Contour extraction method based on inspection image in multi-charged particle beam inspection

Abstract

An improved apparatus and method for extracting pattern contour information from an inspection image in a multi-charged particle beam inspection system is disclosed. An improved method for extracting pattern contour information from an inspection image includes: identifying a first pattern and a second pattern that partially overlap in the inspection image based on the inspection image obtained from the charged particle beam inspection system. The method also includes: generating a first separated image by removing an image area corresponding to the second pattern from the inspection image. The first separated image includes a first pattern, and when the image area corresponding to the second pattern is removed, a first portion of the first pattern is removed. The method also includes: updating the first separated image to include image data representing the removed first portion of the first pattern based on a first reference image corresponding to the first pattern.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to U.S. application 62 / 976,216, filed February 13, 2020, which is incorporated herein by reference in its entirety. Technical Field

[0003] The embodiments provided herein generally relate to contour extraction techniques, and more particularly to pattern contour extraction from inspection images in multi-charged particle beam inspection. Background Art

[0004] During the manufacturing process of integrated circuits (ICs), unfinished or completed circuit components are inspected to ensure they are manufactured according to design and are free of defects. Inspection systems utilizing optical microscopes or charged particle (e.g., electron) beam microscopes, such as scanning electron microscopes (SEMs), may be employed. As the physical dimensions of IC components continue to shrink, the accuracy and yield of defect detection become even more important.

[0005] The critical dimensions of patterns / structures measured from SEM images can be used to detect defects in manufactured ICs. For example, shifts or edge placement variations between patterns can help control the manufacturing process and identify defects. Summary of the Invention

[0006] The embodiments provided herein disclose a particle beam inspection apparatus, and more particularly, disclose an inspection apparatus using multiple charged particle beams.

[0007] In some embodiments, a method for extracting pattern contour information from an inspection image includes: based on an inspection image obtained from a charged particle beam inspection system, identifying a first pattern and a second pattern that partially overlap in the inspection image. The method also includes: generating a first separated image by removing an image region corresponding to the second pattern from the inspection image. The first separated image includes the first pattern, and when the image region corresponding to the second pattern is removed, a first portion of the first pattern is removed. The method also includes: updating the first separated image to include image data representing the removed first portion of the first pattern based on a first reference image corresponding to the first pattern.

[0008] In some embodiments, a contour extraction device includes a memory storing an instruction set and at least one processor, the at least one processor being configured to execute the instruction set so that the device performs the following steps: based on an inspection image obtained from a charged particle beam inspection system, identifying a first pattern and a second pattern that partially overlap in the inspection image. The at least one processor is also configured to execute the instruction set so that the device further performs the following steps: generating a first separated image by removing an image region corresponding to the second pattern from the inspection image. The first separated image includes a first pattern, and when the image region corresponding to the second pattern is removed, a first portion of the first pattern is removed. The at least one processor is also configured to execute the instruction set so that the device further performs the following steps: updating the first separated image based on a first reference image corresponding to the first pattern to include image data representing the removed first portion of the first pattern.

[0009] In some embodiments, a non-transitory computer-readable medium is provided that stores an instruction set that can be executed by at least one processor of a computing device to cause the computing device to perform a method for extracting pattern contour information from an inspection image. The method includes: based on an inspection image obtained from a charged particle beam inspection system, identifying a first pattern and a second pattern that partially overlap in the inspection image. The method also includes: generating a first separated image by removing an image area corresponding to the second pattern from the inspection image. The first separated image includes a first pattern, and when the image area corresponding to the second pattern is removed, a first portion of the first pattern is removed. The method also includes: updating the first separated image to include image data representing the removed first portion of the first pattern based on a first reference image corresponding to the first pattern.

[0010] In some embodiments, a method for measuring overlay error based on an inspection image includes: based on an inspection image obtained from a charged particle beam inspection system, identifying a first pattern and a second pattern that partially overlap in the inspection image. The first pattern is in a buried layer. The method also includes: generating, based on the inspection image, a first separated image including the first pattern and a second separated image including the second pattern. The method also includes: extracting contour information of the first and second patterns based on the first and second separated images. The method also includes: determining the overlay error between the first and second patterns based on the extracted contour information of the first and second patterns.

[0011] In some embodiments, a non-transitory computer-readable medium is provided, the non-transitory computer-readable medium storing an instruction set that can be executed by at least one processor of a computing device to cause the computing device to perform a method for measuring an overlay error based on an inspection image. The method includes: based on an inspection image obtained from a charged particle beam inspection system, identifying a first pattern and a second pattern that partially overlap in the inspection image, the first pattern being in a buried layer. The method also includes: based on the inspection image, generating a first separated image including the first pattern and a second separated image including the second pattern. The method also includes: based on the first separated image and the second separated image, extracting contour information of the first pattern and the second pattern. The method also includes: based on the extracted contour information of the first pattern and the second pattern, determining an overlay error between the first pattern and the second pattern.

[0012] Other advantages of the embodiments of the present disclosure will become apparent from the following description taken in conjunction with the accompanying drawings which illustrate certain embodiments of the invention by way of illustration and example. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 is a schematic diagram illustrating an example electron beam inspection (EBI) system according to an embodiment of the present disclosure.

[0014] Figure 2 It is a diagram illustrating that the embodiment according to the present disclosure may be Figure 1 Schematic diagram of an example electron beam tool of parts of an electron beam inspection system.

[0015] Figure 3 is an example of an inspection image having a first pattern and a second pattern.

[0016] Figure 4 is a block diagram of an example contour extraction device according to an embodiment of the present disclosure.

[0017] Figure 5A is an example for extracting image patches from an inspection image according to an embodiment of the present disclosure.

[0018] Figure 5B is an example of an image patch for labeling an inspection image according to an embodiment of the present disclosure.

[0019] Figure 6 According to an embodiment of the present disclosure Figure 4 Block diagram of an example neural network model provider for a contour extraction device.

[0020] Figure 7A is an example of an inspection image on which a reference image for each pattern is superimposed according to an embodiment of the present disclosure.

[0021] Figure 7B is an example of a first separated image after removing the second pattern from the inspection image according to an embodiment of the present disclosure.

[0022] Figure 7C is an example of a second separated image after removing the first pattern from the inspection image according to an embodiment of the present disclosure.

[0023] Figure 8A is an example of a first restored image after restoring the first pattern in the first separated image according to an embodiment of the present disclosure.

[0024] Figure 8B is an example of a second restored image after restoring the second pattern in the second separated image according to an embodiment of the present disclosure.

[0025] Figure 9 is a process flow diagram representing an example method for extracting contour information from an electron beam image according to an embodiment of the present disclosure.

[0026] Figure 10 is a process flow diagram representing an example method for training a machine learning model according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0027] Reference will now be made in detail to exemplary embodiments, examples of which are illustrated in the accompanying drawings. The following description refers to the accompanying drawings, in which, unless otherwise indicated, the same numbers in different figures represent the same or similar elements. The implementations set forth in the following description of the exemplary embodiments do not represent all implementations. Instead, they are merely examples of devices and methods consistent with the aspects related to the disclosed embodiments recited in the appended claims. For example, although some embodiments are described in the context of using electron beams, the present disclosure is not limited thereto. Other types of charged particle beams can be similarly applied. In addition, other imaging systems can be used, such as optical imaging, photoelectric detection, x-ray detection, etc.

[0028] Electronic devices are made up of circuits formed on a silicon wafer called a substrate. Many circuits can be formed together on the same silicon wafer and are called integrated circuits, or ICs. The size of these circuits has been significantly reduced, so more circuits can be mounted on a substrate. For example, the IC chip in a smartphone can be as small as a thumbnail but may include over 2 billion transistors, each less than 1 / 1000 the size of a human hair.

[0029] Manufacturing these extremely small ICs is a complex, time-consuming, and expensive process, often involving hundreds of individual steps. An error in even a single step can result in a defect in the finished IC, rendering it unusable. Therefore, one goal of the manufacturing process is to avoid such defects in order to maximize the number of functional ICs produced during the process—that is, to improve the overall yield of the process.

[0030] An integral part of improving yield is monitoring the chip manufacturing process to ensure that it is producing a sufficient number of functional integrated circuits. One way to monitor the process is to inspect the chip circuit structures at various stages of their formation. Inspection can be performed using a scanning electron microscope (SEM). The SEM can be used to image these extremely small structures, in effect taking a "picture" of the structure. This image can be used to determine whether the structure was formed correctly and whether it was formed in the correct location. If the structure is defective, the process can be adjusted to make the defect less likely to occur again.

[0031] When identifying defects, critical dimensions of patterns / structures measured from SEM images can be used. For example, variations in offset or edge placement between patterns, as determined by measured critical dimensions, can be used to identify defects in manufactured chips and control their manufacturing process. Such critical dimensions of patterns can be derived from the pattern's outline information on the SEM image.

[0032] Overlapping patterns / structures are common in today's chip designs (e.g., Figure 3 As shown). However, when an SEM image is obtained from overlapping patterns, the signal from the pattern in the buried layer is usually not strong enough, which makes it challenging to extract reliable contour information of the overlapping patterns from the SEM image. In order to obtain reliable contour information of overlapping patterns from the SEM image, an embodiment of the present disclosure can provide a technology for restoring each pattern on the SEM image based on an image restoration algorithm. In the present disclosure, a restored image for each overlapping pattern can be obtained in real time with the help of a GDS file corresponding to the pattern (such as by using machine learning). In the present disclosure, the contour information of each overlapping pattern can be extracted from the restored image.

[0033] For the sake of clarity, the relative sizes of the components in the drawings may be exaggerated. In the following description of the drawings, the same or similar reference numerals refer to the same or similar components or entities, and only the differences with respect to the individual embodiments are described. As used herein, unless otherwise specifically stated, the term "or" encompasses all possible combinations, except where not feasible. For example, if a component is described that can include A or B, then unless otherwise specifically stated or not feasible, the component can include A or B or A and B. As a second example, if a component is described that can include A, B or C, then unless otherwise specifically stated or not feasible, the component can include A, or B, or C, or A and B, or A and C, or B and C, or A and B and C.

[0034] Now refer to Figure 1 , Figure 1 An example electron beam inspection (EBI) system 100 is illustrated consistent with embodiments of the present disclosure. Figure 1 As shown, the charged particle beam inspection system 100 includes a main chamber 10, a load-lock chamber 20, an electron beam tool 40, and an equipment front end module (EFEM) 30. The electron beam tool 40 is located within the main chamber 10. Although the description and drawings refer to electron beams, it should be understood that the embodiments are not intended to limit the present disclosure to specific charged particles.

