Inspection system and method with gray level compensation

Abstract

The invention discloses an inspection system and method with gray level compensation. A method includes: generating a first image by scanning a mask; exposing the mask in a lithographic operation after generating the first image; after exposure, generating a second image by scanning the mask; generating a compensated third image by performing gray scale map local compensation on the second image, the gray scale map local compensation being performed via a transformation function; generating a comparison result by comparing the compensated third image with the first image; and determining whether the mask has defects based on the comparison result.

Description

Technical Field

[0001] The present disclosure relates generally to the field of semiconductor technology, and more particularly to an inspection system and method with grayscale compensation. Background Art

[0002] The semiconductor integrated circuit (IC) industry has experienced exponential growth. Technological advances in IC materials and design have produced successive generations of ICs, each with smaller and more complex circuits than the previous generation. Over the course of IC evolution, functional density (i.e., the number of interconnected devices per chip area) has generally increased, while geometry size (i.e., the smallest component (or line) that can be created using a manufacturing process) has decreased. This shrinking process generally provides benefits by increasing production efficiency and reducing associated costs. This shrinking has also increased the complexity of processing and manufacturing ICs. Summary of the invention

[0003] According to a first aspect of the present disclosure, a method for a semiconductor device is provided, comprising: generating a first image by scanning a mask; after the generating of the first image, exposing the mask in a photolithography operation; after the exposing, generating a second image by scanning the mask; generating a compensated third image by performing a first grayscale map local compensation on the second image, the first grayscale map local compensation being performed via a transformation function; generating a comparison result by comparing the compensated third image with the first image; and determining whether there is a defect on the mask based on the comparison result.

[0004] According to a second aspect of the present disclosure, a method for a semiconductor device is provided, comprising: generating a reference image by scanning a mask; after generating the reference image, generating a second image by scanning the mask; generating a third image by performing grayscale map local compensation on the second image, the grayscale map local compensation being performed via a trained neural network; generating a comparison result by comparing the third image with the reference image; and determining whether there is a defect on the mask based on the comparison result.

[0005] According to a third aspect of the present disclosure, a system for a semiconductor device is provided, comprising: a light source; a mask stage operable to position a mask in an optical path of light generated by the light source; a detector operable to generate a second image by scanning the mask on the mask stage, the mask having been exposed to extreme ultraviolet (EUV) light in a lithography system before scanning the mask; and a controller comprising: a memory storing computer instructions; and a processor configured to execute the computer instructions to perform the following operations: generating a compensated third image by performing local compensation of a grayscale map on the second image; generating a comparison result by comparing the compensated third image with a reference image of the mask; and determining whether there is a defect on the mask based on the comparison result. BRIEF DESCRIPTION OF THE DRAWINGS

[0006] When the following detailed description is read in conjunction with the accompanying drawings, the various aspects of the present disclosure can be best understood from the following detailed description. It should be noted that, in accordance with standard practice in the industry, various features are not drawn to scale. In fact, for the sake of clarity of discussion, the size of various features may be arbitrarily increased or reduced.

[0007] Figures 1A to 1C are diagrams depicting systems and methods for inspecting a substrate and example substrates according to embodiments of the present disclosure.

[0008] Figure 2 is a diagram illustrating compensation of a grayscale sampling map according to various aspects of the present disclosure.

[0009] Figure 3 is a view illustrating compensation of grayscale distribution of multiple layers according to various aspects of the present disclosure.

[0010] Figure 4 is a diagram depicting compensation of grayscale distribution of multiple layers via machine learning according to various embodiments.

[0011] Figure 5A and Figure 5B is a diagram depicting detection of outliers in a gray level distribution according to various embodiments.

[0012] Figure 6 is a diagram depicting point selection based on a mask design file according to various embodiments.

[0013] Figure 7 is a block diagram of a controller according to various embodiments. DETAILED DESCRIPTION

[0014] The following disclosure provides many different embodiments or examples for implementing the different features of the provided subject matter. Specific examples of components and arrangements are described below to simplify the present disclosure. Of course, these are merely examples and are not intended to be limiting. For example, in the subsequent description, forming a first feature above or on a second feature may include an embodiment in which the first feature and the second feature are formed in direct contact, and may also include an embodiment in which an additional feature may be formed between the first feature and the second feature so that the first feature and the second feature may not be in direct contact. In addition, the present disclosure may repeat reference numerals and / or letters in various examples. This repetition is for the purpose of simplicity and clarity, and does not itself dictate the relationship between the various embodiments and / or configurations discussed.

[0015] In addition, for ease of description, spatially relative terms (e.g., "below," "under," "lower," "above," "upper," etc.) may be used herein to describe the relationship of one element or feature shown in the figures relative to another (one or more) elements or (one or more) features. These spatially relative terms are intended to encompass different orientations of the device in use or operation in addition to the orientation depicted in the figures. The device may be otherwise oriented (rotated 90 degrees or at other orientations), and the spatially relative descriptors used herein should be interpreted accordingly.

[0016] For the convenience of description, terms such as "about", "substantially", etc. may be used herein. Those skilled in the art will be able to understand and deduce the meaning of such terms.

[0017] Semiconductor manufacturing generally involves forming electronic circuits by performing multiple depositions, etches, cleanings, anneals, and / or implantations of material layers to form a stacked structure including many semiconductor devices and interconnections therebetween.

[0018] In many cases, etching is performed based on a pattern. The pattern can first be transferred to a resist layer by reflecting light from a mask having a pattern thereon. Defects in the mask are transferred along with the pattern, which can lead to improper etching and failure of the semiconductor device. Therefore, an inspection of the mask is usually performed before the mask is used in semiconductor processing. For example, after the mask is manufactured, a "golden" digital image of a mask without defects can be captured. Then, before the mask is used, another digital image can be captured and compared to the golden digital image to determine whether there are (one or more) defects. In response to the presence of (one or more) defects, the mask can be cleaned, repaired, reworked, or replaced before starting semiconductor processing.

[0019] One of the benefits of inspection is to determine if there are (one or more) defects on the front side of the mask. To achieve this, the inspection tool can compare the scanned image of the mask with a golden image collected before EUV wafer exposure. When the mask is exposed to extreme ultraviolet (EUV) light, the H plasma and EUV light can cause brightness changes on the mask surface. The brightness changes of the mask caused by EUV exposure can cause grayscale changes during inspection. This can lead to false detection of (one or more) defects and may cause scan failures and a large number of false positives. In some cases, the scan failure rate of brightness / darkness (BD) scans can be as high as 2%, which can result in a reduction in tool availability of up to 3.6 hours / week.

