Wafer bump defect detection method, device, equipment, medium and program product

Abstract

One or more embodiments of the present specification provide a wafer bump defect detection method, device, equipment, medium and program product. The defect detection method comprises: acquiring a 2D planar image and a 3D height image of a wafer surface to be measured, wherein the wafer surface to be measured comprises at least one bump; determining the area information of each bump in the image coordinate system according to the 2D planar image; extracting three-dimensional height data corresponding to each bump from the 3D height image according to the area information of each bump in the image coordinate system; and performing defect detection on each of the bumps based on the extracted three-dimensional height data to determine whether the bumps have geometric size defects and / or local topography defects.

Description

Technical Field

[0001] This specification relates to one or more embodiments in the field of semiconductor technology, and more particularly to a method, apparatus, device, medium, and program product for detecting defects in wafer bumps. Background Technology

[0002] In the field of high-performance computing, exemplified by AIGC (Artificial Intelligence Generated Content), the demand for high-efficiency, low-power chips is increasingly urgent. Against this backdrop, chiplet design concepts and advanced 2.5D / 3D packaging technologies (such as Chip-on-Wafer-Side (CoWoS) and High-Bandwidth Memory (HBM)) have emerged and become the mainstream technology direction for high-performance chips. These technologies significantly improve the system's computational density and overall performance through the stacking and bonding integration of multiple computing and storage units.

[0003] As a key microstructure for achieving vertical interconnection between chips in advanced packaging, the uniformity and integrity of microbumps directly determine the performance and reliability of the interconnection. Uneven height can easily lead to connection failure, while local morphological defects (such as voids and cracks) can severely degrade electrical characteristics and threaten the long-term stable operation of the product, becoming a major technical bottleneck restricting the yield and reliability of advanced packaging.

[0004] Currently, existing defect detection technologies for microbumps face significant limitations and cannot meet the urgent needs for high-precision, high-efficiency, and comprehensive defect detection in large-scale production. Summary of the Invention

[0005] In view of the above, one or more embodiments of this specification provide the following technical solutions:

[0006] According to a first aspect of one or more embodiments of this specification, a method for detecting defects in wafer bumps is provided, the method comprising:

[0007] Acquire a 2D planar image and a 3D height image of the surface of the wafer under test, wherein the surface of the wafer under test includes at least one bump;

[0008] Based on the 2D planar image, determine the region information of each bump in the image coordinate system;

[0009] Based on the region information of each bump in the image coordinate system, extract the three-dimensional height data corresponding to each bump from the 3D height image;

[0010] Based on the extracted three-dimensional height data, defect detection is performed on each of the bumps to determine whether the bumps have geometric defects and / or local shape defects.

[0011] Optionally, determining whether the bump has a geometric defect includes:

[0012] Calculate the average height value of the bump based on the three-dimensional height data of the bump;

[0013] Based on the average height value of at least one of the bumps, calculate the theoretical height value corresponding to each bump;

[0014] Based on the height deviation between the theoretical height value and the average height value corresponding to the bump, it is determined whether the bump has a geometric dimensional defect.

[0015] Optionally, calculating the theoretical height value corresponding to each bump based on the average height value of at least one of the bumps includes:

[0016] Global reference surface fitting is performed based on the average height value of at least one of the bumps;

[0017] For each bump, the theoretical height value corresponding to the bump is calculated based on the global reference surface.

[0018] Optionally, determining whether the bump has a geometrical defect based on the height deviation between the theoretical height value and the average height value corresponding to the bump includes:

[0019] When the average height value is greater than the theoretical height value, and the difference between the average height value and the theoretical height value is greater than the first height difference threshold, it is determined that the bump has a geometric dimension defect of excessive height.

[0020] If the average height value is less than the theoretical height value, and the absolute value of the difference between the average height value and the theoretical height value is greater than the second height difference threshold, it is determined that the bump has a geometric dimension defect of being too low in height.

[0021] Optionally, determining the region information of each bump in the image coordinate system includes:

[0022] The two-dimensional center coordinates of each of the bumps are determined based on the theoretical center coordinates of each bump.

[0023] The 2D planar image is segmented using grayscale histograms to obtain multiple foreground regions containing convexities;

[0024] The foreground region is matched with the two-dimensional center coordinates of each bump to determine the foreground region corresponding to each bump;

[0025] The region information of the foreground region is determined as the region information of the bump in the image coordinate system.

[0026] Optionally, determining the region information of each bump in the image coordinate system includes:

[0027] The two-dimensional center coordinates of each of the bumps are determined based on the theoretical center coordinates of each bump.

[0028] For each of the bumps, the region of interest corresponding to the bump is determined in the 2D planar image based on the two-dimensional center coordinates of the bump;

[0029] The region of interest is segmented using a grayscale histogram to obtain a foreground region containing the convex blocks;

[0030] The region information of the foreground region is determined as the region information of the bump in the image coordinate system.

[0031] Optionally, determining whether the bump has local topographic defects includes:

[0032] First, determine whether the bump has any geometric defects;

[0033] For bumps that are not determined to have geometric defects, the determination is then made to determine whether the bumps have local morphological defects.

[0034] Optionally, determining whether the bump has local topographic defects includes:

[0035] Based on the region information of the bump in the image coordinate system, the effective detection region of the bump is determined in the 3D height image;

[0036] Construct a corresponding local theoretical height map for the effective detection area;

[0037] The effective detection area and the local theoretical height map are subjected to pixel-level difference operation to obtain the difference image corresponding to the bump;

[0038] The difference image is segmented to determine whether the bump has local shape defects.

[0039] Optionally, determining the effective detection area of ​​the bump in the 3D height image based on the region information of the bump in the image coordinate system includes:

[0040] In the 3D height image, the region pointed to by the region information is fitted with a shape regularization to obtain the corresponding regular region;

[0041] The rule region is shrunk based on a preset shrinkage rule to obtain an effective detection region.

[0042] Optionally, constructing a corresponding local theoretical height map for the effective detection region includes:

[0043] For each pixel in the effective detection region, the theoretical height value of the pixel is obtained by averaging the height values ​​of multiple other pixels in its neighborhood, thus generating the theoretical height values ​​of all pixels in the effective detection region to construct a local theoretical height map corresponding to the effective detection region; or,

[0044] A local height plane is fitted to each pixel within the effective detection area. Based on the local height plane, the theoretical height value of each pixel within the effective detection area is calculated to construct a local theoretical height map corresponding to the effective detection area.

[0045] Optionally, the step of performing image segmentation on the difference image to determine whether the bump has local shape defects includes:

[0046] For each pixel in the difference image, determine whether the pixel has a protrusion defect or a depression defect based on the height value of the pixel;

[0047] When a pixel in the differential image with a raised defect satisfies the first condition, it is determined that the bump has a raised defect.

[0048] When a pixel in the differential image with a concave defect satisfies the second condition, it is determined that the bump has a concave defect.

[0049] Optionally, determining whether a pixel has a protrusion defect or a depression defect based on the pixel's height value includes:

[0050] If the height value of the pixel is positive, obtain the protrusion threshold corresponding to the pixel; if the height value is greater than the protrusion threshold, determine that the pixel has a protrusion defect.

[0051] If the height value of the pixel is negative, obtain the indentation threshold corresponding to the pixel, and if the absolute value of the height value is greater than the indentation threshold, determine that the pixel has an indentation defect.

[0052] Optionally, obtaining the protrusion threshold or concavity threshold corresponding to the pixel includes:

[0053] Calculate the center distance between the pixel and the center point of the bump;

[0054] Based on the center distance, the protrusion threshold or concavity threshold of the pixel is determined through a preset mapping relationship; wherein the mapping relationship is configured such that: the larger the center distance, the larger the corresponding protrusion threshold or concavity threshold; the smaller the center distance, the smaller the corresponding protrusion threshold or concavity threshold.

[0055] Optionally, the first condition includes one or more of the following: the number of pixels with protrusion defects exceeds a first quantity threshold, the proportion of pixels with protrusion defects exceeds a first proportion threshold, the positional features of pixels with protrusion defects satisfy a first positional constraint condition, and the morphological features of pixels with protrusion defects satisfy a first morphological constraint condition.

[0056] The second condition includes one or more of the following: the number of pixels with dents exceeds a second quantity threshold; the proportion of pixels with dents exceeds a second proportion threshold; the positional features of pixels with dents satisfy a second positional constraint condition; and the morphological features of pixels with dents satisfy a second morphological constraint condition.

[0057] According to a second aspect of one or more embodiments of this specification, a defect detection apparatus for wafer bumps is provided, the apparatus comprising:

[0058] The image acquisition unit acquires a 2D planar image and a 3D height image of the surface of the wafer under test, wherein the surface of the wafer under test includes at least one bump;

[0059] The region determination unit determines the region information of each bump in the image coordinate system based on the 2D planar image;

[0060] The height extraction unit extracts the three-dimensional height data corresponding to each bump from the 3D height image based on the region information of each bump in the image coordinate system.

