Defect detection method and system

Abstract

The invention provides a defect detection method and system. The method comprises the following steps: acquiring defect coordinate point cloud data of a to-be-detected sample; converting the defect coordinate point cloud data into a two-dimensional data point set; dividing the two-dimensional data point set into a plurality of detection areas, wherein the detection areas comprise path points; planning a detection path based on the path points; and performing defect reinspection on the detection area according to the detection path. According to the defect detection method provided by the invention, the defect coordinate point cloud data is converted into the two-dimensional data point set, and the two-dimensional data point set is subjected to region division and clustering processing, so that the detection path is planned. Different from a traditional point-to-point detection mode, the method has the advantages that frequent movement can be avoided, meanwhile, each defect point is accurately detected, the detection efficiency is improved, and the yield of equipment is further improved.

Description

Technical Field

[0001] The present invention relates to the field of wafer defect detection, and more particularly, to a defect detection method and system. Background Art

[0002] In semiconductor manufacturing, a wafer is the basic material for manufacturing chips. Due to the complex wafer manufacturing process, various defects may occur during production. The detection of wafer defects includes primary detection and re-inspection. The primary defect detection tries to find all possible defects on the wafer as comprehensively as possible, but there may be a certain misjudgment rate in the primary detection, or more precise analysis is required for some complex defects. Therefore, it is necessary to re-inspect some defects to determine detailed information such as the true nature, size, and location of the defects.

[0003] Currently, the method for re-inspecting wafer defects mainly targets the defects detected in the primary detection. Some or all of the targeted defects are re-inspected. The designated defects are moved one by one to the detection area for detection, and only one defect is detected at a time. Then, image data is obtained. This process needs to be repeated for each defect point, making the entire detection process long and resulting in a significant increase in time cost. In addition, the detection path for each defect point usually follows the order of moving from left to right first and then from top to bottom. Although this movement rule is simple, it has obvious drawbacks. The point-to-point movement between defects is too frequent, which consumes a large amount of time and energy, seriously affecting the productivity of the defect detection equipment and reducing the overall efficiency of wafer production. Summary of the Invention

[0004] In view of the problems existing in the wafer defect detection method in the above-mentioned prior art, the present application provides a defect detection method, including the following steps:

[0005] Obtain defect coordinate point cloud data of the sample to be tested;

[0006] Convert the defect coordinate point cloud data into a two-dimensional data point set;

[0007] Divide the two-dimensional data point set into several detection regions, where the detection regions contain path points;

[0008] Based on the path points, plan a detection path;

[0009] According to the detection path, re-inspect the defects of the sample to be tested.

[0010] Optionally, converting the defect coordinate point cloud data into a two-dimensional data point set further includes: projecting the defect coordinate point cloud data so that the coordinate values in the height direction of each defect point change to the same value.

[0011] Optionally, the step of dividing the two-dimensional data point set into several detection regions includes:

[0012] Obtain the window size of the camera field of view, including the width of the camera field of view;

[0013] Divide the two-dimensional data point set into the several detection regions, where the width of the detection region is greater than or equal to the width of the camera field of view; and

[0014] Cluster the two-dimensional data points within the detection region.

[0015] Optionally, the step of clustering the two-dimensional data points within the detection region includes:

[0016] Set the clustering window size to be the same as the window size of the camera field of view;

[0017] Determine at least one clustering window within the detection region and perform clustering on the two-dimensional data points within the detection region.

[0018] Optionally, the clustering windows are distributed in parallel within the corresponding detection regions.

[0019] Optionally, the path points are the center points of the clustering windows, and at least one image of the clustering window is obtained when the center of the camera field of view coincides with the path points.

[0020] Optionally, before obtaining the clustering window image, rotate the angle of the clustering window so that the clustering window completely coincides with the detection region.

[0021] Optionally, the step of planning the detection path includes:

[0022] Determine the path points within the several detection regions;

[0023] Determine the detection start point and the detection end point of the path points; and

[0024] Determine the detection order of the path points within the several detection regions according to the detection start point and the detection end point.

[0025] On the other hand, the present application provides a defect detection system, which is characterized by including:

[0026] A moving stage for carrying and moving the sample to be tested;

[0027] A light source for providing illumination light;

[0028] An imaging system including an objective lens, a tube lens, and a camera module, where the imaging system is used to collect a detection image of the detection region of the sample to be tested, and the detection image includes defect coordinate point cloud data of the sample to be tested.

[0029] A computer system projects the defective coordinate point cloud data into two dimensions, divides the two-dimensional projection into several detection regions, clusters the defective data within the detection regions to determine a detection path, and reinspects the defects of the sample to be tested according to the detection path.

