Wafer alignment method for semiconductor devices

Abstract

The application provides a wafer alignment method of a semiconductor device, comprising: wafer up; collecting a template image and obtaining a binary edge image, obtaining a first projection along a horizontal direction and a second projection along a vertical direction of the binary edge image, obtaining a first peak and a second peak in the first projection and the second projection, judging whether the positions of the first peak and the second peak correspond to a center region of the image in the binary edge image, if yes, determining a first template with a preset first size in the template image according to the positions of the first peak and the second peak, if no, moving the wafer based on a preset displacement, re-collecting the template image, until the positions of the first peak and the second peak correspond to the center region; completing primary wafer alignment; completing wafer alignment of all levels; generating a wafer alignment work menu according to the results of wafer alignment of all levels; wafer down, wafer up of the same type, and performing wafer alignment of the same type of wafers based on the wafer alignment work menu.

Description

Technical Field

[0001] This invention relates to the field of semiconductor equipment, and more particularly to a wafer alignment method for semiconductor equipment. Background Technology

[0002] Large-scale integrated circuit (IC) manufacturing and testing processes require the use of semiconductor equipment, and after wafer mounting, the equipment typically needs to perform wafer alignment (WA). In existing technologies, such as... Figure 1As shown, the aforementioned semiconductor device 100 typically includes a basic Equipment Front End Module (EFEM) 110 for wafer loading and unloading, housing a wafer cassette 112 containing a wafer pre-aligner 113 and a robotic arm 114 for loading and unloading (some devices also have one or more robotic arms within a vacuum chamber). The device also includes a mechanical motion platform 115, typically movable in the X, Y, and Z directions and rotating around the Z direction, on which a wafer tray (e-chuck, not shown) can be placed to hold the wafer 120. The semiconductor device further includes an optical microscope (OM) 130, which has lower resolution but a larger field of view (FOV), typically used for primary wafer alignment. During subsequent advanced wafer alignment, the OM130 can be switched to a higher magnification for imaging. More commonly, imaging is performed using the core components 131 of a higher-resolution inspection or measurement device, such as the Time Delayed Integration (TDI) system in a wafer defect inspection device (including bright-field and dark-field devices), or the Scanning Electron Microscopy (SEM) system in an electron beam inspection (EBI) wafer defect detection system. These devices each have different core tasks. Additionally, the semiconductor device 100 includes a computer 140 and software 141 running on it, including a user interface (GUI), system software and algorithm software, hardware control, and communication modules. The devices listed above are all front-end IC manufacturing equipment, with common WA levels at level 3. The semiconductor device also includes back-end IC manufacturing equipment such as probers and dicing equipment, with common WA levels using the OM130 at magnification levels 1-2. The reason why the above equipment needs to perform wafer alignment is as follows: After the wafer is pre-aligned by EFEM 110, it is mounted on the mechanical motion platform 115. The position of the wafer center is not precisely determined. For example, the positional deviation between the actual position and the predetermined position can be within 200μm. The orientation is also not precisely determined. For example, the orientation deviation between the actual position and the predetermined position can be within ±1°. Therefore, wafer alignment must be performed to achieve the accuracy required for the equipment to perform its core tasks such as detection or measurement.

[0003] In addition, semiconductor equipment operation includes two parts: creating a work menu (Recipe) and executing the Recipe. The Recipe includes the core task Recipe and the preparatory work Recipes required before performing the core task. These preparatory work includes the WA Recipe (wafer alignment work menu) and other preparatory tasks. Industry practice typically involves referring to... Figure 2A When creating a WA Recipe, a template image 201 is acquired in an area near the center of the wafer. A template 202 is then selected from this image as a sub-image, unique within the image, and its brightness and contrast must meet predetermined requirements. The wafer is then moved 115 meters by one or more dies, and a target image 204 is acquired. Template 202 is then used to perform template matching within this target image, searching for the optimal matching position 205. (Reference) Figure 2B The figure illustrates a template image acquisition region 206 on the wafer, used to acquire template images at a higher magnification (higher than that used for primary wafer alignment, such as the magnification used for secondary wafer alignment). This region includes a template region 207 (the template obtained from the template image is the sub-image corresponding to this template region 207). Both template image acquisition region 206 and template region 207 are regions on the wafer, not images. The template image can be a higher-magnification OM image or a more common non-OM image, such as the TDI or SEM image (later, when performing wafer alignment at other levels besides primary wafer alignment, wafer alignment can be based on SEM imaging, but it can also be based on imaging from other high-magnification imaging systems such as TDI). Many template matching algorithms are commonly used, such as image similarity algorithms, including Normalized Cross Correlation (NCC) algorithms or feature-based template matching methods, achieving sub-pixel accuracy. Additionally, the grains on the wafer are arranged in the X and Y directions (defined by the grain arrangement orientation, such as...). Figure 2C As shown, the wafer coordinate system typically uses the grain alignment direction as the two coordinate axes. Therefore, the grain alignment direction is usually the Xw and Yw directions of the wafer coordinate system. The coordinate axes of the mechanical motion platform coordinate system are the Xs and Ys directions. The Xs and Xw directions can be parallel or non-parallel, and the Ys and Yw directions can also be parallel or non-parallel. The mechanical motion platform coordinate system and the wafer coordinate system have a fixed transformation relationship (e.g., transformation matrix) and fixed dimensions, strictly speaking, referring to periods. The spaces between grains are represented by streets, and there are industry-standard dimensions. (Reference) Figure 2CThe process involves progressively acquiring more target image acquisition positions located further away from the wafer center, along with the target images and their corresponding matching positions. The figure illustrates different matching positions (i.e., different matching points) along multiple dies in the same row (column), such as matching points 211, 212, 213, 214, 215, and 216. These matching points represent successfully matched points. A straight line is fitted using these points to obtain the wafer orientation angle θ, which is then corrected (e.g., by rotating the wafer in the opposite direction or determining the transformation matrix between the wafer coordinate system and the mechanical motion platform coordinate system). There are various methods for selecting target image acquisition positions on the wafer during wafer alignment, and these methods are not limited to... Figure 2C In the image acquisition process, the positions are all in the horizontal direction, i.e., the X-axis. For example, positions 217, 218, etc. on the wafer can also be used as target image acquisition positions. Each matching position is obtained through template matching, ultimately achieving wafer alignment. Finally, the template (which may also include a template image) and each successfully matched position are saved in the WA Recipe. When the semiconductor device is working, after the wafer is mounted, the WA Recipe is executed first. The steps in the WA Recipe execution are the same as when the Recipe was created, using the template saved in the Recipe to the various matching positions saved in the Recipe (when creating the Recipe, the target image acquisition position is determined based on the die periodicity; when executing the WA Recipe, the matching position saved in the working menu is the target image acquisition position). For example, based on... Figure 2C The matching positions 211 to 216 acquire target images and perform wafer alignment to obtain matching positions that meet the threshold conditions. Together, they determine the current wafer orientation angle θ, and then perform the correction (e.g., angle correction or determination of the transformation matrix). Then, it proceeds to the next level of wafer alignment. The same method is used until all levels of wafer alignment are completed. Then, the device completes the remaining preparation work (if necessary). After the preparation is completed, the device can perform its core tasks such as detection or measurement tasks.

[0004] However, the aforementioned prior art has some serious problems, including:

[0005] 1. The creation of WA Recipes relies on manual selection of templates for template matching at each level of WA. Operators lacking experience or relevant knowledge have a high probability of making errors. If the equipment performs wafer alignment on the production line according to the work menu, the addition of other adverse factors, including equipment drift within a limited range, noise, and changes in the local physical properties of the wafer, further increases the likelihood of wafer alignment failure or errors. If WA fails when the equipment executes a Recipe on the production line, it can lead to significant time and economic losses, especially for online equipment such as the aforementioned inspection equipment, where the consequences are severe.

[0006] 2. Currently, an industry trend is for semiconductor equipment to have the capability to generate work recipes offline, meaning that work recipes can be created without occupying production line machine time. The biggest, and perhaps only, obstacle to offline recipe generation is the generation of wafer alignment work menus. This is because it requires the equipment to acquire template images used for wafer alignment at each stage, as well as target images, to perform template matching and determine the matching position. During this time, the equipment cannot perform other tasks, resulting in significant negative impacts on the production line both operationally and economically. Summary of the Invention

[0007] The purpose of this invention is to provide a wafer alignment method for semiconductor devices, which solves the problems of high error risk caused by the reliance on manual selection of templates for wafer alignment and the inability to generate wafer alignment working menus offline in the prior art.

[0008] To achieve this objective, the present invention provides a wafer alignment method for a semiconductor device, comprising:

[0009] Mount the wafer;

[0010] A template image is acquired at a preset position on the wafer. Edge extraction is performed on the template image to obtain an edge image. The edge image is binarized and segmented based on a preset threshold to obtain a binarized edge image. A first projection along the horizontal direction and a second projection along the vertical direction of the binarized edge image are obtained. A first peak satisfying a preset condition is obtained in the first projection, and a second peak satisfying the preset condition is obtained in the second projection. It is determined whether the positions of the first peak and the second peak both correspond to a preset image center region in the binarized edge image. If yes, a first template with a preset first size is determined in the template image based on the positions of the first peak and the second peak. If no, the wafer is moved based on a preset displacement, and the template image is reacquired at the moved position. The binarized edge image, the first peak, the second peak, and the judgment are reacquired until the positions of the first peak and the second peak both correspond to the image center region.

[0011] At least one target image acquisition position is determined on the wafer according to the periodicity of the grains. The wafer is moved to the target image acquisition position to acquire the target image and the first template is used for template matching to obtain the first matching point. The primary wafer alignment is completed according to the position of the first template and the first matching point.

