Defect detection method for logic chip mask

Abstract

The invention discloses a logic chip mask defect detection method, and relates to the technical field of integrated circuit manufacturing, and the method comprises the following steps: S1, constructing a scattering response model of sub-wavelength inorganic particles on the surface of a mask, and constructing a scattering response model of sub-wavelength inorganic particles on the surface of the mask under a set multi-angle polarized light incidence condition; non-linear disturbance influence of inorganic particles with different particle sizes on the reflected light phase in each incident polarization state is simulated, and a scattering response template containing standard phase disturbance characteristics is generated; and S2, based on a scattering response template, carrying out pixel-by-pixel matching analysis on the acquired multi-angle polarization phase diagram on the surface of the mask, identifying a local area matched with the template, and marking the local area as a suspected scattering interference area. According to the method, through construction of the scattering response template and multi-angle polarization consistency analysis, accurate distinguishing of artifacts and real defects is achieved, the recognition precision and stability under complex interference are improved in combination with structure retention type phase reconstruction and dynamic threshold recognition, and mask quality control and photoetching yield improvement are facilitated.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of integrated circuit manufacturing, in particular to a logic chip mask plate defect detection method. BACKGROUND

[0002] The logic chip mask plate defect detection refers to a technical method for high-precision defect identification and analysis of a glass blank mask used in a photolithography process in the manufacturing process of a logic chip. If there are micro-defects such as particle contamination, bubbles, scratches, metal impurities or uneven thickness on the surface of the glass blank mask as the basic material for subsequent mask pattern preparation, the transfer precision of the mask pattern will be directly affected, and then the integrity and functional reliability of the chip circuit are affected. The logic chip has a complex structure and a high wiring density, and higher requirements are put forward for the optical quality and defect control of the mask plate. Therefore, defect detection needs to use high-resolution imaging, optical interference, deep learning algorithm or self-adaptive recognition model and the like to scan and analyze the entire optical area of the glass blank mask, accurately identify the microscopic defects affecting the pattern transfer quality, realize early screening and quality control, and ensure the yield and consistency of subsequent chip production.

[0003] Multi-angle polarimetric phase analysis is a high-sensitivity auxiliary means for logic chip mask plate defect detection, and is especially suitable for identifying micro-structural defects, internal stress abnormalities and coating uniformity disturbances in the surface and near-surface layer of the glass blank mask that are difficult to identify through traditional bright-field imaging or single-angle reflection detection. The method applies illumination light with different polarization states at multiple incident angles, collects and analyzes the phase delay, polarization state change and polarization elliptical parameter characteristics in the reflected light to capture the perturbation signals of the mask material in the local refractive index, thickness change and stress distribution, and realizes high-precision identification of defects such as surface fine scratches, subsurface optical non-uniform regions and local delamination of thin films. In the manufacturing link of logic chips, such micro-scale defects do not directly change the mask pattern structure, but may induce image distortion, line width deviation or local alignment error in the photolithography projection process, and then affect the functional integrity of the circuit and the yield of finished products. Therefore, multi-angle polarimetric phase analysis can be used as a key supplementary module of the traditional mask detection system, significantly improving the perception ability of potential hidden defects of the glass blank mask, and enhancing the precision and coverage depth of the overall quality control.

[0004] The prior art has the following disadvantages: In the existing logic chip mask defect detection process, especially when high sensitivity recognition is performed by using a multi-angle polarization phase analysis method, if there are inorganic microparticles with a particle size smaller than a detection wavelength (for example, smaller than 193 nm), such as silicon oxide, metal oxide particles or residual cleaning impurities, on the mask surface, a nonlinear light scattering response is easily caused. These sub-wavelength microparticles cannot be directly imaged by traditional optical resolution, but under multi-angle illumination of polarized light, the phase delay and polarization state change of the reflected light will produce a nonlinear enhancement effect, resulting in serious interference in the phase calculation and polarization ellipse reconstruction process of the system. Such scattering disturbance not only has high angle sensitivity and extremely low repeatability, but also easily induces abnormal coherence under a specific polarization state, thereby forming local "interference artifacts" or "phase distortion spots" in the detection image, which are highly similar to the characteristics of real structure defects, seriously interfering with the accuracy and stability of the subsequent defect recognition algorithm. Since such artifacts are not derived from the mask structure itself and lack stable reconfigurability, they often cause false judgment of qualified masks or missed detection of potential defects, ultimately affecting the precision of the photolithography process and the quality of the logic chip pattern transfer.

[0005] The above information disclosed in the background section is only intended to enhance the understanding of the background of the present disclosure, and thus it can include information that does not constitute the prior art known to those of ordinary skill in the art. SUMMARY

[0006] The purpose of the present application is to provide a logic chip mask defect detection method, which realizes accurate differentiation between artifacts and real defects by constructing a scattering response template and multi-angle polarization consistency analysis, improves recognition accuracy and stability under complex interference by combining structure-preserving phase reconstruction and dynamic threshold recognition, and helps mask quality control and photolithography yield improvement, to solve the problems in the above background.

[0007] In order to achieve the above purpose, the present application provides the following technical scheme: a logic chip mask defect detection method, comprising the following steps: S1, a scattering response model of sub-wavelength inorganic microparticles on the mask surface is constructed, under the condition of multi-angle polarization light incidence, the nonlinear disturbance effect of inorganic microparticles with different particle sizes on the phase of reflected light under each incident polarization state is simulated, and a scattering response template containing standard phase disturbance characteristics is generated; S2, based on the scattering response template, the multi-angle polarization phase map of the mask surface collected is analyzed pixel by pixel, the local area matched with the template is identified, and is marked as a suspected scattering interference area; S3, multi-angle polarization synchronous scanning data acquisition is performed on the suspected scattering interference area, the phase change sequence thereof under different polarization angles and incident directions is recorded, and a response consistency index is calculated, which is used to represent the optical behavior stability of the phase disturbance; S4, based on the response consistency index, a non-structural interference discrimination mechanism is established from two dimensions of phase disturbance stability and polarization response nonlinearity degree, suspected scattering interference regions are classified and judged, regions identified as non-structural scattering interference are identified, and an artifact distribution map is formed; S5, based on the artifact distribution map, local phase reconstruction is performed on the corresponding region in the original multi-angle polarization phase image, a structure-preserving filtering model is established by analyzing the phase gradient and trend continuity of the adjacent region of the artifact region, and the phase distribution of the artifact region is compensated and repaired; S6, based on the phase reconstructed image, a defect re-identification process is implemented, an identification threshold based on dynamic adjustment of image features is set, the identification accuracy of real micro-defect features is improved, and stable closed-loop detection control of the mask defects is realized under the condition of sub-wavelength scattering interference.

[0008] Preferably, step S1 comprises: An optical parameter library of inorganic microparticle materials including silicon oxide microparticles, metal oxide particles and sub-nanometer impurity particles is established; Based on the finite difference time domain method or the rigorous coupled wave analysis method, the optical response behavior of inorganic microparticles with different particle sizes under different attachment states on the mask surface and the near surface layer is simulated; The reflection phase distribution under each polarization angle and incident direction combination is extracted, and Fourier inversion and spatial filtering are performed thereon to obtain a local phase delay image and polarization ellipse deformation parameters; Based on the extracted disturbance feature set, a standard phase disturbance template is constructed, and the identification applicability and discrimination degree thereof are quantified by a phase disturbance correlation coefficient and a polarization ellipse similarity index.

