An image accurate alignment method, device, equipment and readable storage medium

Abstract

The application discloses a kind of image accurate alignment method, device, equipment and readable storage medium, it can be applied to semiconductor technical field, the method comprises: obtaining the real shot optical image of wafer to be aligned;Determine the target Kernel image corresponding to real shot optical image;Determine the target alignment template graph corresponding to target Kernel image based on preset alignment database;Pre-set alignment database is based on design layout construction;Target alignment template graph is one of all Target alignment graph in pre-set alignment database;Determine the translation deviation value corresponding to target Kernel image based on target alignment template graph;Compensate all pixels in target Kernel image based on translation deviation value, realize image alignment.So, by introducing pre-set alignment database, design layout and image alignment database are directly associated, improve the accuracy between real shot image and design layout alignment.

Description

Technical Field

[0001] This application relates to the field of semiconductor technology, and in particular to a method, apparatus, device, and readable storage medium for precise image alignment. Background Technology

[0002] Image alignment accuracy in pattern optical bright-field defect inspection equipment is a core indicator for ensuring chip yield. In production line wafer applications, detected defects must be accurately mapped to the specific structure of the design layout; otherwise, it is impossible to analyze the impact of defects on the wafer circuit function.

[0003] For image alignment accuracy, mechanical alignment precision is paramount, with indicators such as linearity and repeatability typically measured in nanometers. However, for advanced manufacturing lines, mechanical alignment precision alone is insufficient; feature-based image alignment algorithms must be developed. Existing image alignment methods usually identify the positions of feature patterns in the acquired image and then calculate the deviation between theoretical and actual coordinates to achieve alignment. However, the feature patterns used for detection are generally optical images, and their association with the design layout is unclear, leading to low alignment accuracy between the captured image and the design layout.

[0004] Therefore, how to improve the alignment accuracy between the actual photographed images and the design layout is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] To address the aforementioned issues, this application provides a method, apparatus, device, and readable storage medium for precise image alignment. By introducing a preset alignment database, the design layout and the image alignment database are directly linked, thereby improving the accuracy of alignment between the actual captured image and the design layout.

[0006] In a first aspect, embodiments of this application provide a method for precise image alignment, including:

[0007] Obtain real-shot optical images of the wafer to be aligned;

[0008] Determine the target kernel image corresponding to the captured optical image;

[0009] A target alignment template image corresponding to the target kernel image is determined based on a preset alignment database; the preset alignment database is constructed based on the design layout; the target alignment template image is one of all target alignment images in the preset alignment database.

[0010] Based on the target alignment template image, determine the translation deviation value corresponding to the target kernel image;

[0011] Based on the translation deviation value, all pixels in the target kernel image are compensated to achieve image alignment.

[0012] Optionally, the preset alignment database is constructed using the following method:

[0013] Construct a target layout corresponding to the wafer to be aligned;

[0014] The target map is segmented to obtain N aligned kernel images;

[0015] Based on the preset judgment criteria, the target feature map and the feature coordinates corresponding to the target feature map in each aligned kernel image are determined respectively.

[0016] Acquire the alignment image corresponding to each feature coordinate;

[0017] Based on the grayscale information of the actual captured image, each target feature map is rendered to obtain the rendered image corresponding to each target feature map.

[0018] Each rendered image is matched with its corresponding real-shot image to obtain the Target bitmap corresponding to each rendered image.

[0019] A preset alignment database is constructed based on the feature coordinates and the target alignment map.

[0020] Optionally, constructing the target layout corresponding to the wafer to be aligned includes:

[0021] Obtain the original design layout corresponding to the wafer to be aligned;

[0022] Based on the target process, extract the target layer corresponding to the target process from the original design layout;

[0023] Construct the target map based on the target layer.

[0024] Optionally, when the number of target layers is greater than or equal to two, constructing the target map based on the target layers includes:

[0025] Merge the data corresponding to each target layer;

[0026] The merged target layers are renumbered and renamed to obtain the target map.

[0027] Optionally, the method further includes:

[0028] Determine the translation and scaling factors between the first coordinate system corresponding to the target layout and the second coordinate system corresponding to the wafer to be aligned;

[0029] The acquisition of the alignment image corresponding to each feature coordinate includes:

[0030] The coordinates of the acquisition point corresponding to each feature coordinate are determined based on the translation coefficient and the scaling coefficient.

[0031] Based on the coordinates of the acquisition points, acquire the corresponding real-shot image for each feature coordinate.

[0032] Optionally, determining the translation and scaling factors between the first coordinate system corresponding to the target layout and the second coordinate system corresponding to the wafer to be aligned includes:

[0033] A first target point and a second target point are determined on the wafer to be aligned, and the coordinates of the first point corresponding to the first target point and the coordinates of the second point corresponding to the second target point are determined based on the second coordinate system.

[0034] On the target map, determine a third target point corresponding to the first target point and a fourth target point corresponding to the second target point;

[0035] The coordinates of the third point corresponding to the third target point and the coordinates of the fourth point corresponding to the fourth target point are determined based on the first coordinate system.

[0036] Based on the coordinate system mapping formula, the translation coefficient and scaling coefficient between the first coordinate system and the second coordinate system are determined using the coordinates of the first point, the second point, the third point, and the fourth point.

[0037] Optionally, determining the target feature map and the corresponding feature coordinates of each aligned kernel image according to a preset judgment criterion includes:

[0038] Based on the principle of uniqueness, candidate small images are determined in each pair of kernel images;

[0039] Based on the preset judgment criteria, target small images are selected from the candidate small images corresponding to each pair of kernel images as target feature maps in each pair of kernel images.

[0040] Determine the feature coordinates of the target feature map corresponding to each pair of kernel images.

