Method, device and equipment for detecting TSV via defects in 3D package and storage medium

Abstract

The application provides a TSV via defect detection method, device and equipment in 3D package, and a storage medium. The method comprises the following steps: positioning the opening center positions of a plurality of silicon vias on a top surface based on a first optical image of the top surface, and acquiring a second optical image corresponding to each silicon via on a bottom surface based on the opening center positions; determining the exit center positions of each silicon via on the bottom surface based on the second optical image, and calculating the axial offset vector of each silicon via in the vertical direction based on the opening center positions and the exit center positions; extracting the sidewall profile line of each silicon via along the depth direction based on a third optical image of a sidewall region, and constructing the target space trajectory of each silicon via based on the axial offset vector and the sidewall profile line; and judging whether each silicon via has a morphology anomaly deviating from a preset space path based on the target space trajectory, and obtaining a silicon via defect detection result. The application improves the recognition accuracy of TSV via defects.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a method, apparatus, device, and storage medium for detecting TSV via defects in 3D stacked packaging. Background Technology

[0002] In three-dimensional stacked packaging structures based on planar characterization, through-silicon vias (TSVs) are key structures for realizing vertical interconnects between chips, and their manufacturing quality directly affects the overall electrical performance and reliability of the package.

[0003] In existing technologies, a common method for detecting TSV (Through-Vacuum Valves) defects involves acquiring a two-dimensional cross-sectional image of the packaged sample, extracting edge contours from this image, and then comparing it with a preset geometric template to determine the presence of morphological anomalies. This method relies on static image information from a single viewpoint, making it difficult to accurately reflect the true morphological characteristics of the TSV in three-dimensional space. Especially when dealing with asymmetric defects such as tilting, offsetting, or local collapse, it is prone to missed detections or false positives. Therefore, the main technical problem with this method is that relying solely on a single-view two-dimensional image for TSV defect judgment cannot effectively capture the spatial morphological deviations of the TSV in a three-dimensional stacked structure, resulting in insufficient defect identification accuracy. Summary of the Invention

[0004] This invention provides a method, apparatus, device, and storage medium for detecting TSV via defects in 3D stacked packaging, in order to improve the accuracy of TSV via defect identification.

[0005] In a first aspect, the present invention provides a method for detecting TSV (Through-Video) defects in 3D stacked packaging, comprising: Based on a first optical image of the top surface of the three-dimensional stacked package sample, the center positions of the openings of multiple through-silicon vias on the top surface are located, and based on the opening center positions, a second optical image corresponding to each through-silicon via is obtained on the bottom surface of the three-dimensional stacked package sample. The exit center position of each through-silicon via on the bottom surface is determined based on the second optical image, and the axial offset vector of each through-silicon via in the vertical direction is calculated based on the opening center position and the exit center position. Based on the third optical image of the sidewall region of the three-dimensional stacked packaged sample, the sidewall contour line of each through-silicon via along the depth direction is extracted, and the target spatial trajectory of each through-silicon via is constructed based on the axis offset vector and the sidewall contour line. Based on the target spatial trajectory, determine whether each through-silicon via has any morphological anomalies that deviate from the preset spatial path, and obtain the through-silicon via defect detection results.

[0006] In a second aspect, the present invention also provides a TSV via defect detection device in 3D stacked packaging, applied to the TSV via defect detection method in 3D stacked packaging as described in the first aspect; the TSV via defect detection device in 3D stacked packaging includes: An optical image acquisition module is used to locate the opening center position of multiple through-silicon vias on the top surface of the three-dimensional stacked package sample based on a first optical image of the top surface of the sample, and to acquire a second optical image corresponding to each through-silicon via on the bottom surface of the sample based on the opening center position. The axis offset module is used to determine the emission center position of each through-silicon via on the bottom surface based on the second optical image, and to calculate the axis offset vector of each through-silicon via in the vertical direction based on the opening center position and the emission center position. The spatial trajectory construction module is used to extract the sidewall contour line of each through-silicon via along the depth direction based on the third optical image of the sidewall region of the three-dimensional stacked package sample, and to construct the target spatial trajectory of each through-silicon via based on the axis offset vector and the sidewall contour line. The defect detection module is used to determine whether each through-silicon via has an abnormal shape that deviates from the preset spatial path based on the target spatial trajectory, and to obtain the through-silicon via defect detection result.

[0007] Thirdly, the present invention also provides an electronic device, comprising: a memory for storing computer software programs; and a processor for reading and executing the computer software programs, thereby realizing the TSV via defect detection method in 3D stacked packaging as described above.

[0008] Fourthly, the present invention also provides a non-transitory computer-readable storage medium storing a computer software program, which, when executed by a processor, implements the TSV via defect detection method in 3D stacked packaging as described above.

[0009] Fifthly, the present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the TSV via defect detection method in 3D stacked packaging as described above.

[0010] The TSV via defect detection method provided in this invention for 3D stacked packaging acquires a first optical image of the top surface of the 3D stacked package sample to accurately locate the opening center positions of multiple TSVs on the top surface. Simultaneously, using this opening center position as a positioning reference, a second optical image of each TSV is acquired on the bottom surface, breaking the limitations of a single-view perspective and establishing a correspondence between the top and bottom of the TSV. Based on the opening center position and the second optical image at the bottom, the emission center position of each TSV on the bottom surface is determined. Combining the spatial coordinate difference between the top opening center and the bottom emission center, the axial offset vector of each TSV in the vertical direction is calculated. Therefore, it can intuitively reflect the spatial position deviation of the TSV in the vertical direction, effectively compensating for the inability of single two-dimensional images to capture spatial offset. Based on the axial offset vector and the sidewall contour lines of each TSV along the depth direction extracted from the third optical image of the sidewall region, the position offset information is combined with the morphological information in the depth direction to construct the target spatial trajectory of each TSV. This achieves a complete restoration of the 3D morphology of the TSV, solving the problem that single-view images cannot reflect the 3D spatial morphology, and clearly presenting the true morphology of asymmetric defects such as tilting and local collapse. The target spatial trajectory is compared with the preset spatial path to determine whether each through-silicon via (TSV) has morphological abnormalities, thus obtaining accurate TSV defect detection results. This effectively avoids the problems of missed detection and misjudgment caused by the inability to capture three-dimensional morphology, thereby improving the accuracy of TSV defect identification. Attached Figure Description

[0011] Figure 1 This is a flowchart illustrating the TSV via defect detection method in 3D stacked packaging provided by an embodiment of the present invention; Figure 2 This is a schematic diagram of the TSV via defect detection device in 3D stacked packaging provided in an embodiment of the present invention; Figure 3 An embodiment diagram of the electronic device provided in this invention; Figure 4 An embodiment diagram of a computer-readable storage medium provided in accordance with the present invention. Detailed Implementation

[0012] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0013] In the description of this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0014] In the description of this invention, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed herein.

