Joint test system, silicon carbide metasurface grating detection method and test device

Abstract

The present invention discloses a joint test system, a silicon carbide metasurface grating detection method and a test device, which belong to the field of semiconductor detection technology. A collaborative excitation signal is generated by regulating a light source and a microwave source, and applied to the silicon carbide metasurface grating to be tested to form an interference pattern of light intensity and microwave field. Multi-channel synchronous data acquisition, a multi-dimensional matrix is ​​constructed through time alignment, marker feature extraction and coordinate mapping, defect features are extracted to establish a benchmark model, the deviation is calculated to determine the quality, and the defect grating is divided into potential defect areas by screening abnormal points, constructing candidate areas and coupling verification. For potential defect areas, the excitation parameters are dynamically adjusted by the gradient descent algorithm, and the defect level is divided in combination with the characteristic defect degree and area. The present invention solves the problems of insufficient single-modal detection, low multi-modal alignment accuracy and blind adjustment of excitation parameters, realizes cross-band collaborative detection and intelligent grading, and improves detection reliability.

Description

Technical Field

[0001] The present invention relates to the field of semiconductor detection technology, and more particularly to a combined testing system, a silicon carbide metasurface grating detection method, and a testing device. Background Art

[0002] The performance of the silicon carbide metasurface grating to be tested is closely related to the structural parameters and material properties. However, the complex cross-band electromagnetic response characteristics and extreme environmental adaptability requirements are constantly increasing the requirements for testing technology.

[0003] At present, structural parameter testing mainly relies on scanning electron microscopes and atomic force microscopes. Scanning electron microscopes require a vacuum environment and the equipment is expensive, and the sample preparation process may introduce damage; the atomic force microscope has a slow detection speed and cannot meet the needs of large-area rapid detection. In optical performance testing, spectrophotometers are susceptible to interference from stray light, spectrometers have difficulty balancing broadband resolution and scanning speed, and polarization analyzers have a slow response speed and cannot analyze dynamic processes in real time. Microwave parameter testing is not sensitive enough to subwavelength structural responses and cannot achieve spatiotemporal synchronization with optical testing. In addition, traditional testing equipment has poor stability when simulating extreme environments such as high and low temperatures, humidity, and radiation, and lacks the ability to dynamically monitor multiple physical fields.

[0004] The above technical bottlenecks lead to many difficulties in the testing of silicon carbide metasurface gratings: there is a contradiction between efficiency and accuracy in the detection of structural parameters, the measurement of optical performance is limited by the principle of the equipment, microwave testing has difficulty in capturing nanoscale defects, and there is a lack of effective means for the evaluation of dynamic characteristics in extreme environments. How to break through the limitations of single modal detection and achieve high-precision detection and extreme environment adaptability evaluation under multi-band collaborative excitation has become a key issue that needs to be solved in this field. Therefore, in order to overcome these limitations, the present invention proposes a combined testing system, a silicon carbide metasurface grating detection method and a testing device. Summary of the Invention

[0005] In view of the shortcomings of the existing technology, the purpose of the present invention is to provide a joint testing system, a silicon carbide metasurface grating detection method and a testing device to solve the problems of insufficient single-modal detection sensitivity, low spatiotemporal alignment accuracy of multimodal data, and inaccurate defect feature extraction caused by blind adjustment of excitation signal parameters in the existing silicon carbide metasurface grating detection, and to achieve high-precision fusion detection of multimodal data and intelligent classification of defect levels under cross-band collaborative excitation.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] Joint test system, including

[0008] Obtaining initial excitation parameters of the optical wave excitation signal and the microwave excitation signal, regulating the optical wave source and the microwave source to generate a cooperative excitation signal, and applying the cooperative excitation signal to the silicon carbide metasurface grating to be measured to form an interference pattern;

[0009] The interference pattern is collected synchronously and multiple times through multiple channels by the detector to obtain multiple interference pattern data and perform preprocessing. The interference pattern data includes light intensity distribution data and microwave field intensity data;

[0010] Preprocessing is used to align the time dimension of the interference pattern data and extract feature points from the markers configured on the silicon carbide metasurface grating to map and match the interference pattern data coordinates to construct a multidimensional data matrix;

[0011] Defect features are extracted from the multidimensional data matrix to construct a multidimensional feature matrix. By establishing a feature benchmark model, the deviation index of the SiC metasurface grating to be tested is calculated to conduct a preliminary quality assessment. For SiC metasurface gratings to be tested that are determined to have defects, multiple rounds of screening for stable abnormal points, edge detection to construct candidate defect connected regions, and region growing to combine abnormal coupling verification to divide potential defect areas.

[0012] If a potential defect area is identified, the initial excitation parameters of the optical wave excitation signal and the microwave excitation signal are dynamically adjusted based on the information of the potential defect area to update and adjust the interference pattern regulated by the collaborative excitation signal. The updated interference pattern data is obtained and a deep quality assessment is performed to classify the defect level.

[0013] Specifically, the specific steps of updating and adjusting the collaborative incentive signal include:

[0014] Acquire the structural parameter data and electromagnetic characteristic data of the silicon carbide metasurface grating to be tested to determine the initial excitation parameters of the optical wave excitation signal and the microwave excitation signal;

[0015] According to the determined initial excitation parameters, the lightwave source and the microwave source are set to generate a cooperative excitation signal;

[0016] Applying a cooperative excitation signal to the silicon carbide metasurface grating to be tested, collecting the resulting interference pattern data and performing preprocessing;

[0017] Extract defect features from the pre-processed interference pattern data, perform preliminary quality assessment based on the extracted defect features, and determine whether the silicon carbide metasurface grating to be tested has defects;

[0018] If it is determined that there is a defect, the potential defect area of ​​the silicon carbide metasurface grating to be tested is divided, the position and range information of the potential defect area is recorded, and the collaborative excitation signal control is triggered.

[0019] Specifically, the specific steps of updating and adjusting the collaborative incentive signal also include:

[0020] The objective function is to maximize the ratio of the edge gradient of the interference pattern data in the potential defect area to the noise standard deviation, which is used to control the excitation parameters of the optical wave excitation signal and the microwave excitation signal.

[0021] Within the adjustment range of the incentive parameters, the incentive parameters are coordinated and regulated by the gradient descent algorithm, specifically including:

[0022] Based on the initial excitation parameters of the optical wave excitation signal and the microwave excitation signal, the initial frequency difference and phase difference are determined, and the objective function value of the potential defect area is obtained;

[0023] Calculate the partial derivatives of the objective function with respect to the frequency difference and phase difference, and update the frequency difference and phase difference according to the gradient direction;

[0024] Recalculating the frequencies and phases of the optical wave excitation signal and the microwave excitation signal based on the updated frequency difference and phase difference, synthesizing the updated optical wave excitation signal and the microwave excitation signal into a new cooperative excitation signal, applying the signal to the silicon carbide metasurface grating to be tested to form an updated interference pattern, and calculating the objective function value of the potential defect area;

[0025] Configure the convergence threshold and calculate the change in the objective function value between two adjacent times. If the change is less than the convergence threshold, stop updating and output the optimized excitation parameters; otherwise, continue iterative updating.

[0026] Specifically, the specific steps of constructing a multidimensional data matrix include:

[0027] Using the timestamp of the synchronous clock signal, the light intensity distribution data collected by the photoelectric detector and the microwave field intensity data collected by the microwave detector are aligned in the time dimension;

[0028] After removing noise from the light intensity distribution data and enhancing the contrast, a scale space is constructed to detect extreme points in the enhanced light intensity distribution data, determine the locations of feature points of the silicon carbide metasurface grating marker to be measured, calculate the scale and orientation information of each feature point, and generate a feature descriptor;

[0029] A threshold segmentation method is used for microwave field intensity data. The segmentation threshold is set to separate the marker area. The outline of the marker area is optimized through morphological processing. The geometric center of the marker area is calculated as the feature point, and the coordinate position of the feature point in the microwave field intensity data is recorded.

