Method and apparatus for detecting arc-shaped scratches in a wafer

Abstract

Embodiments of the present specification provide a method and device for detecting a circular arc scratch in a wafer. A specific embodiment of the method includes: determining at least one abnormal area in a wafer image based on a plurality of abnormal dies contained in the wafer image, wherein the abnormal area is an area connected by abnormal dies; excluding a non-target area from the at least one abnormal area based on a width of the at least one abnormal area; determining a characteristic circle based on position coordinates of dies in a target area remaining in the at least one abnormal area; and determining that the target area forms a circular arc scratch in response to determining that a standard deviation between the dies in the target area and the characteristic circle and a radius of the characteristic circle are within a preset range.

Description

Technical Field

[0001] This specification relates to the field of microelectromechanical system (MEMS) chip manufacturing, and more particularly to a method and apparatus for detecting arc-shaped scratches in a wafer. Background Technology

[0002] In the manufacturing process of microelectromechanical systems (MEMS) chips, the wafer is the fundamental raw material. During chip manufacturing, the etched wafer needs to be polished. If particles fall off or there are residues on the polishing dies, arc scratches will be generated during the relative rotation of the dies and wafer, leading to a sharp drop in chip yield. Therefore, quickly and accurately detecting arc scratches on the wafer and performing root cause analysis as early as possible is a crucial step in controlling the escalation of chip yield decline. Summary of the Invention

[0003] The embodiments described in this specification illustrate a method and apparatus for detecting arc-shaped scratches in a wafer, which can quickly and accurately detect arc-shaped scratches in a wafer.

[0004] According to a first aspect, a method for detecting arc-shaped scratches in a wafer is provided, comprising: determining at least one abnormal region in the wafer image based on a plurality of abnormal dies included in the wafer image, wherein the abnormal region is a region formed by connecting abnormal dies; excluding non-target regions from the at least one abnormal region based on the width of the at least one abnormal region; determining a feature circle for the remaining target region in the at least one abnormal region based on the position coordinates of the dies in the target region; and determining that the target region forms an arc-shaped scratch in response to determining that the standard deviation between the dies in the target region and the feature circle and the radius of the feature circle are within a preset range.

[0005] According to a second aspect, an apparatus for detecting arc-shaped scratches in a wafer is provided, comprising: a first determining unit configured to determine at least one abnormal region in a wafer image based on a plurality of abnormal dies included in the wafer image, wherein the abnormal region is a region formed by connecting abnormal dies; an exclusion unit configured to exclude non-target regions from the at least one abnormal region based on the width of the at least one abnormal region; a second determining unit configured to determine a feature circle for the remaining target regions in the at least one abnormal region based on the position coordinates of dies in the target region; and a third determining unit configured to determine that the target region has an arc-shaped scratch in response to determining that the standard deviation between the die and the feature circle in the target region and the radius of the feature circle are within a preset range.

[0006] According to a third aspect, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed in a computer, causes the computer to perform the method described in any implementation of the first aspect.

[0007] According to a fourth aspect, a computing device is provided, including a memory and a processor, characterized in that the memory stores executable code, and when the processor executes the executable code, it implements the method described in any implementation of the first aspect.

[0008] The method and apparatus for detecting arc-shaped scratches in a wafer according to the embodiments of this specification first determine at least one abnormal region in the wafer image based on multiple abnormal dies included in the wafer image. Then, non-target regions are excluded from the at least one abnormal region based on the width of the abnormal region. A feature circle is determined based on the position coordinates of the dies included in the remaining target region. If the standard deviation between the die and the feature circle in the target region and the radius of the feature circle are within a preset range, it is determined that the target region has formed an arc-shaped scratch. Thus, by excluding some non-target regions that cannot form arc-shaped scratches through the width of the abnormal region, and further determining whether the target region has formed an arc-shaped scratch based on the feature circle determined by the dies in the remaining target region, it is possible to quickly and accurately detect arc-shaped scratches in the wafer. Attached Figure Description

[0009] Figure 1 A schematic diagram is shown illustrating one application scenario in which the embodiments of this specification can be applied;

[0010] Figure 2 A flowchart of a method for detecting arc-shaped scratches in a wafer according to one embodiment is shown;

[0011] Figure 3(a) shows an example diagram of mapping the number of abnormal cores in the abnormal region to the coordinate axis, taking the abnormal region 1011 as an example;

[0012] Figure 3(b) shows an example diagram of mapping the number of abnormal cores in the abnormal region to the coordinate axis, taking the abnormal region 1012 as an example;

[0013] Figure 4 A schematic diagram illustrating an example of a shared edge is shown;

[0014] Figure 5 A schematic diagram showing the smoothing of the position coordinates of multiple dies is shown;

[0015] Figure 6 A schematic diagram is shown illustrating the determination of the circumcircle based on triangle △ABC formed by a certain set of target candidates;

[0016] Figure 7A schematic block diagram of an apparatus for detecting arc-shaped scratches in a wafer according to one embodiment is shown. Detailed Implementation

[0017] The technical solutions provided in this specification will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the relevant invention and not intended to limit the invention. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings. It should be noted that, unless otherwise specified, the embodiments and features described in this specification can be combined with each other.