[0030] This application also provides a computer-readable storage medium, on which computer instructions are stored. When the computer instructions are executed by a processor, the defective detection method described in any one of the above is implemented.

[0031] As described above, the defective detection method provided by the present invention has at least the following beneficial technical effects:

[0032] The detection efficiency is improved. In this application, the defective coordinate point cloud data is converted into a two-dimensional data point set, the two-dimensional data point set is divided into multiple detection regions, the two-dimensional data points within the detection regions are clustered, and the detection order between multiple detection regions and the clustering groups within the detection regions is planned to find the optimal detection path, avoiding the problem of overly frequent point-to-point movement between defects, reducing the unnecessary movement times of the moving stage and the acceleration and deceleration processes, significantly shortening the overall detection time and the corresponding energy consumption, greatly improving the detection efficiency, and thus improving the productivity of the equipment.

[0033] In terms of the clustering strategy, when clustering the data points within the detection region, the window size of the clustering is set to the window size of the camera field of view, and by reasonably moving the position of the clustering window and the clustering algorithm, the number of clustering groups in each detection region is made as small as possible. Moreover, the utilization rate of the camera field of view is improved, ensuring that all defects within the camera field of view can be included in each acquired image, and one or more defects can be detected after each movement. While reducing the movement times of the moving stage, the imaging ability of the camera is also fully utilized, enhancing the overall detection efficiency.

[0034] In addition, since the detection region contains all data points, all defective points can be comprehensively detected, avoiding omissions, thereby improving the detection accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 Shows a flowchart of a defective detection method that is prior art.

[0036] Figure 2a Shows a conventional detection route in the prior art.

[0037] Figure 2b Shows a detection route after algorithm optimization in the prior art.

[0038] Figure 3Shown is a flowchart of a defect detection method provided in the first embodiment.

[0039] Figure 4 Shown is a schematic diagram of projecting defect coordinate point cloud data provided by the present invention.

[0040] Figure 5 Shown is a schematic diagram of dividing a two-dimensional data point set into a detection area.

[0041] Figure 6 Shown is a schematic diagram of clustering defect points.

[0042] Figure 7 Shown is a schematic diagram of the detection path provided in this embodiment.

[0043] Figure 8a Shown is a scanning schematic diagram when the angle of the detection area is the same as the camera view angle.

[0044] Figure 8b Shown is a scanning schematic diagram when the angle of the detection area is different from the camera view angle.

[0045] Figure 9 Shown is a schematic diagram of the defect detection device provided in the second embodiment.

[0046] Reference Numerals

[0047] 01, moving stage; 02, imaging system; 021, objective lens; 022, tube lens; 023, camera; 03, light source; 04, wafer; 05, computer system; 051, control system; 052, data processing unit; 053, data storage unit; 054, display module. Detailed Embodiments

[0048] The following uses specific specific examples to illustrate the embodiments of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention.

[0049] It should be noted that the diagrams provided in this embodiment only illustrate the basic concept of the present invention in a schematic manner. Although only the components related to the present invention are shown in the diagrams and are not drawn according to the number, shape, and size of the components in actual implementation, the actual shape, quantity, positional relationship, and proportion of each component can be arbitrarily changed on the premise of implementing the technical solution of the present invention, and the component layout form may also be more complex.

[0050] Specifically, as Figure 1As shown, it is a flowchart of a defect detection method in the prior art. It includes the following steps:

[0051] S1: Conduct a preliminary inspection on the sample to be tested to obtain defect coordinate point cloud data;

[0052] S2: Determine the defect points that need to be reinspected: According to the detection results of the preliminary defects, select the defect points that need to be reinspected. Specifically, these defect points may be selected according to specific screening criteria (such as defect type, severity, etc.), or all detected defects may be inspected.

[0053] S3: Plan the detection path: Generally, as Figure 2a shown, it shows a conventional detection route in the prior art; the detection follows the coordinate order to move: in the order of first from left to right and then from top to bottom, move each defect point to be detected to the detection area of the imaging system one by one. Optionally, a more complex algorithm, such as the ant colony algorithm, brute force algorithm, nearest neighbor algorithm, etc., is used to find the shortest path between defect points. As Figure 2b shown, it shows the detection route optimized by the algorithm in the prior art.

[0054] S4: Move the defect points to the detection area of the imaging system one by one according to the planned route. Specifically, first perform the initial position movement: The moving stage 01 moves the wafer in the horizontal (x, y axes) and height (z axis) directions according to the coordinates of the first defect point to be detected selected, so that the defect point is within the field of view of the camera. Secondly, according to the planned route, for each defect point, the moving stage 01 calculates the movement path according to its coordinates, so that the defect point is accurately focused on the imaging plane of the camera.