[0030] Using a geometric calibration algorithm, the coordinates of the light intensity distribution data are mapped to the physical coordinates of the silicon carbide metasurface grating to be measured, and the coordinates of the microwave field intensity data are matched with the coordinates of the light intensity distribution data;

[0031] A fixed position on the silicon carbide metasurface grating to be measured is selected as the origin of the spatial coordinate system and the direction of the coordinate axis is determined, and the coordinates of the light intensity distribution data and the microwave field intensity data are converted into coordinates in the spatial coordinate system;

[0032] The light intensity distribution data and microwave field intensity data are integrated according to the spatial coordinate system to construct a multidimensional data matrix, in which the elements of the multidimensional data matrix correspond to the light intensity distribution data and microwave field intensity data at corresponding positions in the silicon carbide metasurface grating to be measured.

[0033] Specifically, the steps for extracting defect features include:

[0034] Obtain the preprocessed multidimensional data matrix and extract the corresponding defect features for the local area at each coordinate position in the multidimensional data matrix. The defect features exist in the form of multidimensional feature vectors, including light intensity distribution features, microwave field features, and cross-modal correlation features.

[0035] The light intensity distribution feature is extracted by demarcating adjacent areas centered on the coordinate position of the multidimensional data matrix and calculating the light intensity contrast index. The actual fringe period corresponding to the coordinate position of the multidimensional data matrix is ​​compared with the configured theoretical period, and the deviation rate is calculated as the fringe spacing uniformity index. The gradient amplitude at the edge position of the multidimensional data matrix is ​​calculated, and the standard deviation of the gradient amplitudes of adjacent edge positions is calculated as the edge gradient consistency index.

[0036] The microwave field characteristics are extracted by dividing the local area centered on the coordinate position in the multidimensional data matrix into analysis units, calculating the variance of the microwave phase value in the local area as an indicator of microwave phase stability; comparing the microwave field intensity in the local area with the standard microwave field intensity, and calculating the ratio of the microwave field intensity to the standard microwave field intensity as a characteristic indicator of the field intensity attenuation rate;

[0037] The cross-modal correlation feature is extracted by calculating the light intensity gradient and microwave phase gradient respectively for the light intensity distribution data and microwave field intensity data corresponding to the coordinate position in the multidimensional data matrix, and calculating the correlation index of the light intensity gradient and microwave phase gradient through the correlation coefficient;

[0038] The extracted defect features are dedimensionalized so that each coordinate position in the multidimensional data matrix generates a set of defect feature vectors to construct a multidimensional feature matrix.

[0039] Specifically, the steps for conducting a preliminary quality assessment include:

[0040] Establishing a standard feature set, the standard feature set includes a multi-dimensional feature matrix of a defect-free standard silicon carbide metasurface grating sample;

[0041] Statistically calculate the mean vector and covariance matrix of each defect feature in the multidimensional feature matrix in the standard feature set, and construct a distribution model of the multidimensional feature matrix in the standard feature set as a feature benchmark model;

[0042] Obtain interference pattern data of the silicon carbide metasurface grating to be tested, and construct a multidimensional data matrix and a multidimensional feature matrix. Calculate the distance between the defect feature vector at each coordinate position in the multidimensional feature matrix of the silicon carbide metasurface grating to be tested and the feature reference model to generate a deviation matrix.

[0043] The deviation matrix is ​​globally averaged to obtain the deviation index of each interference pattern data; a distance threshold is configured and the deviation index is compared with the distance threshold. If the deviation index is greater than the distance threshold, it is determined that the silicon carbide metasurface grating to be tested has defects; otherwise, it is determined to be qualified.

[0044] Specifically, the steps for conducting a preliminary quality assessment include:

[0045] If it is determined that the silicon carbide metasurface grating to be tested has defects, based on the deviation matrix and the distance threshold, the multidimensional data matrix of the interference pattern data collected multiple times is marked with coordinate points whose deviation index is greater than the distance threshold as suspected defect points, forming a binary abnormal point matrix, and the abnormal frequency of the same coordinate point is counted to screen out stable abnormal points whose abnormal frequency is greater than the preset frequency threshold;

[0046] Edge detection is performed on stable outliers to extract continuous edge contours, and candidate defect connected domains are formed through morphological dilation operations;

[0047] Select seed points in the candidate defect connectivity domain by sorting by deviation index, and perform regional growth based on the light intensity distribution characteristics, microwave field characteristics, and cross-modal correlation characteristics with the seed points as the center until the preset growth termination conditions are met;

[0048] Configure the effective threshold and calculate the proportion of coordinates with both abnormal light intensity distribution and abnormal microwave field in each potential defect area. If the proportion is greater than the effective threshold, it is confirmed as a valid potential defect area; otherwise, it is eliminated.

[0049] Specifically, the steps of deep quality assessment include:

[0050] Obtaining a multidimensional data matrix and a multidimensional feature matrix of each updated interference pattern data, mapping the coordinates of the potential defect area to the updated multidimensional data matrix, and generating a potential defect area mask;

[0051] Based on the potential defect area mask, the mean and standard deviation of the defect feature vectors in the potential defect area and the non-potential defect area are calculated respectively, and the difference of each defect feature vector is calculated;

[0052] Based on the calculated mean and standard deviation of the defect feature vectors of the potential defect area and the non-potential defect area, the defect degree of the defect feature vector of each potential defect area is calculated to quantify the degree of difference between the potential defect area and the non-potential defect area in each defect feature dimension;

[0053] According to the defect degree of each defect feature vector in the potential defect area and the area of ​​the potential defect area, the defect level is divided into a mild defect level, a moderate defect level, and a severe defect level.

[0054] The jointly tested SiC metasurface grating detection method includes the following steps:

[0055] Step S1: obtaining initial excitation parameters of the lightwave excitation signal and the microwave excitation signal, regulating the lightwave source and the microwave source to generate a cooperative excitation signal, and applying the cooperative excitation signal to the silicon carbide metasurface grating to be measured to form an interference pattern including light intensity distribution and microwave field intensity distribution;

[0056] Step S2: Performing multi-channel synchronous multiple acquisitions of the interference pattern through a detector to obtain multiple interference pattern data, including light intensity distribution data and microwave field intensity data; aligning the interference pattern data in the time dimension, and mapping and matching the interference pattern data coordinates by extracting the characteristic points of the marker on the silicon carbide metasurface grating to be measured to construct a multidimensional data matrix;

[0057] Step S3: Defect features are extracted from the multidimensional data matrix to construct a multidimensional feature matrix. A feature reference model is established based on the multidimensional feature matrix of a defect-free standard silicon carbide metasurface grating sample. The deviation index of the silicon carbide metasurface grating to be tested is calculated to perform quality judgment. If a defect is determined to be present, the method then divides the potential defect area by screening stable outliers for multiple rounds, constructing a candidate defect connected domain through edge detection, performing region growing based on multimodal feature similarity conditions, and verifying the coupling between light intensity and microwave field anomalies.

[0058] Step S4: If a potential defect area is identified, the initial excitation parameters of the optical wave excitation signal and the microwave excitation signal are dynamically adjusted according to the information of the potential defect area, the collaborative excitation signal is updated, and the updated interference pattern data is obtained; the defect level is divided based on the defect degree of the defect area and the non-defect area in the updated multidimensional feature matrix and the area of ​​the potential defect area.

[0059] A testing device, comprising:

[0060] Lightwave signal source, used to generate lightwave excitation signals of different wavelengths, powers, and polarization states;

[0061] A microwave signal source, used to generate microwave excitation signals of different frequencies, phases, and powers;

[0062] A trigger controller is used to generate a synchronous clock signal to achieve time synchronization between the lightwave signal source and the microwave signal source;

[0063] A signal synthesizer, used to synthesize the optical wave excitation signal and the microwave excitation signal into a cooperative excitation signal;

[0064] Photoelectric detection array, used to collect light intensity distribution data and identify the characteristic points of the marker;

[0065] Microwave scanning probe system, used to collect microwave field intensity data and record scanning coordinates;

[0066] The data processing unit is used to execute the marker feature extraction and coordinate mapping algorithm, and construct a multidimensional data matrix, extract the light intensity distribution characteristics, microwave field characteristics and cross-modal correlation characteristics, generate defect feature vectors, and construct a multidimensional feature matrix;

[0067] Defect detection unit, used to perform potential defect area division and classify defect levels based on characteristic defectivity, area ratio, and cross-modal overlap;

[0068] The parameter control actuator is used to dynamically adjust the excitation parameters according to the information of the potential defect area and optimize the excitation parameters through the gradient descent algorithm.