[0018] In practice, manufactured wafers may contain various defects. Some defects share the same formation mechanism and tend to cluster together, forming specific patterns. Studying these pattern characteristics and analyzing their causes helps improve processes and increase chip yield. As mentioned earlier, during wafer polishing, if particles fall or residue remains on the polishing disc, arc-shaped scratches will be generated during the relative rotation of the disc and wafer. Therefore, quickly and accurately detecting arc-shaped scratches on wafers is beneficial for improving processes and increasing chip yield.

[0019] Therefore, embodiments of this specification provide a method for detecting arc-shaped scratches in a wafer, which can quickly and accurately detect arc-shaped scratches in a wafer. Figure 1 This diagram illustrates an application scenario in which embodiments of this specification can be applied. For example... Figure 1 As shown, firstly, a wafer image 101 can be obtained. Each small grid on the wafer image 101 can represent a die. The wafer image 101 can include multiple abnormal dies, which refer to dies that have been detected to have defects. Figure 1In the example shown, black squares represent abnormal dies, and white squares represent normal dies. Then, based on the multiple abnormal dies contained in wafer image 101, abnormal regions 1011, 1012, and 1013 can be identified in wafer image 101. Abnormal regions 1011, 1012, and 1013 are regions formed by connecting abnormal dies. Since the trace length of an arc-shaped scratch should have a minimum value, the number of abnormal dies contained in the abnormal regions should have a lower threshold. In this example, the number of abnormal dies contained in the abnormal regions is no less than 6. Then, based on the width of abnormal regions 1011, 1012, and 1013, non-target regions are excluded from these three abnormal regions. In this example, since the width of abnormal region 1013 is too wide and clearly does not conform to the characteristics of an arc-shaped scratch, abnormal region 1013 is excluded as a non-target region, while abnormal regions 1011 and 1012 are retained as target regions. Then, it is determined whether arc-shaped scratches have formed in abnormal regions 1011 and 1012 respectively. Taking abnormal region 1011 as an example, a feature circle can be determined based on the position coordinates of the core in abnormal region 1011. When the standard deviation between the core in abnormal region 1011 and the feature circle is within a preset range, and the radius of the feature circle is also within a preset range, it is determined that arc-shaped scratches have formed in abnormal region 1011.

[0020] See also Figure 2 , Figure 2 A flowchart illustrating a method for detecting arc-shaped scratches in a wafer according to one embodiment is shown. It will be understood that this method can be performed by any device, apparatus, platform, or cluster of devices with computing and processing capabilities. Figure 2 As shown, the method for detecting arc-shaped scratches in a wafer may include the following steps 201-204, specifically:

[0021] Step 201: Based on the multiple abnormal dies contained in the wafer image, determine at least one abnormal region in the wafer image.

[0022] In this embodiment, a wafer image can be acquired, which may contain multiple defective dies. Here, the wafer is the carrier for designing integrated circuits; analog or digital circuits designed by designers are ultimately implemented on the wafer. Each small square on each wafer represents a complete chip circuit unit, such as a CPU or a memory, and is called a die. Here, a defective die can refer to a die that has been detected as having defects. These defects can include various types of defects that may occur on the wafer, such as wafer surface redundancy, mechanical scratches, etc. In this example, each die can be displayed as a small square in the wafer image. Based on the multiple defective dies contained in the wafer image, at least one abnormal region in the wafer image can be determined, where the abnormal region can be an area formed by connecting defective dies. For example, connectivity analysis can be performed on multiple small squares representing abnormal dies in a wafer image. For instance, using an 8-connectivity mode, the small square corresponding to an abnormal die can be connected to the small squares corresponding to abnormal dies in the eight directions surrounding it (including: up, down, left, right, upper left, upper right, lower left, and lower right), thereby obtaining at least one abnormal region.

[0023] It is understood that, in order to improve the quality of wafer images, preprocessing can be performed on the wafer images before using the methods described in this embodiment. For example, denoising processing can be performed on the wafer images. In practice, the denoising processing methods used may include, but are not limited to, Fourier transform, convolution kernel filtering, Radon transform, wavelet transform, erosion processing, Gaussian filtering, median filtering, PM (Perona-Malik) equation filtering, etc.

