A mechanical damage defect detection method and apparatus
By utilizing defect distribution maps and dynamically adjusting clustering radii in wafer surface defect detection, mechanical damage defects can be identified, solving the problem of insufficient detection accuracy in existing technologies and achieving more efficient mechanical damage defect identification.
Patent Information
- Application Number
- CN202611187591.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-06
- Publication Date
- 2026-11-03
AI Technical Summary
Existing wafer surface defect detection methods are difficult to accurately identify mechanical damage defects, especially under different starting test sizes, which can easily lead to missed or over-detection.
By using the defect distribution maps of the wafer under test and the sample wafer, the target clustering region of defect points is determined, and the target clustering radius is dynamically adjusted according to the number of defect points, the adjacent distance and the starting size of the test to perform cluster analysis. Combined with the distribution pattern of the defect clusters, mechanical damage defects are determined.
It improves the accuracy of mechanical damage defect detection, takes into account the over-detection and under-detection issues in different detection scenarios, and enhances detection efficiency and accuracy.
Smart Images
Figure CN123032633A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of semiconductor manufacturing technology, and in particular to a method and apparatus for detecting mechanical damage defects. Background Technology
[0002] During wafer inspection, transportation, or processing, wafers are prone to non-ideal contact with components such as chucks, the edges of carriers, and robotic arms, resulting in mechanical damage defects on the wafer surface, which in turn affects the performance of subsequent products.
[0003] Currently, particle counters, such as SP5 and SP7, are commonly used to detect defects on wafer surfaces for surface scanning. However, the detection results vary depending on the starting size. For example, when the starting size is small (e.g., 15nm, 19nm), the detection sensitivity is high, which can easily lead to random particles being misidentified as mechanical damage, resulting in over-detection. When the starting size is large (e.g., 50nm, 90nm), the detection sensitivity decreases, which can easily cause batch-wide missed detections of mechanical damage.
[0004] It is evident that existing wafer surface defect detection methods are insufficient for accurately identifying mechanical damage defects in wafers. Summary of the Invention
[0005] To address the aforementioned technical problems, this invention provides a method and apparatus for detecting mechanical damage defects, which can improve the accuracy of mechanical damage defect detection.
[0006] To achieve the above objectives, the technical solution adopted in the embodiments of the present invention is as follows:
[0007] This invention provides a method for detecting mechanical damage defects, including:
[0008] Based on the first defect distribution map of the wafer under test and the second defect distribution map of the sample wafer, the target aggregation region of defect points in the wafer under test is determined;
[0009] Cluster analysis is performed on the defect points in the target cluster region based on the target cluster radius to obtain multiple defect clusters. The target cluster radius is determined based on the number of defect points in the wafer under test, the distance between adjacent defect points, and the starting size of the defect points.
[0010] The mechanical damage defect detection results of the wafer under test are determined based on the distribution pattern of the defect clusters.
[0011] In some embodiments, determining the target clustering region of defect points in the wafer under test based on a first defect distribution map of the wafer under test and a second defect distribution map of a sample wafer includes:
[0012] The first defect distribution map of the wafer under test is divided into multiple first preset regions of the same area;
[0013] Obtain the defect density in each first preset region of the first defect distribution map and the defect density in each second preset region of the second defect distribution map. The second preset region corresponds to the first preset region. The defect density in any first preset region is determined by scanning and detecting the wafer under test. The defect density in any second preset region is the historical batch average defect density of the sample wafer in that preset region.
[0014] The target clustering area is determined based on the defect density in each first preset area of the first defect distribution map and the defect density in each second preset area of the second defect distribution map.
[0015] In some embodiments, determining the target clustering region based on the defect density in each first preset region of the first defect distribution map and the defect density in each second preset region of the second defect distribution map includes at least one of the following:
[0016] If the defect density in the first target preset area is greater than N times the defect density in the second target preset area, the first target preset area is determined as the target aggregation area, where N is greater than 1.
[0017] If the defect density of the first target preset area and the defect density of the third target preset area are in the same range, and the range is greater than, less than or equal to the defect density of the second target preset area, the first target preset area is determined as the target aggregation area.
[0018] Wherein, the first target preset area is any one of a plurality of first preset areas, the second target preset area is the area corresponding to the first target preset area among a plurality of second preset areas, the third target preset area is the area corresponding to the first target preset area in the third defect distribution map of the first wafer, and the first wafer is a wafer from the same batch as the wafer to be tested and adjacent to the wafer to be tested.
[0019] In some embodiments, determining the mechanical damage defect detection result of the wafer under test based on the distribution pattern of the defect clusters includes:
[0020] Calculate the first standard deviation and the second standard deviation for each defect cluster. The first standard deviation is the standard deviation of the coordinates of each defect point contained in a single defect cluster on the X-axis, and the second standard deviation is the standard deviation of the coordinates of each defect point contained in a single defect cluster on the Y-axis. The X-axis and the Y-axis are perpendicular coordinate axes in a Cartesian coordinate system.