[0069] Beneficial effects of the present invention:

[0070] Through multi-modal collaborative excitation and multi-channel synchronous acquisition, the light intensity and microwave field data are integrated to achieve cross-frequency information complementarity, which solves the problems of insufficient sensitivity of single-modal detection and weak nano-level defect capture capability, and significantly improves detection accuracy; multi-dimensional data matrix construction and multi-modal feature extraction realize comprehensive characterization of grating performance, and combine characteristic benchmark model with deviation analysis to accurately determine the existence of defects; in the process of potential defect area division, stable outlier point screening, regional growth and coupling verification are used to effectively eliminate noise interference and achieve accurate positioning of defects; the closed-loop feedback mechanism of dynamic adjustment of excitation signal parameters can adaptively optimize detection conditions, enhance defect feature contrast, and improve detection reliability; the overall solution breaks through the limitations of traditional equipment in efficiency, accuracy and environmental adaptability, and realizes efficient, accurate detection and intelligent grading of silicon carbide metasurface gratings to be tested, providing an effective means for performance evaluation of silicon carbide metasurface gratings to be tested under extreme environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 This is a structural diagram of a joint testing system of the present invention;

[0072] Figure 2 A flowchart of the specific steps of updating and adjusting the cooperative excitation signal of the present invention;

[0073] Figure 3Flowchart of the specific steps of pre-processing the collected interference pattern data according to the present invention;

[0074] Figure 4 Flowchart for preliminary quality assessment and potential defect area division for the present invention;

[0075] Figure 5 This is a flow chart of a combined testing method for silicon carbide metasurface grating detection according to the present invention. DETAILED DESCRIPTION

[0076] Example 1:

[0077] See also Figure 1 This embodiment introduces a joint test system, including a multi-source excitation module, a signal acquisition module, a signal processing module, and a quality assessment module:

[0078] The multi-source excitation module is designed to generate a cross-band, synergistic excitation signal containing light waves and microwaves, and apply it to the silicon carbide metasurface grating under test to form an interference pattern. Based on information about potential defect areas, the module dynamically adjusts the parameters of the synergistic excitation signal to update and optimize the interference pattern. By changing the period, spatial distribution, and phase distribution of the interference pattern, the module can more clearly reflect potential defects in the target area, improving the sensitivity and accuracy of defect detection.

[0079] Specifically, the structural parameters and electromagnetic properties of the SiC metasurface grating to be tested are obtained. Structural parameters include the period, height, and duty cycle of the SiC metasurface grating to be tested, while electromagnetic properties include the dielectric constant and permeability. Based on the detection requirements for defects of different sizes and types, the excitation signal parameters, such as the frequency range, amplitude range, and phase relationship of the lightwave and microwave, are determined. Multiple markers with distinct features are placed on the surface of the SiC metasurface grating to be tested. These markers must be clearly identifiable by photodetectors and microwave detectors. For example, tiny metal dots or high-contrast patterns can be used as markers. The positions of the markers must be precisely measured and recorded within the actual physical coordinate system of the SiC metasurface grating to be tested. The lightwave and microwave sources are controlled so that each source stably outputs a signal that meets the specified frequency, intensity, and phase requirements. To ensure signal coordination, synchronized clock signals are used to trigger the lightwave and microwave signal sources for time synchronization. Furthermore, by adjusting the frequency difference between the lightwave and microwave frequencies, the period and spatial distribution of the interference pattern are adjusted to meet the detection requirements for defects of different sizes and types. A phase modulator is used to adjust the phase of the light wave and microwave, changing the phase distribution of the interference pattern. The light wave and microwave signals that have been regulated, synchronized and phase-adjusted are input into a signal synthesizer for synthesis to form a cross-band collaborative excitation signal.

[0080] See also Figure 2 Preferably, the specific steps of updating and adjusting the collaborative excitation signal include:

[0081] The structural parameter data and electromagnetic characteristic data of the SiC metasurface grating to be tested are obtained. Using multi-physics simulation software, an interference model of the SiC metasurface grating to be tested is established based on the collected data. Lightwave and microwave excitation simulations are then performed on this interference model under different parameters, analyzing the formation of interference patterns under different excitation parameter combinations. Combining the simulation results with the structural parameter data and electromagnetic characteristic data of the SiC metasurface grating to be tested, the initial excitation parameter values ​​for the lightwave and microwave excitation signals are preliminarily determined. The excitation parameters include frequency, amplitude, and phase.

[0082] Based on the initial excitation parameters, the lightwave and microwave sources are configured. For the lightwave source, parameters such as the laser cavity length and operating current are adjusted to achieve the desired frequency. For the microwave source, the output frequency and amplitude are set using the signal generator's control panel. A synchronous clock signal is also set to ensure time synchronization between the lightwave and microwave sources. The configured lightwave and microwave signals are then fed into a signal synthesizer for synthesis to form a coordinated excitation signal.

[0083] A collaborative excitation signal is applied to the silicon carbide metasurface grating to be tested, and the resulting interference pattern data is collected using photodetectors and microwave detectors. The collected interference pattern data is preprocessed to remove noise interference and normalized to unify the signal data from different acquisition channels and at different acquisition times into the same scale range.

[0084] Extract defect features from the pre-processed interference pattern data, including contrast, fringe spacing, fringe shape, amplitude change, phase change, etc., and perform a preliminary quality assessment based on the extracted defect features to determine whether the silicon carbide metasurface grating to be tested has defects;

[0085] If there are no defects, the quality of the silicon carbide metasurface grating to be tested is determined to be qualified;

[0086] If it is determined that there is a defect, the potential defect area of ​​the silicon carbide metasurface grating to be tested is divided, the location and range information of the potential defect area is recorded, and the cooperative excitation signal control is triggered, that is:

[0087] Configure the excitation parameter adjustment range of the lightwave excitation signal and the microwave excitation signal, including the frequency adjustment range, amplitude adjustment range, and phase adjustment range;

[0088] The objective function is to maximize the ratio of the edge gradient of the interference pattern data in the potential defect area to the noise standard deviation, and to adjust the excitation parameters of the light wave excitation signal and the microwave excitation signal. The expression is as follows:

[0089]

[0090] in, is the frequency difference between the optical wave excitation signal and the microwave excitation signal, is the phase difference between the optical wave excitation signal and the microwave excitation signal, is the intensity distribution of the interference pattern, is the edge gradient of the interference pattern, is the noise standard deviation of the interference pattern. Objective function By adjusting the frequency difference between the light wave excitation signal and the microwave excitation signal and phase difference , maximizes the edge gradient of the interference pattern and noise standard deviation The ratio of is used to enhance the contrast of defect edges and improve the accuracy of defect detection.

[0091] Within the excitation parameter adjustment range of the optical wave excitation signal and the microwave excitation signal, the excitation parameters are coordinated and controlled by the gradient descent algorithm, namely:

[0092] According to the initial excitation parameters of the light wave excitation signal and the microwave excitation signal, the frequency difference and phase difference between the initial light wave excitation signal and the microwave excitation signal are determined, and the objective function value of the potential defect area is obtained.

[0093] Calculate the partial derivatives of the objective function with respect to the frequency difference and phase difference, and update the frequency difference and phase difference according to the gradient direction. The update formula is:

[0094]

[0095]

[0096] in, is the learning rate, ranging from (0, 1), It is The frequency difference of the update, It is The frequency difference of the update, is the partial derivative of the objective function with respect to the frequency difference, It is The updated phase difference, It is The updated phase difference, is the partial derivative of the objective function with respect to the phase difference.

[0097] Obtaining the updated frequency difference and phase difference, recalculating the frequency and phase of the lightwave excitation signal and the microwave excitation signal, inputting the updated lightwave signal and microwave signal into the signal synthesizer for synthesis to form an updated cooperative excitation signal, and applying it to the silicon carbide metasurface grating to be tested to form an updated interference pattern, and calculating the objective function value of the potential defect area;

[0098] Configure a convergence threshold and calculate the change in the objective function value between two adjacent times. If the change is less than the convergence threshold, stop updating and output the optimized excitation parameter values ​​of the optical wave excitation signal and the microwave excitation signal. Otherwise, continue iterative updating.