[0061] The defect detection unit performs defect detection on each of the bumps based on the extracted three-dimensional height data to determine whether the bumps have geometric defects and / or local shape defects.

[0062] According to a third aspect of one or more embodiments of this specification, an electronic device is provided, comprising: a processor; a memory for storing processor-executable instructions; wherein the processor executes the executable instructions to implement the steps of the aforementioned wafer bump defect detection method.

[0063] According to a fourth aspect of one or more embodiments of this specification, a computer-readable storage medium is provided that stores computer instructions thereon, which, when executed by a processor, implement the steps of the aforementioned wafer bump defect detection method.

[0064] According to a fifth aspect of one or more embodiments of this specification, a computer program product is provided, comprising a computer program / instructions that, when executed by a processor, implement the steps of the aforementioned wafer bump defect detection method.

[0065] As can be seen from the above embodiments, by combining and analyzing the 2D planar image and 3D height image of the wafer surface under test, the region information of each bump in the image coordinate system can be determined based on the 2D planar image of the wafer surface under test, thereby achieving precise positioning of each bump through the 2D planar image. Then, based on the region information, the three-dimensional height data corresponding to each bump can be extracted from the 3D height image of the wafer surface under test, and defect detection is performed on each bump based on the extracted three-dimensional height data to determine whether the bump has geometric size defects and / or local morphological defects, thereby achieving high-precision and high-efficiency detection of geometric size defects and local morphological defects, overcoming the limitations of traditional single detection methods. Attached Figure Description

[0066] Figure 1 This is a flowchart of a defect detection method for wafer bumps provided in an exemplary embodiment.

[0067] Figure 2 This is a flowchart of a method for determining bump region information provided in an exemplary embodiment.

[0068] Figure 3 This is a schematic diagram of grayscale histogram segmentation provided in an exemplary embodiment.

[0069] Figure 4 This is a flowchart of another method for determining bump region information provided in an exemplary embodiment.

[0070] Figure 5 This is a flowchart of an exemplary embodiment of a method for determining geometric defects of bumps.

[0071] Figure 6 This is a flowchart of a method for determining bump morphology defects, provided in an exemplary embodiment.

[0072] Figure 7 This is a schematic diagram illustrating the determination of local morphological defects of a bump, provided in an exemplary embodiment.

[0073] Figure 8 This is a schematic diagram of the structure of a device provided in an exemplary embodiment.

[0074] Figure 9 This is a block diagram of a defect detection device for wafer bumps provided in an exemplary embodiment. Detailed Implementation

[0075] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0076] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented, for example, in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “corresponding to,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0077] In the field of high-performance computing, exemplified by AIGC (Artificial Intelligence Generated Content), the demand for high-efficiency, low-power chips is increasingly urgent. Against this backdrop, chiplet design concepts and advanced 2.5D / 3D packaging technologies (such as Chip-on-Wafer-Side (CoWoS) and High-Bandwidth Memory (HBM)) have emerged and become the mainstream technology direction for high-performance chips. These technologies significantly improve the system's computational density and overall performance through the stacking and bonding integration of multiple computing and storage units.

[0078] As a key microstructure for achieving vertical interconnection between chips in advanced packaging, the uniformity and integrity of microbumps directly determine the performance and reliability of the interconnection. Uneven height can easily lead to connection failure, while local morphological defects (such as voids and cracks) can severely degrade electrical characteristics and threaten the long-term stable operation of the product, becoming a major technical bottleneck restricting the yield and reliability of advanced packaging.

[0079] Traditional non-contact wafer surface defect inspection primarily relies on two-dimensional vision imaging technology. This involves acquiring wafer surface images using line scan or area scan cameras under bright or dark lighting conditions, and then identifying defects through high-precision image processing and analysis algorithms. This approach is sensitive to color and texture changes and is suitable for detecting planar defects such as scratches and foreign object contamination. However, two-dimensional vision inspection cannot provide three-dimensional morphological information such as height and depth. For microbumps, key quality parameters include not only surface cleanliness but also three-dimensional deformation features such as bump height, coplanarity, and depressions or protrusions. These defects appear as only slight grayscale or texture changes in two-dimensional images, easily masked by noise. This leads to a significant increase in the false negative and false positive rates of traditional two-dimensional vision inspection methods when dealing with three-dimensional morphological anomalies such as pits and protrusions in microbumps, failing to meet the comprehensive quality control requirements of packaging processes for microbumps.

[0080] In addition, to compensate for the inherent limitations of two-dimensional visual inspection in acquiring topographic information, the industry has introduced three-dimensional scanning technologies, such as white light interferometry, laser confocal microscopy, and triangulation, for the three-dimensional topographic acquisition of microbumps on wafer surfaces. While these technologies can achieve topographic reconstruction at the micro-nano scale and accurately measure key topographic parameters such as bump height, the height of the metallization step beneath the bump, and wafer warpage, each technology approach has significant functional shortcomings when it comes to defect detection of microbump arrays. Specifically:

[0081] White light interferometers, based on the principle of phase-shifting interferometry, can acquire the three-dimensional topography of a full-field micro-bump array within seconds, offering significant efficiency advantages in obtaining geometric parameters such as height measurement and coplanarity analysis. However, their lateral resolution is limited by the optical diffraction limit, and they have low sensitivity to signal reflection in steep sidewall regions, making it easy to overlook local defects such as tiny pits, fine cracks, and surface-attached foreign matter.

[0082] Confocal microscopy, with its unique optical slicing capabilities and high lateral resolution, can effectively suppress stray light from the focal surface and clearly reveal minute scratches and particulate contamination on the surface of micro-bumps, demonstrating a significant advantage in characterizing local morphological defects. However, its point- or line-scanning imaging mechanism limits detection efficiency to the scanning range and density. When dealing with large-area, high-density micro-bump arrays, a single detection takes too long, making it difficult to meet the high-throughput detection requirements of mass production lines. While this technology can detect defects, its efficiency is insufficient.

[0083] Multidimensional triangulation measurement techniques, such as structured light 3D measurement and multi-view photometric stereo methods, calculate the 3D morphology of an object's surface based on triangulation principles or photometric information by projecting specific grating patterns or multi-angle illumination. These techniques are highly adaptable to high-contrast surfaces and low-reflectivity areas, and can reconstruct the morphology of steep sidewalls through multi-view data fusion, offering unique advantages in acquiring the overall contour of micro-bumps. However, their technical bottlenecks lie in the fact that the ability to distinguish local morphological defects is limited by the optical projection resolution and camera sampling rate. For sub-micron level pits or cracks on the surface of micro-bumps, triangulation measurement techniques often struggle to effectively identify them; simultaneously, photometric calculations based on grayscale information are easily affected by the uniformity of the surface material, leading to misjudgments of defects. Furthermore, multi-view data acquisition and fusion involve massive data processing, which to some extent restricts detection efficiency. Improving detection efficiency by increasing computing power would significantly increase detection costs.

[0084] Due to the aforementioned technological limitations, current packaging production lines generally adopt a tandem inspection mode—that is, first, white light interferometers or triangulation measurement technology are used to complete the topographic acquisition and geometric parameter measurement, and then confocal microscopes or high-resolution vision equipment are used for defect re-inspection. This "secondary inspection, multiple devices" approach not only increases equipment procurement and maintenance costs, but also leads to a more complex inspection process and a decrease in production efficiency. Even attempts to achieve full-dimensional inspection through a single technical path often fall into a dilemma between data processing speed and inspection accuracy: pursuing high-precision three-dimensional topographic data processing requires overcoming the efficiency bottleneck caused by massive data computation, while pursuing efficiency inevitably sacrifices the ability to distinguish local topographic defects.

[0085] As advanced packaging evolves towards higher density and smaller bump sizes, the scale of microbump arrays is growing exponentially, further amplifying the inherent shortcomings of existing technologies. The inherent trade-offs between measurement accuracy, detection efficiency, and defect identification capabilities inherent in a single technology path are no longer sufficient to support the future packaging production line's demand for a one-stop solution for high-precision, high-efficiency, and multi-dimensional defect detection.

[0086] Therefore, there is an urgent need for an integrated testing solution that can overcome the physical limitations of various technical paths and achieve complementary advantages. Through multimodal data fusion and intelligent algorithm collaboration, it can balance the accuracy and efficiency of macroscopic morphology measurement and local morphology defect identification, so as to break through the quality control bottleneck in the current advanced packaging process.

[0087] This specification provides a defect detection scheme for wafer bumps, which combines and analyzes 2D planar images and 3D height images of the wafer surface to achieve high-precision and high-efficiency detection of geometric defects and local morphological defects, overcoming the limitations of traditional single detection methods.