[0012] For all other wafer alignment levels, the process includes: obtaining the template image acquisition position for the current wafer alignment from the template image or successfully matched target image in the previous wafer alignment level; acquiring the template image for the current wafer alignment level and determining a second template with a preset second size from it; determining the target image acquisition position in the current wafer alignment level according to the periodicity of the die; acquiring the target image and performing template matching with the second template to obtain a second matching point; and then completing the current wafer alignment level based on the positions of the second template and the second matching point.

[0013] Generate a wafer alignment working menu based on the results of all levels of wafer alignment;

[0014] The wafer is unloaded, and a similar wafer is loaded onto the wafer. Wafer alignment of similar wafers is performed based on the wafer alignment menu.

[0015] The beneficial effects of this invention are:

[0016] 1. Implemented a wafer alignment function that eliminates the need for manual template selection, avoiding the risk of errors when selecting templates when manually creating a wafer alignment menu;

[0017] 2. It eliminates the need to occupy valuable time of semiconductor equipment, especially in-line semiconductor equipment, enabling semiconductor equipment to generate wafer alignment work menus offline. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of a semiconductor device in the prior art;

[0019] Figure 2A This is a schematic diagram of a template image and a target image used for wafer alignment in the prior art;

[0020] Figure 2B This is a schematic diagram of a template image acquisition area and a template area on a wafer during wafer alignment in a prior art.

[0021] Figure 2C This is a schematic diagram of the matching positions on the wafer during a certain level of wafer alignment in a prior art.

[0022] Figure 3 This is a flowchart of the wafer alignment method in an embodiment of the present invention;

[0023] Figure 4 This is a schematic diagram of the projection of the binarized edge image corresponding to the primary WA in the X and Y directions in an embodiment of the present invention;

[0024] Figure 5A This is a schematic diagram of the field of view corresponding to the primary WA at different magnifications on the wafer in an embodiment of the present invention;

[0025] Figure 5B This is a schematic diagram of the template image acquisition area of ​​the primary WA on the wafer in an embodiment of the present invention;

[0026] Figure 5C This is a schematic diagram of the projection of three different binarized edge images in an embodiment of the present invention;

[0027] Figure 6A This is a schematic diagram of the template image acquisition area on the wafer during primary WA template image acquisition in an embodiment of the present invention;

[0028] Figure 6B This is another schematic diagram of the template image acquisition area on the wafer during primary WA template image acquisition in an embodiment of the present invention;

[0029] Figure 6C This is another schematic diagram of the template image acquisition area on the wafer during the primary WA acquisition of the template image in an embodiment of the present invention;

[0030] Figure 7A This is a schematic diagram of obtaining candidate templates based on the first peak in the first projection along the Y direction, as described in an embodiment of the present invention.

[0031] Figure 7B This is a schematic diagram of obtaining candidate templates based on the second peak in the second projection along the X direction, as described in an embodiment of the present invention.

[0032] Figure 8 This is a schematic diagram illustrating a method for determining the acquisition location of a higher-level WA template image within a template image of a certain level WA in an embodiment of the present invention;

[0033] Figure 9 This is a schematic diagram of the projection of a template image of an advanced WA and its corresponding binarized edge image in an embodiment of the present invention. Detailed Implementation

[0034] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0035] The following is an explanation of some of the technical terms used throughout this application:

[0036] Wafer coordinate system: In this scheme, all wafers are patterned wafers. The wafer coordinate system of patterned wafers is typically defined by the grain orientation, for example... Figure 2C As shown, the horizontal and vertical directions of the grain arrangement are taken as the coordinate axes in the wafer coordinate system, which are the mutually perpendicular Xw and Yw directions. The origin of the wafer coordinate system can be a predetermined mark on the wafer or a grain pattern, such as a corner of the grain (as long as it is relatively meaningful).

[0037] Mechanical motion platform coordinate system: The coordinate axes of the mechanical motion platform coordinate system are mutually perpendicular Xs and Ys directions. The Xs and Xw directions may be parallel or non-parallel, and the Ys and Yw directions may be parallel or non-parallel. The mechanical motion platform coordinate system and the wafer coordinate system have a fixed transformation relationship, such as a transformation matrix.

[0038] Image coordinate system: Imaging systems (such as OM or SEM) can acquire images. The two coordinate axes of the image coordinate system are the horizontal and vertical directions of the image. The horizontal direction of the image is parallel to the upper and lower image boundaries, and the vertical direction of the image is parallel to the left and right image boundaries. If the imaging system, such as OM, is properly calibrated, the image coordinate system should be aligned with the wafer coordinate system. This alignment means that the coordinate axes are parallel. The origin of the image coordinate system can be the upper left corner (most common), the lower left corner, or the center of the image, but is not limited to these.

[0039] Template image acquisition location: The location on the wafer used to acquire template images. Imaging systems (such as OM or SEM) acquire template images based on the coordinates of this location (usually coordinates in the wafer coordinate system).

[0040] Template image acquisition area: The area on the wafer used to acquire template images. This area is the area on the wafer that includes the field of view (FOV) of the imaging system (such as OM or SEM) and the template image acquisition position. For example, the above-mentioned template image acquisition area can be obtained on the wafer with the template image acquisition position as the center and the field of view of the imaging system as the range.

[0041] refer to Figure 3 This invention provides a semiconductor device wafer alignment method, comprising the following steps S1-S5:

[0042] S1. Mount the wafer;

[0043] This refers to loading a wafer. The purpose of loading this wafer is to perform wafer alignment without a known working menu. Furthermore, it can be further improved by creating a wafer alignment working menu based on this wafer. For similar wafers, subsequent execution of this wafer alignment working menu can achieve wafer alignment for the same type of wafer, a one-time solution. During its loading process, it undergoes pre-alignment by the pre-aligner 113 in EFEM 110 before reaching the mechanical motion platform 115. It should be noted that at this point, some basic information about the wafer is known, including the wafer size (e.g., diameter of 300mm, 200mm, or 150mm) and die size, at least their theoretical values ​​are known.

[0044] S2. Complete primary wafer alignment;

[0045] The next step is to perform primary (i.e., first-stage) wafer alignment, which is usually performed using an optical microscopy (OM) system with a large field of view (FOV)130, but can also be performed using other imaging systems such as SEM. The most crucial step is to first obtain the template for wafer alignment in the first-stage WA, and then complete the primary WA based on the known grain period.

[0046] In this embodiment of the invention, step S2 includes: acquiring a template image at a preset position on the wafer; performing edge extraction on the template image to obtain an edge image; performing binarization segmentation on the edge image based on a preset threshold to obtain a binarized edge image; acquiring a first projection of the binarized edge image along the horizontal direction and a second projection along the vertical direction; acquiring a first peak satisfying a preset condition in the first projection; acquiring a second peak satisfying the preset condition in the second projection; determining whether the positions of the first peak and the second peak both correspond to a preset image center region in the binarized edge image; if yes, determining a first template with a preset first size in the template image based on the positions of the first peak and the second peak; if no, moving the wafer based on a preset displacement; re-acquiring the template image at the moved position; re-acquiring the binarized edge image, the first peak, the second peak, and performing the determination again, until the positions of the first peak and the second peak both correspond to the image center region;

[0047] At least one target image acquisition position is determined on the wafer according to the periodicity of the grains. The wafer is moved to the target image acquisition position to acquire the target image and the first template is used for template matching to obtain the first matching point. The primary wafer alignment is completed according to the position of the first template and the first matching point.

[0048] As can be seen from the above, the first template in the primary wafer alignment is automatically acquired, rather than manually, thus achieving automatic template acquisition. Therefore, the primary wafer alignment in this embodiment can also be called primary automatic wafer alignment. Due to the aforementioned wafer loading error, automatically implementing primary automatic wafer alignment is the most challenging aspect. In this embodiment, the preset position is the wafer center position, but the wafer loading error is permissible at the wafer center position.

[0049] In this embodiment of the invention, the horizontal direction is parallel to the upper and lower image boundaries of the binarized edge image, the vertical direction is perpendicular to the horizontal direction, and the vertical direction is parallel to the left and right image boundaries of the binarized edge image. That is to say, the above horizontal and vertical directions are the directions of two coordinate axes in the image coordinate system.

[0050] In this embodiment of the invention, step S2 includes steps S21-S22, and step S21 includes steps S211-S215. These steps will be described in detail below.

[0051] S21. Obtain the first template;

[0052] The following are the steps for obtaining the first template in the primary wafer alignment. This is a dynamic template acquisition method, including the following steps S211-S215:

[0053] S211. Acquire template image;

[0054] The wafer (i.e., the moving mechanical motion platform, hereinafter referred to as the moving platform) is moved to the center position of the wafer to acquire the first frame OM image, which serves as the template image. Before the primary WA, the wafer center and orientation inevitably have some uncertainty. For example, as mentioned in the background art, the uncertainty of the wafer center position after loading is in the range of approximately 200 μm, and the wafer orientation uncertainty is in the range of ±1°. These are common situations, and they are all taken into consideration in the embodiments of this invention (a 200 μm offset in the first frame OM image does not affect the wafer alignment method in the embodiments of this invention). It should also be noted that the first frame image is not limited to the OM image. Although using the OM image is the most common, in fact, as long as the field of view (FOV) of the imaging system used in the primary WA is large enough (similar to the FOV of the primary WA OM image), it is acceptable.

[0055] S212. Obtain a binarized edge image and its first and second projections. Obtain a first peak that meets a preset condition in the first projection and a second peak that meets the preset condition in the second projection.