[0099] The signal acquisition module is used to collect the interference pattern data generated by the silicon carbide metasurface grating to be tested under the action of the cooperative excitation signal, including the interference pattern data under the initial excitation parameter value, the interference pattern data under the cooperative excitation signal control state, and the updated interference pattern data under the optimized excitation parameter values ​​of the lightwave excitation signal and the microwave excitation signal. Multiple acquisitions are performed according to the set sampling frequency and sampling time to improve the accuracy of the data and provide data support for subsequent signal processing and quality assessment links.

[0100] The signal acquisition module is equipped with multiple detectors, including photoelectric detectors for collecting light wave signals and microwave detectors for collecting microwave signals, so as to realize multi-channel synchronous acquisition of interference pattern data. The interference pattern data includes the light intensity distribution data collected by the photoelectric detector and the microwave field intensity data collected by the microwave detector.

[0101] A photodetector is used to image the SiC metasurface grating under test, where markers are placed, and collect light intensity distribution data containing the markers. Simultaneously, a microwave detector is used to spatially scan the SiC metasurface grating under test, recording the position information and corresponding microwave field intensity data for each scan point. The scanning path must be consistent and accurate throughout the entire process.

[0102] The collected interference pattern data is transmitted to the signal processing module in real time using a high-speed data transmission interface to ensure stability and reliability. During the data transmission process, the collected signals undergo preliminary preprocessing operations such as denoising and filtering. Data collected from different channels is synchronized and calibrated to ensure consistency and comparability, providing a high-quality data foundation for subsequent processing.

[0103] The signal processing module pre-processes each collected interference pattern data, integrates the light intensity distribution data and microwave field intensity data, constructs a multidimensional data matrix, and extracts defect features that are helpful for quality assessment from the multidimensional data matrix. It constructs a multidimensional feature matrix, and under the control of the collaborative excitation signal, extracts the edge gradient and noise standard deviation of the potential defect area in the interference pattern data in real time, and feeds them back to the multi-source excitation module.

[0104] See also Figure 3 Preferably, the specific steps of preprocessing the collected interference pattern data include:

[0105] By synchronizing the timestamp of the clock signal, the light intensity distribution data collected by the photodetector and the microwave field intensity data collected by the microwave detector are aligned in the time dimension, ensuring that the light intensity distribution data and microwave field intensity data obtained at any same moment correspond to the same physical state of the silicon carbide metasurface grating to be measured, thereby laying the foundation for time consistency for subsequent data processing and analysis.

[0106] The noise of the light intensity distribution data is removed by a filtering algorithm, and the contrast of the light intensity distribution data is enhanced by a contrast enhancement algorithm.

[0107] By constructing a scale space, the extreme points are detected in the enhanced light intensity distribution data to determine the position of the feature points of the marker. For each extracted feature point, its scale and direction information are calculated, and then a feature descriptor is generated to characterize the feature point. The coordinate position of each feature point in the light intensity distribution data is recorded, providing key information for subsequent coordinate conversion and matching.

[0108] A threshold segmentation method is applied to the microwave field intensity data. Based on the characteristics of the microwave field intensity data, a segmentation threshold is set to separate the marker region from the background data, achieving preliminary identification and labeling of the marker region. Morphological processing methods are applied to the segmented marker region to optimize its contours and remove noise and isolated points within the marker region, making the marker region clearer and more accurate. Furthermore, the geometric center of each marker region is calculated and used as the marker's characteristic point in the microwave field intensity data. The coordinate position of each characteristic point in the microwave field intensity data is also recorded.

[0109] Using a geometric calibration algorithm, the coordinates of the light intensity distribution data after feature extraction are mapped to the physical coordinates of the silicon carbide metasurface grating to be measured based on the feature descriptors of the feature points in the light intensity distribution data and the coordinate positions of the feature points in the microwave field intensity data. This allows the light intensity distribution data to be converted from image pixel coordinates to actual physical space coordinates. Simultaneously, the geometric calibration algorithm is used to match the microwave scanning coordinates in the microwave field intensity data with the coordinates of the light intensity distribution data, ensuring an accurate correspondence between the two within the same spatial reference frame, providing the prerequisites for the subsequent establishment of a spatial coordinate system and data integration.

[0110] A fixed and easily identifiable position on the SiC metasurface grating to be tested is selected as the origin of the spatial coordinate system, such as a corner point of the SiC metasurface grating to be tested; and the direction of the coordinate axis is determined based on the structural characteristics and detection requirements of the SiC metasurface grating to be tested.

[0111] The mapped coordinates of the light intensity distribution data and the converted coordinates of the microwave field intensity data are converted to coordinates in the established spatial coordinate system based on the origin position and coordinate axis directions. Through this conversion process, the light intensity distribution data and the microwave field intensity data have consistent coordinate expressions in a unified spatial coordinate system.

[0112] The light intensity distribution data and microwave field intensity data are integrated according to the established spatial coordinate system to construct a multidimensional data matrix containing multimodal features. In this multidimensional data matrix, each element corresponds to a multifaceted data set at a specific coordinate position on the SiC metasurface grating to be tested, including light intensity distribution data, microwave field intensity data, and associated characteristic information. This integration provides a rich and organized data foundation for subsequent comprehensive analysis and defect detection of the SiC metasurface grating to be tested.

[0113] Preferably, the specific steps of extracting defect features include:

[0114] Obtain the pre-processed multidimensional data matrix and extract the corresponding defect features for the local area of ​​each coordinate position in the multidimensional data matrix. The defect features exist in the form of multidimensional feature vectors, specifically including three types: light intensity distribution features, microwave field features, and cross-modal correlation features:

[0115] The light intensity distribution characteristics include: for any coordinate position in the multidimensional data matrix, adjacent square or rectangular areas are delineated with the position as the center, and the light intensity distribution contrast index is obtained by calculating the light intensity difference of the adjacent areas. The contrast index value reflects the degree of brightness difference in the areas surrounding the position.

[0116] Based on the periodic characteristics of the fringes in the light intensity distribution data, the actual fringe period of the area at that position is measured and compared with the theoretical period determined by the grating design parameters. The deviation rate is calculated as an indicator of fringe spacing uniformity, which is used to detect abnormal changes in the fringe period.

[0117] The edge detection algorithm is used to determine whether the position is an edge pixel. If it is an edge pixel, its gradient amplitude is calculated, and the standard deviation of the gradient amplitudes of adjacent edge pixels is calculated. This is used as the edge gradient consistency indicator to characterize the degree of fluctuation of the light intensity change at the edge.

[0118] The microwave field characteristics include: dividing the local area centered on the coordinate position in the multidimensional data matrix into analysis units, calculating the variance of the microwave phase value in the local area as a microwave phase stability index, and reflecting the fluctuation of the microwave phase in the local area.

[0119] The microwave field intensity in the local area is compared with the standard microwave field intensity measured from a defect-free standard sample or calculated from a theoretical model, and the ratio of the microwave field intensity to the standard microwave field intensity is calculated as a field intensity attenuation rate index, which is used to measure the abnormal attenuation degree of the microwave field intensity in the local area.

[0120] Cross-modal correlation features include: calculating the light intensity gradient and microwave phase gradient respectively for the light intensity distribution data and microwave field intensity data corresponding to the coordinate position, where the light intensity gradient reflects the spatial change rate of light intensity, and the microwave phase gradient reflects the spatial change rate of microwave phase. The correlation index between the two is calculated through the correlation coefficient, which is used to identify the abnormal coupling state of the electromagnetic characteristics and geometric structure at that position.

[0121] The extracted defect features are dedimensionalized so that each coordinate position in the multidimensional data matrix generates a set of defect feature vectors of unified dimension containing light intensity distribution, microwave field and cross-modal correlation information, and a multidimensional feature matrix is ​​constructed.

[0122] The quality assessment module extracts a multidimensional feature matrix based on the interference pattern data to evaluate the quality of the silicon carbide metasurface grating under test. If the lightwave and microwave sources are set according to the initial excitation parameters to generate a cooperative excitation signal and collect interference pattern data, the module extracts defect features from the interference pattern data for a preliminary quality assessment, determines the presence of potential defect areas, and identifies and divides these areas. If updated interference pattern data is collected based on an updated cooperative excitation signal, the module extracts defect features from the updated interference pattern data for an in-depth quality assessment.