[0055] S5: The imaging system acquires and processes the image. When the defect moves to the detection area of the imaging system, the camera performs exposure imaging to obtain the image information of the defect point and its surrounding area at one time. During the imaging process, the light source provides illumination, and the light passes through the objective lens and the tube lens and then reaches the two-dimensional sensor array of the camera, is converted into an electrical signal and recorded as a digital image. Then the acquired image is processed to obtain more detailed defect information, providing a basis for subsequent defect classification and statistics.

[0056] S6: Classify, count, and label the defect points. Classification and counting: According to the image processing and analysis results, classify each detected defect point, such as classify and count according to classification criteria such as the type of defect (such as scratches, holes, impurities, etc.), size, severity, etc., and record information such as the number and distribution of various defects. And label the defect positions: In the coordinate system of the wafer, accurately label the position information of each defect point. For subsequent quality analysis and process improvement.

[0057] As can be seen from the wafer defect detection method according to the prior art, the conventional detection route is relatively simple. When detecting defects, the moving stage moves each defect one by one according to the coordinates of the defect points on the wafer, following the order of first from left to right and then from top to bottom. This simple rule does not take into account the actual distribution of defects on the wafer and the movement characteristics of the moving stage, resulting in the moving path often not being optimal.

[0058] Secondly, although complex algorithms such as the ant colony algorithm, brute force algorithm, and nearest neighbor method are used to find the shortest path, the point-to-point movement is frequent and time-consuming, and the wafer detection strategy is inefficient. Every time the moving stage moves a defect point to the detection area of the imaging system for detection, it needs to perform point-to-point movement, that is, directly move from the current defect position to the camera position, and each movement includes acceleration and deceleration processes. This not only consumes a large amount of energy but also significantly increases the movement time. And in the prior art, usually only one defect point is detected at a time. After the moving stage moves a defect point to the detection area of the imaging system to obtain an image and process it, it will move to the next defect point position for detection, with low efficiency.

[0059] Embodiment 1

[0060] In view of the defects existing in the defect detection method in the prior art, such as Figure 3 as described, the figure shows a flowchart of a defect detection method provided by the present invention; in this embodiment, a wafer is taken as the sample to be detected. It can be seen that the defect detection method provided by the present invention includes the following steps: Figure 3

[0061] S1: Obtain the defect coordinate point cloud data of the sample to be detected.

[0062] Adjust the defect detection device to a normal working state. Check and debug each key component in the device, including the moving stage, light source, and imaging system. Generally, the imaging system includes an objective lens, a tube lens, and a line scan / area array camera, etc. Place the wafer to be detected at the designated position on the moving stage. Use the line array camera for initial scanning imaging. Apply image processing technology for preliminary analysis. Obtain the defect coordinate point cloud data of the sample to be detected. Generally, the wafer is a three-dimensional structure with a certain thickness, and defects may exist at different depth positions on the surface or inside. Through the imaging principle of the imaging system and related algorithms, accurately determine the specific position of each defect in three-dimensional space, providing a comprehensive and accurate data basis for subsequent further analysis and re-inspection operations.

[0063] S2: Convert the defect coordinate point cloud data into a two-dimensional data point set: Converting the defect coordinate point cloud data into a two-dimensional data point set further includes: projecting the defect coordinate point cloud data, and changing the height direction coordinates of each defect point to the same value.

[0064] Specifically, the defect coordinate point cloud data obtained from the initial defect detection is projected to obtain the relative position relationship between the defect points. Generally, projection methods include top-down projection, bottom-up projection, non-vertical projection (oblique projection) or projection along a specific feature (for example, projection along the defect density gradient direction). Specifically, in this embodiment, the wafer is projected from top to bottom. In this process, the original three-dimensional coordinates (X, Y, Z) of the defect coordinate point cloud data are converted into two-dimensional coordinates (X, Y), where the coordinates of the horizontal X and Y axes remain unchanged, and the coordinates of the height direction Z axis change to the same value, resulting in a two-dimensional data point set containing a two-dimensional data point set without height difference. As Figure 4 As shown, it is a schematic diagram of projecting defect coordinate point cloud data onto a plane, showing how to convert three-dimensional coordinate points into two-dimensional coordinate points during the top-down projection process, as well as the distribution of defect point cloud data on the plane; Figure 4 For example, suppose there are N defect points (P 1 , P 2 , P 3 ...P N ), whose coordinates are P 1 (X 1 , Y 1 , Z 1 ), P 2 (X 2 , Y 2 , Z 2 )……P N (X N , Y N , Z N ). Generally, the value of the height direction Z axis can be selected according to the actual coordinates. For example, P 1 (X 1 , Y 1 , Z 1 )’s Z-axis height Z 1 As a benchmark, convert the Z-axis coordinates of other defect points into Z 1 The wafer is projected from top to bottom to obtain a two-dimensional data point set after projection, and the three-dimensional defect coordinate point cloud data is converted into a two-dimensional data point cloud. In the two-dimensional data point set, the defect point P 1 The projection coordinates are (X 1 , Y 1 , Z), the projection coordinates of the defect point P2 are (X 2 , Y 2 , Z) ... defect point P N The projection coordinates are (X N , Y N, Z). After projection, the relative positional relationship of all defect points on the two-dimensional plane is preserved, while the height information is uniformly processed, making the data easier to manage and operate.