[0024] In practice, based on the characteristics of arc-shaped scratches, the trace length of arc-shaped scratches should have a minimum value. Therefore, the number of abnormal dies contained in the abnormal region should have a lower threshold. Therefore, in some implementations, step 201 may specifically include: step (1), performing connectivity analysis on multiple abnormal dies to obtain at least one abnormal region formed by connecting abnormal dies; step (2), removing abnormal regions containing fewer than the first threshold number of dies to obtain at least one abnormal region.

[0025] In this implementation, the first threshold in step (2) can be preset according to actual needs. For example, the first threshold can be set to 6, 8, etc. Through this implementation, abnormal areas that do not conform to the characteristics of arc-shaped scratches can be easily and quickly eliminated.

[0026] Step 202: Based on the width of at least one abnormal region, exclude non-target regions from at least one abnormal region.

[0027] In this embodiment, based on the characteristics of arc-shaped scratches, it is known that the width of arc-shaped scratches is usually not too wide. Figure 1 Taking the abnormal region 1013 as an example, its width is clearly too large, failing to meet the characteristics of a circular arc-shaped scratch. Therefore, abnormal region 1013 can be excluded as a non-target region and not processed further, thus improving calculation speed. In practice, there are several ways to determine whether the width of an abnormal region is too wide.

[0028] In some implementations, analysis has revealed a correlation between the width of the abnormal region and the number of abnormal dies contained in the rows and columns within that region. For example, if most rows or columns in the abnormal region contain a large number of abnormal dies, the width of the abnormal region will be larger, and its shape will tend to be sheet-like rather than arc-shaped. Based on this, step 202 can specifically include the following steps (a)-(c):

[0029] Step (a): For each abnormal region in at least one abnormal region, calculate the first proportion of rows in the abnormal region where the number of dies exceeds the second threshold, and the second proportion of columns where the number of dies exceeds the second threshold.

[0030] In this implementation, as a statistical method, as shown in Figure 3(a), Figure 3(a) shows an example diagram of mapping the number of abnormal dies in the abnormal region 1011 to the coordinate axis. In the example shown in Figure 3(a), the number of abnormal dies contained in each column of the abnormal region 1011 can be mapped to the X-axis, and the number of abnormal dies contained in each row can be mapped to the Y-axis. As shown in Figure 3(b), Figure 3(b) shows an example diagram of mapping the number of abnormal dies in the abnormal region 1012 to the coordinate axis. In the example shown in Figure 3(b), the number of abnormal dies contained in each column of the abnormal region 1012 can be mapped to the X-axis, and the number of abnormal dies contained in each row can be mapped to the Y-axis. Figure 3(a) and 3(b) As the example shown illustrates, the greater the number of abnormal dies, the larger the mapped value.

[0031] Then, based on the mapping of abnormal die counts onto the coordinate axes, the first proportion of rows where the die count exceeds the second threshold and the second proportion of columns where the die count exceeds the second threshold in each abnormal region can be calculated. Here, the second threshold can be set according to actual conditions. For example, it can be determined based on statistical values ​​(e.g., mean, median, etc.) of the abnormal die counts in the rows and columns with known arc-shaped scratches. Alternatively, the second threshold can be set based on manual experience, for example, setting it to 2.

[0032] Step (b): Determine the judgment index based on the first ratio and the second ratio.

[0033] In this implementation, the judgment index can be determined in various ways based on the statistically obtained first and second proportions. For example, the smaller of the first and second proportions can be used as the judgment index. Alternatively, the average of the first and second proportions can be used as the judgment index.

[0034] Step (c) is to determine that the width of the abnormal area does not conform to the arc-shaped scratch in response to the determination that the judgment index exceeds the third threshold.

[0035] In this implementation, a third threshold can be preset according to actual needs, for example, the third threshold can be set to 0.3. Then, it is determined whether the judgment index corresponding to each abnormal area exceeds the third threshold. If an abnormal area exceeds the threshold, it means that the width of the abnormal area is too large and does not conform to the arc-shaped scratch, thus determining that the abnormal area is not the target area and can be excluded as a non-target area.

[0036] In other implementations, analysis revealed a correlation between the width of the abnormal region and the number of shared edges between the abnormal dies within that region. Generally, a larger number of shared edges indicates a denser distribution of abnormal dies, a wider abnormal region, and a shape more plate-like than arc-shaped. Based on this, step 202 can further include steps ①-③, specifically:

[0037] Step ①: For each abnormal region in at least one abnormal region, count the total number of dies in that abnormal region and the number of shared edges of each die.