[0123] See also Figure 4 , preferably, the specific steps of conducting preliminary quality assessment and dividing potential defect areas include:

[0124] A standard feature set is established based on multiple defect-free standard silicon carbide metasurface grating samples. The standard feature set contains the multidimensional feature matrices of multiple defect-free standard silicon carbide metasurface grating samples. The mean vector and covariance matrix of each defect feature are statistically analyzed, and a distribution model of the multidimensional feature matrix in the standard feature set is established as a feature benchmark model.

[0125] The multidimensional data matrix and multidimensional feature matrix of each collected interference pattern data are obtained. For each coordinate position in each multidimensional feature matrix, the distance between its defect feature vector and the feature reference model is calculated to generate a deviation matrix. The deviation matrix is ​​globally averaged and used as the deviation index of each interference pattern data. Distance measurement methods include Mahalanobis distance, Euclidean distance, etc.

[0126] According to the distance distribution of defect-free standard silicon carbide metasurface grating samples in the standard feature set, a distance threshold is configured, and the deviation index value of the multidimensional feature matrix of each interference pattern data is compared with the distance threshold. If the deviation index value of the multidimensional feature matrix of the interference pattern data is greater than the distance threshold, it is determined that the silicon carbide metasurface grating to be tested has defects; otherwise, it is determined that the silicon carbide metasurface grating to be tested has passed the test.

[0127] If it is determined that the silicon carbide metasurface grating to be tested has defects, the potential defect area is divided, including:

[0128] For each collected interference pattern data, based on the deviation matrix and the distance threshold, the coordinate points in the multidimensional feature matrix whose deviation index is greater than the distance threshold are marked as suspected defect points to form a binary abnormal point matrix.

[0129] For the interference pattern data collected multiple times, the frequency of occurrence of the same coordinate point in each round of binary outlier matrix is ​​counted. The frequency threshold is configured through experiments or historical data, and the coordinate points with an occurrence frequency greater than the frequency threshold are screened out as stable outliers. A stable outlier matrix is ​​generated to suppress accidental misjudgments caused by single acquisition noise.

[0130] Edge detection is used to extract continuous edge contours from the stable outlier matrix, retaining strong edges and connecting weak edges to ensure the integrity of the defect boundary. A sliding window is then set to perform a morphological dilation operation on the edge contours, connecting adjacent edge points and filling small gaps to form a closed candidate defect connected domain, aggregating discrete outliers into a physically meaningful continuous region.

[0131] Configure seed point ratio , in the candidate defect connectivity domain, sort by deviation index from high to low, and select the top ranked The coordinate points are used as seed points to ensure that the seed points are located at the most significant position of the boundary features between the defect and the normal area, providing a typical starting point for the growth of potential abnormal areas.

[0132] With the seed point as the center, the potential abnormal region is grown. The growth conditions are based on the light intensity distribution characteristics, microwave field characteristics and cross-modal correlation characteristics.

[0133] For the light intensity distribution feature, the difference in light intensity contrast index between the neighborhood pixels and the seed point is within the preset contrast range;

[0134] For microwave field characteristics, the difference in microwave phase stability index and field intensity attenuation rate characteristic index between the neighboring pixels and the seed point does not exceed the preset phase difference threshold and attenuation rate threshold;

[0135] For cross-modal correlation features, the differences in correlation indicators between the light intensity gradient and microwave phase gradient of the neighborhood pixels and the seed point are within the preset range.

[0136] Neighborhood pixels that meet the growth conditions are merged into the current potential anomaly area until no new neighbors meet the conditions or the potential anomaly area reaches the preset minimum detection unit size.

[0137] Configure an effective threshold and calculate the proportion of coordinates with both light intensity distribution anomalies and microwave field anomalies for each potential defect area. If the proportion is greater than the effective threshold, the potential defect area is confirmed as a valid potential defect area; otherwise, it is determined to be a pseudo potential defect area and eliminated to eliminate non-structural defect interference.

[0138] Preferably, the specific steps of deep quality assessment include:

[0139] The multidimensional data matrix and the multidimensional feature matrix of each updated interference pattern data are obtained, and the coordinates of the potential defect area are mapped into the updated multidimensional data matrix to generate a potential defect area mask.

[0140] Based on the potential defect area mask, the mean and standard deviation of the defect feature vectors in the potential defect area and non-potential defect area are calculated respectively to calculate the difference between each defect feature vector; the mean reflects the average performance of the potential defect area and non-potential defect area in each feature dimension, and the standard deviation measures the degree of dispersion of the feature value in the potential defect area and non-potential defect area.

[0141] Based on the calculated mean and standard deviation of the defect feature vectors for potential defect areas and non-potential defect areas, the defectivity of the defect feature vector for each potential defect area is calculated. The defectivity can be calculated using a variety of methods, such as Euclidean distance and cosine similarity. By calculating the defectivity, the degree of difference between potential defect areas and non-potential defect areas in various feature dimensions is quantified.

[0142] Defect levels are divided according to the defect degree of each defect feature vector in the potential defect area and the area of ​​the potential defect area. The defect levels are divided into mild defect level, moderate defect level, and severe defect level. Each level corresponds to a different defect degree range and defect characteristic performance.

[0143] Specifically, the steps for defect grade classification include:

[0144] The defect threshold and area ratio threshold are determined based on the multi-dimensional feature distribution of defect-free standard SiC metasurface grating samples and the specific requirements of actual testing. The defect threshold is divided into upper and lower defect thresholds, which are determined by statistically analyzing the characteristic differences between defect-free areas and known defect areas. The area ratio threshold, which includes upper and lower area ratio thresholds, is set based on the minimum functional unit size of the grating and its impact on performance.

[0145] To determine the level of minor defects, select potential defect regions whose defect feature vector defectivity is less than the lower defectivity threshold and whose area ratio is less than the lower area ratio threshold. Further verification is conducted to determine whether the defect region only exhibits a single abnormal modal characteristic, including deviations from the normal range for either the light intensity distribution or the microwave field characteristics. If these conditions are met, the defect is determined to be minor, and its specific location and deviation in the grating's physical coordinate system are recorded.

[0146] For the determination of moderate defect levels, potential defect areas with defect feature vectors greater than or equal to the lower defect threshold and less than the upper defect threshold, or potential defect areas with an area ratio greater than or equal to the lower area ratio threshold and less than the upper area ratio threshold, are extracted and determined to be moderate defects. For these areas, focus on checking whether at least two types of modal features are abnormal at the same time. For example, the light intensity distribution feature may have a stripe period deviation problem, while the microwave field feature also has field intensity attenuation abnormalities. In addition, it is necessary to check whether these abnormal areas are spatially correlated, that is, the overlapping area ratio of the abnormal light intensity area and the abnormal microwave field area.

[0147] For severe defect classification, potential defect areas with a defect feature vector defectivity greater than or equal to the upper defectivity threshold, or with an area ratio greater than or equal to the upper area ratio threshold, are identified as severe defects. These areas are inspected for abnormalities in the light intensity distribution, microwave field, and cross-modal correlation features, as well as for defect feature deviation.

[0148] Example 2

[0149] See also Figure 5This embodiment introduces a joint testing method for silicon carbide metasurface grating detection, including the following steps:

[0150] Step S1: obtaining initial excitation parameters of the lightwave excitation signal and the microwave excitation signal, regulating the lightwave source and the microwave source to generate a cooperative excitation signal, and applying the cooperative excitation signal to the silicon carbide metasurface grating to be measured to form an interference pattern including light intensity distribution and microwave field intensity distribution;

[0151] Step S2: Performing multi-channel synchronous multiple acquisitions of the interference pattern through a detector to obtain multiple interference pattern data, including light intensity distribution data and microwave field intensity data; aligning the interference pattern data in the time dimension, and mapping and matching the interference pattern data coordinates by extracting the characteristic points of the marker on the silicon carbide metasurface grating to be measured to construct a multidimensional data matrix;

[0152] Step S3: Defect features are extracted from the multidimensional data matrix to construct a multidimensional feature matrix. A feature reference model is established based on the multidimensional feature matrix of a defect-free standard silicon carbide metasurface grating sample. The deviation index of the silicon carbide metasurface grating to be tested is calculated to perform quality judgment. If a defect is determined to be present, the method then divides the potential defect area by screening stable outliers for multiple rounds, constructing a candidate defect connected domain through edge detection, performing region growing based on multimodal feature similarity conditions, and verifying the coupling between light intensity and microwave field anomalies.