[0009] Preferably, step S2 comprises: The multi-angle polarization phase image is normalized and spatially aligned to eliminate background noise and image distortion; According to the standard scattering response template, a sliding window matching operation is performed, the matching score between each position and the template is calculated, and a matching score map is generated; In the matching score map, high matching regions exceeding the threshold value are extracted, spatial clustering analysis and edge continuity evaluation are performed, and noise and isolated regions are removed; The identified high matching region is marked as a suspected scattering interference region, and a corresponding spatial mask map is generated for subsequent processing.

[0010] Preferably, step S3 comprises: For the marked suspected scattering interference region, a plurality of illumination combinations of incident angles and different polarization states are set, and multi-angle polarized light scanning is performed; Collecting the reflection phase image data under each angle and polarization state combination, constructing a time series data set containing the phase response sequence of each pixel point; Calculating the phase response variation amplitude, directionality and stability of the suspected scattering interference region under different polarization conditions, extracting its multi-angle response consistency index; According to the response consistency index, the region is classified, and the low consistency region is identified as a region with high scattering artifact probability and is marked.

[0011] Preferably, step S4 comprises: Extracting the phase response sequence of each suspected scattering interference region under different incident angles and polarization states, and calculating its response consistency index; Extracting the phase disturbance curve of the suspected scattering interference region in the polarization angle change process, and analyzing the nonlinear characteristics of its polarization response; According to the response consistency index and the polarization nonlinear characteristics, a discrimination mechanism is constructed to determine whether the region is a non-structural scattering interference region; The region determined as a non-structural scattering interference region is integrated to generate an artifact distribution map, which is used for subsequent image processing procedures.

[0012] Preferably, step S5 comprises: Load the generated artifact distribution map and align the coordinates with the original multi-angle polarization phase map, extract the region coordinate information marked as non-structural scattering interference, establish a mask index and determine the adjacent reference region; Constructing a plurality of adjacent reference bands in multiple directions around the artifact region, counting the phase change trend and gradient amplitude, and generating a local phase guide map to reflect the phase delay trend and boundary transition mode; According to the phase guide map, a structure-preserving filtering model is constructed to perform phase value compensation and reconstruction on the artifact region, retaining the boundary jump trend and restoring the internal phase continuous distribution; Fuse the compensated artifact region with the original phase map to form a complete repaired polarization phase map as the input basis for defect identification.

[0013] Preferably, step S6 comprises: Load the multi-angle polarization phase map after artifact removal and local phase reconstruction processing and perform preprocessing operation, and extract the phase distribution density, phase gradient change amplitude, edge definition and local peak distribution characteristics in the image; Based on the extracted image features, a dynamic threshold generation strategy is established to divide the image into sub-regions and set their respective recognition sensitivity levels to generate local recognition thresholds; The defect candidate area is extracted by using image gradient enhancement and multi-scale edge extraction method, and the secondary screening is carried out by combining phase change continuity, boundary closure and geometric structure characteristics, so that the false signals are removed and only the real defect area is reserved. The identified defect area is positioned and parameter labeled, and the defect area, shape and distribution position are output, and the defect confirmation process is completed.

[0014] In the above technical solution, the technical effects and advantages provided by the present application are as follows: The present application establishes a discrimination benchmark from physical modeling to image recognition by constructing a standard scattering response template, and introduces multi-angle polarization response consistency analysis to effectively identify the dynamic optical behavior characteristics of false signals, avoid misjudgment and missed detection. At the same time, combined with the phase diagram reconstruction of structure-preserving filtering and the defect re-identification process of dynamic threshold control, the robustness and sensitivity of the defect recognition system in complex interference background are comprehensively improved, realizing the technical leap of mask surface defect detection from static imaging recognition to dynamic behavior modeling. Finally, this method not only improves the detection accuracy and yield screening ability, but also provides high-reliability technical support for mask quality control in advanced photolithography process, which has significant engineering practical value and industrial popularization prospect. BRIEF DESCRIPTION OF DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments or prior art, the drawings needed in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments described in the present application, and other drawings can also be obtained by those skilled in the art according to these drawings.

[0016] Figure 1 The method flow chart of the logic chip mask defect detection method of the present application. DETAILED DESCRIPTION

[0017] Example implementations will now be described more fully with reference to the accompanying drawings. Example implementations may, however, be implemented in many different forms and should not be construed as limited to the examples set forth herein; rather, these example implementations are provided so that this disclosure will be thorough and complete, and will fully convey the inventive aspects to those skilled in the art.

[0018] The present application provides a logic chip mask defect detection method as shown in Figure 1 The present application provides a logic chip mask defect detection method as shown in S1, a scattering response model of subwavelength inorganic microparticles on the mask surface is constructed, under the condition of setting multi-angle polarized light incidence, the nonlinear disturbance influence of different particle sizes of inorganic microparticles on the phase of reflected light under each incident polarization state is simulated, and a scattering response template containing standard phase disturbance characteristics is generated; Aiming at the non-structural scattering interference problem caused by sub-wavelength inorganic particles in the process of logic chip mask defect detection, a method of constructing a mask surface sub-wavelength inorganic particle scattering response model and generating a standard phase disturbance feature template is proposed. This method combines optical simulation and polarization physical principles to completely reproduce the disturbance behavior of different particle sizes of inorganic particles under multi-angle polarized light irradiation conditions on the reflected phase information, including the following steps: Based on the optical electromagnetic field theory and the scattering characteristics of sub-wavelength particles, a material optical parameter library covering a variety of typical inorganic particles is established, including but not limited to silicon oxide particles, metal oxide particles, and sub-nanometer impurity particles that may be left over in the cleaning process. After the material parameter library is constructed, the finite difference time domain method (FDTD) or the rigorous coupled wave analysis method (RCWA) is used to simulate the optical response behavior of various particles in different attachment states on the mask surface and the near-surface layer, especially the phase disturbance patterns formed by particles of different particle sizes (such as 50nm, 100nm, 150nm) on the mask glass surface. In this step, the light irradiation conditions are 193nm wavelength ultraviolet polarized light, the incident angle is set at 10° intervals in the range of 0° to 75°, and various polarization states such as linear polarization, right circular polarization, and left elliptical polarization are applied for each angle to realize incident excitation simulation under optical interference.

[0019] The finite difference time domain method (FDTD) is a full-wave simulation method that uses time as a step variable, establishes a grid model in space, and numerically solves Maxwell's equations. It is suitable for simulating the local light scattering, electric field enhancement, and phase disturbance propagation process of complex three-dimensional microstructures, and can be used for high-precision analysis of the time-domain response of particles to local light fields.

[0020] The rigorous coupled wave analysis method (RCWA) is a frequency-domain analytical method based on Fourier expansion and the continuity of boundary conditions, and is particularly suitable for simulating two-dimensional optical interfaces with periodic or layered structures. It can be used to simulate the light reflection, transmission, and interference characteristics of particle layers in vertical structures and calculate their modulation behavior on the phase distribution of outgoing light.