[0088] It should be noted that the "bump" described in this specification is typically implemented using microbumps in advanced packaging, which are usually micrometer-sized and used to achieve high-density vertical interconnects between chips. However, it is not limited to this structure and may also include other forms of bumps such as conventional solder bumps.

[0089] The "geometric defects" mentioned in this specification refer to the conformity assessment of key dimensional parameters such as height, coplanarity, diameter, and volume of microbumps. These defects alter the "body" and "position" of the bump as an interconnect structure.

[0090] The “local morphological defects” mentioned in this specification refer to abnormalities in the surface material of a single microbump, but the overall shape does not show any depression. The focus is on defects such as scratches, pits, cracks, residues, and oxidation discoloration. These defects affect the integrity and cleanliness of the microstructure of the microbump surface.

[0091] The wafer bump defect detection solution provided in this manual can be applied to electronic devices, such as PCs, mobile phones, tablets, laptops, and PDAs (Personal Digital Assistants).

[0092] The wafer bump defect detection solution provided in this manual can also be applied to either a client-server (CS) architecture or a browser-server (BS) architecture. Taking the CS architecture as an example, the client can upload a 2D planar image and a 3D height image of the wafer under test. The server then performs defect detection on the wafer bumps based on these 2D planar image and 3D height image and returns the detection results to the client for user viewing, etc. This manual does not impose any special limitations on this approach.

[0093] Figure 1 This is a flowchart of a defect detection method for wafer bumps provided in an exemplary embodiment.

[0094] Please refer to Figure 1 The defect detection method for wafer bumps may include the following steps:

[0095] Step 102: Obtain a 2D planar image and a 3D height image of the surface of the wafer to be tested, wherein the surface of the wafer to be tested includes at least one bump.

[0096] In some embodiments, the surface of the wafer under test is typically integrated with a large number of high-density bump structures, such as microbumps. When performing defect detection on the bumps on the surface of the wafer under test, 2D planar images and 3D height images of the surface of the wafer under test can be acquired first.

[0097] For example, a high-resolution area array camera (or machine vision camera) can be used to acquire a 2D planar image of the surface of the wafer under test, and a multi-dimensional structured light camera can be used to acquire a 3D height image of the surface of the wafer under test. See also the Chinese patent application of the same applicant, application number CN202311016168.7, filed on October 13, 2023, the technical solution disclosed in that application, particularly the hardware configuration and measurement data acquisition scheme based on the integration of 3D and 2D data acquisition in the wafer surface three-dimensional measurement device, which can be used as a reference for one embodiment of this case.

[0098] Furthermore, the 2D planar image can be acquired using stroboscopic illumination technology. Specifically, multiple illumination modes, such as coaxial illumination or low-angle lateral illumination, can be triggered sequentially within a very short exposure time, each operating at its specific frequency (which may include parameters such as illumination angle, phase, or encoding method). A high-resolution camera is used to precisely synchronize with each stroboscopic illumination, capturing instantaneous reflection images of the wafer surface under different illumination conditions to obtain a composite image. These composite images can then be processed, selecting the highest quality image as the 2D planar image, or they can be fused together to obtain the 2D planar image, etc.

[0099] In other embodiments, the 3D height image can also be obtained by high-precision optical measurement techniques such as white light interferometry vertical scanning, laser confocal scanning, or multi-source photometric stereoscopic imaging to obtain a 3D height image registered with the 2D planar image.

[0100] Of course, the methods for obtaining the 2D planar image and 3D height image described above are merely illustrative examples and are not intended to limit this specification.

[0101] Step 104: Determine the region information of each bump in the image coordinate system based on the 2D planar image.

[0102] In some embodiments, the region information of each bump in the image coordinate system can be determined based on the theoretical center coordinates of each bump and the 2D planar image. The region information can represent the position of the bump in the 2D planar image or the position of the bump in the 3D height image.

[0103] Step 106: Based on the region information of each bump in the image coordinate system, extract the three-dimensional height data corresponding to each bump from the 3D height image.

[0104] In some embodiments, three-dimensional height data corresponding to each bump can be extracted from the 3D height image based on the determined region information. The three-dimensional height data includes the height value of each pixel in the region where the corresponding bump is located.

[0105] Step 108: Based on the extracted three-dimensional height data, perform defect detection on each of the bumps to determine whether the bumps have geometric defects and / or local shape defects.

[0106] In some embodiments, the height deviation between the theoretical height value and the average height value of each bump can be calculated based on the three-dimensional height data of the bump, and then the presence of geometric defects in the bump at the macroscopic level can be determined based on the height deviation, so as to realize the detection of geometric defects in the bump.

[0107] In some embodiments, pixel-level defect detection can be performed on the effective monitoring area of ​​the bump based on the three-dimensional height data of the bump, thereby determining whether there are morphological defects in the bump at the microscopic level, so as to realize the detection of local morphological defects of the bump.

[0108] In some embodiments, the presence of geometrical defects in the bump can be determined first based on its three-dimensional height data. For bumps not identified as having geometrical defects, the presence of local topographic defects is then determined. For bumps identified as having geometrical defects, since they already have large macroscopic defects, further detection of local topographic defects can be discontinued. This optimizes system resource allocation, significantly improves detection efficiency, and avoids misjudgment interference. Of course, the above is only an illustrative example. In some embodiments, further detection of local topographic defects can be performed on bumps identified as having geometrical defects, which is helpful for defect root cause analysis, process optimization, and comprehensive quality assessment in high-reliability scenarios.

[0109] As can be seen from the above description, by combining and analyzing the 2D planar image and 3D height image of the wafer surface under test, the region information of each bump in the image coordinate system can be determined based on the 2D planar image of the wafer surface under test. This allows for precise positioning of each bump using the 2D planar image. Then, based on the region information, the three-dimensional height data corresponding to each bump can be extracted from the 3D height image of the wafer surface under test. Defect detection is then performed on each bump based on the extracted three-dimensional height data to determine whether the bump has geometric defects and / or local morphological defects. This achieves high-precision and high-efficiency detection of geometric defects and local morphological defects, overcoming the limitations of traditional single detection methods.

[0110] The implementation process of this specification will be described in detail below, focusing on three aspects: determining bump area information, detecting geometric defects, and detecting local morphological defects, in conjunction with specific implementation methods.

[0111] I. Determining the information of the bump area

[0112] In some embodiments, the region information of each bump on the surface of the wafer under test in the image coordinate system can be determined based on a 2D planar image of the surface of the wafer under test.

[0113] For example, please refer to Figure 2 The process of determining the area information may include the following steps:

[0114] Step 202: Determine the two-dimensional center coordinates of each bump based on the theoretical center coordinates of each bump.

[0115] In some embodiments, the theoretical center coordinates of the bumps on the surface of the wafer under test are typically the physical coordinates of the bump center in the corresponding design file of the wafer under test. For ease of description, the theoretical center coordinates of the bump can be denoted as... Next, the theoretical center coordinates can be mapped to the pixel positions in the 2D planar image according to the size of the 2D planar image to obtain the two-dimensional center coordinates of the bump.

[0116] In some embodiments, the actual physical dimensions of each pixel in the 2D planar image in the X and Y directions are denoted as... ,in, The value ranges from 0 to the length of the 2D planar image. The value ranges from 0 to the width of the 2D planar image. (This is achieved through...) The theoretical center coordinates of the convex block can be converted into pixel coordinates in the image coordinate system, resulting in the two-dimensional center coordinates of the convex block, denoted as . The conversion formula is as follows:

[0117]

[0118]

[0119] In the above formula Represents the pixel coordinates of the top-left corner of the current field of view image, and serves as the reference origin for coordinate transformation.

[0120] Step 204: Perform grayscale histogram segmentation on the 2D planar image to obtain multiple foreground regions containing convexities.

[0121] In some embodiments, the number of pixels corresponding to each gray level (typically 0 to 255) in the 2D planar image can be counted to form a gray level distribution histogram. Then, the threshold for pixel segmentation can be determined by using the gray level distribution histogram through algorithms such as Otsu's method or the triangle method. Subsequently, the pixels of the 2D planar image are binarized and segmented according to the threshold to separate the foreground region and the background region.

[0122] Generally, the grayscale representation of bumps and substrates in an image is the result of the combined effect of material reflectivity and local surface morphology. The specific grayscale representation depends on factors such as illumination method, surface roughness, and imaging system parameters. For example, under bright-field illumination, although bumps are made of highly reflective metal, the microscopic irregularities on their top and sidewalls caused by the manufacturing process result in significant diffuse reflection of incident light, reducing the effective light intensity returning to the area array camera. Therefore, they appear as low grayscale (dark / near black) in a two-dimensional grayscale image. Conversely, the wafer substrate surface is relatively smooth, primarily exhibiting specular reflection, with more light returning to the camera along the optical axis, thus presenting a higher grayscale (light gray) background. Thus, the segmented background area often corresponds to the substrate, and the segmented foreground area often corresponds to the bumps.

