A Method for Measuring the Overlay Error of Wafer Based on Halcon

Abstract

The present invention discloses a method for measuring wafer overlay error based on Halcon, belonging to the technical field of computer graphics. Based on the operator functions provided by the Halcon software, this method first extracts the images of different regions of the overlay error markers from the wafer image and generates corresponding templates based on this image. For the wafer image to be measured for overlay error, the overlay error marker template is used to find the corresponding overlay error marker region, realizing sub-pixel level overlay error measurement, and thus calculating the offset between different layers. This method significantly improves the measurement accuracy and efficiency, and can adapt to a variety of complex chip scenarios. Through step-by-step template search, the error interference and calculation cost in the measurement process are effectively reduced. This method performs excellently in a variety of chip environments, and the measurement time for a single wafer image is less than one second, providing an efficient and reliable solution for overlay error measurement, and having broad industrial application prospects.

Description

Technical Field

[0001] The present invention belongs to the technical field of computer graphics, relates to image-based error measurement, and particularly relates to a method for measuring wafer overlay error based on Halcon. Background Art

[0002] Overlay error is a key indicator of the alignment accuracy between lithography layers in the wafer manufacturing process. Its importance lies in directly affecting the chip's function and performance. If the error is too large, it may lead to circuit short circuits or open circuits, seriously affecting the reliability of the device. At the same time, as the process advances to advanced nodes, the demand for precise control of overlay error further increases. Any tiny error will amplify the negative impact on the yield, thereby increasing production costs and extending the development cycle. Therefore, effectively controlling overlay error is crucial for improving the yield of semiconductor manufacturing, ensuring high chip performance, and reducing production costs.

[0003] The measurement of overlay error is usually achieved through overlay marks, which are specific patterns designed on the wafer to assist in detecting the alignment deviation between different lithography layers. Overlay marks are usually arranged in the non-active area of the wafer to avoid interfering with the device function. During the measurement process, the detection device captures the mark pattern through high-precision optical imaging technology, analyzes its displacement and rotation deviation in the X and Y directions, and thus calculates the alignment error between layers.

[0004] Although neural networks have demonstrated powerful learning capabilities in some fields, there are many instability problems in industrial production. First, neural networks rely on large-scale labeled data for training, but it is costly to comprehensively obtain labeled data in industrial production. Second, its measurement results are greatly affected by model parameters, data distribution, and external environment, and it is difficult to ensure stability. In addition, the "black box" characteristic of neural networks makes the results difficult to interpret and has limited adaptability to new scenarios, which may lead to uncontrollability of measurement errors in production. The algorithm based on computer graphics uses a deterministic model, and the measurement results have strong repeatability and a controllable error range. At the same time, this method can achieve sub-pixel-level measurement accuracy, meeting the requirements of real-time and efficiency in industrial scenarios. In addition, the logic of computer graphics technology is clear and the results are interpretable, facilitating the analysis and tracing of problems, and it is a reliable choice for large-scale industrial applications. In order to ensure the reliability and consistency of wafer overlay error measurement results, a measurement method based on computer graphics needs to be designed. Summary of the Invention

[0005] Aiming at the deficiencies of the existing technology, the present invention proposes a method for measuring the overlay error of wafers based on Halcon. By using the functions of the operators provided by the Halcon software, different regions of the overlay error marks are framed and a model is established, realizing the measurement of the overlay error of wafers in computer graphics, and improving the efficiency and stability of the measurement.

[0006] A method for measuring the overlay error of wafers based on Halcon specifically includes the following steps:

[0007] Step 1: Select a wafer image, take the top left vertex of the image as the origin O, the horizontal direction as the X-axis, and the vertical direction as the Y-axis to establish a two-dimensional rectangular coordinate system. Select an overlay error mark from this image and label the rectangular frames 1, 2, and 3. Among them, only one complete overlay error mark is contained inside the rectangular frame 1, only the inner rectangular mark of the overlay error mark is contained inside the rectangular frame 2, and the rectangular frame 3 is inside the outer contour line of the inner rectangular mark of the overlay error mark. Record the position information of the three rectangular frames in the two-dimensional rectangular coordinate system XOY respectively.

[0008] Step 2: Calculate the center point coordinates of the rectangular frames 1, 2, and 3, and then calculate the regions between the rectangular frames 1, 2 and the rectangular frames 2, 3 respectively to obtain the outer rectangular mark frame region and the inner rectangular mark frame region. Use the outer rectangular mark frame region and the inner rectangular mark frame region to crop the enhanced wafer image respectively to obtain the enhanced outer rectangular mark frame image and the enhanced inner rectangular mark frame image.