[0041] Optionally, the step of selecting target feature maps from the candidate image corresponding to each corresponding kernel image as target feature maps in each corresponding kernel image according to a preset judgment criterion includes:

[0042] A scoring function is constructed based on at least one of the following: the distance between the candidate small image and the center of the corresponding Kernel image; the number of polygons within the size range of the candidate small image; and the proportion of polygon area within the candidate small image.

[0043] Based on the scoring function, the candidate image with the highest score is selected from each candidate image corresponding to each pair of kernel images and recorded as the target image, which is then used as the target feature map in each pair of kernel images.

[0044] Optionally, determining the feature coordinates of the target feature map corresponding to each paired kernel image includes:

[0045] Determine the coordinates of the center point of the target feature map corresponding to each pair of kernel images;

[0046] The center point coordinates of the target feature map corresponding to each pair of kernel images are used as the feature coordinates of the target feature map corresponding to each pair of kernel images.

[0047] Optionally, when the distance between the candidate small image and the center of the corresponding kernel image is used as the preset judgment criterion, the scoring function is used as the different regions in the corresponding kernel image that the candidate small image falls into.

[0048] Optionally, when the number of polygons within the size range of the candidate small images is used as a preset judgment criterion, the expression of the scoring function is:

[0049] ;

[0050] In the formula, S2 is the scoring function, A and B are two specific numerical values, and x is the number of polygons.

[0051] Optionally, when the area ratio of polygons within the candidate small image is used as a preset judgment criterion, the expression of the scoring function is:

[0052] ;

[0053] ;

[0054] In the formula This represents the total area of ​​the polygons within the candidate small images. The total area of ​​the candidate small images. For area percentage, This is the scoring function.

[0055] Secondly, embodiments of this application provide an apparatus for precise image alignment, comprising:

[0056] The acquisition module is used to acquire real-shot optical images of the wafer to be aligned;

[0057] The first determining module is used to determine the target Kernel image corresponding to the real-shot optical image;

[0058] The second determining module is used to determine a target alignment template image corresponding to the target kernel image based on a preset alignment database; the preset alignment database is constructed based on the design layout; the target alignment template image is one of all target alignment images in the preset alignment database.

[0059] The third determining module is used to determine the translational deviation value corresponding to the target Kernel image based on the target alignment template image;

[0060] The compensation alignment module is used to compensate all pixels in the target kernel image based on the translation deviation value, thereby achieving image alignment.

[0061] Thirdly, embodiments of this application provide an image precision alignment device, comprising:

[0062] Memory, used to store computer programs;

[0063] A processor for executing the computer program to implement the method of precise image alignment as described above.

[0064] Fourthly, embodiments of this application provide a readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the method for precise image alignment as described above.

[0065] As can be seen from the above technical solutions, compared with the prior art, this application has the following advantages:

[0066] This application provides a method for precise image alignment. First, a real-world optical image of the wafer to be aligned is acquired, and a target kernel image corresponding to the real-world optical image is determined. Then, a target alignment template image corresponding to the target kernel image is determined based on a preset alignment database. The preset alignment database is constructed based on the design layout; the target alignment template image is one of all target alignment images in the preset alignment database. Finally, a translational deviation value corresponding to the target kernel image is determined based on the target alignment template image, and all pixels in the target kernel image are compensated based on the translational deviation value to achieve image alignment. Thus, by introducing a preset alignment database, the design layout and the image alignment database are directly linked, improving the accuracy of alignment between the real-world image and the design layout. Attached Figure Description

[0067] Figure 1 A flowchart illustrating a method for precise image alignment provided in this application embodiment;

[0068] Figure 2 A flowchart illustrating a method for constructing a preset alignment database as provided in an embodiment of this application;

[0069] Figure 3 A schematic diagram illustrating the construction of a target layout provided in an embodiment of this application;

[0070] Figure 4 This application provides a schematic diagram of a target layout segmentation.

[0071] Figure 5 A schematic diagram illustrating a scoring criterion provided in an embodiment of this application;

[0072] Figure 6 A schematic diagram illustrating the location of a target feature map provided in an embodiment of this application;

[0073] Figure 7 This is a schematic diagram illustrating target point selection as provided in an embodiment of this application.

[0074] Figure 8 A rendering schematic diagram of a design layout provided in an embodiment of this application;

[0075] Figure 9 A schematic diagram of a process for generating a Target bitmap provided in an embodiment of this application;

[0076] Figure 10 This is a schematic diagram of the structure of an image precise alignment device provided in an embodiment of this application. Detailed Implementation

[0077] As mentioned earlier, existing image alignment methods suffer from low accuracy in aligning real-world images with the design layout. Specifically, existing methods typically achieve alignment by identifying the positions of feature graphics in the acquired image and then calculating the deviation between theoretical and actual coordinates. However, the feature graphics used for detection are generally optical images, and their correlation with the design layout is not clear. This leads to the problem of low alignment accuracy between real-world images and the design layout.

[0078] To address the aforementioned issues, this application provides a method for precise image alignment, comprising: first, acquiring a real-shot optical image of the wafer to be aligned, and determining a target kernel image corresponding to the real-shot optical image; then, determining a target alignment template image corresponding to the target kernel image based on a preset alignment database. The preset alignment database is constructed based on the design layout; the target alignment template image is one of all target alignment images in the preset alignment database. Finally, determining the translational deviation value corresponding to the target kernel image based on the target alignment template image, and compensating all pixels in the target kernel image based on the translational deviation value to achieve image alignment.

[0079] Thus, by introducing a pre-set alignment database, the design layout and the image alignment database are directly linked, improving the accuracy of alignment between the actual photographed image and the design layout.

[0080] It should be noted that the method, apparatus, device, and readable storage medium for precise image alignment provided in this application can be applied to the field of semiconductor technology. The above are merely examples and do not limit the application field of the method, apparatus, device, and readable storage medium for precise image alignment provided in this application.

[0081] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0082] Figure 1 This is a flowchart illustrating a method for precise image alignment provided in an embodiment of this application. (In conjunction with...) Figure 1 As shown in the embodiments of this application, a method for precise image alignment may include:

[0083] S101: Acquire a real-shot optical image of the wafer to be aligned.