[0021] The distribution pattern of each defect cluster is determined based on the ratio between the first standard deviation and the second standard deviation;
[0022] When there are defect clusters distributed linearly or in a banded manner in the target aggregation region of the wafer under test, and there are defect clusters distributed linearly or in a banded manner in the region corresponding to the target aggregation region of the first wafer, it is determined that the batch of the wafer under test has mechanical damage defects, and the first wafer is a wafer from the same batch as the wafer under test and adjacent to the wafer under test.
[0023] In some embodiments, before performing cluster analysis on defect points in the target clustering region based on the target clustering radius to obtain multiple defect clusters, the method further includes:
[0024] The scanning and detection results of the wafer under test are analyzed to obtain the total number of defects, the average distance between adjacent defects, and the starting size of the defects.
[0025] The target cluster radius corresponding to the wafer to be tested is determined based on the total number, the average distance, and the starting size. The target cluster radius is inversely proportional to the total number and directly proportional to the average distance and the starting size.
[0026] In some embodiments, after determining the mechanical damage defect detection result of the wafer under test based on the distribution pattern of the defect clusters, the method further includes:
[0027] Obtain feedback data corresponding to the mechanical damage defect detection results, wherein the feedback data includes at least one of engineer review results and process verification results;
[0028] If the feedback data indicates that the misjudgment rate or missed judgment rate of mechanical damage defects is greater than or equal to a preset threshold, the target cluster radius is adjusted.
[0029] Establish a mapping relationship between the initial measurement size and the adjusted target cluster radius.
[0030] This invention provides a mechanical damage defect detection device, comprising:
[0031] The first determining module is used to determine the target clustering region of defect points in the wafer under test based on the first defect distribution map of the wafer under test and the second defect distribution map of the sample wafer.
[0032] The analysis module is used to perform cluster analysis on defect points in the target cluster region based on the target cluster radius to obtain multiple defect clusters. The target cluster radius is determined based on the number of defect points in the wafer under test, the distance between adjacent defect points, and the starting size of the defect points.
[0033] The second determining module is used to determine the mechanical damage defect detection result of the wafer under test based on the distribution pattern of the defect clusters.
[0034] This invention also provides an electronic device, including: a processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps of the above-described mechanical damage defect detection method.
[0035] This invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described mechanical damage defect detection method.
[0036] This invention also provides a computer program product, wherein a computer program is stored on a computer-readable storage medium, and the computer program, when executed by a processor, implements the steps of the above-described mechanical damage defect detection method.
[0037] The beneficial effects of this invention are:
[0038] In this embodiment of the invention, a second defect distribution map of a sample wafer is used as a reference benchmark. By comparing the second defect distribution map with the first defect distribution map of the wafer under test, the target clustering region of defect points in the wafer under test is determined, thus narrowing the scope of subsequent analysis. Then, the target clustering radius is dynamically determined based on the number of defect points in the wafer under test, the distance between adjacent defect points, and the starting size of the defect points. Cluster analysis is then performed based on the target clustering radius, which can adapt to different detection scenarios, balance suppressing over-detection and avoiding missed detection, and improve the accuracy of cluster analysis results. Finally, the mechanical damage defect detection result of the wafer under test is determined based on the distribution pattern of the defect clusters, thus improving the accuracy of mechanical damage defect detection. Attached Figure Description
[0039] Figure 1 This is a flowchart illustrating the mechanical damage defect detection method in an embodiment of the present invention.
[0040] Figure 2 This diagram illustrates a linearly distributed defect cluster in an embodiment of the present invention.
[0041] Figure 3 This diagram illustrates the defect clusters with a banded distribution in an embodiment of the present invention.
[0042] Figure 4 This is a schematic diagram of the Chuck contact area in an embodiment of the present invention;
[0043] Figure 5 This is a schematic diagram of the mechanical damage defect detection device in an embodiment of the present invention;
[0044] Figure 6 This is a schematic diagram of the structure of the electronic device in an embodiment of the present invention. Detailed Implementation
[0045] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention are within the scope of protection of the present invention.
[0046] In the manufacturing and inspection of semiconductor silicon wafers, particle counters (such as SP5 and SP7) are commonly used to detect surface defects and generate defect maps and Klarf files. However, during mechanical handling, processing, or inspection, non-ideal contact between the wafer and components such as Chucks, Foup edges, and robotic arms can easily lead to mechanical damage defects with fixed positions and orientations on the wafer surface. Existing wafer surface defect detection methods show differences in detection results under different initial measurement sizes, easily resulting in missed or over-detected defects. Furthermore, mechanical damage defects in single-wafer maps and lot overlays often still fall within the specification limit sample range, making accurate identification difficult by relying solely on defect quantity or simple spatial thresholds. This invention provides a method and apparatus for detecting mechanical damage defects, which can improve the accuracy of mechanical damage defect detection.
[0047] This invention provides a method for detecting mechanical damage defects, such as... Figure 1 As shown, it includes the following steps:
[0048] Step 101: Based on the first defect distribution map of the wafer under test and the second defect distribution map of the sample wafer, determine the target clustering region of defect points in the wafer under test.
