Photomask 3D defect inspection device and method
By establishing a three-dimensional model of the photomask through a 3D defect inspection device and combining EUV or X-ray scanning with holographic image comparison, the problems of low detection efficiency and many false defects in the existing technology are solved, and high-precision and fast photomask inspection is achieved.
Patent Information
- Application Number
- CN202511021529.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-24
- Publication Date
- 2025-09-23
- Estimated Expiration
- Not applicable · inactive patent
Smart Images
Figure CN120689343A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of semiconductor manufacturing, and in particular to a device and method for inspecting 3D defects of a photomask. Background Art
[0002] During the integrated circuit (IC) manufacturing process, patterns are formed on semiconductors using photolithography and etching techniques. To replicate these patterns on wafers, they must be presented through a photomask for lithography wafer tapeout. Therefore, to ensure defect-free wafers and high yields, effective defect control inspections must be performed at the photomask end to ensure that each "master" used for photolithography achieves zero defects within specifications. Currently, the mask industry generally utilizes two mainstream methods for pattern defect inspection: DB and DD. DB defect inspection predominates for complex patterns and high-end OPC products, but the two methods are sometimes used interchangeably and complementary during the mask processing process to improve defect inspection accuracy and capture rates. However, both of these defect inspection methods compare the plane of the mask image and are unable to effectively and accurately identify defects perpendicular to the grooves in the image.
[0003] Existing inspection technologies still have the following technical problems: 1. The structure of photomask patterns is complex and varied, and the line width and groove processing precision of the patterns are becoming increasingly fine. The 28nm, 14nm, 10nm, and 5nm processes will continue to challenge the limits of Arf wavelength inspection equipment. Second, the two-dimensional image comparison method for photomask pattern calibration cannot fully calibrate factors such as the photomask pattern etching depth. As a result, some photomask etching may have defects such as uneven groove depth. These defects will cause uneven substrate transmittance, thereby affecting the scattering and refractive index ratio of the wafer photolithography pattern, and thus affecting the photolithography accuracy. 3. As processing technology becomes increasingly sophisticated and precision continues to increase, the complexity of image correction and comparison algorithms is also increasing, and the inspection process is time-consuming. For some products below 14nm, the inspection efficiency using the most advanced US KLA6 series inspection equipment is an average of 8 hours, resulting in low product output efficiency. 4. At present, defect inspection images are all two-dimensional images. When the photos captured by the laser generator are compared with the design image data, it is easy to produce errors due to the contrast difference between the actual photos and the image design data, resulting in an increase in "false defects" and missed defects, affecting the quality of the finished mask. Summary of the Invention
[0004] The present invention aims to provide a device and method for inspecting 3D defects of photomasks to solve the problems existing in the prior art.
[0005] The specific technical solutions are as follows: A photomask 3D defect inspection device, comprising: A modeling unit (1), for establishing a 3D model of a photomask product; A scanning unit (2) performs 3D scanning on a physical object in a closed chamber environment; The comparison unit (3) is used to compare the results of the 3D model and the 3D object scan, and accurately mark the abnormal position through comparison.
[0006] Furthermore, the modeling unit (1) comprises: A data input module (11) is used to input product image data; AI algorithm processing module (12), used for processing input product image data; The modeling software module (13) is used to perform modeling based on the processed data.
[0007] Furthermore, the modeling unit (1) models the photomask product image data to form a 3D model through an AI algorithm and modeling software; The specific algorithm is as follows: The input image data is ,in and Represents the horizontal and vertical coordinates of the image respectively.
[0008] The convolutional layer of a convolutional neural network can be expressed as:
[0009] in, is the output of the convolutional layer, is the weight of the convolution kernel, is the bias term, Is the activation function, using the ReLU function: Through multiple layers of convolution and pooling operations, the feature representation of the image is extracted. Finally, the features are mapped to the parameter space of the 3D model through a fully connected layer to generate a 3D model.
[0010] Furthermore, the scanning unit (2) comprises: A film loading platform (21) for conveying products to be inspected; a workpiece table (22) for placing the product to be inspected; The scanning device (23) is used to perform 3D scanning of a physical object using EUV or X-rays.
[0011] Furthermore, the comparison unit (3) includes: A holographic image generation module (31) is used to generate a holographic image from the results of scanning the 3D model and the 3D object; The comparison and analysis module (32) is used to compare and overlap the two sets of holographic images and accurately identify the abnormal position through comparison.
[0012] The specific algorithm is as follows: The holographic image of the 3D model is , the holographic image after 3D object scanning is .
[0013] The feature points of the two sets of holographic images are extracted using the Scale Invariant Feature Transform (SIFT) algorithm or the Speeded Up Robust Features (SURF) algorithm. Feature point matching is expressed as:
[0014] in, and They are the feature points in the 3D model and the physical scanned holographic image, is the Euclidean distance between feature points. By minimizing , find the best matching feature point pair.