[0009] Use the create_shape_model operator of the Halcon software to create an outer rectangular mark frame template and an inner rectangular mark frame template from the enhanced outer rectangular mark frame image and the enhanced inner rectangular mark frame image respectively.

[0010] Step 3: For the wafer image for which the overlay error needs to be measured, first perform image enhancement on it and establish a two-dimensional rectangular coordinate system X'O'Y' according to the method in Step 1. Then use the outer rectangular mark frame template obtained in Step 2 to find all the outer rectangular mark frames that match the template from the enhanced wafer image, and obtain the abscissa, ordinate and rotation angle relative to the outer rectangular mark frame template of the outer rectangular mark frame matching result.

[0011] Step 4: According to the position information of the rectangular frame 1 and the outer rectangular mark frame matching result, generate a 2D position transformation matrix 2D1 from the rectangular frame 1 to the outer rectangular mark frame, and use the position transformation matrix 2D1 to generate a rectangular frame on the enhanced wafer image that has the same shape as the rectangular frame 1 and contains the corresponding outer rectangular mark frame matching result inside. .

[0012] Step 5: Use the rectangular frame Crop the enhanced wafer image. On the cropped image, use the inner rectangle marking frame template obtained in step 2 to find the inner rectangle marking frame that matches the template, and obtain the abscissa, ordinate of the matching result of the inner rectangle marking frame, and the rotation angle relative to the inner rectangle marking frame template, then proceed to step 6. If there is no inner rectangle marking frame that matches the template, then proceed to step 9.

[0013] Step 6: According to the position information of rectangle 2 and the matching result of the inner rectangle marking frame, generate the position transformation matrix 2D2 from rectangle 2 to the inner rectangle marking frame. Use the position transformation matrix 2D2 to generate a rectangle on the enhanced wafer image that has the same shape as rectangle 2 and contains the corresponding matching result of the inner rectangle marking frame inside. Calculate the area between rectangle and , and perform Gaussian line segment detection to obtain the outer rectangle marking frame contour line.

[0014] Step 7: According to the position information of rectangle 3 and the position transformation matrix 2D2, generate a rectangle on the enhanced wafer image that has the same shape as rectangle 3 and is located inside the matching result of the inner rectangle marking frame. Calculate the area between rectangle and , and perform Gaussian line segment detection to obtain the inner rectangle marking frame contour line.

[0015] Step 8: Perform minimum rectangle fitting on the outer rectangle marking frame contour line and the inner rectangle marking frame contour line respectively to obtain the outer rectangle marking prediction result and the inner rectangle marking prediction result. Use the area_center operator of the Halcon software to obtain the center coordinates of the inner and outer rectangle marking prediction results, and then calculate the deviation results of the center coordinates of the inner and outer rectangle marking prediction results in the X' axis direction and the Y' axis direction.

[0016] Step 9: Follow the methods in steps 4 - 8 to sequentially traverse each outer rectangle marking frame that matches the template, and output the error measurement results of different overlay error markings on this image.

[0017] The present invention has the following beneficial effects:

[0018] 1. Optimize the calculation process and resource scheduling, so that the overlay error measurement for a single wafer image takes within one second, ensuring that the moving stage can complete target switching in the shortest time, thereby significantly improving the efficiency of industrial production and meeting the production requirements of high beat and high precision.

[0019] 2. After obtaining the matching result of the outer rectangle bounding box, crop the measurement image, and then perform the matching of the inner rectangle bounding box that has been viewed, which can effectively filter out non-target features, focus on the analysis of key areas, ensure the continuity and reliability of the entire production process, and meet the stringent requirements of the industrial environment.

[0020] 3. Enhance the image to accurately identify the target area. Using the operators provided by Halcon software, sub-pixel level deviation calculation can be achieved, improving the measurement resolution. Especially under complex image backgrounds or low-contrast conditions, it can accurately capture target features and achieve high-precision measurement.