[0049] In this step, the first defect distribution map of the wafer under test and the second defect distribution map of the sample wafer are first obtained. The acquisition of the first defect distribution map can be achieved by: using a Particle Counter to scan and inspect the wafer under test to obtain defect information on the wafer surface and generating a corresponding Klarf file; then reading the coordinate position, size, and defect type (e.g., protruding particles or recessed defects) of each defect from the Klarf file, and drawing the first defect distribution map of the wafer under test based on the coordinate information. The sample wafer can be a batch of historical wafers with the same processing technology and testing conditions as the wafer under test. By retrieving the batch defect data of the sample wafer, a corresponding historical defect overlay map, i.e., the second defect distribution map, is generated as the baseline distribution of wafer defect points.
[0050] After obtaining the first defect distribution map of the wafer under test and the second defect distribution map of the sample wafer, the target clustering region of defect points in the wafer under test is further determined, as detailed below:
[0051] In some embodiments, determining the target clustering region of defect points in the wafer under test based on a first defect distribution map of the wafer under test and a second defect distribution map of a sample wafer includes:
[0052] The first defect distribution map of the wafer under test is divided into multiple first preset regions of the same area;
[0053] Obtain the defect density in each first preset region of the first defect distribution map and the defect density in each second preset region of the second defect distribution map. The second preset region corresponds to the first preset region. The defect density in any first preset region is determined by scanning and detecting the wafer under test. The defect density in any second preset region is the historical batch average defect density of the sample wafer in that preset region.
[0054] The target clustering area is determined based on the defect density in each first preset area of the first defect distribution map and the defect density in each second preset area of the second defect distribution map.
[0055] In this embodiment, after obtaining the first defect distribution map of the wafer under test, the first defect distribution map is divided into multiple first preset regions of the same area. For example, during the division of the preset regions, the wafer can be divided along the radial direction based on the concentric ring regions to obtain multiple first preset regions of the same area; and / or, it can be divided along the angular direction based on the fan-shaped regions of the wafer to obtain multiple first preset regions of the same area. Similarly, after obtaining the second defect distribution map of the sample wafer, the second defect distribution map is divided into multiple second preset regions of the same area in the same way, so that each second preset region corresponds one-to-one with each first preset region. This facilitates subsequent defect analysis of local areas of the wafer, improving the ability to identify abnormal mechanical damage at local locations compared to relying solely on the total number of defects on the entire wafer for defect analysis.
[0056] Then, the defect density in each first preset region and each second preset region is statistically analyzed. The defect density in each second preset region can be the historical batch average defect density of the corresponding preset region, while the defect density in each first preset region is determined by scanning and detecting the wafer under test. Thus, after obtaining the defect density in any first preset region, the defect density in the corresponding second preset region can be retrieved as a reference density. By comparing the defect density in the first preset region with the reference density (i.e., the defect density in the corresponding second preset region), it is determined whether the first preset region is a target cluster region. For example, if the defect density in the first preset region is greater than the reference density of the corresponding reference preset region, then the first preset region is determined to be a target cluster region. This method can quickly screen out local areas with abnormally concentrated defects, facilitating further cluster analysis of target cluster regions, improving detection efficiency, and reducing random particle interference.
[0057] Specifically, the determination of the target cluster area can be found in the following description:
[0058] In some embodiments, determining the target clustering region based on the defect density in each first preset region of the first defect distribution map and the defect density in each second preset region of the second defect distribution map includes at least one of the following:
[0059] If the defect density in the first target preset area is greater than N times the defect density in the second target preset area, the first target preset area is determined as the target aggregation area, where N is greater than 1.
[0060] If the defect density of the first target preset area and the defect density of the third target preset area are in the same range, and the range is greater than, less than or equal to the defect density of the second target preset area, the first target preset area is determined as the target aggregation area.
[0061] Wherein, the first target preset area is any one of a plurality of first preset areas, the second target preset area is the area corresponding to the first target preset area among a plurality of second preset areas, the third target preset area is the area corresponding to the first target preset area in the third defect distribution map of the first wafer, and the first wafer is a wafer from the same batch as the wafer to be tested and adjacent to the wafer to be tested.
[0062] In this embodiment, the first defect distribution map of the wafer to be tested is divided into multiple first preset regions of the same area. Taking any one of the first preset regions (denoted as the i-th first preset region, i.e. the first target preset region) as an example, an exemplary description is given, where i is a positive integer.
[0063] The second and third target preset regions are both defect density reference regions corresponding to the first target preset region, used to characterize the baseline level of defect distribution within the first target preset region. Specifically, the second target preset region is the corresponding region at the same location of a sample wafer manufactured under the same process and testing conditions as the wafer under test; the third target preset region is the corresponding region at the same location of the first wafer from the same batch as the wafer under test and adjacent to it.
[0064] The defect density in the first target preset area can be denoted as ρ. i The defect density in the second target preset region can be denoted as ρ. i The baseline, the defect density of the third target preset region can be denoted as ρ. j .
[0065] In one example, if the defect density ρ of the first target preset region on the current wafer under test... i If the defect density exceeds N times the corresponding historical baseline area defect density (i.e., the defect density ρᵢ baseline of the second target preset area) (N>1, usually 2~3), it indicates that the defect points in the first target preset area have a significant local clustering phenomenon and have deviated from the normal fluctuation range. Therefore, the first target preset area can be identified as the target clustering area.