[0015] Then, the anomaly location is determined by calculating the difference between the matching feature point pairs. The anomaly location can be calculated using the following formula:
[0016] in, Indicates the intensity of the abnormal location. By setting the threshold , the abnormal location can be determined:
[0017] A method for inspecting 3D defects of a photomask is provided, which is implemented based on the above-mentioned 3D defect inspection device of the photomask and is characterized by comprising the following steps: Modeling the photomask product image data to form a 3D model through a modeling unit (1); The product to be inspected is transferred to the workpiece table (22) via the film loading table (21); Perform 3D scanning of physical objects in a closed chamber environment; Use holographic imaging to overlap and compare the 3D model and the 3D object scan results; The abnormal position can be accurately identified through comparison.
[0018] In step 1, the photomask product image data is modeled into a 3D model using AI algorithms and modeling software. The specific algorithm is as follows: The input image data is ,in and Represents the horizontal and vertical coordinates of the image respectively.
[0019] The convolutional layer of a convolutional neural network can be expressed as:
[0020] in, is the output of the convolutional layer, is the weight of the convolution kernel, is the bias term, Is the activation function, using the ReLU function:
[0021] Through multiple layers of convolution and pooling operations, the feature representation of the image is extracted. Finally, the features are mapped to the parameter space of the 3D model through a fully connected layer to generate a 3D model.
[0022] In step 3, EUV or X-ray is used for scanning.
[0023] In step 4, the 3D model and the 3D object scanned are overlapped and compared by means of holographic imaging. The specific algorithm is as follows: The holographic image of the 3D model is , the holographic image after 3D object scanning is .
[0024] The feature points of the two sets of holographic images are extracted using the Scale Invariant Feature Transform (SIFT) algorithm or the Speeded Up Robust Features (SURF) algorithm. Feature point matching is expressed as:
[0025] in, and They are the feature points in the 3D model and the physical scanned holographic image, is the Euclidean distance between feature points. By minimizing , find the best matching feature point pair.
[0026] Then, the anomaly location is determined by calculating the difference between the matching feature point pairs. The anomaly location can be calculated using the following formula:
[0027] in, Indicates the intensity of the abnormal location. By setting the threshold , the abnormal location can be determined:
[0028] In step 5, the abnormal position is accurately calibrated through comparison.
[0029] Compared with the prior art, the present invention has the following beneficial effects: The present invention uses X-rays and EUV light with shorter wavelengths, which can meet the limits of the Arf light source of current defect inspection equipment, meaning that it can meet the defect inspection work of higher-level products (below 5nm).
[0030] The inspection process of the present invention uses the method of overall holographic image 3D comparison to detect vertical defects at the bottom of the graphic line width groove that cannot be discovered by the existing technology. This breaks the limitation of the existing technology that can only detect planar defects in the horizontal direction, and is more conducive to ensuring the processing quality of the mask product.
[0031] The inspection process of the present invention adopts the method of overall holographic image 3D comparison, which eliminates the need for graphic comparison by moving the product on the machine workpiece table and light source step scanning. It can greatly reduce the production time of the graphic defect inspection process, improve output efficiency, and shorten the inspection time to less than 1 hour.
[0032] Since the defect capture method is three-dimensional, it can ensure the accurate location of the defect generated intuitively, improve the reliability and accuracy of defect capture, reduce the probability of "false defects" to a certain extent, and do not require more human intervention to judge. It can improve processing efficiency to a certain extent and reduce the dependence on the experience of engineers. BRIEF DESCRIPTION OF THE DRAWINGS
[0033] Figure 1 It is a structural schematic diagram of the photomask 3D defect inspection device of the present invention.
[0034] Figure 2 Flowchart of the photomask 3D defect inspection method of the present invention. DETAILED DESCRIPTION
[0035] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.
[0036] like Figure 1 As shown, the photomask 3D defect inspection device includes a modeling unit 1, a scanning unit 2 and a comparison unit 3.
[0037] Modeling Unit 1: Located on the left side of the device, it includes a data input module 11, an AI algorithm processing module 12, and a modeling software module 13. The data input module 11 is used to receive product image data, the AI algorithm processing module 12 processes the data, and the modeling software module 13 creates a model based on the processed data.
[0038] Scanning unit 2: Located in the middle of the device, it includes a loading table 21, a workpiece table 22, and a scanning device 23. The loading table 21 is used to transport the photomask product to be inspected, the workpiece table 22 is used to place the product to be inspected, and the scanning device 23 uses EUV or X-rays to perform 3D scanning of the physical object.