[0084] In practical applications, a time-delay integration (TDI) camera can be used to acquire real-world optical images of the wafer to be aligned.

[0085] S102: Determine the target kernel image corresponding to the actual optical image.

[0086] In practical applications, based on the detection parameters configured in the detection equipment, such as pixel size, the continuous images acquired by the TDI camera are segmented, and the resulting local images are called kernel images. The segmentation parameters used in this case are the same as those used when dividing the design layout to construct the preset alignment database, and the physical size and position of the kernel images obtained in each die are also consistent with the layout division results. It can be understood that the actual optical image of the wafer to be aligned is composed of a large number of kernel images. The target kernel image is any one of all the kernel images that constitute the entire actual optical image; that is, the precise image alignment method provided in this application precisely aligns each kernel image.

[0087] S103: Determine the target alignment template image corresponding to the target kernel image based on the preset alignment database; the preset alignment database is constructed based on the design layout; the target alignment template image is one of all target alignment images in the preset alignment database.

[0088] In practical applications, a pre-set alignment database records the association between actual optical images and design layouts. Specifically, for each kernel image constituting an actual optical image, there is a uniquely corresponding alignment feature map (target alignment map) in the pre-set alignment database. This target alignment map clearly shows the positional matching between the design layout and the actual optical image within the size range of the corresponding kernel image. Thus, based on this pre-set alignment database, after determining the target kernel image, this embodiment of the application needs to first determine a uniquely corresponding target alignment map, i.e., a target alignment template map, from the pre-set alignment database.

[0089] S104: Determine the translation deviation value corresponding to the target kernel image based on the target alignment template image.

[0090] In practical applications, the target bitmap can clearly define the positional matching between the design layout and the actual optical image within the size range of the corresponding kernel image. Therefore, by using the target bitmap (i.e., the target alignment template image) and the feature coordinates of the design layout corresponding to the target kernel image as inputs to the algorithm model, the translational deviation value corresponding to the target kernel image can be obtained.

[0091] S105: Compensate all pixels in the target kernel image based on the translation deviation value to achieve image alignment.

[0092] In practical applications, the translation deviation value obtained above can be applied to all pixels in the target kernel image to compensate for the alignment deviation and achieve ultra-high precision alignment between the real-shot image and the design layout.

[0093] In summary, the image alignment method provided in this application first acquires a real-shot optical image of the wafer to be aligned and determines the target kernel image corresponding to the real-shot optical image. Then, based on a preset alignment database, a target alignment template image corresponding to the target kernel image is determined. The preset alignment database is constructed based on the design layout; the target alignment template image is one of all target alignment images in the preset alignment database. Finally, based on the target alignment template image, the translational deviation value corresponding to the target kernel image is determined, and all pixels in the target kernel image are compensated based on the translational deviation value to achieve image alignment. Thus, by introducing a preset alignment database, the design layout and the image alignment database are directly linked, improving the accuracy of alignment between the real-shot image and the design layout.

[0094] Figure 2 A flowchart illustrating a method for constructing a preset alignment database as provided in an embodiment of this application. (In conjunction with...) Figure 2 As shown, the method for constructing this preset positional database may include:

[0095] S201: Construct the target layout corresponding to the wafer to be aligned.

[0096] In practical applications, the design layout used in wafer production lines includes information on all process layers of the wafer. However, for inspection equipment, not all layer information is needed; instead, the focus is on one or a few layers that the current process inspection involves. Therefore, it is necessary to reconstruct the target layout corresponding to a specific process for the wafer to be aligned.

[0097] Furthermore, since the methods for constructing the target layout are not entirely the same, this application embodiment can describe one possible construction method.

[0098] In one scenario, S201: Constructing a target layout corresponding to the wafer to be aligned, which may specifically include:

[0099] Obtain the original design layout corresponding to the wafer to be aligned;

[0100] Based on the target process, extract the target layer corresponding to the target process from the original design layout;

[0101] Construct the target map based on the target layer.

[0102] In practical applications, the original design layout corresponds to all process layer information of the wafer to be aligned. To perform targeted image alignment in conjunction with target process detection, the target process must first be determined, and the target layer corresponding to that target process is extracted from the original design layout. Then, based on the target layer and its corresponding data, a target layout corresponding to the target process is constructed.

[0103] Furthermore, when the number of target layers is greater than or equal to two, the step of constructing the target map based on the target layers includes:

[0104] Merge the data corresponding to each target layer;

[0105] The merged target layers are renumbered and renamed to obtain the target map.

[0106] Figure 3 This is a schematic diagram illustrating the construction of a target layout as provided in an embodiment of this application. (In conjunction with...) Figure 3 As shown, the original design layout includes all process layer information for the wafer to be aligned, such as Layer 1, Layer 2, Layer 3, Layer 4, Layer 5, Layer 6, Layer 7, Layer 8, Layer 9, Layer 10, and Layer 11, etc. However, the target layers corresponding to the target process may only be Layer 2, Layer 3, Layer 4, Layer 7, and Layer 9, etc. Therefore, this application first extracts the target layers corresponding to the target process and merges the data of multiple target layers according to actual needs. The merging principle is that the merged target layers have the same or similar optical response, such as Layer 2, Layer 3, and Layer 4. Furthermore, the extracted and merged layers are renumbered (e.g., Layer NO. is New Layer 1, New Layer 2, New Layer 3, and New Layer 4) and named (e.g., Desciption is Metal 1 and AA1, etc.) for easy retrieval later. Finally, the data is saved to obtain the simplified layout data file, i.e., the target layout.

[0107] S202: Segment the target map to obtain N aligned kernel images.