[0038] In this implementation, a shared edge can refer to an adjacent edge between two dies. For example... Figure 4 As shown, Figure 4 A schematic diagram illustrating an example of a shared edge is shown. Figure 4 In the example shown, for abnormal dies A, B, and C in abnormal region 1012, A and B share a common edge, and B and C share a common edge. In this example, gray rectangles are used to represent the common edges.

[0039] Step 2: Select the dies with more than the fourth threshold as feature dies and determine the ratio of the number of feature dies to the total number of dies in the abnormal region.

[0040] In this implementation, the fourth threshold can be set according to actual conditions; for example, it can be set to 3. Here, the ratio between the number of characteristic dies in each abnormal region and the total number of dies in that abnormal region can be calculated.

[0041] Step ③: In response to the determination that the ratio exceeds the fifth threshold, it is determined that the width of the abnormal area does not conform to the arc-shaped scratch, and the abnormal area is determined to be the target area.

[0042] In this implementation, the fifth threshold can be set according to the actual situation; for example, it can be set to 0.2. Then, it is determined whether the ratio corresponding to each abnormal region exceeds the fifth threshold. If it does, it indicates that the width of the abnormal region is too large and does not conform to the arc-shaped scratch. This abnormal region is not the target region and can be excluded as a non-target region.

[0043] As another implementation method, the total number of shared edges contained in the abnormal region can be counted. Then, the ratio between the total number of shared edges and the total number of cores in the abnormal region is calculated. If this ratio is greater than a preset threshold (e.g., 1.05), it indicates that the width of the abnormal region is too large and does not conform to the arc-shaped scratch. The abnormal region is not the target region and can be excluded as a non-target region.

[0044] Step 203: For the remaining target area in at least one abnormal area, determine the feature circle based on the position coordinates of the core in the target area.

[0045] In this embodiment, the wafer image can be mapped onto a two-dimensional coordinate system to obtain the position coordinates of each die in the target area, such as the coordinates of the four vertices of the small square corresponding to the die, and the coordinates of the center point, etc. Then, a feature circle can be determined based on the position coordinates of the dies in the target area.

[0046] Research has found that at least three points are needed to define a circle. Based on this, firstly, multiple coordinate points can be determined based on the positional coordinates of the tube cores in the target area. For example, the center coordinates of each tube core in the target area can be selected as the determined coordinate points. Alternatively, the coordinates of a vertex of each tube core in the target area can be selected as the determined coordinate points. Then, any three of these coordinate points can be grouped together to define a circle. Finally, a feature circle can be determined based on the resulting multiple circles. For example, the center of the feature circle can be determined based on the centers of the multiple circles, and the radius of the feature circle can be determined based on the radii of the multiple circles.

[0047] In some implementations, step 203 above may specifically include the following steps S1-S4, specifically:

[0048] Step S1: Determine multiple coordinate points based on the position coordinates of multiple cores in the target area.

[0049] In this implementation, multiple coordinate points can be determined in various ways based on the position coordinates of multiple dies in the target area. For example, the center point coordinates of multiple dies in the target area can be smoothed to obtain multiple smoothed coordinate points. Here, smoothing the center point coordinates of multiple dies can mean merging the center point coordinates of adjacent dies in the target area into a single coordinate, where the value is the midpoint of the two center point coordinates, i.e., the midpoint of the common edge of adjacent dies. Figure 5 As shown, Figure 5 A schematic diagram is shown showing the smoothing of the position coordinates of multiple tube dies. Figure 5 Taking abnormal dies A, B, and C in abnormal region 1012 as examples, the center point coordinates of adjacent dies A and B are merged into coordinate M, and the center point coordinates of adjacent dies B and C are merged into coordinate N. Similarly, the position coordinates of multiple dies in abnormal region 1012 can be smoothed to obtain multiple coordinate points.

[0050] Step S2: Determine multiple candidate groups based on multiple coordinate points.

[0051] In this implementation, each candidate group can include three coordinate points. For example, any three coordinate points from multiple sets can be selected to form a candidate group, thus obtaining multiple candidate groups. For instance, assuming there are n coordinate points, sorting and combining them yields C... 3 n = n(n-1)(n-2) / 6 candidate groups. It's understandable that some candidate groups may not have enough points to form a circle; therefore, candidate groups that cannot form a circle can be deleted. For example, if three points in a candidate group lie on a straight line, then those three points cannot form a circle, and therefore, that candidate group can be deleted.

