Rapid detection method and system for line deviation of micro-nano grating

Abstract

The invention relates to the technical field of optical detection, in particular to a method and a system for rapidly detecting line deviation of a micro-nano grating. According to the invention, a structured composite light field is constructed by adopting a coding illumination technology, and single-channel detection limitation is broken through through multi-channel parallel illumination; obtaining composite diffraction spot distribution data based on Fourier space multi-frame imaging; obtaining coherence and time correlation characteristics which cannot be obtained by traditional intensity measurement by using a second-order intensity correlation function; constructing a joint forward physical model, reconstructing grating complex amplitude distribution and line edge roughness parameters through an iterative inversion algorithm, and carrying out micro-nano grating line deviation detection; according to the method, multi-channel coded illumination and second-order intensity correlation function analysis are combined, the measurement time is shortened through parallel detection, the detection reliability is improved by utilizing inter-channel redundant information, and the detection efficiency is remarkably improved while the precision is ensured.

Description

Technical Field

[0001] This invention relates to the field of optical detection technology, and in particular to a method and system for rapid detection of deviations in micro / nano grating lines. Background Technology

[0002] With the continuous advancement of semiconductor manufacturing process technology nodes and the sustained increase in integrated circuit functional density, micro / nano gratings, as core structural units of photolithography masks, optical components, and microelectronic devices, directly determine the optical performance, electrical characteristics, and manufacturing yield of these devices through their geometric dimensional accuracy and morphological quality. As process nodes evolve towards 7nm, 5nm, and even more advanced processes, the tolerance requirements for key parameters such as linewidth, spacing, and edge roughness of micro / nano gratings are becoming increasingly stringent. Traditional direct measurement techniques based on scanning electron microscopy (SEM) and atomic force microscopy (AFM) face severe challenges in terms of detection speed, sample damage, and large-area detection capabilities. First, existing detection methods mostly employ contact or destructive measurement modes. Electron beam irradiation and probe contact during the detection process can easily cause irreversible damage to micro- and nano-structures. Furthermore, single-point measurement methods result in low detection efficiency, making it difficult to meet the needs of online real-time monitoring. Second, traditional optical detection techniques are constrained by the diffraction limit, limiting their resolution of subwavelength-scale micro- and nano-gratings. They cannot accurately extract fine morphological features such as line edge roughness and sidewall angles, and lack sensitivity analysis for implicit geometric deviations. In addition, existing detection systems are mostly limited to the measurement of single physical quantities, failing to achieve the collaborative extraction of multi-dimensional grating line deviation information and global quality assessment.

[0003] For example, Chinese patent CN109031894B discloses a method for detecting the bottom morphology of phase-type defects in multilayer films using extreme ultraviolet (EUV) lithography masks. This method involves depositing a multilayer film structure on a mask substrate, utilizing the modulation effect of phase-type defects on the phase of reflected light, and combining this with optical microscopy imaging technology to detect the bottom morphology features of the defects. This patent primarily addresses the problem of insufficient sensitivity of traditional detection methods to buried defects, achieving effective identification and localization of phase-type defects within multilayer films.

[0004] In addition, Chinese patent CN102222329B discloses a grating scanning method for depth detection. This technology scans light onto an object using a grating and detects the light reflected from the object. Based on the time-of-flight principle or spatial difference analysis, it determines the distance information to the object and generates a 3D map of the object. This patent mainly solves the problem of insufficient ranging accuracy of traditional depth cameras in complex scenes, and achieves high-precision 3D reconstruction based on grating scanning.

[0005] The existing technologies mentioned above have the following shortcomings in the detection of micro / nano grating line deviations: limited detection parameter dimensions, making it impossible to achieve collaborative analysis of multi-dimensional grating line deviations; limited information extraction methods, failing to fully utilize the coherence characteristics and spatiotemporal correlation information of the light field; and insufficient physical modeling capabilities, lacking a precise inversion mechanism from diffraction signals to geometric parameters. Therefore, existing solutions still suffer from significant drawbacks in addressing device performance degradation caused by deviations in advanced process micro / nano grating lines, including limited detection accuracy, insufficient detection speed, and a lack of multi-parameter collaborative analysis capabilities. To solve these systemic technical challenges, this invention proposes a rapid detection method and system for micro / nano grating line deviations. Summary of the Invention

[0006] This invention employs coded illumination technology to construct a structured composite light field, overcoming the limitations of traditional single-channel detection. It acquires composite diffraction spot distribution data through multi-frame Fourier space imaging and utilizes a second-order intensity correlation function to deeply mine physical layer information. Furthermore, it constructs a joint forward physical model and reconstructs the grating complex amplitude distribution and line edge roughness parameters from the diffraction information using an iterative inversion algorithm. Finally, it establishes a two-dimensional deviation distribution map and combines it with anomaly region identification to achieve global quality assessment. This invention integrates multi-channel coded illumination with second-order intensity correlation function analysis, significantly improving detection speed and result reliability while ensuring detection accuracy.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A method for rapid detection of deviations in micro / nano grating lines, the method comprising:

[0009] The distribution data of the composite diffraction spot generated by the micro / nano grating under test in the Fourier plane is acquired by multi-frame continuous imaging, and the first-order intensity sequence and its time fluctuation data of the composite diffraction spot distribution data are calculated.

[0010] Based on the first-order intensity sequence and its time fluctuation data, the actual second-order intensity correlation function characteristic of each lighting channel is calculated through channel separation and registration processing.

[0011] The actual second-order intensity correlation function features are input into a preset joint forward physical model, and the reconstructed grating complex amplitude distribution and line edge roughness power spectrum parameters are obtained through iterative inversion.

[0012] Based on the reconstructed complex amplitude distribution of the grating and the power spectrum parameters of the line edge roughness, the grating line deviation at the current coordinate position is calculated.

[0013] Before acquiring composite diffraction spot distribution data, the method further includes:

[0014] The incident light is spatially structured and modulated using coded illumination technology to generate a structured composite light field comprising multiple coherent channels. This spatially structured modulation includes phase-coded modulation and amplitude-coded modulation, wherein:

[0015] The phase-coded modulation applies a differentiated phase delay to different spatial regions of the incident light by loading a preset phase mask pattern to generate a first illumination channel, wherein the phase distribution of the first illumination channel is adjusted according to the phase delay distribution of the phase mask pattern.

[0016] The amplitude coding modulation selectively transmits the incident light to different spatial regions by loading a preset amplitude mask pattern to generate a second illumination channel, wherein the amplitude distribution of the second illumination channel is adjusted by the transmittance distribution of the amplitude mask pattern.

[0017] Based on the first and second illumination channels, multiple parallel illumination channels with preset spatial spectrum distributions are generated, wherein the coherence characteristics of the parallel illumination channels are different from each other. Based on the parallel illumination channels and coherence characteristics, a structured composite light field is generated by spatial superposition.

[0018] The process of acquiring composite diffraction spot distribution data generated by the micro / nano grating under test in the Fourier plane through multi-frame continuous imaging, and calculating the first-order intensity sequence and its temporal fluctuation data of the composite diffraction spot distribution data, includes:

[0019] Based on the composite diffraction spot distribution data under the structured composite light field illumination, continuous multi-frame real-time acquisition is performed through preset sampling frequency and total frame count requirements to obtain continuous multi-frame Fourier plane images of the composite diffraction spot distribution data.

[0020] Based on the spatial distribution functions of the first and second illumination channels, the diffraction spot regions corresponding to each illumination channel are extracted from the multi-frame Fourier plane images using an image segmentation algorithm. Based on the diffraction spot regions, the integrated light intensity value of each diffraction spot region at each time is calculated using an integration operation method.

[0021] Based on the integrated light intensity value, a light intensity response sequence for each illumination channel is constructed in chronological order to form a first-order intensity sequence that includes at least two illumination channels.

[0022] Based on the first-order intensity sequence, the intensity variance and time autocorrelation function of each lighting channel are calculated using statistical analysis methods. Based on the intensity variance and time autocorrelation function, the time fluctuation data of each lighting channel are obtained.

[0023] The step involves calculating the actual second-order intensity correlation function characteristic quantities of each lighting channel based on the first-order intensity sequence and its temporal fluctuation data, through channel separation and registration processing, including:

[0024] Based on the center position coordinates and spatial distribution function of the first and second lighting channels, the theoretical diffraction position of each lighting channel in the Fourier plane is determined by a spatial positioning algorithm.

[0025] The multi-frame Fourier plane images are processed by channel separation. The diffraction spot regions corresponding to each illumination channel are identified by the image segmentation algorithm. The actual center position of each diffraction spot region is located by the sub-pixel level registration algorithm.

[0026] Based on the actual center position and the theoretical diffraction position, the first-order intensity sequence is corrected using a displacement correction algorithm to obtain the corrected first-order intensity sequence.

[0027] Based on the corrected first-order intensity sequence, the second-order intensity correlation function between any two illumination channels is calculated through cross-correlation operation, and the peak value, full width at half maximum (FWHM), decay time constant, and coherence of the second-order intensity correlation function are extracted as feature parameters.

[0028] Based on the aforementioned characteristic parameters, combined with the coherence characteristics and relative light intensity weights of the illumination channel, the actual second-order intensity correlation function characteristic quantities are obtained through normalization processing.

[0029] The step of inputting the actual second-order intensity correlation function feature quantity into a preset joint forward physical model, and obtaining the reconstructed grating complex amplitude distribution and line edge roughness power spectrum parameters through iterative inversion includes:

[0030] Based on the center position coordinates, spatial distribution function, coherence characteristics and relative light intensity weights of the first and second illumination channels, a joint forward physical model of micro-nano grating diffraction under coded illumination is constructed based on scalar diffraction theory.

[0031] Based on the joint forward physical model, a mathematical mapping relationship is constructed from the complex amplitude distribution of the grating, the power spectrum parameters of the line edge roughness to the characteristic quantities of the theoretical second-order intensity correlation function;

[0032] Based on the mathematical mapping relationship and the actual second-order intensity correlation function characteristic quantities, the residual between the theoretical second-order intensity correlation function characteristic quantities and the actual second-order intensity correlation function characteristic quantities is minimized through a nonlinear least squares iterative inversion algorithm, thereby obtaining the reconstructed grating complex amplitude distribution and line edge roughness power spectrum parameters.