[0015] See Figure 1 , Figure 1 This is a flowchart illustrating the TSV via defect detection method in 3D stacked packaging provided by the present invention. In this embodiment of the invention, the executing entity of the TSV via defect detection method in 3D stacked packaging is a defect detection device. Therefore, the TSV via defect detection method in 3D stacked packaging includes: Step 10: Based on the first optical image of the top surface of the three-dimensional stacked package sample, locate the opening center position of multiple through-silicon vias on the top surface, and based on the opening center position, acquire a second optical image corresponding to each through-silicon via on the bottom surface of the three-dimensional stacked package sample.

[0016] Optionally, the defect detection device acquires a first optical image of the top surface of the three-dimensional stacked package sample. The first optical image refers to an image obtained by the defect detection device through an optical imaging component, capturing the top surface of the three-dimensional stacked package sample and clearly showing the top surface morphology and through-silicon via (TSV) opening features. The first optical image undergoes preprocessing, including image noise reduction, contrast enhancement, and edge sharpening. Image noise reduction removes specks and noise from the first optical image caused by factors such as the shooting environment and errors in the imaging component itself, ensuring clear features of the TSV opening area in the image. Contrast enhancement increases the brightness difference between the TSV opening area and the background area of ​​the sample's top surface in the first optical image, making the outline of the opening area more prominent. Edge sharpening enhances the clarity of the TSV opening edges in the first optical image, facilitating subsequent precise positioning of the opening center.

[0017] Furthermore, the defect detection device performs contour extraction on the first optical image to locate the opening area of ​​each through-silicon via on the top surface. Contour extraction refers to identifying all areas in the first optical image that conform to the opening characteristics of through-silicon vias through an image recognition algorithm. The opening characteristics of through-silicon vias are specifically closed areas that are circular or have a preset shape, and their size is consistent with the preset opening size of the through-silicon via. At the same time, it excludes interfering areas such as scratches and stains on the top surface of the sample to ensure that the extracted opening areas are all real openings of through-silicon vias.

[0018] Furthermore, the defect detection device calculates the opening center position of each opening region. The opening center position refers to the geometric center of the through silicon via opening region. Optionally, the calculation process in this embodiment of the invention is as follows: the defect detection device first determines all pixels on the edge contour of each opening region, and then calculates the average horizontal coordinate and the average vertical coordinate of these pixels. The coordinate point corresponding to the average horizontal coordinate and the average vertical coordinate is the opening center position of the through silicon via on the top surface.

[0019] Furthermore, based on the center position of each opening, the defect detection device acquires a second optical image corresponding to each through-silicon via (TSV) on the bottom surface of the three-dimensional stacked package sample. The second optical image refers to an image obtained by the defect detection device through an optical imaging component, which clearly shows the bottom surface morphology and the emission area of ​​the TSV. Each second optical image corresponds one-to-one with a TSV. The correspondence is as follows: the area captured by the second optical image is the area on the bottom surface of the three-dimensional stacked package sample that corresponds to the center position of the TSV opening in the vertical direction, ensuring that the captured second optical image can accurately include the emission area of ​​the TSV on the bottom surface.

[0020] In one embodiment, five through-silicon vias (TSVs) are distributed on the top surface of the three-dimensional stacked packaged sample. An optical image of the top surface is captured using an optical imaging component, obtaining a first optical image with a resolution of 1024×768 pixels, clearly showing the opening areas of the five TSVs. The first optical image undergoes noise reduction processing to remove image artifacts; contrast enhancement is performed to increase the brightness difference between the opening areas and the background area by 30%; and edge sharpening is applied to improve the clarity of the opening edges by 25%. Subsequently, a contour extraction algorithm is used to identify five circular opening areas, eliminating two scratch interference areas on the top surface of the sample. For each opening area, all pixels on its edge contour are collected, and the average horizontal and vertical coordinates are calculated to obtain the center positions of the five TSV openings as (120, 150), (230, 210), (350, 300), (480, 270), and (590, 180), respectively. Based on the center positions of these 5 openings, the shooting angle and position of the optical imaging component are adjusted to capture images of the area on the bottom surface of the sample that is perpendicular to the center position of each opening, resulting in 5 second optical images. Each second optical image clearly contains the emission area of ​​the corresponding through-silicon via on the bottom surface.

[0021] Step 20: Determine the emission center position of each through-silicon via on the bottom surface based on the second optical image, and calculate the axial offset vector of each through-silicon via in the vertical direction based on the opening center position and the emission center position.

[0022] Optionally, each second optical image is preprocessed, and the preprocessing process is the same as that for the first optical image in step 10, namely, image noise reduction, contrast enhancement, and edge sharpening are performed sequentially to ensure that the emission region features of the through-silicon vias on the bottom surface are clear in the second optical image, avoiding deviations in the positioning of the emission center due to image quality issues. The defect detection device extracts the contour of each second optical image to locate the emission region of the corresponding through-silicon via on the bottom surface. The emission region refers to the opening region formed on the bottom surface of the sample after the through-silicon via penetrates the three-dimensional stacked packaged sample. Its contour features are consistent with the opening region on the top surface, both being circular or pre-defined closed regions. During the contour extraction process, the defect detection device uses an image recognition algorithm to identify regions in the second optical image that match the emission region features, eliminating interference regions such as stains and scratches on the bottom surface of the sample.

[0023] Furthermore, the defect detection device calculates the emission center position of each emission region. The emission center position refers to the geometric center of the emission region of the through silicon via. The calculation process is the same as the calculation process of the opening center position in step 10: the defect detection device first determines all the pixels on the edge contour of each emission region, and then calculates the average horizontal coordinate and the average vertical coordinate of these pixels. The coordinate point corresponding to the average horizontal coordinate and the average vertical coordinate is the emission center position of the through silicon via on the bottom surface.

[0024] Furthermore, based on these two center positions, the defect detection device calculates the axial offset vector of each through-silicon via in the vertical direction. The vertical direction refers to the direction perpendicular to the top and bottom surfaces of the three-dimensional stacked package sample. The axial offset vector is a vector used to characterize the deviation of the through-silicon via axis from the ideal axis in the vertical direction. The calculation process is as follows: a spatial rectangular coordinate system is established with the top surface of the three-dimensional stacked package sample as the reference plane. The normal of the reference plane is the vertical direction (i.e., the z-axis direction), and the plane on the reference plane is the xy plane. The coordinates of the opening center position in this coordinate system are (x1, y1, 0), and the coordinates of the emission center position in this coordinate system are (x2, y2, h). h is the thickness of the three-dimensional stacked package sample, i.e., the vertical distance between the top and bottom surfaces. The method for calculating the axis offset vector is as follows: calculate the coordinate differences between the ejection center position and the opening center position in the x-axis direction, y-axis direction, and z-axis direction respectively. The coordinate difference in the x-axis direction is x2-x1, the coordinate difference in the y-axis direction is y2-y1, and the coordinate difference in the z-axis direction is h-0=h. By combining the three coordinate differences, the axis offset vector of the through-silicon via in the vertical direction can be obtained. The vector can accurately reflect the offset of the axis of the through-silicon via in the vertical direction.