[0153] Step S4: If a potential defect area is identified, the initial excitation parameters of the optical wave excitation signal and the microwave excitation signal are dynamically adjusted according to the information of the potential defect area, the collaborative excitation signal is updated, and the updated interference pattern data is obtained; the defect level is divided based on the defect degree of the defect area and the non-defect area in the updated multidimensional feature matrix and the area of ​​the potential defect area.

[0154] Preferably, the specific steps of updating and adjusting the collaborative excitation signal include:

[0155] Acquiring structural parameter data and electromagnetic characteristic data of the silicon carbide metasurface grating to be tested to determine initial excitation parameter values ​​of the optical wave excitation signal and the microwave excitation signal;

[0156] According to the determined initial excitation parameters, the lightwave source and the microwave source are set to generate a cooperative excitation signal;

[0157] Applying a cooperative excitation signal to the silicon carbide metasurface grating to be tested, collecting the resulting interference pattern data and performing preprocessing;

[0158] Extract defect features from the pre-processed interference pattern data, perform preliminary quality assessment based on the extracted defect features, and determine whether the silicon carbide metasurface grating to be tested has defects;

[0159] If it is determined that there is a defect, the potential defect area of ​​the silicon carbide metasurface grating to be tested is divided, the location and range information of the potential defect area is recorded, and the cooperative excitation signal control is triggered;

[0160] The objective function is to maximize the ratio of the edge gradient of the interference pattern data in the potential defect area to the noise standard deviation, which is used to control the excitation parameters of the optical wave excitation signal and the microwave excitation signal.

[0161] Within the adjustment range of the incentive parameters, the incentive parameters are coordinated and regulated by the gradient descent algorithm, specifically including:

[0162] Based on the initial excitation parameters of the optical wave excitation signal and the microwave excitation signal, the initial frequency difference and phase difference are determined, and the objective function value of the potential defect area is obtained;

[0163] Calculate the partial derivatives of the objective function with respect to the frequency difference and phase difference, and update the frequency difference and phase difference according to the gradient direction;

[0164] Based on the updated frequency difference and phase difference, the frequencies and phases of the optical wave excitation signal and the microwave excitation signal are recalculated, the updated optical wave excitation signal and the microwave excitation signal are synthesized into a new cooperative excitation signal, and the signal is applied to the silicon carbide metasurface grating to be tested to form an updated interference pattern, and the objective function value of the potential defect area is calculated;

[0165] Configure the convergence threshold and calculate the change in the objective function value between two adjacent times. If the change is less than the convergence threshold, stop updating and output the optimized excitation parameters; otherwise, continue iterative updating.

[0166] Preferably, the specific steps for conducting a preliminary quality assessment include:

[0167] Establishing a standard feature set, the standard feature set includes a multi-dimensional feature matrix of a defect-free standard silicon carbide metasurface grating sample;

[0168] Statistically calculate the mean vector and covariance matrix of each defect feature in the multidimensional feature matrix in the standard feature set, and construct a distribution model of the multidimensional feature matrix in the standard feature set as a feature benchmark model;

[0169] Obtain the interference pattern data of the silicon carbide metasurface grating to be tested, and construct a multidimensional data matrix and a multidimensional feature matrix. Calculate the distance between the defect feature vector at each coordinate position in the multidimensional feature matrix and the feature reference model to generate a deviation matrix.

[0170] Perform a global average of the deviation matrix to obtain the deviation index of each interference pattern data; configure a distance threshold and compare the deviation index with the distance threshold. If the deviation index is greater than the distance threshold, it is determined that the silicon carbide metasurface grating to be tested has a defect; otherwise, it is determined to be qualified;

[0171] If it is determined that the silicon carbide metasurface grating to be tested has defects, based on the deviation matrix and the distance threshold, the multidimensional data matrix of the interference pattern data collected multiple times is marked with coordinate points whose deviation index is greater than the distance threshold as suspected defect points, forming a binary abnormal point matrix, and the abnormal frequency of the same coordinate point is counted to screen out stable abnormal points whose abnormal frequency is greater than the preset frequency threshold;

[0172] Edge detection is performed on stable outliers to extract continuous edge contours, and a closed candidate defect connected domain is formed through morphological dilation operation;

[0173] Select seed points in the candidate defect connectivity domain by sorting by deviation index, and perform regional growth based on the light intensity distribution characteristics, microwave field characteristics, and cross-modal correlation characteristics with the seed points as the center until the preset growth termination conditions are met;

[0174] Configure the effective threshold and calculate the proportion of coordinates with both abnormal light intensity distribution and abnormal microwave field in each potential defect area. If the proportion is greater than the effective threshold, it is confirmed as a valid potential defect area; otherwise, it is eliminated.

[0175] Example 3

[0176] This embodiment introduces a testing device, including hardware equipment for generating cross-band excitation signals, a detection unit for realizing multimodal data acquisition, a computing unit for performing data processing and feature analysis, and a control actuator for completing defect detection and parameter regulation. These parts work together to realize joint testing of the silicon carbide metasurface grating to be tested.

[0177] Specifically, a testing device includes:

[0178] The lightwave signal source uses a tunable semiconductor laser that can output lightwave excitation signals with controllable wavelength, power, and polarization state. By adjusting the laser cavity length and drive current, precise control of the lightwave parameters can be achieved.

[0179] The microwave signal source adopts a high-stability microwave signal generator, which can generate microwave excitation signals with adjustable frequency, phase and power. Its output parameters are dynamically adjusted by receiving control instructions through a digital interface.

[0180] The trigger controller serves as the synchronization core, generating a high-precision synchronous clock signal and simultaneously triggering the optical wave signal source and the microwave signal source to ensure that their output signals are strictly synchronized in the time dimension, providing a benchmark for the subsequent time alignment of multimodal data.

[0181] The signal synthesizer integrates a free-space coupling device to synthesize light waves and microwave excitation signals into a cooperative excitation signal. The device focuses the light waves through a collimating lens and uses a microstrip antenna to realize the spatial radiation of the microwave signal, so that the synthesized cooperative excitation signal can be evenly applied to the silicon carbide metasurface grating to be tested, forming an interference pattern containing the light intensity distribution and the microwave field intensity distribution.

[0182] The photoelectric detection array uses a high-resolution camera with a microscope objective to image the grating surface, collect light intensity distribution data including the marker, and extract the feature points of the marker through an image recognition algorithm, providing a key reference for subsequent coordinate mapping.

[0183] The microwave scanning probe system consists of a three-dimensional translation stage and a near-field microwave probe. Under the precise control of the translation stage, the grating surface is scanned point by point, and the microwave field intensity data at each position is collected in real time and the coordinate information is recorded. During the scanning process, the distance between the probe and the grating surface is kept constant to ensure the consistency of microwave signal acquisition.

[0184] The data processing unit preprocesses the acquired interference pattern data, aligning the time dimension of the light intensity and microwave field intensity data using the timestamp of the synchronized clock signal. Edge detection and feature matching algorithms are then used to extract the sub-pixel coordinates of the markers from the light intensity distribution data. Combined with the microwave scan position information, a geometric calibration algorithm is used to uniformly map the light intensity distribution data coordinates and microwave scan coordinates to the grating physical coordinate system, constructing a multidimensional data matrix containing multimodal information. Light intensity distribution features, microwave field features, and cross-modal correlation features are then extracted from the multidimensional data matrix to generate standardized defect feature vectors, ultimately constructing a multidimensional feature matrix for quality assessment.

[0185] The defect detection unit uses a multi-dimensional feature matrix to perform potential defect area division and level assessment. A feature reference model is established based on a defect-free standard silicon carbide metasurface grating sample. The deviation matrix is ​​generated by calculating the distance between each coordinate point of the silicon carbide metasurface grating to be tested and the feature reference model, and suspected defect points with excessive deviation are marked. Then, stable outliers are screened out through multiple rounds of data statistics, and discrete outliers are aggregated into closed candidate defect connected domains using edge detection and morphological operations. Regional growth is performed based on multi-modal feature similarity conditions, and the coordinate ratio of light intensity and microwave field anomalies is combined to verify the regional validity and divide the real potential defect area. For the identified potential defect area, the defect level is divided into mild, moderate and severe according to the defect degree of the defect feature vector, the regional area ratio and the spatial overlap of cross-modal anomalies. Among them, the defects in key functional areas are automatically upgraded to reflect the significant impact on performance.