[0056] Edge extraction

[0057] Image processing is performed on the template image, including edge detection. Specifically, an edge detection kernel, such as the first or second derivative of a Gaussian function, is convolved with the current template image, and the absolute value is then taken to obtain the edge image, as shown in the following formula:

[0058]

[0059] in, This represents the gradient operation or the Laplacian operation, where G is a Gaussian function or a low-pass filter function similar to a Gaussian function. Let G be the symbol for convolution operation, I be the template image, and E be the edge image. The shape parameters of G (primarily the 2σ scale) are determined empirically or experimentally. It's crucial that any template contains sufficient features, and image edges are its primary dependent feature. The template needs sufficient features in both the X and Y directions, or in two non-parallel directions. Typically, IC circuits on a wafer are horizontally and vertically aligned, i.e., in the Xw and Yw directions. As mentioned earlier, the two axes of the wafer coordinate system are usually parallel to the two axes of the image coordinate system. Therefore, the Xw and Yw directions here usually refer to the two axes of the image coordinate system (as mentioned earlier, in the image coordinate system, the horizontal direction, i.e., the X direction, is parallel to the upper and lower image boundaries, and the vertical direction, i.e., the Y direction, is parallel to the left and right image boundaries). Furthermore, the edges originate from grayscale gradients; strong edges necessarily have high grayscale contrast, ensuring the uniqueness of the template within the image—a necessary and sufficient condition for template formation.

[0060] Image binarization

[0061] In this embodiment of the invention, the edge image is first binarized, and then the image projection described later is performed based on the binarized edge image.

[0062] In this embodiment of the invention, the edge image E is binarized and segmented using a preset threshold. The purpose is to select the feature-rich (edge) portion from the image. Many methods exist for obtaining the preset threshold, such as algorithms based on fixed thresholds, algorithms that automatically search for thresholds like Otsu's, and algorithms based on region growing. In one embodiment of the invention, two or more methods can be used simultaneously. For example, one algorithm can be used to obtain the preset threshold for the edge image E, and then another algorithm can be used to verify the preset threshold. Note that the result is not the binarization result of a typical grayscale image, but rather an edge binary image (divided into edge pixels and non-edge pixels; for example, the image content of edge pixels is 1, and the image content of non-edge pixels is 0), which can be called a binarized edge image. In addition, in this embodiment of the invention, after binarizing the edge image, there is a step of "removing small edge regions", that is, using pixel clustering or connected domain detection algorithms to filter out small connected regions. For example, for connected regions of the binarized edge image obtained by the 8 connected domain detection algorithm, if the area of ​​the connected region is less than a fixed number, such as 16 pixels, it is discarded.

[0063] Image projection

[0064] Then, two projections of the aforementioned binarized edge image along the X and Y directions (i.e., the horizontal and vertical directions in the image coordinate system) are obtained. Obtaining the projection along the X direction involves summing the binarized edge images along the X direction to obtain the first projection, and obtaining the projection along the Y direction involves summing the binarized edge images along the Y direction to obtain the second projection. Clearly, the edge image obtained after the above threshold filtering includes the edges of the patterns in the template image (the patterns originate from the boundaries of the grains themselves, or possibly from structural boundaries such as lines within the grains), and these edges correspond to stronger peaks in the projections. Figure 4 For example, the region corresponding to the binarized edge image 401 mentioned above is exactly in the four-grain corner region of the wafer. It should be noted that... Figure 4 The binarized edge image 401 contains the corners of four adjacent grains (this is just an example to illustrate the image projection method; the binarized edge image is not limited to this). Figure 4 (The case in the middle). Then, the binarized edge image 401 is obtained along the horizontal direction, i.e., along the X direction, first projection 402, and along the vertical direction, i.e., along the Y direction, second projection 412.

[0065] In this embodiment of the invention, the projection method is faster than searching for the template directly in the template image. As mentioned earlier, this is possible mainly because the search object is a semiconductor wafer, and the IC circuit layout is mostly horizontal and vertical.

[0066] Obtain peaks in the projection that meet preset conditions

[0067] Then, the peaks that meet the preset conditions in the first projection 402 and the second projection 412 are obtained. The peaks that meet the preset conditions in the first projection are named the first peaks, and the peaks that meet the preset conditions in the second projection are named the second peaks.

[0068] In this embodiment of the invention, the preset conditions include: peak threshold conditions, which include absolute threshold conditions and / or relative threshold conditions, wherein the absolute threshold conditions include: P≥Ta, where P is the peak value and Ta is a preset absolute threshold; the relative threshold conditions include: P / Pmin≥Tr, where P is the peak value, Pmin is the minimum value in the projection (first projection or second projection), and Tr is a preset relative threshold.

[0069] by Figure 4For example, comparing the peak value with the absolute threshold Ta, in the first projection 402, peaks 403, 404, 405, 406, 407, and 408 are all greater than or equal to the threshold Ta; these peaks are all first peaks. Additionally, in the second projection 412, peaks 413, 414, 415, 416, and 417 are all greater than or equal to the threshold Ta; these peaks are all second peaks. In this embodiment of the invention, the value of the absolute threshold Ta is based on the image size combined with experimental experience, and has different values ​​for different image widths, heights (number of pixels), and data types (8 / 10 / 12 / 16 bits, etc.).

[0070] In this embodiment of the invention, Tr can be an empirical value based on experiments, for example, Tr = 3.5.

[0071] In the embodiments of the invention, in addition to the two threshold conditions for peak intensity mentioned above, the peak also satisfies other conditions, including peak shape and position requirements, referred to as peak shape condition and position condition, respectively.

[0072] In an embodiment of the invention, the preset conditions further include: peak shape conditions and position conditions. The peak shape conditions include: the half-width at half-maximum (WHM) of the peak is between a preset upper and lower threshold. The position conditions include: the distance between the peak (position) in the first projection and the upper and lower image boundaries of the binarized edge image in the vertical direction is greater than or equal to GB pixels; the distance between the peak (position) in the second projection and the left and right image boundaries of the binarized edge image in the horizontal direction is greater than or equal to GB pixels, wherein 3 ≤ GB ≤ 9.

[0073] In this embodiment of the invention, the full width at half maximum (FWHM) of the peak is finite, requiring FWHM to be between an upper and lower threshold, i.e., between [FWHMmin, FWHMmax]. The upper and lower thresholds are usually empirical values. This condition is not demanding and is easily met in most practical applications. On the one hand, it helps to exclude peaks with excessively low peak values, as the corresponding regions on the wafer are not suitable template regions. On the other hand, it excludes peaks with excessively high peak values, because such peaks often form plateaus due to multiple close peaks, indicating that the gray levels of pixels in the corresponding region of the template image are relatively similar, lacking gray-level differences, and therefore unsuitable as templates.

[0074] In this embodiment of the invention, the peak position should not be too close to the image boundary; it should be at least gb pixels away from the image edge. Since the template image, edge image, and binarized edge image are from the same source, they have the same size. Therefore, the image edge described here can be the edge of the template image, the edge image, or the edge of the binarized edge image. This is because image processing, including filtering and edge extraction, typically requires some edge pixels of the image, determined by the previous image processing content. For OM images, this is usually only a few pixels, for example, gb = 5 pixels. It should be noted that in this embodiment of the invention, including all subsequent descriptions, peaks within a gb range from the image boundary are unusable and are therefore considered peakless.

[0075] It should be noted that the FOV of an imaging system (such as the OM in this embodiment) may not include the four-grain corner region. For example, it may only include the boundaries of some grains, which is quite probable. This is just an example of the projection method for the binarized edge image. As will be seen later, the method in the embodiment is not limited to the case of the four-grain corner region. That is to say, in this embodiment, the projection of the binarized edge image is not limited to simultaneously including the first peak and the second peak that meet the preset conditions. It is also possible that the first projection does not include the first peak and / or the second projection does not include the second peak. In this case, the number of the first peak and / or the second peak obtained is 0.

[0076] Thus, the first and second peaks that satisfy the above preset conditions are obtained. The first projection includes npx first peaks, and the second projection includes npy second peaks, for a total of np = npx + npy, where npx and npy are greater than or equal to 0. In this embodiment of the invention, 0 ≤ npx ≤ npmax, 0 ≤ npy ≤ npmax, where npmax is an upper limit determined by experience (to save time and speed up the process; more is unnecessary because a template image usually does not necessarily contain and does not need too many templates, often only one is needed). If the number of first or second peaks exceeds npmax, the npmax first peaks or the npmax second peaks with the strongest peak values ​​are selected.

[0077] by Figure 4Taking the preceding steps as an example, the horizontal projection, i.e., the first projection 402, contains six peaks that meet the aforementioned threshold, peak shape, and position requirements, including peaks 403, 404, 405, 406, 407, and 408. The distances of these peaks from the edge of the binarized edge image are all greater than GB (GB is small and not shown in the figure). All six peaks can be used as the first peak. In this embodiment of the invention, npmax = 3 is selected, so the strongest peaks 403, 404, and 405 can be selected from the first projection 402 as the first peaks, resulting in a total of three first peaks. Similarly, in the projection along the vertical direction, i.e., the second projection 412, there are 5 peaks that satisfy the above-mentioned threshold conditions, peak shape conditions, and position conditions, including peaks 413, 414, 415, 416, and 417. The distances from the edges of the binarized edge image are all greater than GB. All 5 peaks can be used as the second peak. In this embodiment, taking npmax = 3 as an example, the three strongest peaks, 413, 414, and 415, are finally selected as the second peaks. Thus, a total of np = 3 + 3 = 6 peaks are available for the next step, namely, the step of determining whether both the first and second peaks correspond to the central region of the image.

[0078] S213. Determine whether the positions of the first peak and the second peak both correspond to the image center region of the binarized edge image;

[0079] It should be noted that the template acquisition method in this embodiment of the invention is dynamic. Its core idea is to determine the subsequent scheme based on the relevant information obtained from the initially acquired template image, including the wafer movement direction and movement step size required to acquire the next frame template image, so as to obtain a qualified first template with fewer wafer movements and image acquisitions.