[0066] In another example, if the defect density ρ of the first target preset region on the current wafer under test... i The defect density ρ in the corresponding region of adjacent wafers in the same batch j They are in the same numerical range, and ρ i With ρj The deviation direction remains consistent with that of the ρᵢ baseline, for example, ρ i With ρ j If all values are greater than, less than or equal to ρᵢ baseline, it indicates that the defect points in the first target preset area exhibit local clustering, and are not random phenomena on a single piece, but rather appear repeatedly within a batch, exhibiting systematic defect characteristics. Therefore, the first target preset area can be identified as the target clustering area.
[0067] In yet another example, if ρ i Greater than N times ρᵢ baseline, and ρ i With ρ j By maintaining the same deviation direction as the ρᵢ baseline, the first target preset area can also be determined as the target aggregation area.
[0068] In this way, by identifying the target clustering area through the above method, the accuracy of identifying abnormal areas with high local defect density is improved. At the same time, the number of missed detections of synchronous abnormal areas in the same position of multiple wafers in a batch is reduced, temporary fluctuations caused by random particles are effectively eliminated, and suspected clustering areas corresponding to mechanical damage defects caused by factors such as mechanical contact and equipment interference are accurately identified, which facilitates further clustering analysis of the target clustering area.
[0069] Step 102: Perform cluster analysis on the defect points in the target cluster region based on the target cluster radius to obtain multiple defect clusters. The target cluster radius is determined based on the number of defect points in the wafer to be tested, the distance between adjacent defect points, and the starting size of the defect points.
[0070] In this step, after identifying the target cluster region, cluster analysis is further performed on the defect points within the target cluster region. First, the detection mode of the wafer under test is determined based on the number of defect points in the wafer under test, the distance between adjacent defect points, and the starting size of the defect points.
[0071] For example, when the defect detection size is small (e.g., 19nm), the equipment has high detection sensitivity, a significantly higher total number of defect points per wafer, and a smaller average adjacent distance between defect points. Based on this, the detection mode of the wafer under test can be determined to be a high-density detection mode. In high-density detection mode, a large number of random microparticles exist. If the cluster radius remains unchanged, random particles are easily mis-clustered into "false stripes". Based on this, the original cluster radius can be reduced. Based on the reduced target cluster radius, cluster analysis can be performed on the defect points in the target cluster region. This can filter out randomly distributed particles, reduce the situation of mis-clustering random particles into mechanical damage defect clusters, and only aggregate spatially highly continuous defects, thereby improving the accuracy of detecting mechanical damage defects.
[0072] For example, when the defect size is large (e.g., 90nm), the equipment has low detection sensitivity, only identifying large-sized defects. The total number of defect points on a single wafer is small, and the average distance between adjacent defect points is large. Based on this, the detection mode of the wafer under test can be determined to be a low-density detection mode. In low-density detection mode, repeated defects at the same mechanical contact point on different wafers will be broken up and cannot be identified. Therefore, the original cluster radius can be expanded. Based on the expanded target cluster radius, cluster analysis can be performed on defect points in the target cluster region. Defects that are spatially close and oriented in the same direction can be identified as having the same mechanical interference source, thereby capturing mechanical damage recurring across wafers and improving the accuracy of detecting mechanical damage defects.
[0073] Thus, the detection mode of the wafer under test is determined based on the number of defect points M, the distance D̄ between adjacent defect points, and the starting measurement size Size of the defect points, thereby determining the target clustering radius R. The target clustering radius R can be expressed as: R = f(M, D̄, Size), realizing dynamic adjustment of the clustering radius. When the wafer under test is in high-density detection mode, the system automatically reduces the clustering radius R to improve spatial resolution; when the wafer under test is in low-density detection mode, the system automatically expands the clustering radius R to enhance the correlation ability of defects with similar positions.
[0074] To achieve adaptive adjustment of the target cluster radius and ensure that cluster analysis can adapt to different detection conditions while balancing overdetection suppression and missed detection prevention, the target cluster radius corresponding to the wafer under test is determined before performing cluster analysis on defect points within the target cluster region. See the following description:
[0075] In some embodiments, before performing cluster analysis on defect points in the target clustering region based on the target clustering radius to obtain multiple defect clusters, the method further includes:
[0076] The scanning and detection results of the wafer under test are analyzed to obtain the total number of defects, the average distance between adjacent defects, and the starting size of the defects.
[0077] The target cluster radius corresponding to the wafer to be tested is determined based on the total number, the average distance, and the starting size. The target cluster radius is inversely proportional to the total number and directly proportional to the average distance and the starting size.