[0021] 4. Traditional methods only optimize for a single target. For a wafer image with multiple overlay error marks existing simultaneously, only a single target can be detected each time, while this method can directly detect multiple targets, improving the measurement efficiency. Brief Description of the Drawings

[0022] Figure 1 It is a schematic diagram of the overlay error mark;

[0023] Figure 2 It is a schematic diagram of the two-dimensional rectangular coordinate system XOY in the embodiment;

[0024] Figure 3 They are the rectangular boxes 1, 2, and 3 marked in the embodiment;

[0025] Figure 4 They are the enhanced inner and outer rectangular bounding box images;

[0026] Figure 5 It is a schematic diagram of the wafer image to be measured in the embodiment. Detailed Implementation Manner

[0027] The following describes the specific implementation manner of this method with reference to the accompanying drawings.

[0028] A method for measuring wafer overlay error based on Halcon, aiming at Figure 1 performing feature extraction on the overlay error mark composed of an inner and an outer rectangle mark as shown, and measuring the overlay error. The specific steps are as follows:

[0029] Step 1. For wafer images of the same batch, the shapes of the overlay error marks are the same. Therefore, it is necessary to first make an overlay error mark template for the wafers of the same batch, and then find the overlay error marks in other images according to the template and measure the errors.

[0030] Select a wafer image. First, obtain its height and width, and then establish a two-dimensional rectangular coordinate system XOY with the top left vertex of the image as the origin O, the horizontal direction as the X-axis, and the vertical direction as the Y-axis, asFigure 2 As shown. Select one of the overlay error marks and mark the rectangular frames 1, 2, and 3. Among them, only one complete overlay error mark is contained inside the rectangular frame 1, only the inner rectangular mark of the overlay error mark is contained inside the rectangular frame 2, and the rectangular frame 3 is located inside the outer contour line of the inner rectangular mark of the overlay error mark, as Figure 3 shown. Record the coordinates of the upper left vertex and the lower right vertex of the three rectangular frames in the two-dimensional rectangular coordinate system XOY respectively.

[0031] Step 2: According to the coordinate information obtained in Step 1, use the area_center operator of the Halcon software to calculate the center point coordinates of the rectangular frames 1, 2, and 3 respectively, and then use the difference operator to calculate the areas between the rectangular frames 1, 2 and the rectangular frames 2, 3 respectively to obtain the outer rectangular mark frame area and the inner rectangular mark frame area. The reduce_domain operator crops the enhanced wafer image through the outer rectangular mark frame area and the inner rectangular mark frame area to obtain the enhanced outer rectangular mark frame image and the inner rectangular mark frame image, as Figure 4 shown.

[0032] Use the create_shape_model operator of the Halcon software to create shape-based templates, an outer rectangular mark frame template and an inner rectangular mark frame template, from the enhanced outer rectangular mark frame image and the inner rectangular mark frame image respectively.

[0033] In this embodiment, first convert the wafer image into a grayscale image :

[0034]

[0035] where R, G, and B respectively represent the R, G, and B channel values of the wafer image. Then use a 3×3 filter to perform image enhancement on the grayscale image with an enhancement factor of 1 to obtain the enhanced wafer image.

[0036] Step 3: For the wafer pictures that need to measure the overlay error, first perform image enhancement on them and establish a two-dimensional rectangular coordinate system X'O'Y' according to the method in Step 1. Then, according to the outer rectangular mark frame template obtained in Step 2, use the find_shape_model operator of the Halcon software to find all the outer rectangular mark frames that match the template in the enhanced wafer pictures, and obtain the abscissa, ordinate, and rotation angle relative to the outer rectangular mark frame template of the outer rectangular mark frame matching results.

[0037] In this embodiment, only the matching results with a confidence level higher than 0.9 are retained.

[0038] Step 4: According to the coordinate information of the rectangle 1 saved in Step 1 and the position information of the outer rectangle marking box matching result, generate a 2D position transformation matrix 2D1 from rectangle 1 to the outer rectangle marking box. Use the position transformation matrix 2D1 to generate a rectangle on the enhanced wafer image that is the same shape as rectangle 1. , the rectangle contains the corresponding outer rectangle marking box matching result inside.

[0039] Step 5: Use the rectangle to crop the enhanced wafer image. On the cropped image, according to the inner rectangle marking box template obtained in Step 2, use the find_shape_model operator again to find the inner rectangle marking box that matches the template, and obtain the abscissa, ordinate of the inner rectangle marking box matching result, and the rotation angle relative to the inner rectangle marking box template, then enter Step 6. If there is no inner rectangle marking box that matches the template, enter Step 9.

[0040] Searching for the inner rectangle marking box on the cropped image can avoid getting incorrect inner rectangle marking box matching results when directly performing template matching on the wafer image.