[0080] In this embodiment of the invention, determining whether the positions of the first peak and the second peak both correspond to a preset image center region in the binarized edge image includes:

[0081] The system determines whether the distances between the position of the first peak and the upper and lower image boundaries of the binarized edge image in the vertical direction are both greater than or equal to a first preset distance, and whether the distances between the position of the second peak and the left and right image boundaries of the binarized edge image in the horizontal direction are both greater than or equal to a second preset distance. The distances between the upper boundary and the upper image boundary, and the lower boundary and the lower image boundary of the image center region are both the first preset distances, and the distances between the left boundary and the left image boundary, and the right boundary and the right image boundary of the image center region are both the second preset distances.

[0082] Therefore, the image center region is determined in the binarized edge image based on the first and second preset distances. In this embodiment of the invention, the image center region is a rectangular region including the center point of the binarized edge image. Furthermore, determining whether the positions of the first peak and the second peak both correspond to the preset image center region in the binarized edge image means determining whether the positions of the first peak and the second peak both fall within the image center region. In this embodiment of the invention, when the distance between the position of the first peak and the upper and lower image boundaries is greater than or equal to the first preset distance, the position of the first peak corresponds to the image center region; when the distance between the position of the second peak and the left and right image boundaries is greater than or equal to the second preset distance, the position of the second peak corresponds to the image center region.

[0083] In this embodiment of the invention, the first preset distance is max(g*FWHMmax1, gc), and the second preset distance is max(g*FWHMmax2, gc), where 2≤g≤5, FWHMmax1 is the maximum value of the full width at half maximum (FWHM) of all the first peaks, and FWHMmax1 is 0 when there is no first peak, FWHMmax2 is the maximum value of the full width at half maximum (FWHM) of all the second peaks, and FWHMmax2 is 0 when there is no second peak, and gc is a preset positive number.

[0084] Various situations in the image

[0085] It should be noted that the die size on a 300mm wafer (in this invention, die size refers to the die period) is typically 2mm to 6mm (single side length). The equipment, in its design, is usually oriented to ensure a sufficiently large field of view (FOV) to accommodate wafer loading errors, while also achieving the highest possible magnification (to reduce the total number of wafer handling stages, thereby reducing time consumption / ensuring equipment throughput, and guaranteeing the success rate of subsequent wafer handling stages). Therefore, the FOV of the optical module (OM) is typically around 1mm to 10mm (camera image width and height are typically 1k pixels or higher), although it can be smaller / larger (as can be achieved with ordinary optical lenses and industrial cameras), while still accommodating most wafers with different die sizes. In short, the common situation is that the two are similar, with FOV < die size, but the difference is not by a factor of several, remaining within the same order of magnitude. (Reference) Figure 5A The diagram shows areas with multiple dies on the wafer, including FOV 501, whose size is close to the die period; it also shows rare FOVs such as FOV 502, whose size is much larger than the die period. Although it is easy to find four-die corner areas after wafer mounting, if the final WA requires high precision, the WA may need more stages, which will increase the time consumption and affect the throughput of the device. Figure 5AThe study also revealed rare FOVs, such as FOV 503, whose size is much smaller than the grain period. Due to the relatively small FOV, finding the four-corner regions after wafer mounting becomes extremely difficult, not only is template matching with known templates challenging, but the maximum error during wafer mounting is also difficult because these two rare cases are usually encountered due to improper equipment configuration or use and are not common. Therefore, we will first consider the more common grain size and FOV relationship.

[0086] refer to Figure 5B , Figure 5B The diagram illustrates different template image acquisition regions determined at different locations on the wafer based on the same FOV, including template image acquisition regions 511-514. Based on the location of the template image acquisition region (i.e., the template image acquisition position), the template image and its corresponding binarized edge image are divided into cases 1-4, as follows:

[0087] Case 1: Both the first projection along the X direction and the second projection along the Y direction contain peaks corresponding to the image center region. In other words, both the first and second peaks correspond to the image center region simultaneously. In this case, the judgment above is correct. As we will see later, there can be multiple peaks satisfying the conditions of Case 1. Case 1 is as follows... Figure 5B As shown in the template image acquisition area 511, both the first projection and the second projection have strong peaks (the grain edge usually constitutes a strong peak, and the strong peaks here are the first peak and the second peak that meet the preset conditions). For example, there are two strong peaks in both the first projection and the second projection.

[0088] This paragraph will provide a specific example of scenario 1. For example... Figures 5B-5C As shown, Figure 5B The template image acquisition area 511 is illustrated. A template image is acquired at the template image acquisition area 511. After edge extraction and binarization processing of the template image, the result is as follows. Figure 5C The binarized edge image 520 shown has the following projections: The first projection 521 is projected along the X direction and has two first peaks satisfying preset conditions, including first peaks 522 and 523; the second projection 524 is projected along the Y direction and has two second peaks satisfying preset conditions, including second peaks 525 and 526. Figure 5CThe diagram illustrates a binarized edge image 520 and its image center, as well as an image center region 527. The vertical distances in the vertical direction between the upper boundary of the image center region 527 and the upper image boundary of the binarized edge image 520, and between the lower boundary of the image center region 527 and the lower image boundary of the binarized edge image 520, are both a first preset distance τ1. The vertical distances in the vertical direction between the left boundary of the image center region 527 and the left image boundary of the binarized edge image 520, and between the right boundary of the image center region 527 and the right image boundary of the binarized edge image 520, are both a second preset distance τ2. Figure 5C It can be seen that the existence of a first peak and a second peak allows both to simultaneously correspond to the image's central region. The first peaks 522 and 523, and the second peaks 525 and 526, all correspond to the image's central region. For example, the positions of the first peaks 522 and 523 are marked with dashed lines in the figure, and the positions of the second peaks 525 and 526 are also marked with dashed lines. In this embodiment of the invention, the fact that the positions of the first peak and the second peak both correspond to the image's central region means that both the positions of the first peak and the second peak fall within the image's central region. The position of the first peak is a vertical coordinate position, and the position of the second peak is a horizontal coordinate position, both within the image coordinate system. It should be noted that if the position of a peak is the same as the boundary of the image's central region—for example, if the position of the first peak is the same as the upper or lower boundary of the image's central region—it also falls within the image's central region.

[0089] This section provides a general overview of cases 2, 3, and 4. Cases 2 through 4 are all non-case 1; in all three cases, the first and second peaks do not simultaneously correspond to the center region of the image. Later, it will be explained that for cases 2 through 4, it is necessary to move the wafer to transform them into case 1.

[0090] Case 2: The projection along the X direction (first projection) contains a first peak, and the projection along the Y direction (second projection) contains a second peak, but neither the first peak nor the second peak simultaneously corresponds to the image center region. The number of the first peak and the second peak is one or more. Case 2 includes two sub-cases: Sub-case 2.1: The first peak exists in both the first and second projections, but a peak corresponding to the image center region exists only in either the first or second projection; sub-case 2.2: The first peak exists in both the first and second projections, but no peak corresponding to the image center region exists in either the first or second projection; neither the first peak nor the second peak corresponds to the image center region.

[0091] This section will introduce a specific example of subcase 2.1. For example... Figures 5B-5C As shown, Figure 5BTwo template image acquisition areas 512 are illustrated. A template image is acquired at one template image acquisition area 512. After edge extraction and binarization processing of the template image, the result is as follows. Figure 5C The binarized edge image 528 shown has the following projections: The first projection 529, projected along the X direction, has two first peaks satisfying preset conditions, including first peaks 530 and 531; the second projection 532, projected along the Y direction, has two second peaks satisfying preset conditions, including second peaks 533 and 534. Figure 5C The diagram illustrates the image center and image center region 535 of the binarized edge image 528, which are... Figure 5C It can be seen that only one first peak 531 corresponds to the central region 535 of the image, while the second peaks 533 and 534 do not correspond to the central region 535 of the image. The positions of the first peaks 530 and 531 and the second peaks 533 and 534 are marked with dashed lines in the figure.

[0092] This paragraph will provide a general overview of scenarios 3 and 4. As discussed earlier, the number of the first peak and / or the second peak may be zero. When the number of the first peak and / or the second peak is zero, the above judgment will undoubtedly be incorrect. Specifically, when the number of either the first peak or the second peak is zero, this corresponds to scenario 3; when the number of both the first peak and the second peak is zero, this corresponds to scenario 4.

[0093] Case 3: The first peak is present only in the projection along the X direction (i.e., the first projection), or the second peak is present only in the projection along the Y direction (i.e., the second projection). The number of the first or second peak can be one or more, and the first or second peak may or may not correspond to the central region of the image. Therefore, Case 3 includes two sub-cases: Sub-case 3.1 is: the first projection has a first peak and the second projection does not have a second peak; Sub-case 3.2 is: the first projection does not have a first peak and the second projection has a second peak.

[0094] This paragraph will introduce a specific example of subcase 3.1. Figure 5B The diagram illustrates two template image acquisition areas 513, according to Figure 5BThe template image acquisition area 513 on the left side can be used to acquire a template image. Edge extraction and binarization are then performed on the template image to obtain a binarized edge image. In this binarized edge image, there is a first peak in the projection along the X direction, i.e., the first projection (the template image acquisition area 513 shows the edges of two grains. Grain edges usually form strong peaks. Both grain edges are parallel to the X direction, so the first projection includes at least one first peak). There is no second peak in the projection along the Y direction, i.e., the second projection (there are no grain edges parallel to the Y direction in the template image acquisition area 513, so there is no second peak formed by the grain edge projection. In addition, it is assumed here that the internal structure of the grain does not form a second peak in the second projection).