[0065] S3: Divide the two-dimensional data point set into several detection regions, where the detection regions contain path points;

[0066] First, obtain the window size of the camera field of view, including the width of the camera field of view. Divide the two-dimensional data point set into several detection regions. Specifically, the width of the detection region is greater than or equal to the width of the camera field of view to ensure that each detection can make full use of the camera field of view. Specifically, in this embodiment, the width of the camera field of view is used as the width of the detection region to divide the two-dimensional data point set, so that the image captured by the camera each time is fully docked with the detection region, avoiding repeated or insufficient coverage. Specifically, the length of the detection region can be set according to the positions of the defect points on the actual wafer. And each detection region contains at least one of the two-dimensional data points (i.e., defect points). Since the camera can only image and detect the region within its field of view during the detection process, dividing in this way avoids the problems of omission or re-detection. Generally, the algorithms for dividing the two-dimensional data point set include the hough transform and the Radon transform, so that the number of detection regions in the two-dimensional data point set is as small as possible. Specifically, since the size of the camera field of view is determined by the type of the camera, the focal length of the lens, and the size of the sensor, the size of the camera field of view is not limited here.

[0067] In particular, all the detection regions should be able to completely cover all the two-dimensional data points to ensure that no defect points are missed during subsequent clustering and detection.

[0068] In particular, there may be overlap between detection regions. For the data points in the middle of two detection regions, in order to avoid the moving stage from frequently accelerating and decelerating between two adjacent detection regions, these two detection regions may be partially overlapped in space, so that the moving stage can transition more smoothly from one region to another during movement, in order to better process this data point. However, too much overlapping detection will cause the data points in the overlapping part to be detected repeatedly, which not only wastes the detection time, but may also cause confusion in the data processing process due to multiple detections and analyses of the same data point. Therefore, overlapping of regions should be avoided as much as possible when designing the detection regions.

[0069] As Figure 6 shown, it is a schematic diagram of dividing the two-dimensional data point set into detection regions; assuming the width of the camera field of view is W, then starting from one side of the two-dimensional data point set, starting from the first data point, multiple detection regions are sequentially divided with W as the width until the entire data point range in the two-dimensional data point set is covered. Specifically, asFigure 5 As shown, in this embodiment, the two-dimensional data point set is divided into 4 detection regions.

[0070] The steps of clustering the two-dimensional data points in the detection region include: setting the clustering window size to be the same as the window size of the camera field of view; performing clustering on the two-dimensional data points in the detection region by moving the position of the clustering window in the detection region.

[0071] Specifically, the size of the clustering window is set to be the same as the window size of the camera field of view. The clustering windows are distributed in parallel within the corresponding detection regions. Determining the size of the clustering window based on the size of the camera field of view aims to enable all the data points within each clustering group to be included in one camera shot, thereby improving the detection efficiency. When multiple two-dimensional data points are grouped together, when the moving stage is moved subsequently, only this group of data points needs to be moved as a whole to the center of the camera field of view for detection, and the information of multiple data points can be obtained. By continuously moving the position of the clustering window in the detection region, the number of clustering groups is optimized as much as possible to further reduce the moving operations of the moving stage, because the fewer the number of clustering groups, the fewer the number of times the clustering groups need to be moved during the detection process. Generally, the window size of the camera field of view is determined by the type of camera, the focal length of the lens, and the size of the sensor, so the window size of the camera field of view is not limited here.

[0072] Generally, to optimize the number of clustering groups, factors such as the distribution density and position relationship of the data points need to be comprehensively considered to find a reasonable clustering method that minimizes the number of clustering groups while ensuring that all defect points can be effectively detected. For example, some clustering algorithms (such as the mean shift clustering algorithm), optimization objective functions based on the intra-cluster distance and inter-cluster distance, and metrics based on information entropy and other clustering methods can be used to assist in determining the position and number of clustering windows.