[0025] Continuing with the above embodiment, the five second optical images are preprocessed sequentially, including image noise reduction, contrast enhancement, and edge sharpening, to ensure that the emission regions of the corresponding through-silicon vias (TSVs) are clearly identifiable in each second optical image. Using a contour extraction algorithm, the emission regions of the five TSVs on the bottom surface are identified, eliminating one area of ​​interference from dirt on the bottom surface. The emission center positions of each emission region are calculated, resulting in five emission center positions: (121, 150), (229, 211), (350, 301), (481, 269), and (589, 180) (unit: pixels). Given that the thickness h of the three-dimensional stacked packaged sample is 100 micrometers, a spatial rectangular coordinate system is established with the top surface of the sample as the reference plane. The coordinates of the opening center position are (120, 150, 0), (230, 210, 0), (350, 300, 0), (480, 270, 0), (590, 180, 0) (unit: pixels, z-axis unit: micrometers). The coordinates of the emission center position are (121, 150, 100), (229, 211, 100), (350, 301, 100), (481, 269, 100), (589, 180, 100). Calculate the axis offset vector for each through-silicon via: For the first through-silicon via, the x-axis difference is 121-120=1, the y-axis difference is 150-150=0, and the z-axis difference is 100, so the axis offset vector is (1, 0, 100); For the second through-silicon via, the x-axis difference is 229-230=-1, the y-axis difference is 211-210=1, and the z-axis difference is 100, so the axis offset vector is (-1, 1, 100); For the third through-silicon via, the x-axis difference is 350-350=0, and the y-axis difference is... The value of the fourth through-silicon via is 301-300=1, the z-axis difference is 100, and the axis offset vector is (0, 1, 100); the x-axis difference of the fourth through-silicon via is 481-480=1, the y-axis difference is 269-270=-1, the z-axis difference is 100, and the axis offset vector is (1, -1, 100); the x-axis difference of the fifth through-silicon via is 589-590=-1, the y-axis difference is 180-180=0, the z-axis difference is 100, and the axis offset vector is (-1, 0, 100).

[0026] Step 30: Extract the sidewall contour line of each through-silicon via along the depth direction based on the third optical image of the sidewall region of the three-dimensional stacked package sample, and construct the target spatial trajectory of each through-silicon via based on the axis offset vector and the sidewall contour line.

[0027] Optionally, the defect detection device acquires a third optical image of the sidewall region of the three-dimensional stacked package sample. The sidewall region refers to the side surface area of ​​the three-dimensional stacked package sample, which reflects the sidewall morphology of the through-silicon vias (TSVs) along the depth direction. The third optical image is an image obtained by the defect detection device through an optical imaging component, capturing the sidewall region of the three-dimensional stacked package sample and clearly showing the sidewall features of the TSVs along the depth direction. The defect detection device preprocesses the third optical image, including image noise reduction, contrast enhancement, and edge sharpening, consistent with the image preprocessing processes in steps 10 and 20. The purpose is to remove noise from the third optical image and enhance the contrast between the sidewall contours and the background.

[0028] Furthermore, the defect detection device extracts the sidewall contour line of each through-silicon via along the depth direction based on the third optical image. The sidewall contour line refers to the boundary line between the sidewall of the through-silicon via and other areas inside the three-dimensional stacked package sample. The extraction process in this embodiment of the invention is as follows: by using an image edge detection algorithm, the edge pixels of the sidewall of each through-silicon via in the third optical image are identified, and these edge pixels are connected sequentially along the depth direction to form a continuous closed contour line, which is the sidewall contour line of the through-silicon via along the depth direction.

[0029] Furthermore, the defect detection device constructs the target spatial trajectory of each through-silicon via (TSV) based on the axial offset vector of each TSV and the extracted sidewall contour line, as described in steps 301 to 303. The target spatial trajectory refers to the actual extension trajectory of the TSV in three-dimensional space.

[0030] Step 40: Based on the target spatial trajectory, determine whether each through-silicon via has any morphological abnormalities that deviate from the preset spatial path, and obtain the through-silicon via defect detection results.

[0031] Optionally, the preset spatial path refers to the three-dimensional spatial extension path of the through-silicon via (TSV) that is preset during the design process and conforms to production standards. This path is determined based on the design parameters of the TSV (including opening size, ejection size, depth, preset axial direction, etc.). The defect detection device compares and analyzes the target spatial trajectory of each TSV with the preset spatial path to determine whether there are any morphological anomalies in the target spatial trajectory of each TSV that deviate from the preset spatial path, thereby obtaining the TSV defect detection result. Among them, morphological anomalies refer to deviations between the actual spatial trajectory of the TSV and the preset spatial path that exceed the allowable deviation range. The TSV defect detection results in this embodiment of the invention include tilting anomalies caused by etching depth control failure and serpentine anomalies caused by uneven sidewall stress release.

[0032] The embodiments of the present invention effectively avoid the problems of missed detection and misjudgment caused by the inability to capture three-dimensional morphology, thereby improving the identification accuracy of TSV through-hole defects.

[0033] Optionally, steps 301 to 303 include: Step 301: Construct a straight line segment connecting the opening center position indicated by the axis offset vector and the exit center position in the preset three-dimensional space to obtain the initial axial direction. Starting from the top depth along the depth direction, determine the first candidate trajectory point of the current depth layer based on the projection point set of the contour points in the sidewall contour line of the current depth layer on the plane perpendicular to the initial axial direction and the initial axial direction.

[0034] Optionally, the preset three-dimensional space refers to a virtual space that is completely consistent with the actual three-dimensional space size and orientation of the three-dimensional stacked packaging sample. This space is used to carry out the construction process of the target spatial trajectory of the through silicon via, and its spatial coordinate system is consistent with the spatial rectangular coordinate system established when calculating the axis offset vector in step 20.

[0035] Optionally, the defect detection device constructs a straight line segment in a preset three-dimensional space that connects the opening center position and the exit center position indicated by the axis offset vector. The two endpoints of the straight line segment correspond to the opening center position and the exit center position, respectively. The extension direction of the straight line segment is the initial axial direction, which refers to the initial reference direction of the through-silicon via axis.

[0036] Furthermore, the defect detection device performs depth layering processing on the through-silicon vias (TSVs) starting from the top depth along the depth direction. Here, the top depth refers to the starting depth of the TSV corresponding to the top surface of the three-dimensional stacked package sample, the depth direction refers to the extension direction from the top opening of the TSV to the bottom exit port, and the depth layering processing refers to uniformly dividing the TSV into multiple continuous depth layers along the depth direction. The thickness of each depth layer is consistent, and the thickness value must be selected to ensure that the sidewall morphology changes of the TSV along the depth direction can be completely captured.