[0095] This section will introduce a specific example of subcase 3.2. For example... Figures 5B-5C As shown, Figure 5B The diagram illustrates two template image acquisition areas 513, in Figure 5B A template image was captured at location 513 in the template image acquisition area on the right side of the image. After edge extraction and binarization processing of the template image, the result is as follows: Figure 5C The binarized edge image 536 shown has the following projections: The first projection 537 is projected along the X direction, and the first projection 537 does not contain a first peak satisfying a preset condition; the second projection 538 is projected along the Y direction, and the second projection 538 has two second peaks satisfying the preset condition, including second peaks 539 and 540. Figure 5C The image shows the central region 541, which is... Figure 5C It can be seen that the number of the first peak is 0, and there is no first peak corresponding to the central region of the image. The second peaks 539 and 540 do not correspond to the central region of the image either.

[0096] Case 4: There are no peaks in either the first projection along the X direction or the second projection along the Y direction; that is, the number of both the first and second peaks is 0. For example... Figure 5B As shown in the two template image acquisition areas 514, it can be imagined that the first projection of the binarized edge image corresponding to any template image acquisition area 514 has no first peak and the second projection has no second peak. Similarly, it is also assumed here that the structure inside the grain does not form a first peak in the first projection and does not form a second peak in the second projection.

[0097] The above four cases cover all possible scenarios in the template image and its corresponding binarized edge image.

[0098] Regarding situation 1 above, since the first peak and the second peak both correspond to the central region of the image, the result of the above judgment is "yes", and proceed to step S215 below to search for candidate templates and determine the template.

[0099] For situations 2 to 4 above, if the result of the above judgment is "no", it is necessary to proceed to the next step S214, that is, move the wafer to re-acquire the template image and repeat the above process to proceed to situation 1.

[0100] S214. Move the wafer and then proceed to step S211;

[0101] For cases 2 to 4 above, it is necessary to determine the wafer displacement as a preset displacement, and then move the wafer according to the preset displacement.

[0102] In some embodiments of the present invention, for cases 2 to 4, the preset displacement includes a preset step size and a preset moving direction. Since step S211 is entered after moving the wafer, the template image is re-acquired, and it is determined whether the first and second peaks correspond to the center region of the image. This process is repeated, and by moving the wafer once or multiple times and acquiring the same number of template images, the result of the above determination can be made to be "yes." Then, step S215 is entered to determine the first template. However, this scheme does not consider the differences between cases 2 to 4, wasting easily obtainable information in cases 2 and 3. For cases 2 and 3, the preset displacement described in this paragraph causes cases 2 and 3 to take a considerable amount of time to become case 1, thus making the above determination result "yes."

[0103] In this embodiment of the invention, if no, and when the first projection has a first peak and the second projection has a second peak (i.e., case 2), moving the wafer based on a preset displacement includes:

[0104] Obtain the vertical distance dy1 of the position of the first peak relative to the center point of the binarized edge image, and obtain the vertical movement direction that enables the position of the first peak to move towards the center point;

[0105] Obtain the horizontal distance dx1 of the position of the second peak relative to the center point of the binarized edge image, and obtain the horizontal movement direction that can move the position of the second peak toward the center point;

[0106] The wafer is moved based on the horizontal distance dx1 and the horizontal moving direction, and the wafer is moved based on the vertical distance dy1 and the vertical moving direction.

[0107] like Figure 5CAs shown in the binarized edge image 528, the position of the first peak is a vertical coordinate position, i.e., the coordinate position in the Y direction, and the position of the second peak is a horizontal coordinate position, i.e., the coordinate position in the X direction. The positions of the first peak 530 and the first peak 531 can be obtained from the first projection 529, and the positions of the second peak 533 and the second peak 534 can be obtained from the second projection 532. Taking the first peak 531 and the second peak 533 as examples, if the first peak 531 in the binarized edge image 528 is moved downward along the y direction by a vertical distance dy1 and the second peak 533 is moved to the left along the x direction by a horizontal distance dx1, then the positions of the first peak 531 and the second peak 533 after the movement correspond to the center point in the binarized edge image. That is to say, the intersection point of the first peak 531 and the second peak 533 in the binarized edge image is moved to this center point. At this time, the grain boundary moves to the center of the image in the binarized edge image. In other words, it is only necessary to move the wafer once based on the horizontal distance dx1, the horizontal movement direction, the vertical distance dy1, and the vertical movement direction to change case 2 into case 1, which can save time. Please refer to Figure 6A The template image acquisition area 601 (whose size is the field of view of the imaging system) is the template image acquisition area corresponding to the binarized edge image 528. After the wafer is moved, the area corresponding to the field of view of the imaging system changes, so that the template image acquisition area 601 becomes the template image acquisition area 602. The template image acquisition area 602 includes the four-grain corner area located in the image center area (the image center area is not marked in the figure).

[0108] When the situation changes from case 2 to case 1, in the template image corresponding to the moved binarized edge image, the grain boundary is located at the center of the template image. The center of the template image forms a four-grain corner region as described above. The template image has a large number of grain boundaries. From these grain boundaries, a large number of features (such as edges and corners) that can be used as templates can be obtained, forming a template required for good matching.

[0109] In this embodiment of the invention, if there are multiple first peaks in the first projection, the vertical distance dy1 and the vertical movement direction are obtained using the first peak closest to the image center; if there are multiple second peaks in the second projection, the horizontal distance dx1 and the horizontal movement direction are obtained using the second peak closest to the image center.

[0110] In the actual movement of the wafer, the horizontal distance dx1 and the vertical distance dy1 (both distances in the image coordinate system) need to be converted to the coordinate system of the mechanical motion platform to obtain the converted horizontal distance δx and vertical distance δy. The mechanical motion platform is then moved based on these converted horizontal and vertical distances δx and δy to move the wafer. Furthermore, the horizontal distance dx1 and vertical distance dy1 mentioned later are also transformed using a coordinate transformation matrix P. For example, one conversion formula is as follows:

[0111]

[0112] Where P is a known 3x3 coordinate transformation matrix.

[0113] In this embodiment of the invention, when actually moving the wafer, the vertical movement direction determined from the image is converted to the vertical movement direction in the coordinate system of the mechanical motion platform using a known coordinate transformation matrix, and the horizontal movement direction determined from the image is converted to the horizontal movement direction in the coordinate system of the mechanical motion platform using a known coordinate transformation matrix. This will not be elaborated further here.

[0114] In this embodiment of the invention, if not and when the first projection has a first peak and the second projection does not have a second peak (sub-case 3.1 of case 3), moving the wafer based on a preset displacement includes:

[0115] Obtain the vertical distance dy1 of the position of the first peak relative to the center point of the binarized edge image, and obtain the vertical movement direction that enables the position of the first peak to move towards the center point;

[0116] Obtain the preset horizontal step size dx2 and horizontal movement direction;

[0117] The wafer is moved based on the horizontal step size dx2 and the horizontal moving direction, and the wafer is moved based on the vertical distance dy1 and the vertical moving direction.

[0118] In this embodiment of the invention, dx2 = a * min(FOVx, Dx), 0 < a < 1, FOVx is the field of view size of the imaging system currently used to acquire template images and target images in the horizontal direction, and Dx is the size of a grain in the wafer in the horizontal direction (in pixels).

[0119] Alternatively, if not, and when there is no first peak in the first projection and a second peak in the second projection (i.e., sub-case 3.2 of case 3), moving the wafer based on a preset displacement includes:

[0120] Obtain the horizontal distance dx1 of the position of the second peak relative to the center point of the binarized edge image, and obtain the horizontal movement direction that can move the position of the second peak toward the center point;

[0121] Obtain the preset vertical step size dy2 and vertical movement direction;

[0122] The wafer is moved based on the horizontal distance dx1 and the horizontal moving direction, and the wafer is moved based on the vertical step size dy2 and the vertical moving direction;

[0123] Where dy2 = a * min(FOVy, Dy), 0 < a < 1, FOVy is the field of view size in the vertical direction of the imaging system currently used to acquire template and target images, and Dy is the size of a single grain in the wafer in the vertical direction (in pixels).

[0124] In this embodiment of the invention, similarly, the vertical distance dy1 and horizontal distance dx1 mentioned above need to be converted to the corresponding vertical distance δy and horizontal distance δx in the coordinate system of the mechanical motion platform. The horizontal and vertical movement directions also need to be converted to the corresponding horizontal and vertical movement directions in the coordinate system of the mechanical motion platform.

[0125] Furthermore, since the preset horizontal step size dx2, preset horizontal movement direction, preset vertical step size dy2, and preset vertical movement direction are all preset contents, they can all be contents under the image or wafer coordinate system, or they can all be contents under the mechanical motion platform coordinate system. Undoubtedly, when the preset horizontal step size dx2, preset horizontal movement direction, preset vertical step size dy2, and preset vertical movement direction are all contents under the image or wafer coordinate system, they also need to be converted to the mechanical motion platform coordinate system, as mentioned above, and will not be repeated here.

[0126] This paragraph will introduce a specific example of sub-case 3.2 of case 3. For example... Figure 5C As shown in the binarized edge image 536, the position of the first peak is a vertical coordinate position, i.e., the coordinate position in the Y direction, and the position of the second peak is a horizontal coordinate position, i.e., the coordinate position in the X direction. The first projection 537 does not show the first peak, and the positions of the second peak 539 and the second peak 540 can be obtained from the second projection 538. Taking the second peak 539 as an example, if the second peak 539 in the binarized edge image 536 is moved horizontally to the left by a distance dx1 along the x direction, the position of the moved second peak 539 corresponds to the center point in the binarized edge image. At this time, the two coincide in the horizontal direction. It is also necessary to move the wafer based on the preset vertical step size dy2 and the vertical movement direction. After moving the wafer once or multiple times, the sub-case 3.2 of case 3 becomes case 1, and the result of the above judgment changes from no to yes. For sub-case 3.2, the wafer only needs to be moved once in the x-direction, and moved in the y-direction according to a preset vertical step size dy2 and a preset vertical movement direction. Similarly, for sub-case 3.1, the wafer only needs to be moved once in the y-direction, and moved in the x-direction according to a preset horizontal step size dx2 and a preset horizontal movement direction. Therefore, for case 3, compared to moving the wafer in both the x and y directions with preset step sizes and preset movement directions, time can be saved. Please refer to [reference needed]. Figure 6BThe template image acquisition area 603 (its size is the field of view) is the template image acquisition area corresponding to the binarized edge image 536. After moving the wafer, the area corresponding to the field of view of the imaging system changes, causing the template image acquisition area 603 to become the template image acquisition area 604. The template image acquisition area 604 includes a four-grain corner region located in the image center region (the image center region is not labeled in the figure). Similarly, when changing from case 3 to case 1, the center of the template image forms a four-grain corner region as described above, and its beneficial effects are as described above, and will not be repeated here.