[0021] In this step: FDTD is mainly used to simulate the scattering phase influence of random-shaped, non-periodic particles in any attachment state with high precision; RCWA is more suitable for simulating the light field modulation of periodic, regularly arranged particle structures, such as multi-particle arrays or thin film layered structures.

[0022] Specific implementation steps (FDTD / RCWA simulation as the core) A geometric model containing mask substrate and inorganic particles is constructed, the mask substrate is set as an optical glass material with a known refractive index, the thickness is set to be more than 500 μm, and the surface part is set as a scanning area. On the surface or near the surface layer, representative inorganic particles are placed in turn, including silica spheres of different particle sizes (such as 50 nm, 100 nm, and 150 nm), non-spherical metal oxide particles, and multiple groups of impurity particles with varying heights. Each particle model is set with different attachment states according to the contact angle, burial ratio, or attachment angle to simulate the particle deposition characteristics under real pollution conditions.

[0023] According to the lithography wavelength, a 193 nm wavelength plane polarized light is set as the incident light source, multiple incident angles are set at intervals of 10° between 0° and 75°, and the polarization state covers linear polarization, right-handed circular polarization, and left-handed elliptical polarization. The simulation area is divided into a three-dimensional grid area (FDTD) or a Fourier expansion cross-section (RCWA) according to the layers, the material parameters are assigned with optical constants, including the real part of the refractive index and the imaginary part of the extinction coefficient, and the optical contrast between the particles and the mask substrate is ensured to truly reflect the scattering phenomenon.

[0024] The full-wave simulation solving process is performed, the electric field and the magnetic field at each time are iteratively updated in the three-dimensional space in FDTD, and the complex amplitude change of the unit area on the reflection surface is observed; in RCWA, the multi-order diffraction components after the interaction of the incident light and the microstructure are solved, and the complex phase change of the reflected wave under different spatial frequency components is calculated. The phase distribution of the reflected light field is extracted through simulation, and the equal phase line image is drawn on the simulation surface to intuitively observe the local phase delay, deflection, and scattering distortion characteristics caused by the particles under different polarization incidence.

[0025] The obtained complex phase distribution results are processed, the local disturbance area is extracted through a high-pass spatial filter, and the phase shift peak value, gradient change range, and disturbance scale caused by the particles are quantitatively analyzed. Further, the simulation results under multiple polarization angles are combined, the multi-angle phase disturbance sequence corresponding to each particle size is calculated, and the Stokes parameter sequence and polarization principal axis rotation trend under elliptical polarization are derived to construct the standard disturbance feature vector required for subsequent template matching.

[0026] Through the above method, the finite difference time domain method (FDTD) and the rigorous coupled wave analysis method (RCWA) can accurately simulate the influence of different sub-wavelength particles on the phase characteristics of the mask surface in terms of local microstructure detail reconstruction and overall polarization interference modeling. With the aid of the simulation process, not only the mapping relationship between particle size and phase disturbance can be obtained, but also the characteristic polarization response curve can be extracted, providing high credibility of optical physics basis for subsequent artifact discrimination and defect identification.

[0027] The simulation results under each of the above incident conditions are high-precision phase calculation, by Fourier inversion and spatial filtering of the complex amplitude distribution of reflected light, extract the local phase delay image caused by scatterers, and further calculate the polarization elliptical deformation parameters, including ellipticity, polarization axis direction change and Stokes parameter change trend. For each set of simulation data, the most significant disturbance area for identification is extracted, and the geometric distribution, intensity gradient and spatial phase mutation characteristics are processed by feature vector coding. Through this process, the original three-dimensional phase disturbance data is converted into a measurable and different phase response feature set.

[0028] Fourier inversion and spatial filtering is a basic method in optical signal processing, which is used to extract local phase disturbance features from complex amplitude information and suppress background interference, especially in multi-angle polarization detection. The core purpose is to convert the original interference pattern or complex amplitude image into a spatial distribution map of distinguishable local phase changes, and calculate the polarization state change parameters accordingly.

[0029] The following are the specific steps of extracting the local phase delay image caused by scatterers by Fourier inversion and spatial filtering, and further calculating the polarization elliptical deformation parameters: Obtain the reflected complex amplitude distribution image of the mask surface under multi-angle polarized light irradiation, which is reconstructed from the interference signal and contains real and imaginary part information, recording the amplitude and phase state of the complex light field at any point in space.

[0030] Perform two-dimensional Fourier transform on the complex amplitude image to transform it from the spatial domain to the frequency domain. In the frequency domain, high-frequency information usually corresponds to edge features and rapidly changing phase regions, while low-frequency information reflects background light intensity or global slow trends. By selectively retaining specific frequency bands (such as high-pass or band-pass) and performing frequency domain filtering, large-scale optical background disturbances and systematic noise can be effectively suppressed.

[0031] Perform Fourier inversion (i.e. inverse Fourier transform) on the filtered frequency spectrum data to convert it back to the spatial domain and obtain a complex amplitude image that enhances local structural disturbance features. In this image, the phase delay distribution is obtained by analyzing the complex phase angle (such as the arctangent function to calculate the ratio of imaginary part to real part), which reflects the phase change characteristics caused by local particle disturbance.

[0032] The complex amplitudes of the same region under multiple polarization states are jointly analyzed by combining the reflection patterns under different polarization angles, and the polarization state change behavior of each pixel point is extracted. By calculating the Jones vector or the Stokes vector, the polarization elliptical deformation parameters of the pixel can be further obtained, including the ellipticity (ratio of major axis to minor axis), the polarization axis direction change (principal axis rotation angle), and the variation trend of the Stokes parameter with the polarization angle, so as to completely describe the polarization state distortion characteristics of the point. These polarization parameter changes are the optical abnormality caused by scattering interference, which can be used as important discriminant features in subsequent template comparison and interference identification.

[0033] Based on the extracted disturbance feature set, a response index matrix in the three-dimensional parameter space of particle size-polarization angle-incident direction is constructed. This matrix uniformly encodes all simulated particle sizes and their induced phase disturbance patterns under specific polarization light conditions, forming a standard feature sequence with discriminant ability. To improve the representativeness and robustness of the template, additive noise simulation under simulated disturbance and local light intensity gradient disturbance test are further introduced to ensure the adaptability of the template under common interference backgrounds in actual polarization detection systems. For each typical particle type and attachment state, at least 5 feature samples under different polarization angle combinations are extracted, and finally a high-dimensional standard phase disturbance template library is formed.

[0034] The generated standard templates are verified by a template training evaluation method, which specifically matches the constructed disturbance templates on a set of simulated defect phase maps point by point. By calculating the phase disturbance correlation coefficient and the polarization ellipse similarity index, the applicability and discriminability of the template in identifying subwavelength particle-induced phase distortion are quantified. If the correlation index is stable and higher than the set threshold (such as 0.9), the template is confirmed to have significant discriminant ability and is included in the standard template library; if it does not meet the discrimination requirements, it is fed back to the simulation link for material model or incident parameter modification and iteration. The final set of standard scattering response templates constructed will serve as a key reference for identifying non-structural scattering interference in subsequent polarization phase detection, providing stable, reliable, and quantifiable optical prior support for the mask defect identification process.