[0033] The nonlinear least squares iterative inversion algorithm includes:

[0034] Based on the preset initial grating complex amplitude distribution and initial line edge roughness power spectrum parameters, the characteristic quantity of the theoretical second-order intensity correlation function is calculated through the joint forward physical model.

[0035] The theoretical second-order intensity correlation function eigenvalues ​​and the actual second-order intensity correlation function eigenvalues ​​are compared to obtain the residual vector, and the corresponding Jacobian matrix is ​​calculated using the numerical differentiation method.

[0036] Based on the residual vector and Jacobian matrix, the parameter update amount is calculated by the least squares method. Based on the parameter update amount, the updated grating complex amplitude distribution and line edge roughness power spectrum parameters are obtained.

[0037] Based on the updated grating complex amplitude distribution and line edge roughness power spectrum parameters, the iteration is repeated. When the relative change in the L2 norm of the residual vector is less than a preset threshold, the iteration is terminated and the final grating complex amplitude distribution and line edge roughness power spectrum parameters are output.

[0038] Based on the reconstructed complex amplitude distribution of the grating and the power spectrum parameters of the line edge roughness, the grating line deviation at the current coordinate position is calculated, including:

[0039] Based on the reconstructed complex amplitude distribution of the grating, the amplitude distribution and phase distribution of the micro / nano grating under test are extracted, and the transmittance distribution and phase delay distribution of the micro / nano grating under test are calculated. Through threshold segmentation and edge detection algorithms, the geometric contour and boundary position information of the grating lines are extracted.

[0040] Based on the geometric contour and boundary position information of the grating lines, the position deviation of the micro-nano grating lines at the current coordinate position is calculated, and the statistical results of the grating line deviation are obtained through statistical analysis.

[0041] Based on the line edge roughness power spectrum parameters, the line edge roughness characteristic parameters are calculated by function integration.

[0042] Based on the statistical results of the grating line deviation and the line edge roughness characteristic parameters, the grating line deviation at the current coordinate position is obtained through a preset quality standard.

[0043] After calculating the grating line deviation at the current coordinate position, the method further includes:

[0044] Multiple preset scanning positions of the micro / nano grating under test are scanned. A deviation distribution map is constructed based on the grating line deviation and edge roughness index at each scanning position. The grating line deviation detection result is obtained through abnormal region identification, and a corresponding adjustment command is generated. The construction of the deviation distribution map based on the grating line deviation and edge roughness index at each scanning position includes:

[0045] Based on multiple preset scanning positions, a two-dimensional regular grid covering the entire detection area is constructed;

[0046] Based on the two-dimensional regular grid and multiple scanning positions, the grating line deviation and line edge roughness characteristic parameters are obtained, and the interpolated regular grid is obtained through a spatial interpolation algorithm.

[0047] Based on the interpolated regular grid, a line width deviation distribution map, a position deviation distribution map, a spacing deviation distribution map, and a line edge roughness distribution map are constructed. A two-dimensional deviation distribution map is then constructed using a weighted fusion algorithm.

[0048] The step of obtaining grating line deviation detection results through abnormal region identification and generating corresponding adjustment instructions includes:

[0049] Based on the two-dimensional deviation distribution map, an anomaly threshold is set, and abnormal grid nodes are located and merged into connected abnormal regions through spatial clustering.

[0050] Based on the area and shape characteristics of the connected abnormal regions, the anomaly type is determined through pattern recognition;

[0051] A re-examination strategy is formulated based on the anomaly type to obtain the re-examination result. The re-examination strategy includes performing local high-density scanning and extending the exposure time.

[0052] Based on the re-inspection results, the true abnormal areas are determined through confidence testing, and the grating line deviation detection results and final adjustment instructions are generated.

[0053] A rapid detection system for micro / nano grating line deviation is provided to implement a rapid detection method for micro / nano grating line deviation. The rapid deviation detection system includes a light source illumination module, an encoding and modulation module, a scanning control module, an optical imaging module, and a data processing module, wherein:

[0054] The light source illumination module is used to provide a partially coherent light source;

[0055] The encoding and modulation module includes a phase modulator and an amplitude modulator, used to perform phase encoding modulation and amplitude encoding modulation on the incident light to generate a first illumination channel and a second illumination channel;

[0056] The scanning control module includes a scanning drive unit and a position configuration unit. The scanning drive unit is used to control the structured composite light field to perform fixed-point scanning at multiple preset scanning positions of the micro-nano grating under test. The position configuration unit is used to output preset position coordinates to provide a position reference for fixed-point scanning.

[0057] The optical imaging module includes a Fourier lens, a channel-separating optical element, and an imaging detector. The Fourier lens is used to focus multi-level diffracted light onto the Fourier plane. The channel-separating optical element is used to separate the diffracted light of each parallel illumination channel. The imaging detector is used to acquire composite diffraction spot distribution data of the Fourier plane in multiple frames.

[0058] The data processing module is used to perform rapid detection of grating line deviation, output the grating line deviation detection result, and generate the final adjustment instruction.

[0059] Compared with the prior art, the beneficial effects of the present invention are:

[0060] This invention utilizes multi-channel coded illumination technology and second-order intensity correlation function analysis, combined with joint forward physical model inversion, to obtain coherent characteristics of the diffraction field that are unavailable in traditional single-channel detection. Fourier spatial imaging and iterative inversion algorithms are employed to achieve accurate reconstruction of the grating complex amplitude distribution and line edge roughness parameters. Two-dimensional deviation distribution maps and anomaly region identification significantly improve detection reliability. Compared to existing methods, this invention overcomes the information acquisition limitations of traditional single-channel detection, avoids the time consumption of point-by-point scanning, and achieves large-area parallel detection, greatly improving detection efficiency while maintaining high accuracy. Attached Figure Description

[0061] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0062] Figure 1 This is a flowchart illustrating a rapid detection method for micro / nano grating line deviation in an embodiment of the present invention.

[0063] Figure 2 This is a flowchart illustrating the system scanning process at multiple preset scanning positions of the micro / nano grating to be measured, as described in an embodiment of the present invention.

[0064] Figure 3 This is a schematic diagram of the generation of structured composite light field in an embodiment of the present invention;

[0065] Figure 4 This is an overall framework diagram of a micro / nano grating line deviation rapid detection system according to an embodiment of the present invention. Detailed Implementation

[0066] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations thereof.

[0067] The term "and / or" in the following text is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. Additionally, the character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0068] This application provides an optical detection method, including:

[0069] The distribution data of the composite diffraction spot generated by the sample under test in the Fourier plane is acquired by multi-frame continuous imaging, and the first-order intensity sequence and its time fluctuation data of the composite diffraction spot distribution data are calculated.

[0070] Based on the first-order intensity sequence and its time fluctuation data, the actual second-order intensity correlation function characteristic of each lighting channel is calculated through channel separation and registration processing.

[0071] The actual second-order intensity correlation function features are input into a preset joint forward physical model, and the reconstructed sample structure parameters and surface feature parameters are obtained through iterative inversion.

[0072] Based on the reconstructed sample structure parameters and surface feature parameters, the sample feature deviation at the current coordinate position is calculated.

[0073] For example, this embodiment uses the detection of linewidth deviation of micro-nano gratings as an example to specifically illustrate the above optical detection method.

[0074] In existing technologies, the detection of deviations in micro / nano grating lines typically employs two methods. The first method uses traditional single-channel optical imaging, which involves vertically illuminating the grating and then using image processing algorithms to extract the edge information of the grating lines to calculate the linewidth deviation. This method ensures that, with a relatively regular grating structure, basic geometric dimension measurement functions can be achieved while maintaining system simplicity. However, this method suffers from insufficient sensitivity to minute structural changes, and the system's detection accuracy is primarily affected by the imaging resolution and the accuracy of the edge extraction algorithm. Insufficient sensitivity severely impacts the system's detection performance under nanometer-level precision requirements.

[0075] Method two involves using a contact detection method based on a scanning probe microscope to measure grating line deviation. This method acquires the three-dimensional topographic information of the grating surface by scanning point by point with a probe. Although this method theoretically has extremely high spatial resolution, in practical applications, it is extremely slow due to the need for point-by-point mechanical scanning and complex data processing. Furthermore, it remains inefficient and susceptible to environmental vibration interference when detecting large areas.

[0076] Therefore, existing technologies often struggle to achieve a perfect balance between improving detection accuracy, reducing detection time, and minimizing system complexity. This is especially true in applications such as micro-nano manufacturing, where high efficiency and accuracy are required. Existing grating deviation detection methods are unable to meet the dual demands of rapid detection and high-precision measurement.

[0077] To achieve efficient and accurate detection of grating line deviations in micro-nano manufacturing lines and improve the adaptability of the detection system to complex grating structures, while also maximizing product yield through rapid response to quality anomalies, this application provides a rapid detection method for micro-nano grating line deviations based on the aforementioned optical detection method. Figure 1 As shown, the execution steps of this method include:

[0078] S1: Acquire the composite diffraction spot distribution data generated by the micro / nano grating under test in the Fourier plane through multi-frame continuous imaging, and calculate the first-order intensity sequence and its time fluctuation data of the composite diffraction spot distribution data.

[0079] In this step, the micro / nano grating under test is illuminated by a structured composite light field. The generated multi-order diffraction light is used as input, and the composite diffraction spot distribution data is acquired using Fourier space multi-frame continuous imaging technology. Furthermore, image segmentation algorithms and integral calculation methods are used to extract the light intensity response sequences of each illumination channel, and the first-order intensity sequence and time fluctuation data are calculated. This composite diffraction data acquisition method, by integrating multi-channel parallel detection and imaging technologies, significantly improves the ability to acquire information and the detection efficiency of minute structural changes in the grating, providing a richer and more reliable raw data foundation for subsequent accurate analysis.

