Method, product and apparatus for correcting scanning electron microscope images

Abstract

This invention provides a method, product, and device for correcting electron microscope (SEM) scan images. The method includes: acquiring the SEM scan image to be corrected and its corresponding design layout; extracting an ideal SEM scan image of the corresponding region of the SEM scan image to be corrected from the design layout; determining corner matching pairs between the SEM scan image to be corrected and the ideal SEM scan image; estimating the perspective transformation matrix between the two images based on the corner matching pairs; performing perspective correction on the SEM scan image to be corrected using the perspective transformation matrix to obtain a perspective-corrected SEM scan image; and outputting the perspective-corrected SEM scan image. This method enables rapid analysis and compensation of SEM images, outputting geometrically distortion-free images without relying on hardware calibration parameters, thus significantly improving the spatial accuracy and automated correction capabilities of SEM images.

Description

Technical Field

[0001] This invention relates to the field of semiconductor technology, and in particular to a method, product, and apparatus for correcting electron microscope scan images. Background Technology

[0002] With the continuous development of semiconductor technology, the feature size of integrated circuits is shrinking, reaching the nanometer and even sub-nanometer scale. Current advanced process nodes place stringent demands on the precision and yield control of manufacturing processes. In this context, any minute physical defects, particulate contamination, pattern distortion, or electrical anomalies can lead to device performance degradation or even functional failure. Therefore, defect detection has become an indispensable and critical step in the semiconductor manufacturing process.

[0003] In advanced semiconductor manufacturing, electron beam inspection (EBI), as a core method for identifying nanoscale defects, faces a fundamental trade-off between throughput and image quality. As process nodes continue to shrink, the number of chip layers increases significantly, requiring multiple defect inspections at each layer to ensure yield. However, the electron beam in an EBI machine must scan the sample surface point by point, resulting in a significantly lower single-shot imaging speed compared to optical inspection. Therefore, increasing wafers per hour (WPH) while maintaining inspection accuracy has become a key bottleneck for the large-scale deployment of EBI technology in mass production environments.

[0004] The most direct strategy to improve wafer per inch (WPH) is to expand the field of view (FOV). A larger FOV means that each scan can cover a larger wafer area, significantly reducing electron beam movement time and feature matching time, thereby reducing the number of scans and total time required to complete the entire wafer inspection and significantly improving inspection efficiency. However, the introduction of a large FOV introduces severe image distortion problems, which seriously restrict the accuracy and reliability of defect detection. Existing technologies face multiple technical bottlenecks in addressing geometric distortion in scanning electron microscopy (SEM) images, including cumbersome operation, low automation, poor timeliness, limited accuracy, and high cost. Therefore, there is an urgent need to propose a technical solution that can achieve automatic, accurate, and hierarchical distortion correction based on real design information without relying on hardware calibration parameters. Summary of the Invention

[0005] In view of the above problems, the present invention proposes a method, product and device for correcting electron microscope scanning images that overcomes or at least partially solves the above problems.

[0006] One object of the present invention is to eliminate or reduce geometric distortion in SEM images; Another further objective of this invention is to reduce the cost of SEM image correction and achieve automated correction.

[0007] In particular, the present invention provides a method for correcting electron microscope scanned images, comprising: Acquire the electron microscopy scan image to be calibrated and the corresponding design layout, and extract the ideal electron microscopy scan image of the region corresponding to the electron microscopy scan image to be calibrated from the design layout; Determine the corner matching pairs between the electron microscope scan image to be corrected and the ideal electron microscope scan image, and estimate the perspective transformation matrix between the electron microscope scan image to be corrected and the ideal electron microscope scan image based on the corner matching pairs; A perspective correction electron microscope image is obtained by using a perspective transformation matrix to perform perspective correction on the electron microscope scan image to be corrected. Output fluoroscopic corrected electron microscope scan images.

[0008] Optionally, after obtaining the transmission-corrected electron microscope scan image, the method further includes: A distortion compensation model was established based on the fluoroscopically corrected electron microscope scan image and the ideal electron microscope scan image; Global distortion compensation was performed on the fluoroscopic correction electron microscope scan image using a distortion compensation model to obtain a distortion-compensated electron microscope scan image. Output distortion-compensated electron microscope scan images.

[0009] Optionally, the steps of establishing a distortion compensation model based on the perspective-corrected electron microscope scan image and the ideal electron microscope scan image include: Extract feature point matching pairs between the perspective-corrected electron microscopy scan image and the ideal electron microscopy scan image; Obtain residual distortion displacement information corresponding to feature point matching pairs in the fluoroscopic corrected electron microscope scan image; The distortion compensation model is obtained by fitting and solving the feature point matching pairs and residual distortion displacement information.

[0010] Optionally, the steps of fitting and solving the distortion compensation model based on the feature point matching pairs and residual distortion displacement information include: obtaining the radial distortion coefficient and tangential distortion coefficient by fitting and solving the feature point matching pairs and residual distortion displacement information using the least squares method; and establishing the distortion compensation model based on the radial distortion coefficient and tangential distortion coefficient. The steps for global distortion compensation of fluoroscopic corrected electron microscope scanned images using a distortion compensation model include: calculating the distortion compensation parameters corresponding to the fluoroscopic corrected electron microscope scanned images using the distortion compensation model; and applying the distortion compensation parameters to the fluoroscopic corrected electron microscope scanned images to achieve global distortion compensation.