[0123] Therefore, by performing grayscale histogram segmentation on the 2D planar image, multiple foreground regions can be obtained, each of which includes a bump on the surface of the wafer to be tested.

[0124] Step 206: Match the foreground region with the two-dimensional center coordinates of each bump to determine the foreground region corresponding to each bump.

[0125] In some embodiments, multiple foreground regions containing bumps have been segmented by the aforementioned grayscale histogram segmentation. Then, the two-dimensional center coordinates of each bump determined above can be matched with the segmented foreground regions to determine the bumps contained in each foreground region, that is, to determine the correspondence between the bumps and the foreground regions.

[0126] Step 208: Determine the region information of the foreground region as the region information of the bump in the image coordinate system.

[0127] In some embodiments, after determining the correspondence between bumps and foreground regions, for each bump, the region information of the foreground region corresponding to the bump can be determined as the region information of the bump in the image coordinate system. The region information represents the position of the foreground region, and the bump region where the corresponding bump is located can be determined in the 3D height image using this region information.

[0128] Please refer to Figure 3 ,in, Figure 3 Image (a) shows a 2D planar image of the surface of a wafer under test. Figure 3 (b) shows the result of grayscale histogram segmentation of the 2D planar image. Figure 3 (b) macroscopically shows the segmented region with bumps, specifically the red highlighted area.

[0129] Specifically, Figure 3Image (a) shows a 2D planar grayscale image of a wafer surface under bright-field illumination. This image was acquired using an optical imaging system (such as an area array camera) and reflects the surface morphology of a local area of ​​the wafer. Due to the unevenness of the bump surface, significant diffuse reflection occurs, causing most of the incident light to be scattered, with only a small amount of light returning directly to the area array camera. Therefore, in the final acquired 2D image, the bump appears as a low grayscale (dark / near black). In contrast, the wafer substrate surface is relatively smooth and can effectively perform specular reflection, directly reflecting more incident light back to the camera, thus appearing as a higher grayscale value (light gray) in the 2D image. Figure 3 In (a), a large number of regularly arranged dark strip structures can be observed. These structures are bumps formed during the wafer manufacturing process. Their overall distribution is highly ordered and they are periodically arranged along a specific direction (such as the vertical direction) to form a grid-like array layout.

[0130] Furthermore, in Figure 3 In the image (a), region D is marked in the upper right corner and shown in a magnified view. The magnified view of region D shows that each bump can adopt a rectangular structure, with consistent size, clear boundaries, and strict alignment in both the horizontal and vertical directions. This regularly arranged bump structure is commonly used for solder ball connections in chip packaging (such as in the Flip-Chip process), therefore, the consistency of its height, position, and shape is a key quality control parameter.

[0131] Figure 3 (b) shows the... Figure 3 The result of grayscale histogram segmentation of the 2D planar image in (a) is shown. This method is based on the difference in grayscale values ​​between different regions of the image. Specifically, the algorithm analyzes the grayscale distribution histogram of the original image, identifies the low grayscale intervals corresponding to convexities (because convexities are usually less reflective than the substrate or have darker shadows), and determines the low grayscale intervals corresponding to the convexities as the segmentation threshold. After segmentation, all regions determined to contain convexities are highlighted in red, while the remaining background regions remain gray. From a macroscopic perspective, Figure 3 In (b), the red area shows the same... Figure 3 The regular stripe distribution in (a) where the dark stripes completely correspond to the segmentation algorithm verifies its effectiveness. Meanwhile, the continuity and integrity of the red area also indicate that this segmentation method can effectively preserve the overall spatial distribution characteristics of the convexities, making it suitable for subsequent 3D point cloud data matching.

[0132] For example, please refer to Figure 4 The process of determining the area information may include the following steps:

[0133] Step 402: Determine the two-dimensional center coordinates of each bump based on the theoretical center coordinates of each bump.

[0134] In some embodiments, the process of determining the two-dimensional center coordinates of each bump can refer to the foregoing. Figure 2 The specific description of step 202 in the embodiment will not be repeated here.

[0135] Step 404: For each bump, determine the region of interest corresponding to the bump in the 2D planar image based on the two-dimensional center coordinates of the bump.

[0136] In some embodiments, for each bump, the region of interest corresponding to the bump can be determined in a 2D planar image based on the two-dimensional center coordinates of the bump and the design dimensions of the bump. The shape of the region of interest can be rectangular, circular, elliptical, etc.

[0137] Taking a rectangle as an example, the design dimensions of the bump can be combined with the actual physical dimensions of each pixel in the X and Y directions in the 2D planar image. This converts the design length to pixel dimensions. Assume the design dimension of the bump is its length. ,width It can be converted to pixel length. and pixel width ,in, , Next, using the two-dimensional center of the convex block... Centered on, with pixel length and pixel width Given the side length, the corresponding rectangular region is extracted from the 2D planar image as the region of interest for the convex block.

[0138] Step 406: Perform grayscale histogram segmentation on the region of interest to obtain a foreground region containing the bumps.

[0139] In some embodiments, for each bump's region of interest, grayscale histogram segmentation can be performed on the region of interest to obtain the foreground region of the bump. Thus, by performing multiple grayscale histogram segments, the foreground region of each bump can be obtained.

[0140] For details on the grayscale histogram segmentation method, please refer to the aforementioned methods. Figure 2 Step 204 in the illustrated embodiment will not be repeated here.

[0141] Step 408: Determine the region information of the foreground region as the region information of the bump in the image coordinate system.

[0142] In some embodiments, for each bump, the region information of the foreground region of the bump can be determined as the region information of the bump in the image coordinate system. The region information represents the position of the foreground region, and the bump region where the corresponding bump is located can be determined in the 3D height image using this region information.

[0143] II. Detection of Geometric Dimensional Defects

[0144] Figure 5 This is a flowchart of an exemplary embodiment of a method for determining geometric defects of bumps.

[0145] Please refer to Figure 5 The method for determining geometric defects of bumps may include the following steps:

[0146] Step 502: Calculate the average height value of the bump based on the three-dimensional height data of the bump.

[0147] In some embodiments, the three-dimensional height data of the bump includes the height values ​​of each pixel in the bump region where the bump is located, and the average height value of the corresponding bump can be calculated based on the height values ​​of these pixels.

[0148] For example, for each bump, the average height value of all pixels in the bump region can be calculated as the average height value of the bump.

[0149] For example, for each bump, a portion of pixels can be sampled from the bump region, and the average height of these sampled pixels can be calculated as the average height of the bump. This specification does not limit this method. The pixel sampling method can be random sampling or interval sampling, etc.

[0150] Therefore, for each bump on the surface of the wafer under test, its average height value can be calculated, and this average height value can be denoted as... This allows us to obtain the three-dimensional pixel coordinates of each bump. ).

[0151] Step 504: Calculate the theoretical height value corresponding to each bump based on the average height value of at least one of the bumps.

[0152] In some embodiments, after calculating the average height value of each bump on the surface of the wafer under test, the theoretical height value corresponding to each bump can be calculated based on the average height value of at least one bump. Specifically, the theoretical height value corresponding to each bump can be calculated based on the average height value of all bumps on the surface of the wafer under test, or it can be calculated based on the average height value of some bumps on the surface of the wafer under test. For example, the bumps on the surface of the wafer under test can be sampled to determine the bumps used to calculate the theoretical height value. This specification does not impose any special limitations on this.

[0153] In some embodiments, when calculating the theoretical height value corresponding to each bump, a global reference surface can be fitted based on the average height value of at least one bump. Then, for each bump, the theoretical height value corresponding to the bump is calculated based on the fitted global reference surface.

[0154] Taking the global reference plane fitting of the average height value of all bumps on the surface of the wafer under test as an example, methods such as least squares method or singular value decomposition method can be used, based on the three-dimensional pixel coordinates of each bump ( A global reference surface is fitted, and its equation can be: The coefficients can be calculated through fitting. , and The value of this global reference plane can reflect the overall trend of height variations of all bumps on the surface of the wafer under test.

[0155] The above global reference surface was obtained through fitting. Then, for each bump, the two-dimensional center coordinates of that bump can be used as a reference. The theoretical height of the bump can be calculated. That is, the two-dimensional center coordinates of the bump can be determined. Substitute into the global reference surface equation Its theoretical height value is obtained and denoted as ,Right now .

[0156] Of course, the above-described method of fitting the global reference surface is only an example. In other examples, polynomial surface fitting can be performed, or the theoretical height value of each bump can be calculated based on machine learning models (such as random forest regression models, neural network models, etc.). This specification does not limit this.