[0080] S2: Based on the first-order intensity sequence and its time fluctuation data, the actual second-order intensity correlation function characteristic of each lighting channel is calculated through channel separation and registration processing;

[0081] In this step, the first-order intensity sequence and temporal fluctuation data are used as input. Channel separation and sub-pixel-level registration techniques are employed to locate the actual center positions of each diffraction spot. Furthermore, cross-correlation calculations and feature parameter extraction methods are used to calculate the actual second-order intensity correlation function characteristics of each illumination channel. The second-order intensity correlation function analysis method, by integrating spatial positioning correction and coherence characteristic deep mining techniques, significantly improves the sensitivity to optical field coherence information and the accuracy of feature extraction, providing more accurate and stable feature inputs for physical model inversion.

[0082] S3: Input the actual second-order intensity correlation function feature quantity into the preset joint forward physical model, and obtain the reconstructed grating complex amplitude distribution and line edge roughness power spectrum parameters through iterative inversion;

[0083] In this step, the actual second-order intensity correlation function features are used as input. A mapping relationship between diffraction information and grating structure parameters is established through a joint forward physical model. Furthermore, a nonlinear least squares iterative inversion algorithm is used to reconstruct the grating complex amplitude distribution and line edge roughness power spectrum parameters. The joint physical model inversion method, by integrating multi-physics coupled modeling and intelligent optimization algorithms, significantly improves the reconstruction accuracy and computational stability of grating structure parameters, providing a more reliable and complete structural information foundation for accurate deviation calculation.

[0084] S4: Calculate the grating line deviation at the current coordinate position based on the reconstructed grating complex amplitude distribution and the line edge roughness power spectrum parameters;

[0085] In this step, the reconstructed complex amplitude distribution of the grating and the power spectrum parameters of the line edge roughness are used as inputs. The spatial distribution characteristics of the grating lines are quantified using a geometric contour extraction algorithm, and the grating line deviation at the current coordinate position is further obtained using a multi-dimensional deviation calculation method. The grating line deviation calculation method, by integrating complex amplitude information analysis and multi-dimensional geometric feature quantization techniques, significantly improves the detection accuracy and parameter characterization capability of deviations in nanoscale structures, providing more accurate and comprehensive deviation assessment results for quality control in micro-nano manufacturing.

[0086] On one hand, based on the reconstructed complex amplitude distribution of the grating, the amplitude distribution and phase distribution of the micro / nano grating under test are extracted, and the transmittance distribution and phase delay distribution of the micro / nano grating under test are calculated. Then, through threshold segmentation and edge detection algorithms, the geometric contours and boundary position information of the grating lines are extracted. Specifically, amplitude components and phase components are separated and extracted from the reconstructed complex amplitude distribution of the grating. Based on the amplitude components, the amplitude distribution is obtained through modulus calculation. Based on the phase components, a continuous phase distribution is obtained through phase angle calculation combined with a phase unwinding algorithm. Based on the optical transmission characteristics of the grating, the amplitude distribution is converted into a transmittance distribution, and the phase delay distribution is directly obtained. A threshold segmentation algorithm is used to binarize the transmittance distribution to obtain a transmittance image. The transmittance image is segmented into high-transmittance grating line regions and low-transmittance background regions using an adaptive threshold, and segmentation noise is eliminated by morphological filtering. An edge detection algorithm is applied to the segmented transmittance image, and the boundary between the grating lines and the background is detected by a gradient operator. After non-maximum suppression and double thresholding, edge lines with a width of one pixel are obtained, forming the boundary contour of the grating lines. The edge lines with a width of one pixel are connected and organized by a contour tracking algorithm to extract the complete geometric contour of each grating line. At the same time, sub-pixel positioning technology is used to accurately determine the left and right boundary positions of each line on the transmittance profile, and finally, boundary position information including the center contour of the line, boundary coordinates, and geometric parameters is obtained.

[0087] On the other hand, based on the geometric contour and boundary position information of the grating lines, the positional deviation of the micro / nano grating lines at the current coordinate position is calculated, and statistical analysis is used to obtain the statistical results of the grating line deviation. Specifically, based on the geometric contour and boundary position information of the grating lines, the center axis of each grating line is fitted using the least squares fitting method to obtain the center position coordinates of the grating lines; the actual linewidth value of each grating line is calculated based on the boundary position information, and the actual spacing value is calculated by the distance between the center axes of adjacent grating lines; the center position coordinates, actual linewidth values, and actual spacing values ​​are compared with preset design nominal values ​​to calculate the values ​​of linewidth deviation, positional deviation, and spacing deviation. Through statistical analysis, the mean, standard deviation, maximum value, and minimum value of each type of deviation are calculated to obtain the statistical results of the grating line deviation.

[0088] Furthermore, based on the power spectrum parameters of the line edge roughness, the line edge roughness characteristic parameters are calculated through function integration. Specifically, a power spectral density function is constructed based on the reconstructed power spectrum parameters of the line edge roughness; the power spectral density function is integrated over the entire frequency domain using a numerical integration method to calculate the root mean square value of the line edge roughness; the spatial correlation length is extracted from the full width at half maximum (FWHM) feature of the power spectrum by analyzing the frequency characteristics of the power spectral density function; the spectral index is calculated based on the slope of the power spectral density function in logarithmic coordinates; and the root mean square value, spatial correlation length, and spectral index are combined to form the line edge roughness characteristic parameters.

[0089] Finally, based on the statistical results of the grating line deviation and the line edge roughness characteristic parameters, the grating line deviation at the current coordinate position is obtained through a preset quality standard. Specifically, based on the statistical results of the grating line deviation and the line edge roughness characteristic parameters, a preset quality standard is constructed, including line width deviation tolerance, position deviation tolerance, spacing deviation tolerance, and line edge roughness tolerance. The statistical values ​​of each deviation are compared with the corresponding quality tolerances to calculate the quality score of each indicator. A weighted average algorithm is used to integrate the line width deviation score, position deviation score, spacing deviation score, and roughness score to calculate a comprehensive quality score. Based on the comparison between the comprehensive quality score and the preset quality level threshold, the grating line deviation level at the current coordinate position is determined, generating a grating line deviation at the current coordinate position that includes the specific deviation value, quality score, and level determination.

[0090] like Figure 2As shown, in this embodiment, multiple preset scanning positions of the micro / nano grating under test are systematically scanned. At each scanning position, grating line deviation and edge roughness indices are obtained according to steps S1-S4, constructing a two-dimensional deviation distribution map. Specifically, firstly, based on the preset scanning position distribution, a two-dimensional regular grid covering the entire detection area is constructed. The spacing between grid nodes in the two-dimensional regular grid is set according to the grating feature size and detection accuracy requirements to ensure sufficient spatial resolution. Then, based on the grating line deviation and edge roughness feature parameters obtained at each scanning position, a bicubic spline interpolation algorithm is used to numerically interpolate the grid nodes, obtaining an interpolated regular grid. This interpolated regular grid contains continuous distribution information of various deviation parameters within the entire detection area. Next, based on the interpolated regular grid, a linewidth deviation distribution map, a position deviation distribution map, a spacing deviation distribution map, and an edge roughness distribution map are constructed, respectively. Each distribution map displays the deviation magnitude using pseudo-color encoding. Finally, a weighted fusion algorithm is used to synthesize the four distribution maps. The weighting coefficients for synthesis are set according to the degree of influence of different deviation parameters on grating performance, constructing a two-dimensional deviation distribution map.

[0091] Furthermore, based on the two-dimensional deviation distribution map, the grating line deviation detection result is obtained through anomaly region identification, and corresponding adjustment instructions are generated. Specifically, based on the numerical distribution characteristics in the two-dimensional deviation distribution map, anomaly thresholds are set using statistical methods, and grid nodes with deviation values ​​exceeding the anomaly threshold range are marked as anomaly nodes; a density-based spatial clustering algorithm is used to perform cluster analysis on the anomaly nodes, merging adjacent anomaly nodes into connected anomaly regions. The clustering parameters are optimized based on the grid resolution and the minimum detection size of the anomaly region; for the identified connected anomaly regions, their area, aspect ratio, shape complexity, and other geometric feature parameters are calculated, and pattern recognition is performed using a machine learning classifier, based on the geometric shape characteristics, spatial distribution patterns, and deviation intensity of the anomaly regions. Anomalies are classified based on multi-dimensional features such as size and positional relationship within the grating structure. A corresponding re-inspection strategy is developed based on the classification results, including adjusting scanning resolution, optimizing exposure parameters, and employing multi-angle scanning. After re-inspection, the results are compared with the initial results, and confidence is verified using statistical testing methods. The true anomaly region is determined through statistical significance assessment. Finally, a grating line deviation detection result containing information such as the anomaly region's location coordinates, anomaly type, and severity is generated. Based on the grating line deviation detection result, corresponding process adjustment instructions are generated, including at least lithography parameter optimization and process condition correction.

[0092] The specific steps of S1 are as follows:

[0093] S1.1: Spatial structured modulation of incident light is performed using coded illumination technology to generate a structured composite light field including multiple coherent channels. The spatial structured modulation includes phase-coded modulation and amplitude-coded modulation.

[0094] In this embodiment, before acquiring the composite diffraction spot distribution data, the incident light is first spatially structured and modulated using coded illumination technology to generate a structured composite light field including multiple coherent channels. For example... Figure 3 As shown, the spatial structured modulation includes two basic modulation methods: phase-coded modulation and amplitude-coded modulation.

[0095] In one embodiment, the phase-encoded modulation applies differentiated phase delays to different spatial regions of the incident light by loading a preset phase mask pattern, generating a first illumination channel. The phase mask pattern is designed based on the structural characteristics and detection requirements of the micro / nano grating to be detected and includes spatially distributed phase delay information. The phase mask pattern is implemented by loading a liquid crystal spatial light modulator, where each pixel unit applies a corresponding phase delay to the incident light according to the requirements of the phase mask pattern. The liquid crystal material changes its molecular orientation under electric field drive, thereby altering the effective refractive index of the region and generating phase modulation of the passing light wave. The phase distribution of the first illumination channel is adjusted according to the phase delay distribution of the phase mask pattern, forming a modulated beam with a specific wavefront distribution. In another embodiment, the phase mask pattern is implemented using diffractive optical elements, employing a surface relief structure to create a preset depth distribution on a transparent substrate. Different depth regions generate different optical path differences for the incident light, achieving phase modulation.