[0039] Comparison unit 3: Located on the right side of the device, it includes a holographic image generation module 31 and a comparison and analysis module 32. Holographic image generation module 31 generates a holographic image from the scanned 3D model and the 3D object. Comparison and analysis module 32 compares the two sets of holographic images and accurately locates the abnormal location through comparison.
[0040] Modeling unit 1, used to build a 3D model of the photomask product; Scanning unit 2, performs 3D scanning of the object in a closed chamber environment; The comparison unit 3 is used to compare the scanning results of the 3D model and the 3D object, and accurately mark the abnormal position through comparison.
[0041] Furthermore, the modeling unit 1 includes: Data input module 11, used for inputting product image data; AI algorithm processing module 12, used to process input product image data; The modeling software module 13 is used to perform modeling based on the processed data.
[0042] Furthermore, the modeling unit 1 models the photomask product image data into a 3D model using an AI algorithm and modeling software; The specific algorithm is as follows: The input image data is ,in and Represents the horizontal and vertical coordinates of the image respectively.
[0043] The convolutional layer of a convolutional neural network is represented as:
[0044] in, is the output of the convolutional layer, is the weight of the convolution kernel, is the bias term, Is the activation function, using the ReLU function:
[0045] Through multiple layers of convolution and pooling operations, the feature representation of the image is extracted. Finally, the features are mapped to the parameter space of the 3D model through a fully connected layer to generate a 3D model.
[0046] Furthermore, the scanning unit (2) comprises: A film loading platform (21) for conveying products to be inspected; a workpiece table (22) for placing the product to be inspected; The scanning device (23) is used for performing 3D scanning of a physical object using EUV or X-rays.
[0047] Furthermore, the comparison unit (3) includes: A holographic image generation module (31) is used to generate a holographic image from the results of scanning the 3D model and the 3D object; The comparison and analysis module (32) is used to compare and overlap the two sets of holographic images and accurately identify the abnormal position through comparison.
[0048] The specific algorithm is as follows: The holographic image of the 3D model is , the holographic image after 3D object scanning is .
[0049] The feature points of the two sets of holographic images are extracted using the Scale Invariant Feature Transform (SIFT) algorithm or the Speeded Up Robust Features (SURF) algorithm. Feature point matching is expressed as:
[0050] in, and They are the feature points in the 3D model and the physical scanned holographic image, is the Euclidean distance between feature points. By minimizing , find the best matching feature point pair.
[0051] Then, the anomaly location is determined by calculating the difference between the matching feature point pairs. The anomaly location can be calculated using the following formula:
[0052] in, Indicates the intensity of the abnormal location. By setting the threshold , the abnormal location can be determined:
[0053] Figure 2 The specific steps of the photomask 3D defect inspection method are demonstrated.
[0054] Step 1: Modeling The product image data is modeled into a 3D model by the modeling unit 1. The specific process is as follows: The data input module 11 receives the photomask product image data, which may be the design drawings of the photomask or other related image data.
[0055] The AI algorithm processing module 12 processes the input data and uses a convolutional neural network (CNN) to extract and process the image data. The specific algorithm is as follows: The input image data is ,in and Represents the horizontal and vertical coordinates of the image respectively.
[0056] The convolutional layer of a convolutional neural network is represented as:
[0057] in, is the output of the convolutional layer, is the weight of the convolution kernel, is the bias term, Is the activation function, using the ReLU function:
[0058] Through multiple layers of convolution and pooling operations, the feature representation of the image is extracted. Finally, the features are mapped to the parameter space of the 3D model through a fully connected layer to generate a 3D model.
[0059] The modeling software module 13 performs modeling based on the processed data to generate a 3D model of the photomask.
[0060] Step 2: Send the product to be inspected The photomask product to be inspected is transferred to the workpiece stage 22 via the loading stage 21. The loading stage 21 is responsible for transferring the product from the storage location to the workpiece stage 22, ensuring the stability and accuracy of the product during the transfer process.
[0061] Step 3: 3D Scanning Perform 3D scanning of objects in a closed chamber environment. The specific process is as follows: The product to be inspected is placed on the workpiece table 22 .
[0062] The scanning device 23 is activated to perform a 3D scan of the object using EUV or X-rays. EUV or X-rays have shorter wavelengths and can provide higher resolution, making them suitable for detecting subtle defects in photomasks.
[0063] The scanning process is carried out in a closed chamber environment to ensure the accuracy and stability of the scanning. The scanning device 23 generates 3D object scanning data based on the scanning results.
[0064] Step 4: Holographic image generation and overlay comparison The 3D model and the 3D object scan results are overlapped and compared through holographic imaging. The specific process is as follows: The holographic image generation module 31 generates a holographic image from the scanned 3D model and the 3D object. The holographic image can provide a more intuitive three-dimensional view, which is convenient for subsequent comparison and analysis.