[0157] Step 506: Determine whether the bump has a geometric defect based on the height deviation between the theoretical height value and the average height value corresponding to the bump.

[0158] In some embodiments, the geometrical defects of the bump typically include excessively high geometrical defects and excessively low geometrical defects. When detecting bump geometrical defects, the average height value of the bump can be calculated. and theoretical height value The height deviation is calculated, and then it can be used to determine whether the corresponding bump has a geometric defect of excessive height or excessive height.

[0159] For example, when the average height value of the bump Greater than its theoretical height When this occurs, it indicates that the actual height of the corresponding bump is higher than the theoretical height, and the difference between its average height and the theoretical height can be calculated. ,Right now And determine the difference. and the first height difference threshold The size relationship between them. If the difference between the average height of the bump and the theoretical height is greater than the first height difference threshold, that is... If so, it is determined that the corresponding bump has a geometric defect of excessive height.

[0160] For example, when the average height value of the bump Less than its theoretical height value This indicates that the actual height of the corresponding bump is lower than the theoretical height. The difference between its average height and the theoretical height can also be calculated. ,Right now Then, determine the difference. Second height difference threshold The size relationship between them. If the absolute value of the difference between the average height value and the theoretical height value of the bump is greater than the second height difference threshold, that is... If so, it is determined that the corresponding bump has a geometric dimension defect of being too low in height.

[0161] Among them, the first height difference threshold Second height difference threshold They can be the same or different; there are no restrictions on this.

[0162] Therefore, for each bump on the surface of the wafer under test, the presence of geometrical defects can be determined based on the height deviation between the theoretical height value and the average height value of the bump, thus accurately determining whether there are macroscopic geometrical defects. Furthermore, this method can accurately identify whether macroscopic geometrical defects are excessively high or excessively low, achieving refined identification of macroscopic defects.

[0163] III. Detection of Local Morphological Defects

[0164] Figure 6 This is a flowchart of a method for determining local morphological defects of a bump, provided in an exemplary embodiment.

[0165] Please refer to Figure 6 and Figure 7 The illustrated method for determining local morphological defects in bumps may include the following steps:

[0166] Step 602: Based on the region information of the bump in the image coordinate system, determine the effective detection area of ​​the bump in the 3D height image.

[0167] In some embodiments, bumps are typically columnar. During bump fabrication, the edges of the columnar body may have irregular steep slopes, which are not considered defects. To avoid misjudging these steep slopes during local topographic defect detection, the corresponding bump area can be slightly recessed to exclude them, thus determining the effective detection area for local topographic defect detection. Figure 7 (a) shows the effective detection area of ​​a certain bump.

[0168] In some embodiments, an effective detection region for local shape defect detection can be determined for each bump in the 3D height image based on the region information of each bump in the image coordinate system. Considering that the foreground region including the bump obtained by the aforementioned gray-level histogram segmentation method may be an irregularly shaped region, that is, the region pointed to by the region information of the bump may be an irregularly shaped region, in order to facilitate the determination of the effective detection region, the region pointed to by the region information in the 3D height image can first be fitted with a shape regularization to fit the region pointed to by the region information into a region with regular shapes such as rectangle, circle, ellipse, etc., and then the fitted regular shape region can be shrunk based on a preset shrinkage rule to obtain the effective detection region of the corresponding bump.

[0169] In some embodiments, the preset shrinkage rule can be a shrinkage ratio, such as 10%. When shrinking, the major axis and minor axis of the fitted rule region can be shrunk inward by 10% each to obtain an effective detection region.

[0170] In some embodiments, the preset shrinkage rule can also be multiple different shrinkage ratios. For example, different shrinkage ratios can be set for the major axis and the minor axis, such as setting a shrinkage ratio of 10% for the major axis and a shrinkage ratio of 8% for the minor axis. When shrinking, the major axis of the fitted rule region can be shrunk inward by 10%, and the minor axis can be shrunk inward by 8%.

[0171] In some embodiments, the preset shrinkage rule may also be a shrinkage size and a number of iterations. When shrinking, the fitted rule region can be morphologically eroded according to the shrinkage size (e.g., the number of shrinkage pixels), and the erosion can be stopped when the number of erosions reaches the number of iterations, so as to obtain an effective detection region.

[0172] Of course, the above-described shrinkage method is merely an illustrative example and is not intended to limit this specification.

[0173] Step 604: Construct a corresponding local theoretical height map for the effective detection area.

[0174] In some embodiments, the local theoretical height map is an ideal, defect-free height map constructed for the effective detection area, which can be used for the detection of local topographic defects in bumps.

[0175] In some embodiments, when constructing the local theoretical height map for the effective detection region of a bump, a neighborhood can be defined for each pixel within the effective detection region. Then, the theoretical height value of that pixel is determined based on the height values ​​of multiple other pixels within its neighborhood, thereby determining the theoretical height value of each pixel within the effective detection region to construct the corresponding local theoretical height map. Figure 7 (b) shows the local theoretical height map of the effective detection area of ​​the aforementioned bump.

[0176] The neighborhood of a pixel can be square, for example, a square neighborhood with a preset length and width centered on the corresponding pixel. The neighborhood can also be circular, for example, a circular neighborhood with a preset radius centered on the corresponding pixel.

[0177] Each pixel's neighborhood can include multiple other pixels. When calculating the theoretical height of a pixel, the average height of its neighbors can be used, or the average height of all pixels in its neighborhood, including the pixel itself, can be used. Alternatively, other pixels in its neighborhood can be sampled, and the average height of the sampled pixels can be calculated as the theoretical height. Of course, weighted averaging or other methods can also be used to calculate the theoretical height, and this specification does not impose any restrictions on this approach.

[0178] For example, the neighborhood of each pixel includes the other 8 pixels around it, specifically the 8 pixels above, below, left, right, upper left, lower left, upper right, and lower right. When calculating the theoretical height value of the pixel, the average height value of the 8 pixels around it can be calculated as the theoretical height value.

[0179] Therefore, in the process of detecting local morphological defects, for each pixel, the theoretical height value of the pixel is calculated using the height values ​​of multiple pixels in its neighborhood. This can effectively suppress height abrupt changes caused by isolated noise points or local morphological anomalies, achieve smooth reconstruction of bump morphology and correction of reference surface, and enhance the accuracy, robustness and tolerance to process fluctuations of subsequent differential detection. This provides a key data foundation for high-precision and high-reliability local morphological defect judgment.

[0180] In some embodiments, when constructing the local theoretical height map for the effective detection area of ​​the bump, a local height plane may be fitted for each pixel in the effective detection area, and then the theoretical height value of each pixel in the effective detection area may be calculated based on the local height plane to construct the local theoretical height map corresponding to the effective detection area.

[0181] In fitting a local height plane for each pixel within the effective detection area, a corresponding fitting window can be defined for each pixel. This fitting window includes the pixel itself and all surrounding pixels. Similar to the aforementioned pixel neighborhood method, the fitting window is determined centered on the corresponding pixel, based on a preset size and window shape. Then, methods such as least squares or singular value decomposition can be used to fit a local height plane for the pixel based on the height values ​​of all pixels within the fitting window. The theoretical height value of the pixel is then calculated based on the fitted local height plane. The specific process can be referenced in the previous sections on fitting the global reference surface and calculating the theoretical height value of the bump; these details will not be elaborated upon here.

[0182] Step 606: Perform pixel-level difference operation on the effective detection area and the local theoretical height map to obtain the difference image corresponding to the bump.

[0183] In some embodiments, the pixel-level difference operation includes: for each pixel within the effective detection area, obtaining the actual height value of that pixel within the effective detection area and the corresponding theoretical height value in the local theoretical height map, and calculating the difference between the actual height value and the theoretical height value as the difference result for that pixel. By traversing all pixels within the effective detection area, the difference results for all pixels can be obtained, thereby obtaining a difference image that can characterize the deviation between the surface topography of the corresponding bump and the local theoretical topography, such as... Figure 7 The difference image shown in (c) is shown in the figure.

[0184] Step 608: Perform image segmentation on the differential image to determine whether the bump has local morphological defects.

[0185] In some embodiments, the height value of each pixel in the differential image is the difference between its actual height value and its theoretical height value. Therefore, the differential image can be directly segmented based on the height value of each pixel in the differential image to determine whether the corresponding pixel has a protrusion defect or a depression defect.

[0186] In the difference image, if the height value of a pixel is positive, it indicates that the actual height of the pixel is greater than the theoretical height. This pixel may have a protrusion defect but not a depression defect. A protrusion threshold can be obtained for this pixel, and if the height value of the pixel in the difference image is greater than the protrusion threshold, the pixel is determined to have a protrusion defect. Conversely, if the height value of a pixel in the difference image is negative, it indicates that the actual height of the pixel is less than the theoretical height. This pixel may have a depression defect but not a protrusion defect. A depression threshold can be obtained for this pixel, and if the absolute value of the height value of the pixel in the difference image is greater than the depression threshold, the pixel is determined to have a depression defect.