[0078] In this embodiment, a Particle Counter can be used to scan and inspect the wafer under test to obtain the scan results. The scan results are then parsed to obtain defect information on the wafer surface and generate a corresponding Klarf file. The detection formula field in the Klarf file is then read to obtain the starting size (Size) of the defect points. The starting size can be the minimum defect detection threshold set by the inspection equipment (e.g., 19nm, 90nm), with different starting sizes corresponding to different defect detection ranges. The total number of valid defect entries in the Klarf file is read to determine the total number (M) of defect points on the wafer under test. The total number of defect points characterizes the overall density of defect points in the current inspection mode. After reading the coordinates of each defect from the Klarf file, the spatial distance between each defect and its nearest neighbor is calculated. The average distance of all nearest neighbor distances is then calculated to obtain the average distance (D̄) between adjacent defect points, representing the sparseness of the spatial distribution of defect points on the wafer surface. Thus, after obtaining the above three parameters, the target clustering radius (R) of the wafer under test can be calculated comprehensively based on the preset mapping relationship R=f(M, D̄, Size).
[0079] Among them, the target cluster radius is inversely proportional to the total number of defect points: the more total number of defects, the denser the defect distribution and the more random particles. At this time, it is necessary to reduce the cluster radius to avoid mis-clustering random particles into defect clusters, thereby suppressing over-detection.
[0080] The target cluster radius is proportional to the average distance between adjacent defect points: the larger the average distance between defect points, the more dispersed the defect distribution. In this case, it is necessary to expand the cluster radius to effectively cluster defects with similar locations and origins into a cluster to avoid missed detection.
[0081] The target cluster radius is directly proportional to the starting size of the defect points: the larger the starting size, the larger the detected defect size, the fewer the number, and the farther the spacing, which corresponds to expanding the cluster radius; the smaller the starting size, the more tiny defects and random particles are detected, which corresponds to shrinking the cluster radius.
[0082] Thus, cluster analysis of defect points within the target clustering region based on a dynamically determined target clustering radius can adapt to different detection conditions, balancing over-detection and under-detection, thereby improving the accuracy of cluster analysis results. Under the dynamically determined clustering radius, cluster analysis of defect points within the target clustering region identifies defect clustering structures in regions of different densities, resulting in multiple defect clusters. Further, the mechanical damage defect detection results of the wafer under test are determined based on the distribution morphology of the defect clusters, as detailed below:
[0083] Step 103: Determine the mechanical damage defect detection result of the wafer under test based on the distribution pattern of the defect clusters.
[0084] In this step, the distribution pattern of defect clusters can be determined to determine whether the defect clusters exhibit a linear or banded distribution. Then, by combining the distribution pattern and positional repeatability of defect clusters in the corresponding areas of adjacent wafers in the same batch, it can be finally determined whether the wafer under test has mechanical damage defects caused by structural interference such as chucks, carriers, or robotic arms, and the test results can be generated.
[0085] In some embodiments, determining the mechanical damage defect detection result of the wafer under test based on the distribution pattern of the defect clusters includes:
[0086] Calculate the first standard deviation and the second standard deviation for each defect cluster. The first standard deviation is the standard deviation of the coordinates of each defect point contained in a single defect cluster on the X-axis, and the second standard deviation is the standard deviation of the coordinates of each defect point contained in a single defect cluster on the Y-axis. The X-axis and the Y-axis are perpendicular coordinate axes in a Cartesian coordinate system.
[0087] The distribution pattern of each defect cluster is determined based on the ratio between the first standard deviation and the second standard deviation;
[0088] When there are defect clusters distributed linearly or in a banded manner in the target aggregation region of the wafer under test, and there are defect clusters distributed linearly or in a banded manner in the region corresponding to the target aggregation region of the first wafer, it is determined that the batch of the wafer under test has mechanical damage defects, and the first wafer is a wafer from the same batch as the wafer under test and adjacent to the wafer under test.
[0089] In this embodiment, for each identified defect cluster, the first standard deviation and the second standard deviation of the defect cluster are calculated. The first standard deviation σ... x It can be represented as:
[0090] σ x = std(x1…x n );
[0091] Where x1…x n Let σ be the coordinates of n defect points contained in a defect cluster on the X-axis. x This represents the first standard deviation of the defect cluster.
[0092] Second standard deviation σ y It can be represented as:
[0093] σ y = std(y1…y n );
[0094] Among them, y1…y nLet σ be the Y-coordinate of the n defect points contained in this defect cluster. y This represents the second standard deviation of the defect cluster.
[0095] After calculating the first and second standard deviations of the defect clusters, the distribution pattern of each defect cluster can be determined based on the ratio between the first and second standard deviations. For example, if the first standard deviation σ... x With the second standard deviation σ y The ratio between them is greater than a preset threshold (e.g., the preset threshold is 2), i.e., σ x / σ y A value greater than 2 indicates that the defect points within this defect cluster have a large distribution range along the X-axis, while their distribution along the Y-axis is relatively concentrated, exhibiting an overall linear distribution. Figure 2 As shown. If the second standard deviation σ y Compared with the first standard deviation σ x The ratio between them is greater than a preset threshold, i.e., σ y / σ x A value greater than 2 indicates that the defect points within this defect cluster have a large distribution range along the Y-axis, while their distribution along the X-axis is relatively concentrated, exhibiting an overall banded distribution. Figure 3 As shown, the linear and banded distribution patterns closely match the distribution characteristics of mechanical damage defects caused by mechanical contact, scratching, and other mechanical damage. Based on this, the distribution pattern of defect clusters is determined according to the ratio between the first and second standard deviations. If linear or banded defect clusters exist in the target cluster area of the wafer under test, it can be preliminarily determined that mechanical damage defects exist in the target cluster area.