[0020] Brightness may change during EUV light exposure due to contamination being removed from the surface by hydrogen plasma cleaning and EUV light. For example, a new mask may be contaminated with hydrocarbons before a golden image is collected. Hydrocarbons on the EUV mask surface may undergo photodissociation when EUV photons are absorbed by hydrocarbon molecules. This absorption excites the molecules to a higher electronic state, which can then dissociate into smaller fragments. The exact mechanism of photodissociation can vary depending on the hydrocarbon molecules involved, but can generally include one or more steps as described below. EUV photons are absorbed by hydrocarbon molecules, exciting them to a higher electronic state. The excited molecules undergo internal vibrational relaxation, causing the absorbed energy to be redistributed between different vibrational modes of the molecule. The excited molecules then dissociate into smaller fragments. Photodissociation of hydrocarbons on the EUV mask surface can cause a number of problems, including mask contamination, mask damage, and yield loss. The dissociated fragments can be deposited on the mask surface, forming a contamination layer. This contamination layer can absorb EUV light, reducing the efficiency of the exposure process. The dissociated debris can also damage the mask surface, causing defects that can be printed on the wafer. Both mask contamination and mask damage can lead to wafer yield loss.

[0021] A variety of techniques have been developed to mitigate the effects of photodissociation on EUV masks, including the use of low outgassing materials, protective coatings, vacuum baking, laser cleaning, etc. Low outgassing materials are less likely to release hydrocarbons that form a contaminant layer on the mask surface. A protective coating can be applied to the mask surface to protect it from the dissociation debris. Vacuum baking can be used to remove hydrocarbons from the mask surface before exposure. Laser cleaning can be used to remove contaminants from the mask surface during exposure.

[0022] Light calibration can be used to match the grayscale between the scanned image and the golden image. Light calibration can be performed based on two or three points on the mask. Two-point or three-point light calibration may not compensate for the local brightness of the mask, thereby failing to adequately mitigate brightness changes that cause scanning failures. For example, if there are measurement errors at two or three points, this may result in unstable light calibration results. When the mask scan fails due to changes in mask brightness or grayscale, recapturing the golden image of the mask can effectively restore the stability of the calibration results. That is, the new golden image can better match the current brightness or grayscale of the mask. For different types of inspection tools, recapturing the golden image may take 2 hours or up to 4 hours. In addition, the mask that failed to scan can be scanned via a tool that compares the image with a mask graphic design system (GDS) file (e.g., mask design), which may take up to 12 hours for a single mask.

[0023] Embodiments of the present disclosure include various beneficial features for reducing false positives due to grayscale variations, resulting in increased tool usability. In some embodiments, in order to compensate for local brightness variations, a brightness map of the mask is collected before the mask is scanned. By measuring the grayscale levels of various positions distributed on the mask, local brightness information of the mask can be collected for each scan. In some embodiments, grayscale compensation includes compensation via a transformation equation, a machine learning model, or both. In some embodiments, in addition to or instead of collecting spatial information at each position on the mask, the grayscale distribution of the material of the mask can also be collected, which is conducive to removing outliers of the calibration points. These verification methods can also serve as the basis for z-map measurement error checks. In some embodiments, material information from the mask GDS file is used as the basis for determining the material type associated with each position, which can help reduce grayscale compensation errors caused by misjudgment of material type.

[0024] Figure 1A and Figure 1B is a diagram of a system 10 and method 11 operable to inspect a substrate according to an embodiment of the present disclosure. In some embodiments, the substrate is a mask, such as a mask used to perform EUV lithography.

[0025] Figure 1B is a flow chart of a method of performing sensitivity verification according to various embodiments. Figure 1B The actions shown in Figure 1A , Figure 1C and Figures 2 to 6 The system 10 described is implemented. Figure 1BA flow chart of method 11 for performing sensitivity verification according to one or more aspects of the present disclosure is shown. Method 11 is an example and is not intended to limit the present disclosure to what is explicitly shown in method 11. Additional actions may be provided before, during, and after method 11, and some of the actions described may be replaced, eliminated, or moved for additional embodiments of the method. For example, method 11 may be used to inspect masks, wafers, etc. For simplicity, not all actions are described in detail herein. Figure 1A , Figure 1C and Figures 2 to 6 The actions of method 11 are described by referring to the elements of system 10. Many actions can be referred to as Figure 1A The controller 120 described here is executed, and the embodiment of the controller 120 is referenced Figure 6 For example, the controller 60 may execute instructions to perform the actions of the method 11. It should be understood that the method 11 is not limited to being performed by the system 10 and / or the controller 120, 60, and in other embodiments may be performed by a system and / or controller that is different from the system 10 and / or the controller 120, 60 in one or more aspects.

[0026] exist Figure 1A In the embodiment of the present invention, the system 10 is operable to generate optical image data associated with the substrate 14 and generate inspection results based on the optical image data. The optical image data may include an optical image of the substrate 14.

[0027] Examples of the substrate 14 include a mask used for photolithography. The mask has a pattern arranged on its surface. Figure 1C An example of a mask 14 according to various embodiments is depicted in FIG.

[0028] Figure 1CA mask 14 including four different material layers or multilayers 142, 144, 146, 148 is depicted. In some embodiments, layer 142 may be a substrate. For example, layer 142 may be or include glass, such as fused silica or ultra-low expansion (ULE) glass. Multilayer 144 on substrate 142 may be a reflective multilayer stack that includes alternating layers of two or more different materials, such as silicon and molybdenum. Layer 146 on multilayer 144 may be an absorber layer that can absorb any unreflected light and may include tantalum, boron nitride, etc. Layer 148 on absorber layer 146 may be a capping layer, which may be a thin layer of a material such as ruthenium or silicon that is deposited on top of the absorber layer to protect mask 14 from contamination and / or oxidation. Fewer or additional layers may be included in mask 14, such as adhesion layers, buffer layers, anti-reflective coating (ARC) layers, etch stop layers, etc. One or more of layers 142, 144, 146, 148 may have an exposed surface. For example, the openings 145 in the absorber layer 146 can expose the multiple layers 144 so that light incident on the multiple layers 144 can be reflected while light incident on the absorber layer 146 can be absorbed. In this way, the openings 145 of the mask 14 can form a pattern that can be used to expose a photoresist layer on a wafer during a photolithography operation.