[0186] Based on the potential defect area information output by the defect detection unit, the parameter control actuator drives the parameter adjustment modules of the lightwave and microwave signal sources. Using a gradient descent algorithm to iteratively optimize the excitation signal parameters, the parameter control actuator dynamically adjusts the frequency, phase, and power differences between the lightwave and microwave sources, enhancing the contrast of defect features and improving the sensitivity and accuracy of defect detection, with the objective function being to maximize the ratio of the edge gradient to the noise standard deviation of the interference pattern in the potential defect area.

[0187] Working principle and its effect:

[0188] The working principle and effect of this application are as follows: by regulating the light wave source and the microwave source to generate a cooperative excitation signal containing light waves and microwaves and applying it to the silicon carbide metasurface grating to be tested, the cross-band coupling of light waves and microwaves is used to form an interference pattern reflecting the grating structure and electromagnetic characteristics. The interference pattern also contains information on the light intensity distribution and the microwave field intensity distribution, breaking through the limitation of insufficient response of single-mode detection to nanoscale defects, and realizing the comprehensive excitation of the cross-band electromagnetic response characteristics of the grating. The interference pattern data is collected synchronously multiple times through multiple channels of the detector, and the time dimension of the light intensity and microwave field data is aligned by the synchronized clock signal timestamp. The light intensity pixel coordinates and microwave scanning coordinates are uniformly mapped to the grating physical coordinate system through the marker feature point extraction and geometric calibration algorithm, and a multidimensional data matrix integrating multimodal information is constructed, which solves the problem of time and space asynchrony and coordinate mismatch of multi-source data in traditional detection, and provides a high-precision data foundation with consistent time and space for subsequent analysis.

[0189] Defect features such as light intensity distribution, microwave field, and cross-modal correlation are extracted from a multidimensional data matrix. A multidimensional feature matrix is ​​constructed, and a feature benchmark model of a defect-free standard silicon carbide metasurface grating sample is established. Quality is determined by calculating the deviation index between the grating to be tested and the feature benchmark model. This method can accurately identify weak anomalies that are difficult to detect with a single modality. For gratings determined to have defects, stable anomalies are screened through multiple rounds of data collection, edge detection is used to construct candidate defect connected domains, and region growth is performed based on multimodal feature similarity conditions. The validity of the potential defect region is verified by combining the coordinate ratio of light intensity and microwave field anomalies. Noise interference is effectively suppressed and pseudo potential defect regions are eliminated, achieving accurate positioning of real defects.

[0190] When a potential defect area is identified, the objective function is to maximize the ratio of the edge gradient of the interference pattern in the defect area to the noise standard deviation. A gradient descent algorithm is used to dynamically adjust excitation parameters such as the frequency and phase difference between the light and microwave waves to generate a targeted, enhanced, synergistic excitation signal. This optimizes the period, phase distribution, and defect feature contrast of the interference pattern in real time, forming a closed "detection-feedback-control" loop. Based on the updated interference pattern data, the characteristic defect degree and area of ​​the defective and non-defective areas are calculated, and the defect level is classified as mild, moderate, or severe, achieving a quantitative assessment of the degree of defect impact.

[0191] The overall solution breaks through the bottlenecks of traditional detection technology in efficiency, accuracy and environmental adaptability through multi-modal collaborative excitation, multi-source data fusion, dynamic parameter optimization and graded evaluation. It can efficiently capture nanoscale defects and accurately evaluate their impact on grating performance. It significantly improves the detection reliability and scientific grading of the silicon carbide metasurface grating to be tested in complex electromagnetic environments and extreme conditions, and provides advanced technical means for quality control in the manufacturing of high-end optoelectronic devices.

[0192] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiment. All technical solutions based on the concept of the present invention are within the scope of protection of the present invention. It should be noted that for those skilled in the art, various improvements and modifications that do not depart from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. Joint testing system, characterized in that, include Obtaining initial excitation parameters of the lightwave excitation signal and the microwave excitation signal, regulating the lightwave source and the microwave source to generate a cooperative excitation signal, wherein the cooperative excitation signal is applied to the silicon carbide metasurface grating to be measured to form an interference pattern; Performing multi-channel synchronous multiple acquisitions on the interference pattern through a detector to obtain multiple interference pattern data and perform pre-processing, wherein the interference pattern data includes light intensity distribution data and microwave field intensity data; The preprocessing is used to align the interference pattern data in the time dimension, and extract feature points of the markers configured on the silicon carbide metasurface grating to be tested to map and match the interference pattern data coordinates to construct a multidimensional data matrix; Extracting defect features from the multidimensional data matrix to construct a multidimensional feature matrix, establishing a feature benchmark model, calculating the deviation index of the silicon carbide metasurface grating to be tested to perform a preliminary quality assessment, and screening the silicon carbide metasurface grating to be tested for defects through multiple rounds of stable abnormal point screening, edge detection to construct candidate defect connected domains, performing region growing, and combining abnormal coupling verification to divide potential defect areas; If a potential defect area is identified, the initial excitation parameters of the optical wave excitation signal and the microwave excitation signal are dynamically adjusted based on the information of the potential defect area to update and adjust the interference pattern regulated by the collaborative excitation signal, obtain the updated interference pattern data and perform in-depth quality assessment to classify the defect level.

2. The joint testing system according to claim 1, wherein: The specific steps of updating and adjusting the cooperative excitation signal include: Acquire the structural parameter data and electromagnetic characteristic data of the silicon carbide metasurface grating to be tested to determine the initial excitation parameters of the optical wave excitation signal and the microwave excitation signal; According to the determined initial excitation parameters, the lightwave source and the microwave source are set to generate a cooperative excitation signal; Applying the cooperative excitation signal to the silicon carbide metasurface grating to be measured, collecting the interference pattern data formed and preprocessing it; Extract defect features from the pre-processed interference pattern data, perform preliminary quality assessment based on the extracted defect features, and determine whether the silicon carbide metasurface grating to be tested has defects; If it is determined that there is a defect, the potential defect area of ​​the silicon carbide metasurface grating to be tested is divided, the location and range information of the potential defect area is recorded, and the cooperative excitation signal control is triggered; The specific steps of updating and adjusting the cooperative excitation signal further include: The objective function is to adjust the excitation parameters of the optical wave excitation signal and the microwave excitation signal by maximizing the ratio of the edge gradient of the interference pattern data of the potential defect area to the noise standard deviation; Within the adjustment range of the excitation parameters, the excitation parameters are collaboratively controlled by the gradient descent algorithm, specifically including: Determining an initial frequency difference and phase difference based on initial excitation parameters of the optical wave excitation signal and the microwave excitation signal, and obtaining an objective function value of a potential defect area; Calculating the partial derivatives of the objective function with respect to the frequency difference and the phase difference, and updating the frequency difference and the phase difference according to the gradient direction; recalculating the frequencies and phases of the optical wave excitation signal and the microwave excitation signal based on the updated frequency difference and phase difference, synthesizing the updated optical wave excitation signal and the microwave excitation signal into a new cooperative excitation signal, applying the signal to the silicon carbide metasurface grating to be tested to form an updated interference pattern, and calculating the objective function value of the potential defect area; Configure the convergence threshold and calculate the change in the objective function value between two adjacent times. If the change is less than the convergence threshold, stop updating and output the optimized excitation parameters; otherwise, continue iterative updating.

3. The joint testing system according to claim 1, wherein: The specific steps of constructing the multidimensional data matrix include: Using the timestamp of the synchronous clock signal, the light intensity distribution data collected by the photoelectric detector and the microwave field intensity data collected by the microwave detector are aligned in the time dimension; After removing noise from the light intensity distribution data and enhancing the contrast, a scale space is constructed to detect extreme points in the enhanced light intensity distribution data, determine the locations of feature points of the silicon carbide metasurface grating marker to be measured, calculate the scale and orientation information of each feature point, and generate a feature descriptor; A threshold segmentation method is used for microwave field intensity data. The segmentation threshold is set to separate the marker area. The outline of the marker area is optimized through morphological processing. The geometric center of the marker area is calculated as the feature point, and the coordinate position of the feature point in the microwave field intensity data is recorded. Using a geometric calibration algorithm, based on the feature descriptors of the feature points in the light intensity distribution data and the coordinate positions of the feature points in the microwave field intensity data, the coordinates of the light intensity distribution data are mapped to the physical coordinate positions of the silicon carbide metasurface grating to be measured, and the coordinates of the microwave field intensity data are matched with the coordinates of the light intensity distribution data; Selecting a fixed position on the silicon carbide metasurface grating to be measured as the origin of a spatial coordinate system and determining the direction of the coordinate axis, and converting the coordinates of the light intensity distribution data and the microwave field intensity data into coordinates in the spatial coordinate system; The light intensity distribution data and the microwave field intensity data are integrated according to the spatial coordinate system to construct a multidimensional data matrix, wherein the elements of the multidimensional data matrix correspond to the light intensity distribution data and the microwave field intensity data of the corresponding positions in the silicon carbide metasurface grating to be measured.