[0037] Furthermore, the defect detection device acquires the contour points in the sidewall contour line corresponding to each depth layer. The contour points refer to the pixel points on the sidewall contour line of the through-silicon via (TSV) along the depth direction determined in step 30. The contour points corresponding to each depth layer are the boundary points between the TSV sidewall and other areas inside the three-dimensional stacked package sample at that depth layer location, which can accurately reflect the sidewall position of the TSV at that depth layer.

[0038] Furthermore, the defect detection device projects all contour points of the current depth layer onto a plane perpendicular to the initial axial direction to obtain a projection point set. The plane perpendicular to the initial axial direction refers to a plane that forms a 90-degree angle with the initial axial direction. This plane matches the spatial position of the current depth layer. The projection point set refers to the collection of points formed after all contour points are projected, which can reflect the distribution of the sidewall contour of the current depth layer in the direction perpendicular to the initial axial direction.

[0039] Furthermore, based on the obtained set of projection points and the initial axial direction, the defect detection device determines the first candidate trajectory point of the current depth layer, as described in steps 3021 to 3023. The first candidate trajectory point refers to the candidate position point of the target spatial trajectory of the through-silicon via at the current depth layer.

[0040] Step 302: Based on the first candidate trajectory point of the current depth layer and the second candidate trajectory point of the previous depth layer, determine the current axial direction, and based on the distribution of the contour points of the current depth layer on the plane perpendicular to the current axial direction, determine the current principal normal direction.

[0041] Optionally, the previous depth layer refers to the depth layer located above and adjacent to the current depth layer along the depth direction, and the second candidate trajectory point refers to the candidate position point of the through-silicon via target spatial trajectory obtained at the previous depth layer according to the same determination logic as the first candidate trajectory point of the current depth layer.

[0042] After acquiring two candidate trajectory points, the defect detection device determines the current axial direction based on the first candidate trajectory point at the current depth layer and the second candidate trajectory point at the previous depth layer. The current axial direction refers to the actual reference direction of the through-silicon via (TSV) axis at the current depth layer. The determination logic of this embodiment is as follows: a straight line segment is constructed with the first and second candidate trajectory points as the two endpoints. The extension direction of this straight line segment is the current axial direction. In this way, the axial direction can be dynamically adjusted according to the actual morphological changes of the TSV along the depth direction, so that the axial direction is more in line with the actual extension trend of the TSV and the trajectory construction deviation caused by fixing the initial axial direction can be avoided.

[0043] After determining the current axial direction, the defect detection device acquires all contour points of the current depth layer, projects these contour points onto a plane perpendicular to the current axial direction, and analyzes the distribution of the contour points on this plane. The plane perpendicular to the current axial direction refers to a plane that forms a 90-degree angle with the current axial direction. This plane matches the spatial position of the current depth layer, and the distribution of the contour points on this plane can reflect information such as the symmetry features and dimensional changes of the through-silicon via sidewalls at the current depth layer.

[0044] Furthermore, based on the distribution of contour points on a plane perpendicular to the current axial direction, the defect detection device determines the current main normal direction. The current main normal direction refers to the direction perpendicular to the current axial direction, passing through the central region of the through-silicon via at the current depth layer, and reflecting the dominant direction of the sidewall contour distribution. The determination logic of this embodiment is as follows: based on the distribution density and distribution range of contour points on the plane, the direction in which the contour points are most widely distributed is determined. This direction is the current main normal direction, which can accurately reflect the main extension direction of the through-silicon via sidewall at the current depth layer.

[0045] Step 303: Based on the cross product of the current axial direction and the current principal normal direction, the current sub-normal direction is obtained, and based on the difference between the first maximum spacing of the contour points of the current depth layer in the current principal normal direction and the second maximum spacing in the current sub-normal direction, the target spatial trajectory is constructed.

[0046] Optionally, the defect detection device performs a cross product between the current axial direction and the current principal normal direction to obtain the current sub-normal direction. The cross product operation is a spatial vector operation, and the operation logic is as follows: taking the current axial direction as the first vector and the current principal normal direction as the second vector, a new vector is obtained through the cross product operation. This new vector is perpendicular to both the current axial direction and the current principal normal direction, which is the current sub-normal direction. The current sub-normal direction, together with the current axial direction and the current principal normal direction, constitutes a three-dimensional orthogonal coordinate system, which can comprehensively characterize the spatial orientation of the through-silicon via at the current depth layer.

[0047] After determining the current secondary normal direction, the defect detection device acquires all contour points of the current depth layer and calculates the first maximum spacing of these contour points in the current principal normal direction. The first maximum spacing refers to the distance between the two farthest projection points among all contour points of the current depth layer in the current principal normal direction. The calculation process of this embodiment is as follows: project all contour points onto the current principal normal direction to obtain the projection coordinates of each contour point, find the maximum and minimum values ​​of the projection coordinates, and the difference between the maximum and minimum values ​​is the first maximum spacing. The first maximum spacing can reflect the sidewall size of the through-silicon via in the current depth layer in the principal normal direction. Simultaneously, the defect detection device calculates the second maximum spacing of all contour points in the current depth layer in the current sub-normal direction. The second maximum spacing refers to the distance between the two farthest projection points among all contour points in the current depth layer in the current sub-normal direction. Its calculation process is the same as that of the first maximum spacing: project all contour points onto the current sub-normal direction to obtain the projection coordinates of each contour point, find the maximum and minimum values ​​of the projection coordinates, and the difference between the maximum and minimum values ​​is the second maximum spacing. The second maximum spacing can reflect the sidewall size of the through-silicon via in the current depth layer in the sub-normal direction.

[0048] After calculating the first maximum spacing and the second maximum spacing, the defect detection device calculates the spacing difference between the two, that is, the value obtained by subtracting the second maximum spacing from the first maximum spacing. Then, based on the spacing difference, the target spatial trajectory of each through-silicon via is constructed, as in steps 3031 to 3034.

[0049] The embodiments of this invention achieve accurate reconstruction of the three-dimensional morphology of through-silicon vias (TSVs). By deeply integrating the spatial position information of the axial offset with the sidewall morphology details in the depth direction, it effectively compensates for the deficiency that relying solely on a single axial reference or sidewall contour cannot completely capture the three-dimensional extension trajectory of TSVs. It can accurately present the spatial distribution characteristics of asymmetric defects such as tilting and local collapse of TSVs along the depth direction, improve the three-dimensional characterization capability of TSV defect detection, and thus improve the identification accuracy of TSV defects.

[0050] Optionally, steps 3011 to 3013 include: Step 3011: Take the midpoint of the line connecting the projection points of the contour points of the current depth layer with the farthest target projection point as the top local center point, take the direction of the line as the initial principal normal direction, and perform a cross product between the initial axial direction and the initial principal normal direction to obtain the initial secondary normal direction.