[0035] The EFEM 30 includes a first load port 30a and a second load port 30b. The EFEM 30 may include additional load ports. The first load port 30a and the second load port 30b receive wafer front opening unified pods (FOUPs) containing wafers (e.g., one or more semiconductor wafers made of (a variety of) other materials) or samples (wafers and samples are collectively referred to as "wafers") to be inspected. One or more robotic arms (not shown) in the EFEM 30 transfer the wafers to the load-lock chamber 20.

[0036] The load-lock chamber 20 is connected to a load / lock vacuum pump system (not shown) that removes gas molecules in the load-lock chamber 20 to reach a first pressure lower than atmospheric pressure. After reaching the first pressure, one or more robotic arms (not shown) transfer the wafer from the load-lock chamber 20 to the main chamber 10. The main chamber 10 is connected to a main chamber vacuum pump system (not shown) that removes gas molecules in the main chamber 10 to reach a second pressure lower than the first pressure. After reaching the second pressure, the wafer is subjected to inspection by an electron beam tool 40. In some embodiments, the electron beam tool 40 may include a single-beam inspection tool. In other embodiments, the electron beam tool 40 may include a multi-beam inspection tool.

[0037] The controller 50 may be electrically connected to the electron beam tool 40 and may also be electrically connected to other components. The controller 50 may be a computer configured to perform various controls for the charged particle beam inspection system 100. The controller 50 may also include processing circuitry configured to perform various signal and image processing functions. Although the controller 50 is Figure 1 3. The controller 50 is shown external to the structure including the main chamber 10, the load-lock chamber 20, and the EFEM 30, but it will be appreciated that the controller 50 may be part of the structure.

[0038] Although the present disclosure provides an example of a main chamber 10 housing an electron beam inspection system, it should be noted that aspects of the present disclosure in its broadest sense are not limited to chambers housing electron beam inspection systems. Instead, it should be understood that the principles described above may also be applied to other chambers.

[0039] Now refer to Figure 2 , Figure 2 A schematic diagram consistent with an embodiment of the present disclosure is shown, which illustrates a Figure 1 The example electron beam tool 40 is a portion of the example charged particle beam inspection system 100. The electron beam tool 40 (also referred to herein as the apparatus 40) includes an electron source 101, a gun aperture plate 171 having a gun aperture 103, a pre-beamlet forming mechanism 172, a condenser lens 110, a source conversion unit 120, a primary projection optical system 130, a sample stage (not shown), and a plurality of other components. Figure 2 ), a secondary optical system 150, and an electronic detection device 140. The primary projection optical system 130 may include an objective lens 131. The electronic detection device 140 may include a plurality of detection elements 140_1, 140_2, and 140_3. The beam splitter 160 and the deflection scanning unit 132 may be placed within the primary projection optical system 130. It will be appreciated that other known components of the device 40 may be appropriately added or omitted.

[0040] The electron source 101, the gun aperture plate 171, the condenser lens 110, the source conversion unit 120, the beam splitter 160, the deflection scanning unit 132, and the primary projection optical system 130 may be aligned with the primary optical axis 100_1 of the apparatus 100. The secondary optical system 150 and the electron detection apparatus 140 may be aligned with the secondary optical axis 150_1 of the apparatus 40.

[0041] The electron source 101 may include a cathode, an extractor, or an anode, wherein primary electrons may be emitted from the cathode and extracted or accelerated to form a primary electron beam 102 forming a cross (virtual or real) 101s. The primary electron beam 102 may be visualized as being emitted from the cross 101s.

[0042] The source conversion unit 120 may include an image forming element array (not shown). Figure 2 ), an aberration compensator array (not shown), a beam limiting aperture array (not shown), and a pre-bent micropolarizer array (not shown). The image forming element array may include a plurality of micropolarizers or microlenses to form a plurality of parallel images (virtual or real) of the cross 101s with the plurality of beamlets of the primary electron beam 102. Figure 2 Three sub-bundles 102_1 , 102_2 , and 102_3 are shown as an example, and it should be understood that the source conversion unit 120 can process any number of sub-bundles.

[0043] In some embodiments, the source conversion unit 120 may be provided with a beam-limiting aperture array and an image forming element array (neither of which is shown). The beam-limiting aperture array may include a beam-limiting aperture. It should be understood that any number of apertures may be used appropriately. The beam-limiting aperture may be configured to limit the size of the sub-beams 102_1, 102_2, and 102_3 of the primary electron beam 102. The image forming element array 122 may include an image forming deflector (not shown) configured to deflect the sub-beams 102_1, 102_2, and 102_3 by changing the angle toward the primary optical axis 100_1. In some embodiments, a deflector further away from the primary optical axis 100_1 may deflect the sub-beams to a greater extent. In addition, the image forming element array 122 may include multiple layers (not shown), and the deflectors may be arranged in separate layers. The deflectors may be configured to be individually controlled independently of each other. In some embodiments, the deflector can be controlled to adjust the spacing of the detection spots (e.g., 102_1S, 102_2S, and 102_3S) formed on the surface of the sample 1. As mentioned herein, the spacing of the detection spots can be defined as the distance between two adjacent detection spots on the surface of the sample 1.

[0044] The centrally located deflector of the image forming element array can be aligned with the primary optical axis 100_1 of the electron beam tool 40. Therefore, in some embodiments, the central deflector can be configured to maintain the trajectory of the sub-beam 102_1 as straight. In some embodiments, the central deflector can be omitted. However, in some embodiments, the primary electron source 101 may not necessarily be aligned with the center of the source conversion unit 120. Furthermore, it should be understood that although Figure 2 A side view of the device 40 is shown, wherein the beamlet 102_1 is on the primary optical axis 100_1. When viewed from different sides, the beamlet 102_1 may deviate from the primary optical axis 100_1. That is, in some embodiments, all beamlets 102_1, 102_2, and 102_3 may be off-axis. The off-axis component may be offset relative to the primary optical axis 100_1.

[0045] The deflection angle of the deflected beamlets can be set based on one or more criteria. In some embodiments, the deflector can deflect the off-axis beamlets radially outward or away from the primary optical axis 100_1 (not shown). In some embodiments, the deflector can be configured to deflect the off-axis beamlets radially inward or toward the primary optical axis 100_1. The deflection angle of the beamlets can be set so that the beamlets 102_1, 102_2, and 102_3 fall perpendicularly on the sample 1. The off-axis aberration of the image caused by the lens (such as the objective lens 131) can be reduced by adjusting the path of the beamlets through the lens. Therefore, the deflection angle of the off-axis beamlets 102_2 and 102_3 can be set so that the detection spots 102_2S and 102_3S have small aberrations. The beamlets can be deflected so as to pass through or approach the front focus of the objective lens 131 to reduce the aberrations of the off-axis detection spots 102_2S and 102_3S. In some embodiments, the deflector may be arranged such that the beamlets 102_1 , 102_2 and 102_3 fall perpendicularly on the sample 1 while the detection spots 102_1S, 102_2S and 102_3S have small aberrations.

[0046] The condenser lens 110 is configured to focus the primary electron beam 102. The current of the beamlets 102_1, 102_2, and 102_3 downstream of the source conversion unit 102 can be varied by adjusting the focusing power of the condenser lens 110 or by changing the radial size of the corresponding beam-limiting apertures within the beam-limiting aperture array. The current can be varied by changing the radial size of the beam-limiting apertures and the focusing power of the condenser lens 110. The condenser lens 110 can be an adjustable condenser lens that can be configured to move the position of its first principal plane. The adjustable condenser lens can be configured to be magnetic, which can cause the off-axis beamlets 102_2 and 102_3 to illuminate the source conversion unit 120 at a rotated angle. The rotation angle can vary depending on the focusing power or position of the first principal plane of the adjustable condenser lens. Therefore, the condenser lens 110 can be an anti-rotation condenser lens that can be configured to maintain a constant rotation angle while the focusing power of the condenser lens 110 is changed. In some embodiments, the condenser lens 110 may be an adjustable anti-rotation condenser lens, wherein the rotation angle does not change when the focusing power and the position of the first principal plane of the condenser lens 110 are changed.

[0047] The electron beam tool 40 may include a pre-beamlet forming mechanism 172. In some embodiments, the electron source 101 may be configured to emit primary electrons and form a primary electron beam 102. In some embodiments, the gun aperture plate 171 may be configured to block peripheral electrons in the primary electron beam 102 to reduce the Coulomb effect. In some embodiments, the pre-beamlet forming mechanism 172 further cuts off peripheral electrons of the primary electron beam 102 to further reduce the Coulomb effect. The primary electron beam 102 may be trimmed into three primary electron beamlets 102_1, 102_2, and 102_3 (or any other number of beamlets) after passing through the pre-beamlet forming mechanism 172. The electron source 101, the gun aperture plate 171, the pre-beamlet forming mechanism 172, and the condenser lens 110 may be aligned with the primary optical axis 100_1 of the electron beam tool 40.

[0048] The pre-beamlet forming mechanism 172 may include a Coulomb aperture array. The central aperture (also referred to herein as the on-axis aperture) of the pre-beamlet forming mechanism 172 and the central deflector of the source conversion unit 120 may be aligned with the primary optical axis 100_1 of the electron beam tool 40. The pre-beamlet forming mechanism 172 may be provided with a plurality of pre-trimmed apertures (e.g., a Coulomb aperture array). Figure 2 , three sub-beams 102_1, 102_2, and 102_3 are generated when the primary electron beam 102 passes through the three pre-trimmed apertures, and many of the remaining parts of the primary electron beam 102 are cut off. That is, the pre-beamlet forming mechanism 172 can trim many or most of the electrons from the primary electron beam 102 that do not form the three sub-beams 102_1, 102_2, and 102_3. The pre-beamlet forming mechanism 172 can cut off electrons that will not eventually be used to form the detection spots 102_1S, 102_2S, and 102_3S before the primary electron beam 102 enters the source conversion unit 120. In some embodiments, the gun aperture plate 171 can be disposed near the electron source 101 to cut off electrons at an early stage, while the pre-beamlet forming mechanism 172 can also be configured to further cut off electrons around the plurality of beamlets. Although Figure 2 Three apertures of the pre-beamlet forming mechanism 172 are shown, but it will be appreciated that any number of apertures may be present as appropriate.