4. The joint testing system according to claim 1, wherein: The specific steps of extracting defect features include: Obtaining a preprocessed multidimensional data matrix, and extracting corresponding defect features for a local area at each coordinate position in the multidimensional data matrix, wherein the defect features include light intensity distribution features, microwave field features, and cross-modal correlation features; The light intensity distribution feature is extracted by: demarcating adjacent areas with the coordinate position of the multidimensional data matrix as the center and calculating the light intensity contrast index; comparing the actual fringe period corresponding to the coordinate position of the multidimensional data matrix with the configured theoretical period, and calculating the deviation rate as the fringe spacing uniformity index; calculating the gradient amplitude of the edge position of the multidimensional data matrix and calculating the standard deviation of the gradient amplitude of adjacent edge positions as the edge gradient consistency index; The microwave field feature is extracted by dividing a local area centered on a coordinate position in the multidimensional data matrix into an analysis unit, calculating the variance of the microwave phase value in the local area as a microwave phase stability index; comparing the microwave field intensity in the local area with the standard microwave field intensity, and calculating the ratio of the microwave field intensity to the standard microwave field intensity as a field intensity attenuation rate characteristic index; The cross-modal correlation feature is extracted by calculating the light intensity gradient and microwave phase gradient respectively for the light intensity distribution data and microwave field intensity data corresponding to the coordinate position in the multidimensional data matrix, and calculating the correlation index of the light intensity gradient and the microwave phase gradient through the correlation coefficient; The extracted defect features are dedimensionalized so that each coordinate position in the multidimensional data matrix generates a set of defect feature vectors to construct a multidimensional feature matrix.

5. The joint testing system according to claim 2, wherein: The specific steps for conducting the preliminary quality assessment include: Establishing a standard feature set, wherein the standard feature set includes a multi-dimensional feature matrix of a defect-free standard silicon carbide metasurface grating sample; Counting the mean vector and covariance matrix of each defect feature in the multidimensional feature matrix in the standard feature set, and constructing a distribution model of the multidimensional feature matrix in the standard feature set as a feature benchmark model; Obtaining interference pattern data of the silicon carbide metasurface grating to be tested, and constructing a multidimensional data matrix and a multidimensional feature matrix, calculating the distance between the defect feature vector at each coordinate position in the multidimensional feature matrix of the silicon carbide metasurface grating to be tested and the feature reference model, and generating a deviation matrix; The deviation matrix is ​​globally averaged to obtain a deviation index for each interference pattern data; a distance threshold is configured, and the deviation index is compared with the distance threshold. If a deviation index is greater than the distance threshold, it is determined that the silicon carbide metasurface grating to be tested has defects, otherwise it is determined to be qualified.

6. The joint testing system according to claim 5, wherein: The specific steps of conducting the preliminary quality assessment also include: If it is determined that the silicon carbide metasurface grating to be tested has defects, based on the deviation matrix and the distance threshold, the multidimensional data matrix of the interference pattern data collected multiple times is marked, and the coordinate points with deviation indicators greater than the distance threshold are marked as suspected defect points, forming a binary abnormal point matrix, and the abnormal occurrence frequency of the same coordinate point is counted, and stable abnormal points with abnormal occurrence frequency greater than the preset frequency threshold are screened out; Perform edge detection on the stable outlier points to extract continuous edge contours, and form candidate defect connected domains through morphological dilation operations; Select seed points in the candidate defect connected domain according to the deviation index, and perform regional growth based on the light intensity distribution characteristics, microwave field characteristics, and cross-modal correlation characteristics with the seed points as the center until a preset growth termination condition is met; Configure the effective threshold and calculate the proportion of coordinates with both abnormal light intensity distribution and abnormal microwave field in each potential defect area. If the proportion is greater than the effective threshold, it is confirmed as a valid potential defect area; otherwise, it is eliminated.

7. The joint testing system according to claim 1, wherein: The specific steps of the depth quality assessment include: Obtaining a multidimensional data matrix and a multidimensional feature matrix of each updated interference pattern data, mapping the coordinates of the potential defect area to the updated multidimensional data matrix, and generating a potential defect area mask; Based on the potential defect area mask, respectively calculating the mean and standard deviation of the defect feature vectors in the potential defect area and the non-potential defect area, and calculating the difference between the defect feature vectors; Calculate the defect degree of the defect feature vector of each potential defect area according to the calculated mean and standard deviation of the defect feature vectors of the potential defect area and the non-potential defect area; According to the defect degree of each defect feature vector in the potential defect area and the area of ​​the potential defect area, the defect level is divided into a mild defect level, a moderate defect level, and a severe defect level.

8. A method for detecting silicon carbide metasurface gratings by joint testing, which is implemented based on the joint testing system according to any one of claims 1 to 7, characterized in that: The following steps are involved: Step S1: obtaining initial excitation parameters of the lightwave excitation signal and the microwave excitation signal, regulating the lightwave source and the microwave source to generate a cooperative excitation signal, and applying the cooperative excitation signal to the silicon carbide metasurface grating to be measured to form an interference pattern including light intensity distribution and microwave field intensity distribution; Step S2: performing multi-channel synchronous acquisition of the interference pattern multiple times through a detector to obtain a plurality of interference pattern data, wherein the interference pattern data includes light intensity distribution data and microwave field intensity data; Aligning the interference pattern data in the time dimension, and mapping and matching the interference pattern data coordinates by extracting the characteristic points of the marker on the silicon carbide metasurface grating to be measured to construct a multidimensional data matrix; Step S3: extracting defect features from the multidimensional data matrix to construct a multidimensional feature matrix, establishing a feature reference model based on the multidimensional feature matrix of a defect-free standard silicon carbide metasurface grating sample, and calculating the deviation index of the silicon carbide metasurface grating to be tested for quality judgment; If a defect is determined to exist, the system will use multiple rounds of screening for stable outliers, edge detection to construct candidate defect connected domains, and region growing based on multimodal feature similarity conditions. The system will then verify the coupling between light intensity and microwave field anomalies to delineate potential defect areas. Step S4: If the potential defect area is identified, the initial excitation parameters of the optical wave excitation signal and the microwave excitation signal are dynamically adjusted according to the information of the potential defect area, the collaborative excitation signal is updated and the updated interference pattern data is obtained; the defect level is divided based on the defect degree of the defect area and the non-defect area in the updated multidimensional feature matrix and the area of ​​the potential defect area.

9. A testing device for implementing the joint testing system according to any one of claims 1 to 7, characterized in that: include: Lightwave signal source, used to generate lightwave excitation signals of different wavelengths, powers, and polarization states; A microwave signal source, used to generate microwave excitation signals of different frequencies, phases, and powers; A trigger controller is used to generate a synchronous clock signal to achieve time synchronization between the lightwave signal source and the microwave signal source; A signal synthesizer, used to synthesize the optical wave excitation signal and the microwave excitation signal into a cooperative excitation signal; Photoelectric detection array, used to collect light intensity distribution data and identify the characteristic points of the marker; Microwave scanning probe system, used to collect microwave field intensity data and record scanning coordinates; The data processing unit is used to execute the marker feature extraction and coordinate mapping algorithm, and construct a multidimensional data matrix, extract the light intensity distribution characteristics, microwave field characteristics and cross-modal correlation characteristics, generate defect feature vectors, and construct a multidimensional feature matrix; Defect detection unit, used to perform potential defect area division and classify defect levels based on characteristic defectivity, area ratio, and cross-modal overlap; The parameter control actuator is used to dynamically adjust the excitation parameters according to the information of the potential defect area and optimize the excitation parameters through the gradient descent algorithm.

Images