[0035] In order to quantify the applicability and discriminability of the standard scattering template in identifying subwavelength inorganic particle-induced phase distortion, the "phase disturbance correlation coefficient" and "polarization ellipse similarity index" are introduced as evaluation means for matching degree analysis from two dimensions of phase distribution pattern and polarization state change. This process does not rely on subjective judgment and has objective quantitative basis and strong discrimination ability, which includes the following steps: For each suspected phase disturbance region in the image to be detected, the local phase distribution map of the region under multiple angle polarization conditions is extracted, and is compared one by one with the typical phase disturbance map of the corresponding particle size and polarization angle in the standard scattering template. During the comparison process, the degree of coincidence of the two images in the spatial distribution form, the light and dark change trend, the edge transition position and the like is analyzed, whether the similarity is stable and concentrated, whether there is obvious pattern structure deviation or local mutation, so as to evaluate whether the phase distribution of the detection region conforms to the scattering disturbance mode described by the template.

[0036] The polarization state information corresponding to the above detection region is extracted, that is, the polarization response trajectory of the reflected light under illumination at multiple polarization angles, and is converted into a standardized polarization ellipse shape. The ellipse shape is compared with the polarization ellipse under the corresponding condition in the template, mainly focusing on whether the geometric characteristics such as the ratio of the major and minor axes of the ellipse, the rotation angle of the major axis of the ellipse, and the shift degree of the whole ellipse are consistent. If the geometric appearances of the two are basically coincident, it indicates that the polarization change process caused by scattering has high consistency.

[0037] The matching results of the phase pattern and the polarization ellipse in the first two steps are comprehensively evaluated, and a certain weight ratio is given, so as to obtain the overall matching score between each suspected region and the template. The higher the score, the closer the phase disturbance and the polarization state change of the region to the standard template characteristics, and the greater the possibility of being caused by subwavelength particles. For the regions with low scores, it indicates that the deviation is large, which may belong to other types of interference or not real scattering disturbance, and should not be included in the subsequent template matching and recognition process.

[0038] In actual engineering application, a plurality of templates of different types, different particle sizes and different attachment states can be subjected to the above evaluation process, and the matching accuracy and interference recognition ability are repeatedly tested in a plurality of groups of real detection images, the average recognition coverage and false positive rate are counted, and the universality and stability of the template are further confirmed. Through such a quantitative matching process, not only the effective screening and precise adaptation of the standard template are realized, but also a highly reliable optical prior basis is provided for the subsequent defect recognition algorithm, and the distinguishing ability and discrimination accuracy of the system for phase distortion caused by subwavelength scattering interference are significantly improved.

[0039] The main role of this step is to provide a physical accurate, structure identifiable, and algorithm identifiable standard reference basis for logic chip mask defect detection, to solve the problem that non-structural optical interference caused by sub-wavelength particles is difficult to distinguish in traditional defect recognition. In multi-angle polarization phase detection, since the inorganic particle size is smaller than the detection wavelength, the phase disturbance caused by the particle cannot be clearly identified by direct imaging, but the phase delay and polarization state change of the reflected light have obvious nonlinear characteristics. Therefore, by establishing an accurate optical simulation model, the phase disturbance behavior of particles of different sizes, different materials, and different attachment states under various polarization light irradiation conditions is simulated, and the characteristics of this non-structural interference are digitized and patterned, forming a "standard scattering template" with discriminant significance. This template not only contains the local phase delay pattern caused by the scatterer, but also records the modulation characteristics of the polarization elliptical shape, which serves as an important basis for identifying artifacts, excluding interference, and distinguishing real defects in subsequent image recognition. In short, this step not only realizes the conversion from "intangible disturbance" to "tangible template", but also provides a reliable physical prior and comparison basis for the entire defect detection process, which is a key technical basis for high-precision recognition, error control, and closed-loop detection.

[0040] S2, based on the scattering response template, pixel-by-pixel matching analysis is performed on the collected multi-angle polarization phase image of the mask surface, and the local area matched with the template is identified and marked as a suspected scattering interference area; Based on the constructed sub-wavelength inorganic particle scattering response template of the mask surface, pixel-by-pixel matching analysis is performed on the multi-angle polarization phase image obtained by actual detection to identify possible non-structural scattering interference areas in the image and provide basic support for subsequent discrimination and suppression. This step emphasizes high consistency retrieval of pixel-level accuracy and template feature space, which includes the following steps: Load and preprocess the multi-angle polarization phase image data to be analyzed. This data is usually a sequence of reflected phase images obtained by illuminating the mask surface at different polarization angles, with a multi-channel, multi-frame structure. To eliminate the interference of environmental light noise, system distortion, and uneven light intensity, the image is first normalized and spatially aligned to ensure consistency in spatial coordinates and gray scale amplitude. Then, use image enhancement methods to highlight local phase mutation areas, and calculate the initial disturbance response graph through spatial gradient calculation to provide a candidate area basis for matching analysis.

[0041] According to the standard scattering response template generated by the previous simulation, the normalized polarization phase map is subjected to a sliding window matching operation. This process takes the typical phase disturbance pattern in the template as a reference, slides pixel by pixel and region by region within the entire detection image range according to the set pixel window size, and calculates the similarity score between the current window content and the template. The scoring method is based on the consistency of pattern structure features, including the comparison of edge direction, phase gradient distribution, and local extreme position of the disturbance pattern. A matching score is obtained for each position, and the score value and coordinate information are recorded, forming a matching score map of the entire image scale.

[0042] In the obtained matching score map, all regions with scores exceeding the set threshold are selected as suspected scattering interference candidate regions. These regions usually exhibit small-scale pattern blocks with high similarity and high concentration, and have obvious non-natural phase disturbance characteristics. To further improve accuracy, clustering analysis and edge continuity analysis can be performed on these candidate regions to eliminate isolated points and noise interference regions. The clustering process considers the spatial connectivity of adjacent high-matching regions, merges regions with relevance into a whole, and evaluates the internal matching stability to ensure that the final screening result has structural similarity and spatial consistency.

[0043] The finally identified high-matching regions are uniformly marked as suspected scattering interference regions, and the corresponding spatial mask map is generated for subsequent processing. The mask map corresponds to the original phase map in size, where the suspected scattering interference regions are assigned a specific marker value to facilitate subsequent targeted implementation of multi-angle verification, response consistency detection, and phase reconstruction repair operations. This identification and marking process not only achieves efficient preliminary screening of complex scattering interference, but also greatly reduces the waste of subsequent computing resources in invalid regions, improving the accuracy and efficiency of the overall identification process.