[0029] exist Figure 1A In the embodiment, the optical image data corresponding to the pattern may be generated, for example, by the sensor or detector 100. The controller 120 may detect defects of the pattern of the mask by comparing the golden or reference image generated via the golden or reference image generation function with the optical image data output by the detector 100.

[0030] The system 10 includes a platform 12 that can translate in one or more horizontal directions (e.g., X-axis directions and Y-axis directions) and a third direction (which can be a vertical direction (e.g., Z-axis direction) or a rotational direction (e.g., θ-axis direction). The system 10 may include a monitoring device that determines the position of the platform 12. The system 10 includes a light source 16 that generates light 13 of a selected wavelength (e.g., an EUV wavelength (e.g., 13.5 nanometers)). The system 10 includes an illumination optical device 18 that guides and / or collects light 13 from the light source 16 so that light 15 can be irradiated on a mask 14. The imaging device 100 generates optical image data associated with the mask 14. The system 10 may include a loader that may have a robotic arm that is operable to position the mask 14 on the platform 12 and / or remove the mask 14 from the platform 12.

[0031] The system may include one or more controllers 120 that may control various operations of the system 10. The controller 120 may be in data communication with the detector 100, the light source 16, the illumination optics 18, and the platform 12.

[0032] The light source 16 is operable to generate light 13 of a single wavelength ω, such as a laser, and may be a laser light source 16. The output wavelength range of the light source 16 may be in a wavelength domain, such as a visible light domain, an ultraviolet (UV) domain, a mid-ultraviolet (MUV) domain, a deep ultraviolet (DUV) domain, an extreme ultraviolet (EUV) domain, etc. The light source 16 may include a visible domain laser, such as a helium-neon laser, an argon laser, a krypton laser, etc. In some embodiments, the light source 16 includes an excimer laser, a helium-cadmium laser, a nitrogen laser, an argon fluoride laser, a krypton fluoride laser, a free electron laser, etc.

[0033] The detector 100 may include an imaging device and / or a photodiode. The controller 120 may be operable to calibrate or select parameters of the imaging device, which may include a charge coupled device (CCD) imaging device or a complementary metal oxide semiconductor (CMOS) imaging device. Both the CCD imaging device and the CMOS imaging device may be a time delay integration (TDI) imaging device. In some embodiments, the controller 120 includes a microcontroller unit (MCU), a processor, a multiprocessor, etc. The controller 120 may communicate data with the detector 100 to select its parameters.

[0034] Various parameters of CCD, CMOS, TDI CCD and TDI CMOS sensors can be controlled in real time via, for example, an MCU (e.g., controller 120). The gain of the sensor can control the degree to which the signal from each pixel of the detector 100 is amplified. Increasing the gain can increase the sensitivity of the detector 100, but it can also increase the noise. The offset of the detector 100 controls the black level of the image. Increasing the offset can reduce the noise in the image, but it can also make it more difficult to detect small defects in the mask. The integration time can refer to the amount of time the detector 100 is exposed to light. Increasing the integration time can improve the signal-to-noise ratio (SNR) of the image, but it can also make the detector 100 more susceptible to motion blur. The detector 100 can be triggered to start and stop integration in real time, which can be helpful in synchronizing the detector 100 with other devices in the mask scanner or inspection system.

[0035] In addition to these parameters, some detectors 100 may also allow real-time control of other features, such as the number of TDI levels or the shift frequency. The parameters that can be controlled or selected in real time can vary depending on the detector 100 and the MCU. In some embodiments, the selection of various parameters or characteristics can be performed by the controller 120.

[0036] The following are examples of how real-time control or selection of sensor parameters may be used in semiconductor mask inspection. The gain of the detector 100 may be adjusted to compensate for changes in lighting conditions. The offset of the detector 100 may be adjusted to compensate for changes in background noise levels. The integration time of the detector 100 may be adjusted to improve the SNR of different mask features. The detector 100 may be triggered to start and stop integration at selected points in the mask scanning process, which may be useful for synchronizing the detector 100 with a mask scanner or capturing images of selected areas on a mask.

[0037] CCD, CMOS, TDI CCD and / or TDI CMOS sensors can include various features that are beneficial to semiconductor mask inspection. For example, CCD sensors can have features such as pixel size, well depth, and read noise. The smaller size of each pixel on the detector 100 can be beneficial to higher resolution images, but can also result in more noise. Well depth can refer to the maximum number of photoelectrons that a pixel can accommodate before saturation, which is beneficial for low light imaging and high contrast scene imaging. Read noise can refer to the amount of noise introduced into the image during the readout process. Reducing read noise can be beneficial for low light imaging and high dynamic range scene imaging.

[0038] For CMOS sensors, pixel size, fill factor, and dark current may be characteristics that are beneficial for semiconductor mask inspection. As previously mentioned, smaller pixels may allow for higher resolution images, but may also result in more noise. Fill factor may refer to the percentage of pixel area that is sensitive to light. A higher fill factor may result in a higher sensitivity detector 100. Dark current may refer to the amount of current that flows through a CMOS sensor even when the CMOS sensor is not exposed to light. Reduction in dark current is beneficial for low light imaging and high dynamic range scene imaging.

[0039] In a TDI CCD sensor, beneficial characteristics may include the number of stages and the shift frequency. The number of stages is related to the number of times the image charge packets are shifted along the rows of the CCD sensor. A higher number of stages may result in a higher signal-to-noise ratio (SNR) image. The shift frequency may refer to the frequency at which the image charge packets are shifted along the rows of the CCD sensor. Matching the shift frequency to the speed of the object being imaged (e.g., a mask) is beneficial for improving image quality.

[0040] In a TDI CMOS sensor, the number of stages, shift frequency, and rolling shutter may be features that are advantageous for mask inspection in one or more ways. A CMOS sensor may use a rolling shutter, which refers to the detector 100 being read out one row at a time. A rolling shutter can cause image distortion when the object being imaged (e.g., a mask) moves. A TDI CMOS sensor may include a global shutter, which reads out the entire detector 100 at a time. This is advantageous in reducing or eliminating image distortion, but may increase the cost and / or complexity of the CMOS sensor.

[0041] In the context of semiconductor mask inspection, the following characteristics may be particularly beneficial: pixel size, well depth, read noise, TDI number of levels, shift frequency, rolling shutter, and integration time. The pixel size can be selected small enough to resolve features on the mask. The well depth can be selected high enough to avoid saturating the pixel when imaging bright features on the mask. The read noise can be selected low enough to avoid masking small defects on the mask. The number of levels can be selected high enough to achieve a selected SNR. The shift frequency can be selected to match the speed of the mask scanner. A global shutter can be selected to avoid image distortion. Semiconductor mask inspection can be performed under low light conditions, so that a detector 100 with a long integration time can be advantageously selected. In some embodiments, the selection of various parameters or characteristics can be performed by the controller 120.