[0052] Analysis revealed that candidate groups with relatively large and evenly distributed coordinate points (e.g., two coordinate points at the beginning and end of the target area, and one coordinate point at the center of the target area) more closely resembled the morphology of arc-shaped scratches on a wafer. Furthermore, the distance and distribution of coordinate points are related to their values ​​on the X and Y axes; therefore, candidate groups can be determined based on the X and Y axis values ​​of the coordinate points.

[0053] In one implementation, step S2 above can be achieved through the following steps 1)-3), specifically:

[0054] Step 1) Sort the multiple coordinate points based on their X-axis coordinates. For example, they can be sorted in ascending or descending order of their X-axis coordinates.

[0055] Step 2): Based on the sorting results, select the coordinates of the preset positions (previous, next, and middle) on the X-axis.

[0056] Step 3): Based on the coordinate points selected in Step 2), a candidate group is formed.

[0057] For example, suppose we have a set set1 for coordinates at the first preset position on the X-axis, a set set2 for coordinates at the next preset position, and a set set3 for coordinates at the middle preset position. Then, we can take one coordinate point from each set set1, set2, and set3 to form a candidate group. Therefore, by selecting multiple times from sets set1, set2, and set3, we can obtain multiple candidate groups.

[0058] Using the same method as the X-axis, multiple coordinate points can also be sorted based on the size of the Y-axis coordinates. Then, based on the sorting results, coordinate points that are sorted in the first, last, and middle preset positions on the Y-axis are selected. Finally, a candidate group is formed based on the selected coordinate points.

[0059] It is understandable that in practice, one can use the X-axis-based method alone, or the Y-axis-based method alone, or combine the X-axis and Y-axis methods to obtain more alternative groups.

[0060] In another implementation, step S2 above can also be achieved through the following steps (1)-(3), specifically:

[0061] Step 1) For each of the multiple coordinate points, calculate the calculated value of that coordinate point on the X-axis and Y-axis (e.g., sum, product, weighted sum, etc.).

[0062] Step 2) Sort the multiple calculated values ​​and select the coordinates of the preset positions (first, last, and middle) based on the sorting results. For example, the multiple calculated values ​​can be sorted in descending or ascending order.

[0063] Step 3) Based on the coordinate points selected in Step 2), a candidate group is formed.

[0064] In another implementation, step S2 can also be achieved through the following steps I) and II), specifically:

[0065] Step I): Determine the target coordinate point based on multiple coordinate points.

[0066] In this implementation, the X-axis coordinate value of the target point can be the average of the X-axis coordinate values ​​of multiple points, and the Y-axis coordinate value of the target point can be the average of the Y-axis coordinate values ​​of multiple points.

[0067] Step II) Select coordinate points from multiple coordinate points based on their distance from the target coordinate point to form a candidate group.

[0068] For example, assuming the target coordinates are (μ1, μ2), we can select a set of coordinates near (μ1, μ2), a set of coordinates furthest to the left of (μ1, μ2), and a set of coordinates furthest to the right of (μ1, μ2). Then, we select one coordinate point from each of these three sets to form a candidate set. By selecting multiple times from these three sets, we can obtain multiple candidate sets.

[0069] Step S3: Based on the feature information of the triangles formed by each candidate group, select at least one target candidate group from multiple candidate groups.

[0070] Analysis revealed that when the three coordinate points in a candidate group are close together, the circle determined by this candidate group is either too small or too large, neither of which conforms to the shape of an arc-shaped scratch in a wafer. Therefore, it is necessary to further eliminate unreasonable candidate groups and retain the target candidate group. For example, at least one target candidate group can be selected from multiple candidate groups based on the characteristic information of the triangles formed by the three coordinate points in the candidate groups. Here, the characteristic information of the triangle can include, but is not limited to, the minimum side length, perimeter, area, etc. Taking the minimum side length of the triangle as an example, assuming that the three coordinate points in each candidate group can form triangle △ABC, and the three sides of triangle △ABC are a, b, and c, then the characteristic distance of each candidate group is d, where d = min(a, b, c). Then, the multiple characteristic distances corresponding to multiple candidate groups are sorted, and the candidate group with the larger characteristic distance is selected as the target candidate group. For example, multiple candidate groups are sorted in ascending order of their corresponding feature distances, and the top E% of candidate groups are selected as the target candidate groups. Here, E can be set according to actual needs.

[0071] Step S4: Determine the feature circle based on the circumcircle corresponding to the triangle formed by each target candidate group.

[0072] In this implementation, three coordinate points from each candidate target group can be used to form a triangle, and the circumcircle of each triangle can be determined. For example... Figure 6 As shown, Figure 6 This diagram illustrates the process of determining the circumcircle of triangle △ABC formed by a set of target candidates. Subsequently, a characteristic circle can be determined based on at least one circumcircle. For example, a circle can be randomly selected from at least one circumcircle as the characteristic circle.