[0096] If the ratio between the first standard deviation and the second standard deviation is less than or equal to the preset threshold, it indicates that the defect points are evenly distributed in the X-axis and Y-axis directions. The defect clusters are mostly circular or irregular clusters formed by random particle noise, which can rule out the possibility of mechanical damage.
[0097] In the case where it is initially determined that there are mechanical damage defects in the target cluster area, that is, when there are linear and / or banded defect clusters in the current wafer under test, the repeatability characteristics of the linear and / or banded defect clusters in multiple wafers are further determined.
[0098] Specifically, the wafer under test is compared with other adjacent wafers in the same batch. If multiple wafers show linear and / or banded defect clusters with consistent shape and distribution trend in a preset area at similar positions (preset area within a similar coordinate position or angle range), it can be determined that the defect cluster is a batch-specific, systematic mechanical damage defect, rather than random particle interference from a single wafer. This ultimately confirms that the batch of the wafer under test contains mechanical damage defects caused by contact with equipment such as chucks, carriers, or robotic arms.
[0099] In this embodiment of the invention, a second defect distribution map of a sample wafer is used as a reference benchmark. By comparing the second defect distribution map with the first defect distribution map of the wafer under test, the target clustering region of defect points in the wafer under test is determined, thus narrowing the scope of subsequent analysis. Then, the target clustering radius is dynamically determined based on the number of defect points in the wafer under test, the distance between adjacent defect points, and the starting size of the defect points. Cluster analysis is then performed based on the target clustering radius, which can adapt to different detection scenarios, balance suppressing over-detection and avoiding missed detection, and improve the accuracy of cluster analysis results. Finally, the mechanical damage defect detection result of the wafer under test is determined based on the distribution pattern of the defect clusters, thus improving the accuracy of mechanical damage defect detection.
[0100] In some embodiments, after determining the mechanical damage defect detection result of the wafer under test based on the distribution pattern of the defect clusters, the method further includes:
[0101] Obtain feedback data corresponding to the mechanical damage defect detection results, wherein the feedback data includes at least one of engineer review results and process verification results;
[0102] If the feedback data indicates that the misjudgment rate or missed judgment rate of mechanical damage defects is greater than or equal to a preset threshold, the target cluster radius is adjusted.
[0103] Establish a mapping relationship between the initial measurement size and the adjusted target cluster radius.
[0104] In this embodiment, after determining the mechanical damage defect detection result of the wafer under test, the engineer's re-judgment result of the current mechanical damage defect detection result or the subsequent process verification result can be used as feedback data to optimize and adjust the clustering parameters under different size ranges. Specifically, the false positive rate and false negative rate in each size range are statistically analyzed, and the clustering radius R and related clustering parameters are regressed and corrected accordingly. That is, when the feedback data indicates that the false positive rate or false negative rate of mechanical damage defects is greater than or equal to a preset threshold, the target clustering radius is adjusted so that R=f(M, D̄, Size) gradually converges to the optimal value range. Through long-term operation and multiple batch statistical regressions, a mapping relationship between the initial test size and the optimal target clustering radius is formed, realizing the continuous self-optimization of the mechanical damage judgment model.
[0105] For example, in the 19nm starting size mode, the total number M of defects on a single wafer can be 5000. The detection mode for the wafer under test is determined to be high-density detection mode based on the number M of defects in the wafer under test, the distance D̄ between adjacent defects, and the starting size of the defects. The initial clustering radius R0 can be set according to the equipment calibration value. In high-density mode, R is adjusted to: R = 0.6 × R0. Thus, after reducing R, random particles cannot form stable clusters due to insufficient spatial continuity, only retaining linear defect clusters with consistent spatial and directional height. If these linear defects are present in multiple consecutive wafers and are similar in position (±5mm) and consistent in direction, they can be identified as potential mechanical damage in the Chuck contact area. Figure 4 As shown. In addition, the judgment logic for Foup corners can be as follows: if the above method detects that a cluster recurs in a specific angular region at the edge of the wafer, and it is consistent with the Foup loading direction or contact point, then the association is judged as Foup corner interference risk.
[0106] This invention also provides a mechanical damage defect detection device, such as... Figure 5 As shown, it includes:
[0107] The first determining module 51 is used to determine the target aggregation region of defect points in the wafer under test based on the first defect distribution map of the wafer under test and the second defect distribution map of the sample wafer.
[0108] Analysis module 52 is used to perform cluster analysis on defect points in the target cluster region based on the target cluster radius to obtain multiple defect clusters. The target cluster radius is determined based on the number of defect points in the wafer under test, the distance between adjacent defect points, and the starting size of the defect points.
[0109] The second determining module 53 is used to determine the mechanical damage defect detection result of the wafer under test based on the distribution pattern of the defect clusters.