[0096] On the other hand, the amplitude-encoded modulation selectively transmits the incident light to different spatial regions by loading a preset amplitude mask pattern, generating a second illumination channel. The amplitude mask pattern design includes spatially distributed transmittance information to control the transmission intensity of the incident light in different regions. The amplitude mask pattern is implemented using a digital micromirror device (DMM), which includes a large number of independently controllable micromirror units. Each micromirror can switch between two angular positions, achieving spatially selective reflection by controlling the state of each micromirror. When the micromirror is at its working angle, the incident light is reflected into the effective optical path; when it is at its non-working angle, the incident light is guided to the optical trap. The amplitude distribution of the second illumination channel is adjusted by the transmittance distribution of the amplitude mask pattern, forming a modulated beam with a specific intensity distribution. In other embodiments, the amplitude-encoded modulation is achieved using a liquid crystal light valve combined with a polarization device. The electrically controlled birefringence of the liquid crystal is used to adjust the polarization state of the transmitted light, and then an analyzer is used to achieve amplitude modulation.

[0097] Furthermore, based on the first and second illumination channels, multiple parallel illumination channels with preset spatial spectral distributions are generated. These parallel illumination channels are implemented using an optical beam splitting system, decomposing the input beam into multiple independent sub-beams, each of which undergoes independent modulation processing. Each parallel illumination channel employs different combinations of phase and amplitude mask patterns to form illumination beams with differentiated spatial spectral characteristics. The coherence characteristics of the parallel illumination channels are distinct, achieved through optical path control and coherence adjustment. The optical path control uses an adjustable delay line to precisely adjust the optical path length of each illumination channel, controlling the phase relationship between the channels. The coherence characteristic adjustment is achieved through partially coherent light sources or coherence modulation devices, giving each parallel illumination channel different coherence characteristics.

[0098] Based on the parallel illumination channels and their coherence characteristics, a structured composite light field is generated through spatial superposition. This spatial superposition is achieved using a beam combiner or interferometer configuration, spatially overlapping the parallel illumination channels within the target illumination area. The light fields of each parallel illumination channel undergo coherent or incoherent superposition in the superposition region according to their coherence characteristics, forming a structured composite light field with a complex spatial structure. This structured composite light field contains information from multiple coherent channels, possesses rich spatial spectral content and tunable illumination characteristics, providing optimized illumination conditions for multi-dimensional illumination of micro / nano gratings.

[0099] S1.2: Based on the composite diffraction spot distribution data under the structured composite light field illumination, perform continuous multi-frame real-time acquisition through preset sampling frequency and total frame count requirements to obtain continuous multi-frame Fourier plane images of the composite diffraction spot distribution data.

[0100] In this embodiment, the continuous multi-frame real-time acquisition is a process of obtaining temporal diffraction information through imaging technology. Specifically, the sampling frequency is determined by analyzing the temporal spectral characteristics of the composite diffraction spot distribution data. The total number of frames is determined based on the sample size requirements and measurement accuracy indicators of statistical analysis. The sample size requirements are obtained through theoretical analysis and numerical simulation calculations. In some optional implementations, the determination of the total number of frames also needs to consider practical constraints such as system stability, storage capacity, and processing efficiency.

[0101] Furthermore, based on the sampling frequency and total number of frames, Fourier plane images are continuously acquired at preset time intervals using imaging devices such as high-speed CCD cameras, CMOS image sensors, or photodiode arrays. The imaging device receives the diffracted light generated after the structured composite light field illuminates the micro / nano grating under test and converts the diffracted light into digital image data. During the acquisition process, precise timing control ensures time synchronization of each frame's acquisition. A master clock generator provides a stable reference frequency as the time base, a trigger signal distributor generates acquisition trigger signals for each frame based on the sampling frequency, and a frame synchronization monitor records the timestamps of each frame's image in real time.

[0102] Finally, the digital image data is transmitted in real time through a data transmission interface, and continuous acquisition and storage are achieved by using a circular buffer and pipeline processing technology to obtain the continuous multi-frame Fourier plane images containing time series information.

[0103] S1.3: Based on the spatial distribution functions of the first and second illumination channels, the diffraction spot regions corresponding to each illumination channel are extracted from the multi-frame Fourier plane images using an image segmentation algorithm. Based on the diffraction spot regions, the integrated light intensity value of each diffraction spot region at each time is calculated using an integral operation method.

[0104] In this embodiment, spatial distribution functions for each illumination channel are established based on the coding and modulation process. For the first illumination channel, a spatial distribution function is constructed based on the phase delay distribution of the phase mask pattern. This function includes the spatial intensity distribution and phase distribution characteristics of the light field after phase coding and modulation. For the second illumination channel, a spatial distribution function is constructed based on the transmittance distribution of the amplitude mask pattern. This function also includes the spatial intensity distribution and phase distribution characteristics of the light field after amplitude coding and modulation. Based on the spatial distribution functions, the diffraction transformation results of the spatial distribution functions of each illumination channel in the Fourier plane are calculated using Fraunhofer diffraction theory. This yields the spatial spectral distribution of each illumination channel in the Fourier plane, determining the theoretical center coordinates, shape characteristics, size range, and intensity distribution patterns of the diffraction spots generated by each illumination channel in the Fourier plane. This provides a theoretical prediction benchmark and a basis for locating segmented regions for subsequent image segmentation algorithms.

[0105] Furthermore, based on the spatial distribution function, a theoretical diffraction model for each illumination channel is constructed using diffraction optics theory, establishing a mapping relationship from illumination channel parameters to the characteristics of the diffraction spots in the Fourier plane. The theoretical diffraction model includes prediction of the position distribution of the diffraction spots in the Fourier plane for each illumination channel, calculation of the spot boundary contour, simulation of light intensity distribution, and analysis of interference effects between adjacent spots. For each frame of the multi-frame Fourier plane images, a region-based image segmentation algorithm is used to extract the diffraction spot region corresponding to each illumination channel. Specifically, based on the theoretical position coordinates determined by the spatial distribution function, a search window for each diffraction spot is established in the Fourier plane image to determine the approximate location area of ​​each diffraction spot. Within each search window, an adaptive threshold segmentation method is used to automatically determine the segmentation threshold by analyzing local gray-level distribution characteristics, effectively separating the diffraction spot area from the background noise. Simultaneously, a seed point-based region growing algorithm is used, with the theoretical center position of each diffraction spot as the seed point. Based on pixel gray-level similarity criteria and spatial connectivity constraints, the growing area is gradually expanded to accurately determine the actual boundary contour of each diffraction spot area.

[0106] Furthermore, based on the diffraction spot regions, the integrated light intensity value of each diffraction spot region at each time step is calculated using an integration method. Specifically, for each identified diffraction spot region, spatial integration is performed on the grayscale values ​​of all pixels within that region to obtain the total light intensity value of the diffraction spot region. Based on the pixel response characteristics and quantum efficiency of the imaging device, the total light intensity value of the diffraction spot region is calibrated and corrected to obtain the corrected integrated light intensity value. The above integration operation is repeated for each diffraction spot region, and the results are recorded in chronological order to obtain the integrated light intensity value of each diffraction spot region at each time step.

[0107] S1.4: Based on the integrated light intensity value, construct the light intensity response sequence of each illumination channel in chronological order to form a first-order intensity sequence including at least two illumination channels;

[0108] In this embodiment, based on the sampling timestamp, the integrated light intensity values ​​of each lighting channel at different times are arranged in chronological order, establishing a correspondence between time indexes and light intensity data. Each lighting channel corresponds to the light intensity data at the same physical time point under the same time index, ensuring time synchronization between multi-channel data.

[0109] For each illumination channel, a complete light intensity response sequence is constructed according to the chronological order of sampling times. The light intensity response sequence is a one-dimensional array comprising all integrated light intensity values ​​of that illumination channel over the entire measurement time period, with the elements of the array arranged chronologically. For systems with multiple illumination channels, a light intensity response sequence is constructed for each channel separately, with each sequence maintaining the same time length and sampling interval.

[0110] Furthermore, the light intensity response sequences of each lighting channel are organized into a matrix-like multi-channel time series data structure, where each row corresponds to the light intensity response sequence of one lighting channel, and each column corresponds to the integrated light intensity value of all lighting channels at a specific time. The number of rows in the multi-channel time series data structure is equal to the total number of lighting channels, and the number of columns is equal to the total number of sampling times.

[0111] Finally, the light intensity response sequence of each illumination channel is defined as a first-order intensity sequence, which directly reflects the change of the integrated light intensity value over time. In this way, a first-order intensity sequence including at least two illumination channels is constructed, providing a time-series data foundation for subsequent light intensity deviation analysis.

[0112] S1.5: Based on the first-order intensity sequence, the intensity variance and time autocorrelation function of each lighting channel are calculated using statistical analysis methods. Based on the intensity variance and time autocorrelation function, the time fluctuation data of each lighting channel are obtained.

[0113] In this embodiment, firstly, based on the first-order intensity sequence, the intensity variance is calculated through variance statistics to obtain the intensity characteristic information of time fluctuations. Specifically, based on the first-order intensity sequence of each lighting channel, its average value in the time dimension is calculated as the light intensity reference level for each lighting channel; the deviation of each light intensity sample value in the first-order intensity sequence from the light intensity reference level is compared to obtain the deviation value, and the squared deviation value is obtained by squaring; the intensity variance of each lighting channel is calculated by averaging the squared deviation values.

[0114] Subsequently, a time autocorrelation function is calculated based on the first-order intensity sequence to reveal the temporal characteristics of the light intensity signal. Specifically, the first-order intensity sequence is first subjected to mean-reduction processing. Then, a series of time interval parameters are set, starting from zero time interval, with the sampling time interval of the first-order intensity sequence as the step size, gradually increasing to a certain proportion of the total sequence duration as the maximum analysis time interval. For each set time interval, the correlation coefficient between the mean-reduced first-order intensity sequence and its sequence shifted by that time interval is calculated. By systematically traversing all set time intervals, a complete time autocorrelation function curve is constructed. The calculation result of the time autocorrelation function reflects the correlation degree of the first-order intensity sequence of the same illumination channel at different time intervals, and its decay characteristics reveal the temporal correlation scale of light intensity fluctuations.