[0051] Optionally, the defect detection device traverses all projection points in the projection point set and calculates the distance between any two projection points. The distance calculation adopts the spatial two-point distance calculation method. After the traversal is completed, the two projection points with the farthest distance are selected and these two projection points are determined as the target projection point pair. The target projection point pair refers to the two projection points with the largest spacing in the projection point set, which can reflect the maximum distribution range of the current depth layer sidewall contour in the direction perpendicular to the initial axial direction.

[0052] Furthermore, the defect detection device calculates the midpoint of the line connecting the target projection point pair and determines this midpoint as the top local center point. The top local center point refers to the local center reference point of the through-silicon via sidewall profile on the plane perpendicular to the initial axial direction at the current depth layer. The calculation process is as follows: calculate the average coordinates of the target projection point pair on each coordinate axis in the spatial coordinate system. The spatial point formed by the average coordinates on each coordinate axis is the midpoint of the line connecting the target projection point pair, which is the top local center point.

[0053] Furthermore, the defect detection device determines the direction of the line connecting the target projection point pair as the initial principal normal direction. The initial principal normal direction refers to the direction that is perpendicular to the initial axial direction, passes through the top local center point, and can reflect the dominant distribution direction of the current depth layer sidewall profile in the direction perpendicular to the initial axial direction. This direction is completely consistent with the direction of the line connecting the target projection point pair and can accurately reflect the main extension direction of the current depth layer sidewall profile on the plane.

[0054] Furthermore, the defect detection device performs a cross product operation on these two directions to obtain the initial subnormal direction. The cross product operation is a spatial vector operation, and the operation logic is as follows: taking the initial axial direction as the first vector and the initial principal normal direction as the second vector, a new vector is obtained through the cross product operation. This new vector is perpendicular to both the initial axial direction and the initial principal normal direction, which is the initial subnormal direction. The initial subnormal direction, together with the initial axial direction and the initial principal normal direction, constitutes a three-dimensional orthogonal coordinate system, which can comprehensively characterize the spatial orientation of the through-silicon via at the current depth layer in the plane perpendicular to the initial axial direction.

[0055] Step 3012: Project each contour point of the current depth layer along the initial principal normal direction and the initial subnormal direction to obtain the principal normal projection extreme point and the subnormal projection extreme point.

[0056] Optionally, the defect detection device projects each contour point of the current depth layer along the initial principal normal direction to obtain the projection point of each contour point in the initial principal normal direction. It then iterates through all these projection points and selects the projection point with the largest projection coordinate and the projection point with the smallest projection coordinate. These two projection points are determined as the principal normal projection extreme points. The principal normal projection extreme points refer to the extreme position points of the contour points of the current depth layer projected in the initial principal normal direction, which can reflect the maximum extension range of the sidewall contour of the current depth layer in the initial principal normal direction.

[0057] Furthermore, the defect detection device projects each contour point of the current depth layer along the initial subnormal direction, obtaining the projection point of each contour point in the initial subnormal direction. It iterates through all these projection points, selecting the projection point with the largest and smallest projection coordinates. These two projection points are designated as subnormal projection extrema points. The subnormal projection extrema points refer to the extreme positions of the contour points of the current depth layer projected in the initial subnormal direction, reflecting the maximum extension range of the sidewall contour of the current depth layer in the initial subnormal direction. It should be noted that during the projection process, it is ensured that the projection operation of each contour point is performed in the same three-dimensional coordinate system, and the projection direction follows the initial principal normal direction and the initial subnormal direction to avoid errors in identifying projection extrema points due to projection direction deviations or inconsistencies in the coordinate system.

[0058] Step 3013: Based on the average position of the first midpoint between the principal normal projection extrema and the second midpoint between the secondary normal projection extrema, determine the first candidate trajectory point of the current depth layer.

[0059] Optionally, there are two extreme points for the principal normal projection, corresponding to the maximum and minimum projected coordinates in the initial principal normal direction, respectively. There are also two extreme points for the secondary normal projection, corresponding to the maximum and minimum projected coordinates in the initial secondary normal direction, respectively.

[0060] For the extreme points of the principal normal projection, the defect detection device calculates the first midpoint between the two extreme points. The first midpoint refers to the midpoint of the line connecting the two extreme points of the principal normal projection. The calculation process is as follows: calculate the average coordinates of the two extreme points of the principal normal projection on each coordinate axis in the spatial coordinate system. The spatial point formed by the average coordinates on each coordinate axis is the first midpoint. The first midpoint can reflect the center position of the sidewall profile of the current depth layer in the initial principal normal direction.

[0061] For the extreme points of the subnormal projection, the same calculation method as for the first midpoint is used to calculate the second midpoint between the two extreme points. The second midpoint refers to the midpoint of the line connecting the two extreme points of the subnormal projection, which can reflect the center position of the sidewall profile of the current depth layer in the initial subnormal direction.

[0062] Furthermore, the defect detection device calculates the average position of the two midpoints and determines the average position as the first candidate trajectory point of the current depth layer. The calculation process of the average position is as follows: calculate the average coordinates of the first midpoint and the second midpoint on each coordinate axis in the spatial coordinate system. The spatial point formed by the average coordinates on each coordinate axis is the average position of the two midpoints, which is the first candidate trajectory point of the current depth layer. This first candidate trajectory point can comprehensively reflect the central features of the sidewall contour of the current depth layer in both the initial principal normal and the initial secondary normal directions.

[0063] The embodiments of the present invention can accurately capture the local center features of each layer along the depth direction of the through-silicon via (TSV), combine the distribution range of the sidewall contour with the spatial orientation information, improve the accuracy of the restoration of the three-dimensional spatial trajectory of the TSV, ensure the accuracy of defect judgment based on the trajectory, make up for the defect that a single reference positioning cannot accurately capture the local changes of the sidewall contour, optimize the detail accuracy of the three-dimensional morphology representation of the TSV, and thus improve the identification accuracy of TSV defects.

[0064] Optionally, steps 3031 to 3034 include: Step 3031: Based on the spacing difference, the first candidate trajectory point of the current depth layer is offset by a preset offset distance along the current principal normal direction to the side with smaller spacing, so as to obtain the current corrected trajectory point.

[0065] Optionally, the defect detection device determines the dimensional deviation of the sidewall profile of the current depth layer in the principal normal direction based on the spacing difference, and identifies the side with the smaller spacing. The side with the smaller spacing is defined as follows: if the spacing difference is positive, it means that the first maximum spacing is greater than the second maximum spacing, that is, the sidewall size in the current principal normal direction is greater than that in the secondary normal direction, and the side with the smaller spacing is the sidewall corresponding to the current secondary normal direction; if the spacing difference is negative, it means that the first maximum spacing is less than the second maximum spacing, that is, the sidewall size in the current principal normal direction is less than that in the secondary normal direction, and the side with the smaller spacing is the sidewall corresponding to the current principal normal direction; if the spacing difference is zero, it means that the sidewall size in both directions is the same, and no offset operation is required.