[0073] In some implementations, step S4 above may specifically include the following steps (1)-(4):

[0074] Step (1): Calculate the center and radius of the circumcircle corresponding to the triangle formed by each target candidate group.

[0075] Step (2): Determine the target center based on the multiple center points. Here, the X-axis coordinate of the target center can be a statistical value of the X-axis coordinates of the multiple center points, such as the mean, median, etc. Similarly, the Y-axis coordinate of the target center can be a statistical value of the Y-axis coordinates of the multiple center points, such as the mean, median, etc.

[0076] Step (3): Determine the target radius based on multiple radii. Here, the target radius can be a statistical value of the above multiple radii, such as the mean, median, etc.

[0077] Step (4): Determine the feature circle based on the target circle center and target radius. This implementation method can eliminate the influence of noise data on the results, making the obtained feature circle more accurate.

[0078] Step 204: In response to determining that the standard deviation between the core and the feature circle in the target area and the radius of the feature circle are within a preset range, determine that an arc-shaped scratch is formed in the target area.

[0079] In this embodiment, the presence or absence of an arc-shaped scratch in the target area can be determined based on the obtained feature circle. Specifically, the standard deviation between the die position coordinates in the target area and the feature circle is first calculated, and then it is determined whether the standard deviation is within a preset range. Furthermore, analysis has revealed that the radius of curvature of an arc-shaped scratch on a wafer is typically neither particularly large nor particularly small, generally falling between one and several wafer radii, for example, [0.8, 5]. Therefore, the range of the radius of curvature can be preset based on experience. Then, it can be determined whether the radius of the feature circle is within the preset range. If both the standard deviation and the radius of the feature circle are within their respective ranges, it can be determined that an arc-shaped scratch has formed in the target area.

[0080] In some implementations, the method for detecting arc-shaped scratches in a wafer may further include: Figure 2 The following content, not shown in the image, describes how, in response to determining a target area and forming an arc-shaped scratch, annotations are made on the wafer image based on the center coordinates and radius of the feature circle. For example, the center coordinates and radius of the feature circle can be annotated on the wafer image as text. Alternatively, the feature circle can be drawn on the wafer image, and its center and radius can be displayed.

[0081] Reviewing the above process, in the embodiments of this specification, firstly, at least one abnormal region in the wafer image is determined based on multiple abnormal dies included in the wafer image. Then, non-target regions are excluded from the at least one abnormal region based on the width of the abnormal region. A feature circle is determined based on the position coordinates of the dies included in the remaining target region. If the standard deviation between the die and the feature circle in the target region and the radius of the feature circle are within a preset range, it is determined that the target region forms an arc-shaped scratch. Thus, by excluding some non-target regions that cannot form arc-shaped scratches through the width of the abnormal region, and further determining whether the target region forms an arc-shaped scratch based on the feature circle determined from the dies in the remaining target region, it is possible to quickly and accurately detect arc-shaped scratches in the wafer.

[0082] According to another embodiment, an apparatus for detecting arc-shaped scratches in a wafer is provided. The above-described apparatus for detecting arc-shaped scratches in a wafer can be deployed in any device, platform, or cluster of devices with computing and processing capabilities.

[0083] Figure 7 A schematic block diagram of an apparatus for detecting arc-shaped scratches in a wafer according to one embodiment is shown. Figure 7 As shown, the apparatus 700 for detecting arc-shaped scratches in a wafer includes: a first determining unit 701 configured to determine at least one abnormal region in the wafer image based on a plurality of abnormal dies included in the wafer image, wherein the abnormal region is a region formed by connecting abnormal dies; an exclusion unit 702 configured to exclude non-target regions from the at least one abnormal region based on the width of the at least one abnormal region; a second determining unit 703 configured to determine a feature circle for the remaining target regions in the at least one abnormal region based on the position coordinates of the dies in the target region; and a third determining unit 704 configured to determine that the target region has an arc-shaped scratch in response to determining that the standard deviation between the die and the feature circle in the target region and the radius of the feature circle are within a preset range.

[0084] In some optional implementations of this embodiment, the second determining unit 703 includes: a coordinate point determining unit (not shown in the figure), configured to determine multiple coordinate points based on the position coordinates of multiple cores in the target area; a candidate group determining unit (not shown in the figure), configured to determine multiple candidate groups based on the multiple coordinate points, wherein each candidate group includes three coordinate points; a selecting unit (not shown in the figure), configured to select at least one target candidate group from the multiple candidate groups based on the feature information of the triangles formed by each candidate group; and a feature circle determining unit (not shown in the figure), configured to determine a feature circle based on the circumcircle corresponding to the triangles formed by each target candidate group.