[0042] One or more of the controller 120 and the detector 100 may be in data communication with a data storage device or system 124, which may be a database, a server, a data center, etc. The data storage device 124 may store digital data generated by the controller 120, the detector 100, an external device, a combination thereof, etc. For example, the digital data may include one or more of a recipe, a golden or reference image, an inspection image, sensor data (in addition to data from the detector 100), a combination thereof, etc. The digital data may include one or more inspection algorithms that may be used by the system 10 to inspect the mask 14 and may include a defect detection algorithm. The digital data may include one or more mask designs or design files (e.g., GDS files), which may include reference designs and mask patterns that may be compared to the mask 14 being inspected. The digital data may include historical inspection data, which may include past inspection results, including detected anomalies or defects. The digital data may include one or more inspection settings, such as parameters and / or configurations for different types of masks, inspection modes, or other hardware settings. The digital data may include an inspection report, which may be a detailed report including metrics from the inspection, detected defects, and images or scans. The digital data may include system logs, which may be related to system performance, errors, and other events of the system 10. The digital data may include calibration data, which may be or include data related to the calibration of various hardware components (e.g., the detector 100 and the light source 16). The digital data may include device status data, such as information associated with the status of various system components, including, for example, predictive / scheduled maintenance data. Other data included in the digital data may include vendor and component information, such as details about hardware components, their vendors, their firmware versions, etc. Quality control data may be stored as digital data, including information related to quality control procedures that the mask 14 may undergo and results from the quality control procedures. Job queues, such as schedules and / or queues for upcoming inspections, may be stored as digital data in the data storage system 124. The data storage system 124 may store raw data, such as raw data collected by the detector 100 during an inspection. Additional or less data may be stored in and by the data storage system 124.

[0043] The digital data of the data storage system 124 may be accessed by the controller(s) 120. The controller 120 may store at least some of the digital data to the data storage system 124 and may receive at least some of the digital data from the data storage system 124. For example, the controller 120 may receive a recipe from the data storage system 124 and may control the operation of the platform 12, the light source 16, and the detector 100 based on the recipe. In another example, the controller 120 may receive a historical or reference or golden image and an inspection image from the data storage system 124 and may compare the images to determine if there are defects in the mask 14. The controller 120 may generate a comparison result based on the comparison and may store the comparison result in the data storage system 124.

[0044] Figure 1B A flow chart of a method 11 for inspecting a substrate (eg, a mask) according to various embodiments is depicted.

[0045] Method 11 begins at act 102, which includes collecting one or more golden images or reference images associated with one or more corresponding masks. In the context of mask inspection, a "golden" image can be a reference image of a perfect (or near-perfect) mask. The golden image is used to compare with an image of the actual mask to identify any defects. The golden image can be generated by imaging a mask that is known to produce high-quality wafers. The golden image can be taken under the same or similar conditions as the image of the actual mask to be inspected. For example, the same illumination source 16, optical device 18, and detector 100 can be used. Once the golden image has been created, it can be used to inspect the actual mask for defects. The difference between the golden image and the image of the actual mask can indicate defects.

[0046] Action 102 is followed by action 104. In action 104, the mask 14 is exposed to ultraviolet light (e.g., EUV light), which can be performed during wafer processing (e.g., a lithography process). In EUV lithography, a mask or photomask 14 is used to transfer a circuit pattern onto a semiconductor wafer. Prior to the lithography process, the mask 14 can be cleaned to remove any particles or contaminants. The mask can then be loaded into an EUV scanner and aligned to ensure that the pattern can be accurately transferred. A laser plasma (LLP) source then generates EUV light. A collection mirror focuses and directs the light to the mask 14. The EUV light illuminates the mask 14. The mask 14 contains a pattern made of materials that alternately absorb or reflect EUV light. The light reflected by the mask 14 carries the pattern information. The patterned EUV light is then projected onto a semiconductor wafer coated with photoresist, which can be accomplished through a series of lenses called projection optics or a projection optics box (POB). In most cases, the features of the mask 14 are scaled down by projection optics, for example by a factor of 4 to 5, in order to accurately project extremely small features onto the wafer.

[0047] After exposure, the mask 14 may be inspected for defects or particle contamination that may have occurred during processing in order to improve the quality of future wafer exposures. The inspection is described in more detail with reference to actions 106, 108, 112, 114, which are described in detail in Figure 1B 14 is highlighted by a dashed box. Metrics related to the performance and accuracy of the mask 14 can be collected and sent to a database 124 for analysis by the controller 120 or another analysis system. Once the exposure cycle is completed and the mask 14 has been inspected and cleaned, the mask 14 can be returned to a protective environment to reduce the risk of contamination.

[0048] Action 104 is followed by action 106. Action 106 includes collecting and / or generating a grayscale map associated with the mask 14, and may include performing local compensation of the collected or generated grayscale map. To compensate for local brightness variations, a brightness or grayscale map of the mask 14 may be collected or generated before scanning the mask 14. By measuring grayscale or brightness levels distributed over the entire mask 14 or a large portion of the mask 14, local brightness information of the EUV mask 14 may be generated for each scan.

[0049] Figure 2 An example of sampling points of a grayscale map of the mask 14 is depicted. Sampling the points may be performed via operations 200, 210, 220, which will be described in detail below.

[0050] Operation 200 is an example of sampling points 202, 204, 206 of a grayscale map. Each point on the map has a z-height value and a grayscale value. For example, a first point 202 is associated with a boundary (or "black boundary") region of mask 14, a second point 204 is associated with an absorber region of mask 14, and a third point 206 is associated with a multilayer region of mask 14. The absorber region may be a region of mask 14 where an absorber layer 146 is present. The multilayer region may be a region of mask 14 where an opening 145 is present and exposes multilayer 144. In operation 200, the material of mask 14 may be classified or identified based on the z-axis height at each location. For example, the height of the absorber region may exceed the height of the multilayer region, such as Figure 1C shown.