[0049] In some embodiments, the pre-beamlet former 172 can be placed below the condenser lens 110. Placing the pre-beamlet former 172 closer to the electron source 101 can more effectively reduce the Coulomb effect. In some embodiments, the gun aperture plate 171 can be omitted when the pre-beamlet former 172 can be placed close enough to the source 101 while still being manufacturable.

[0050] The objective lens 131 can be configured to focus the sub-beams 102_1, 102_2, and 102_3 onto the sample 1 for inspection, and can form three detection spots 102_1s, 102_2s, and 102_3s on the surface of the sample 1. The gun aperture plate 171 can block unused peripheral electrons in the primary electron beam 102 to reduce Coulomb interaction effects. The Coulomb interaction effects can increase the size of each of the detection spots 102_1s, 102_2s, and 102_3s, thereby degrading inspection resolution.

[0051] The beam splitter 160 may be a Wien filter type beam splitter, comprising a generator that generates an electrostatic dipole field E1 and a magnetic dipole field B1 (both of which are not in the Figure 2 1 ). If applied, the force exerted by the electrostatic dipole field E1 on the electrons of beamlets 102_1, 102_2, and 102_3 is equal in magnitude and opposite in direction to the force exerted by the magnetic dipole field B1 on the electrons. Therefore, beamlets 102_1, 102_2, and 102_3 can pass directly through beam splitter 160 with zero deflection angle.

[0052] The deflection scanning unit 132 may deflect the sub-beams 102_1, 102_2, and 102_3 to scan the detection spots 102_1s, 102_2s, and 102_3s over three small scanning areas in a section of the surface of the sample 1. In response to the incidence of the sub-beams 102_1, 102_2, and 102_3 at the detection spots 102_1s, 102_2s, and 102_3s, three secondary electron beams 102_1se, 102_2se, and 102_3se may be emitted from the sample 1. Each of the secondary electron beams 102_1se, 102_2se, and 102_3se may include electrons having an energy distribution including secondary electrons (energy ≤ 50 eV) and backscattered electrons (energy between 50 eV and the landing energy of the sub-beams 102_1, 102_2, and 102_3). The beam splitter 160 can direct the secondary electron beams 102_1se, 102_2se, and 102_3se to the secondary imaging system 150. The secondary imaging system 150 can focus the secondary electron beams 102_1se, 102_2se, and 102_3se onto the detection elements 140_1, 140_2, and 140_3 of the electron detection device 140. The detection elements 140_1, 140_2, and 140_3 can detect the corresponding secondary electron beams 102_1se, 102_2se, and 102_3se and generate corresponding signals for constructing an image of the corresponding scan area of the sample 1.

[0053] exist Figure 2, three secondary electron beams 102_1se, 102_2se, and 102_3se generated by the three detection spots 102_1S, 102_2S, and 102_3S, respectively, travel upward along the primary optical axis 100_1 toward the electron source 101, successively passing through the objective lens 131 and the deflection scanning unit 132. The three secondary electron beams 102_1se, 102_2se, and 102_3se are steered by a beam splitter 160 (such as a Wien filter) to enter the secondary imaging system 150 along the secondary optical axis 150_1 of the secondary imaging system 150. The secondary imaging system 150 focuses the three secondary electron beams 102_1se to 102_3se onto an electron detection device 140 including three detection elements 140_1, 140_2, and 140_3. Therefore, the electronic detection device 140 can simultaneously generate images of three scanning areas scanned by the three detection spots 102_1S, 102_2S, and 102_3S, respectively. In some embodiments, the electronic detection device 140 and the secondary imaging system 150 form a detection unit (not shown). In some embodiments, the electron optical device components in the secondary electron beam path (such as, but not limited to, the objective lens 131, the deflection scanning unit 132, the beam splitter 160, the secondary imaging system 150, and the electronic detection device 140) can form a detection system.

[0054] In some embodiments, the controller 50 may include an image processing system comprising an image capture device (not shown) and a storage device (not shown). The image capture device may include one or more processors. For example, the image capture device may include a computer, a server, a mainframe, a terminal, a personal computer, any type of mobile computing device, or a combination thereof. The image capture device may be communicatively coupled to the electronic detection device 140 of the apparatus 40 via a medium such as an electrical conductor, a fiber optic cable, a portable storage medium, infrared (IR), Bluetooth, the internet, a wireless network, radio, or a combination thereof. In some embodiments, the image capture device may receive signals from the electronic detection device 140 and may construct an image. The image capture device may thereby capture an image of the sample 1. The image capture device may also perform various post-processing functions, such as generating outlines, overlaying indicators on the captured image, and the like. The image capture device may be configured to adjust the brightness and contrast of the captured image, for example. In some embodiments, the storage device may be a storage medium such as a hard drive, a flash drive, cloud storage, random access memory (RAM), or other types of computer-readable memory. A storage device may be coupled to the image acquirer and may be used to save the scanned raw image data as an initial image and to save a post-processed image.

[0055] In some embodiments, the image acquirer can acquire one or more images of the sample based on one or more imaging signals received from the electronic detection device 140. The imaging signal can correspond to a scanning operation for performing charged particle imaging. The acquired image can be a single image including multiple imaging areas or can involve multiple images. A single image can be stored in a storage device. A single image can be an initial image that can be divided into multiple regions. Each of these regions can include an imaging area containing features of sample 1. The acquired image can include multiple images of a single imaging area of sample 1 sampled multiple times in a time series or can include multiple images of different imaging areas of sample 1. Multiple images can be stored in a storage device. In some embodiments, the controller 50 can be configured to perform image processing steps on multiple images of the same position of sample 1.

[0056] In some embodiments, the controller 50 may include measurement circuitry (e.g., an analog-to-digital converter) to obtain a distribution of the detected secondary electrons. The electron distribution data collected during the detection time window, combined with the corresponding scan path data for each of the primary beamlets 102_1, 102_2, and 102_3 incident on the wafer surface, can be used to reconstruct an image of the inspected wafer structure. The reconstructed image can be used to reveal various features of the internal or external structure of the sample 1 and, therefore, any defects that may be present in the wafer.

[0057] In some embodiments, the controller 50 can control a motorized stage (not shown) to move the sample 1 during inspection. In some embodiments, the controller 50 can enable the motorized stage to continuously move the sample 1 in a certain direction at a constant speed. In other embodiments, the controller 50 can enable the motorized stage to change the movement speed of the sample 1 over time depending on the steps of the scanning process. In some embodiments, the controller 50 can adjust the configuration of the primary projection optical system 130 or the secondary imaging system 150 based on the images of the secondary electron beams 102_1se, 102_2se, and 102_3se.

[0058] although Figure 2 The electron beam tool 40 is shown using three primary electron beams, but it should be understood that the electron beam tool 40 can use two or more primary electron beams. The present disclosure does not limit the number of primary electron beams used in the apparatus 40.

[0059] Now refer to Figure 3 , Figure 3 is an example of an inspection image 300 having a first pattern A and a second pattern B, the first pattern A having a rectangular shape and the second pattern B having a circular shape. According to an embodiment of the present disclosure, the inspection image 300 may be obtained by a charged particle beam inspection system (e.g., Figure 1 For example, the inspection image 300 may be an electron beam image generated based on the electron detection signal from the electron detection element 140. Figure 3 , a first pattern A and a second pattern B overlap. For example, the first pattern A in the inspection image 300 may be generated based on an electron detection signal from a first layer of the sample structure, and the second pattern B may be generated based on an electron signal from a second layer of the structure located on top of the first layer. In this example, the first pattern A is in a buried layer. Therefore, the electron beam signal for the first pattern A may be weaker than the electron beam signal for the second pattern B. It is also possible that the portion of the first pattern A covered by the second pattern B is not visible or discernible in the inspection image 300. Even for the second pattern B, the overlapping area with the first pattern A may be blurred due to various reasons, such as a different signal-to-noise ratio compared to the non-overlapping area. Therefore, when multiple patterns are overlapped in the inspection image, it is challenging to extract contour information of the patterns from the inspection image.

[0060] Figure 4 is a block diagram of an example contour extraction device consistent with embodiments of the present disclosure. It should be understood that in various embodiments, the contour extraction device 400 may be a charged particle beam inspection system (e.g., Figure 1 The contour extraction device 400 may be part of the electron beam inspection system 100 or may be separate from the charged particle beam inspection system. In some embodiments, the contour extraction device 400 may be part of the controller 50 and may include an image acquisition device, measurement circuitry, or storage device. In some embodiments, the contour extraction device 400 may include an image processing system and may include an image acquisition device, storage device, etc.

[0061] In some embodiments, as Figure 4 As shown, the contour extraction apparatus 400 may include an inspection image acquirer 410 , a pattern identifier 420 , a separation image generator 430 , a pattern compensator 440 , and a pattern measurer 450 .

[0062] According to an embodiment of the present disclosure, the inspection image acquirer 410 can acquire an inspection image of a sample to be inspected. For the purpose of illustration and simplicity, Figure 3The inspection image 300 shown will be used as an example of an inspection image acquired by the inspection image acquirer 410. In some embodiments, the inspection image acquirer 410 can generate the inspection image 300 based on the detection signal from the electron detection device 140 of the electron beam tool 40. In some embodiments, the inspection image acquirer 410 can be part of the image acquirer included in the controller 50 or can be separate therefrom. In some embodiments, the inspection image acquirer 410 can acquire the inspection image 300 generated by the image acquirer included in the controller 50. In some embodiments, the inspection image acquirer 410 can acquire the inspection image 300 from a storage device or system that stores the inspection image 300.