[0108] In practical applications, to correspond the N alignment kernel images obtained by segmenting the target layout with the N target kernel images obtained by segmenting the actual optical image corresponding to the wafer to be aligned, both should use the same segmentation parameters. Based on the actual detection parameters configured in the detection equipment, such as pixel size, the size of the kernel image during the detection process is determined, and then the target layout is segmented according to this size to obtain N alignment kernel images. Here, N is an integer greater than or equal to 1. Figure 4 This is a schematic diagram illustrating the segmentation of a target layout as provided in an embodiment of this application. (In conjunction with...) Figure 4 As shown, the target layout is segmented starting from the top left corner according to the segmentation size. If the horizontal or vertical dimension of the target layout is not divisible by the kernel size, the last column / row of aligned kernel images is shifted forward to ensure edge alignment, resulting in N aligned kernel images. Figure 4 (N=30).

[0109] S203: Based on the preset judgment criteria, determine the Target feature map and the feature coordinates corresponding to the Target feature map in each paired Kernel image.

[0110] In practical applications, for N alignment kernel images segmented from the target map, a feature region needs to be selected from each alignment kernel image for subsequent alignment. This feature region can be called a target feature map, and its size must be smaller than the size of the alignment kernel image; the specific size can be user-defined. Furthermore, the selection of this target feature map must adhere to preset criteria. After determining the target feature map corresponding to each alignment kernel image, the feature coordinates corresponding to each target feature map are saved accordingly.

[0111] Furthermore, since the methods for determining the Target feature map and the corresponding feature coordinates in each paired Kernel image are not entirely the same, this application embodiment can describe one possible determination method.

[0112] In one scenario, determining the target feature map and the corresponding feature coordinates of each aligned kernel image based on a preset judgment criterion includes:

[0113] Based on the principle of uniqueness, candidate small images are determined in each pair of kernel images;

[0114] Based on the preset judgment criteria, target small images are selected from the candidate small images corresponding to each pair of kernel images as target feature maps in each pair of kernel images.

[0115] Determine the feature coordinates of the target feature map corresponding to each pair of kernel images.

[0116] In practical applications, the selection of target feature maps needs to follow the principle of uniqueness, meaning that each target feature map is unique within a certain range around it (the range is determined by the alignment accuracy of the detection device, such as 10 pixels). In this case, a single alignment kernel image may contain multiple small images that satisfy the above principle; these small images can be called candidate small images. Furthermore, a ranking mechanism needs to be introduced to select the best small image from the candidate small images corresponding to each alignment kernel image as the target feature map based on preset judgment criteria, and then record the feature coordinates of each target feature map.

[0117] Furthermore, since there are different ways to select the target feature map, this application embodiment can describe one possible selection method.

[0118] In one scenario, selecting a target feature map from the candidate images corresponding to each paired kernel image as the target feature map in each paired kernel image, based on a preset judgment criterion, includes:

[0119] A scoring function is constructed based on at least one of the following: the distance between the candidate small image and the center of the corresponding Kernel image; the number of polygons within the size range of the candidate small image; and the proportion of polygon area within the candidate small image.

[0120] Based on the scoring function, the candidate image with the highest score is selected from each candidate image corresponding to each pair of kernel images and recorded as the target image, which is then used as the target feature map in each pair of kernel images.

[0121] Furthermore, when the distance between the candidate small image and the center of the corresponding kernel image is used as the preset judgment criterion, the different regions in which the candidate small image falls within the corresponding kernel image are used as the scoring function.

[0122] When the number of polygons within the size range of the candidate small images is used as the preset judgment criterion, the expression of the scoring function is:

[0123] ;

[0124] In the formula, S2 is the scoring function, A and B are two specific quantitative values, and x is the number of polygons;

[0125] When the area ratio of polygons within the candidate small image is used as the preset judgment criterion, the expression of the scoring function is:

[0126] ;

[0127] ;

[0128] In the formula This represents the total area of ​​the polygons within the candidate small images. This represents the total area of ​​the candidate small images. For area percentage, This is the scoring function.

[0129] In practical applications, within each aligning kernel image, a scoring function can be constructed based on the distance between the candidate small image and the center of the aligning kernel image, the number of polygons within the candidate small image's size range, and the polygon area ratio. The candidate small image with the higher score is then designated as the target small image and used as the target feature map corresponding to that aligning kernel image. Specifically, Figure 5 This is a schematic diagram illustrating a scoring criterion provided in an embodiment of this application. (In conjunction with...) Figure 5 As shown, when using the distance from the center of the candidate small image to the center of the corresponding kernel image as the preset judgment criterion, different regions can be divided in the corresponding kernel image, and the region where the candidate small image falls can be used as the scoring function. For example, the closer the candidate small image is to the center of the corresponding kernel image, the higher the score. The very center receives 100 points, then 0-100 points, and the very edge receives 0 points. When using the number of polygons within the size range of the candidate small image as the preset judgment criterion, the more polygons within the candidate small image, the higher the score. The corresponding scoring function is as follows:

[0130] ;

[0131] In the formula, S2 is the scoring function, A and B are two specific numerical values, and x is the number of polygons. When using the proportion of polygon area within the candidate small image as the preset judgment criterion, the larger the proportion, the higher the score. The corresponding scoring function is as follows:

[0132] ;

[0133] ;

[0134] In the formula This represents the total area of ​​the polygons within the candidate small images. This represents the total area of ​​the candidate small images. For area percentage, Let this be the scoring function. Thus, if two or more criteria are selected to score the selected small image, the scoring function is as follows:

[0135] ;

[0136] In the formula, For the scoring function, This represents the individual score of the candidate small image under each judgment criterion. This indicates the weight of each judgment criterion. Different judgment criteria can be used to indicate different judgment criteria. In addition, other judgment criteria can be introduced in the embodiments of this application, such as "the number of polygon corner points contained in the candidate small image" and "the line density of the polygons in the candidate small image".

[0137] Furthermore, since the methods for determining the feature coordinates of the target feature map corresponding to each pair of kernel images are not entirely the same, this application embodiment can describe one possible determination method.