[0041] Step 6: According to the coordinate information of the rectangle 2 saved in Step 1 and the position information of the inner rectangle marking box matching result, generate a 2D position transformation matrix 2D2 from rectangle 2 to the inner rectangle marking box. Use the position transformation matrix 2D2 to generate a rectangle on the enhanced wafer image that is the same shape as rectangle 2. , the rectangle contains the corresponding inner rectangle marking box matching result inside.

[0042] Calculate the area between the rectangles and to obtain the outer rectangle marking box prediction area. Crop the enhanced wafer image, and then use the operator lines_gauss to perform Gaussian line segment detection to obtain the outer rectangle marking box contour line.

[0043] Step 7: According to the coordinate information of the rectangle 3 saved in Step 1 and the 2D position transformation matrix 2D2 from rectangle 2 to the inner rectangle marking box, generate a rectangle on the enhanced wafer image that is the same shape as rectangle 3. , so as to ensure that the relative position relationship between the rectangles and is the same as the relative position relationship between rectangles 2 and 3. The rectangle is located inside the inner rectangle marking box matching result.

[0044] Calculate the rectangles and The area between them is used to obtain the inner rectangular marking box prediction area. The enhanced wafer image is cropped, and then the operator lines_gauss is used to detect Gaussian line segments to obtain the contour line of the inner rectangular marking box.

[0045] Step 8: Perform minimum rectangle fitting on the outer rectangular marking box contour line and the inner rectangular marking box contour line respectively to obtain the outer rectangular marking prediction result and the inner rectangular marking prediction result. Use the area_center operator of the Halcon software to obtain the center coordinates of the inner and outer rectangular marking prediction results, accurate to three decimal places. Then calculate the deviation results of the center coordinates of the inner and outer rectangular marking prediction results in the X' axis direction and the Y' axis direction, and store them in the arrays Res_dx and Res_dy.

[0046] Step 9: According to the methods in Steps 4 to 8, sequentially traverse each outer rectangular marking box that meets the template matching to obtain the inner and outer rectangular marking prediction results, and calculate the error measurement results of different sets of registration error marks on this image.

[0047] For the wafer image to be measured as shown in Figure 5 , first use the outer rectangular marking box template obtained in Step 2 to find all outer rectangular marking boxes that meet the template matching. Then sequentially traverse the interiors of the outer rectangular marking box matching results a, b, and c, and use the inner rectangular marking box template obtained in Step 2 to find the inner rectangular marking boxes that meet the template matching. Since there is no inner rectangular marking box that meets the template matching in the outer rectangular marking box matching result b, it indicates that the outer rectangular marking box matching result b is not a correct registration error mark. Therefore, finally output the error calculation results of the 2 registration error marks in the figure.

[0048] The registration error measurement time of this method for a single wafer image is within 1 second, and multiple registration error marks in the image can be measured at one time. Its stability and measurement efficiency can both meet the application requirements of the industrial level. For the method based on neural network, not only a large amount of labeled data is required for model training, but also for a wafer image with multiple registration error marks, its measurement time is about 2 seconds. At the same time, the interpretability of the neural network is relatively weak, which also limits its popularization in industrial scenarios.

Claims

1. A method for measuring the overlay error of wafers based on Halcon, which extracts the features of the overlay error marks composed of two inner and outer rectangular marks in the wafer image and measures the overlay error, is characterized in that: Specifically, it includes the following steps: Step 1: Use rectangular frames 1, 2, and 3 to label a registration error mark in the wafer image. The area between rectangular frames 2 and 3 only contains the contour line of the inner rectangular mark, denoted as the inner rectangular mark frame area; the area between rectangular frames 1 and 2 only contains the contour line of the outer rectangular mark, denoted as the outer rectangular mark frame area; record the position information of each rectangular frame. Step 2: For the images of the inner and outer rectangular mark frame areas, use the create_shape_model operator of Halcon software to create inner and outer rectangular mark templates. Step 3: For the wafer image that needs to measure the registration error, first use the outer rectangular mark template to find all outer rectangular mark frames that match the template, and record the position information; then use the inner rectangular mark template to find the inner rectangular mark frames that match the template within the image range of the outer rectangular mark frames that match the template, record the position information, and discard the outer rectangular mark frames without inner rectangular mark frames inside. Step 4: For the matching results of the inner and outer rectangular marking frames after Step 3, generate rectangular frame 1 on the wafer image where overlay error measurement is required according to the relative position relationship between the rectangular frame defined in Step 1 and the matching results. ′ , 2 ′ , 3 ′ . For rectangular frame 2 ′ , 3 ′ and rectangular frame 1 ′ , 2 ′ , perform Gaussian line segment detection on the area between them to obtain the prediction results of the inner and outer rectangular markings, and calculate the offset between the center coordinates of the prediction results of the inner and outer rectangular markings to complete the overlay error measurement.