[0063] The pattern identifier 420 is configured to identify a pattern on the inspection image 300 consistent with embodiments of the present disclosure. According to embodiments of the present disclosure, the pattern identifier 420 can extract image patches as portions of the image from the inspection image 300 and can identify patterns corresponding to the extracted image patches as features on the die or in a layout database (e.g., a GDS graphic data system). In some embodiments, each image patch can include at least a portion of a pattern that does not overlap with other patterns. Figure 5A Four patches P1 to P4 extracted from the inspection image 300 are illustrated as an example. Figure 5A As shown, each of the first image patch P1 and the second image patch P2 covers the portion of the first pattern A that does not overlap with the second pattern B. Similarly, each of the third image patch P3 and the fourth image patch P4 covers the portion of the second pattern B that does not overlap with the first pattern A. According to some embodiments of the present disclosure, an image patch as large as possible may be selected to increase the pattern recognition ratio. For example, in Figure 5A , for the area covered by the first image patch P1, only one image patch is extracted instead of two or more image patches for the same area. In some embodiments of the present disclosure, the image patches may be selected so that each image patch includes (multiple) boundary lines of the corresponding pattern, since the boundary lines can be used to identify the pattern. In some embodiments, the boundary lines of the pattern may be lines for determining the outer shape of the pattern, lines for determining the inner shape of the pattern, boundary lines between different textures in the pattern, or other types of lines that can be used to identify the pattern. For example, Figure 5A The diagram shows that the first image patch P1 and the second image patch P2 include partial lines for determining the outer shape of the first pattern A (e.g., a rectangle), and the third image patch P3 and the fourth image patch P4 include partial lines for determining the outer shape of the second pattern B (e.g., a circle).

[0064] Return Reference Figure 4According to some embodiments of the present disclosure, the pattern identifier 420 may be configured to associate each image patch with a corresponding pattern based on a machine learning model. For example, by using the extracted image patches P1 to P4 as input to the machine learning model, the pattern identifier 420 may determine that each of the first image patch P1 and the second image patch P2 is part of a first pattern A and each of the third image patch P3 and the fourth image patch P4 is part of a second pattern B. In some embodiments, the pattern identifier 420 may automatically label each image patch with a corresponding pattern identification. For example, Figure 5B As shown, the first image patch P1 and the second image patch P2 are labeled A1 and A2 to represent the first pattern A, and the third image patch P3 and the fourth image patch P4 are labeled B1 and B2 to represent the second pattern B.

[0065] According to an embodiment of the present disclosure, the pattern identifier 420 may obtain a machine learning model from the machine learning model provider 460. In some embodiments, the machine learning model provider 460 may train the machine learning model to identify patterns of image patches. According to an embodiment of the present disclosure, the machine learning model provider 460 may pre-train the machine learning model and provide the trained machine learning model to the pattern identifier 420 on demand. In some embodiments, the machine learning model obtained from the machine learning model provider 460 may be stored in a storage medium (not shown) and may be accessed by the pattern identifier 420. In some embodiments, the machine learning model provider 460 may update the machine learning model when a new inspection image is acquired.

[0066] Now refer to Figure 6 , Figure 6 FIGURE 1 illustrates a block diagram of an example neural network model provider consistent with an embodiment of the present disclosure. Figure 6 As shown, the machine learning model provider 460 may include an image patch extractor 461, a machine learning model trainer 462 and an information file 463.

[0067] The image patch extractor 461 may receive a training inspection image IM obtained from an inspection system such as, but not limited to, the tool 40 or the system 100. In some embodiments, the training inspection image IM may include Figure 3 When the training inspection image IM includes multiple patterns, the image patch extractor 461 can extract image patches such as Figure 5A For the purpose of illustration and for simplicity, the Figure 3 The inspection image 300 is used as the training inspection image IM and by using Figure 5AImage patches P1 to P4 are used as training image patches to illustrate the process of training a machine learning model.

[0068] According to an embodiment of the present disclosure, the machine learning model trainer 462 may be configured to train a machine learning model to be provided to the pattern identifier 420 by using image patches P1 to P4 extracted from the training inspection image IM as input to the machine learning model. According to an embodiment of the present disclosure, the machine learning model trainer 462 may train the machine learning model to predict a pattern associated with each image patch by referring to a reference image contained in the information file 463.

[0069] According to an embodiment of the present disclosure, information file 463 may include reference images corresponding to patterns included in training inspection image IM. For example, information file 463 may include a first reference image corresponding to first pattern A and a second reference image corresponding to second pattern B. In some embodiments, the reference images included in information file 463 may be ground-truth images of the corresponding patterns. The ground-truth images may include original images of a wafer or die containing the corresponding patterns, or may include ground-truth wafer maps measured from a wafer or die containing the corresponding patterns, etc. In some embodiments, the reference images included in information file 463 may be in a Graphic Database System (GDS) format, a Graphic Database System II (GDSII) format, an Open Artwork System Interchange Standard (OASIS) format, a Caltech Intermediate Format (CIF), etc. In some embodiments, the reference images included in information file 463 may include a wafer design layout of the corresponding pattern. The wafer design layout may be based on a pattern layout used to construct the wafer. The wafer design layout may correspond to one or more photolithography masks or reticles used to transfer features from the photolithography masks or reticles to the wafer. In some embodiments, a reference image of a GDS or OASIS, etc., may include feature information stored in a binary file format representing planar geometry, text, and other information related to the wafer design layout.

[0070] In some embodiments, the machine learning model trainer 462 can be configured to train the machine learning model through supervised learning. In supervised learning, the training data fed to the machine learning model includes the desired solution. For example, under the condition that the machine learning model trainer 462 knows the patterns corresponding to the image patches P1 to P4, the machine learning model can be trained using input training data such as image patches P1 to P4. During training, the weights of the machine learning model can be updated or modified so that the machine learning model can provide inference results corresponding to the known solutions. For training purposes, when extracting image patches from the training inspection image IM, the image patch extractor 461 can also refer to a reference image contained in the information file 463. For example, the image patch extractor 461 can extract image patches corresponding to a certain pattern by referring to a reference image of the pattern.

[0071] While the training process has been described for a single training inspection image IM comprising two patterns, it should be understood that embodiments of the present disclosure can be applied to scenarios involving two or more training inspection images, each of which may include one or more patterns. In these scenarios, a reference image for each pattern included in the two or more training inspection images can be included in information file 463. It should be noted that when the number of reference images and patterns is large, the process of searching for a corresponding pattern for a particular image patch is time-consuming and resource-intensive. Therefore, pre-training a machine learning model allows for real-time identification of inspection image patterns.

[0072] Return Reference Figure 4 According to an embodiment of the present disclosure, the separated image generator 430 is configured to separate the patterns of the inspection image 300 and generate a separated image for each pattern of the inspection image 300. According to some embodiments, the separated image generator 430 may generate the separated image based on the pattern identification from the pattern identifier 420 and based on a reference image corresponding to the pattern. The reference image used to separate the pattern of the inspection image may be a reference image described with respect to the information file 463. In some embodiments, the separated image generator 430 may access the information file 463 included in the machine learning model provider 460, or have a separation database (not shown) containing reference images.

[0073] Will refer to Figure 7A 、 Figure 7B and Figure 7C A process for separating a pattern on the inspection image 300 is described. Figure 7A is an example of an inspection image on which a reference image for each pattern is superimposed, consistent with an embodiment of the present disclosure. Figure 5BThe pattern marker 420 shown can be used to overlay reference images on the corresponding patterns A and B. When the reference images are overlaid on the inspection image 300, the reference images can be aligned to match the corresponding patterns as closely as possible. In some embodiments, the reference images can be aligned to match the corresponding patterns in terms of angular position, outer boundaries, etc. For example, the first reference image A R is superimposed on the first pattern A, such as Figure 7A Similarly, the second reference image B R Superimposed on Figure 7A On the second pattern B.

[0074] According to an embodiment of the present disclosure, when a reference image is superimposed on the inspection image 300, the separation image generator 430 can match the reference image with the image patch determined to correspond to the reference image. For example, by comparing the first image patch P1 and the second image patch P2 determined to correspond to the first pattern A with the first reference image A in terms of position, shape, texture, etc. R For comparison, the first reference image A R Superimposed on the inspection image 300. Similarly, by comparing the third image patch P3 and the fourth image patch P4 determined to correspond to the second pattern B in terms of position, shape, texture, etc. R For comparison, the second reference image B R Superimposed on the inspection image 300.

[0075] According to an embodiment of the present disclosure, the separation image generator 430 may generate a first separation image by removing an image region covered by the second pattern B from the inspection image 300 . Figure 7B An example of a first separated image 310 is shown. Figure 7B In the image, the removed area is represented as I B In some embodiments, the removal of the image area covered by the second pattern B can be performed based on the image being superimposed on the image. Figure 7A The second reference image B on the inspection image 300 is shown R In some embodiments, when the image area covered by the second pattern B is removed, the removed image area I B Can be larger than the second reference image B R , to ensure that the second pattern B is removed from the first separated image 310. For example, the removed image area I B The outer boundary of the second reference image B R The outside. Figure 7B As shown, it should be noted that the portion of the first pattern A that overlaps with the second pattern B is also removed from the inspection image 300 .

[0076] Similarly, the separated image generator 430 may generate a second separated image by removing the image region covered by the first pattern A from the inspection image 300 . Figure 7C An example of a second separated image 320 is shown. Figure 7C In the image, the removed area is represented by 1 A According to some embodiments of the present disclosure, the image area covered by the first pattern A can be removed according to the following method: Figure 7A The first reference image A shown is superimposed on the inspection image 300. R In some embodiments, when the image area covered by the first pattern A is removed, the removed image area I A Can be larger than the first reference image A R , to ensure that the first pattern A is removed from the second separated image 320. For example, the removed image area I A The outer boundary of the first reference image A R The outside. Figure 7C As shown, it should be noted that the portion of the second pattern B that overlaps with the first pattern A is also removed from the inspection image 300 .

[0077] Although the process of pattern separation is described by using an inspection image including two patterns, and thus two separated images are generated, it should be noted that the present disclosure is also applicable when the inspection image includes three or more patterns. For example, when the inspection image includes three patterns (e.g., a first pattern to a third pattern), three separated images can be generated. The first separated image can be generated by removing the image area covered by the second pattern and the third pattern, the second separated image can be generated by removing the image area covered by the first pattern and the third pattern, and the third separated image can be generated by removing the image area covered by the first pattern and the second pattern.

[0078] As mentioned above, when Figure 7B When the second pattern B is removed from the inspection image 300 as shown, a portion of the first pattern A is also removed in the first separated image 310, and when Figure 7C As shown, when the first pattern A is removed from the inspection image 300, a portion of the second pattern B is also removed in the second separated image 320. Therefore, the separated image may include an incomplete pattern. Figure 4 According to embodiments of the present disclosure, the pattern compensator 440 may be configured to update or restore an incomplete pattern in the separated image. According to some embodiments of the present disclosure, the pattern compensator 440 may update or restore the pattern based on an image restoration algorithm. The image restoration algorithm may be used to reconstruct lost / missing or degraded portions of an image. In some embodiments, when restoring the incomplete pattern, the pattern compensator 440 may utilize structural information, textual information, or contextual information from the separated image and the reference image.