[0051] In operation 210 after operation 200, a target gray level of each material may be calculated based on the gray level distribution of each material. In some embodiments, an average gray level of each material is calculated. Figure 2 In the depicted example, a grayscale map 212 can be generated that includes only the grayscale levels 216 of the third point 206 associated with the multi-layer region of the mask 14. Some of the grayscale levels 216 may exceed the average value of the grayscale levels, while other grayscale levels in the grayscale levels 216 may not exceed (e.g., may be lower than) the average value of the grayscale levels. In some embodiments, the target grayscale level may be the average value, while in other embodiments, another grayscale level that exceeds or is lower than the average value may be used as the target grayscale level. The grayscale map 212 may also be referred to as a first or pre-compensated grayscale map 212.

[0052] In operation 220 after operation 210, a second grayscale map 222 may be generated, wherein each point on the map 222 has been compensated. For example, the second grayscale map 222 may include a compensated grayscale 226 of the third point 206. In some embodiments, the compensated grayscale 226 has the same value as each other, such as a target grayscale (e.g., an average value). In some embodiments, the compensated grayscale 226 has a value within a range smaller than the range of the grayscale 216. For example, a first range of grayscale values ​​of the grayscale 216 may exceed a second range of grayscale values ​​of the compensated grayscale 226 by about 2, about 3, about 4, about 5 times, or another suitable multiple. That is, the second range may be about half the size of the first range or less. In some embodiments, the generation of the compensated grayscale 226 is by machine learning, such as based on a trained neural network.

[0053] Figure 3 is a diagram depicting a process 30 for performing grayscale compensation according to various embodiments. Process 30 may be a reference Figure 2 Embodiments of operations 210, 220 are described.

[0054] In process 30, the first grayscale image 300 may represent Figure 2 The first grayscale map 300 may be referred to as a pre-compensated grayscale map. The vertical axis of the first grayscale map 300 may represent the count or number of third points 206 associated with a grayscale level or a narrow range of grayscale levels. The horizontal axis of the first grayscale map 300 may represent the grayscale value associated with each bar of the first grayscale map 300. Figure 3 As shown, before compensation, the range of gray levels of the first gray level map 300 may have a size D1, eg, approximately 60 levels, which may be centered around the first gray level (eg, approximately 2130).

[0055] The transformation function 310 may be applied to the first grayscale map 300 to generate a second grayscale map 360. The second grayscale map 360 may be referred to as a post-compensation grayscale map 360. The range of grayscale levels of the post-compensation of the second grayscale map 360 may have a size D2, for example, about 10 levels, which may be centered on a second grayscale level (e.g., about 2130). The second grayscale level and the first grayscale level may be the same as or different from each other. Figure 3 In the example, the first and second gray levels are substantially the same as each other.

[0056] In some embodiments, the transform function 310 may be In the transform function 310, each pre-compensated individual gray level may be represented by “GL”, and “GL mean " can refer to the average gray level, "GL std ” may refer to the standard deviation of the pre-compensated histogram, “k” may be the compensation factor, and “GL new " may refer to post-compensation individual gray levels. Increasing the compensation factor k reduces the size of dimension D2. In some embodiments, the compensation factor k is selected based on scanning performance.

[0057] Figure 4 is a diagram depicting a process 40 for performing grayscale compensation according to various embodiments. Process 40 may be a reference Figure 2 Embodiments of operations 210, 220 are described.

[0058] In process 40, gray level compensation can be performed via machine learning, for example based on a trained neural network. The gray level of mask 14 can gradually change due to repeated exposure to EUV light. A model including a convolutional neural network (CNN) can be trained based on the gray level distribution of historical scans and golden or reference scans. In some embodiments, the model is operable to match the gray level of a new scan to the gray level of a golden or reference scan.

[0059] like Figure 4As shown, CNN model 430 can be trained based on training data, which can include historical scans 410 and golden or reference scans 420. In some embodiments, historical scans 410 and golden scans 420 are for a single mask, such as mask 14. In some embodiments, historical scans 410 and (one or more) golden scans 420 are for a large number of masks, which can include mask 14. In embodiments where CNN model 430 is trained on a large number of masks, historical scans 410 and corresponding golden scans 420 can be associated with each other. CNN model 430 can also be referred to as trained CNN model 430. Trained CNN model 430 can be used to compensate the grayscale of new scan 400 to generate compensated new scan 440. The new scan can be Figure 2 The first grayscale map 212 is depicted in the embodiment and the compensated new scan 440 may be Figure 2 An embodiment of the second grayscale map 212 is depicted in FIG.

[0060] Figure 5A and Figure 5B A histogram or graph 50 is depicted. Figure 5A The histogram 50 includes gray levels 500 , 510 , 520 for more than one material, such as an absorber gray level 500 , a black boundary gray level 520 , and a multi-layer gray level 510 . Figure 5B Figure 50 depicts only the absorber gray level 500 and its outliers 500X. In addition to the spatial information of each point, the gray level distribution of each material of the mask 14 can also be determined, which can be helpful in removing or repairing the outliers of the calibration points. This verification method can also be helpful in improving the z-map measurement error check.

[0061] Figure 5A The intensity distribution of all materials is depicted, where each material may have a different gray level 500, 510, 520. In some cases, due to defocus during the measurement of the z-axis height, one or more points or outliers 500X may be identified as incorrect materials. Another reason for incorrect identification may be that a single point may be located at the boundary of two different materials, such that both materials are measured for the single point. After determining the intensity distribution of each material, outliers for each material can be detected relatively easily. Figure 5B An example of an outlier 500X of an absorber gray level 500 is depicted. For an outlier point 500X, the point 500X can be resampled and the correct z-axis height of the point 500X can be determined. Thus, the gray level distribution of each material can serve as the basis for a measurement validation method.

[0062] Figure 6 An example of a point 60 comprising multiple materials 610, 620 is depicted. Figure 6In the example of FIG. 6 , point 60 may include both absorbent material region 610 and adjacent multilayer material region 620. Identifying the material at each location based on z-axis height may be subject to measurement fluctuations, which may lead to false identifications. Figure 5A and Figure 5B As described, some incorrectly identified points can be rejected based on the gray level distribution of each material, but an additional or alternative method to avoid such errors is based on a mask design file (e.g., a GDS file) that contains pattern information for each point on the mask 14. Based on the information in the GDS file, the method 11 can avoid measuring points containing multiple materials, and can avoid erroneous pattern recognition due to measurement errors. For example, for one or more points, such as point 60, each point includes two or more regions associated with different materials, these points can be excluded from the gray level map local compensation operation 106.