[0138] In one case, determining the feature coordinates of the target feature map corresponding to each paired kernel image includes:

[0139] Determine the coordinates of the center point of the target feature map corresponding to each pair of kernel images;

[0140] The center point coordinates of the target feature map corresponding to each pair of kernel images are used as the feature coordinates of the target feature map corresponding to each pair of kernel images.

[0141] Figure 6 This is a schematic diagram illustrating the location of a target feature map according to an embodiment of this application. (In conjunction with...) Figure 6 As shown, the yellow pentagram in each paired kernel image represents the position of the target feature map selected from each candidate small image. Furthermore, the center point coordinates of each target feature map are determined, and the center point coordinates of the target feature map corresponding to each paired kernel image are used as the feature coordinates of the target feature map corresponding to each paired kernel image.

[0142] S204: Acquire the alignment image corresponding to each feature coordinate.

[0143] In practical applications, alignment images need to be acquired from the wafer to be aligned, and then matched with the design layout to determine and save the deviations. Specifically, the alignment images correspond to the feature coordinates, requiring adjustments to the detection equipment parameters, such as wavelength, polarization, and pixel size, to ensure consistency with the actual detection process. Images at the feature coordinates of each target feature map are acquired as alignment images. Furthermore, during the acquisition process, the size of the acquired images needs to be controlled according to the alignment accuracy of the equipment hardware, ensuring they are larger than the size of the target feature map. That is, the alignment images should be slightly larger than the size of the target feature map.

[0144] Furthermore, since the methods for acquiring alignment images are not entirely the same, this application embodiment can describe one possible acquisition method.

[0145] In one instance, the method further includes:

[0146] Determine the translation and scaling factors between the first coordinate system corresponding to the target layout and the second coordinate system corresponding to the wafer to be aligned;

[0147] The acquisition of the alignment image corresponding to each feature coordinate includes:

[0148] The coordinates of the acquisition point corresponding to each feature coordinate are determined based on the translation coefficient and the scaling coefficient.

[0149] Based on the coordinates of the acquisition points, acquire the corresponding real-shot image for each feature coordinate.

[0150] In practical applications, the feature coordinates are based on the coordinate system of the design layout, while the alignment image corresponding to the wafer to be aligned is based on the coordinate system of the inspection equipment. Although the design layout and the alignment image correspond, the coordinates of the two corresponding points on each may differ due to factors such as equipment motion errors. Therefore, it is necessary to determine the mapping relationship between the two coordinate systems before acquiring the alignment image, that is, to determine the translation and scaling factors between the first coordinate system corresponding to the target layout and the second coordinate system corresponding to the wafer to be aligned. Then, based on the translation and scaling factors, the coordinates of the acquisition point corresponding to each feature coordinate are determined, and then the image within a certain area is acquired as the alignment image corresponding to the feature coordinate using the acquisition point coordinates as a reference. For example, if the coordinates (0,0) in the first coordinate system correspond to the coordinates (1,1) in the second coordinate system, and the feature coordinate is (0,0), then when acquiring the alignment image corresponding to this feature coordinate, it is necessary to acquire the image centered at (1,1).

[0151] Furthermore, since the methods for determining the translation and scaling factors between the first and second coordinate systems are not entirely the same, this application embodiment can describe one possible determination method.

[0152] In one case, determining the translation and scaling factors between the first coordinate system corresponding to the target layout and the second coordinate system corresponding to the wafer to be aligned includes:

[0153] A first target point and a second target point are determined on the wafer to be aligned, and the coordinates of the first point corresponding to the first target point and the coordinates of the second point corresponding to the second target point are determined based on the second coordinate system.

[0154] On the target map, determine a third target point corresponding to the first target point and a fourth target point corresponding to the second target point;

[0155] The coordinates of the third point corresponding to the third target point and the coordinates of the fourth point corresponding to the fourth target point are determined based on the first coordinate system.

[0156] Based on the coordinate system mapping formula, the translation coefficient and scaling coefficient between the first coordinate system and the second coordinate system are determined using the coordinates of the first point, the second point, the third point, and the fourth point.

[0157] In practical applications, the wafer to be aligned is first placed on the testing equipment to complete the rotation, alignment and other operations and establish the (x, y) coordinate system, that is, the second coordinate system. Figure 7 This is a schematic diagram illustrating target point selection as provided in an embodiment of this application. (In conjunction with...) Figure 7 As shown, a Reticle (photomask) region is selected on the wafer to be aligned, typically near the center of the wafer. Further, a first target point and a second target point are determined within this Reticle region. The first target point can be a unique, sharp, and easily distinguishable corner point in the lower left corner of the Reticle region, with its coordinates recorded as (x1, y1). The second target point can be a unique, sharp, and easily distinguishable corner point in the upper right corner of the Reticle region, with its coordinates recorded as (x2, y2). Further, a third target point corresponding to the first target point and a fourth target point corresponding to the second target point are found on the target layout through position matching. The coordinates of the third target point are (x1, y1). gds1 y gds1 The coordinates of the fourth point corresponding to the fourth target point are (x... gds2 y gds2 Furthermore, by substituting the two sets of points mentioned above into the coordinate system mapping formula, the corresponding scaling and translation coefficients can be calculated. The corresponding coordinate system mapping formula is as follows:

[0158] ;

[0159] In the formula, x and y correspond to the x-coordinate and y-coordinate of the first and second points, respectively. gds and y gds The x and y coordinates correspond to the coordinates of the third and fourth points, respectively. and These correspond to the scaling factors on the x-axis and y-axis of the two coordinate systems, respectively. and These correspond to the translation coefficients of the two coordinate systems on the x-axis and y-axis, respectively.

[0160] S205: Render each Target feature map based on the grayscale information of the actual captured image to obtain the rendering image corresponding to each Target feature map.