[0073] Specifically, the method for determining the number and position of clustering groups using the mean shift clustering method is as follows:

[0074] (1) Construct the original clustering set Region{i}: Using any data point P in the detection region as the center point, calculate the distance d(i, x) between all other points P in the detection region and the center point P, and classify the points with d(i, x) less than the distance threshold Md(i, x) into the original clustering set Region{i}, that is: i For the center point, calculate the distance between all other points in the detection region and the center point, and classify the points with the distance less than the distance threshold into the original clustering set, that is: x from the center point, calculate the distance between all other points in the detection region and the center point, and classify the points with the distance less than the distance threshold into the original clustering set, that is: i from the center point, calculate the distance between all other points in the detection region and the center point, and classify the points with the distance less than the distance threshold into the original clustering set, that is:

[0075] Region{i} = {Pi|1 ≤ x ≤ K, d(i, x) < Md(i, x)}

[0076] where K is the total number of data points in the detection region;

[0077] (2) Construct the extended clustering set Region{exp}: Calculate each point P in the original clustering set Region{i} y and the data points P outside the boundary of the detection region j The distance d(y, j) is calculated. Points with a distance less than the distance threshold Md(y, j) are extended into the original clustering set Region{i} to form the extended clustering set Region{exp}, that is:

[0078] Region{exp} = {Pj|1≤j≤M, Py∈Region{i}, d(y, j)<Md(y, j)}

[0079] where d(y, j) is the target point P in the original clustering set Region{i} y and the target point P outside the boundary of the detection region j The distance, and M is the total number of data points outside the boundary of the detection region.

[0080] (3) Take the extended clustering set as a clustering group, remove the clustered data points from the data point cloud, and repeat steps (1) and (2) to complete the clustering grouping of all data points in the detection region.

[0081] Specifically, the method for determining the number and location of clustering groups based on the intra-cluster distance and the inter-cluster distance is as follows:

[0082] (1) Define the objective function: Suppose there are a total of N data points, divided into K clustering groups, P i represents the i-th data point, and M j represents the center of the j-th clustering group (which can be represented by the mean of the data points within the cluster). Define an objective function:

[0083]

[0084] where C j represents the set of data points included in the j-th clustering group, ‖·‖ represents a measure of a certain distance, and λ represents a balancing parameter.

[0085] The goal is to find a suitable value of K such that M(k) is minimized. By calculating and comparing different values of K (for example, starting from a smaller value of K and gradually increasing), the K that makes M(k) reach the minimum value is found. The clustering situation corresponding to this K is the result that, while minimizing the number of clustering groups as much as possible, tries to ensure that the data points within the cluster are relatively close and the clusters are relatively well separated from each other.

[0086] Such as Figure 6As shown, it is a schematic diagram after clustering the defect points; select a suitable clustering method to cluster the data points. The clustering group should contain at least one two-dimensional data point.

[0087] The steps of planning the detection path include: determining the path points in the several detection regions; determining the detection start point and the detection end point of all detection regions; and determining the detection order of the path points in the several detection regions according to the detection start point and the detection end point.

[0088] First, determine the path points in the detection region. The path points are the center points of the clustering window. As Figure 6 shown, the path points are the positions of the red points in the figure; when performing detection, at least one image of the clustering window is obtained when the center of the camera field of view coincides with the path points.

[0089] Then, determine the positions of the key clustering groups: determine the positions of the first clustering group and the last clustering group in each detection region. These two positions will be used as important reference points for calculating the detection order between detection regions and within the regions. And determine the detection start point and the detection end point of all detection regions.

[0090] Finally, determine the detection order of all path points according to the detection start point and the detection end point of the detection regions.

[0091] Secondly, plan the detection order between detection regions: to find the optimal order that minimizes the total moving distance of the moving stage or the total detection time. Generally, a numerical operation model (such as ant colony algorithm, brute force algorithm, genetic algorithm, etc.) is used to plan the detection order between multiple detection regions.

[0092] Specifically, taking the ant colony algorithm as an example, assume there are three detection regions A, B, and C. Consider the center position of each detection region as a node on the ant path. The ants move between these nodes and mark the path by releasing pheromones. The higher the pheromone concentration, the better the path. After multiple iterations, more pheromones are left by the ants passing through the A - B - C path, and the optimal detection order is determined. The brute force algorithm will list all possible detection order combinations of the detection regions, and then calculate the moving cost (such as moving distance, moving time, etc.) of the moving stage for each order. For example, for three detection regions A, B, and C, it will calculate the moving costs for the six orders A - B - C, A - C - B, B - A - C, B - C - A, C - A - B, and C - B - A respectively, and select the order with the lowest cost as the optimal detection order. Select a suitable numerical operation module to determine the optimal detection route.