2. The method for measuring the overlay error of a wafer based on Halcon according to claim 1, characterized in that: For a wafer image, establish a two-dimensional rectangular coordinate system XOY with its upper left vertex as the origin O, the horizontal direction as the X-axis, and the vertical direction as the Y-axis, and record the coordinates of the upper left vertex and the lower right vertex of the rectangular frame in the two-dimensional rectangular coordinate system XOY as the position information of the rectangular frame.

3. The method for measuring the overlay error of a wafer based on Halcon according to claim 1, characterized in that: Perform image enhancement operations on the wafer image. First, convert it from an RGB image to a grayscale image Gray, and then use a filter with a size of 3×3 and an enhancement factor of 1 to process and enhance the grayscale image Gray.

4. The method for measuring the overlay error of a wafer based on Halcon according to claim 3, wherein: The grayscale image Gray is: Gray = 0.299×R + 0.587×G + 0.114×B; where R, G, and B respectively represent the R, G, and B channel values of the original wafer image.

5. The method for measuring the overlay error of a wafer based on Halcon according to claim 1, wherein: Only one complete registration error mark is contained inside rectangular frame 1, only the inner rectangular mark of the registration error mark is contained inside rectangular frame 2, and rectangular frame 3 is inside the outer contour line of the inner rectangular mark of the registration error mark.

6. The method for measuring the overlay error of a wafer based on Halcon according to claim 1 or 5, characterized in that: Use the area_center operator of Halcon software to calculate the center coordinates of rectangular frames 1, 2, and 3 respectively, and then use the difference operator to calculate the areas between rectangular frames 1, 2 and rectangular frames 2, 3 respectively to obtain the outer rectangular mark frame area and the inner rectangular mark frame area; then use the reduce_domain operator to crop the wafer image to obtain the images of the outer rectangular mark frame area and the inner rectangular mark frame area.

7. The method for measuring the overlay error of a wafer based on Halcon according to claim 1, wherein: Use the find_shape_model operator to find the inner and outer rectangular mark frames that match the template from the enhanced wafer image.

8. The method for measuring the overlay error of a wafer based on Halcon according to claim 1 or 7, characterized in that: Only retain the matching results with a confidence level higher than 0.

9.

9. The method for measuring the overlay error of a wafer based on Halcon according to claim 1, characterized in that: Based on the position information of the matching result between the rectangular box 1 and the outer rectangular marking box, generate a 2D position transformation matrix 2D1 from the rectangular box 1 to the outer rectangular marking box, and use the position transformation matrix 2D1 to generate a rectangular box 1 with the same shape as the rectangular box 1 and containing the matching result of the corresponding outer rectangular marking box inside ′ ; According to the position information of the matching result between the rectangular box 2 and the inner rectangular marking box, generate a 2D position transformation matrix 2D2 from the rectangular box 2 to the inner rectangular marking box, and use the position transformation matrix 2D2 to generate a rectangular box 2 with the same shape as the rectangular box 2 and containing the corresponding matching result of the inner rectangular marking box inside ′ ; Generate a rectangle 3 that has the same shape as rectangle 3 and is within the matching result of the inner rectangle marking box based on the position information of rectangle 3 and the position transformation matrix 2D2 ′ .

10. The method for measuring the overlay error of a wafer based on Halcon according to claim 1 or 9, characterized in that: Calculate the area between rectangle 1 ′ and 2 ′ , and use the operator lines_gauss to perform Gaussian line segment detection to obtain the contour line of the outer rectangle marking box; Calculate the area between rectangle 2 ′ and 3 ′ , and use the operator lines_gauss to perform Gaussian line segment detection to obtain the contour line of the inner rectangle marking box; Perform the minimum rectangle fitting on the outer contour line of the outer rectangular mark frame and the inner contour line of the inner rectangular mark frame respectively to obtain the outer rectangular mark prediction result and the inner rectangular mark prediction result. Use the area_center operator to obtain the center coordinates of the inner and outer rectangular mark prediction results, and then calculate the offset results of the center coordinates of the inner and outer rectangular mark prediction results in the X-axis direction and the Y-axis direction.

Images