[0127] In this embodiment of the invention, if no, and when there is no first peak in the first projection and no second peak in the second projection (i.e., case 4), moving the wafer based on a preset displacement includes:

[0128] Obtain the preset horizontal step size dx2, vertical step size dy2, horizontal movement direction, and vertical movement direction;

[0129] The wafer is moved based on preset horizontal step size dx2, vertical step size dy2, horizontal movement direction, and vertical movement direction.

[0130] Where dx2 = a*min(FOVx, Dx), dy2 = a*min(FOVy, Dy), 0 < a < 1, FOVx and FOVy are the field of view sizes in the horizontal and vertical directions of the imaging system currently used to acquire template and target images, respectively, and Dx and Dy are the dimensions (in pixels) of a single grain in the wafer in the horizontal and vertical directions, respectively.

[0131] Similarly, after moving the wafer once or multiple times, situation 4 can be transformed into situation 1, such as... Figure 6C As shown, after moving the wafer, the area corresponding to the field of view of the imaging system changes. Taking the wafer moving twice to change situation 4 to situation 1 as an example, the wafer is moved once, so that the template image acquisition area 605 becomes the template image acquisition area 606. The template image is acquired in the template image acquisition area 606. The projection of the binarized edge image corresponding to the template image does not change to situation 1, that is, the result of the above judgment is still negative. Therefore, the wafer is moved again, so that the template image acquisition area 606 becomes the template image acquisition area 607. The template image is acquired in the template image acquisition area 607. The projection of the binarized edge image corresponding to the template image changes to situation 1, that is, the result of the above judgment is positive. The template image and its binarized edge image corresponding to the template image acquisition area 607 form the above situation 1. The template image acquisition area 607 has the four-grain corner region as mentioned above. Its beneficial effects are as mentioned above and will not be repeated here.

[0132] S215. Determine the first template in the template image based on the positions of the first peak and the second peak;

[0133] Search candidate templates

[0134] As mentioned above, when condition 1 is reached, step S215 will be executed. At this time, the image center region of the binarized edge image corresponds to both a first peak and a second peak (both peaks satisfying the initial selection conditions, which will not be emphasized again), and there is at least one first peak and one second peak. In this embodiment of the invention, it is assumed that the number of second peaks mpx satisfies 1 ≤ mpx ≤ mpmax and the number of first peaks mpy satisfies 1 ≤ mpy ≤ mpmax. mpmax is the maximum number of peaks in a single direction, determined by experience. In this embodiment of the invention, the application effect is sufficient when mpmax = 2, and a larger mpmax is also acceptable, for example, mpmax = 4, since it is all software calculation and will not significantly increase the time. In one embodiment, there is no need to filter by mpmax, and all the current first peaks and second peaks are used directly to obtain candidate templates; in another embodiment, after filtering by mpmax, at most mpmax second peaks and mpmax first peaks remain, and then candidate templates are obtained based on the filtered first peaks and second peaks.

[0135] A simple method is to backproject the first and second peaks in the central region of the image onto the binarized edge image. The intersection points form intersection points, resulting in mp2d = mpx × mpy intersection points. Then, using these intersection points as centers and (w, h) as the template size, the template with the richest features (edges and corners) is selected. Furthermore, (w, h) can be varied within a predetermined template size range (wmin, hmin) and (wmax, hmax) to obtain the template with the richest features.

[0136] In this embodiment of the invention, in each direction of X and Y (so as not to miss the best / feature-rich template candidate region in each direction), the mpx second peaks and the mpy first peaks are traversed to perform a full search.

[0137] In this embodiment of the invention, determining a first template with a preset first size in the template image based on the positions of the first peak and the second peak includes:

[0138] Obtain all first candidate templates corresponding to the first peak, wherein, for any first peak, in the binarized edge image, search in the horizontal width according to the position of the first peak according to a preset first step length, first starting point and first ending point to obtain multiple candidate regions with the preset first size, obtain the candidate region with the largest feature concentration among all candidate regions corresponding to a first peak as the first candidate region, and obtain the region in the template image that has the same position and size as the first candidate region as the first candidate template;

[0139] Obtain all second candidate templates corresponding to the second peak. For any second peak, in the binarized edge image, multiple candidate regions with the preset first size are searched in the vertical direction according to the position of the second peak, with a preset second step size, a second starting point, and a second ending point. The candidate region with the highest feature concentration among all candidate regions corresponding to a second peak is obtained as the second candidate region. In the template image, a region with the same position and size as the second candidate region is obtained as the second candidate template.

[0140] Use at least one of the first candidate template and the second candidate template as the first template;

[0141] Wherein, the candidate region at the first starting point shares an edge with the left or right image boundary in the binarized edge image, and the position of the first ending point is obtained based on the first starting point and the first step length, such that the first ending point and the first starting point are located at opposite ends in the binarized edge image, and the candidate region at the first ending point is closest to the right or left image boundary in the binarized edge image; the candidate region at the second starting point shares an edge with the upper or lower image boundary in the binarized edge image, and the position of the second ending point is obtained based on the second starting point and the second step length, such that the second ending point and the second starting point are located at opposite ends in the binarized edge image, and the candidate region at the second ending point is closest to the lower or upper image boundary in the binarized edge image; the feature concentration = feature content within the region / region area.

[0142] refer to Figure 7AThe diagram illustrates a binarized edge image 701, whose vertical projection is a second projection 702. This second projection contains multiple second peaks corresponding to the image's central region (not shown), such as second peaks 703, 705, and 709. Taking second peak 703 as an example, with the position of second peak 703 (i.e., its position xp in the X direction) as the center, a candidate region 704 of size w×h is selected in the binarized edge image 701. The search is conducted along the direction corresponding to the peak, i.e., the orthogonal direction of the projection direction where the peak is located (e.g., along the Y direction in Figure 7), within the image boundary range. The movement, or full-process search, involves a movement step of s pixels (s≥1). For example, when s=1, it requires (Hh) moves, resulting in (H-h+1) candidate regions, where (H-h+1)≥2. H is the height of the binarized edge image. Among the multiple candidate regions, the candidate region with the highest feature (edge ​​and corner) concentration is selected as the second candidate region. The position of the second candidate region is, for example, (xp, xyp), where xyp is the position in the Y direction, and its range is the range centered on this position and determined by a preset first size (including width w and height h). If no candidate region meeting the criteria is encountered during the search process (one of the main reasons being that the edge content contributed in another direction, i.e., the direction orthogonal to the search movement, does not reach the predetermined threshold), the dimensions of width w and / or height h can be changed. For example, the width w can be enlarged to the maximum allowable size, and the height h can be scaled, as long as the width w and height h are within the predetermined ranges (wmin, hmin) and (wmax, hmax), i.e., wmin ≤ w ≤ wmax, hmin ≤ h ≤ hmax, respectively. This continues until a candidate region meeting the feature concentration requirement is found as a second candidate region. The same method can be used for other second peaks in the second projection 702, such as second peaks 705 and 709. Finally, mpx second candidate regions are obtained.

[0143] Then the same method can be used in another direction to search at the first peak corresponding to the image center region (not shown) in the first projection of the binarized edge image 701. For example, refer to... Figure 7B The horizontal projection, i.e., the first projection 701, contains multiple first peaks corresponding to the central region of the image. Taking the first peak 707 as an example, starting from its position yp, a candidate region search is performed throughout the binarized edge image. This includes taking a candidate region 708 of size w×h in the image centered at position yp, along the direction corresponding to the peak, i.e., the orthogonal direction of the projection direction of the peak (i.e., in the direction of the projection direction of the peak). Figure 7BThe search in the X direction (with a step size s) yields the first candidate region, where s ≥ 1. For example, when s = 1, it requires (Ww) moves, resulting in (W-w+1) candidate regions, where (W-w+1) ≥ 2, and W is the width of the binarized edge image. The method is exactly the same as above, ultimately obtaining mpy first candidate regions. Furthermore, in this embodiment of the invention, if a candidate region is close to the image edge, even if its peaks satisfy all the aforementioned conditions, it can still be translated within a limited range.

[0144] Thus, the search results in two directions yield mpx second candidate regions and mpy first candidate regions, resulting in a total of mp = mpx + mpy candidate regions. Since this embodiment searches for the region with the richest features (highest feature concentration) throughout the entire process as the candidate region, it obtains the relatively optimal candidate regions. These candidate regions are all regions in the binarized edge image, and the region corresponding to this candidate region in the template image is called a candidate template. One candidate region in the binarized edge image corresponds to a candidate template in the template image with the same position and size. Therefore, there are a total of mpx second candidate templates and mpy first candidate templates.

[0145] Determine template

[0146] In this embodiment of the invention, at least one of the first candidate template and the second candidate template is used as the first template, including:

[0147] The candidate template with the highest feature concentration is selected from the first candidate template and the second candidate template as the first template.

[0148] In this embodiment of the invention, since the feature concentration is the feature concentration in the candidate region of the binarized edge image, the candidate template with the largest feature concentration refers to the candidate template corresponding to the candidate region with the largest feature concentration, which must satisfy the uniqueness requirement when selecting the template.