[0011] Optionally, after the step of outputting the distortion-compensated electron microscope scan image, the method further includes: The perspective transformation matrix and distortion compensation model are fed back to the electron beam scanning hardware system; In the electron beam scanning hardware system, the parameters of the electron beam scanning hardware system are modified by applying the inverse perspective transformation based on the perspective transformation matrix and the displacement calculated by applying the distortion compensation model in the radial and tangential distortion.

[0012] Optionally, before determining the corner matching pairs between the electron microscope scan image to be corrected and the ideal electron microscope scan image, the method further includes: preprocessing the electron microscope scan image to be corrected to reduce noise and improve contrast in the electron microscope scan image to be corrected; After performing global distortion compensation on the perspective-corrected electron microscope scan image using the distortion compensation model, the process further includes: re-extracting the correction feature point matching pairs between the distortion-compensated electron microscope scan image and the ideal electron microscope scan image; verifying whether the distortion-compensated electron microscope scan image meets the preset conditions based on the correction feature point matching pairs; if so, then executing the step of outputting the distortion-compensated electron microscope scan image.

[0013] Optionally, the step of determining the corner matching pairs between the electron microscopy scan image to be corrected and the ideal electron microscopy scan image includes: Multiple key feature points were identified in the electron microscope scanned image to be corrected using a corner detection algorithm; Based on the geometric structure of the ideal electron microscope scan image, multiple matching feature points corresponding to the key feature points were simulated and calculated. Key feature points and matching feature points are matched to obtain multiple sets of corner point matching pairs.

[0014] Optionally, the step of matching key feature points and matching feature points includes: matching key feature points and matching feature points using a nearest neighbor search method to obtain multiple sets of corner point matching pairs; Algorithms for estimating perspective transformation matrices include: random sampling consensus algorithm.

[0015] According to another aspect of the present invention, a computer program product is also provided, comprising a computer program that, when executed by a processor, implements the steps of the electron microscope scanning image correction method described above.

[0016] According to another aspect of the present invention, a computer device is also provided, including a memory, a processor, and a machine-executable program stored in the memory and running on the processor, wherein the processor executes the machine-executable program to implement the steps of the correction method for electron microscope scanning images described above.

[0017] The electron microscope (SEM) image correction method of the present invention first acquires the SEM image to be corrected and the corresponding design layout, and extracts the ideal SEM image of the corresponding region of the SEM image to be corrected from the design layout; then, it determines the corner matching pairs between the SEM image to be corrected and the ideal SEM image, and estimates the perspective transformation matrix between the SEM image to be corrected and the ideal SEM image based on the corner matching pairs; it then uses the perspective transformation matrix to perform perspective correction on the SEM image to be corrected, obtaining a perspective-corrected SEM image; and finally, it outputs the perspective-corrected SEM image. This method enables rapid analysis and compensation of distortion under the current imaging state, and ultimately outputs the corrected image without relying on hardware calibration parameters, thereby significantly improving the spatial accuracy and automated correction capability of SEM images.

[0018] Furthermore, the electron microscope scanning image correction method of the present invention, after obtaining the perspective-corrected electron microscope scanning image, can also establish a distortion compensation model based on the perspective-corrected electron microscope scanning image and the ideal electron microscope scanning image; use the distortion compensation model to perform global distortion compensation on the perspective-corrected electron microscope scanning image to obtain a distortion-compensated electron microscope scanning image; and finally output the distortion-compensated electron microscope scanning image. This method enables further fine compensation of the perspective-corrected electron microscope scanning image, ultimately outputting a geometrically distortion-free image, thereby further improving the spatial accuracy and automated correction capability of SEM images.

[0019] The above and other objects, advantages and features of the present invention will become more apparent to those skilled in the art from the following detailed description of specific embodiments of the invention in conjunction with the accompanying drawings. Attached Figure Description

[0020] The following sections will describe some specific embodiments of the invention in a detailed manner by way of example and not limitation, with reference to the accompanying drawings. The same reference numerals in the drawings denote the same or similar parts or portions. Those skilled in the art should understand that these drawings are not necessarily drawn to scale. In the drawings: Figure 1 This is a schematic flowchart of a method for correcting electron microscope scanned images according to an embodiment of the present invention; Figure 2 This is a schematic flowchart of a method for correcting electron microscope scanned images according to another embodiment of the present invention; Figure 3 This is a schematic diagram of a computer program product according to an embodiment of the present invention; Figure 4 This is a schematic diagram of a computer-readable storage medium according to an embodiment of the present invention; and Figure 5 This is a schematic diagram of a computer device according to an embodiment of the present invention. Detailed Implementation

[0021] Those skilled in the art should understand that the embodiments described below are merely a part of the embodiments of the present invention, and not all of the embodiments of the present invention. These partial embodiments are intended to explain the technical principles of the present invention and are not intended to limit the scope of protection of the present invention. Based on the embodiments provided by the present invention, all other embodiments obtained by those skilled in the art without creative effort should still fall within the scope of protection of the present invention.