[0187] In some embodiments, the indentation threshold of different pixels in the differential image may be the same.

[0188] In some embodiments, considering that the edges of bumps will naturally form a certain slope due to processes such as electroplating and etching during the bump manufacturing process, if the same concavity threshold is used, normal deformation at the edge of the effective detection area may be misjudged as a concavity defect. Therefore, different concavity thresholds are set for pixels at different locations to avoid such misjudgments as much as possible.

[0189] For example, when obtaining the concavity threshold corresponding to a pixel, the center distance between the pixel and the center point of the bump can be calculated. Then, based on the center distance, the concavity threshold of the pixel can be determined through a preset mapping relationship. The mapping relationship can be configured such that: the larger the center distance, the larger the corresponding concavity threshold; the smaller the center distance, the smaller the corresponding concavity threshold.

[0190] In one example, a base concavity value can be preset. After calculating the center distance between the pixel and the center point of the bump, the concavity threshold of the pixel can be calculated based on the center distance and the base concavity value. For example, the ratio between the center distance and the preset distance can be calculated, and this ratio can be determined as the weight for calculating the concavity threshold. Then, the product of this weight and the base concavity value is calculated to obtain the concavity threshold of the pixel. The preset distance can be the maximum distance between the center point of the bump and the edge, etc., and is not limited here.

[0191] In another example, multiple center distance intervals and their corresponding concavity thresholds can be preset. After calculating the center distance between the pixel and the center point of the bump, the concavity threshold corresponding to the distance interval to which the center distance belongs can be obtained as the concavity threshold corresponding to the pixel.

[0192] Of course, the above method for calculating the indentation threshold is merely an illustrative example and should not be construed as a limitation of this specification.

[0193] In some embodiments, the bulge threshold of different pixels in the differential image may be the same.

[0194] In some embodiments, considering that the center of the bump is the core functional area of ​​the electrical connection, any minor morphological anomaly near the center may have a direct impact on the reliability of the device. Therefore, a more stringent bump threshold can be set to achieve highly sensitive defect detection.

[0195] For example, when obtaining the protrusion threshold corresponding to a pixel, the center distance between the pixel and the center point of the protrusion can be calculated. Then, based on the center distance, the protrusion threshold of the pixel can be determined through a preset mapping relationship. The mapping relationship can be configured such that: the larger the center distance, the larger the corresponding protrusion threshold; the smaller the center distance, the smaller the corresponding protrusion threshold.

[0196] Similar to the aforementioned concavity threshold, a base protrusion value can be preset. After calculating the center distance between the pixel and the center point of the protrusion, the protrusion threshold of the pixel can be calculated based on the center distance and the base protrusion value. Alternatively, multiple center distance intervals and their corresponding protrusion thresholds can be preset. After calculating the center distance between the pixel and the center point of the protrusion, the protrusion threshold corresponding to the distance interval to which the center distance belongs can be obtained as the protrusion threshold for that pixel, etc., which will not be elaborated further here.

[0197] In some embodiments, after determining whether each pixel in the differential image has a protrusion defect, the protrusion can be determined to have a protrusion defect when the pixel with the protrusion defect meets the first condition.

[0198] The first condition includes one or more of the following: the number of pixels with protrusion defects exceeds a first quantity threshold; the proportion of pixels with protrusion defects exceeds a first proportion threshold; the positional features of pixels with protrusion defects satisfy a first positional constraint condition; and the morphological features of pixels with protrusion defects satisfy a first morphological constraint condition.

[0199] In one example, after determining whether each pixel in the differential image has a protrusion defect, the number of pixels with protrusion defects can be counted, and it can be determined whether the number exceeds a preset first number threshold. If it exceeds the first number threshold, it means that there are a large number of pixels with protrusion defects in the bump, and thus it can be determined that the bump has a microscopic protrusion defect.

[0200] In another example, after determining whether each pixel in the differential image has a protrusion defect, the proportion of pixels with protrusion defects can be calculated, that is, the ratio of the number of pixels with protrusion defects to the total number of pixels. It is then determined whether this proportion exceeds a preset first proportion threshold. If it exceeds the first proportion threshold, it means that there are a large number of pixels with protrusion defects in the bump, and thus it can be determined that the bump has a microscopic protrusion defect.

[0201] In another example, after determining whether each pixel in the differential image has a protrusion defect, the positional features of the pixels with protrusion defects can be identified, and it can be determined whether the positional features satisfy a first positional constraint condition. The positional feature can be the center distance between the pixel and the center point of the protrusion. The first positional constraint condition typically indicates that the pixel with the protrusion defect is relatively close to the center point of the protrusion. In other words, when the positional features of the pixel with the protrusion defect satisfy the first positional constraint condition, it means that the pixel with the protrusion defect appears in the central region of the protrusion, and thus it can be determined that the protrusion has a microscopic protrusion defect. Specifically, the first positional constraint condition can be a restriction on the sum of the center distances between the pixel with the protrusion defect and the center point of the protrusion, etc., which is not limited here.

[0202] In the next example, after determining whether each pixel in the difference image has a protrusion defect, the morphological features of the pixels with protrusion defects can be identified, and it can be determined whether the morphological features satisfy a first morphological constraint condition. The morphological features can represent the shape formed by the pixels with protrusion defects, specifically including continuous shapes, discrete shapes, and clustered shapes. The first morphological constraint condition typically indicates that the pixels with protrusion defects have a continuous or clustered shape. In other words, when the shape of the pixels with protrusion defects is a continuous or clustered shape, it can be determined that the protrusion has a microscopic protrusion defect.

[0203] Of course, the above-mentioned method for determining microscopic protrusion defects of bumps is only an example. One method can be selected to determine protrusion defects, or different methods can be combined to determine protrusion defects. For example, when the proportion of pixels with protrusion defects exceeds the first proportion threshold, and the positional characteristics of pixels with protrusion defects satisfy the first positional constraint condition, it is determined that the bump has microscopic protrusion defects. This specification does not limit this.

[0204] In some embodiments, after determining whether each pixel in the differential image has a concave defect, the bump can be determined to have a concave defect when the pixel with the concave defect meets the second condition.

[0205] The second condition includes one or more of the following: the number of pixels with dents exceeds a second quantity threshold; the proportion of pixels with dents exceeds a second proportion threshold; the positional features of pixels with dents satisfy a second positional constraint condition; and the morphological features of pixels with dents satisfy a second morphological constraint condition.

[0206] In one example, after determining whether each pixel in the differential image has a concave defect, the number of pixels with concave defects can be counted, and it can be determined whether the number exceeds a preset second number threshold. If it exceeds the second number threshold, it means that there are a large number of pixels with concave defects in the bump, and thus it can be determined that the bump has a microscopic concave defect.

[0207] In another example, after determining whether each pixel in the differential image has a concave defect, the proportion of pixels with concave defects can be calculated, that is, the ratio of the number of pixels with concave defects to the total number of pixels. It is then determined whether this proportion exceeds a preset second proportion threshold. If it exceeds the second proportion threshold, it means that there are a large number of pixels with concave defects in the bump, and thus it can be determined that the bump has a microscopic concave defect.

[0208] In another example, after determining whether each pixel in the differential image has a concave defect, the positional features of the pixels with concave defects can be identified, and it can be determined whether the positional features satisfy the second positional constraint condition. The determination process for the second constraint condition can refer to the aforementioned first positional constraint condition, and will not be repeated here.

[0209] In the next example, after determining whether each pixel in the differential image has a concave defect, the morphological features of the pixels with concave defects can be identified, and it can be determined whether the morphological features satisfy the second morphological constraint condition. The determination process of the second morphological constraint condition can refer to the aforementioned first morphological constraint condition, and will not be repeated here.

[0210] Similar to the aforementioned method for determining microscopic protrusion defects of bumps, one method can be selected alone to determine dent defects, or different methods can be combined to determine dent defects. This specification does not impose any restrictions on this.

[0211] Please refer to Figure 7 In (d) and (e), by segmenting the difference image showing the bump, it can be determined that there is a protrusion defect at the center of the bump. Figure 7 In the detection result image (e), the area circled by the red irregular shape is its protruding defect.

[0212] Therefore, for each bump on the surface of the wafer under test, a corresponding local theoretical height map can be constructed for the effective detection area of ​​the bump. Then, pixel-level difference operations are performed on the effective detection area and the local theoretical height map to obtain the difference image corresponding to the bump. By performing image segmentation on the difference image, it can be determined whether the bump has local topographic defects. Furthermore, through the above scheme, it is also possible to accurately identify whether the topographic defects at the microscopic level are protrusion defects or depression defects, achieving refined identification of local topographic defects.