[0161] In practical applications, the actual captured image corresponds to the target feature map, but is slightly larger. Because the target feature map is cropped from the target image, its contrast differs from the actual captured optical image. Therefore, to better match the two, the target feature map needs to be rendered based on the grayscale information of the actual captured image. Specifically, Figure 8 This is a rendered schematic diagram of a design layout provided for an embodiment of this application. (In conjunction with...) Figure 8 As shown, for each target feature map, its coordinate position is aligned with its corresponding real-world image according to the coordinate system mapping relationship described above (e.g., ...). Figure 8 (a) Then, based on the grayscale information of the actual captured image, the Target feature map is processed to obtain its corresponding rendering image (e.g., Figure 8 (b) in the middle.

[0162] S206: Match the positions of each rendered image with the corresponding real-shot image to obtain the Target bitmap corresponding to each rendered image.

[0163] In practical applications, since the target layout only contains a few selected target layers corresponding to the target process, and the actual wafer structure optical imaging is complex, the rendered image cannot be directly used as the alignment template image (the template image corresponding to the target kernel image during the precise alignment process of the image). The alignment template image needs to come from the actual optical imaging results. Figure 9 This is a schematic flowchart illustrating a process for generating a Target bitmap, provided in an embodiment of this application. (In conjunction with...) Figure 9As shown, 'a' represents a set of target feature maps, extracted from the alignment kernel image, and 'b' represents a set of rendered images, obtained by rendering each target feature map based on the grayscale information of its corresponding alignment real-world image. It is understandable that since the target alignment maps used to construct the preset alignment database need to be optical real-world images, and the grayscale information of the target feature maps differs significantly from that of the optical real-world images, errors can easily be introduced during position matching. Therefore, this embodiment first renders each target feature map to obtain its corresponding rendered image, and then, based on the alignment real-world image corresponding to each target feature map, performs position template matching using the rendered image and its corresponding alignment real-world image, such as... Figure 9 As shown in 'c', the yellow box indicates the position of the rendered image. Since the alignment image is acquired based on the position of the Target feature map, continuing with the example above, given the mapping relationship of a point coordinate (0,0) in the first coordinate system corresponding to a point coordinate (1,1) in the second coordinate system, if there exists a Target feature map with a corresponding feature coordinate of (0,0), then when acquiring the alignment image corresponding to this feature coordinate, it needs to be centered at (1,1). It can be understood that in the Target alignment map, the center point coordinate (0,0) of the yellow box should coincide with the center point coordinate (1,1) of the alignment image, indicating no deviation. When the center of the yellow box does not coincide with the center of the alignment image, or in other words, with the center of the Target alignment map, it indicates a translational deviation. In other words, by using each Target alignment map, it can be determined whether the target kernel image corresponding to that Target alignment map has a translational deviation.

[0164] S207: Construct a preset alignment database based on the feature coordinates and the Target alignment map.

[0165] In practical applications, the feature coordinates of each target alignment bitmap and its corresponding target feature map are saved to generate a preset alignment database. During actual alignment, the target alignment template map and feature coordinates corresponding to the target kernel image can be determined from the preset alignment database. Then, the target alignment template map and feature coordinates are used as inputs into the algorithm model to obtain the translation deviation value.

[0166] In summary, this embodiment of the application, when constructing a preset alignment database, firstly constructs a target layout corresponding to the wafer to be aligned, and segments the target layout to obtain N alignment kernel images. Then, based on preset judgment criteria, it determines the target feature map and the corresponding feature coordinates in each alignment kernel image, and acquires an alignment real-shot image corresponding to each feature coordinate. Next, it renders each target feature map based on the grayscale information of the alignment real-shot images, obtaining a rendered image corresponding to each target feature map. Finally, it performs position matching between each rendered image and the corresponding alignment real-shot image to obtain the target alignment map corresponding to each rendered image, and constructs a preset alignment database based on the feature coordinates and the target alignment map. In this way, the alignment database is automatically built, reducing the burden of manual operation, avoiding the introduction of human error, and supporting the configuration of high-density, large-scale feature graphics, further improving the accuracy of alignment between real-shot images and the design layout.

[0167] Figure 10 This is a schematic diagram of a device for precise image alignment provided in an embodiment of this application. (Combined with...) Figure 10 As shown, the image precise alignment device 1000 may include:

[0168] The acquisition module 1001 is used to acquire real-shot optical images of the wafer to be aligned;

[0169] The first determining module 1002 is used to determine the target kernel image corresponding to the real-shot optical image;

[0170] The second determining module 1003 is used to determine a target alignment template image corresponding to the target kernel image based on a preset alignment database; the preset alignment database is constructed based on the design layout; the target alignment template image is one of all target alignment images in the preset alignment database.

[0171] The third determining module 1004 is used to determine the translation deviation value corresponding to the target Kernel image based on the target alignment template image;

[0172] The compensation alignment module 1005 is used to compensate all pixels in the target kernel image based on the translation deviation value to achieve image alignment.

[0173] As one implementation method, the above-mentioned image precise alignment device 1000, which is designed to construct a preset alignment database, further includes: a first construction module, a segmentation module, a fourth determination module, an acquisition module, a rendering module, a position matching module, and a second construction module.

[0174] The first construction module is used to construct the target layout corresponding to the wafer to be aligned;

[0175] The segmentation module is used to segment the target map to obtain N aligned kernel images;

[0176] The fourth determining module is used to determine the Target feature map and the feature coordinates corresponding to the Target feature map in each paired Kernel image according to a preset judgment criterion.

[0177] The acquisition module is used to acquire the alignment image corresponding to each feature coordinate.

[0178] The rendering module is used to render each Target feature map based on the grayscale information of the actual captured image, and obtain the rendering map corresponding to each Target feature map respectively.

[0179] The position matching module is used to match the position of each rendered image with the corresponding real-shot image to obtain the Target bitmap corresponding to each rendered image;

[0180] The second construction module is used to construct a preset alignment database based on the feature coordinates and the Target alignment map.