[0022] It should be noted that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus or device (such as a computer-based system, a processor-included system or other system that can fetch and execute instructions from an instruction execution system, apparatus or device).

[0023] In the current semiconductor technology field, the introduction of large field of view (FOV) has led to serious image distortion problems, severely limiting the accuracy and reliability of defect detection. These distortions mainly stem from the following physical and engineering factors: 1. Sample surface unevenness and tilting effect: The actual wafer surface is not an ideal plane, but has microscopic undulations or slight tilting. At the same time, the platform on which the wafer is mounted is not completely perpendicular to the electron beam direction. Under a large FOV, this height difference is magnified, producing linear distortion similar to perspective transformation. In addition, due to the height inconsistency, some areas are out of focus, resulting in image blurring and distortion.

[0024] 2. Scanning coil linearity error: When using magnetic deflection to change the deflection angle of the electron beam, when acquiring images of the lower edge of a large FOV, the electron beam deflection control system has difficulty maintaining a completely linear response, which will cause nonlinear image deformation and produce radial distortion similar to that caused by imperfections in optical lenses, such as pincushion or barrel distortion.

[0025] 3. Non-uniformity error at the edge of the electric deflection field: When the deflection angle of the electron beam is changed by using an electric field, when acquiring an edge image under a large FOV, there is an electric field edge effect at the edge of the electron beam deflection control system, that is, the electric field is non-uniform at the edge of the electric field. This will cause the electron beam to fail to acquire the specified position information when acquiring the edge image, and the image will produce non-linear deformation, similar to radial distortion in optical distortion.

[0026] 4. Electron beam misalignment error: When the optical axis of the electron beam is not accurately aligned with the geometric center of the electromagnetic deflection field in the zero-deflection state, there is an initial position offset between the electron beam and the center of the deflection field. At this time, the scanning trajectory of the electron beam is asymmetrically distributed relative to the deflection field. When acquiring a large field of view image, this eccentricity error will cause the image distortion to exhibit spatial nonlinear distortion characteristics that are highly similar to tangential distortion in the optical system.

[0027] In traditional methods, distortion and jitter in optical images are typically corrected through camera calibration. However, this approach has limited applicability in SEM imaging scenarios. Such methods usually rely on precise intrinsic parameters of the imaging system (e.g., focal length, principal point coordinates, distortion coefficients) and extrinsic parameters (e.g., rotation matrix and translation vector), and use a calibration board (e.g., a checkerboard) to acquire multi-view images to solve for the parameters. However, in scanning electron microscope systems, there is a lack of an equivalent imaging model parameter output interface to an optical camera, and the electron beam scanning mechanism differs fundamentally from the optical imaging principle, making it impossible to directly obtain or define standard camera parameters. Therefore, traditional calibration-based distortion modeling and correction methods are difficult to apply to geometric distortion correction in SEM images.

[0028] At the hardware level, distortion effects can be mitigated by periodically performing system calibration operations. For example, optical path alignment and focusing calibration can be implemented, electron beam alignment and astigmatism correction can be performed to ensure the electron beam is perpendicularly incident on the sample surface, thereby suppressing elliptic distortion caused by lens field asymmetry; the working distance can be optimized and stabilized to avoid geometric deformation caused by focal length deviation; and the tilt, rotation, and displacement accuracy of the sample stage can be calibrated to reduce image distortion introduced by inaccurate mechanical movements. However, the above hardware calibration methods have many limitations, including being time-consuming, requiring the intervention of professional technicians, having long calibration cycles, and affecting the normal use of the equipment during execution.

[0029] To overcome the above-mentioned technical deficiencies, this invention proposes a method for correcting electron microscope scanning images, the specific process of which is as follows: Figure 1 As shown, Figure 1 This is a schematic flowchart of a method for correcting electron microscope scanned images according to an embodiment of the present invention. The method for correcting electron microscope scanned images includes at least the following steps S101 to S106.

[0030] Step S101: Obtain the electron microscope scan image to be corrected and the corresponding design layout. The electron microscope scan image to be corrected is an actual image generated by scanning the surface of a semiconductor wafer with an electron beam. Due to factors such as uneven wafer surface and errors in the electron beam deflection system, image distortion exists and needs to be corrected using this method.

[0031] The design layout refers to the Graphical Data System (GDS) file output during the chip design phase. It is an ideal physical blueprint that records the precise geometric information of the target area of ​​the chip. It includes the circuit layout, transistor positions, pattern size, and the absolute physical coordinates of each element, and serves as the core reference benchmark for distortion correction.

[0032] Step S102: Extract the ideal SEM image of the region corresponding to the SEM image to be calibrated from the design layout. The ideal SEM image is a large field-of-view SEM image extracted from the design layout that corresponds to the same wafer region and the same process layer as the SEM image to be calibrated. Its geometry completely replicates the design layout without any distortion, and it serves as the direct basis for comparing the distortion degree of the actual SEM image (i.e., the SEM image to be calibrated) and constructing the calibration model.