[0093] Meanwhile, the detection order of the clustering groups within the detection area is also determined: within each detection area, there are two options for the detection order of the clustering groups, namely from the first clustering group to the last clustering group, or from the last clustering group to the first clustering group. It is necessary to determine according to the specific situation which order can make the moving path of the moving stage within this area the shortest and the detection efficiency the highest.

[0094] According to the finally determined detection path, the detection paths of the determined data points are compared one by one with the defect points on the wafer to detect the detection paths between the defect points. Specifically, as Figure 7 shown, it is a schematic diagram of the defect detection path provided by this embodiment.

[0095] S5: Detect the data points according to the detection path;

[0096] Specifically, move the detection starting point on the wafer to the detection area and detect: The detection system sequentially detects the defect points according to the pre-planned detection path. The moving stage moves each clustering group of each detection area to under the camera field of view in turn according to the determined path. The detection system starts triggering image acquisition from the detection starting point and triggers the camera to perform image acquisition. At the same time, the detection system obtains the detection images of each clustering group through the camera. These images contain information such as the shape, size, and position of the defect points within the clustering group, which is the basis for subsequent defect analysis and processing. Classify, count, and label the defect points according to the obtained detection images.

[0097] In particular, attention needs to be paid to the relationship between the camera field of view and the detection area and related issues. As Figure 8a shown, it is a scanning schematic diagram when the angle of the detection area is the same as the camera viewing angle; it can be seen that when the width of the detection area is equal to the width of the camera field of view and when the angle of the detection area and the camera field of view angle are the same, the camera field of view is completely within the detection area, that is, the camera field of view can better cover the detection area. At this time, the field of view can be completely covered by the detection area (although it is not necessary to be completely covered), and in this case, the field of view utilization rate is relatively high. As Figure 8a shown, the angle of the detection area A where the 5 clustering groups A1 - A5 are located is consistent with the camera field of view angle, and the camera field of view can be utilized more fully, and defect images can be obtained more efficiently during detection.

[0098] Furthermore, as Figure 8b shown, it is a scanning schematic diagram when the angle of the detection area is different from the camera viewing angle; as Figure 8bIt can be seen that when the detection area angle is different from the camera field of view angle, according to the different clustering situations of the two-dimensional data points in the detection area, the camera field of view may completely cover all the two-dimensional data points in the corresponding detection area, and at this time, the utilization rate of the camera field of view is relatively low; it may also partially cover the two-dimensional data points in the detection area, and in this case, the system needs to re-cluster or move the camera field of view multiple times. For example, as can be seen in Figure 8b, for the same 5 clustering groups A1 - A5, when there is a difference between the detection area angle and the camera field of view angle and the camera field of view can only cover part of the clustering window, more movements of the moving stage may be required during the detection process to obtain images of all defects, which reduces the detection efficiency.

[0099] Optionally, in order to maximize the utilization rate of the camera field of view, when detecting wafer defects, before taking a picture of the first clustering group of each detection area, before obtaining the image of the clustering window, first move the center of the first clustering group to the center of the camera field of view, complete the trigger for taking a picture, then rotate the moving stage at a specified angle so that the angle of the detection area is the same as the camera viewing angle, and then perform scanning and taking pictures, thereby maximizing the optimization of the equipment productivity; or before obtaining the image of the clustering window, rotate the angle of the clustering window so that the clustering window completely coincides with the detection area.

[0100] Embodiment 2

[0101] This embodiment provides a defect detection system, as Figure 9 shown, which shows a schematic diagram of the defect detection device provided in this embodiment; the defect detection device includes a moving stage 01, an imaging system 02, a light source 03, and a computer system 05. The moving stage 01 is used to carry and move the sample to be tested; the light source 03 is used to provide illumination light; the imaging system 02 is used to collect the detection image of the detection area of the sample to be tested, and the detection image includes the defect coordinate point cloud data of the sample to be tested; the computer system 05 is used to perform two-dimensional projection on the defect coordinate point cloud data, divide the two-dimensional projection into several detection areas, determine the detection path after clustering the defect data in the detection areas, and re-inspect the defects of the sample to be tested according to the detection path. Specifically, in this embodiment, the wafer 04 is used as the sample to be tested, and the defects in the wafer 04 are detected, but the defect detection device provided in this embodiment is not limited to wafer detection.

[0102] The moving stage 01 is used to carry the wafer 04 and drive the wafer 04 to move under the imaging system, and can achieve multi-dimensional movements, including up and down movement of the height position, horizontal position, and rotation and other movements. Optionally, the moving stage is equipped with a high-precision positioning system, such as a servo motor, a linear motor, or a stepping motor, which can perform precise fine-tuning within the nanometer or micrometer range to ensure that each point on the wafer surface can be clearly captured by the imaging system.