[0149] Additionally, it should be noted that in the event of the following two rare situations (as mentioned above, rare situations may occur when the device is misconfigured or used), the countermeasures described above in this embodiment of the invention can also be used to address them.

[0150] Rare Case 1: When FOV > n grain sizes, for example, when n > 2. Figure 5A The field of view is 502. The first template is obtained using the same method described above. The image will now contain multiple qualified templates; selecting one is sufficient. Since the grain period is known, subsequent steps will not result in errors.

[0151] Rare Case 2: The opposite of the above, this occurs when the grain size is greater than m FOVs, and m > 2. In this case, the field of view is equivalent to a much smaller grain size. It's essentially Case 4 above, but requires more wafer movement and template image acquisition. The corresponding countermeasures for Case 4 are then used to finally obtain the first template.

[0152] Therefore, it can be seen that these two rare cases are also covered by the countermeasures in the embodiments of the present invention.

[0153] S22. Acquire the target image and perform template matching to obtain the first matching point. Complete the primary wafer alignment based on the position of the first template and the first matching point.

[0154] In this embodiment of the invention, at least one target image acquisition position is determined on the wafer according to the periodicity of the grains, the wafer is moved to the target image acquisition position to acquire the target image, and the first template is used for template matching to obtain a first matching point. Primary wafer alignment is completed according to the position of the first template and the first matching point.

[0155] Currently, we are still in the current level WA (i.e., primary wafer alignment, not yet switched to a higher level WA). Next, we need to determine the acquisition positions of subsequent target images so that template matching can be performed, thus completing the current level WA. Similar to the method described in the background art, the template image contains a first template, automatically acquired using the above method, located near the wafer center. The target image acquisition positions are, for example, separated from the template image positions by one or more grain cycles in the X direction. The remaining target image acquisition positions are based on the current wafer orientation and are integer multiples of the grain cycle along the X direction, gradually moving outwards from the wafer center, and so on. The number of target images is, for example, one frame, usually at least two frames. At most one first matching point is determined in each target image frame. At least one successfully matched matching position, i.e., at least one first matching point, can be determined in these target images (e.g., at least two first matching points are determined). Based on the position of the first template and the position of the first matching point, the wafer position is corrected (as described in the background art, by rotating the wafer in the opposite direction by a determined angle θ) or the transformation matrix between the wafer coordinate system and the mechanical motion platform coordinate system is determined, thus completing the primary wafer alignment.

[0156] S3. Complete the alignment of the remaining wafers at each stage;

[0157] In this embodiment of the invention, for each of the remaining wafer alignment stages, the process includes: obtaining the template image acquisition position for the current wafer alignment from the template image or successfully matched target image in the previous wafer alignment stage; acquiring the template image for the current wafer alignment stage and determining a second template with a preset second size therefrom; determining the target image acquisition position in the current wafer alignment stage based on the periodicity of the die; acquiring the target image and performing template matching with the second template to obtain a second matching point; and then completing the current wafer alignment based on the positions of the second template and the second matching point.

[0158] In this embodiment of the invention, since it is a fully automatic wafer alignment (WA), preparation for template searching in the primary WA is required for the next level WA. For all other levels of wafer alignment, the template image acquisition position needs to be determined by the previous level WA for the current level WA. Therefore, in this embodiment of the invention, the selection method for the first template is automatic, and the template image acquisition positions required for all other levels of wafer alignment are also automatically selected.

[0159] In this embodiment of the invention, a template image acquisition area with a feature content that reaches or exceeds a preset content threshold within the field of view (FOV) required for the current wafer alignment is obtained from the template image or the successfully matched target image in the previous wafer alignment. The position of the template image acquisition area is used as the template image acquisition position for the current wafer alignment. The template image acquisition position may be one or more.

[0160] For example Figure 8 As shown, the template image 801 acquired during the previous stage WA (such as primary wafer alignment) contains the template 802 used. In regions 803, 804, and 805 that have the field of view required for the current stage wafer alignment (such as second stage wafer alignment), the feature content reaches or exceeds a preset content threshold. These three regions can all be used as template image acquisition regions for the current stage wafer alignment (such as second stage wafer alignment). The position of at least one of these three template image acquisition regions is taken as the template image acquisition position for the current stage wafer alignment (such as second stage wafer alignment).

[0161] In this embodiment of the invention, when performing current-level wafer alignment (such as second-level wafer alignment), a template image is acquired at the template image acquisition position of the current-level wafer alignment, and a second template is determined from it.

[0162] The method for determining the second template can be the same as the method used for searching the primary WA template. This involves extracting edges from the template image, binarizing it to obtain a binarized edge image, and then projecting it to obtain horizontal and vertical projections. This process is repeated until the second template is finally obtained. (Reference) Figure 9The current-stage wafer alignment (e.g., the second-stage wafer alignment) acquires a template image 901. Template image 901 corresponds to a binarized edge image (not shown). Since the field of view (FOV) of the imaging system (e.g., OM or SEM) is smaller than that of the primary wafer alignment, template image 901 contains fewer features (edges and corners). The horizontal projection of its corresponding binarized edge image is the first projection 902, and the vertical projection is the second projection 903. Following the above method, the second template 904 and / or 905 can ultimately be obtained from template image 901. Besides this, other methods can determine the second template, including feature-based template selection methods. In these methods, the second template is extracted from the template image used in the current-stage wafer alignment using a feature extraction algorithm. Because the template image has a higher magnification and a smaller field of view, it contains far fewer features than in the primary WA. Some commercial image processing or machine vision software has corresponding feature-based automatic template selection functions, which are less prone to errors. This is existing technology and will not be elaborated further. Therefore, in this embodiment of the invention, whether the second template is selected by projection method or by feature extraction algorithm, it is an automatic selection method. In other words, the selection method for the second template is also automatic.

[0163] In summary, in the embodiments of the present invention, whether it is the selection method of the first template, the selection method of the template image acquisition position in the other stages of wafer alignment, or the selection method of the second template, all are automatic selection methods. Therefore, the wafer alignment method provided in the embodiments of the present invention can be called the automatic wafer alignment method.

[0164] For any of the remaining wafer alignment levels, the wafer alignment is completed by determining at least one successfully matched position, i.e., at least one second matching point, in the target image based on the second template (the number of target images is, for example, one frame, usually at least two frames, and at most one second matching point can be determined in each frame of the target image, for example, at least two second matching points are determined in these target images). The wafer position is corrected or the transformation matrix between the wafer coordinate system and the mechanical motion platform coordinate system is determined based on the position of the second template and the position of the second matching point, thus completing the wafer alignment of that level.

[0165] It should be noted that the templates used in the remaining wafer alignment stages are collectively referred to as the second templates, which does not mean that the same templates are used in the remaining wafer alignment stages. In fact, those skilled in the art know that the field of view of the imaging system is different in all wafer alignment stages (including primary wafer alignment and other stages), and the field of view decreases step by step (that is, the magnification of the imaging system increases step by step). The template images acquired in all wafer alignment stages are different, and undoubtedly, the templates extracted from the template images are different.

[0166] S4. Generate a wafer alignment working menu based on the results of all levels of wafer alignment;

[0167] Then, the results of each level of wafer alignment are saved, for example, to the database of the device's computer. The results include the templates successfully used for each level of wafer alignment (WA), template images, and successfully matched positions on the wafer, i.e., matching points (including the first matching point corresponding to the primary wafer alignment and the second matching point corresponding to the remaining levels of wafer alignment). This generates a wafer alignment working menu Recipe, so subsequent wafers of the same type do not need to undergo all the above wafer alignment steps, but can execute a predetermined Recipe as in the prior art. Obviously, it can be used on the current semiconductor device, and also on other similar semiconductor devices processing similar wafers. In fact, only the first wafer of all similar wafers needs to be automatically aligned as described above. Once completed, saving the templates, template images, and successfully matched positions from each level of wafer alignment process generates a Recipe for the WA of other similar wafers. This Recipe can be used not only on the current device, but also for similar work (using the same Recipe) on similar devices.

[0168] As can be seen, the method in this embodiment of the invention does not require the prior creation of a WA Recipe, and most importantly, it does not require prior acquisition of any images, including template images, to determine the template required for WA. Upon encountering the first wafer, WA can be completed, and a WA Recipe can be generated for wafers of the same type as the first wafer. Therefore, the WA Recipe can be generated offline, without prior occupation of the equipment. For the first wafer, there is no need to acquire template images beforehand; only basic information related to the first wafer is required, such as wafer size, die size or cycle, WA stages, and configuration parameters such as FOV for each stage. This completely eliminates the need to occupy the equipment beforehand. This differs from existing technologies that require prior generation of images in the WA Recipe using the equipment, saving equipment time. If the equipment's core functions and other preparatory work can also be set using a similar method without occupying equipment time, then the equipment's Recipe can be generated completely offline, which is highly significant. Therefore, the beneficial effects are obvious while the cost is minimal.

[0169] S5. Unmount the wafer, mount similar wafers, and perform wafer alignment for similar wafers based on the wafer alignment menu;

[0170] The wafer is unloaded. Then, similar wafers are mounted, and automatic alignment of similar wafers is performed based on the Recipe. If on a production line, the current wafer is unloaded, and a new wafer is mounted. Since the Recipe generated from the automatic alignment results already exists, there is no need to perform new automatic wafer alignment. Instead, as in the prior art, WA is performed based on a known WARecipe.