[0033] Step S103: Determine the corner matching pairs between the SEM image to be corrected and the ideal SEM image. Each corner matching pair is a one-to-one correspondence between actual key feature points detected in the SEM image to be corrected and simulated ideal matching feature points in the ideal SEM image. Multiple sets of corner matching pairs can achieve a one-to-one correspondence between the SEM image to be corrected and the ideal SEM image, facilitating subsequent operations such as calculating the perspective transformation matrix and analyzing distortion displacement.

[0034] In some optional embodiments, the step of determining corner matching pairs between the electron microscope scan image to be corrected and the ideal electron microscope scan image generally includes: identifying multiple key feature points in the electron microscope scan image to be corrected using a corner detection algorithm (such as Harris corner detection or FAST corner detection); simulating and calculating multiple matching feature points corresponding to the key feature points based on the geometric structure of the ideal electron microscope scan image; and matching the key feature points and matching feature points to obtain multiple sets of corner matching pairs.

[0035] The steps for matching key feature points and matching feature points generally include: matching key feature points with matching feature points using a nearest neighbor search method to obtain multiple sets of corner point matching pairs. Nearest neighbor search is an efficient matching method based on spatial distance, which can quickly find the ideal corner point with the closest spatial distance for each actual corner point, ensuring the rationality of the matching.

[0036] In addition to the steps of determining the corner matching pairs between the SEM image to be corrected and the ideal SEM image, the process typically includes preprocessing the SEM image to be corrected to reduce noise and improve contrast. This is because during SEM image generation, factors such as electron beam scanning noise and surface charge accumulation can lead to image artifacts and uneven brightness, directly affecting the accuracy of corner detection (e.g., false corner detection or missed detection of real corners). Therefore, preprocessing is necessary to reduce noise and improve contrast in the SEM image to be corrected. Specific processing operations can include applying grayscale normalization, Gaussian filtering, etc., to reduce noise and improve contrast, thereby providing high-quality image input for subsequent corner detection and reducing the probability of false matches.

[0037] Step S104: Estimate the perspective transformation matrix between the SEM image to be corrected and the ideal SEM image based on corner matching pairs. The perspective transformation matrix is ​​generally a mathematical matrix describing the perspective distortion relationship between the SEM image to be corrected and the ideal image. It can be solved by at least 4 sets of corner matching pairs and is used to compensate for large-scale linear distortions caused by wafer tilt, non-perpendicular electron beam incident direction, etc.

[0038] The estimation algorithm for the perspective transformation matrix generally includes the Random Sampling Consensus (RANSAC) algorithm. RANSAC typically estimates an initial matrix by randomly selecting a small number of corner point matching pairs (usually ≥4 pairs), and then uses this matrix to verify all matching pairs: those with deviations within the allowable range are considered inliers (valid matches), and those exceeding the range are considered outliers (incorrect matches). The matrix with the most inliers is ultimately retained as the optimal perspective transformation matrix. This algorithm effectively eliminates interference from incorrect matches, ensuring that the perspective transformation matrix accurately describes the perspective distortion relationship between the SEM image to be corrected and the ideal image, providing reliable data support for subsequent perspective correction. Furthermore, it can select as few as four pairs of points for distortion correction, requiring less information and having a faster computation speed. Those skilled in the art can determine the number of corner point matching pairs according to the actual situation.

[0039] Step S105: Perspective correction is performed on the SEM image to be corrected using a perspective transformation matrix to obtain a perspective-corrected SEM image. The main large-scale distortion in the SEM image to be corrected is perspective distortion (caused by factors such as uneven wafer surface, tilted mounting platform, and non-perpendicular electron beam incident direction). If this type of distortion is not eliminated first, it will mask subsequent radial and tangential distortions, making precise fine-tuning impossible. Therefore, the method of this invention first performs perspective correction, applying the perspective transformation matrix to the entire SEM image to be corrected. Through mathematical transformation, perspective distortion is canceled out, aligning the large-scale geometric structure of the image with the ideal image. The resulting perspective-corrected SEM image generally retains only small-scale nonlinear distortions such as radial and tangential distortions, facilitating subsequent fine-tuning compensation.

[0040] Step S106: Output the fluoroscopic corrected electron microscope scan image.

[0041] In some alternative embodiments, the perspective-corrected electron microscope (EM) image is generally used for preliminary correction. To further improve the correction effect of the SEM image, after obtaining the perspective-corrected EM image, the process may further include: establishing a distortion compensation model based on the perspective-corrected EM image and the ideal EM image; using the distortion compensation model to perform global distortion compensation on the perspective-corrected EM image to obtain a distortion-compensated EM image; and outputting the distortion-compensated EM image. The distortion compensation model is a nonlinear mathematical model that integrates radial and tangential distortion coefficients to compensate for residual nonlinear distortions (such as pincushion / barrel distortion and asymmetric offset distortion) after perspective correction, thereby achieving further refined correction.

[0042] In some optional embodiments, the step of establishing a distortion compensation model based on a perspective-corrected electron microscope (EM) image and an ideal EM image generally includes: extracting feature point matching pairs between the perspective-corrected EM image and the ideal EM image; obtaining residual distortion displacement information corresponding to the feature point matching pairs in the perspective-corrected EM image; and fitting and solving the distortion compensation model based on the feature point matching pairs and the residual distortion displacement information. The residual distortion displacement information refers to the remaining positional deviation between the feature points in the perspective-corrected EM image and the corresponding feature points in the ideal image after perspective correction. It is direct data reflecting the degree of radial and tangential distortion and is used to participate in subsequent fitting of distortion coefficients, thereby achieving refined correction.