[0044] The purpose of this step is to accurately screen and spatially locate the non-structural scattering interference area caused by sub-wavelength inorganic particles in the multi-angle polarization phase map, providing key input basis for subsequent interference elimination, response consistency verification and phase map reconstruction. In the process of mask defect detection, due to the instability, easy deformation and similarity with real defect characteristics of scattering artifacts, conventional full-image recognition methods cannot effectively distinguish the subtle differences between them and real defects, which can easily cause misjudgment or missed detection. By pre-constructing a standard scattering response template, the typical phase disturbance behavior of sub-wavelength particles at different polarization angles can be standardized, characterized, and made comparable and pattern recognizable. Therefore, this step identifies the area highly similar to the known scattering characteristics by scanning and comparing the template on the detection image in the image scale, essentially establishing a bridge from "random disturbance" to "template recognition". The labeled suspected interference area not only narrows down the interference evaluation range, reduces the overall computational complexity, but also improves the accuracy and efficiency of the subsequent processing flow. In addition, through pixel-by-pixel spatial matching, the boundary fuzzy, small-scale or sparse distributed interference patterns can be accurately extracted, solving the problem of weak disturbance signal recognition that traditional global threshold or overall structure recognition methods cannot solve. Therefore, this step is one of the core links to realize the closed loop of sub-wavelength optical interference recognition, with outstanding technical pertinence and practical value.

[0045] S3, multi-angle polarization synchronous scanning data acquisition is performed on the suspected scattering interference area, the phase change sequence of the area at different polarization angles and incident directions is recorded, and a response consistency index is calculated to represent the stability of the phase disturbance optical behavior of the area; In order to further confirm the optical properties of the suspected scattering interference area labeled and exclude the confusion with real structural defects, a data acquisition and analysis method based on multi-angle polarization synchronous scanning is proposed. This method analyzes the phase response change trend of the area at different polarization angles and incident directions, calculates the response consistency index to quantify the stability of the optical disturbance behavior, and thus assists in identifying whether the area belongs to non-structural scattering interference. This process not only has the advantages of non-contact and high resolution, but also has the ability of time sequence consistency analysis, which includes the following steps: On the mask surface to be detected, the suspected scattering interference area identified in the previous stage is repositioned and selected as the target acquisition area, a plurality of fixed incident angles (for example, from 0° to 75°, taking an angle every 10°) are set, and a set of light combinations with different polarization states is set for each angle, including typical states such as linear polarization, right circular polarization, and elliptical polarization. By precisely controlling the incident angle and polarization state of the light source, each sampling point can receive uniform and systematic light conditions in the area to maximize its scattering response behavior. The goal of this step is to establish a multi-dimensional scanning condition covering the optical angle space to capture the overall response of the microparticle to optical excitation.

[0046] Under the above conditions, the reflection phase image data of the area under each angle and each polarization state combination is sequentially acquired, and each group of images is time-sequentially archived to ensure the spatiotemporal consistency and data integrity of the sampling. The acquisition process uses a high-resolution phase interference imaging device to record the phase response values of each pixel point under different conditions, and constructs these values into a phase change curve in time sequence according to the incident angle and polarization state. The core of this step is to extract the multi-dimensional response trajectory of the target area to provide a comprehensive data basis for subsequent analysis of its behavior consistency.

[0047] The consistency of the collected phase change sequence data is calculated to analyze whether the phase response of each pixel point or area under different polarization illumination conditions has stability and regularity. The specific analysis method includes calculating the change amplitude, directional shift trend, and disturbance velocity gradient of the phase response of the area between different angles to determine whether it shows strong angle sensitivity, periodic jump, or extremely unstable response. If the response trajectory of a certain area under each condition is highly inconsistent, it indicates that its response behavior lacks repeatability and tends to be a non-structural scattering artifact; on the contrary, if the phase disturbance shows consistent trend, stable amplitude, and clear edge under multiple angles, it may be caused by a real structural defect. The calculation result is output in the form of a "response consistency index" to measure the stability of its behavior characteristics.

[0048] According to the response consistency index of each suspected area, it is divided into three categories: high consistency, partial consistency, and low consistency, and the low consistency area is further marked as a target area with a high probability of scattering artifact. This result can be used for subsequent non-structural response screening and artifact mask construction, and also provides quantitative input basis for image reconstruction and defect identification process. The entire acquisition and calculation process takes into account the time continuity and angle breadth to ensure the capture of complex nonlinear behavior caused by microparticle scattering, significantly improving the reliability of identifying artifacts.

[0049] The role of this step is to reveal the phase disturbance behavior of the suspected scattering interference area under different polarization illumination and incident direction conditions through dynamic and full-angle optical response observation, to calculate its response consistency index, to accurately judge whether the disturbance has the characteristics of non-structural scattering interference, and to provide a quantitative basis for accurate differentiation between real defects and artifacts. In actual detection, scattering artifacts caused by sub-wavelength inorganic particles often show strong angle sensitivity and polarization dependence, that is, their phase disturbance behavior does not have a stable and repeatable response mode under different illumination conditions; while the phase distortion caused by real structural defects often has high trend consistency and structural stability under different polarization or incident angles. Therefore, by implementing multi-angle, synchronized polarization scanning acquisition on the target area, this step can not only capture the phase evolution process in space, but also construct a complete phase response trajectory. Further, by analyzing the consistency of these trajectories, the response consistency index reflecting the stability of the optical behavior can be extracted. As an important basis for subsequent interference identification, this index can significantly enhance the objectivity, reliability and resolution of the identification process, avoiding misjudgment and missed detection caused by relying on single-angle or static images. At the same time, this method successfully transfers the scattering interference from the "appearance similarity" level to the "behavior difference" dimension for identification through dynamic and multi-dimensional observation means, greatly expanding the ability boundary of defect identification methods. It is one of the core technical links to realize high-precision and strong-robustness closed-loop control of mask defect detection, and has important engineering application value and creative significance.

[0050] S4, based on the response consistency index, establishing a non-structural interference discrimination mechanism from the two dimensions of phase disturbance stability and polarization response nonlinearity, classifying and judging the suspected scattering interference area, identifying the area as non-structural scattering interference, and forming an artifact distribution map; For the acquired suspected scattering interference area, in order to effectively distinguish real structural defects from non-structural scattering interference, a interference identification method based on response consistency index for classification and discrimination is proposed. This method constructs a discrimination logic from two key dimensions of phase disturbance stability and polarization response nonlinearity, combines actual acquisition data, accurately identifies and classifies the interference area, and finally outputs an artifact distribution map for subsequent repair processing. This method not only relies on the quantitative characterization of optical behavior, but also introduces dynamic analysis ideas, significantly improving the identification accuracy of scattering artifacts, including the following steps: On the basis of the completion of multi-angle polarization phase data acquisition, the phase response sequence recorded under all incidence conditions of each suspected area is called, and the corresponding response consistency index is calculated. The consistency index is used to measure the stability of the phase disturbance change of the area under different polarization angles and incidence angles, that is, whether there is a continuous, regular and repeatable change trend. Under normal circumstances, structural defects will produce relatively consistent phase disturbance at all observation angles due to the stability of physical entities; and non-structural scattering interference shows dramatic jumps, distorted shapes or irregular distribution of phase extreme values. Therefore, the calculation of the index provides a basic evaluation dimension for subsequent classification.