[0171] In this embodiment of the invention, the semiconductor equipment includes all semiconductor equipment that requires a wafer inspection (WA), including IC front-end process equipment such as inspection equipment for patterned wafers, such as bright field inspection (BFI) equipment, dark field inspection (DFI) equipment, electron beam inspection (EBI) equipment, electron beam review (EBR) equipment, macro defect inspection equipment, electron beam critical dimension measurement (CD-SEM) equipment, and many IC back-end process equipment such as wafer dicing equipment, probe equipment, etc. These semiconductor devices generally have the following characteristics: Figure 1 The structure shown in the semiconductor device 100 illustrates this. It can be seen that any semiconductor device requiring wafer alignment of patterned wafers can utilize the aforementioned automatic wafer alignment method and thus benefit from it.

[0172] Therefore, the embodiments of the present invention provide a wafer alignment method for semiconductor devices, the beneficial effects of which include: 1. It realizes wafer alignment function without manual template selection, avoiding the risk of errors in template selection when manually creating wafer alignment working menus; 2. It does not occupy the valuable machine time of semiconductor devices, especially online semiconductor devices, enabling semiconductor devices to generate wafer alignment working menus offline; 3. In addition, in the embodiments of the present invention, because the preset displacement of the moving wafer is designed for different situations, a qualified template, i.e., the first template, can be obtained in the fastest time with the fewest wafer movements and image acquisitions during the first implementation, resulting in lower time costs.

[0173] The above embodiments only illustrate the basic principles and characteristics of the present invention. The present invention is not limited to the above embodiments. Without departing from the spirit and scope of the present invention, all other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

Claims

1. A wafer alignment method for a semiconductor device, characterized in that, include: Mount the wafer; A template image is acquired at a preset position on the wafer. Edge extraction is performed on the template image to obtain an edge image. The edge image is binarized and segmented based on a preset threshold to obtain a binarized edge image. A first projection along the horizontal direction and a second projection along the vertical direction of the binarized edge image are obtained. A first peak satisfying a preset condition is obtained in the first projection, and a second peak satisfying the preset condition is obtained in the second projection. It is determined whether the positions of the first peak and the second peak both correspond to a preset image center region in the binarized edge image. If yes, a first template with a preset first size is determined in the template image based on the positions of the first peak and the second peak. If no, the wafer is moved based on a preset displacement, and the template image is reacquired at the moved position. The binarized edge image, the first peak, the second peak, and the judgment are reacquired until the positions of the first peak and the second peak both correspond to the image center region. At least one target image acquisition position is determined on the wafer according to the periodicity of the grains. The wafer is moved to the target image acquisition position to acquire the target image and the first template is used for template matching to obtain the first matching point. The primary wafer alignment is completed according to the position of the first template and the first matching point. For all other wafer alignment levels, the process includes: obtaining the template image acquisition position for the current wafer alignment from the template image or successfully matched target image in the previous wafer alignment level; acquiring the template image for the current wafer alignment level and determining a second template with a preset second size from it; determining the target image acquisition position in the current wafer alignment level according to the periodicity of the die; acquiring the target image and performing template matching with the second template to obtain a second matching point; and then completing the current wafer alignment level based on the positions of the second template and the second matching point. Generate a wafer alignment working menu based on the results of all levels of wafer alignment; The wafer is unloaded, and a similar wafer is loaded onto the wafer. Wafer alignment of similar wafers is performed based on the wafer alignment menu.

2. The method according to claim 1, characterized in that, The preset conditions include: peak threshold conditions, which include absolute threshold conditions and / or relative threshold conditions. The absolute threshold conditions include: P≥Ta, where P is the peak value and Ta is the preset absolute threshold. The relative threshold conditions include: P / Pmin≥Tr, where P is the peak value, Pmin is the minimum value in the projection, and Tr is the preset relative threshold.

3. The method according to claim 1 or 2, characterized in that, The preset conditions include: peak shape conditions and position conditions. The peak shape conditions include: the half-width at half-maximum (WHM) of the peak is between a preset upper and lower threshold. The position conditions include: the distance between the peak in the first projection and the upper and lower image boundaries of the binarized edge image in the vertical direction is greater than or equal to GB pixels, and the distance between the peak in the second projection and the left and right image boundaries of the binarized edge image in the horizontal direction is greater than or equal to GB pixels, wherein 3 ≤ GB ≤ 9.

4. The method according to claim 1, characterized in that, Determining whether the positions of the first peak and the second peak both correspond to a preset image center region in the binarized edge image includes: The system determines whether the distances between the position of the first peak and the upper and lower image boundaries of the binarized edge image in the vertical direction are both greater than or equal to a first preset distance, and whether the distances between the position of the second peak and the left and right image boundaries of the binarized edge image in the horizontal direction are both greater than or equal to a second preset distance. The distances between the upper boundary and the upper image boundary, and the lower boundary and the lower image boundary of the image center region are both the first preset distances, and the distances between the left boundary and the left image boundary, and the right boundary and the right image boundary of the image center region are both the second preset distances.

5. The method according to claim 4, characterized in that, The first preset distance is max(g*FWHMmax1, gc), and the second preset distance is max(g*FWHMmax2, gc), where 2≤g≤5, FWHMmax1 is the maximum value of the full width at half maximum (FWHM) of all the first peaks, and FWHMmax1 is 0 when there is no first peak, FWHMmax2 is the maximum value of the full width at half maximum (FWHM) of all the second peaks, and FWHMmax2 is 0 when there is no second peak, and gc is a preset positive number.

6. The method according to claim 1, characterized in that, Determining a first template with a preset first size in the template image based on the positions of the first and second peaks includes: Obtain all first candidate templates corresponding to the first peak, wherein, for any first peak, in the binarized edge image, search in the horizontal width according to the position of the first peak according to a preset first step length, first starting point and first ending point to obtain multiple candidate regions with the preset first size, obtain the candidate region with the largest feature concentration among all candidate regions corresponding to a first peak as the first candidate region, and obtain the region in the template image that has the same position and size as the first candidate region as the first candidate template; Obtain all second candidate templates corresponding to the second peak. For any second peak, in the binarized edge image, multiple candidate regions with the preset first size are searched in the vertical direction according to the position of the second peak, with a preset second step size, a second starting point, and a second ending point. The candidate region with the highest feature concentration among all candidate regions corresponding to a second peak is obtained as the second candidate region. In the template image, a region with the same position and size as the second candidate region is obtained as the second candidate template. Use at least one of the first candidate template and the second candidate template as the first template; Wherein, the candidate region at the first starting point shares an edge with the left or right image boundary in the binarized edge image, and the position of the first ending point is obtained based on the first starting point and the first step length, such that the first ending point and the first starting point are located at opposite ends in the binarized edge image, and the candidate region at the first ending point is closest to the right or left image boundary in the binarized edge image; the candidate region at the second starting point shares an edge with the upper or lower image boundary in the binarized edge image, and the position of the second ending point is obtained based on the second starting point and the second step length, such that the second ending point and the second starting point are located at opposite ends in the binarized edge image, and the candidate region at the second ending point is closest to the lower or upper image boundary in the binarized edge image; the feature concentration = feature content within the region / region area.

7. The method according to claim 6, characterized in that, Using at least one of the first candidate template and the second candidate template as the first template includes: The candidate template with the highest feature concentration is selected from the first candidate template and the second candidate template as the first template.

8. The method according to claim 1, characterized in that, If not, and when the first projection has a first peak and the second projection has a second peak, moving the wafer based on a preset displacement includes: Obtain the vertical distance dy1 of the position of the first peak relative to the center point of the binarized edge image, and obtain the vertical movement direction that enables the position of the first peak to move towards the center point; Obtain the horizontal distance dx1 of the position of the second peak relative to the center point of the binarized edge image, and obtain the horizontal movement direction that can move the position of the second peak toward the center point; The wafer is moved based on the horizontal distance dx1 and the horizontal moving direction, and the wafer is moved based on the vertical distance dy1 and the vertical moving direction.

9. The method according to claim 1, characterized in that, If not, and when the first projection has a first peak and the second projection does not have a second peak, moving the wafer based on a preset displacement includes: Obtain the vertical distance dy1 of the position of the first peak relative to the center point of the binarized edge image, and obtain the vertical movement direction that enables the position of the first peak to move towards the center point; Obtain the preset horizontal step size dx2 and horizontal movement direction; The wafer is moved based on the horizontal step size dx2 and the horizontal moving direction, and the wafer is moved based on the vertical distance dy1 and the vertical moving direction. Alternatively, if not, and when there is no first peak in the first projection and a second peak in the second projection, moving the wafer based on a preset displacement includes: Obtain the horizontal distance dx1 of the position of the second peak relative to the center point of the binarized edge image, and obtain the horizontal movement direction that can move the position of the second peak toward the center point; Obtain the preset vertical step size dy2 and vertical movement direction; The wafer is moved based on the horizontal distance dx1 and the horizontal moving direction, and the wafer is moved based on the vertical step size dy2 and the vertical moving direction.

10. The method according to claim 1, characterized in that, If not, and when there is no first peak in the first projection and no second peak in the second projection, moving the wafer based on the preset displacement includes: Obtain the preset horizontal step size dx2, vertical step size dy2, horizontal movement direction, and vertical movement direction; The wafer is moved based on preset horizontal step size dx2, vertical step size dy2, horizontal movement direction, and vertical movement direction.

11. The method according to claim 9 or 10, characterized in that, dx2 = a*min(FOVx, Dx), dy2 = a*min(FOVy, Dy), 0 < a < 1, FOVx and FOVy are the field of view sizes in the horizontal and vertical directions of the imaging system currently used to acquire template and target images, respectively, and Dx and Dy are the dimensions of a single die in the wafer in the horizontal and vertical directions, respectively. The horizontal direction is parallel to the upper and lower image boundaries of the binarized edge image, respectively, and the vertical direction is perpendicular to the horizontal direction. The vertical direction is parallel to the left and right image boundaries of the binarized edge image, respectively. The second template is a template extracted from the template image used for wafer alignment in the current stage based on a feature extraction algorithm. The preset position is the center position of the wafer.

Images