[0063] return Figure 1B , act 106 is followed by act 108. In act 108, the mask is inspected. For example, the mask 14 can be inspected in act 108. When the mask 14 has been repeatedly exposed to extreme ultraviolet (EUV) light (e.g., during wafer exposure), the H plasma and EUV light can cause brightness changes at the surface of the mask 14. The brightness changes of the mask 14 caused by the EUV exposure can cause grayscale changes picked up by the detector 100 during inspection. This may result in false detection of (one or more) defects and may cause scan failures and a large number of false positives. In some cases, the scan failure rate of the brightness / darkness (BD) scan can be as high as 2%, which can result in a reduction in tool availability by up to 3.6 hours / week.

[0064] The mask inspection in action 108 may include various operations. Initially, the mask 14 may be loaded into the inspection system 10, which may be in a controlled environment to reduce the risk of contamination. When the mask 14 is loaded into the system 10, the mask 14 may be placed on the platform 12. Before the inspection begins, the mask 14 may be precisely aligned within the system 10. Then, the mask 14 may be illuminated for the purpose of inspection. Various light sources such as the light source 16 may be used for inspection. The mask 14 may be scanned by moving the mask 14 relative to the focused light source 15 via the platform 12, or by moving the light source 16 and the optical device 18 relative to the mask. As the mask 14 is scanned, the light 17 reflected by the mask 14 is collected as data by (one or more) detectors 100. The data may be compared with an expected pattern based on, for example, design data of the mask 14. Any variation between the captured data and the expected pattern may be marked as a potential defect. These defects may include particle contamination, pinholes, scratches, or problems with the pattern itself, such as missing or deformed features. The identified defects can be stored as digital data based on type and size, which can be useful in determining whether the mask 14 is suitable for use or whether to repair or replace the mask 14. As previously described, brightness changes in the mask 14 caused by EUV exposure can cause grayscale changes that are picked up as defects by the detector 100 during inspection. This can result in false detection of defect(s) and can lead to scan failures and a large number of false positives.

[0065] Act 108 is followed by act 112. In act 112, it is determined whether the mask 14 fails the inspection scan. In response to the mask 14 not failing the inspection scan, the mask 14 can be used in another exposure operation 104 to pattern a photoresist layer on a wafer. Before returning to the lithography system, the mask 14 can be cleaned to remove any contaminants that may have landed on the mask 14 during the inspection process 108. In response to the mask 14 failing the inspection scan, the method 11 can continue to perform act 114.

[0066] Action 112 is followed by action 114. In action 114, the weight of the compensation is adjusted. In some embodiments, the weight of the compensation refers to the reference Figure 3 As previously described, the transformation function 310 may be applied to the first grayscale image 300 to generate the second grayscale image 360. In some embodiments, the transformation function 310 may be When the scan of the mask 14 fails in the inspection of act 108, the failure may be due to insufficient compensation of the grayscale of the first grayscale map 300 in act 106. Therefore, the first grayscale map 300 or the second grayscale map 360 may undergo another compensation in act 114. For example, in act 106, before the first execution of act 108, the second grayscale map 360 may be generated based on the compensation factor k as a first value (e.g., 2). After the inspection of act 108 fails, the second grayscale map 360 may be regenerated in act 106 using a new second value (e.g., 3) of the compensation factor k that exceeds the first value. The second value may be selected in act 112. The selection of the second value of the compensation factor k may be via the controller 120 and may be based on an algorithm. For example, the controller 120 may select the second value by increasing the compensation factor k by 1. In another example, the controller 120 may select the second value by increasing the compensation factor k by a certain percentage (such as 5%, 10%, or another suitable percentage). That is, in one example, the first value may be 2, and the second value may be 2.1 or 2.2. Increasing the compensation factor k is beneficial to tightening the histogram of grayscale distribution, which may reduce the occurrence of false positive defect detection due to brightness variation of the mask 14 caused by repeated EUV exposure.

[0067] In some embodiments, the gray levels of the second gray level map 360 are initially generated via a trained neural network. In these embodiments, after the inspection scan fails, the second gray level map 360 can be subjected to the transformation function 310 to generate a third gray level map 360 having a narrower gray level distribution. That is, the weight adjustment of action 114 can be a different type from the initial compensation performed in action 106.

[0068] like Figure 1B As shown, actions 106, 108, 112, 114 may be repeated until the inspection scan in action 108 passes. In some embodiments, adjusting the weight of the compensation in action 114 may be performed based on a threshold value associated with the number of defects detected on the mask 14. For example, it may be expected that brightness variations of the mask 14 may result in a large number of false positive defect detections in a single inspection scan, since brightness variations are essentially a global phenomenon. Thus, when the number of defect detections does not exceed the threshold value, it may be determined to rework, repair, or replace the mask 14 due to the presence of actual defects on the surface of the mask 14. When the number of defect detections does exceed the threshold value, it may be determined to proceed to action 114 to adjust the gray level distribution by increasing the compensation factor k.

[0069] By adjusting the grayscale value distribution before performing mask inspection, compensation for local brightness variations caused by repeated EUV exposure can be achieved, which helps reduce false positives and improves tool availability.

[0070] Figure 7 is a schematic diagram of a controller 70 according to various embodiments. The controller 70 may be a reference Figure 1A The controller 70 may execute the embodiment of the controller 120 described in ... Figures 1A to 6 The controller 70 is described with reference to the system 10 and the method 11 .

[0071] exist Figure 7 , a controller or control system 70 includes a processor 702, a memory 700, a data interface 710, and a network interface 712. To simplify the illustration, some elements of the controller 70 may be omitted from the view. For example, the controller 70 may include a power supply (e.g., a voltage regulator), an analog-to-digital converter (ADC), a digital-to-analog converter (DAC), a pulse width modulation (PWM) controller, a clock and timing circuit, and the like.

[0072] Control system 70 generates output control signals for controlling the operation of one or more components of system 10. Control system 70 may receive input signals from one or more components of system 10.

[0073] The control system 70 includes a processor 702 and a non-transitory computer-readable storage medium 704 encoded with (i.e., storing) computer program code 706 (i.e., a set of executable instructions). The computer-readable storage medium 704 is also encoded with instructions 707 for interfacing with components of the system 10. The processor 702 is electrically coupled to the computer-readable storage medium 704 via a bus 708. The processor 702 is also electrically coupled to an I / O interface 710 via the bus 708. A network interface 712 is also electrically connected to the processor 702 via the bus 708. The network interface 712 is connected to a network 714, enabling the processor 702 and the computer-readable storage medium 704 to connect to external elements via the network 714. The processor 702 is configured to execute the computer program code 706 encoded in the computer-readable storage medium 704 so that the control system 70 can be used to perform a portion or all of the operations as described with respect to the system 10.