[0065] The comparison and analysis module 32 compares the two sets of holographic images. The specific algorithm is as follows: The holographic image of the 3D model is , the holographic image after 3D object scanning is .
[0066] First, extract the feature points of the two sets of holographic images using the Scale Invariant Feature Transform (SIFT) algorithm or the Speeded Up Robust Features (SURF) algorithm. Feature point matching is expressed as:
[0067] in, and They are the feature points in the 3D model and the physical scanned holographic image, is the Euclidean distance between feature points. By minimizing , find the best matching feature point pair.
[0068] Then, the anomaly location is determined by calculating the difference between the matching feature point pairs. The anomaly location can be calculated using the following formula:
[0069] in, Indicates the intensity of the abnormal location. By setting the threshold , the abnormal location can be determined:
[0070] Step 5: Abnormal location calibration The abnormal position is accurately calibrated through comparison. The specific process is as follows: The comparison and analysis module 32 marks the abnormal position in the holographic image according to the abnormal position calculated by the above algorithm.
[0071] The marked holographic image can intuitively show the defect location of the photomask, facilitating subsequent repair and improvement.
[0072] Finally, it should be noted that the above preferred embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail through the above preferred embodiments, those skilled in the art should understand that various changes, combinations and deletions can be made in form and details without departing from the patent scope defined by the claims of the present invention.
Claims
1. A photomask 3D defect inspection device, characterized in that: include: A modeling unit (1), for establishing a 3D model of a photomask product; A scanning unit (2) performs 3D scanning of an object in a closed chamber environment; The comparison unit (3) is used to compare the results of the 3D model and the 3D object scan, and accurately identify the abnormal position through comparison; The modeling unit (1) models the photomask product image data into a 3D model using an AI algorithm and modeling software; The scanning unit (2) comprises: A film loading platform (21) for conveying products to be inspected; a workpiece table (22) for placing the product to be inspected; The scanning device (23) is used for performing 3D scanning of a physical object using EUV or X-rays.
2. The photomask 3D defect inspection device according to claim 1, wherein: The modeling unit (1) comprises: A data input module (11) is used to input product image data; AI algorithm processing module (12), used for processing input product image data; The modeling software module (13) is used to perform modeling based on the processed data.
3. The photomask 3D defect inspection device according to claim 2, wherein the modeling unit (1) models the photomask product image data into a 3D model using an AI algorithm and modeling software; the modeling unit (1) comprises the following steps: The input image data is ,in and Represent the horizontal and vertical coordinates of the image respectively; The convolutional layer of a convolutional neural network can be expressed as: ; in, is the output of the convolutional layer, is the weight of the convolution kernel, is the bias term, Is the activation function, using the ReLU function: ; Through multi-layer convolution and pooling operations, the feature representation of the image is extracted, and the features are mapped to the parameter space of the 3D model through the fully connected layer to generate a 3D model.
4. The photomask 3D defect inspection device according to claim 1, wherein: The comparison unit (3) comprises: A holographic image generation module (31) is used to generate a holographic image from the results of scanning the 3D model and the 3D object; The comparison and analysis module (32) is used to compare and overlap the two sets of holographic images and accurately identify the abnormal position through comparison.
5. The photomask 3D defect inspection device according to claim 4, wherein: The anomaly location is determined by calculating the difference between the matching feature point pairs in the 3D model and the scanned holographic image of the physical object.
6. A method for inspecting 3D defects of a photomask based on the device according to any one of claims 1 to 5, characterized in that: The following steps are involved: 1) Modeling: Modeling the photomask product image data into a 3D model through the modeling unit (1); 2) Transporting the product to be inspected: transporting the product to be inspected to the workpiece table (22) via the film loading table (21); 3) 3D scanning: 3D scanning of physical objects in a closed chamber environment; 4) Holographic image generation and overlay comparison: Overlay and compare the 3D model and the 3D object scan results using holographic images; 5) Abnormal position calibration: Accurately calibrate the abnormal position through comparison.
7. The photomask 3D defect inspection method according to claim 6, wherein: In step 1, the photomask product image data is modeled into a 3D model using an AI algorithm and modeling software.
8. The photomask 3D defect inspection method according to claim 6, wherein: In step 3, EUV or X-ray is used for scanning.
9. The photomask 3D defect inspection method according to claim 6, wherein: In step 4, the scanning results of the 3D model and the 3D object are overlapped and compared by means of holographic imaging.
10. A computer-readable medium having a program stored thereon, characterized in that: When the program is executed, at least part of the steps in the photomask 3D defect inspection method according to any one of claims 6 to 9 are performed.