[0043] Optionally, the steps for obtaining the distortion compensation model by fitting the feature point matching pairs and residual distortion displacement information generally include: obtaining the radial distortion coefficient and tangential distortion coefficient by fitting the feature point matching pairs and residual distortion displacement information using the least squares method; and establishing the distortion compensation model based on the radial distortion coefficient and tangential distortion coefficient. Since radial distortion (generally caused by linearity error of the scanning coil and non-uniformity at the edge of the electric deflection field, manifesting as pincushion / barrel distortion) and tangential distortion (generally caused by misalignment between the electron beam and the center of the deflection field, manifesting as asymmetric offset) are superimposed nonlinear distortions, their patterns need to be accurately described through mathematical modeling. The least squares method (e.g., the Levenberg-Marquardt algorithm) can fit the optimal distortion coefficients using multiple sets of residual displacement data, making the distortion displacement calculated by the model as consistent as possible with the actual residual displacement.

[0044] Specifically, the steps for global distortion compensation of a fluoroscopically corrected electron microscope (TEM) image using a distortion compensation model generally include: calculating the distortion compensation parameters corresponding to the TEM image using the distortion compensation model; and applying the distortion compensation parameters to the TEM image to achieve global distortion compensation. Since the distortion compensation model clearly defines the laws governing nonlinear distortion, by calculating the compensation parameters to offset the distortion in reverse and applying them to the TEM image, residual radial and tangential distortions can be eliminated, achieving accurate global correction.

[0045] To ensure the correction effect meets the precision requirements of semiconductor manufacturing (such as nanometer-level thresholds), a secondary matching verification is needed to confirm whether the distortion-compensated image meets the standards. If it does not meet the standards, the distortion coefficients can be refitted or the correction parameters adjusted to avoid outputting unqualified images that could affect subsequent defect detection. Therefore, in some optional embodiments, after the step of performing global distortion compensation on the transmission-corrected electron microscope scan image using the distortion compensation model, the following steps are generally included: re-extracting the correction feature point matching pairs between the distortion-compensated electron microscope scan image and the ideal electron microscope scan image; verifying whether the distortion-compensated electron microscope scan image meets preset conditions based on the correction feature point matching pairs; if so, then executing the step of outputting the distortion-compensated electron microscope scan image. The preset conditions can generally be a deviation less than a preset precision threshold (such as 10 nanometers), and those skilled in the art can set specific rules for the preset conditions according to actual needs.

[0046] In some alternative embodiments, the step of outputting distortion-compensated SEM images may further include: feeding back the perspective transformation matrix and distortion compensation model to the electron beam scanning hardware system; and modifying the parameters in the electron beam scanning hardware system by applying the inverse perspective transformation based on the perspective transformation matrix and the displacements calculated by the distortion compensation model for radial and tangential distortions. Traditional correction methods involve post-image correction, which is inefficient and cannot meet the high-throughput requirements of semiconductor mass production. Therefore, this invention chooses to feed back the correction parameters to the hardware system. After receiving the perspective transformation matrix and distortion compensation model, the electron beam scanning hardware system can first apply the inverse perspective transformation matrix to compensate for perspective distortion in advance when acquiring similar SEM images; then, based on the radial and tangential distortion coefficients, it compensates for nonlinear distortion, ultimately directly obtaining low-distortion images without repeating the entire correction process.

[0047] The electron microscopy scanning image correction method of the present invention acquires an SEM image of the region corresponding to the design layout, extracts feature point pairs using feature matching, performs perspective transformation modeling sequentially to correct global distortion, and then fits radial and tangential distortion coefficients based on residual displacement to construct a complete distortion correction model. This model is then applied to an electron beam scanning system to achieve rapid analysis and compensation of distortion under the current imaging state, outputting a geometrically distortion-free image. This method does not rely on hardware calibration parameters and significantly improves the spatial accuracy and automated correction capability of SEM images.

[0048] Figure 2 This is a schematic flowchart of a method for correcting electron microscope scanned images according to another embodiment of the present invention, as shown below. Figure 2 As shown, the correction method for the electron microscope scan image includes at least the following steps S201 to S210.

[0049] Step S201: Obtain the electron microscope scan image to be calibrated and the corresponding design layout.

[0050] Step S202: Extract the ideal electron microscope scan image of the region corresponding to the electron microscope scan image to be corrected from the design layout.

[0051] Step S203 involves preprocessing the SEM image to be calibrated. Because the SEM image to be calibrated suffers from electron beam scanning noise and uneven brightness due to surface charge accumulation, it directly affects the accuracy of corner detection (leading to false corner detection and missed detection of real corners). Preprocessing is a necessary prerequisite for improving the accuracy of subsequent matching and modeling. Specifically, grayscale normalization can unify the image brightness range and eliminate local differences in brightness; Gaussian filtering can smooth the image and filter high-frequency noise, making key features such as line corners and intersections in the SEM image clearer. The preprocessed image provides high-quality input for subsequent corner detection, reducing the probability of false matching.