[0066] After determining the side with smaller spacing, the defect detection device calculates a preset offset distance, which is one-quarter of the spacing difference. The calculation process is as follows: divide the absolute value of the spacing difference by 4, and the resulting value is the preset offset distance. The direction of the offset distance is consistent with the side with smaller spacing, ensuring that the position of the first candidate trajectory point can be corrected after the offset, so that it is closer to the actual center trajectory of the through silicon via.

[0067] Furthermore, the defect detection device shifts the first candidate trajectory point of the current depth layer along the current principal normal direction to the side with smaller spacing by a preset offset distance to obtain the current corrected trajectory point. The current corrected trajectory point is the trajectory point after deviation correction of the first candidate trajectory point, which can more accurately represent the actual center position of the silicon via in the current depth layer. During the correction process, the offset direction follows the current principal normal direction, and the offset distance is executed according to the calculated preset offset distance.

[0068] Step 3032: Sequentially determine the corrected trajectory points of each depth layer from the top depth to the bottom depth to obtain a preliminary spatial trajectory composed of discrete trajectory points. Based on the spatial vector between the first point of the preliminary spatial trajectory and the center position of the opening, translate the preliminary spatial trajectory so that the first point of the translated trajectory coincides with the center position of the opening, and obtain the translated spatial trajectory.

[0069] Optionally, following the operation logic of step 3031, each depth layer from the top depth to the bottom depth is processed sequentially to determine the corrected trajectory point of each depth layer. The top depth refers to the starting depth of the through-silicon via corresponding to the top surface of the three-dimensional stacked package sample, and the bottom depth refers to the ending depth of the through-silicon via corresponding to the bottom surface of the three-dimensional stacked package sample. Each depth layer is a continuous depth layer formed by uniformly dividing the through-silicon via along the depth direction in step 301, and each depth layer corresponds to a corrected trajectory point.

[0070] After all the corrected trajectory points for all depth layers are determined, the defect detection device arranges these corrected trajectory points sequentially from top to bottom in the depth direction to obtain a preliminary spatial trajectory composed of discrete trajectory points. The preliminary spatial trajectory is composed of multiple discrete corrected trajectory points and can initially reflect the three-dimensional extension trend of the through-silicon via. The arrangement order of the discrete trajectory points is completely consistent with the order of the through-silicon via depth layers to ensure that the trajectory can correspond to the actual extension direction of the through-silicon via from top to bottom.

[0071] Furthermore, the defect detection device acquires the starting point and opening center position of the preliminary spatial trajectory. The starting point of the preliminary spatial trajectory refers to the corrected trajectory point in the depth layer corresponding to the top depth in the preliminary spatial trajectory. The spatial vector between the starting point of the preliminary spatial trajectory and the opening center position is calculated. The spatial vector refers to the vector with the starting point of the preliminary spatial trajectory as the starting point and the opening center position as the ending point. The calculation process is as follows: calculate the coordinate difference between the opening center position and the starting point of the preliminary spatial trajectory on each coordinate axis in the spatial coordinate system. The vector formed by each coordinate difference is the spatial vector, which can reflect the positional deviation between the starting point of the preliminary spatial trajectory and the opening center position.

[0072] Furthermore, the defect detection device performs a translation operation on the preliminary spatial trajectory based on the spatial vector. The translation direction is consistent with the direction of the spatial vector, and the translation distance is equal to the magnitude of the spatial vector, so that the first point of the translated preliminary spatial trajectory completely coincides with the center of the opening, thus obtaining the translated spatial trajectory. It should be noted that during the translation operation, all discrete trajectory points in the preliminary spatial trajectory are translated according to the same spatial vector to ensure that the overall extension trend of the trajectory does not change. Only the spatial position of the trajectory is adjusted to precisely align it with the center of the top opening of the through silicon via.

[0073] Step 3033: Based on the positional deviation between the end point of the translated spatial trajectory and the launch center position, calculate the minimum rotation angle around the center position of the opening, so that the rotated end point coincides with the launch center position, and apply the minimum rotation angle to the translated spatial trajectory to obtain the aligned spatial trajectory point sequence.

[0074] Optionally, the defect detection device determines the endpoint of the translated spatial trajectory. This endpoint refers to the corrected trajectory point corresponding to the depth layer at the bottom depth in the translated spatial trajectory, i.e., the last discrete trajectory point arranged along the depth direction in the translated spatial trajectory. The positional deviation between the endpoint of the translated spatial trajectory and the exit center position is calculated. This positional deviation refers to the difference in spatial distance and direction between the endpoint and the exit center position. The calculation process is as follows: calculate the coordinate difference between the exit center position and the endpoint on each coordinate axis in the spatial coordinate system, and determine the direction and magnitude of the positional deviation based on the coordinate difference. This positional deviation can reflect the offset between the endpoint of the translated spatial trajectory and the exit center at the bottom of the through-silicon via.

[0075] Based on this positional deviation, the defect detection device calculates the minimum rotation angle around the center of the opening. The minimum rotation angle refers to the minimum angle required to rotate the translated spatial trajectory with the center of the opening as the rotation center so that the end point of the translated spatial trajectory coincides with the exit center position. The calculation process is as follows: a local coordinate system is established with the center of the opening as the origin. Based on the coordinates of the end point and the exit center position in the local coordinate system, the angle between the two points is calculated. This angle is the minimum rotation angle. The rotation direction is selected so that the end point can reach the exit center position as quickly as possible, ensuring that the rotation angle is minimized and avoiding the trajectory deviating from the actual extension trend of the through-silicon via due to an excessively large rotation angle.

[0076] The defect detection device applies the minimum rotation angle to the translated spatial trajectory, and rotates it in a determined direction with the center of the opening as the rotation center. This ensures that the end point of the translated spatial trajectory after rotation completely coincides with the exit center, resulting in an aligned spatial trajectory point sequence. This aligned spatial trajectory point sequence is a discrete trajectory point sequence that has been corrected by translation and rotation, with the first point coinciding with the center of the opening and the end point coinciding with the exit center. This allows for a more accurate fit to the actual spatial extension range of the through-silicon via.

[0077] Step 3034: Based on the aligned spatial trajectory point sequence, connect adjacent trajectory points sequentially to form polyline segments, and insert a transition point at the connection of every two adjacent polyline segments to obtain the target spatial trajectory.