[0079] Figure 8A is an example of a first restored image after restoring the first pattern A in the first separated image according to an embodiment of the present disclosure. Figure 8A As shown, the first restored image 311 can be generated by restoring or filling the removed portion of the first pattern A in the first separated image 310. The removed portion of the first pattern A can be a portion of the first pattern A that is removed when the first separated image 310 (e.g., Figure 7B According to some embodiments, the pattern compensator 440 may remove the portion of the first pattern A based on the first reference image A. R ( Figure 7A ) in the first separated image 310 to restore or fill the removed portion of the first pattern A. For example, the pattern compensator 440 may restore or fill the removed portion of the first pattern A in the first separated image 310 by comparing the unremoved portion of the first pattern A with the first reference image A. R Compare to determine the first reference image A R Which part of the first pattern A in the first separated image 310 corresponds to the removed part of the first pattern A in the first separated image 310. For example, the pattern compensator 440 can compare the non-removed part of the first pattern A in the first separated image 310 with the first reference image A in terms of position, shape, texture, etc. R Based on the first reference image A R The first restored portion Ac may be filled in the removed portion of the first pattern A, and thus the first restored image 311 including the restored first pattern A may be as follows: Figure 8A Generated as shown.

[0080] Similarly, Figure 8B is an example of a second restored image after restoring the second pattern in the second separated image according to an embodiment of the present disclosure. Figure 8B As shown, the second restored image 321 can be generated by restoring or filling the removed portion of the second pattern B in the second separated image 320. The removed portion of the second pattern B can be a portion of the second pattern B that is removed when the first pattern A is removed in generating the second separated image 320 (e.g., Figure 7C Similarly, the pattern compensator 440 can remove the portion of the second pattern B by comparing the non-removed portion of the second pattern B with the second reference image B. R Compare and determine the second reference image B R Which part of the second reference image B corresponds to the removed part of the second pattern B in the second separated image 320. RThe determination of the portion of the second pattern B corresponding to the removed portion in the second separated image 320, the second restored portion B C can be used to fill the removed portion of the second pattern B, so it can be Figure 8B As shown, a second restored image 321 including a restored second pattern B is generated.

[0081] like Figure 8A and Figure 8B As shown, a first restored image 311 including the entire outline of the first pattern A and a second restored image 321 including the entire outline of the second pattern B are generated by the pattern compensator 440. Return to Reference Figure 4 The pattern measurer 450 may be configured to obtain measurement results of the first pattern A and the second pattern B from the first restored image 311 and the second restored image 321, respectively. For example, the pattern measurer 450 may extract contour information of the first pattern A from the first restored image 311, and may extract contour information of the second pattern B from the second restored image 321.

[0082] According to some embodiments of the present disclosure, pattern profile information can be used to obtain the critical dimensions of the pattern to determine pattern offset, edge placement changes, and the like. In some embodiments, overlay error can be measured based on the pattern profile information. For example, in the above example, the centroid of first pattern A and the centroid of second pattern B can be determined based on the extracted profile information of first pattern A and second pattern B. By comparing the distance between the two centroids of first pattern A and second pattern B with an expected distance, the overlay error between the two patterns in the inspection image can be determined. In some embodiments, the expected distance between the two patterns can be obtained, for example, based on the wafer design layout.

[0083] Figure 9 is a process flow diagram illustrating an example method for extracting contour information from an inspection image according to an embodiment of the present disclosure. Figure 4 A method for extracting contour information from an inspection image is described with reference to the contour extraction apparatus 400 .

[0084] In step S910, an inspection image of the sample may be obtained. Step S910 may be performed by, for example, the inspection image acquirer 410. For the purpose of illustration and simplicity, Figure 3 The inspection image 300 shown will be used as an example inspection image. In some embodiments, the inspection image 300 can be generated based on the detection signal from the electron detection device 140 of the electron beam tool 40. In some embodiments, the inspection image 300 is generated by an image acquisition device included in the controller 50. In some embodiments, the inspection image 300 can be obtained from a storage device that stores the inspection image 300.

[0085] In step S920, a pattern is identified on the inspection image 300. Step S920 may be performed, for example, by the pattern identifier 420. According to an embodiment of the present disclosure, image patches may be extracted from the inspection image 300, and patterns corresponding to the extracted image patches may be identified. In some embodiments, each image patch may include at least a portion of a pattern that does not overlap with other patterns. Figure 5A As an example, four image patches P1 to P4 extracted from the inspection image 300 are illustrated. Figure 5A As shown, each of the first image patch P1 and the second image patch P2 covers the portion of the first pattern A that does not overlap with the second pattern B. Similarly, each of the third image patch P3 and the fourth image patch P4 covers the portion of the second pattern B that does not overlap with the first pattern A. According to some embodiments of the present disclosure, an image patch as large as possible may be selected to increase the pattern recognition ratio. For example, in Figure 5A , for the area covered by the first image patch P1, only one image patch is extracted instead of two or more image patches for the same area. In some embodiments of the present disclosure, the image patches may be selected so that each image patch includes (multiple) boundary lines of the corresponding pattern, since the boundary lines can be used to identify the pattern. In some embodiments, the boundary lines of the pattern may be lines for determining the outer shape of the pattern, lines for determining the inner shape of the pattern, boundary lines between different textures in the pattern, or other types of lines that can be used to identify the pattern. For example, Figure 5A The diagram shows that the first image patch P1 and the second image patch P2 include partial lines for determining the outer shape of the first pattern A (e.g., a rectangle), and the third image patch P3 and the fourth image patch P4 include partial lines for determining the outer shape of the second pattern B (e.g., a circle).

[0086] In step S920, according to some embodiments of the present disclosure, each image patch may be associated with a corresponding pattern based on a machine learning model. For example, by using the extracted image patches P1 to P4 as input to the machine learning model, each of the first image patch P1 and the second image patch P2 may be determined as part of a first pattern A, and each of the third image patch P3 and the fourth image patch P4 may be determined as part of a second pattern B. In some embodiments, each image patch may be automatically labeled with a corresponding pattern identifier. For example, Figure 5B As shown, the first image patch P1 and the second image patch P2 are labeled A1 and A2 to represent the first pattern A, and the third image patch P3 and the fourth image patch P4 are labeled B1 and B2 to represent the second pattern B.

[0087] According to an embodiment of the present disclosure, a pre-trained machine learning model may be provided. The machine learning model may be provided by, for example, a machine learning model provider 460 or the like. In some embodiments, the machine learning model may be trained to identify patterns of image patches. According to an embodiment of the present disclosure, the machine learning model may be pre-trained and may be provided to identify patterns on demand. In some embodiments, the machine learning model may be stored at a storage medium (not shown) and may be accessed to identify patterns. In some embodiments, the machine learning model may be updated when a new inspection image is acquired.

[0088] Now refer to Figure 10 , Figure 10 is a process flow diagram representing an example method for training a machine learning model according to an embodiment of the present disclosure. For illustrative purposes, reference will be made to Figure 6 Explain the process of training a machine learning model.

[0089] In step S1100, a training inspection image IM is acquired from an inspection system such as, but not limited to, tool 40 or system 100. In some embodiments, the training inspection image IM may include two or more patterns, such as in Figure 3 In some embodiments, the training inspection image IM may include a plurality of patterns.

[0090] In step S1200, training image patches are extracted. Step S1200 may be performed by, for example, the image patch extractor 461. In some embodiments, the training image patches may be extracted as Figure 5A For illustration purposes and for simplicity, we will use Figure 3 The inspection image 300 is used as the training inspection image IM and by using Figure 5A Image patches P1 to P4 are used as training image patches to illustrate the process of training a machine learning model.

[0091] In step S1300, according to an embodiment of the present disclosure, a machine learning model is trained by using the training image patches P1 to P4 extracted from the training inspection image IM as input to the machine learning model. Step S1300 can be performed by, for example, the machine learning model trainer 462. According to an embodiment of the present disclosure, the machine learning model can be trained to predict the pattern associated with each image patch by referring to a reference image contained in an information file (e.g., the information file 463).

[0092] According to an embodiment of the present disclosure, the information file 463 may include reference images corresponding to the patterns included in the training inspection image IM. For example, the information file 463 may include a first reference image corresponding to the first pattern A and a second reference image corresponding to the second pattern B. In some embodiments, the reference images included in the information file 463 may be ground-truth images of the corresponding patterns. In some embodiments, the reference images included in the information file 463 may be in a Graphic Database System (GDS) format, a Graphic Database System II (GDSII) format, an Open Artwork System Interchange Standard (OASIS) format, a Caltech Intermediate Format (CIF), or the like. In some embodiments, the reference images included in the information file 463 may include a wafer design layout for the corresponding pattern. The wafer design layout may be based on a pattern layout used to construct the wafer. The wafer design layout may correspond to one or more photolithography masks or reticles used to transfer features from the photolithography mask or reticle to the wafer. In some embodiments, the reference images in the GDS or OASIS format, etc., may include feature information stored in a binary file format that represents planar geometry, text, and other information related to the wafer design layout.

[0093] In some embodiments, the machine learning model can be trained by supervised learning. In supervised learning, the training data fed to the machine learning model includes the desired solution. For example, under the condition that the patterns corresponding to the image patches P1 to P4 are known, the machine learning model can be trained using input training data such as image patches P1 to P4. During training, the weights of the machine learning model can be updated or modified so that the machine learning model can provide inference results corresponding to the known solutions. For training purposes, when extracting image patches from the training inspection image IM, reference images contained in the information file 463 can be referenced. For example, image patches corresponding to a certain pattern can be extracted by referring to a reference image of the pattern.

[0094] While the training process has been described for a single training inspection image IM comprising two patterns, it should be understood that embodiments of the present disclosure can be applied to scenarios involving two or more training inspection images, each of which may include one or more patterns. In these scenarios, a reference image for each pattern included in the two or more training inspection images can be included in the information file 463. It should be noted that when the number of reference images and patterns is large, the process of searching to locate a corresponding pattern for a particular image patch is time-consuming and resource-intensive. Therefore, pre-training the machine learning model allows for real-time identification of inspection image patterns.