[0213] In a typical application scenario of this specification, when performing defect detection on bumps on the surface of a wafer under test, a 2D planar image and a 3D height image of the surface of the wafer under test can be acquired. Then, the aforementioned scheme provided in this specification is used to first determine whether there are geometric defects on the bumps on the surface of the wafer under test. If it is determined that there are geometric defects, the threshold judgment in the specific process of determining geometric defects can be used to determine whether the bump has a geometric defect of excessive height or a geometric defect of excessive height. Then, the detection result of the presence of geometric defects and the specific geometric defect type (geometric defect of excessive height or geometric defect of excessive height) can be output.

[0214] If the bumps on the surface of the wafer under test are not determined to have geometric defects, the aforementioned scheme provided in this specification can be used to further determine whether the bumps have local topographic defects. If local topographic defects exist, the threshold judgment in the specific process of determining local topographic defects can be used to determine whether the local topographic defects are protrusion defects or depression defects. Then, the detection result can be output that there are no geometric defects, but there are local topographic defects and the specific type of local topographic defects (protrusion defects or depression defects).

[0215] Therefore, the wafer bump defect detection scheme provided in this specification, by combining 2D planar images and 3D height images of the wafer surface under test, can achieve high-precision judgment and classification of bump geometric defects and local morphological defects. Specifically, the 2D planar image can be used to accurately locate and segment the bumps, and the height characteristics of the bumps can be quantitatively analyzed based on the 3D height image. This overcomes the limitations of traditional two-dimensional vision inspection, such as high false negative rate for geometric defects and inability to provide quantitative height information. Furthermore, during the defect detection process, threshold judgment can accurately distinguish various defect types, such as excessively high geometric defects, excessively low geometric defects, micro-bump defects, and micro-depression defects. This achieves high-precision and high-efficiency comprehensive wafer bump defect detection, thereby significantly improving the yield and reliability of advanced packaging.

[0216] This invention leverages the complementary advantages of two-dimensional and three-dimensional technologies to simultaneously achieve precise identification of micro-bump surface defects and accurate measurement of morphological parameters, meeting the urgent need of advanced packaging for one-stop, multi-dimensional inspection.

[0217] Figure 8 This is a schematic structural diagram of a device provided in an exemplary embodiment. Please refer to... Figure 8 At the hardware level, the device includes a processor 802, an internal bus 804, a network interface 806, memory 808, and non-volatile memory 810, and may also include other hardware required for its functions. One or more embodiments of this specification can be implemented in software, for example, the processor 802 reads the corresponding computer program from the non-volatile memory 810 into memory 808 and then runs it. Of course, besides software implementation, one or more embodiments of this specification do not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. That is to say, the execution entity of the following processing flow is not limited to individual logic units, but can also be hardware or logic devices.

[0218] Please refer to Figure 9 The wafer bump defect detection device 900 can be applied to, for example... Figure 8 The device shown implements the technical solution described in this specification. The wafer bump defect detection device 900 may include:

[0219] Image acquisition unit 901 acquires a 2D planar image and a 3D height image of the surface of the wafer under test, wherein the surface of the wafer under test includes at least one bump;

[0220] The region determination unit 902 determines the region information of each bump in the image coordinate system based on the 2D planar image;

[0221] The height extraction unit 903 extracts the three-dimensional height data corresponding to each bump from the 3D height image based on the region information of each bump in the image coordinate system.

[0222] The defect detection unit 904 performs defect detection on each of the bumps based on the extracted three-dimensional height data to determine whether the bumps have geometric defects and / or local shape defects.

[0223] Optionally, determining whether the bump has a geometric defect includes:

[0224] Calculate the average height value of the bump based on the three-dimensional height data of the bump;

[0225] Based on the average height value of at least one of the bumps, calculate the theoretical height value corresponding to each bump;

[0226] Based on the height deviation between the theoretical height value and the average height value corresponding to the bump, it is determined whether the bump has a geometric dimensional defect.

[0227] Optionally, calculating the theoretical height value corresponding to each bump based on the average height value of at least one of the bumps includes:

[0228] Global reference surface fitting is performed based on the average height value of at least one of the bumps; and

[0229] For each bump, the theoretical height value corresponding to the bump is calculated based on the global reference surface.

[0230] Optionally, determining whether the bump has a geometrical defect based on the height deviation between the theoretical height value and the average height value corresponding to the bump includes:

[0231] When the average height value is greater than the theoretical height value, and the difference between the average height value and the theoretical height value is greater than a first height difference threshold, it is determined that the bump has a geometric dimensional defect of excessive height; and

[0232] If the average height value is less than the theoretical height value, and the absolute value of the difference between the average height value and the theoretical height value is greater than the second height difference threshold, it is determined that the bump has a geometric dimension defect of being too low in height.

[0233] Optionally, determining the region information of each bump in the image coordinate system includes:

[0234] The two-dimensional center coordinates of each of the bumps are determined based on the theoretical center coordinates of each bump.

[0235] The 2D planar image is segmented using grayscale histograms to obtain multiple foreground regions containing convexities;

[0236] Match the foreground region with the two-dimensional center coordinates of each bump to determine the foreground region corresponding to each bump; and

[0237] The region information of the foreground region is determined as the region information of the bump in the image coordinate system.

[0238] Optionally, determining the region information of each bump in the image coordinate system includes:

[0239] The two-dimensional center coordinates of each of the bumps are determined based on the theoretical center coordinates of each bump.

[0240] For each of the bumps, the region of interest corresponding to the bump is determined in the 2D planar image based on the two-dimensional center coordinates of the bump;

[0241] The region of interest is segmented using a grayscale histogram to obtain a foreground region containing the convex blocks; and

[0242] The region information of the foreground region is determined as the region information of the bump in the image coordinate system.

[0243] Optionally, determining whether the bump has local topographic defects includes:

[0244] First, determine whether the bump has any geometric defects;

[0245] For bumps that are not determined to have geometric defects, the determination is then made to determine whether the bumps have local morphological defects.

[0246] Optionally, determining whether the bump has local topographic defects includes:

[0247] Based on the region information of the bump in the image coordinate system, the effective detection region of the bump is determined in the 3D height image;

[0248] Construct a corresponding local theoretical height map for the effective detection area;

[0249] Perform pixel-level difference operations on the effective detection area and the local theoretical height map to obtain the difference image corresponding to the bump; and

[0250] The difference image is segmented to determine whether the bump has local shape defects.

[0251] Optionally, determining the effective detection area of ​​the bump in the 3D height image based on the region information of the bump in the image coordinate system includes:

[0252] In the 3D height image, the region pointed to by the region information is fitted with a shape regularization method to obtain the corresponding regular region; and

[0253] The rule region is shrunk based on a preset shrinkage rule to obtain an effective detection region.

[0254] Optionally, constructing a corresponding local theoretical height map for the effective detection region includes:

[0255] For each pixel in the effective detection region, the theoretical height value of the pixel is obtained by averaging the height values ​​of multiple other pixels in its neighborhood, thus generating the theoretical height values ​​of all pixels in the effective detection region to construct a local theoretical height map corresponding to the effective detection region; or,

[0256] A local height plane is fitted to each pixel within the effective detection area. Based on the local height plane, the theoretical height value of each pixel within the effective detection area is calculated to construct a local theoretical height map corresponding to the effective detection area.

[0257] Optionally, the step of performing image segmentation on the difference image to determine whether the bump has local shape defects includes:

[0258] For each pixel in the difference image, determine whether the pixel has a protrusion defect or a depression defect based on the height value of the pixel;

[0259] When a pixel in the differential image with a raised defect satisfies the first condition, it is determined that the bump has a raised defect.

[0260] When a pixel in the differential image with a concave defect satisfies the second condition, it is determined that the bump has a concave defect.

[0261] Optionally, determining whether a pixel has a protrusion defect or a depression defect based on the pixel's height value includes:

[0262] If the height value of the pixel is positive, obtain the protrusion threshold corresponding to the pixel; if the height value is greater than the protrusion threshold, determine that the pixel has a protrusion defect; and

[0263] If the height value of the pixel is negative, obtain the indentation threshold corresponding to the pixel, and if the absolute value of the height value is greater than the indentation threshold, determine that the pixel has an indentation defect.

[0264] Optionally, obtaining the protrusion threshold or concavity threshold corresponding to the pixel includes:

[0265] Calculate the center distance between the pixel and the center point of the bump; and

[0266] Based on the center distance, the protrusion threshold or concavity threshold of the pixel is determined through a preset mapping relationship; wherein the mapping relationship is configured such that: the larger the center distance, the larger the corresponding protrusion threshold or concavity threshold; the smaller the center distance, the smaller the corresponding protrusion threshold or concavity threshold.