[0051] The response consistency index of the suspected scattering interference area is calculated, aiming to measure whether the phase disturbance behavior of the area under different polarization angles and incidence directions has stability and regularity, and the specific process is as follows: Extract the sequence of reflection phase values of the area under all polarization illumination conditions, and construct a multi-dimensional phase change trajectory graph with "polarization angle / incidence direction" as the horizontal axis and phase response value as the vertical axis; Perform trend analysis on the sequence, focusing on whether the phase value presents the characteristics of smooth continuity, consistent trend or shape repetition with angle change. If there are mutation points, inversion phenomena or irregular fluctuations, it is marked as unstable response behavior; Calculate the change amplitude difference, change direction consistency, mean deviation rate and other statistics between adjacent points in the entire sequence, to reflect whether the response curve of the area has internal coordination; The above analysis results are normalized to convert a response consistency index value between 0 and 1. The closer the value is to 1, the more stable and regular the phase response of the area under different illumination conditions is; on the contrary, if the value is significantly low, it may be caused by random scattering or interference artifacts. The generation of this index provides a quantitative basis for the identification of non-structural interference in the subsequent process, and is a key link to realize interference classification and artifact exclusion.

[0052] For the response data of each suspected area under different polarization illumination conditions, further extract the nonlinear features of the polarization response behavior, including the monotonicity, volatility, curvature change rate of the response curve and discontinuity in the polarization state conversion process. This process focuses on whether there is a mutation, inversion or high sensitivity feature in the phase disturbance of the area during the change of polarization angle. If a region shows nonlinear mutation behavior during the polarization state change process, it usually means that it comes from light scattering interference of subwavelength particles, rather than stable structural defects.

[0053] Based on the two dimensions of data, a non-structural interference discrimination mechanism is constructed. The mechanism is based on the pre-set discrimination threshold and logical judgment rule to comprehensively score and classify each suspected area. The judgment basis includes whether the response consistency index is lower than the stability threshold, whether the nonlinear feature of the polarization response is higher than the judgment boundary, and if the above conditions are met at the same time, the area is determined as a non-structural scattering interference area. In order to enhance the adaptability and robustness of the discrimination mechanism, a multi-area cross-validation and local mean comparison strategy can be introduced to further reduce the risk of misjudgment caused by accidental fluctuations.

[0054] All areas determined as non-structural scattering interference are integrated to form a ghost distribution map corresponding to the original phase map, and the position, boundary range and interference intensity level of each ghost area are marked in the map. The distribution map serves as the input basis for subsequent phase map repair and defect identification, making the processing flow have regional selectivity and response adaptability, avoiding indiscriminate processing of the entire image, and improving the calculation efficiency and identification accuracy. Through the implementation, the suspected disturbance in the static image is converted into a classified object based on dynamic optical behavior, so as to realize the exclusion of ghost and interference before real defect identification, and to provide stable and reliable data support for subsequent image reconstruction and high-precision identification.

[0055] The role of this step is to construct a non-structural scattering interference discrimination mechanism with quantifiable characteristics and logical judgment ability by introducing a two-dimensional optical behavior analysis method, so as to realize effective classification and screening of suspected scattering interference areas, and output clear artifact distribution map in space, laying a precise foundation for subsequent image repair and real defect identification. In actual mask defect detection, the scattering artifacts caused by sub-wavelength inorganic microparticles often exhibit non-structural characteristics such as unstable morphology, position drift, sensitivity to polarization conditions, and are easily confused with real structural defects, leading to misjudgment or missed detection. Therefore, relying only on image static features for identification often fails to achieve ideal results. This step comprehensively evaluates the phase disturbance stability of each suspected area under different incident angles and polarization states by introducing the "response consistency index", and assists with nonlinear behavior analysis of the polarization response curve to capture typical interference characteristics such as dramatic jumps, morphological mutations, or irregular changes in response direction. Through joint analysis of these two dimensions, real defect areas with stable and continuous response and scattering interference areas with irregular and non-repeating response can be effectively distinguished, significantly improving the reliability and resolution of the identification process. The final artifact distribution map not only clearly marks the location and shape range of the interference area, but also provides accurate mask information for subsequent steps, allowing phase map reconstruction and real defect identification to avoid interference sources and focus on effective signal processing. The implementation of this step enables the detection system to make judgments based on the nature of optical behavior when faced with complex and variable micro-scale interference factors, representing a key technical breakthrough from "static image judgment" to "dynamic response modeling", with high engineering practicality and innovative value.

[0056] S5, based on the artifact distribution map, locally reconstructing the phase of the corresponding area in the original multi-angle polarization phase map, establishing a structure-preserving filtering model by analyzing the phase gradient and trend continuity of the adjacent area of the artifact area, and compensating and repairing the phase distribution of the artifact area; In order to eliminate the influence of local phase distortion caused by non-structural scattering interference and restore the structural integrity and continuity of the original mask surface reflection phase map, a local phase reconstruction method based on artifact distribution map is proposed. This method constructs a structure-preserving filtering model by analyzing the phase gradient relationship and change trend between the artifact area and its adjacent area, and compensates and repairs the phase data of the disturbed area. This method not only effectively suppresses the interference of scattering artifacts on the overall identification accuracy, but also maintains the boundary clarity and detail continuity of the image structure, including the following steps: Load the artifact distribution map generated in the previous stage and align the coordinates with the original multi-angle polarization phase map to accurately extract the coordinates of all areas marked as non-structural scattering interference. For each artifact area, establish a corresponding mask index and use it as the repair target in phase map reconstruction. At the same time, extract the normal area adjacent to the artifact area as a reference area for subsequent phase trend and gradient analysis to obtain spatially continuous phase information support.

[0057] For each artifact area, construct a plurality of adjacent reference bands in multiple directions around it, and count the phase change trend and gradient amplitude in each direction to form a local phase guide map. This guide map reflects the phase change logic of the area in spatial distribution, focusing on identifying the trend direction, slope change, and boundary transition mode of phase delay. During extraction, extreme outliers can be excluded to ensure that the established phase gradient trend is physically interpretable and structurally reasonable. The key to this step is to capture the phase continuity characteristics of the real structure area and transfer them to the artifact area as a basis for repair.

[0058] According to the established phase guide map, construct a structure-preserving filtering model and apply it to the phase reconstruction of the artifact area. This filtering model is different from traditional mean filtering or Gaussian filtering. Its design idea is to "preserve boundaries and smooth interiors", that is, to preserve the phase jump trend consistent with the adjacent area at the edge of the artifact area and gradually restore the distribution pattern similar to the surrounding phase change in the interior. The model can be executed based on iterative estimation, which propagates reference gradient values layer by layer to reconstruct the phase value of each pixel in the artifact area, ensuring local continuity while avoiding false intervention on the edges of real structures.

[0059] After the phase reconstruction of all artifact areas is completed, the compensated areas are fused with the original phase map to generate a complete, continuous, and structure-preserving repaired polarization phase map. This map serves as the input image for subsequent defect identification, effectively eliminating artifact interference, reducing false positives, and preserving the integrity of original structural defect information. In practical engineering applications, this reconstruction process not only improves the overall quality and data reliability of multi-angle polarization detection images, but also provides a clearer and more stable input basis for defect feature extraction algorithms.