[0110] In some embodiments, the first determining module 51 is specifically used for:
[0111] The first defect distribution map of the wafer under test is divided into multiple first preset regions of the same area;
[0112] Obtain the defect density in each first preset region of the first defect distribution map and the defect density in each second preset region of the second defect distribution map. The second preset region corresponds to the first preset region. The defect density in any first preset region is determined by scanning and detecting the wafer under test. The defect density in any second preset region is the historical batch average defect density of the sample wafer in that preset region.
[0113] The target clustering area is determined based on the defect density in each first preset area of the first defect distribution map and the defect density in each second preset area of the second defect distribution map.
[0114] In some embodiments, determining the target clustering region based on the defect density in each first preset region of the first defect distribution map and the defect density in each second preset region of the second defect distribution map includes at least one of the following:
[0115] If the defect density in the first target preset area is greater than N times the defect density in the second target preset area, the first target preset area is determined as the target aggregation area, where N is greater than 1.
[0116] If the defect density of the first target preset area and the defect density of the third target preset area are in the same range, and the range is greater than, less than or equal to the defect density of the second target preset area, the first target preset area is determined as the target aggregation area.
[0117] Wherein, the first target preset area is any one of a plurality of first preset areas, the second target preset area is the area corresponding to the first target preset area among a plurality of second preset areas, the third target preset area is the area corresponding to the first target preset area in the third defect distribution map of the first wafer, and the first wafer is a wafer from the same batch as the wafer to be tested and adjacent to the wafer to be tested.
[0118] In some embodiments, the second determining module 53 is specifically used for:
[0119] Calculate the first standard deviation and the second standard deviation for each defect cluster. The first standard deviation is the standard deviation of the coordinates of each defect point contained in a single defect cluster on the X-axis, and the second standard deviation is the standard deviation of the coordinates of each defect point contained in a single defect cluster on the Y-axis. The X-axis and the Y-axis are perpendicular coordinate axes in a Cartesian coordinate system.
[0120] The distribution pattern of each defect cluster is determined based on the ratio between the first standard deviation and the second standard deviation;
[0121] When there are defect clusters distributed linearly or in a banded manner in the target aggregation region of the wafer under test, and there are defect clusters distributed linearly or in a banded manner in the region corresponding to the target aggregation region of the first wafer, it is determined that the batch of the wafer under test has mechanical damage defects, and the first wafer is a wafer from the same batch as the wafer under test and adjacent to the wafer under test.
[0122] In some embodiments, the apparatus further includes:
[0123] The first acquisition module is used to parse the scanning and detection results of the wafer under test to obtain the total number of defect points, the average distance between adjacent defect points, and the starting size of the defect points;
[0124] The third determining module is used to determine the target cluster radius corresponding to the wafer to be tested based on the total quantity, the average distance, and the starting size. The target cluster radius is inversely proportional to the total quantity and directly proportional to the average distance and the starting size.
[0125] In some embodiments, the apparatus further includes:
[0126] The second acquisition module is used to acquire feedback data corresponding to the mechanical damage defect detection results, wherein the feedback data includes at least one of engineer review results and process verification results.
[0127] The adjustment module is used to adjust the target cluster radius when the feedback data indicates that the misjudgment rate or missed judgment rate of mechanical damage defects is greater than or equal to a preset threshold.
[0128] A module is established to create a mapping relationship between the initial measurement size and the adjusted target cluster radius.
[0129] The mechanical damage defect detection device is capable of implementing each process of the above-described method embodiments, with one-to-one correspondence of technical features and achieving the same technical effect. To avoid repetition, it will not be described in detail here.
[0130] Please refer to Figure 6The present invention also provides an electronic device 60, including a processor 61, a memory 62, and a computer program stored in the memory 62 and executable on the processor 61. When the computer program is executed by the processor 61, it implements the various processes of the above-described mechanical damage defect detection method embodiments and can achieve the same technical effect. To avoid repetition, it will not be described again here.
[0131] This invention also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the various processes of the above-described mechanical damage defect detection method embodiments and achieves the same technical effect. To avoid repetition, it will not be described again here.
[0132] The computer-readable storage medium, as defined herein, includes both permanent and non-permanent, removable and non-removable media, and information storage can be achieved by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage, or any other non-transfer medium that can be used to store information accessible to the terminal device under test. As defined herein, computer-readable storage media does not include transient media, such as modulated data signals and carrier waves.
[0133] This invention also provides a computer program product, including computer instructions that, when executed by a processor, implement the above-described... Figure 1 The various processes of the method embodiments shown can achieve the same technical effect, and will not be described again here to avoid repetition.
[0134] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.
[0135] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of the present invention.
[0136] In the various method embodiments of this disclosure, the sequence numbers of each step are not intended to limit the order of the steps. For those skilled in the art, any changes in the order of the steps are within the scope of protection of this disclosure without any creative effort.
[0137] It should be noted that the various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, since the embodiments are basically similar to the product embodiments, the descriptions are relatively simple, and the relevant parts can be referred to the descriptions of the product embodiments.