[0115] Furthermore, by combining the intensity variance and the time autocorrelation function, temporal fluctuation data for each illumination channel is extracted. Specifically, intensity characteristic information of the time fluctuations is obtained from the intensity variance, and temporal characteristic information of the time fluctuations is extracted from the time autocorrelation function. The consistency of the calculation is verified by analyzing the relationship between the value of the time autocorrelation function at zero time interval and the intensity variance. Combining the intensity characteristic information and the temporal characteristic information forms complete temporal fluctuation data for each illumination channel, providing the necessary fluctuation characteristic information basis for subsequent light intensity deviation analysis and micro / nano grating parameter measurement.

[0116] The specific steps of S2 are as follows:

[0117] S2.1: Based on the center position coordinates and spatial distribution function of the first and second lighting channels, the theoretical diffraction position of each lighting channel in the Fourier plane is determined by a spatial positioning algorithm;

[0118] In this embodiment, the spatial geometric characteristics of the lighting channels are extracted based on the center position coordinates and spatial distribution function of the first and second lighting channels. The theoretical diffraction positions of each lighting channel in the Fourier plane are determined by theoretical calculation, providing a theoretical reference for subsequent diffraction spot positioning and displacement correction.

[0119] Specifically, firstly, the center position coordinates of each illumination channel are extracted based on the preset phase mask pattern and amplitude mask pattern. For the first illumination channel, phase delay distribution data is extracted from the preset phase mask pattern. By identifying the effective region with phase modulation effect, the geometric center of the effective region is determined as the center position coordinates of the first illumination channel using a geometric centroid calculation method. For the second illumination channel, transmittance distribution data is extracted from the preset amplitude mask pattern. By setting a transmittance discrimination threshold, the effective transmission region is identified. The centroid position of the transmission region is determined as the center position coordinates of the second illumination channel using a centroid calculation method.

[0120] Furthermore, based on the equipment manual and technical specifications of the optical inspection system, a geometric parameter model of the optical system is constructed. Key system parameters, including the nominal wavelength of the incident light source, objective lens focal length, numerical aperture specification, and working distance range, are obtained from the technical specifications provided by the optical system manufacturer. By consulting the equipment manual, the design distance from the objective lens to the Fourier plane, the relative positions of optical elements, and the geometric configuration of the optical path are obtained. Based on these system parameters and geometric configuration, a geometric parameter model of the optical system is constructed, defining the coordinate transformation relationship from the object plane to the Fourier plane and the optical propagation characteristics.

[0121] Finally, based on the center position coordinates and spatial distribution function of each illumination channel, and combined with the optical system geometric parameter model, the theoretical diffraction position of each illumination channel in the Fourier plane is calculated. On one hand, for the first illumination channel, the center position coordinates and corresponding spatial distribution function of this channel are used as input. Utilizing the objective lens focal length parameter and numerical aperture specification in the optical system geometric parameter model, the correspondence between the object plane coordinates and the Fourier plane coordinates is established through Fourier optics theory. Using the diffraction integral calculation method, combined with the incident light source wavelength parameter, the light field distribution of this illumination channel after diffraction by the micro / nano grating under test is calculated in the Fourier plane. Based on the design distance parameter from the objective lens to the Fourier plane, the calculated spatial frequency distribution is converted into the physical coordinate position of the Fourier plane. By analyzing the complex amplitude distribution characteristics of the light field distribution, the intensity centroid positioning algorithm is used to extract the intensity centroid position of the diffracted light field as the theoretical diffraction position of the first illumination channel.

[0122] On the other hand, for the second illumination channel, the center coordinates of the channel and the corresponding spatial distribution function are used as input. Similarly, the objective lens focal length, numerical aperture specifications, and the design distance from the objective lens to the Fourier plane in the optical system's geometric parameter model are used for coordinate transformation and physical position mapping. Using the same diffraction integral calculation method, combined with the nominal wavelength of the incident light source and the relative positional parameters of the optical elements, the light path propagation path is corrected to determine the theoretical diffraction position of the second illumination channel in the Fourier plane.

[0123] The above calculation process uses the optical system geometric parameter model to achieve an accurate conversion from theoretical calculation results to actual physical coordinates, obtains the theoretical diffraction position coordinates of each illumination channel in the Fourier plane, and forms a theoretical diffraction position dataset, which provides an accurate theoretical reference benchmark for the actual diffraction spot positioning and displacement correction algorithm in subsequent steps.

[0124] S2.2: Perform channel separation processing on the multi-frame Fourier plane image, identify the diffraction spot region corresponding to each illumination channel through image segmentation algorithm, and use sub-pixel level registration algorithm to locate the actual center position of each diffraction spot region.

[0125] In this embodiment, a sub-pixel-level registration algorithm fused with multiple algorithms is used to achieve high-precision center position identification of complex diffraction spots, breaking through the accuracy limitations of traditional pixel-level positioning. The multi-algorithm fusion includes the synergistic application of centroid calculation, Gaussian fitting, and ellipse fitting methods.

[0126] First, the multi-frame Fourier plane images undergo preprocessing operations, including adaptive noise filtering, dynamic background correction, and grayscale normalization. The adaptive noise filtering dynamically selects the filter type and parameters based on the local signal-to-noise ratio characteristics of the image, removing noise while maintaining the sharpness of the diffraction spot edges. The dynamic background correction employs a combination of temporal filtering and spatial polynomial fitting to eliminate background inhomogeneities caused by environmental interference and device drift.

[0127] Then, intelligent channel separation processing is performed on the multi-frame Fourier plane images. An adaptive region of interest is defined in each frame based on the theoretical diffraction positions of the first and second illumination channels. A hierarchical threshold segmentation method is employed, using global threshold coarse segmentation and local adaptive threshold fine segmentation to achieve accurate separation of diffraction spots of different intensities.

[0128] Furthermore, a multi-feature fusion image segmentation algorithm is used to identify the diffraction spot regions corresponding to each illumination channel. An improved connected component analysis algorithm is employed to identify continuous high-brightness pixel regions as candidate regions. By calculating the multi-dimensional geometric feature parameters of each connected component, including area, aspect ratio, orientation angle, and compactness, a multi-feature weighted matching criterion is established to screen the true diffraction spot regions. For multiple illumination channels, a parallel recognition process is executed separately to achieve efficient and independent separation of the diffraction spots in each illumination channel.

[0129] Finally, a sub-pixel-level registration algorithm fused with multiple algorithms is used to locate the actual center position of each diffraction spot region. First, a weighted centroid calculation method is used to determine a coarse center position. Within the adaptive neighborhood, multiple mathematical fitting models are constructed in parallel, including a two-dimensional Gaussian function model and an elliptic Gaussian model. The parameters of each model are adjusted using a multi-objective optimization algorithm, and a hybrid optimization strategy using least squares and gradient descent is employed to minimize the fitting error. Finally, a weighted fusion algorithm is used to calculate the weighted average center position coordinates based on the accuracy and reliability of each fitting model, obtaining the sub-pixel-precision actual center position of each diffraction spot region.

[0130] S2.3: Based on the actual center position and the theoretical diffraction position, the first-order intensity sequence is corrected using a displacement correction algorithm to obtain the corrected first-order intensity sequence;

[0131] In this embodiment, based on the actual center position and the theoretical diffraction position, a multi-level displacement correction algorithm is used to systematically correct the first-order intensity sequence, eliminating measurement errors caused by spot position offset and achieving sub-pixel level high-precision correction.

[0132] First, the two-dimensional spatial displacement vector of the diffraction spot for each illumination channel is calculated. Specifically, for each illumination channel, the actual center position is compared with the corresponding theoretical diffraction position, and the horizontal and vertical displacement components are calculated to form a two-dimensional spatial displacement vector. Then, the light intensity integration region is dynamically corrected based on the two-dimensional spatial displacement vector. Based on the preset light intensity integration region, for each frame of the Fourier plane image in the time series, the center coordinates of the original light intensity integration region are translated according to the corresponding two-dimensional spatial displacement vector to obtain the corrected light intensity integration region for each illumination channel at each time point.

[0133] Furthermore, the integral light intensity value at each moment is recalculated based on the corrected integral light intensity region. For each moment of each illumination channel, the pixel grayscale value is accumulated within the corresponding corrected integral light intensity region to obtain the corrected integral light intensity value sequence for each illumination channel. This correction process eliminates the systematic error in light intensity measurement caused by the offset of the light spot position, ensuring that the integral light intensity value accurately reflects the actual intensity change of the diffracted light spot.

[0134] Finally, the corrected first-order intensity sequence for each lighting channel is reconstructed in chronological order. The corrected integral light intensity values ​​for each lighting channel are then arranged chronologically to obtain the final corrected first-order intensity sequence. This time sequence eliminates the influence of spot displacement errors, providing accurate and reliable input data for subsequent calculations of the second-order intensity correlation function.

[0135] S2.4: Based on the corrected first-order intensity sequence, the second-order intensity correlation function between any two illumination channels is calculated through cross-correlation operation, and the peak value, full width at half maximum (FWHM), decay time constant, and coherence of the second-order intensity correlation function are extracted as feature parameters.

[0136] In this embodiment, based on the corrected first-order intensity sequence, the second-order intensity correlation characteristics between multiple channels are calculated using a systematic cross-correlation analysis method, and key feature parameters are extracted for subsequent physical parameter inversion and system performance evaluation.

[0137] First, the corrected first-order intensity sequence is preprocessed using standardization. Specifically, for each lighting channel, the average value over time is calculated, and this average value is subtracted to obtain the intensity fluctuation sequence, eliminating common-mode noise caused by factors such as light source power fluctuations, system gain changes, and ambient light interference. Further, the intensity fluctuation sequence is normalized using the standard deviation of the intensity fluctuations for each lighting channel as the normalization factor to obtain a standardized intensity fluctuation sequence. This normalization process eliminates amplitude differences and gain inconsistencies between different channels.

[0138] Then, a cross-correlation operation is used to calculate the second-order intensity correlation function between any two illumination channels based on the normalized intensity fluctuation sequence. According to quantum optics theory, a normalized second-order intensity correlation function between any two illumination channels is defined, which includes a time delay parameter to characterize the correlation properties at different times. A sliding time window method is used to progressively calculate the cross-correlation coefficient within a preset time delay range, and a fast Fourier transform algorithm is used to accelerate the cross-correlation operation, resulting in the second-order intensity correlation function containing time delay information.