[0103] The imaging system 02 magnifies and images the surface of the wafer 04. The imaging system 02 includes an objective lens 021, a tube lens 022, and a camera module 023. The imaging system 02 can collect the light reflected or scattered from the surface of the wafer 04 and focus this light into a clear image. The tube lens 022 plays an auxiliary imaging role in the imaging system, working in cooperation with the objective lens 021 to ensure that the light can accurately reach the imaging plane of the camera 023 to obtain high-quality images. Generally, the objective lens 021 includes a low-magnification objective lens (such as 1× - 20×) and a high-magnification objective lens (such as 30× - 100×).

[0104] Specifically, the objective lens 021 includes a low-magnification objective lens suitable for large-range surface scanning to detect larger-sized defects; during the preliminary detection, the low-magnification objective lens has a large depth of field and can maintain a relatively clear imaging effect within a large range. Therefore, when using the low-magnification objective lens, the autofocus function of the imaging system 02 can handle the focal length change without much manual intervention. The low-magnification objective lens is suitable for detecting larger defects and can quickly scan the entire surface of the wafer to improve the detection efficiency; the high-magnification objective lens is used for more refined surface defect detection; the high-magnification objective lens has a small depth of field, which makes its imaging area more limited, so the accuracy requirement for the autofocus system is higher. Usually, during the re-inspection process, in order to ensure high-precision measurement, it may be necessary to measure the height of the wafer first and then perform autofocus. The high-magnification objective lens can capture tiny particles, scratches, microcracks and other small defects, which is crucial for ensuring the integrity and quality of the wafer surface.

[0105] The camera 023 is responsible for converting the optical signal processed by the objective lens 021 and the tube lens 022 into an electrical signal and recording it as a digital image. Specifically, the camera 023 includes a line scan camera and an area array camera. The light source 03 provides the necessary lighting conditions for the entire detection process and uniformly transmits it to the detection area of the defect detection system. Generally, the camera module 023 includes: a line scan camera: This camera is suitable for large-range rapid scanning, can read image data row by row in real time and process it, and is suitable for detecting defects in a wide area; an area array camera: The area array camera has a fixed pixel array and can capture image data of multiple points simultaneously. It is suitable for obtaining high-resolution, static images and is especially suitable for precision defect detection, such as tiny particles or cracks.

[0106] The computer system 05 is used to control the imaging system 02 to obtain the detection image of the detection area of the wafer 04, perform data processing on the detection image to determine the re-inspection route of the defect points, and control the imaging system 02 to perform defect re-inspection on the wafer 04 according to the re-inspection route.

[0107] Specifically, the computer system 05 is configured with a control system 051, a data processing unit 052, a data storage unit 053, and a display module 054. The control system 051 controls the movement of the moving stage 01 and commands the imaging system 02 to perform high-precision image acquisition according to the planned route; the data processing unit 052 is used to receive the defective coordinate point cloud data collected from the imaging system 02, process the defective point cloud data, and formulate a re-inspection route; the control system 051 commands the imaging system 02 to re-inspect the defective points on the wafer 04; at the same time, the data storage unit 053 records all the detection data; the display module 054 displays information such as the defect detection results, re-inspection paths, and data charts in real time, and also has some interaction functions, such as allowing the operator to dynamically adjust the re-inspection path, select different detection areas, set image processing algorithms, etc.

[0108] Specifically, the data processing unit 052 preprocesses the defective point cloud data through image processing algorithms, such as denoising, enhancement, edge detection, etc., in order to better extract the defect features; projects the three-dimensional defective coordinate point cloud data into a two-dimensional data point set; divides the two-dimensional data point set into multiple detection areas; clusters the defective points in each detection area; and formulates a reasonable detection path for re-inspection according to the clustered defective data.

[0109] The defect detection device provided in this embodiment greatly improves the detection accuracy. Through the mutual cooperation of the computer system 05, the moving stage 01, and the imaging system 02, it can accurately detect the position distribution of defective points on the sample to be tested (such as a wafer). Secondly, the high-precision imaging system can efficiently scan the entire surface of the wafer, ensuring the comprehensiveness and meticulousness of defect detection. Improve the re-inspection efficiency: The data processing unit 052 clusters the defects and optimizes the re-inspection path, reducing the ineffective re-inspection area, improving the re-inspection efficiency, and ensuring the accurate re-inspection of the defects.