[0052] Step S204: Determine the corner point matching pairs between the electron microscope scan image to be corrected and the ideal electron microscope scan image. Corner points are geometrically stable coordinate anchor points in the image (such as right angles or intersections of lines), and their positional deviations can directly quantify the degree of distortion.

[0053] Step S204 typically includes: First, corner detection: A corner detection algorithm (e.g., Harris corner detection or FAST corner detection) is executed in the preprocessed SEM image to identify key feature points in the image. Then, corner simulation is performed in the ideal electron microscope scan image, simulating and calculating the corresponding corner positions based on the geometry in the design layout. Finally, the corners detected in the SEM image are matched with the simulated corners in the design layout, and the best matching pair is determined using a nearest neighbor search method.

[0054] By establishing corner point matching pairs, a correspondence can be established between the SEM image to be corrected and the ideal electron microscope scan image, thereby providing data support for subsequent perspective transformation matrix estimation.

[0055] Step S205 involves estimating the perspective transformation matrix between the electron microscope scan image to be corrected and the ideal electron microscope scan image based on corner matching pairs. Typically, the number of corner matching pairs is no less than four. Subsequently, based on the successfully matched corner pairs, the transformation matrix between the two images, including parameters such as translation, rotation, and scaling, is estimated using methods such as the random sampling consensus algorithm. This matrix accurately describes the large-scale linear distortion caused by wafer tilt, non-perpendicular electron beam incidence, etc., providing a reliable mathematical tool for subsequent perspective correction and ensuring that the large-scale geometric structure of the corrected image is aligned with the ideal image.

[0056] Step S206 involves performing perspective correction on the SEM image to be corrected using a perspective transformation matrix, resulting in a perspective-corrected SEM image. The perspective transformation matrix is ​​applied to the entire SEM image to be corrected, and mathematical transformations are used to cancel perspective distortion, ensuring the global geometry of the image matches the ideal image. The resulting perspective-corrected image retains only small-scale nonlinear distortions such as radial distortion (pincushion / barrel) and tangential distortion (asymmetric offset). This perspective correction step is the first correction in the layered decoupled correction strategy of this invention. Step S206 provides a clean input without large-scale interference for subsequent lens distortion modeling.

[0057] Step S207 involves repeating feature point matching and modeling lens distortion using the least squares method. After perspective correction in step S206, the geometric shape of the image has changed, and the original corner point matching pairs are no longer accurate. Therefore, it is necessary to re-extract feature point matching pairs between the perspective-corrected image and the ideal image, and calculate the residual displacement (i.e., the positional deviation caused by nonlinear distortion) of each matching pair to ensure that the input modeling data can accurately reflect the residual distortion law. Subsequently, based on the residual displacement information, the radial distortion coefficient and tangential distortion coefficient are solved by fitting using the least squares method. The coefficients are then substituted into a preset nonlinear model to form a complete lens distortion compensation model. This model can accurately replicate the physical law of nonlinear distortion, providing mathematical support for subsequent fine correction.

[0058] Step S208: Apply lens distortion correction to the perspective-corrected electron microscope scan image and repeat the feature point matching operation to verify the correction result.

[0059] After lens distortion correction, it is necessary to verify whether the correction effect meets the preset accuracy threshold (e.g., nanometer level). Repeated feature point matching is the most direct verification method: if the deviation between the corrected feature points and the ideal points is within the allowable range, the model is effective; if the deviation exceeds the standard, iterative correction can be triggered to avoid outputting unqualified images.

[0060] Step S209 outputs a distortion-compensated electron microscope scan image and feeds back the perspective transformation matrix and the modeling results of radial and tangential distortion. The ultimate goal of the correction is to provide high-quality images for semiconductor defect detection. In addition, to improve the efficiency of subsequent image acquisition, this invention chooses to feed back the correction parameters to the hardware system to achieve pre-correction before image capture, thereby solving the problem of low efficiency in traditional post-processing correction and adapting to the high-throughput requirements of mass production.

[0061] Step S210 involves applying the modeling results sequentially during electron beam acquisition, first for perspective and then for distortion. Perspective distortion is a large-scale linear distortion (caused by wafer and equipment mounting), while nonlinear distortion is a small-scale superimposed distortion (caused by the electron beam scanning system). Applying the compensation in this order avoids interference between the two types of compensation, ensuring distortion is canceled at its source and directly obtaining a low-distortion image.

[0062] This invention provides a design layout-guided SEM image correction and registration method. By extracting geometric features from the design layout and simulating corner positions, combined with preprocessing and corner detection of the actual SEM image, feature matching between the two is achieved. An initial transformation model is estimated using a random sampling consensus algorithm, and then radial and tangential distortions are accurately compensated through iterative perspective transformation correction and least-squares optimization-based lens distortion modeling. Finally, the optimized transformation parameters are fed back to the hardware system to achieve real-time distortion pre-correction during image acquisition. This significantly improves the spatial alignment accuracy and imaging geometric fidelity of the SEM image and the design layout, providing a reliable data foundation for high-precision measurement and automatic defect identification in semiconductor process inspection.