[0095] Return Reference Figure 9In step S930, according to an embodiment of the present disclosure, the patterns of the inspection image 300 may be separated, and a separated image may be generated for each pattern of the inspection image 300. Step S930 may be performed, for example, by the separated image generator 430. According to some embodiments, the separated image may be generated based on the pattern identification in step S920 and based on a reference image corresponding to the pattern. The reference image used to separate the pattern on the inspection image may be the reference image described in the information file 463.

[0096] Figure 7A is an example of an inspection image on which a reference image for each pattern is superimposed, consistent with an embodiment of the present disclosure. Figure 5B The pattern identification shown, the reference image can be superimposed on the corresponding pattern A and B. When the reference image is superimposed on the inspection image 300, the reference image can be aligned to match the corresponding pattern as closely as possible. In some embodiments, the reference image can be aligned to match the corresponding pattern in terms of angular position, outer boundaries, etc. For example, the first reference image A R is superimposed on the first pattern A, such as Figure 7A Similarly, the second reference image B R Superimposed on Figure 7A On the second pattern B.

[0097] According to an embodiment of the present disclosure, when a reference image is superimposed on the inspection image 300, the reference image can be matched with the image patch determined to correspond to the reference image. For example, by comparing the first image patch P1 and the second image patch P2 determined to correspond to the first pattern A with the first reference image A in terms of position, shape, texture, etc. R For comparison, the first reference image A R Superimposed on the inspection image 300. Similarly, by comparing the third image patch P3 and the fourth image patch P4 determined to correspond to the second pattern B in terms of position, shape, texture, etc. R For comparison, the second reference image B R Superimposed on the inspection image 300.

[0098] According to an embodiment of the present disclosure, the first separated image may be generated by removing the image region covered by the second pattern B from the inspection image 300 . Figure 7B An example of a first separated image 310 is shown. Figure 7B In the image, the removed area is represented as I B In some embodiments, the removal of the image area covered by the second pattern B can be performed based on the image being superimposed on the image. Figure 7A The second reference image B on the inspection image 300 is shown RIn some embodiments, when the image area covered by the second pattern B is removed, the removed image area I B Can be larger than the second reference image B R , to ensure that the second pattern B is removed from the first separated image 310. For example, the removed image area I B The outer boundary of the second reference image B R The outside. Figure 7B As shown, it should be noted that the portion of the first pattern A that overlaps with the second pattern B is also removed from the inspection image 300 .

[0099] Similarly, a second separated image may be generated by removing the image region covered by the first pattern A from the inspection image 300 . Figure 7C An example of a second separated image 320 is shown. Figure 7C In the image, the removed area is represented by 1 A According to some embodiments of the present disclosure, the image area covered by the first pattern A can be removed according to the following method: Figure 7A The first reference image A shown is superimposed on the inspection image 300. R In some embodiments, when the image area covered by the first pattern A is removed, the removed image area I A Can be larger than the first reference image A R , to ensure that the first pattern A is removed from the second separated image 320. For example, the removed image area I A The outer boundary of the first reference image A R The outside. Figure 7C As shown, it should be noted that the portion of the second pattern B that overlaps with the first pattern A is also removed from the inspection image 300 .

[0100] Although the process of pattern separation is described by using an inspection image including two patterns, and thus two separated images are generated, it should be noted that the present disclosure is also applicable when the inspection image includes three or more patterns. For example, when the inspection image includes three patterns (e.g., a first pattern to a third pattern), three separated images can be generated. The first separated image can be generated by removing the image area covered by the second pattern and the third pattern, the second separated image can be generated by removing the image area covered by the first pattern and the third pattern, and the third separated image can be generated by removing the image area covered by the first pattern and the second pattern.

[0101] As mentioned above, when Figure 7B When the second pattern B is removed from the inspection image 300 as shown, a portion of the first pattern A is also removed in the first separated image 310, and when Figure 7CAs shown, when the first pattern A is removed from the inspection image 300, a portion of the second pattern B is also removed in the second separated image 320. Therefore, the separated image may include an incomplete pattern. Figure 4 In step S940, the incomplete pattern in the separated image can be restored according to some embodiments of the present disclosure. Step S940 can be performed, for example, by pattern compensator 440. According to some embodiments of the present disclosure, the pattern can be restored based on an image restoration algorithm. In some embodiments, structural information, text information, or contextual information from the separated image and the reference image can be utilized when restoring the incomplete pattern.

[0102] Figure 8A is an example of a first restored image after restoring the first pattern A in the first separated image according to an embodiment of the present disclosure. Figure 8A As shown, the first restored image 311 can be generated by restoring or filling the removed portion of the first pattern A in the first separated image 310. The removed portion of the first pattern A can be a portion of the first pattern A that is removed when the first separated image 310 (e.g., Figure 7B According to some embodiments, the first reference image A may be used to remove the portion of the first pattern A. R The removed portion of the first pattern A is restored or filled in the first separated image 310. For example, the removed portion of the first pattern A may be restored or filled in the first separated image 310 by comparing the removed portion of the first pattern A with the first reference image A. R (like Figure 7A ) to determine the first reference image A R Which part of the image corresponds to the removed part of the first pattern A in the first separated image 310. For example, the non-removed part of the first pattern A in the first separated image 310 can be compared with the first reference image A in terms of position, shape, texture, etc. R Based on the first reference image A R The first restored portion Ac may be filled in the removed portion of the first pattern A, and thus the first restored image 311 including the restored first pattern A may be as shown in FIG. Figure 8A Generated as shown.

[0103] Similarly, Figure 8B is an example of a second restored image after restoring the second pattern in the second separated image according to an embodiment of the present disclosure. Figure 8BAs shown, the second restored image 321 can be generated by restoring or filling the removed portion of the second pattern B in the second separated image 320. The removed portion of the second pattern B can be a portion of the second pattern B that is removed when the first pattern A is removed in generating the second separated image 320 (e.g., Figure 7C Similarly, the pattern compensator 440 can remove the portion of the second pattern B by comparing the non-removed portion of the second pattern B with the second reference image B. R Compare and determine the second reference image B R Which part of the second reference image B corresponds to the removed part of the second pattern B in the second separated image 320. R The determination of the portion of the second pattern B corresponding to the removed portion in the second separated image 320, the second restored portion B C can be used to fill the removed portion of the second pattern B, and thus can be Figure 8B As shown, a second restored image 321 including the restored second pattern B is generated. Figure 8A and Figure 8B As shown, a first restored image 311 including the entire outline of the first pattern A and a second restored image 321 including the entire outline of the second pattern B are generated.

[0104] Return Reference Figure 9 In step S950, pattern measurement results can be obtained from the restored separated image compensated in step S940. Step S950 can be performed by, for example, pattern measurer 450. In some embodiments, measurement results for first pattern A and second pattern B can be obtained from first restored image 311 and second restored image 321, respectively. For example, contour information of first pattern A can be extracted from first restored image 311, and contour information of second pattern B can be obtained from second restored image 321. According to some embodiments of the present disclosure, pattern contour information can be used to obtain a critical dimension of the pattern to determine pattern offset, edge placement changes, and the like.

[0105] These embodiments can be further described using the following terms:

[0106] 1. A method for extracting pattern contour information from an inspection image, the method comprising:

[0107] identifying, based on an inspection image obtained from a charged particle beam inspection system, a first pattern and a second pattern that partially overlap in the inspection image;

[0108] generating a first separated image by removing an image region corresponding to the second pattern from the inspection image, wherein the first separated image includes the first pattern, a first portion of the first pattern being removed when the image region corresponding to the second pattern is removed; and

[0109] Based on a first reference image corresponding to the first pattern, the first separate image is updated to include image data representing the removed first portion of the first pattern.

[0110] 2. The method according to clause 1, further comprising:

[0111] Contour information of the first pattern is extracted from the updated first separated image.

[0112] 3. The method according to clause 1 or 2, wherein updating the first separate image is performed by filling the first portion in the first separate image using an image inpainting algorithm.

[0113] 4. The method of any one of clauses 1 to 3, wherein identifying the first pattern and the second pattern comprises:

[0114] extracting a first image patch from the inspection image, the first image patch covering a portion of the first pattern, wherein the portion of the first pattern does not overlap with the second pattern;

[0115] extracting a second image patch from the inspection image, the second image patch covering a portion of the second pattern, wherein the portion of the second pattern does not overlap with the first pattern; and

[0116] The first image patch is associated with a first reference image, and the second image patch is associated with a second reference image corresponding to a second pattern.

[0117] 5. A method according to clause 4, wherein associating the first image patch with the first reference image and associating the second image patch with the second reference image is performed by using the first image patch and the second image patch as input to a machine learning model.

[0118] 6. The method of any one of clauses 1 to 5, wherein identifying the first pattern and the second pattern is performed in real time.

[0119] 7. The method of any one of clauses 1 to 3, wherein generating the first separated image comprises:

[0120] On the inspection image, superimposing a first reference image on the first pattern and superimposing a second reference image on the second pattern, wherein the second reference image corresponds to the second pattern; and

[0121] Based on the second reference image superimposed on the second pattern, an image region corresponding to the second pattern is removed.

[0122] 8. A method according to any one of clauses 1 to 7, wherein the first reference image is in a Graphics Database System (GDS) format, a Graphics Database System II (GDSII) format, an Open Artwork Systems Interchange Standard (OASIS) format, or a Caltech Intermediate Format (CIF).

[0123] 9. The method according to any one of clauses 2 to 8, further comprising:

[0124] extracting contour information of the second pattern; and

[0125] Based on the extracted contour information of the first and second patterns, an overlay error between the first and second patterns is determined.

[0126] 10. The method of clause 9, wherein the overlay error is determined based on the centroid of the first pattern and the centroid of the second pattern, and

[0127] The centroids of the first pattern and the second pattern are determined based on the extracted contour information of the first pattern and the second pattern.

[0128] 11. A method according to any of clauses 1 to 10, wherein the first pattern is in a buried layer.

[0129] 12. A contour extraction device comprising:

[0130] a memory storing an instruction set; and

[0131] At least one processor configured to execute a set of instructions to cause the device to:

[0132] identifying, based on an inspection image obtained from a charged particle beam inspection system, a first pattern and a second pattern that partially overlap in the inspection image;

[0133] generating a first separated image by removing an image region corresponding to the second pattern from the inspection image, wherein the first separated image includes the first pattern, a first portion of the first pattern being removed when the image region corresponding to the second pattern is removed; and

[0134] Based on a first reference image corresponding to the first pattern, the first separate image is updated to include image data representing the removed first portion of the first pattern.