[0060] The purpose of this step is to perform a targeted, structure-preserving phase map repair process on the regions identified as non-structural scattering interference to eliminate the impact of artifacts on the overall detection image, restore the spatial continuity and structural consistency of the phase data, and thus provide a high-quality image input basis for subsequent real defect identification. In the multi-angle polarization phase detection process, the scattering artifacts caused by sub-wavelength inorganic microparticles will form random and irregular phase disturbance regions in the image, often appearing as local phase jump anomalies or reflected phase offset distortions. If these interference regions are not processed, they will not only affect the defect extraction algorithm based on image gradient or morphological features, but also may be misidentified as effective structural defects in subsequent feature classification, thereby causing identification errors or process evaluation deviations. Therefore, this step is based on the artifact distribution map identified in the previous stage to accurately lock the disturbed regions in the image and extract the phase gradient and change trend of the adjacent regions to analyze the phase extension rules of normal structures in spatial distribution. Subsequently, a structure-preserving filtering model with the dual capabilities of "boundary preservation and internal smoothing" is established to specifically compensate for the phase of the artifact regions. In the compensation process, the model does not simply replace the surrounding values with an average, but intelligently restores the distorted or contaminated phase information in the disturbance region while maintaining the coherence of the original structure profile and phase trend, ensuring that the repaired image has no interference residues and does not damage the real structure features. Finally, the reconstructed image presents clearer phase structure boundaries in the visual level and has higher feature recognition accuracy in the algorithm level, greatly improving the robustness and reliability of the entire defect detection system. As can be seen, this step is not only a processing link after artifact identification, but also a key bridge connecting artifact removal and real defect extraction, with outstanding technical specificity and engineering practical value.

[0061] S6, on the basis of the phase reconstructed image, a defect re-identification process is implemented, an identification threshold based on dynamic adjustment of the detection image features is set, the identification accuracy of real small defects is improved, and stable closed-loop detection control of the mask defects is realized under the condition of sub-wavelength scattering interference. In order to further improve the identification accuracy and stability of real defects on the basis of the phase map after the compensation and repair of the artifact region, a re-identification process based on dynamic adjustment of the identification threshold of the image features is proposed. This process fully utilizes the high signal-to-noise ratio and interference suppression effect of the phase reconstructed image, realizes the accurate extraction of real small defects by constructing an adaptive threshold control mechanism, and is particularly suitable for logic chip mask quality detection under the background of sub-wavelength scattering interference. The entire implementation process not only ensures the stable operation of the identification algorithm in complex background, but also builds a closed-loop control path in the detection logic, which includes the following steps: The multi-angle polarization phase map after artifact removal and local phase reconstruction processing is loaded as an input image, and the image is preprocessed, including brightness equalization, edge enhancement, noise suppression, etc., to further optimize the image quality. Subsequently, key statistical feature information is extracted in the global range of the image, such as phase value distribution density, gradient change amplitude, edge contour definition, and local phase peak distribution, which will be used as the basis for dynamic recognition threshold regulation to judge the defect feature saliency and abnormality degree of different regions.

[0062] Based on the spatial feature distribution extracted from the image, a dynamic threshold generation strategy is established, and the traditional fixed threshold method is no longer used for recognition and judgment. Specifically, the image is divided into several sub-regions, and each sub-region is automatically matched with the corresponding recognition sensitivity level according to its phase gradient intensity, noise background level, and structural texture complexity, and a local recognition threshold is generated accordingly. In low-noise, clear structure areas, higher sensitivity is set to enhance the ability to capture small defects; in texture complex or background fluctuation strong areas, relatively loose threshold is set to avoid false detection. This dynamic adjustment mechanism can adapt to the changes of detection difficulty in different regions and improve the overall recognition accuracy.

[0063] An image gradient enhancement and multi-scale edge extraction-based method is used for defect target extraction operation, and all abnormal phase structures that exceed the local recognition threshold range are extracted as potential defect candidate regions. To further remove false recognition areas, phase change continuity, boundary closure, and geometric structure features can be combined to perform secondary screening on the candidate regions, excluding scattered noise residues or non-structural disturbance-induced false signals, and only retaining real defect feature regions with clear structure boundaries and stable optical response. This recognition process adds a multi-verification mechanism based on dynamic regulation, effectively improving the detection ability of weak defects, especially suitable for small abnormality recognition in high-density mask patterns.

[0064] The finally recognized defect regions are positioned and output, and are superimposed and labeled with the original image coordinate system, and key parameters such as defect area, shape, distribution position, etc. are output, providing data basis for subsequent defect statistics, classification, grading, and production process feedback. Combined with the whole process of artifact recognition, phase reconstruction, and dynamic re-recognition, this process completes the full closed-loop technical path from interference removal to defect confirmation, significantly improving the practicality, reliability, and recognition accuracy of the detection system in the presence of sub-wavelength interference.

[0065] The role of this step is to carry out a targeted defect re-identification operation on the processed image after completing the non-structural scattering interference identification and local phase map reconstruction, so as to realize accurate extraction and judgment of real micro defects, and thus to construct a complete detection closed loop with stable identification capability. In the process of mask defect detection, affected by the sub-wavelength particle scattering artifact, the traditional identification method is easy to misjudge the noise as a defect or to bury the real defect in the background disturbance, especially in the area with high image noise intensity fluctuation and high structure density, which is more likely to cause the recognition accuracy to decrease. After the phase map is reconstructed by the previous step, the image has clearer structure edge and more stable phase gradient distribution. At this time, if the fixed identification threshold is still used, the local characteristics of different regions in the image cannot be considered. Therefore, this step introduces a dynamic identification threshold mechanism driven by image features, combines the texture complexity, phase gradient strength and structure continuity in different regions of the image, and adaptively adjusts the sensitivity of defect identification, so that the identification algorithm can capture subtle defects in low interference areas and avoid misidentification in high background fluctuation areas, thereby maximizing the accuracy and robustness of defect extraction. In addition, this step also introduces multi-scale image analysis and edge enhancement strategy to further enhance the response ability to small and local asymmetric structure defects. Finally, through spatial positioning, attribute extraction and output labeling of the identification results, high reliability data support is provided for subsequent quality evaluation and process adjustment, and the closed loop control goal of realizing stable output capability of the detection process under the condition of interference is achieved.

[0066] The present application establishes a discrimination benchmark from physical modeling to image recognition by constructing a standard scattering response template, and also introduces multi-angle polarization response consistency analysis to effectively identify the dynamic optical behavior characteristics of artifacts and avoid misjudgment and missed detection. At the same time, combined with the phase map reconstruction of structure-preserving filtering and the defect re-identification process of dynamic threshold regulation, the robustness and sensitivity of the defect identification system in the complex interference background are comprehensively improved, realizing the technical leap of mask surface defect detection from static imaging recognition to dynamic behavior modeling. Finally, this method not only improves the detection accuracy and yield screening ability, but also provides high reliability technical support for mask quality control in advanced photolithography process, which has significant engineering practical value and industrial popularization prospect.

[0067] The above only describes certain exemplary embodiments of the present application by way of illustration, and it is needless to say that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present application. Therefore, the above figures and description are illustrative in nature and should not be understood as limiting the scope of protection of the claims of the present application.