[0078] Optionally, the defect detection device connects two adjacent trajectory points sequentially according to the arrangement order of discrete trajectory points in the aligned spatial trajectory point sequence to form a broken line segment. Here, adjacent trajectory points refer to two discrete trajectory points that are adjacent in position and arranged sequentially in the depth direction in the sequence, and a broken line segment refers to a straight line segment connecting two adjacent trajectory points. After all adjacent trajectory points are connected, a broken line trajectory composed of multiple broken line segments is formed. This broken line trajectory can initially present the three-dimensional extension shape of the through silicon via, but there are sharp corners at the connection of the broken lines, which cannot accurately reflect the smooth extension characteristics of the through silicon via trajectory.

[0079] The defect detection device inserts a transition point at the junction of every two adjacent broken line segments. The transition point is the midpoint used to connect two adjacent broken line segments and smooth the trajectory. The position of the transition point satisfies two conditions: First, the transition point lies on the angle bisector of the plane formed by the preceding and following broken line segments. The plane formed by the preceding and following broken line segments refers to the plane jointly determined by the preceding and following broken line segments, and the angle bisector is the straight line in this plane that bisects the angle between the preceding and following broken line segments. Second, the distance from the transition point to the connecting vertex is equal to one-tenth of the harmonic average of the lengths of the preceding and following broken line segments. The connecting vertex refers to the junction of two adjacent broken line segments, and the lengths of the preceding and following broken line segments refer to the lengths of the preceding and following broken line segments. The harmonic average is calculated as follows: first, calculate the reciprocal of the lengths of the preceding and following broken line segments; then, calculate the average of the two reciprocals; finally, take the reciprocal of this average to obtain the harmonic average; and then divide the harmonic average by 10 to obtain the distance from the transition point to the connecting vertex.

[0080] After the transition points at the connection of all adjacent polyline segments are inserted, the defect detection device will connect the discrete trajectory points in the aligned spatial trajectory point sequence and the inserted transition points in sequence to form a smooth continuous trajectory, which is the target spatial trajectory.

[0081] This invention combines sidewall size differences, spatial position deviations, and trajectory smoothness optimization to gradually correct trajectory deviations, align trajectory endpoints, and optimize trajectory morphology. This effectively solves the defects in trajectory construction, such as inaccurate positioning, rough trajectories, and discrepancies with the actual extension trend of TSVs. It achieves accurate and smooth construction of TSV target spatial trajectories, improves the accuracy and reliability of TSV defect detection, optimizes the complete characterization capability of TSV three-dimensional morphology, and enhances the accuracy of TSV defect identification.

[0082] Optionally, steps 401 to 404 include: Step 401: Based on the spatial trajectory points arranged in depth order in the target spatial trajectory and the reference points at the corresponding depth positions in the preset spatial path, perform layer-by-layer position deviation calculation to obtain the three-dimensional offset vector between the spatial trajectory point and the reference point at each depth position.

[0083] Optionally, the defect detection device determines spatial trajectory points arranged in depth order in the target spatial trajectory. The spatial trajectory points refer to discrete points that are uniformly distributed on the target spatial trajectory and can characterize the trajectory extension shape. Arranging in depth order means that the spatial trajectory points are sorted in sequence from the top depth to the bottom depth. Each spatial trajectory point corresponds to a specific depth position, which corresponds one-to-one with the depth layer position after the through silicon via is processed in step 301.

[0084] Furthermore, the defect detection device determines a reference point at a corresponding depth position in the preset spatial path. The corresponding depth position refers to a position that is completely consistent with the depth position of the spatial trajectory point in the target spatial trajectory. The reference point refers to a benchmark point on the preset spatial path that is located at the corresponding depth position and meets the design standards. Each reference point corresponds one-to-one with a spatial trajectory point in the target spatial trajectory.

[0085] Furthermore, the defect detection device performs layer-by-layer positional deviation calculations between the spatial trajectory point and the reference point at each depth location to obtain a three-dimensional offset vector at each depth location. The layer-by-layer positional deviation calculation refers to calculating the positional deviation between the spatial trajectory point and the reference point at each depth location sequentially according to the depth order. The three-dimensional offset vector is a vector used to characterize the offset of the spatial trajectory point at that depth location relative to the reference point in three-dimensional space. The calculation process is as follows: calculate the coordinate difference between the spatial trajectory point and the reference point at that depth location on each coordinate axis in the spatial coordinate system. The vector formed by the coordinate differences is the three-dimensional offset vector at that depth location. This vector can intuitively reflect the offset direction and magnitude of the actual trajectory of the through-silicon via at each depth location from the preset path.

[0086] Step 402: Perform vector projection decomposition based on the three-dimensional offset vector at each depth position and the tangent direction of the preset spatial path at the corresponding depth position to obtain the axial offset component along the tangent direction of the preset spatial path and the lateral offset component perpendicular to the tangent direction.

[0087] Optionally, the defect detection device acquires the tangent direction of the preset spatial path at the corresponding depth position, wherein the corresponding depth position refers to the position that is completely consistent with the depth position of the three-dimensional offset vector, and the tangent direction refers to the direction pointed to by the tangent at the corresponding depth position on the preset spatial path, which can reflect the extension direction of the preset spatial path at the depth position.

[0088] The process of obtaining the tangent direction is as follows: extract two adjacent reference points at the depth position on the preset spatial path, construct a straight line segment connecting the two reference points, and the extension direction of the straight line segment is the tangent direction of the preset spatial path at the depth position.

[0089] Furthermore, the defect detection device performs vector projection decomposition on the three-dimensional offset vector and the tangent direction at each depth position. Vector projection decomposition involves decomposing the three-dimensional offset vector into two mutually perpendicular components: an axial offset component along the tangent direction of a preset spatial path and a lateral offset component perpendicular to the tangent direction. The decomposition process follows the principle of spatial vector projection, ensuring that the vector sum of the two components equals the original three-dimensional offset vector. The axial offset component refers to the projection of the three-dimensional offset vector onto the tangent direction of the preset spatial path, reflecting the offset of the actual trajectory of the through-silicon via (TSV) along the extension direction of the preset path, i.e., the offset along the TSV axis. The lateral offset component refers to the projection of the three-dimensional offset vector onto the direction perpendicular to the tangent direction of the preset spatial path, reflecting the offset of the actual trajectory of the TSV along the extension direction of the preset path, i.e., the offset perpendicular to the TSV axis. The two components are mutually perpendicular and together constitute the complete offset information of the original three-dimensional offset vector.

[0090] Step 403: Calculate the modulus based on the lateral offset vector at each depth position to obtain the lateral offset component sequence, and sum the axial offset components at each depth position and all axial offset components from the top depth to the current depth to obtain the axial offset cumulative value sequence.

[0091] Optionally, the modulus of the lateral offset component at each depth position is calculated to obtain a sequence of lateral offset components. The modulus of the lateral offset component refers to the vector length of the lateral offset component, which reflects the actual magnitude of the lateral offset at that depth position. The sequence of lateral offset components refers to the sequence formed by arranging the modulus of the lateral offset components calculated at each depth position in order from the top depth to the bottom depth, which can completely reflect the lateral offset variation of the through-silicon via along the depth direction.

