature and stress changes to modulate the phase of the light beam. The temperature change, through the thermo-optic effect, and the stress change, through the photoelastic effect, together change the refractive index of the graphene, thereby modulating the phase of the light beam transmitted in the sensing arm, creating a phase difference between the light beams of the reference arm and the sensing arm. The light beam from the 1550nm light source is distributed by the optical beam splitter and transmitted along the reference arm and the sensing arm respectively. After being reflected by the metal reflective layer at the end of the reference arm and the sensing arm, the two reflected light beams merge again, and interference occurs due to the phase difference, forming an interference light intensity signal.

3. The dual-wavelength velocimetry device integrating on-chip optical waveguide and graphene according to claim 1, characterized in that, The reference arm of the interferometer arm consists of a first curved section and a first straight section; one end of the first curved section is connected to the optical beam splitter, and the other end is connected to one end of the first straight section; the other end of the first straight section is provided with a metal reflective layer.

4. The dual-wavelength velocimetry device integrating on-chip optical waveguide and graphene according to claim 3, characterized in that, The sensing arm of the interferometer arm is composed of a second curved section and a second straight section; one end of the second curved section is connected to the optical beam splitter, and the other end is connected to one end of the second straight section; the other end of the second straight section is provided with a metal reflective layer.

5. The dual-wavelength velocimetry device integrating on-chip optical waveguide and graphene according to claim 3 or 4, characterized in that, The first straight line segment and the second straight line segment are set in parallel.

6. A dual-wavelength velocimetry method integrating on-chip optical waveguides and graphene, characterized in that, The dual-wavelength velocity measuring device according to any one of claims 1-5 includes the following steps: S1: The dual-wavelength speed measuring device is fixedly installed on the surface of the moving object to be measured. The installation direction needs to ensure that the graphene on the surface of the sensing arm is perpendicular to the direction of movement of the moving object to maximize the perception of relative wind speed. S2: A 980nm light source and a 1550nm light source are respectively coupled into the input / output waveguides of the dual-wavelength velocimetry device. The 980nm light source is used to photothermally excite the graphene on the surface of the sensing arm, placing the graphene in a preset thermal bias state to enhance its sensitivity to wind speed. When the moving object generates relative wind speed, the graphene on the surface of the sensing arm senses the wind speed and generates temperature and stress changes, thereby modulating the phase of the beam transmitted in the sensing arm. The 1550nm light source is used to form an interference light intensity signal to achieve speed measurement. The beam of the 1550nm light source is distributed by an optical beam splitter and transmitted along the reference arm and the sensing arm respectively. After being reflected by the metal reflective layer at the end of the reference arm and the sensing arm, a dual-path reflection Michelson interferometer is formed. The 980nm beam is distributed by an optical beam splitter and transmitted only along the sensing arm. S3: A photodetector is used to synchronously acquire the interference output light intensity signal corresponding to a wavelength of 1550nm, forming an interference light intensity time series. After preprocessing the acquired interference light intensity time series, the preprocessed interference light intensity time series is input into a pre-trained Mamba model. The Mamba model, based on the state space equation, models the long-range time dependence of the interference light intensity signal, extracts the phase change and frequency domain features in the interference output light intensity signal, and mines the interference dynamic features. The phase change and frequency domain features in the interference dynamic features are positively correlated with the velocity of the moving object under test. At the same time, the state transition relationship of the state space equation is used to recursively update the target velocity and position state, and output the velocity estimate of the moving object under test. S4: Construct a speed correction model. During actual speed measurement, perform error compensation and calibration on the speed estimate based on the speed correction model to obtain the speed prediction value.

7. The dual-wavelength velocimetry method using on-chip integrated optical waveguides and graphene according to claim 6, characterized in that, In step S3, the specific preprocessing process is as follows: the time window is divided, the amplitude is normalized, and the difference is processed on the acquired 1550nm interference output light intensity time series to form a feature sequence containing information on the change of interference light intensity; the training process of the Mamba model is as follows: based on the standard wind tunnel platform, interference light intensity time series samples corresponding to different known velocities are obtained, the interference light intensity time series samples are divided into training set and test set, the state transition matrix, Kalman filter parameters and feature extraction weights inside the Mamba model are calibrated through the training set, and the accuracy of the Mamba model is verified and optimized using the test set until the prediction error of the Mamba model meets the preset threshold; when the model test error is lower than the preset threshold, the training of the Mamba model is determined to be complete.