[0181] As one implementation method, regarding how to construct the target layout corresponding to the wafer to be aligned, the first construction module mentioned above is specifically used for:

[0182] Obtain the original design layout corresponding to the wafer to be aligned;

[0183] Based on the target process, extract the target layer corresponding to the target process from the original design layout;

[0184] Construct the target map based on the target layer.

[0185] Furthermore, when the number of target layers is greater than or equal to two, the step of constructing the target map based on the target layers includes:

[0186] Merge the data corresponding to each target layer;

[0187] The merged target layers are renumbered and renamed to obtain the target map.

[0188] As one implementation method, the above-mentioned image precise alignment device 1000, which is used to acquire aligned real-shot images, further includes: a fifth determining module;

[0189] The fifth determining module is used to determine the translation coefficient and scaling coefficient between the first coordinate system corresponding to the target layout and the second coordinate system corresponding to the wafer to be aligned;

[0190] The aforementioned data acquisition module is specifically used for:

[0191] The coordinates of the acquisition point corresponding to each feature coordinate are determined based on the translation coefficient and the scaling coefficient.

[0192] Based on the coordinates of the acquisition points, acquire the corresponding real-shot image for each feature coordinate.

[0193] As one implementation method, the fifth determining module is specifically used to determine the translation and scaling factors:

[0194] A first target point and a second target point are determined on the wafer to be aligned, and the coordinates of the first point corresponding to the first target point and the coordinates of the second point corresponding to the second target point are determined based on the second coordinate system.

[0195] On the target map, determine a third target point corresponding to the first target point and a fourth target point corresponding to the second target point;

[0196] The coordinates of the third point corresponding to the third target point and the coordinates of the fourth point corresponding to the fourth target point are determined based on the first coordinate system.

[0197] Based on the coordinate system mapping formula, the translation coefficient and scaling coefficient between the first coordinate system and the second coordinate system are determined using the coordinates of the first point, the second point, the third point, and the fourth point.

[0198] As one implementation method, the fourth determining module mentioned above, which addresses how to determine the target feature map and feature coordinates, includes: a first determining submodule, a selection module, and a second determining submodule;

[0199] The first determination submodule is used to determine the candidate small images in each pair of kernel images based on the principle of uniqueness;

[0200] The selection module is used to select target small images from the candidate small images corresponding to each pair of kernel images as target feature maps in each pair of kernel images according to preset judgment criteria.

[0201] The second determining submodule is used to determine the feature coordinates of the target feature map corresponding to each paired kernel image.

[0202] As one implementation method, regarding how to select a target small image from the candidate small images as the target feature map in each corresponding kernel image, the above-mentioned selection module is specifically used for:

[0203] A scoring function is constructed based on at least one of the following: the distance between the candidate small image and the center of the corresponding Kernel image; the number of polygons within the size range of the candidate small image; and the proportion of polygon area within the candidate small image.

[0204] Based on the scoring function, the candidate image with the highest score is selected from each candidate image corresponding to each pair of kernel images and recorded as the target image, which is then used as the target feature map in each pair of kernel images.

[0205] Furthermore, when the distance between the candidate small image and the center of the corresponding kernel image is used as the preset judgment criterion, the different regions in which the candidate small image falls within the corresponding kernel image are used as the scoring function.

[0206] When the number of polygons within the size range of the candidate small images is used as the preset judgment criterion, the expression of the scoring function is:

[0207] ;

[0208] In the formula, S2 is the scoring function, A and B are two specific quantitative values, and x is the number of polygons;

[0209] When the area ratio of polygons within the candidate small image is used as the preset judgment criterion, the expression of the scoring function is:

[0210] ;

[0211] ;

[0212] In the formula This represents the total area of ​​the polygons within the candidate small images. This represents the total area of ​​the candidate small images. For area percentage, This is the scoring function.

[0213] As one implementation method, regarding how to determine the feature coordinates of the target feature map corresponding to each kernel image, the second determining submodule is specifically used for:

[0214] Determine the coordinates of the center point of the target feature map corresponding to each pair of kernel images;

[0215] The center point coordinates of the target feature map corresponding to each pair of kernel images are used as the feature coordinates of the target feature map corresponding to each pair of kernel images.

[0216] In summary, this application first acquires a real-world optical image of the wafer to be aligned and determines the target kernel image corresponding to the real-world optical image. Then, based on a preset alignment database, a target alignment template image corresponding to the target kernel image is determined. The preset alignment database is constructed based on the design layout; the target alignment template image is one of all target alignment images in the preset alignment database. Finally, based on the target alignment template image, the translational deviation value corresponding to the target kernel image is determined, and all pixels in the target kernel image are compensated based on the translational deviation value to achieve image alignment. Thus, by introducing a preset alignment database, the design layout and the image alignment database are directly linked, improving the accuracy of alignment between the real-world image and the design layout.

[0217] In addition, this application also provides an image precision alignment device, comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of the image precision alignment method described above.

[0218] In addition, this application also provides a readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the method for precise image alignment as described above.

[0219] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for accurate image alignment, characterized in that, The method comprises: acquiring a real optical image of a wafer to be aligned; determining a target Kernel image corresponding to the real optical image; determining a target alignment template image corresponding to the target Kernel image based on a preset alignment database; the preset alignment database is constructed based on a design layout; the target alignment template image is one of all Target alignment images in the preset alignment database; determining a translation deviation value corresponding to the target Kernel image based on the target alignment template image; compensating for all pixels in the target Kernel image based on the translation deviation value to achieve image alignment; the preset alignment database is constructed by the following method: constructing a target layout corresponding to the wafer to be aligned; segmenting the target layout to obtain N alignment Kernel images; determining a Target feature image in each alignment Kernel image and a feature coordinate corresponding to the Target feature image according to a preset judgment basis; collecting an alignment real image corresponding to each feature coordinate; rendering each Target feature image based on the gray scale information of the alignment real image to obtain a rendering image corresponding to each Target feature image; positionally matching each rendering image with the corresponding alignment real image to obtain a Target alignment image corresponding to each rendering image; constructing a preset alignment database based on the feature coordinates and the Target alignment images.