[0139] On the other hand, the decay time constant is extracted based on the decay characteristics of the second-order intensity correlation function using an exponential fitting algorithm. A suitable fitting window is selected near the peak of the correlation function, and an exponential function is fitted using a nonlinear least squares method to obtain the decay time constant, which reflects the time decay rate of the inter-channel correlation.

[0140] On the other hand, the normalized correlation value under zero time delay is calculated as the coherence. The coherence is calculated by the correlation of the intensities of different illumination channels at the same time, reflecting the instantaneous coherence between channels and the synchronicity of the light field distribution.

[0141] Finally, the above calculation process is systematically performed on all possible lighting channel pair combinations, and the peak value, full width at half maximum (FWHM), decay time constant, and coherence are combined to construct a complete feature matrix containing four types of feature parameters, which are used as feature parameters.

[0142] S2.5: Based on the aforementioned characteristic parameters, and combined with the coherence characteristics and relative light intensity weights of the illumination channel, the actual second-order intensity correlation function characteristic quantities are obtained through normalization processing.

[0143] In this embodiment, a quantitative evaluation system for the coherence characteristics of the lighting channels is first established. Specifically, for each lighting channel, the coherence length, coherence time, and spatial coherence parameters are determined based on its light source characteristics, transmission path, and modulation method. By analyzing the spectral characteristics and temporal stability of each lighting channel, the temporal coherence function and spatial coherence function are calculated. Combining the coherence properties of the light source and the geometric configuration of the system, the coherence characteristic matrix between each lighting channel is obtained, describing the strength of coherent coupling between different channel pairs.

[0144] Furthermore, the relative luminous intensity weighting coefficients for each lighting channel are determined. Specifically, a feature parameter correction model is constructed based on the coherence characteristic matrix and the relative luminous intensity weighting coefficients. Using this model, the peak feature parameters are corrected by weighting the relative coherence intensities between lighting channels to eliminate systematic biases caused by differences in coherence; the full width at half maximum (FWHM) is normalized by combining the coherence time characteristics of each lighting channel to reflect the actual coherence time scale of the system; the decay time constant is weighted and averaged based on the relative luminous intensity weighting coefficients; and the coherence degree is transformed using the coherence characteristic matrix to obtain the corrected feature parameter set.

[0145] Furthermore, a multi-level normalization strategy is employed to standardize the modified feature parameter set. In the first level of normalization, each type of modified feature parameter is standardized separately to eliminate the influence of dimensions and differences in numerical range. In the second level of normalization, principal component analysis is used to extract the main feature components, eliminating the correlation effects between feature parameters. In the third level of normalization, environmental adaptability corrections are performed based on the actual operating conditions of the system, resulting in a standardized feature parameter set.

[0146] Finally, based on the standardized feature parameter set, the actual second-order intensity correlation function feature quantities are constructed. Specifically, using a weight allocation method driven by a physical model, weight coefficients are determined according to the contribution of each feature parameter to system performance. The standardized feature parameter set is then weighted and combined to obtain the actual second-order intensity correlation function feature quantities.

[0147] The specific steps for S3 are as follows:

[0148] S3.1: Based on the center position coordinates, spatial distribution function, coherence characteristics and relative light intensity weights of the first and second illumination channels, a joint forward physical model of micro-nano grating diffraction under coded illumination is constructed based on scalar diffraction theory.

[0149] In this embodiment, firstly, a spatial field distribution model for multi-channel coded illumination is established. Specifically, based on the center position coordinates of each illumination channel, a spatial coordinate system is established on the grating plane to determine the illumination area of ​​each illumination channel. For each illumination channel, based on the spatial intensity distribution and phase distribution characteristics in its spatial distribution function, a Gaussian function is used to construct the overall illumination field distribution function for the multi-channel superposition.

[0150] Then, a coherence matrix of the illumination field based on coherence characteristics is established. Specifically, according to the coherence characteristics of each illumination channel, a mutual coherence function between channels is constructed using partial coherence light theory. Based on the mutual coherence function, a complex coherence function between each illumination channel is obtained through normalization; by aggregating the complex coherence functions between each pair of illumination channels, a complex coherence function matrix is ​​formed.

[0151] Next, based on scalar diffraction theory, a diffraction calculation model for micro / nano gratings is constructed. Specifically, based on the grating's geometric morphology information and material optical property parameters, a complex amplitude transmission function is constructed using optical propagation theory. This function comprehensively reflects the geometric structural characteristics of the grating and the material's optical response characteristics. For periodic grating structures, the transmission function is expressed as a series using Fourier series expansion based on its periodicity, obtaining the Fourier coefficients corresponding to each diffraction order to describe the grating's transmission characteristics. For non-periodic gratings or those with defects, the transmission function is directly solved using numerical methods such as finite difference or finite element analysis based on the irregularity of their structure, resulting in a spatially discretized transmission function distribution. Combining the line-edge roughness characteristics during grating manufacturing, a random phase modulation term is introduced into the transmission function based on the statistical distribution law of roughness. A phase perturbation that conforms to the actual roughness characteristics is generated through statistical modeling methods, ultimately obtaining a complete transmission function model that considers the influence of manufacturing defects.

[0152] Finally, a joint forward physical model under coded illumination is constructed. Specifically, based on the relative intensity weights of each illumination channel, the multi-channel illumination field and the complex amplitude transmission function are coupled and calculated using the principle of linear superposition. Based on scalar diffraction theory and combined with the aforementioned complex coherence function matrix, the diffraction field distribution generated by each illumination channel under the action of the micro / nano grating is calculated. Through a hybrid processing method of coherent superposition and incoherent superposition, the diffraction contribution of each illumination channel is weighted and synthesized according to the coherence characteristics between channels to form the total diffraction field distribution under coded illumination conditions. Based on the total diffraction field distribution, a complete forward mapping relationship from illumination parameters and grating geometry information to the diffraction field distribution is established, constituting a joint forward physical model of micro / nano grating diffraction under coded illumination.

[0153] S3.2: Based on the joint forward physical model, construct a mathematical mapping relationship from the grating complex amplitude distribution, the power spectrum parameters of the line edge roughness to the characteristic quantities of the theoretical second-order intensity correlation function;

[0154] In this embodiment, firstly, a statistical characteristic description of the diffracted light field is established based on the joint forward physical model. Specifically, the total diffracted light field distribution under coded illumination conditions is represented as a superposition of deterministic and random components, where the deterministic component is determined by the complex amplitude distribution of the grating, and the random component is controlled by the power spectrum parameter of the line edge roughness. Through statistical optics theory, a statistical distribution model of the complex amplitude of the diffracted light field at different spatial locations is established to obtain the first and second order statistical moment characteristics of the light field.

[0155] Then, a theoretical calculation model for the second-order intensity correlation function is constructed. Specifically, based on the definition of the second-order intensity correlation function, a statistical average expression for the product of the intensities of the diffracted light field at two spatial locations is established. Based on the statistical distribution characteristics of the complex amplitude of the light field, the second-order intensity correlation function is expanded into a fourth-order moment function of the complex amplitude of the light field using probability theory and stochastic process theory. Combining the Gaussian stochastic process characteristics of line edge roughness, the fourth-order moment is decomposed into a product combination of second-order moments using Isserlis' theorem, thus establishing an analytical relationship between the second-order intensity correlation function and the second-order statistical moments of the light field.

[0156] Next, the characteristic parameters of the second-order intensity correlation function are extracted. Specifically, spatial frequency domain analysis is performed on the theoretical second-order intensity correlation function, and its power spectrum distribution characteristics are obtained through Fourier transform. Key characteristic parameters such as peak position, full width at half maximum (FWHM), and attenuation constant of the correlation function are extracted as characteristic quantities describing the grating structure and roughness characteristics. Functional relationships are established between the characteristic parameters and structural parameters such as periodicity and modulation depth in the complex amplitude distribution of the grating, as well as statistical parameters such as correlation length and root mean square roughness in the power spectrum parameters of the line edge roughness.

[0157] Finally, a complete mathematical mapping model is constructed. Specifically, based on the aforementioned theoretical analysis, an explicit mathematical expression is established, taking the grating complex amplitude distribution and line edge roughness power spectrum parameters as input variables and the characteristic quantities of the theoretical second-order intensity correlation function as output variables. Through parameter sensitivity analysis, the influence weights and coupling relationships of each input parameter on the characteristic quantities are determined. A parameterized mapping function model is established to achieve quantitative prediction from grating physical structure parameters to observable optical characteristic quantities, forming a complete mathematical mapping relationship from the grating complex amplitude distribution, line edge roughness power spectrum parameters to the characteristic quantities of the theoretical second-order intensity correlation function.

[0158] S3.3: Based on the mathematical mapping relationship and the actual second-order intensity correlation function characteristic quantity, the residual between the theoretical second-order intensity correlation function characteristic quantity and the actual second-order intensity correlation function characteristic quantity is minimized through a nonlinear least squares iterative inversion algorithm to obtain the reconstructed grating complex amplitude distribution and line edge roughness power spectrum parameters.

[0159] In this embodiment, a complete nonlinear least squares iterative inversion algorithm framework is established. Based on preset initial parameters, the theoretical second-order intensity correlation function characteristic quantities are calculated using a joint forward physical model. These are then compared with the actual measured characteristic quantities to construct a standardized residual vector. The Jacobian matrix is ​​calculated using numerical differentiation, and the parameter update quantity is solved using the least squares method. Under physical constraints, the grating complex amplitude distribution and line edge roughness power spectrum parameters are iteratively updated. When the relative change of the L2 norm of the residual vector reaches the convergence threshold, the final reconstructed parameters are output. This technical solution accurately describes the combined physical effects of grating diffraction and line edge roughness through precise modeling of the joint forward physical model. It improves the robustness and accuracy of parameter estimation by using characteristic quantity matching and weighted residual analysis. The numerical stability and convergence of the iterative process are ensured by combining numerical differentiation Jacobian matrix calculation and adaptive parameter update strategy. The physical rationality and measurement consistency of the reconstruction results are guaranteed through physical constraint verification and quality assessment mechanisms. This achieves high-precision inversion reconstruction from second-order intensity correlation function measurement data to grating microstructure parameters, providing a reliable technical means for grating manufacturing process optimization and performance evaluation.