[0267] Optionally, the first condition includes one or more of the following: the number of pixels with protrusion defects exceeds a first quantity threshold; the proportion of pixels with protrusion defects exceeds a first proportion threshold; the positional features of pixels with protrusion defects satisfy a first positional constraint condition; and the morphological features of pixels with protrusion defects satisfy a first morphological constraint condition.

[0268] The second condition includes one or more of the following: the number of pixels with dents exceeds a second quantity threshold; the proportion of pixels with dents exceeds a second proportion threshold; the positional features of pixels with dents satisfy a second positional constraint condition; and the morphological features of pixels with dents satisfy a second morphological constraint condition.

[0269] Based on the same concept as the methods described above, this specification also provides an electronic device, including: a processor; a memory for storing processor-executable instructions; wherein the processor performs the steps of the method as described in any of the above embodiments by executing the executable instructions.

[0270] Based on the same concept as the methods described above, this specification also provides a computer-readable storage medium having computer instructions stored thereon that, when executed by a processor, implement the steps of the methods as described in any of the above embodiments.

[0271] Based on the same concept as the methods described above, this specification also provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the methods as described in any of the above embodiments.

Claims

1. A method for detecting defects in wafer bumps, characterized in that, The method includes: Acquire a 2D planar image and a 3D height image of the surface of the wafer under test, wherein the surface of the wafer under test includes at least one bump; Based on the 2D planar image, determine the region information of each bump in the image coordinate system; Based on the region information of each bump in the image coordinate system, extract the three-dimensional height data corresponding to each bump from the 3D height image; Based on the extracted three-dimensional height data, defect detection is performed on each of the bumps to determine whether the bumps have geometric size defects and local shape defects. The determination of whether the bump has local morphological defects includes: Based on the region information of the bump in the image coordinate system, the effective detection region of the bump is determined in the 3D height image; Construct a corresponding local theoretical height map for the effective detection area; Perform pixel-level difference operations on the effective detection area and the local theoretical height map to obtain the difference image corresponding to the bump; and The difference image is segmented to determine whether the bump has local shape defects.

2. The method according to claim 1, characterized in that, The determination of whether the bump has geometric defects includes: Based on the three-dimensional height data of the bump, calculate the average height value of the bump; and Based on the average height value of at least one of the bumps, calculate the theoretical height value corresponding to each bump; Based on the height deviation between the theoretical height value and the average height value corresponding to the bump, it is determined whether the bump has a geometric dimensional defect.

3. The method according to claim 2, characterized in that, The step of calculating the theoretical height value corresponding to each bump based on the average height value of at least one of the bumps includes: Global reference surface fitting is performed based on the average height value of at least one of the bumps; and For each bump, the theoretical height value corresponding to the bump is calculated based on the global reference surface.

4. The method according to claim 2, characterized in that, The step of determining whether the bump has a geometric dimensional defect based on the height deviation between the theoretical height value and the average height value corresponding to the bump includes: When the average height value is greater than the theoretical height value, and the difference between the average height value and the theoretical height value is greater than a first height difference threshold, it is determined that the bump has a geometric dimensional defect of excessive height; and When the average height value is less than the theoretical height value, and the absolute value of the difference between the average height value and the theoretical height value is greater than the second height difference threshold, it is determined that the bump has a geometric dimension defect of being too low in height.

5. The method according to claim 1, characterized in that, Determining the region information of each bump in the image coordinate system includes: The two-dimensional center coordinates of each of the bumps are determined based on the theoretical center coordinates of each bump. The 2D planar image is segmented using grayscale histograms to obtain multiple foreground regions containing convexities; Match the foreground region with the two-dimensional center coordinates of each bump to determine the foreground region corresponding to each bump; and The region information of the foreground region is determined as the region information of the bump in the image coordinate system.

6. The method according to claim 1, characterized in that, Determining the region information of each bump in the image coordinate system includes: The two-dimensional center coordinates of each of the bumps are determined based on the theoretical center coordinates of each bump. For each of the bumps, the region of interest corresponding to the bump is determined in the 2D planar image based on the two-dimensional center coordinates of the bump; The region of interest is segmented using a grayscale histogram to obtain a foreground region containing the convex blocks; and The region information of the foreground region is determined as the region information of the bump in the image coordinate system.

7. The method according to claim 1, characterized in that, The determination of whether the bump has local morphological defects includes: First, determine whether the bump has any geometric defects; For bumps that are not determined to have geometric defects, the determination is then made to determine whether the bumps have local morphological defects.

8. The method according to claim 1, characterized in that, The step of determining the effective detection area of ​​the bump in the 3D height image based on the region information of the bump in the image coordinate system includes: In the 3D height image, the region pointed to by the region information is fitted with a shape regularization method to obtain the corresponding regular region; and The rule region is shrunk based on a preset shrinkage rule to obtain an effective detection region.

9. The method according to claim 1, characterized in that, The construction of a corresponding local theoretical height map for the effective detection area includes: For each pixel in the effective detection region, the theoretical height value of the pixel is obtained by averaging the height values ​​of multiple other pixels in its neighborhood, thus generating the theoretical height values ​​of all pixels in the effective detection region to construct a local theoretical height map corresponding to the effective detection region; or, A local height plane is fitted to each pixel within the effective detection area. Based on the local height plane, the theoretical height value of each pixel within the effective detection area is calculated to construct a local theoretical height map corresponding to the effective detection area.

10. The method according to claim 1, characterized in that, The step of performing image segmentation on the difference image to determine whether the bump has local shape defects includes: For each pixel in the difference image, determine whether the pixel has a protrusion defect or a depression defect based on the height value of the pixel; When a pixel in the differential image with a raised defect satisfies a first condition, it is determined that the raised block has a raised defect; and When a pixel in the differential image with a concave defect satisfies the second condition, it is determined that the bump has a concave defect.

11. The method according to claim 10, characterized in that, The step of determining whether a pixel has a protrusion defect or a depression defect based on the height value of the pixel includes: If the height value of the pixel is positive, obtain the protrusion threshold corresponding to the pixel; if the height value is greater than the protrusion threshold, determine that the pixel has a protrusion defect; and If the height value of the pixel is negative, obtain the indentation threshold corresponding to the pixel, and if the absolute value of the height value is greater than the indentation threshold, determine that the pixel has an indentation defect.

12. The method according to claim 11, characterized in that, The step of obtaining the protrusion threshold or concavity threshold corresponding to the pixel includes: Calculate the center distance between the pixel and the center point of the bump; and Based on the center distance, the protrusion threshold or concavity threshold of the pixel is determined through a preset mapping relationship; wherein the mapping relationship is configured such that: the larger the center distance, the larger the corresponding protrusion threshold or concavity threshold; the smaller the center distance, the smaller the corresponding protrusion threshold or concavity threshold.

13. The method according to claim 10, characterized in that, The first condition includes one or more of the following: the number of pixels with protrusion defects exceeds a first quantity threshold, the proportion of pixels with protrusion defects exceeds a first proportion threshold, the positional features of pixels with protrusion defects satisfy a first positional constraint condition, and the morphological features of pixels with protrusion defects satisfy a first morphological constraint condition. as well as The second condition includes one or more of the following: the number of pixels with dents exceeds a second quantity threshold; the proportion of pixels with dents exceeds a second proportion threshold; the positional features of pixels with dents satisfy a second positional constraint condition; and the morphological features of pixels with dents satisfy a second morphological constraint condition.

14. A defect detection device for wafer bumps, characterized in that, The device includes: The image acquisition unit acquires a 2D planar image and a 3D height image of the surface of the wafer under test, wherein the surface of the wafer under test includes at least one bump; The region determination unit determines the region information of each bump in the image coordinate system based on the 2D planar image; The height extraction unit extracts the three-dimensional height data corresponding to each bump from the 3D height image based on the region information of each bump in the image coordinate system. The defect detection unit performs defect detection on each of the bumps based on the extracted three-dimensional height data to determine whether the bumps have geometric defects and local shape defects. The determination of whether the bump has local morphological defects includes: Based on the region information of the bump in the image coordinate system, the effective detection region of the bump is determined in the 3D height image; Construct a corresponding local theoretical height map for the effective detection area; Perform pixel-level difference operations on the effective detection area and the local theoretical height map to obtain the difference image corresponding to the bump; and The difference image is segmented to determine whether the bump has local shape defects.

15. An electronic device, characterized in that, include: processor; A memory for storing processor-executable instructions; wherein the processor implements the steps of the wafer bump defect detection method as described in any one of claims 1 to 13 by running the executable instructions.

16. A computer-readable storage medium, characterized in that, It stores computer instructions that, when executed by a processor, implement the steps of the defect detection method for wafer bumps as described in any one of claims 1 to 13.

17. A computer program product, characterized in that, Includes a computer program / instructions that, when executed by a processor, implement the steps of the defect detection method for wafer bumps as described in any one of claims 1 to 13.

Images