2. The method of claim 1, wherein, The construction of the target layout corresponding to the wafer to be aligned comprises: acquiring an original design layout corresponding to the wafer to be aligned; extracting a target layer corresponding to the target process from the original design layout based on the target process; constructing a target layout based on the target layer.

3. The method of claim 2, wherein, When the number of target layers is greater than or equal to two, the construction of the target layout based on the target layer comprises: merging the data corresponding to each target layer; re-numbering and naming the merged target layers to obtain a target layout.

4. The method of claim 1, wherein, The method further comprises: determining a translation coefficient and a scaling coefficient between a first coordinate system corresponding to the target layout and a second coordinate system corresponding to the wafer to be aligned; The collection of an alignment real image corresponding to each feature coordinate comprises: determining a collection point coordinate corresponding to each feature coordinate based on the translation coefficient and the scaling coefficient; collecting an alignment real image corresponding to each feature coordinate based on the collection point coordinate.

5. The method of claim 4, wherein, The determination of a translation coefficient and a scaling coefficient between a first coordinate system corresponding to the target layout and a second coordinate system corresponding to the wafer to be aligned comprises: determining a first target point and a second target point on the wafer to be aligned, and determining a first point coordinate corresponding to the first target point and a second point coordinate corresponding to the second target point based on the second coordinate system; determining a third target point corresponding to the first target point and a fourth target point corresponding to the second target point on the target layout; determining a third point coordinate corresponding to the third target point and a fourth point coordinate corresponding to the fourth target point based on the first coordinate system; The translation coefficient and the scaling coefficient between the first coordinate system and the second coordinate system are determined based on a coordinate system mapping formula, by using the first point coordinate, the second point coordinate, the third point coordinate and the fourth point coordinate.

6. The method of claim 1, wherein, The Target feature map in each alignment Kernel image and the feature coordinates corresponding to the Target feature map are determined according to preset judgment basis, including: Based on the uniqueness principle, determine the candidate subgraphs in each alignment Kernel image; According to the preset judgment basis, the target subgraph is selected from the candidate subgraphs corresponding to each alignment Kernel image as the Target feature map in each alignment Kernel image; Determine the feature coordinates of the Target feature map corresponding to each alignment Kernel image.

7. The method of claim 6, wherein, The target subgraph is selected from the candidate subgraphs corresponding to each alignment Kernel image as the Target feature map in each alignment Kernel image according to the preset judgment basis, including: At least one of the distance of the candidate subgraph from the center of the alignment Kernel image, the number of polygons in the size range of the candidate subgraph, and the area proportion of the polygons in the candidate subgraph is used as the preset judgment basis to construct a scoring function; The candidate subgraph with the highest score is selected from each candidate subgraph corresponding to each alignment Kernel image based on the scoring function, and is recorded as the target subgraph and is used as the Target feature map in each alignment Kernel image.

8. The method of claim 6, wherein, The feature coordinates of the Target feature map corresponding to each alignment Kernel image are determined, including: Determine the center point coordinates of the Target feature map corresponding to each alignment Kernel image; The center point coordinates of the Target feature map corresponding to each alignment Kernel image are used as the feature coordinates of the Target feature map corresponding to each alignment Kernel image.

9. The method of claim 7, wherein, When the distance of the candidate subgraph from the center of the alignment Kernel image is used as the preset judgment basis, the different regions in the alignment Kernel image where the candidate subgraph falls are used as the scoring function.

10. The method of claim 7, wherein, When the number of polygons in the size range of the candidate subgraph is used as the preset judgment basis, the expression of the scoring function is: ； In the formula, S2 is the scoring function, A and B are two specific numerical values, and x is the number of polygons.

11. The method of claim 7, wherein, When the area proportion of the polygons in the candidate subgraph is used as the preset judgment basis, the expression of the scoring function is: ； ； wherein is the total area of the polygons within the candidate tile, is the total area of the candidate tile, is the area fraction, is the score function.

12. An image precision alignment device, comprising: Including: An acquisition module is configured to acquire a real optical image of a wafer to be aligned; A first determination module is configured to determine a target Kernel image corresponding to the real optical image; A second determination module is configured to determine a target alignment template corresponding to the target Kernel image based on a preset alignment database; The preset alignment database is constructed based on a design layout; The target alignment template is one of all Target alignment images in the preset alignment database; A third determination module is configured to determine a translation deviation value corresponding to the target Kernel image based on the target alignment template; A compensation alignment module is configured to compensate all pixels in the target Kernel image based on the translation deviation value, so as to realize image alignment. The preset alignment database is constructed by the following method: Constructing a target layout corresponding to the wafer to be aligned; Segmenting the target layout to obtain N alignment Kernel images; According to a preset judgment basis, determining a Target feature map in each alignment Kernel image and a feature coordinate corresponding to the Target feature map; Collecting an alignment actual image corresponding to each feature coordinate; Rendering each Target feature map based on the gray information of the alignment actual image to obtain a rendering image corresponding to each Target feature map; Matching each rendering image with a corresponding alignment actual image to obtain a Target alignment image corresponding to each rendering image; Constructing a preset alignment database based on the feature coordinate and the Target alignment image.

13. A device for precise image alignment, characterized in that, Comprise: A memory for storing a computer program; A processor for executing the computer program to realize the steps of the image accurate alignment method according to any one of claims 1 to 11.

14. A readable storage medium, characterized by, The computer program is stored on the readable storage medium, and the computer program is executed by the processor to realize the steps of the image accurate alignment method according to any one of claims 1 to 11.

Images