8. The dual-wavelength velocimetry method using on-chip integrated optical waveguides and graphene according to claim 6, characterized in that, In step S4, the speed correction model is constructed as follows: the true wind speed values ​​under different wind speeds and airflow directions are obtained in advance through a standard wind tunnel platform, the speed estimates of the Mamba model under the corresponding working conditions are collected simultaneously, and the mapping correction relationship between the speed estimates and the true speed values ​​is established by using the least squares method, thereby obtaining the speed correction model.

9. The dual-wavelength velocimetry method using on-chip integrated optical waveguides and graphene according to claim 6, characterized in that, The method also includes step S5, which involves: weighting and fusing the motion velocity value obtained by physical inversion of the interference light intensity signal with the velocity prediction value output by the Mamba model; and dynamically adjusting the weighting coefficients of the motion velocity value and the velocity prediction value based on the real-time quality of the interference light intensity signal to obtain the final velocity value.

10. A method for manufacturing a dual-wavelength velocimetry device integrating an on-chip optical waveguide and graphene, characterized in that, The method for manufacturing the dual-wavelength velocity measuring device according to any one of claims 1-5 comprises the following steps: Step 1: Select a 100-oriented silicon wafer as the silicon wafer substrate, and use plasma cleaning process to activate the surface of the silicon wafer to remove surface impurities and oxide layer. The cleaning time is controlled at 5-10 minutes, and the plasma power is set to 100-200W. Step 2: Apply EpoCladd material uniformly to the pretreated silicon wafer surface using a spin coating process. The coating speed is 3000-5000 r / min, and the coating thickness is controlled at 1-2 μm. After coating, use a 365nm ultraviolet lamp to irradiate for 10-20s for curing to form the lower cladding layer. Step 3: Apply SU-8 material as the waveguide core layer to the lower cladding surface using a spin coating process. The coating speed is 2000-4000 r / min, and the coating thickness is controlled at 3-5 μm. After coating, transfer the pre-designed mask patterns of the input / output waveguides, beam splitters, reference arms, and sensing arms (excluding the metal reflective layer) onto the SU-8 thin film using a photolithography machine. The exposure parameters of the photolithography machine are set as follows: exposure power 100-150 mW, exposure time 20-30 s, to ensure that the mask patterns are clearly transferred onto the SU-8 thin film. Step 4: The waveguide height of the SU-8 thin film is finely adjusted using reactive ion etching. After etching, the ends of the reference arm and the sensor arm are naturally exposed as vertical surfaces, serving as the deposition substrate for the metal reflective layer. Step 5: Deposit a metal film on the vertical surface of the end of the reference arm and the end of the sensor arm using evaporation or sputtering processes to form a metal reflective layer; the metal film is made of gold, silver or aluminum, and the film thickness is controlled between 50-100nm to ensure that the beam reflection efficiency is not less than 95%; Step 6: Graphene grown by chemical vapor deposition is transferred to the surface of the waveguide core layer of the sensing arm using a wet transfer process. Then, EpoCladd material, the same as the lower cladding, is spin-coated onto the waveguide core layer and the graphene surface as an upper cladding to restrict light transmission. The spin coating speed of the upper cladding is 3000-5000 r / min, and the coating thickness is controlled at 1-2 μm. After coating, it is cured by irradiation with a 365nm wavelength, 100-150mW ultraviolet lamp for 10-20 seconds. Specifically, the graphene transfer involves transferring the CVD-grown graphene film to the surface of the waveguide core layer of the sensing arm using polymethyl methacrylate (PMMA). After transfer, PMMA is removed by dissolving in acetone, and annealing is performed in a nitrogen or argon inert atmosphere at a temperature of 150-200℃ for 10-15 minutes to ensure tight adhesion between the graphene and the waveguide core layer.

Images