[0063] In summary, the electron microscope scanning image correction method of the present invention achieves the following technical effects through a closed-loop design of hierarchical decoupled correction + iterative optimization + hardware pre-correction: High correction accuracy: It adopts a layered strategy of first perspective correction (to solve large-scale linear distortion) and then nonlinear distortion compensation (to solve small-scale nonlinear distortion), combined with precise tools such as random sampling consensus algorithm and least squares method, to ensure that the spatial alignment accuracy between the corrected image and the design layout reaches the nanometer level, which meets the defect detection requirements of advanced processes of 3nm and below.

[0064] High degree of automation: No human intervention is required throughout the entire process. From image preprocessing, corner detection and matching, to matrix estimation, model building, and hardware parameter feedback, all processes are automated. This avoids the problems of traditional hardware calibration, which relies on professional personnel and is highly subjective, and ensures the standardization and repeatability of the calibration process.

[0065] High detection efficiency: By using a hardware pre-calibration closed loop, post-processing calibration is transformed into feedforward pre-calibration. When acquiring images later, there is no need to repeat the complete calibration process. Low-distortion images are directly output, which significantly improves the throughput of electron beam detection, reduces equipment downtime, and reduces production capacity loss in mass production environments.

[0066] High applicability: It does not rely on the calibration parameters of SEM equipment (solving the problem that traditional camera calibration methods are not applicable in SEM scenarios), and can cover perspective distortion, radial distortion and tangential distortion commonly found in semiconductor manufacturing. It is suitable for large field of view (FOV) inspection scenarios and has good practicality and scalability.

[0067] Low maintenance cost: It avoids the shortcomings of traditional hardware calibration, such as "long cycle, cumbersome operation and frequent repetition", and realizes automatic calibration and parameter update through software algorithm, which significantly reduces the equipment maintenance burden and labor costs.

[0068] In summary, this method effectively balances the accuracy and efficiency of SEM image correction, providing reliable technical support for yield control in advanced semiconductor processes and possessing significant industrial application value.

[0069] It should be understood that in some embodiments, the various parts can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. Any method that uses SEM images and design layouts to construct distortion information points and uses perspective transformation, radial distortion, and tangential distortion for modeling and correcting the electron beam should be considered an alternative to this solution.

[0070] This embodiment also provides a computer program product 10, a computer-readable storage medium 20, and a computer device 30. Figure 3 This is a schematic diagram of a computer program product 10 according to an embodiment of the present invention. Figure 4 This is a schematic diagram of a computer-readable storage medium 20 according to an embodiment of the present invention. Figure 5 This is a schematic diagram of a computer device 30 according to an embodiment of the present invention. The computer program product 10 includes a computer program 11, which, when executed by the processor 32, implements the steps of the electron microscope scanning image correction method described above. A computer-readable storage medium 20 stores the computer program 11 thereon, which, when executed by the processor 32, implements the steps of the electron microscope scanning image correction method described above. The computer device 30 may include a memory 31, a processor 32, and the computer program 11 stored in the memory 31 and running on the processor 32.

[0071] The computer program 11 used to perform the operations of this invention may be assembly instructions, Instruction Set Architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, integrated circuit configuration data, or source code or object code written in any combination of one or more programming languages ​​and procedural programming languages. The computer program 11 may execute entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the latter case, the remote computer may be connected to the user's computer via any type of network, including a Local Area Network (LAN) or Wide Area Network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, to perform aspects of this invention, electronic circuits, including, for example, programmable logic circuits, Field-Programmable Gate Arrays (FPGAs), or Programmable Logic Arrays (PLAs), may execute computer-readable program instructions to personalize the electronic circuits by utilizing state information from computer-readable program instructions.

[0072] For the purposes of this embodiment, computer program product 10 is a related product containing computer program 11. For the purposes of this embodiment, computer-readable storage medium 20 is a tangible device capable of holding and storing computer program 11, and can be any device capable of containing, storing, communicating, propagating, or transmitting program 11 for use by or in conjunction with an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable storage medium 20 include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable optical disc read-only memory (CD-ROM), digital versatile disc (DVD), memory stick, floppy disk, mechanical encoding device, and any suitable combination thereof.

[0073] Computer device 30 can be, for example, a server, desktop computer, laptop computer, tablet computer, or smartphone. In some examples, computer device 30 can be a cloud computing node. Computer device 30 can be described in the general context of computer system executable instructions (such as program modules) executed by a computer system. Typically, program modules can include routines, programs, object programs, components, logic, data structures, etc., that perform specific tasks or implement specific abstract data types. Computer device 30 can be implemented in a distributed cloud computing environment where tasks are performed by remote processing devices linked through a communication network. In a distributed cloud computing environment, program modules can reside on local or remote computing system storage media, including storage devices.

[0074] Computer device 30 may include a processor 32 adapted to execute stored instructions and a memory 31 that provides temporary storage space for the operation of said instructions during operation. The processor 32 may be a single-core processor, a multi-core processor, a computing cluster, or any other configuration. The memory 31 may include random access memory (RAM), read-only memory, flash memory, or any other suitable storage system.

[0075] Computer device 30 may also include a network adapter / interface and an input / output (I / O) interface. The I / O interface allows external devices that can be connected to the computer device to input and output data. The network adapter / interface provides communication between the computer device and a network, typically represented as a communication network.