[0138] The above description represents the preferred embodiments of this disclosure. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles described herein, and these improvements and modifications should also be considered within the scope of protection of this disclosure.
Claims
1. A method for detecting mechanical damage defects, characterized in that, include: Based on the first defect distribution map of the wafer under test and the second defect distribution map of the sample wafer, the target aggregation region of defect points in the wafer under test is determined; Cluster analysis is performed on the defect points in the target cluster region based on the target cluster radius to obtain multiple defect clusters. The target cluster radius is determined based on the number of defect points in the wafer under test, the distance between adjacent defect points, and the starting size of the defect points. The mechanical damage defect detection results of the wafer under test are determined based on the distribution pattern of the defect clusters.
2. The method according to claim 1, characterized in that, The determination of target cluster regions of defect points in the wafer under test, based on the first defect distribution map of the wafer under test and the second defect distribution map of the sample wafer, includes: The first defect distribution map of the wafer under test is divided into multiple first preset regions of the same area; Obtain the defect density in each first preset region of the first defect distribution map and the defect density in each second preset region of the second defect distribution map. The second preset region corresponds to the first preset region. The defect density in any first preset region is determined by scanning and detecting the wafer under test. The defect density in any second preset region is the historical batch average defect density of the sample wafer in that preset region. The target clustering area is determined based on the defect density in each first preset area of the first defect distribution map and the defect density in each second preset area of the second defect distribution map.
3. The method according to claim 2, characterized in that, Determining the target clustering region based on the defect density in each first preset region of the first defect distribution map and the defect density in each second preset region of the second defect distribution map includes at least one of the following: If the defect density in the first target preset area is greater than N times the defect density in the second target preset area, the first target preset area is determined as the target aggregation area, where N is greater than 1. If the defect density of the first target preset area and the defect density of the third target preset area are in the same range, and the range is greater than, less than or equal to the defect density of the second target preset area, the first target preset area is determined as the target aggregation area. Wherein, the first target preset area is any one of a plurality of first preset areas, the second target preset area is the area corresponding to the first target preset area among a plurality of second preset areas, the third target preset area is the area corresponding to the first target preset area in the third defect distribution map of the first wafer, and the first wafer is a wafer from the same batch as the wafer to be tested and adjacent to the wafer to be tested.
4. The method according to claim 1, characterized in that, The determination of the mechanical damage defect detection result of the wafer under test based on the distribution pattern of the defect clusters includes: Calculate the first standard deviation and the second standard deviation for each defect cluster. The first standard deviation is the standard deviation of the coordinates of each defect point contained in a single defect cluster on the X-axis, and the second standard deviation is the standard deviation of the coordinates of each defect point contained in a single defect cluster on the Y-axis. The X-axis and the Y-axis are perpendicular coordinate axes in a Cartesian coordinate system. The distribution pattern of each defect cluster is determined based on the ratio between the first standard deviation and the second standard deviation; When there are defect clusters distributed linearly or in a banded manner in the target aggregation region of the wafer under test, and there are defect clusters distributed linearly or in a banded manner in the region corresponding to the target aggregation region of the first wafer, it is determined that the batch of the wafer under test has mechanical damage defects, and the first wafer is a wafer from the same batch as the wafer under test and adjacent to the wafer under test.
5. The method according to any one of claims 1 to 4, characterized in that, Before performing cluster analysis on defect points in the target clustering region based on the target clustering radius to obtain multiple defect clusters, the method further includes: The scanning and detection results of the wafer under test are analyzed to obtain the total number of defects, the average distance between adjacent defects, and the starting size of the defects. The target cluster radius corresponding to the wafer to be tested is determined based on the total number, the average distance, and the starting size. The target cluster radius is inversely proportional to the total number and directly proportional to the average distance and the starting size.
6. The method according to any one of claims 1 to 4, characterized in that, After determining the mechanical damage defect detection result of the wafer under test based on the distribution pattern of the defect clusters, the method further includes: Obtain feedback data corresponding to the mechanical damage defect detection results, wherein the feedback data includes at least one of engineer review results and process verification results; If the feedback data indicates that the misjudgment rate or missed judgment rate of mechanical damage defects is greater than or equal to a preset threshold, the target cluster radius is adjusted. Establish a mapping relationship between the initial measurement size and the adjusted target cluster radius.
7. A mechanical damage defect detection device, characterized in that, include: The first determining module is used to determine the target aggregation region of defect points in the wafer under test based on the first defect distribution map of the wafer under test and the second defect distribution map of the sample wafer. The analysis module is used to perform cluster analysis on defect points in the target cluster region based on the target cluster radius to obtain multiple defect clusters. The target cluster radius is determined based on the number of defect points in the wafer under test, the distance between adjacent defect points, and the starting size of the defect points. The second determining module is used to determine the mechanical damage defect detection result of the wafer under test based on the distribution pattern of the defect clusters.
8. An electronic device, characterized in that, include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps of the method as described in any one of claims 1 to 6.
9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the method as described in any one of claims 1 to 6.
10. A computer program product, characterized in that, Includes computer instructions that, when executed by a processor, implement the steps of the method as described in any one of claims 1 to 6.