[0085] In some optional implementations of this embodiment, the candidate group determining unit is further configured to: sort the plurality of coordinate points based on the magnitude of the X-axis coordinates; select coordinate points that are sorted at the first, last, and middle preset positions on the X-axis according to the sorting result; form a candidate group based on the selected coordinate points; and / or, sort the plurality of coordinate points based on the magnitude of the Y-axis coordinates; select coordinate points that are sorted at the first, last, and middle preset positions on the Y-axis according to the sorting result; form a candidate group based on the selected coordinate points.

[0086] In some optional implementations of this embodiment, the candidate group determination unit is further configured to: calculate the calculated value of each of the above multiple coordinate points on the X-axis and Y-axis; sort the multiple calculated values, and select the coordinate points whose calculated values ​​are in the first preset position, the last preset position, and the middle preset position according to the sorting result; and form a candidate group based on the selected coordinate points.

[0087] In some optional implementations of this embodiment, the candidate group determining unit is further configured to: determine a target coordinate point based on the plurality of coordinate points, wherein the X-axis coordinate value of the target coordinate point is the average of the X-axis coordinate values ​​of the plurality of coordinate points, and the Y-axis coordinate value of the target coordinate point is the average of the Y-axis coordinate values ​​of the plurality of coordinate points; and select coordinate points from the plurality of coordinate points based on the distance between the target coordinate point and the target coordinate point to form a candidate group.

[0088] In some optional implementations of this embodiment, the feature circle determination unit is further configured to: calculate the center and radius of the circumcircle corresponding to the triangle formed by each target candidate group; determine the target circle center based on the plurality of circle centers, wherein the X-axis coordinate value of the target circle center is a statistical value of the X-axis coordinate values ​​of the plurality of circle centers, and the Y-axis coordinate value of the target circle center is a statistical value of the Y-axis coordinate values ​​of the plurality of circle centers; determine the target radius based on the plurality of radii, wherein the target radius is a statistical value of the plurality of radii; and determine the feature circle based on the target circle center and the target radius.

[0089] In some optional implementations of this embodiment, the apparatus 700 further includes a marking unit (not shown in the figure), configured to mark the wafer image according to the center coordinates and radius of the feature circle in response to determining that the target area forms an arc-shaped scratch.

[0090] In some optional implementations of this embodiment, the first determining unit 701 is further configured to: perform connectivity analysis on the above-mentioned plurality of abnormal dies to obtain at least one abnormal region formed by the connection of abnormal dies; remove abnormal regions containing fewer than a first threshold dies to obtain at least one abnormal region.

[0091] In some optional implementations of this embodiment, the exclusion unit 702 is further configured to: for each abnormal region in the at least one abnormal region, calculate the first proportion of rows in the abnormal region where the number of dies exceeds the second threshold, and the second proportion of columns where the number of dies exceeds the second threshold; determine a judgment index based on the first proportion and the second proportion; in response to determining that the judgment index exceeds the third threshold, determine that the width of the abnormal region does not conform to the arc-shaped scratch, and determine that the abnormal region is not the target region.

[0092] In some optional implementations of this embodiment, the exclusion unit 702 is further configured to: for each abnormal region in the at least one abnormal region, count the total number of dies in the abnormal region and the number of common edges of each die, wherein a common edge refers to an adjacent edge between two dies; take dies with a number of common edges exceeding a fourth threshold as feature dies, and determine the ratio between the number of feature dies and the total number; in response to determining that the ratio exceeds a fifth threshold, determine that the width of the abnormal region does not conform to the arc-shaped scratch, and determine that the abnormal region is not the target region.

[0093] According to another embodiment, a computer-readable storage medium is also provided, on which a computer program is stored, which, when executed in a computer, causes the computer to perform... Figure 2 The method described.

[0094] According to another embodiment, a computing device is also provided, including a memory and a processor, characterized in that the memory stores executable code, and when the processor executes the executable code, it implements... Figure 2 The method described.

[0095] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0096] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented in hardware, processor-executed software modules, or a combination of both. The software modules can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, CD-ROMs, or any other form of storage medium known in the art.

[0097] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for detecting arc-shaped scratches in a wafer, comprising: Based on multiple abnormal dies contained in a wafer image, at least one abnormal region in the wafer image is determined, wherein the abnormal region is a region formed by connecting abnormal dies. Based on the width of the at least one abnormal region, exclude non-target regions from the at least one abnormal region; For the remaining target area in the at least one abnormal region, a feature circle is determined based on the position coordinates of the core in the target area; In response to determining that the standard deviation between the core and the feature circle in the target area and the radius of the feature circle are within a preset range, it is determined that an arc-shaped scratch is formed in the target area.