[0068] It should be noted that, in the present document, relational terms are used to convey a relationship of one entity or action to another entity or action. For example, without necessarily implying any actual relationship or order between entities or actions, the terms "first," "second," "top" and "bottom" are used to name different entities and actions, and are used to distinguish one entity or action from another entity or action, without necessarily conveying any actual relationship or order between such entities or actions. Furthermore, the terms "comprise," "include," and "have," and variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises, includes, or has a list of elements is not necessarily limited to those elements, but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without limitation, an element preceded by "comprises... a," "includes... a," or "has... a" does not, without more constraints, foreclose the existence of additional identical elements other than the one listed or other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0069] It should be understood that the sequence numbers of the processes described above do not mean execution sequence in various embodiments of the present application, and the execution sequence of the processes should be determined according to their functions and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0070] Those skilled in the art can clearly understand that the units and algorithm steps of the examples described in combination with the embodiments disclosed in the present document can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software mode depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0071] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0072] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e. they can be located in one place or distributed on multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiments of the present application.

[0073] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit.

[0074] The above merely provides the specific implementation of the present application, but the protection scope of the present application is not limited to this. Any person skilled in the art can easily think of the changes or replacements within the technical range disclosed by the present application, which should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0075] The foregoing merely describes certain exemplary embodiments of this application by way of illustration. Obviously, for those skilled in the art, modifications and changes can be made to the described embodiments in various different ways without departing from the spirit and scope of the application. Therefore, the above drawings and descriptions are illustrative in nature, and should not be construed as limiting the scope of the claims of the present application.

Claims

1. A method of detecting defects in a logic chip mask, characterized by, The method comprises the following steps: S1, constructing a scattering response model of the mask surface subwavelength inorganic microparticles, simulating the nonlinear disturbance effect of inorganic microparticles of different particle sizes on the reflected light phase under different incident polarization states under the condition of setting multi-angle polarized light incidence, and generating a scattering response template containing standard phase disturbance characteristics; S2, based on the scattering response template, performing pixel-by-pixel matching analysis on the collected multi-angle polarization phase map of the mask surface, identifying the local area matched with the template as a suspected scattering interference area, and marking the suspected scattering interference area; S3, performing multi-angle polarization synchronous scanning data acquisition on the suspected scattering interference area, recording the phase change sequence under different polarization angles and incident directions, and calculating a response consistency index for characterizing the optical behavior stability of the phase disturbance; S4, based on the response consistency index, establishing a non-structural interference discrimination mechanism from the two dimensions of phase disturbance stability and polarization response nonlinearity, classifying and judging the suspected scattering interference area, identifying the area as non-structural scattering interference, and forming an artifact distribution map; S5, based on the artifact distribution map, performing local phase reconstruction on the corresponding area in the original multi-angle polarization phase map, establishing a structure-preserving filtering model by analyzing the phase gradient and trend continuity of the adjacent area of the artifact area, and compensating and repairing the phase distribution of the artifact area; S6, based on the phase reconstructed image, performing a defect re-identification process, setting an identification threshold based on the dynamic adjustment of the image features, improving the identification accuracy of the real micro defect features, and realizing stable closed-loop detection control of the mask defects under the condition of subwavelength scattering interference.

2. The method of claim 1, wherein, Step S1 comprises: Establishing an inorganic microparticle material optical parameter library including silicon oxide microparticles, metal oxide particles and sub-nanometer impurity particles; Based on the finite difference time domain method or the rigorous coupled wave analysis method, the optical response behavior of inorganic microparticles of different particle sizes under different attachment states on the mask surface and the near surface is simulated; Extracting the reflection phase distribution under each polarization angle and incident direction combination, and performing Fourier inversion and spatial filtering to obtain a local phase delay image and a polarization ellipse deformation parameter; Based on the extracted disturbance feature set, a standard phase disturbance template is constructed, and the identification applicability and discrimination degree are quantified by phase disturbance correlation coefficient and polarization ellipse similarity index.

3. The method of claim 1, wherein the method further comprises: Step S2 comprises: Normalizing and spatially aligning the multi-angle polarization phase image to eliminate background noise and image distortion; According to the standard scattering response template, perform sliding window matching operation, calculate the matching score between each position and the template, and generate a matching score map; Extract the high matching area exceeding the threshold value in the matching score map, perform spatial clustering analysis and edge continuity evaluation, and eliminate noise and isolated areas; The identified high matching area is marked as a suspected scattering interference area, and a corresponding spatial mask map is generated for subsequent processing.

4. The method of claim 1, wherein, Step S3 comprises: For the marked suspected scattering interference area, set a plurality of illumination combinations of incident angles and different polarization states, and perform multi-angle polarization light scanning; Collecting the reflection phase image data under each angle and polarization state combination, constructing a time series data set containing the phase response sequence of each pixel point; Calculating the phase response change amplitude, directionality and stability of the suspected scattering interference region under different polarization conditions, extracting its multi-angle response consistency index; According to the response consistency index, the region is classified, and the low consistency region is identified as a region with high scattering artifact probability and is marked.

5. The method of claim 1, wherein, Step S4 includes: Extracting the phase response sequence of each suspected scattering interference region under different incident angles and polarization states, calculating its response consistency index; Extracting the phase disturbance curve of the suspected scattering interference region in the polarization angle change process, analyzing the nonlinear characteristics of its polarization response; According to the response consistency index and polarization nonlinear characteristics, a discrimination mechanism is constructed to determine whether the region is a non-structural scattering interference region; The region determined as a non-structural scattering interference region is integrated to generate an artifact distribution map, which is used for subsequent image processing.

6. The method of logic reticle defect detection of claim 1, wherein, Step S5 includes: Load the generated artifact distribution map and align the coordinates with the original multi-angle polarization phase map, extract the region coordinate information marked as non-structural scattering interference, establish a mask index and determine the adjacent reference region; Constructing a plurality of adjacent reference bands in multiple directions around the artifact region, counting the phase change trend and gradient amplitude, and generating a local phase guide map to reflect the phase delay trend and boundary transition mode; According to the phase guide map, a structure-preserving filtering model is constructed to perform phase value compensation and reconstruction on the artifact region, retain the boundary jump trend and restore the internal phase continuous distribution; Fuse the compensated artifact region with the original phase map to form a complete repaired polarization phase map as the input basis for defect identification.

7. The method of logic reticle defect detection of claim 1, wherein, Step S6 includes: Load the multi-angle polarization phase map after artifact removal and local phase reconstruction processing and perform preprocessing operations, and extract the phase distribution density, phase gradient change amplitude, edge definition and local peak distribution characteristics in the image; Based on the extracted image features, a dynamic threshold generation strategy is established to divide the image into sub-regions and set their respective recognition sensitivity levels to generate local recognition thresholds; Use image gradient enhancement and multi-scale edge extraction method to extract defect candidate area, and combine phase change continuity, boundary closure and geometric structure features for secondary screening to remove false signals and only keep real defect regions; Coordinate positioning and parameter labeling are performed on the identified defect regions, and the defect area, shape and distribution position are output, completing the defect confirmation process.

Images