[0160] The specific steps of S3.3 are as follows:

[0161] S3.3.1: Based on the preset initial grating complex amplitude distribution and initial line edge roughness power spectrum parameters, the characteristic quantity of the theoretical second-order intensity correlation function is calculated through the joint forward physical model;

[0162] In this embodiment, firstly, based on the grating design parameters and manufacturing process information, the initial grating complex amplitude distribution is set, wherein the amplitude distribution uses an ideal rectangular wave function as an initial approximation, the duty cycle is determined according to the design value, and the phase distribution calculates the initial estimate of the phase delay based on the refractive index and thickness information of the grating material, thus obtaining the initial grating complex amplitude distribution containing amplitude and phase information; based on the typical characteristics of the manufacturing process, the initial values ​​of the relevant length, roughness amplitude, and spectral index are set, thus obtaining the initial line edge roughness power spectrum parameters.

[0163] Then, based on the minimum feature size of the grating and the required accuracy, the initial complex amplitude distribution of the grating is discretized into a numerical array on a spatial grid. The amplitude and phase information of each grid point are stored in complex form, resulting in a numericalized grating complex amplitude distribution matrix. Based on the physical meaning of the line-edge roughness power spectrum parameters, a vector representation including key parameters such as correlation length, roughness amplitude, and spectral index is established, yielding normalized line-edge roughness power spectrum parameters. Based on the numericalized grating complex amplitude distribution matrix and the normalized line-edge roughness power spectrum parameters, a grating transmission function matrix is ​​constructed. Based on the grating transmission function matrix, a random phase distribution conforming to the power spectrum characteristics is generated using the Monte Carlo method. A random phase modulation term is introduced into the transmission function to obtain a grating transmission function incorporating the line-edge roughness effect.

[0164] Furthermore, based on the spatial distribution and coherence characteristic parameters of the multi-channel coded illumination, the incident field distribution of each illumination channel in the grating plane is calculated, and the multi-channel incident field distribution matrix is ​​obtained based on the amplitude, phase, and polarization characteristics of each illumination channel in the incident field distribution.

[0165] Furthermore, based on the incident field distribution and the grating transmission function, the transmitted field distribution is obtained. Based on the transmitted field distribution, the distribution of the diffraction field on the observation plane is calculated using the angular spectrum propagation method or Fresnel diffraction integral. Combined with the coherence matrix, the total diffraction field distribution is obtained through the coherent superposition effect between multiple channels. Based on the spatial sampling distribution of the Fourier plane, the complex amplitude signal at each spatial location is extracted from the total diffraction field distribution. According to the response characteristics and noise features of the imaging process, the complex amplitude is converted into an intensity signal. Combined with the time modulation sequence of the coded illumination, the time variation process is simulated to obtain the intensity time series data for each spatial location.

[0166] Finally, based on the intensity time series data, the second-order intensity correlation function between each detection channel is calculated using a cross-correlation algorithm. By utilizing the finite sampling effect and statistical fluctuation characteristics, a correction model is constructed to correct the second-order intensity correlation function. Furthermore, by combining the multi-realization averaging method and performing multiple independent calculations, a theoretical second-order intensity correlation function with reduced statistical error is obtained.

[0167] Based on the theoretical second-order intensity correlation function data, peak value, full width at half maximum (FWHM), decay time constant, and coherence are extracted using a feature extraction algorithm. The reliability of these parameters is monitored and calculated according to quality assessment metrics. The extracted parameters are then organized into feature vectors to obtain the theoretical second-order intensity correlation function features, which are used for subsequent comparison and optimization calculations.

[0168] S3.3.2: Compare the theoretical second-order intensity correlation function eigenvalues ​​and the actual second-order intensity correlation function eigenvalues ​​to obtain the residual vector, and use the numerical differentiation method to calculate the corresponding Jacobian matrix;

[0169] In this embodiment, firstly, a correspondence between theoretical and actual characteristic quantities is established. The theoretical second-order intensity correlation function characteristic quantities are matched one by one with the actual second-order intensity correlation function characteristic quantities to obtain a characteristic quantity correspondence table. Based on the correspondence between the theoretical and actual second-order intensity correlation function characteristic quantities, the difference between the theoretical and actual characteristic quantities is calculated one by one to obtain the original residual vector. Based on the measurement uncertainty and physical importance information of each characteristic quantity, the weight coefficients corresponding to the original residual vector are calculated, and the original residual vector is weighted and normalized to obtain a standardized residual vector.

[0170] Then, based on the standardized residual vector, the statistical characteristic parameters of the residuals are calculated, including the L2 norm, mean, and variance. Based on the statistical characteristic parameters, a quantitative index of the fitting error is obtained. Based on the distribution characteristics of the residual vector and combined with the quantitative index of the fitting error, an anomaly detection algorithm is used to identify significantly deviating data points, and a data quality assessment result is obtained. Based on the historical residual data during the iteration process and the data quality assessment result, a time series analysis model is established to obtain the residual change trend and convergence assessment result.

[0171] Furthermore, based on the physical characteristics and numerical range of the parameters to be optimized, a numerical differential calculation scheme for the Jacobian matrix is ​​established, and the perturbation strategy corresponding to each parameter is determined. According to the numerical magnitude and physical dimensions of the parameters, an adaptive method is used to determine the perturbation step size, and a small perturbation is applied to each parameter. The characteristic quantities of the theoretical second-order intensity correlation function are recalculated using a joint forward physical model to obtain the characteristic quantity change data after parameter perturbation. Based on the characteristic quantity change data and the perturbation step size information, the partial derivatives are calculated using the forward difference formula to obtain the elements of the Jacobian matrix. Based on the elements of the Jacobian matrix, the corresponding Jacobian matrix is ​​calculated.

[0172] S3.3.3: Based on the residual vector and Jacobian matrix, the parameter update amount is calculated by the least squares method. Based on the parameter update amount, the updated grating complex amplitude distribution and line edge roughness power spectrum parameters are obtained.

[0173] In this embodiment, firstly, a linearized least squares problem is established based on the standardized residual vector and the Jacobian matrix, and a normal equation system is constructed. Based on the condition number and numerical stability characteristics of the Jacobian matrix, a suitable solution algorithm is selected, and the normal equation system is solved using QR decomposition or singular value decomposition to obtain the parameter update vector.

[0174] Then, based on the numerical value and physical rationality of the parameter update vector, constraints on the update amount are set to prevent excessive parameter updates from causing the physical model to fail. The parameter update amount is adjusted using the trust region method or line search strategy to ensure that the updated parameters are still within the physically permissible range. Based on the adjusted parameter update amount, the updated grating complex amplitude distribution and line edge roughness power spectrum parameters are calculated.

[0175] Furthermore, based on the updated grating complex amplitude distribution and line edge roughness power spectrum parameters, their physical consistency and numerical rationality are verified; for the grating complex amplitude distribution parameters, the range of amplitude and phase values ​​is checked to ensure compliance with optical physics constraints; for the line edge roughness power spectrum parameters, the positive values ​​and dimensional consistency of relevant parameters such as length and roughness amplitude are verified; based on the verification results, unreasonable parameter values ​​are corrected to obtain the final updated parameters.

[0176] S3.3.4: Based on the updated grating complex amplitude distribution and line edge roughness power spectrum parameters, repeat the iteration. When the relative change of the L2 norm of the residual vector is less than a preset threshold, terminate the iteration and output the final grating complex amplitude distribution and line edge roughness power spectrum parameters.

[0177] In this embodiment, firstly, the updated grating complex amplitude distribution and line edge roughness power spectrum parameters are used as new initial parameters, and the iterative calculation process of steps S3.3.1 to S3.3.3 is repeated; in each iteration, the second norm value of the current residual vector is recorded to establish an iterative history data sequence.

[0178] Then, based on the L2 norm of the residual vector of the current iteration and the result of the previous iteration, the relative change is calculated; the relative change is compared with the preset convergence threshold to determine whether the iteration has reached the convergence condition, and the convergence judgment result is obtained; when the relative change is less than the preset threshold, the optimization process is considered to have converged, and the iterative calculation is terminated.

[0179] Furthermore, based on the convergence judgment result, the final optimized grating complex amplitude distribution and line edge roughness power spectrum parameters are output; the quality of the final optimized grating complex amplitude distribution and line edge roughness power spectrum parameters is evaluated, and the fitting accuracy index and parameter uncertainty information are calculated; based on the fitting accuracy index and parameter uncertainty information, the final grating complex amplitude distribution and line edge roughness power spectrum parameters are obtained through the joint forward physical model.

[0180] Finally, based on the convergence characteristics of the iterative process and the final fitting error, a diagnostic report of the optimization process is generated; key intermediate results and calculated statistical information are recorded to provide reference data for subsequent result analysis and method improvement; the final grating complex amplitude distribution and line edge roughness power spectrum parameters are output in a standardized format for grating performance evaluation and manufacturing process optimization.

[0181] Please see Figure 4 This embodiment provides a rapid detection system for micro / nano grating line deviations. The rapid deviation detection system includes a light source illumination module, an encoding modulation module, a scanning control module, an optical imaging module, and a data processing module, wherein:

[0182] The light source illumination module is used to provide a partially coherent light source;

[0183] The encoding and modulation module includes a phase modulator and an amplitude modulator, used to perform phase encoding modulation and amplitude encoding modulation on the incident light to generate a first illumination channel and a second illumination channel;

[0184] The scanning control module includes a scanning drive unit and a position configuration unit. The scanning drive unit is used to control the structured composite light field to perform fixed-point scanning at multiple preset scanning positions of the micro-nano grating under test. The position configuration unit is used to output preset position coordinates to provide a position reference for fixed-point scanning.

[0185] The optical imaging module includes a Fourier lens, a channel-separating optical element, and an imaging detector. The Fourier lens is used to focus multi-level diffracted light onto the Fourier plane. The channel-separating optical element is used to separate the diffracted light of each parallel illumination channel. The imaging detector is used to acquire composite diffraction spot distribution data of the Fourier plane in multiple frames.