2. The method according to claim 1, wherein, Determining the feature circle based on the position coordinates of the core in the target region includes: Based on the position coordinates of multiple cores in the target area, multiple coordinate points are determined; Based on the multiple coordinate points, multiple candidate groups are determined, wherein each candidate group includes three coordinate points; Based on the feature information of the triangles formed by each candidate group, at least one target candidate group is selected from the plurality of candidate groups. The characteristic circle is determined based on the circumcircle corresponding to the triangle formed by each target candidate group.

3. The method according to claim 2, wherein, The step of determining multiple candidate groups based on the multiple coordinate points includes: The multiple coordinate points are sorted based on the magnitude of the X-axis coordinates; Based on the sorting results, select the coordinates of the preset positions on the X-axis: the first preset position, the last preset position, and the middle preset position. Based on the selected coordinate points, a candidate group is formed; and / or, The multiple coordinate points are sorted based on the magnitude of the Y-axis coordinates; Based on the sorting results, select the coordinates of the preset positions on the Y-axis: the first preset position, the last preset position, and the middle preset position. Based on the selected coordinate points, a candidate group is formed.

4. The method according to claim 2, wherein, The step of determining multiple candidate groups based on the multiple coordinate points includes: For each of the plurality of coordinate points, calculate the calculated values ​​of that coordinate point on the X-axis and Y-axis; Sort multiple operation values ​​and select the coordinates of the operation values ​​in the first, last, and middle preset positions according to the sorting results; Based on the selected coordinate points, a candidate group is formed.

5. The method according to claim 2, wherein, The step of determining multiple candidate groups based on the multiple coordinate points includes: Based on the plurality of coordinate points, a target coordinate point is determined, wherein the X-axis coordinate value of the target coordinate point is the average of the X-axis coordinate values ​​of the plurality of coordinate points, and the Y-axis coordinate value of the target coordinate point is the average of the Y-axis coordinate values ​​of the plurality of coordinate points; Based on the distance between the coordinate points and the target coordinate point, coordinate points are selected from the plurality of coordinate points to form a candidate group.

6. The method according to claim 2, wherein, The determination of the feature circle based on the circumcircle corresponding to the triangle formed by each target candidate group includes: Calculate the center and radius of the circumcircle corresponding to the triangle formed by each target candidate group; Based on the plurality of circle centers, a target circle center is determined, wherein the X-axis coordinate value of the target circle center is a statistical value of the X-axis coordinate values ​​of the plurality of circle centers, and the Y-axis coordinate value of the target circle center is a statistical value of the Y-axis coordinate values ​​of the plurality of circle centers; A target radius is determined based on the plurality of radii, wherein the target radius is a statistical value of the plurality of radii; The feature circle is determined based on the target center and the target radius.

7. The method according to claim 1, wherein, The method further includes: In response to determining that the target area forms an arc-shaped scratch, a mark is made on the wafer image according to the center coordinates and radius of the feature circle.

8. The method according to claim 1, wherein, The step of determining at least one abnormal region in a wafer image based on multiple abnormal dies included in the wafer image includes: Connectivity analysis is performed on the multiple abnormal dies to obtain at least one abnormal region formed by the connection of the abnormal dies; Remove abnormal regions containing fewer than a first threshold number of dies to obtain at least one abnormal region.

9. The method according to claim 1, wherein, The step of excluding non-target regions from the at least one abnormal region based on the width of the at least one abnormal region includes: For each of the at least one abnormal region, a first proportion of rows in the abnormal region where the number of dies exceeds the second threshold, and a second proportion of columns where the number of dies exceeds the second threshold are calculated. The judgment index is determined based on the first ratio and the second ratio; In response to determining that the judgment index exceeds the third threshold, it is determined that the width of the abnormal area does not conform to the arc-shaped scratch, and the abnormal area is determined to be not the target area.

10. The method according to claim 1, wherein, The step of excluding non-target regions from the at least one abnormal region based on the width of the at least one abnormal region includes: For each of the at least one abnormal region, the total number of dies in the abnormal region and the number of shared edges of each die are counted, where a shared edge refers to the edge between two adjacent dies. The die with a number of shared edges exceeding the fourth threshold is designated as a feature die, and the ratio between the number of feature dies and the total number is determined. In response to determining that the ratio exceeds the fifth threshold, it is determined that the width of the abnormal region does not conform to the arc-shaped scratch, and the abnormal region is determined not to be the target region.

Images