[0076] Therefore, those skilled in the art should recognize that although numerous exemplary embodiments of the present invention have been shown and described in detail herein, many other variations or modifications conforming to the principles of the present invention can be directly determined or derived from the disclosure of the present invention without departing from the spirit and scope of the invention. Thus, the scope of the present invention should be understood and construed as covering all such other variations or modifications.

Claims

1. A method for correcting electron microscope scanning images, comprising: Acquire the electron microscopy scan image to be corrected and the design layout corresponding to the electron microscopy scan image to be corrected, and extract the ideal electron microscopy scan image of the region corresponding to the electron microscopy scan image to be corrected from the design layout; Determine the corner matching pairs between the electron microscope scan image to be corrected and the ideal electron microscope scan image, and estimate the perspective transformation matrix between the electron microscope scan image to be corrected and the ideal electron microscope scan image based on the corner matching pairs; The electron microscope scan image to be corrected is subjected to perspective correction using the perspective transformation matrix to obtain a perspective-corrected electron microscope scan image. Output the fluoroscopic corrected electron microscope scan image.

2. The method for correcting electron microscope scanned images according to claim 1, wherein, The step of obtaining the fluoroscopic corrected electron microscope scan image further includes: A distortion compensation model is established based on the corrected electron microscope scan image and the ideal electron microscope scan image; The distortion compensation model is used to perform global distortion compensation on the perspective-corrected electron microscope scan image to obtain a distortion-compensated electron microscope scan image; Output the distortion-compensated electron microscope scan image.

3. The method of correcting a scanning image of an electron microscope according to claim 2, wherein The step of establishing a distortion compensation model based on the corrected electron microscope image and the ideal electron microscope image includes: Extract feature point matching pairs between the perspective-corrected electron microscope scan image and the ideal electron microscope scan image; Obtain the residual distortion displacement information corresponding to the feature point matching pair from the fluoroscopic corrected electron microscope scan image; The distortion compensation model is obtained by fitting and solving the feature point matching pairs and the residual distortion displacement information.

4. The method for correcting electron microscope scanned images according to claim 3, wherein, The step of fitting and solving the distortion compensation model based on the feature point matching pairs and the residual distortion displacement information includes: obtaining the radial distortion coefficient and the tangential distortion coefficient by fitting and solving the feature point matching pairs and the residual distortion displacement information using the least squares method; and establishing the distortion compensation model based on the radial distortion coefficient and the tangential distortion coefficient. The step of using the distortion compensation model to perform global distortion compensation on the fluoroscopy-corrected electron microscope scan image includes: calculating the distortion compensation parameters corresponding to the fluoroscopy-corrected electron microscope scan image through the distortion compensation model; and applying the distortion compensation parameters to the fluoroscopy-corrected electron microscope scan image to achieve global distortion compensation.

5. The method of claim 2, wherein, The step of outputting the distortion-compensated electron microscope image further includes: The perspective transformation matrix and the distortion compensation model are fed back to the electron beam scanning hardware system; In the electron beam scanning hardware system, the parameters of the electron beam scanning hardware system are modified by applying the inverse perspective transformation based on the perspective transformation matrix and by applying the displacement calculated by the distortion compensation model in radial and tangential distortion.

6. The method for correcting electron microscope scanned images according to claim 2, wherein, Before the step of determining the corner matching pair between the electron microscopy scan image to be corrected and the ideal electron microscopy scan image, the method further includes: preprocessing the electron microscopy scan image to be corrected to reduce noise and improve contrast in the electron microscopy scan image to be corrected. After the step of performing global distortion compensation on the perspective-corrected electron microscope scan image using the distortion compensation model, the method further includes: re-extracting the correction feature point matching pairs between the distortion-compensated electron microscope scan image and the ideal electron microscope scan image; checking whether the distortion-compensated electron microscope scan image meets preset conditions based on the correction feature point matching pairs; if so, then executing the step of outputting the distortion-compensated electron microscope scan image.

7. The method of claim 1, wherein, The step of determining the corner matching pair between the electron microscopy scan image to be corrected and the ideal electron microscopy scan image includes: Multiple key feature points were identified in the electron microscope scanned image to be corrected using a corner detection algorithm; Based on the geometric structure of the ideal electron microscope scan image, multiple matching feature points corresponding to the key feature points are calculated. The key feature points and the matching feature points are matched to obtain multiple sets of corner point matching pairs.

8. The method for correcting electron microscope scanned images according to claim 7, wherein, The step of matching the key feature points and the matching feature points includes: matching the key feature points and the matching feature points using a nearest neighbor search method to obtain multiple sets of corner point matching pairs; The estimation algorithm for the perspective transformation matrix includes: a random sampling consensus algorithm.

9. A computer program product comprising a computer program that, when executed by a processor, implements the steps of the correction method for electron microscope scanned images according to any one of claims 1 to 8.

10. A computer device comprising a memory, a processor, and a machine-executable program stored in the memory and running on the processor, wherein the processor, when executing the machine-executable program, implements the steps of the correction method for electron microscope scanned images according to any one of claims 1 to 8.

Images