[0186] The data processing module is used to perform rapid detection of grating line deviation, output the grating line deviation detection result, and generate the final adjustment instruction.

[0187] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0188] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0189] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0190] 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 embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for rapid detection of deviations in micro / nano grating lines, characterized in that, The rapid deviation detection method includes: The distribution data of the composite diffraction spot generated by the micro / nano grating under test in the Fourier plane is acquired by multi-frame continuous imaging, and the first-order intensity sequence and its time fluctuation data of the composite diffraction spot distribution data are calculated. Based on the first-order intensity sequence and its time fluctuation data, the actual second-order intensity correlation function characteristic of each lighting channel is calculated through channel separation and registration processing. The actual second-order intensity correlation function features are input into a preset joint forward physical model, and the reconstructed grating complex amplitude distribution and line edge roughness power spectrum parameters are obtained through iterative inversion. Based on the reconstructed complex amplitude distribution of the grating and the power spectrum parameters of the line edge roughness, the grating line deviation at the current coordinate position is calculated.

2. The method for rapid detection of micro / nano grating line deviation according to claim 1, characterized in that, Before acquiring composite diffraction spot distribution data, the method further includes: The incident light is spatially structured and modulated using an coded illumination method to generate a structured composite optical field comprising multiple coherent channels. This spatially structured modulation includes phase-coded modulation and amplitude-coded modulation, wherein: The phase-coded modulation applies a differentiated phase delay to different spatial regions of the incident light by loading a preset phase mask pattern to generate a first illumination channel, wherein the phase distribution of the first illumination channel is adjusted according to the phase delay distribution of the phase mask pattern. The amplitude coding modulation selectively transmits the incident light to different spatial regions by loading a preset amplitude mask pattern to generate a second illumination channel, wherein the amplitude distribution of the second illumination channel is adjusted by the transmittance distribution of the amplitude mask pattern. Based on the first and second illumination channels, multiple parallel illumination channels with preset spatial spectrum distributions are generated, wherein the coherence characteristics of the parallel illumination channels are different from each other. Based on the parallel illumination channels and coherence characteristics, a structured composite light field is generated by spatial superposition.

3. The method for rapid detection of micro / nano grating line deviation according to claim 2, characterized in that, The process of acquiring composite diffraction spot distribution data generated by the micro / nano grating under test in the Fourier plane through multi-frame continuous imaging, and calculating the first-order intensity sequence and its temporal fluctuation data of the composite diffraction spot distribution data, includes: Based on the composite diffraction spot distribution data under the structured composite light field illumination, continuous multi-frame real-time acquisition is performed through preset sampling frequency and total frame count requirements to obtain continuous multi-frame Fourier plane images of the composite diffraction spot distribution data. Based on the spatial distribution functions of the first and second illumination channels, the diffraction spot regions corresponding to each illumination channel are extracted from the multi-frame Fourier plane images using an image segmentation algorithm. Based on the diffraction spot regions, the integrated light intensity value of each diffraction spot region at each time is calculated using an integration operation method. Based on the integrated light intensity value, a light intensity response sequence for each illumination channel is constructed in chronological order to form a first-order intensity sequence that includes at least two illumination channels. Based on the first-order intensity sequence, the intensity variance and time autocorrelation function of each lighting channel are calculated using statistical analysis methods. Based on the intensity variance and time autocorrelation function, the time fluctuation data of each lighting channel are obtained.

4. The method for rapid detection of micro / nano grating line deviation according to claim 3, characterized in that, Based on the first-order intensity sequence and its temporal fluctuation data, the actual second-order intensity correlation function characteristic quantities of each lighting channel are calculated through channel separation and registration processing, including: Based on the center position coordinates and spatial distribution function of the first and second lighting channels, the theoretical diffraction position of each lighting channel in the Fourier plane is determined by a spatial positioning algorithm. The multi-frame Fourier plane images are processed by channel separation. The diffraction spot regions corresponding to each illumination channel are identified by the image segmentation algorithm. The actual center position of each diffraction spot region is located by the sub-pixel level registration algorithm. Based on the actual center position and the theoretical diffraction position, the first-order intensity sequence is corrected using a displacement correction algorithm to obtain the corrected first-order intensity sequence. Based on the corrected first-order intensity sequence, the second-order intensity correlation function between any two illumination channels is calculated through cross-correlation operation, and the peak value, full width at half maximum (FWHM), decay time constant, and coherence of the second-order intensity correlation function are extracted as feature parameters. Based on the aforementioned characteristic parameters, combined with the coherence characteristics and relative light intensity weights of the illumination channel, the actual second-order intensity correlation function characteristic quantities are obtained through normalization processing.

5. The method for rapid detection of micro / nano grating line deviation according to claim 4, characterized in that, The step of inputting the actual second-order intensity correlation function feature quantity into a preset joint forward physical model, and obtaining the reconstructed grating complex amplitude distribution and line edge roughness power spectrum parameters through iterative inversion includes: Based on the center position coordinates, spatial distribution function, coherence characteristics and relative light intensity weights of the first and second illumination channels, a joint forward physical model of micro-nano grating diffraction under coded illumination is constructed based on scalar diffraction theory. Based on the joint forward physical model, a mathematical mapping relationship is constructed from the complex amplitude distribution of the grating, the power spectrum parameters of the line edge roughness to the characteristic quantities of the theoretical second-order intensity correlation function; Based on the mathematical mapping relationship and the actual second-order intensity correlation function characteristic quantities, the residual between the theoretical second-order intensity correlation function characteristic quantities and the actual second-order intensity correlation function characteristic quantities is minimized through a nonlinear least squares iterative inversion algorithm, thereby obtaining the reconstructed grating complex amplitude distribution and line edge roughness power spectrum parameters.

6. The method for rapid detection of micro / nano grating line deviation according to claim 5, characterized in that, The nonlinear least squares iterative inversion algorithm includes: Based on the preset initial grating complex amplitude distribution and initial line edge roughness power spectrum parameters, the characteristic quantity of the theoretical second-order intensity correlation function is calculated through the joint forward physical model. The theoretical second-order intensity correlation function eigenvalues ​​and the actual second-order intensity correlation function eigenvalues ​​are compared to obtain the residual vector, and the corresponding Jacobian matrix is ​​calculated using the numerical differentiation method. Based on the residual vector and Jacobian matrix, the parameter update amount is calculated by the least squares method. Based on the parameter update amount, the updated grating complex amplitude distribution and line edge roughness power spectrum parameters are obtained. Based on the updated grating complex amplitude distribution and line edge roughness power spectrum parameters, the iteration is repeated. When the relative change in the L2 norm of the residual vector is less than a preset threshold, the iteration is terminated and the final grating complex amplitude distribution and line edge roughness power spectrum parameters are output.

7. The method for rapid detection of micro / nano grating line deviation according to claim 6, characterized in that, Based on the reconstructed complex amplitude distribution of the grating and the power spectrum parameters of the line edge roughness, the grating line deviation at the current coordinate position is calculated, including: Based on the reconstructed complex amplitude distribution of the grating, the amplitude distribution and phase distribution of the micro / nano grating under test are extracted, and the transmittance distribution and phase delay distribution of the micro / nano grating under test are calculated. Through threshold segmentation and edge detection algorithms, the geometric contour and boundary position information of the grating lines are extracted. Based on the geometric contour and boundary position information of the grating lines, the position deviation of the micro-nano grating lines at the current coordinate position is calculated, and the statistical results of the grating line deviation are obtained through statistical analysis. Based on the line edge roughness power spectrum parameters, the line edge roughness characteristic parameters are calculated by function integration. Based on the statistical results of the grating line deviation and the line edge roughness characteristic parameters, the grating line deviation at the current coordinate position is obtained through a preset quality standard.

8. The method for rapid detection of micro / nano grating line deviation according to claim 7, characterized in that, After calculating the grating line deviation at the current coordinate position, the method further includes: scanning multiple preset scanning positions of the micro / nano grating to be tested; constructing a two-dimensional deviation distribution map based on the grating line deviation and line edge roughness index at each scanning position; obtaining the grating line deviation detection result through abnormal region identification; and generating corresponding adjustment instructions.

9. The method for rapid detection of micro / nano grating line deviation according to claim 8, characterized in that, The step of obtaining grating line deviation detection results through abnormal region identification and generating corresponding adjustment instructions includes: Based on the two-dimensional deviation distribution map, anomaly thresholds are set, and abnormal nodes are located and merged into connected abnormal regions through spatial clustering. Based on the area and shape characteristics of the connected abnormal regions, the anomaly type is determined through pattern recognition; A re-examination strategy is formulated based on the anomaly type to obtain the re-examination result. The re-examination strategy includes performing local high-density scanning and extending the exposure time. Based on the re-inspection results, the true abnormal areas are determined through confidence testing, and the grating line deviation detection results and final adjustment instructions are generated.

10. A rapid detection system for micro / nano grating line deviation, used to implement the rapid detection method for micro / nano grating line deviation as described in any one of claims 1-9, characterized in that, The rapid deviation detection system includes a light source illumination module, an encoding and modulation module, a scanning control module, an optical imaging module, and a data processing module, wherein: The light source illumination module is used to provide a partially coherent light source; The encoding and modulation module includes a phase modulator and an amplitude modulator, used to perform phase encoding modulation and amplitude encoding modulation on the incident light to generate a first illumination channel and a second illumination channel; The scanning control module includes a scanning drive unit and a position configuration unit. The scanning drive unit is used to control the structured composite light field to perform fixed-point scanning at multiple preset scanning positions of the micro-nano grating under test. The position configuration unit is used to output preset position coordinates to provide a position reference for fixed-point scanning. The optical imaging module includes a Fourier lens, a channel-separating optical element, and an imaging detector. The Fourier lens is used to focus multi-level diffracted light onto the Fourier plane. The channel-separating optical element is used to separate the diffracted light of each parallel illumination channel. The imaging detector is used to acquire composite diffraction spot distribution data of the Fourier plane in multiple frames. The data processing module is used to perform rapid detection of grating line deviation, output the grating line deviation detection result, and generate the final adjustment instruction.

Images