[0074] In some embodiments, processor 702 is a central processing unit (CPU), a multiprocessor, a distributed processing system, an application specific integrated circuit (ASIC), and / or a suitable processing unit.

[0075] In some embodiments, the computer-readable storage medium 704 is an electrical, magnetic, optical, electromagnetic, infrared, and / or semiconductor system (or device or apparatus). For example, the computer-readable storage medium 704 includes a semiconductor or solid-state memory, a magnetic tape, a removable computer floppy disk, a random access memory (RAM), a read-only memory (ROM), a hard disk, and / or an optical disk. In some embodiments using optical disks, the computer-readable storage medium 704 includes a compact disk read-only memory (CD-ROM), a compact disk read / write (CD-R / W), and / or a digital video disk (DVD).

[0076] In some embodiments, the storage medium 704 stores computer program code 706 configured to cause the control system 70 to perform operations as described with respect to the system 10 and method 11. In some embodiments, the storage medium 704 also stores information required to perform operations as described with respect to the system 10 and method 11, such as a set of executable instructions for performing operations as described with respect to the system 10 and method 11. In some embodiments, the storage medium 704 stores computer program code 706 configured to cause the control system 70 to perform operations as described with respect to the system 10 and method 11. Figure 4 A trained neural network 430 is described.

[0077] In some embodiments, storage medium 704 stores instructions 707 for interfacing with system 10. Instructions 707 enable processor 702 to generate operational instructions readable by elements of system 10 to effectively implement operations as described with respect to system 10 and method 11.

[0078] The control system 70 includes an I / O interface 710. The I / O interface 710 is coupled to external circuits. In some embodiments, the I / O interface 710 includes a keyboard, a keypad, a mouse, a trackball, a trackpad, and / or cursor direction keys for transmitting information and commands to the processor 702.

[0079] The control system 70 also includes a network interface 712 coupled to the processor 702. The network interface 712 allows the control system 70 to communicate with a network 714 to which one or more other computer systems are connected. The network interface 712 includes a wireless network interface, such as BLUETOOTH, WIFI, WIMAX, GPRS, or WCDMA; or a wired network interface, such as ETHERNET, USB, or IEEE-1394.

[0080] The control system 70 may be configured to receive information related to the grayscale map and / or distribution and the GDS file via the I / O interface 710. The information is transmitted to the processor 702 via the bus 708 and then stored in the computer-readable medium 704 as a parameter 716. The control system 70 may be configured to receive information related to the threshold via the I / O interface 710. In some embodiments, the threshold is received from an operator. The information is stored in the computer-readable medium 704 as a threshold parameter 718. For example, the threshold parameter 718 may be associated with the number of defects detected in a single scan. In one example, the number of defects detected in action 108 may be compared to the threshold parameter 718, and a determination may be made as to whether to adjust one or more parameters of the grayscale map 300.

[0081] During operation, in some embodiments, the processor 702 executes a set of instructions to determine whether the number of detected defects has exceeded a threshold or is below a threshold. Based on the above determination, the processor 702 generates a control signal to instruct one or more components of the system 10 to adjust its parameters. In some embodiments, the control signal is sent using the I / O interface 710. In some embodiments, the control signal is sent using the network interface 712. In some embodiments, the control signal is used internally by the processor 702 to trigger the execution of the action 114 to adjust the compensation weight of the first grayscale map 300 or the second grayscale map 360.

[0082] Embodiments may provide advantages. By adjusting the weights of the compensation of the grayscale map or its distribution, false positive defect detections due to brightness variations in a mask repeatedly exposed to EUV light may be reduced. Reducing the number of false positive defect detections may save downtime associated with repairing and / or replacing masks, thereby improving tool availability.

[0083] According to at least one embodiment, a method includes: generating a first image by scanning a mask; after the generating of the first image, exposing the mask in a photolithography operation; after the exposing, generating a second image by scanning the mask; generating a compensated third image by performing grayscale image local compensation on the second image, the grayscale image local compensation being performed via a transformation function; generating a comparison result by comparing the compensated third image with the first image; and determining whether there is a defect on the mask based on the comparison result.

[0084] According to at least one embodiment, a method includes: generating a reference image by scanning a mask; after generating the reference image, generating a second image by scanning the mask; generating a third image by performing grayscale image local compensation on the second image, wherein the grayscale image local compensation is performed via a trained neural network; generating a comparison result by comparing the third image with the reference image; and determining whether there is a defect on the mask based on the comparison result.

[0085] According to at least one embodiment, a system includes: a light source; a mask stage operable to position a mask in an optical path of light generated by the light source; a detector operable to generate a second image by scanning a mask on the mask stage, the mask having been exposed to extreme ultraviolet (EUV) light in a lithography system prior to the scanning; and a controller including: a memory storing computer instructions; and a processor configured to execute the computer instructions to perform the following operations: generate a compensated third image by performing local compensation of a grayscale map on the second image; generate a comparison result by comparing the compensated third image with a reference image of the mask; and determine whether there is a defect on the mask based on the comparison result.

[0086] The features of several embodiments are summarized above so that those skilled in the art can better understand the various aspects of the present disclosure. Those skilled in the art should understand that they can easily use the present disclosure as a basis for designing or modifying other processes and structures to achieve the same purpose of the embodiments introduced herein and / or achieve the same advantages of the embodiments introduced herein. Those skilled in the art should also recognize that such equivalent constructions do not depart from the spirit and scope of the present disclosure, and they can make various changes, substitutions and modifications herein without departing from the spirit and scope of the present disclosure.

[0087] Example

[0088] Example 1. A method for a semiconductor device, comprising: generating a first image by scanning a mask; after generating the first image, exposing the mask in a photolithography operation; after the exposure, generating a second image by scanning the mask; generating a compensated third image by performing a first grayscale image local compensation on the second image, wherein the first grayscale image local compensation is performed via a transformation function; generating a comparison result by comparing the compensated third image with the first image; and determining whether there is a defect on the mask based on the comparison result.

[0089] Example 2. The method according to Example 1, wherein the generating of the compensated third image is performed via the transformation function, wherein the transformation function is GL is the gray level at a certain position in the second image, GL mean is the average value of multiple gray levels at multiple positions in the second image, GL std is the standard deviation of the multiple gray levels, GL new is the compensated gray level and k is a compensation factor, which is at least 1.