[0110] Embodiment III

[0111] This embodiment provides a computer-readable storage medium, on which computer instructions are stored. The computer instructions, when executed by a processor, implement the defect detection method described in any one of Embodiment 1 or Embodiment 2.

[0112] Specifically, the computer instructions include but are not limited to: motion control instructions, imaging control instructions, image acquisition and processing instructions, data preprocessing instructions, feature extraction instructions, data analysis instructions, re-inspection path optimization instructions, result display and interaction instructions, data storage and management instructions.

[0113] Motion control instructions to precisely control the moving stage to control the displacement, positioning, and scanning path of the sample to be tested, ensuring the positioning accuracy during the defect detection process; imaging control instructions to control the acquisition parameters of the imaging system, including exposure time, focal length, imaging resolution, etc., ensuring that the detected defect images are clear enough; image acquisition and processing instructions, including image acquisition instructions for capturing image data in different detection areas; data preprocessing instructions, such as denoising, enhancing contrast, edge detection, etc., for improving image quality and enhancing defect features; feature extraction instructions to perform defect localization, classification, and annotation to identify potential defects on the wafer surface or in the sample to be tested; data analysis instructions, including instructions for clustering, coordinate transformation, projection, etc. of defect data; re-inspection path optimization instructions to automatically plan the re-inspection path based on the clustering analysis of defect points; result display and interaction instructions to control the result display module, real-time display of defect data, defect distribution maps, re-inspection paths, and related chart information, allowing users to perform interactive operations in the display module, adjust detection parameters, select detection areas, modify re-inspection paths, etc.; data storage and management instructions responsible for storing and managing detection results, defect data, image information, etc.

[0114] The above embodiments are only illustrative of the principles and effects of the present invention and are not intended to limit the present invention. Any person familiar with this technology can modify or change the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or changes completed by those with ordinary knowledge in the relevant technical field without departing from the spirit and technical ideas disclosed by the present invention should still be covered by the claims of the present invention.

Claims

1. A defect detection method, characterized in that: The following steps are involved: Obtain the defect coordinate point cloud data of the sample to be tested; Converting the defect coordinate point cloud data into a two-dimensional data point set; Dividing the two-dimensional data point set into a plurality of detection areas, wherein the detection areas contain path points; Based on the path points, planning a detection path; According to the detection path, the defects of the sample to be tested are re-inspected.

2. The defect detection method according to claim 1, characterized in that: Converting the defect coordinate point cloud data into a two-dimensional data point set includes: projecting the defect coordinate point cloud data, and changing the height direction coordinates of each defect point to the same value.

3. The defect detection method according to claim 1, characterized in that: The step of dividing the two-dimensional data point set into a plurality of detection areas comprises: Get the window size of the camera's field of view, including the width of the camera's field of view; Dividing the two-dimensional data point set into the plurality of detection areas, wherein the width of the detection area is greater than or equal to the width of the camera field of view; and The two-dimensional data points within the detection area are clustered.

4. The defect detection method according to claim 3, characterized in that: The step of clustering the two-dimensional data points within the detection area comprises: Setting the clustering window size to be the same as the window size of the camera field of view; At least one clustering window is determined within the detection area, and clustering is performed on the two-dimensional data points within the detection area.

5. The defect detection method according to claim 4, characterized in that: The clustering windows are distributed in parallel within the corresponding detection area.

6. The defect detection method according to claim 4, characterized in that: The path point is the center point of the clustering window, and at least one image of the clustering window is acquired when the center of the camera field of view coincides with the path point.

7. The defect detection method according to claim 6, characterized in that: Before acquiring the cluster window image, the detection area angle is rotated so that the camera field of view is located within the detection area.

8. The defect detection method according to claim 1, characterized in that: The steps to plan the inspection path include: Determine the path points within the plurality of detection areas; Determine a detection starting point and a detection end point of the path point; and The detection order of the path points in the plurality of detection areas is determined according to the detection starting point and the detection end point.

9. A defect detection system, characterized in that: include: A motion table, used to carry and move the sample to be tested; A light source, for providing lighting; An imaging system, comprising an objective lens, a tube lens and a camera module, wherein the imaging system is used to collect a detection image of the detection area of ​​the sample to be tested, wherein the detection image includes defect coordinate point cloud data of the sample to be tested; The computer system performs two-dimensional projection on the defect coordinate point cloud data, divides the two-dimensional projection into a number of detection areas, clusters the defect data in the detection areas and determines a detection path, and re-inspects the defects of the sample to be tested according to the detection path.

10. A computer-readable storage medium having computer instructions stored thereon, characterized in that: When the computer instructions are executed by a processor, the defect detection method according to any one of claims 1 to 8 is implemented.

Images