[0135] 13. The apparatus of clause 12, at least one processor being configured to execute a set of instructions causing the apparatus to further perform:

[0136] Contour information of the first pattern is extracted from the updated first separated image.

[0137] 14. Apparatus according to clause 12 or 13, wherein updating the first separate image is performed by filling the first portion in the first separate image using an image inpainting algorithm.

[0138] 15. The apparatus of any one of clauses 12 to 14, wherein identifying the first pattern and the second pattern comprises:

[0139] extracting a first image patch from the inspection image, the first image patch covering a portion of the first pattern, wherein the portion of the first pattern does not overlap with the second pattern;

[0140] extracting a second image patch from the inspection image, the second image patch covering a portion of the second pattern, wherein the portion of the second pattern does not overlap with the first pattern; and

[0141] The first image patch is associated with a first reference image, and the second image patch is associated with a second reference image corresponding to a second pattern.

[0142] 16. A device according to clause 15, wherein associating the first image patch with the first reference image and associating the second image patch with the second reference image is performed by using the first image patch and the second image patch as input to a machine learning model.

[0143] 17. Apparatus according to any of clauses 12 to 16, wherein identifying the first pattern and the second pattern is performed in real time.

[0144] 18. The apparatus of any one of clauses 12 to 14, wherein generating the first separated image comprises:

[0145] On the inspection image, superimposing a first reference image on the first pattern and superimposing a second reference image on the second pattern, wherein the second reference image corresponds to the second pattern; and

[0146] Based on the second reference image superimposed on the second pattern, an image region corresponding to the second pattern is removed.

[0147] 19. An apparatus according to any one of clauses 12 to 18, wherein the first reference image is in a Graphics Database System (GDS) format, a Graphics Database System II (GDSII) format, an Open Artwork Systems Interchange Standard (OASIS) format, or a Caltech Intermediate Format (CIF).

[0148] 20. A non-transitory computer-readable medium storing a set of instructions executable by at least one processor of a computing device to cause the computing device to perform a method for extracting pattern contour information from an inspection image, the method comprising:

[0149] identifying, based on an inspection image obtained from a charged particle beam inspection system, a first pattern and a second pattern that partially overlap in the inspection image;

[0150] generating a first separated image by removing an image region corresponding to the second pattern from the inspection image, wherein the first separated image includes the first pattern, a first portion of the first pattern being removed when the image region corresponding to the second pattern is removed; and

[0151] Based on a first reference image corresponding to the first pattern, the first separate image is updated to include image data representing the removed first portion of the first pattern.

[0152] 21. The computer-readable medium of clause 20, the set of instructions being executable by at least one processor of a computing device to cause the computing device to further perform:

[0153] Contour information of the first pattern is extracted from the updated first separated image.

[0154] 22. The computer-readable medium according to clause 20 or 21, wherein updating the first pattern in the first separate image is performed by filling the first portion in the first separate image using an image inpainting algorithm.

[0155] 23. The computer-readable medium of any one of clauses 20 to 22, wherein identifying the first pattern and the second pattern comprises:

[0156] extracting a first image patch from the inspection image, the first image patch covering a portion of the first pattern, wherein the portion of the first pattern does not overlap with the second pattern;

[0157] extracting a second image patch from the inspection image, the second image patch covering a portion of the second pattern, wherein the portion of the second pattern does not overlap with the first pattern; and

[0158] The first image patch is associated with a first reference image, and the second image patch is associated with a second reference image corresponding to a second pattern.

[0159] 24. A computer-readable medium according to clause 23, wherein associating the first image patch with the first reference image and associating the second image patch with the second reference image is performed by using the first image patch and the second image patch as inputs to a machine learning model.

[0160] 25. The computer-readable medium of any one of clauses 20 to 24, wherein identifying the first pattern and the second pattern is performed in real time.

[0161] 26. The computer-readable medium of any one of clauses 20 to 22, wherein generating the first separated image comprises:

[0162] On the inspection image, superimposing a first reference image on the first pattern and superimposing a second reference image on the second pattern, wherein the second reference image corresponds to the second pattern; and

[0163] Based on the second reference image superimposed on the second pattern, an image region corresponding to the second pattern is removed.

[0164] 27. A computer-readable medium according to any one of clauses 20 to 26, wherein the first reference image is in a Graphics Database System (GDS) format, a Graphics Database System II (GDSII) format, an Open Artwork Systems Interchange Standard (OASIS) format, or a Caltech Intermediate Format (CIF).

[0165] 28. A method for measuring overlay error based on an inspection image, the method comprising:

[0166] identifying, based on an inspection image obtained from a charged particle beam inspection system, a first pattern and a second pattern that partially overlap in the inspection image, the first pattern being in the buried layer;

[0167] generating, based on the inspection image, a first separated image including the first pattern and a second separated image including the second pattern;

[0168] extracting contour information of the first pattern and the second pattern based on the first separated image and the second separated image; and

[0169] Based on the extracted contour information of the first and second patterns, an overlay error between the first and second patterns is determined.

[0170] 29. The method of clause 28, wherein the overlay error is determined based on the centroid of the first pattern and the centroid of the second pattern, and

[0171] The centroids of the first pattern and the second pattern are determined based on the extracted contour information of the first pattern and the second pattern.

[0172] 30. The method according to clause 28 or 29, wherein generating the first separated image comprises:

[0173] generating a first separated image by removing an image region corresponding to the second pattern from the inspection image, wherein the first separated image includes the first pattern, a first portion of the first pattern being removed when the image region corresponding to the second pattern is removed; and

[0174] Based on a first reference image corresponding to the first pattern, the first separate image is updated to include image data representing the removed first portion of the first pattern.

[0175] 31. The method of any one of clauses 28 to 30, wherein identifying the first pattern and the second pattern comprises:

[0176] extracting a first image patch from the inspection image, the first image patch covering a portion of the first pattern, wherein the portion of the first pattern does not overlap with the second pattern;

[0177] extracting a second image patch from the inspection image, the second image patch covering a portion of the second pattern, wherein the portion of the second pattern does not overlap with the first pattern; and

[0178] The first image patch is associated with a first reference image, and the second image patch is associated with a second reference image corresponding to a second pattern.

[0179] 32. A method according to clause 31, wherein associating the first image patch with the first reference image and associating the second image patch with the second reference image is performed by using the first image patch and the second image patch as input to a machine learning model.

[0180] 33. A non-transitory computer-readable medium storing a set of instructions executable by at least one processor of a computing device to cause the computing device to perform a method for measuring overlay error from an inspection image, the method comprising:

[0181] identifying, based on an inspection image obtained from a charged particle beam inspection system, a first pattern and a second pattern that partially overlap in the inspection image, the first pattern being in the buried layer;

[0182] generating, based on the inspection image, a first separated image including the first pattern and a second separated image including the second pattern;

[0183] extracting contour information of the first pattern and the second pattern based on the first separated image and the second separated image; and

[0184] Based on the extracted contour information of the first and second patterns, an overlay error between the first and second patterns is determined.

[0185] 34. The computer-readable medium of clause 33, wherein the overlay error is determined based on a centroid of the first pattern and a centroid of the second pattern, and

[0186] The centroids of the first pattern and the second pattern are determined based on the extracted contour information of the first pattern and the second pattern.

[0187] 35. The computer-readable medium of clause 33 or 34, wherein generating the first separated image comprises:

[0188] generating a first separated image by removing an image region corresponding to the second pattern from the inspection image, wherein the first separated image includes the first pattern, a first portion of the first pattern being removed when the image region corresponding to the second pattern is removed; and

[0189] Based on a first reference image corresponding to the first pattern, the first separate image is updated to include image data representing the removed first portion of the first pattern.

[0190] 36. The computer-readable medium of any one of clauses 33 to 35, wherein identifying the first pattern and the second pattern comprises:

[0191] extracting a first image patch from the inspection image, the first image patch covering a portion of the first pattern, wherein the portion of the first pattern does not overlap with the second pattern;

[0192] extracting a second image patch from the inspection image, the second image patch covering a portion of the second pattern, wherein the portion of the second pattern does not overlap with the first pattern; and

[0193] The first image patch is associated with a first reference image, and the second image patch is associated with a second reference image corresponding to a second pattern.

[0194] 37. A computer-readable medium according to clause 36, wherein associating the first image patch with the first reference image and associating the second image patch with the second reference image is performed by using the first image patch and the second image patch as inputs to a machine learning model.

[0195] A non-transitory computer readable medium may be provided that stores information for use with a controller (e.g., Figure 1 The processor of the controller 50 executes the following instructions: image inspection, image acquisition, stage positioning, beam focusing, electric field adjustment, beam bending, bunching lens adjustment, activation of the charged particle source, beam deflection, and methods 900 and 1000. Common forms of non-transitory media include, for example, floppy disks, flexible disks, hard disks, solid-state drives, magnetic tape or any other magnetic data storage medium, compact disk read-only memory (CD-ROM), any other optical data storage medium, any physical medium with a pattern of holes, random access memory (RAM), programmable read-only memory (PROM), and erasable programmable read-only memory (EPROM), FLASH-EPROM or any other flash memory, non-volatile random access memory (NVRAM), cache, registers, any other memory chip or cartridge, and networked versions thereof.

[0196] It should be understood that the embodiments of the present disclosure are not limited to the exact configurations described above and shown in the accompanying drawings, and that various modifications and changes may be made without departing from the scope thereof. The present disclosure has been described in conjunction with various embodiments, and other embodiments of the present invention will be apparent to those skilled in the art in view of the specification and practice of the invention disclosed herein. The description and examples are to be considered as exemplary only, with the true scope and spirit of the invention being indicated by the following claims.

[0197] The above description is intended to be illustrative, and not restrictive. It will therefore be apparent to one skilled in the art that modifications as described can be made without departing from the scope of the claims set forth below.

Claims

1. A method for extracting pattern contour information from an inspection image, the method comprising: identifying, based on an inspection image obtained from a charged particle beam inspection system, a first pattern and a second pattern that partially overlap in the inspection image; generating a first separated image by removing an image region corresponding to the second pattern from the inspection image, wherein the first separated image includes the first pattern, and a first portion of the first pattern is removed when the image region corresponding to the second pattern is removed; as well as updating the first separated image to include image data representing the removed first portion of the first pattern based on a first reference image corresponding to the first pattern, Wherein identifying the first pattern and the second pattern comprises: extracting a first image patch from the inspection image, the first image patch covering a portion of the first pattern, wherein the portion of the first pattern does not overlap with the second pattern; extracting a second image patch from the inspection image, the second image patch covering a portion of the second pattern, wherein the portion of the second pattern does not overlap with the first pattern; and The first image patch is associated with the first reference image, and the second image patch is associated with a second reference image corresponding to the second pattern.