[0090] Example 3. The method of Example 2, wherein the compensation factor exceeds about 2.

[0091] Example 4. The method according to Example 1 further includes: in response to determining that there is a defect on the mask: generating a compensated fourth image by performing a second grayscale map local compensation on the second image, the second grayscale map local compensation being different from the first grayscale map local compensation.

[0092] Example 5. A method according to Example 4, wherein a second value of the compensation factor for local compensation of the second grayscale image is greater than a first value of the compensation factor for local compensation of the first grayscale image.

[0093] Example 6. The method of Example 5, wherein the second value is greater than the first value by a selected percentage.

[0094] Example 7. The method of Example 1, wherein generating the compensated third image comprises generating an intensity distribution associated with a material of the mask.

[0095] Example 8. The method of Example 7, wherein generating the compensated third image comprises determining outliers of the intensity distribution and resampling points in the second image associated with the outliers.

[0096] Example 9. The method of Example 7, wherein generating a compensated third image includes determining whether points of the second image are associated with a number of materials greater than two.

[0097] Example 10. The method of Example 9, wherein generating the compensated third image comprises: in response to the number exceeding two, removing grayscale information associated with the point from the compensated third image.

[0098] Example 11. A method for a semiconductor device, comprising: generating a reference image by scanning a mask; after generating the reference image, generating a second image by scanning the mask; generating a third image by performing grayscale image local compensation on the second image, wherein the grayscale image local compensation is performed via a trained neural network; generating a comparison result by comparing the third image with the reference image; and determining whether there is a defect on the mask based on the comparison result.

[0099] Example 12. The method of Example 11, further comprising generating the trained neural network based on multiple historical scans of the mask.

[0100] Example 13. The method of Example 12, wherein the generating the trained neural network comprises training a neural network based on the plurality of historical scans of the mask and a second plurality of historical scans associated with a plurality of second masks other than the mask.

[0101] Example 14. The method according to Example 11 further includes: in response to determining that there is a defect on the mask: generating a compensated fourth image by performing local compensation of a second grayscale map on the second image, wherein the local compensation of the second grayscale map is performed via a transformation function.

[0102] Example 15. A system for a semiconductor device, comprising: a light source; a mask stage operable to position a mask in an optical path of light generated by the light source; a detector operable to generate a second image by scanning the mask on the mask stage, the mask having been exposed to extreme ultraviolet (EUV) light in a lithography system prior to the scanning; and a controller comprising: a memory storing computer instructions; and a processor configured to execute the computer instructions to perform the following operations: generating a compensated third image by performing local compensation of a grayscale image on the second image; generating a comparison result by comparing the compensated third image with a reference image of the mask; and determining whether there is a defect on the mask based on the comparison result.

[0103] Example 16. A system according to Example 15, wherein the processor, when executing the computer instructions to generate the compensated third image, executes the computer instructions to generate the compensated third image via a transformation function.

[0104] Example 17. The system of Example 16, wherein the transformation function is GL is the gray level at a certain position in the second image, GL mean is the average value of multiple gray levels at multiple positions in the second image, GLstd is the standard deviation of the multiple gray levels, GL new is the compensated gray level and k is a compensation factor, which is a first value when generating the compensated third image.

[0105] Example 18. A system according to Example 17, wherein the processor is further configured to execute the computer instructions to perform the following operations: in response to determining that there is a defect on the mask, generating a compensated fourth image by performing local compensation of a second grayscale image on the second image, the local compensation of the second grayscale image is performed via the transformation function, and the compensation factor when generating the compensated fourth image is a second value exceeding the first value.

[0106] Example 19. A system according to Example 15, wherein the processor, when executing the computer instructions to generate the compensated third image, executes the computer instructions to generate the compensated third image via a trained neural network.

[0107] Example 20. A system according to Example 15, wherein the processor, when executing the computer instructions to generate the compensated third image, executes the computer instructions to generate the compensated third image by narrowing the histogram associated with the second image.

Claims

1. A method for a semiconductor device, comprising: generating a first image by scanning the mask; After said generating the first image, exposing said mask in a photolithography operation; After the exposing, generating a second image by scanning the mask; generating a compensated third image by performing a first grayscale map local compensation on the second image, the first grayscale map local compensation being performed via a transformation function; generating a comparison result by comparing the compensated third image with the first image; as well as Based on the comparison result, it is determined whether the mask has a defect.

2. The method according to claim 1, wherein: The generating of the compensated third image is performed via the transformation function, which is GL is the gray level at a certain position in the second image, GL mean is the average value of multiple gray levels at multiple positions in the second image, GL std is the standard deviation of the multiple gray levels, GL new is the compensated gray level and k is a compensation factor, which is at least 1.

3. The method according to claim 2, wherein: The compensation factor exceeds about 2.

4. The method according to claim 1, further comprising: In response to determining that the mask has a defect: A compensated fourth image is generated by performing a second grayscale map local compensation on the second image, the second grayscale map local compensation being different from the first grayscale map local compensation.

5. The method according to claim 4, wherein: A second value of the compensation factor for local compensation of the second grayscale image is greater than a first value of the compensation factor for local compensation of the first grayscale image.

6. The method according to claim 5, wherein: The second value is greater than the first value by a selected percentage.

7. The method according to claim 1, wherein: The generating a compensated third image includes generating an intensity distribution associated with a material of the mask.

8. The method according to claim 7, wherein: The generating a compensated third image includes determining outliers of the intensity distribution and resampling points in the second image associated with the outliers.

9. A method for a semiconductor device, comprising: Generate a reference image by scanning the mask; After said generating the reference image, generating a second image by scanning said mask; generating a third image by performing grayscale map local compensation on the second image, the grayscale map local compensation being performed via a trained neural network; generating a comparison result by comparing the third image with the reference image; as well as Based on the comparison result, it is determined whether the mask has a defect.

10. A system for a semiconductor device, comprising: light source; a mask stage operable to position a mask in the path of light generated by the light source; a detector operable to generate a second image by scanning a mask on the mask stage, the mask having been exposed to extreme ultraviolet (EUV) light in a lithography system prior to said scanning the mask; as well as Controller, including: memory for storing computer instructions; and A processor is configured to execute the computer instructions to perform the following operations: Generate a compensated third image by performing local grayscale compensation on the second image; generating a comparison result by comparing the compensated third image with a reference image of the mask; and Based on the comparison result, it is determined whether the mask has a defect.

Images