2. The method according to claim 1, further comprising: Contour information of the first pattern is extracted from the updated first separated image. 3 . The method of claim 1 , wherein updating the first separate image is performed by filling the first portion in the first separate image using an image inpainting algorithm.

4. The method of claim 1 , wherein associating the first image patch with the first reference image and associating the second image patch with the second reference image is performed by using the first image patch and the second image patch as input to a machine learning model. The method of claim 1 , wherein identifying the first pattern and the second pattern is performed in real time.

6. The method of claim 1 , wherein generating the first separated image comprises: On the inspection image, superimposing the first reference image on the first pattern and superimposing a second reference image on the second pattern, wherein the second reference image corresponds to the second pattern; as well as Based on the second reference image superimposed on the second pattern, an image region corresponding to the second pattern is removed.

7. The method of claim 1, wherein the first reference image is in a Graphics Database System (GDS) format, a Graphics Database System II (GDSII) format, an Open Artwork Systems Interchange Standard (OASIS) format, or a Caltech Intermediate Format (CIF).

8. The method according to claim 2, further comprising: extracting contour information of the second pattern; as well as An overlay error between the first pattern and the second pattern is determined based on the extracted contour information of the first pattern and the extracted contour information of the second pattern.

9. The method of claim 8, wherein the overlay error is determined based on a centroid of the first pattern and a centroid of the second pattern, and The centroid of the first pattern and the centroid of the second pattern are determined based on the extracted contour information of the first pattern and the extracted contour information of the second pattern.

10. The method of claim 1, wherein the first pattern is in a buried layer.

11. A contour extraction device comprising: Memory, which stores instruction sets; as well as at least one processor configured to execute the set of instructions to cause the device to perform: identifying, based on an inspection image obtained from a charged particle beam inspection system, a first pattern and a second pattern that partially overlap in the inspection image; generating a first separated image by removing an image region corresponding to the second pattern from the inspection image, wherein the first separated image includes the first pattern, and a first portion of the first pattern is removed when the image region corresponding to the second pattern is removed; as well as updating the first separated image to include image data representing the removed first portion of the first pattern based on a first reference image corresponding to the first pattern, Wherein identifying the first pattern and the second pattern comprises: extracting a first image patch from the inspection image, the first image patch covering a portion of the first pattern, wherein the portion of the first pattern does not overlap with the second pattern; extracting a second image patch from the inspection image, the second image patch covering a portion of the second pattern, wherein the portion of the second pattern does not overlap with the first pattern; and The first image patch is associated with the first reference image, and the second image patch is associated with a second reference image corresponding to the second pattern.

12. The device of claim 11 , wherein the at least one processor is configured to execute the set of instructions so that the device further performs: Contour information of the first pattern is extracted from the updated first separated image. 13 . The apparatus of claim 11 , wherein updating the first separate image is performed by filling the first portion in the first separate image using an image inpainting algorithm.

14. The apparatus of claim 11, wherein associating the first image patch with the first reference image and associating the second image patch with the second reference image is performed by using the first image patch and the second image patch as input to a machine learning model.

15. The apparatus of claim 11, wherein identifying the first pattern and the second pattern is performed in real time.

16. The apparatus of claim 11 , wherein generating the first separated image comprises: On the inspection image, superimposing the first reference image on the first pattern and superimposing a second reference image on the second pattern, wherein the second reference image corresponds to the second pattern; as well as Based on the second reference image superimposed on the second pattern, an image region corresponding to the second pattern is removed.

17. The apparatus of claim 11, wherein the first reference image is in a Graphics Database System (GDS) format, a Graphics Database System II (GDSII) format, an Open Artwork Systems Interchange Standard (OASIS) format, or a Caltech Intermediate Format (CIF).

18. A non-transitory computer-readable medium storing a set of instructions executable by at least one processor of a computing device to cause the computing device to perform a method for extracting pattern contour information from an inspection image, the method comprising: identifying, based on an inspection image obtained from a charged particle beam inspection system, a first pattern and a second pattern that partially overlap in the inspection image; generating a first separated image by removing an image region corresponding to the second pattern from the inspection image, wherein the first separated image includes the first pattern, and a first portion of the first pattern is removed when the image region corresponding to the second pattern is removed; as well as updating the first separated image to include image data representing the removed first portion of the first pattern based on a first reference image corresponding to the first pattern, Wherein identifying the first pattern and the second pattern comprises: extracting a first image patch from the inspection image, the first image patch covering a portion of the first pattern, wherein the portion of the first pattern does not overlap with the second pattern; extracting a second image patch from the inspection image, the second image patch covering a portion of the second pattern, wherein the portion of the second pattern does not overlap with the first pattern; and The first image patch is associated with the first reference image, and the second image patch is associated with a second reference image corresponding to the second pattern.

19. The computer-readable medium of claim 18, the set of instructions being executable by at least one processor of the computing device to cause the computing device to further perform: Contour information of the first pattern is extracted from the updated first separated image.

20. The computer-readable medium of claim 18, wherein updating the first pattern in the first separate image is performed by filling the first portion in the first separate image using an image inpainting algorithm.

21. The computer-readable medium of claim 18, wherein associating the first image patch with the first reference image and associating the second image patch with the second reference image is performed by using the first image patch and the second image patch as input to a machine learning model.

22. The computer-readable medium of claim 18, wherein identifying the first pattern and the second pattern is performed in real time.

23. The computer-readable medium of claim 18, wherein generating the first separated image comprises: On the inspection image, superimposing the first reference image on the first pattern and superimposing a second reference image on the second pattern, wherein the second reference image corresponds to the second pattern; as well as Based on the second reference image superimposed on the second pattern, an image region corresponding to the second pattern is removed.

24. The computer-readable medium of claim 18, wherein the first reference image is in a Graphics Database System (GDS) format, a Graphics Database System II (GDSII) format, an Open Artwork Systems Interchange Standard (OASIS) format, or a Caltech Intermediate Format (CIF).

25. A method for measuring overlay error based on an inspection image, the method comprising: identifying, based on an inspection image obtained from a charged particle beam inspection system, a first pattern and a second pattern that partially overlap in the inspection image, the first pattern being in the buried layer; generating, based on the inspection image, a first separated image including the first pattern and a second separated image including the second pattern; updating the first separated image based on a first reference image corresponding to the first pattern, and updating the second separated image based on a second reference image corresponding to the second pattern; extracting contour information of the first pattern and the second pattern based on the updated first separated image and the updated second separated image; as well as An overlay error between the first pattern and the second pattern is determined based on the extracted contour information of the first pattern and the second pattern.

26. The method of claim 25, wherein the overlay error is determined based on a centroid of the first pattern and a centroid of the second pattern, and The centroid of the first pattern and the centroid of the second pattern are determined based on the extracted contour information of the first pattern and the second pattern.

27. The method of claim 25, wherein generating the first separated image comprises: generating the first separated image by removing an image region corresponding to the second pattern from the inspection image, wherein the first separated image includes the first pattern, and a first portion of the first pattern is removed when the image region corresponding to the second pattern is removed; as well as Wherein, updating the first separated image based on a first reference image corresponding to the first pattern includes: The first separate image is updated to include image data representing the removed first portion of the first pattern based on a first reference image corresponding to the first pattern.

28. The method of claim 25, wherein identifying the first pattern and the second pattern comprises: extracting a first image patch from the inspection image, the first image patch covering a portion of the first pattern, wherein the portion of the first pattern does not overlap with the second pattern; extracting a second image patch from the inspection image, the second image patch covering a portion of the second pattern, wherein the portion of the second pattern does not overlap with the first pattern; as well as The first image patch is associated with a first reference image, and the second image patch is associated with a second reference image corresponding to the second pattern.

29. The method of claim 28, wherein associating the first image patch with the first reference image and associating the second image patch with the second reference image is performed by using the first image patch and the second image patch as input to a machine learning model.

30. A non-transitory computer-readable medium storing a set of instructions executable by at least one processor of a computing device to cause the computing device to perform a method for measuring overlay error from an inspection image, the method comprising: identifying, based on an inspection image obtained from a charged particle beam inspection system, a first pattern and a second pattern that partially overlap in the inspection image, the first pattern being in the buried layer; generating, based on the inspection image, a first separated image including the first pattern and a second separated image including the second pattern; updating the first separated image based on a first reference image corresponding to the first pattern, and updating the second separated image based on a second reference image corresponding to the second pattern; extracting contour information of the first pattern and the second pattern based on the updated first separated image and the updated second separated image; as well as An overlay error between the first pattern and the second pattern is determined based on the extracted contour information of the first pattern and the second pattern.

31. The computer-readable medium of claim 30, wherein the overlay error is determined based on a centroid of the first pattern and a centroid of the second pattern, and The centroid of the first pattern and the centroid of the second pattern are determined based on the extracted contour information of the first pattern and the second pattern.

32. The computer-readable medium of claim 30, wherein generating the first separated image comprises: generating the first separated image by removing an image region corresponding to the second pattern from the inspection image, wherein the first separated image includes the first pattern, and a first portion of the first pattern is removed when the image region corresponding to the second pattern is removed; as well as Wherein, updating the first separated image based on a first reference image corresponding to the first pattern includes: The first separate image is updated to include image data representing the removed first portion of the first pattern based on a first reference image corresponding to the first pattern.

33. The computer-readable medium of claim 30, wherein identifying the first pattern and the second pattern comprises: extracting a first image patch from the inspection image, the first image patch covering a portion of the first pattern, wherein the portion of the first pattern does not overlap with the second pattern; extracting a second image patch from the inspection image, the second image patch covering a portion of the second pattern, wherein the portion of the second pattern does not overlap with the first pattern; as well as The first image patch is associated with a first reference image, and the second image patch is associated with a second reference image corresponding to the second pattern.

34. The computer-readable medium of claim 33, wherein associating the first image patch with the first reference image and associating the second image patch with the second reference image is performed by using the first image patch and the second image patch as input to a machine learning model.

Images