[0092] Simultaneously, the axial offset components at each depth position are summed. The summation range is all axial offset components from the top depth to the current depth, resulting in the cumulative axial offset value at each depth position. Here, the top depth refers to the starting depth of the through-silicon via, and the current depth refers to the depth position corresponding to the current summation operation. The summation process is as follows: starting from the axial offset component corresponding to the top depth, the axial offset components corresponding to each subsequent depth position are added sequentially until the axial offset component corresponding to the current depth position is obtained. The sum is the cumulative axial offset value at the current depth position.

[0093] Furthermore, the defect detection device arranges the cumulative axial offset values ​​at each depth position in order from the top depth to the bottom depth to obtain a sequence of cumulative axial offset values. This sequence can fully reflect the cumulative change of axial offset along the depth direction of the through silicon via.

[0094] Step 404: Perform morphological anomaly detection on each through-silicon via based on the lateral offset component sequence and the axial offset cumulative value sequence to obtain the through-silicon via defect detection results.

[0095] Optionally, each through-silicon via (TSV) is subjected to morphological anomaly detection based on the lateral offset component sequence and the axial offset cumulative value sequence to obtain TSV defect detection results, as detailed in steps 4041 to 4044.

[0096] This invention decomposes and integrates three-dimensional offset information in layers and directions, effectively solving the problems of traditional detection methods that cannot accurately distinguish offset directions and are difficult to capture offset changes along the depth direction. It can accurately identify various morphological anomalies such as tilting, local collapse, and excessive axis offset of TSVs, improving the accuracy and comprehensiveness of TSV defect detection, avoiding missed detections and misjudgments, and ensuring that the detection results can truly reflect the actual quality status of TSVs, thereby improving the accuracy of TSV defect identification.

[0097] Optionally, steps 4041 to 4044 include: Step 4041: Calculate the first-order difference of the cumulative axial offset value based on the cumulative axial offset value sequence and the depth spacing between adjacent depth positions to obtain the cumulative axial offset rate of change sequence.

[0098] Optionally, the defect detection device performs first-order difference calculation on the axial offset cumulative value sequence. The calculation process combines the depth spacing between adjacent depth positions to obtain a sequence of axial offset cumulative change rates. The first-order difference calculation refers to sequentially calculating the difference between two adjacent axial offset cumulative values ​​in the axial offset cumulative value sequence; that is, subtracting the axial offset cumulative value of the previous depth position from the axial offset cumulative value of the later depth position to obtain the axial offset cumulative difference between two adjacent depth positions. The axial offset cumulative change rate refers to the change in the axial offset cumulative value per unit depth. The calculation process is as follows: divide the axial offset cumulative difference between each pair of adjacent depth positions by the depth spacing between the adjacent depth positions to obtain the axial offset cumulative change rate between those two depth positions. Each adjacent depth position corresponds to one axial offset cumulative change rate. Further, these change rates are arranged sequentially according to their corresponding depth positions to obtain an axial offset cumulative change rate sequence. The length of this sequence is one less than the length of the axial offset cumulative value sequence. Each change rate corresponds to the axial offset cumulative change between two adjacent depth positions, which can intuitively reflect the rate and direction of change of the axial offset cumulative value along the depth direction of the through-silicon via.

[0099] Step 4042: Based on the lateral offset component sequence, the set of depth locations where the lateral offset component value exceeds the preset lateral anomaly threshold is determined as the lateral over-limit region. Based on the axial offset cumulative change rate sequence, the continuous depth interval where the axial offset cumulative change rate has the same sign and a value greater than zero at three or more consecutive depth locations is determined as the axial continuous offset region.

[0100] Optionally, the preset lateral anomaly threshold refers to a pre-set critical value used to determine whether the lateral offset component exceeds the normal range. Determined based on the design standards and production allowable deviations of the through-silicon via (TSV), it serves as the benchmark for judging whether the lateral offset is abnormal. When the lateral offset component value exceeds this threshold, it indicates that the lateral offset at the corresponding depth position exceeds the normal range, posing a potential morphological anomaly risk. The defect detection device traverses each lateral offset component value in the lateral offset component sequence, compares each lateral offset component value with the preset lateral anomaly threshold, and filters out all depth positions where the lateral offset component value exceeds the preset lateral anomaly threshold. These depth positions are integrated into a set and identified as the lateral over-limit region. The lateral over-limit region refers to the set of all depth positions where the lateral offset of the TSV exceeds the normal range along the depth direction; this region reflects the specific location distribution of the lateral offset anomaly of the TSV.

[0101] Furthermore, the defect detection device iterates through each cumulative axial offset rate of change in the sequence, analyzes its sign and magnitude, and filters out continuous depth intervals that meet specific conditions. These conditions are: the cumulative axial offset rates of change at three or more consecutive depth positions have the same sign, and the value of the cumulative axial offset rate of change at each depth position is greater than zero. The same sign means that the cumulative axial offset rates of change at all consecutive depth positions are either all positive or all all negative, reflecting a consistent direction of change in the cumulative axial offset value. A value greater than zero means that the cumulative axial offset rate of change is positive, reflecting a continuous increase in the cumulative axial offset value within that depth interval. The defect detection device defines these continuous depth intervals as axially continuous offset regions. An axially continuous offset region refers to a continuous depth interval where the cumulative axial offset value of the through-silicon via (TSV) continuously increases along the depth direction and maintains a consistent direction of change. This region can reflect the continuous abnormal change in the cumulative axial offset value of the TSV.

[0102] Step 4043: Perform a region intersection operation based on the lateral over-limit region and the axial continuous offset region, and determine the depth interval belonging to the lateral over-limit region and the axial continuous offset region as the composite deviation region.

[0103] Optionally, the defect detection device performs a region intersection operation on the lateral over-limit region and the axial continuous offset region. The region intersection operation refers to finding the depth interval that belongs to both the lateral over-limit region and the axial continuous offset region. That is, all depth positions within the depth interval belong to both the lateral over-limit region (lateral offset exceeds the normal range) and the axial continuous offset region (axial offset cumulative value continuously increases and changes in the same direction). The resulting intersection depth interval is determined as the composite deviation region.

[0104] Step 4044: Based on the lateral offset component values ​​corresponding to each depth position in the composite deviation region and the two consecutive depth positions before and after it, morphological anomaly detection is performed to obtain the through-silicon via defect detection results.

[0105] Optionally, morphological anomaly detection is performed based on the lateral offset component values ​​corresponding to each depth position in the composite deviation region and the two consecutive depth positions before and after it to obtain the through-silicon via defect detection results, as specifically in steps 40441 to 40444.

[0106] The embodiments of the present invention effectively solve the problems of inaccurate positioning of abnormal areas and easy omission of compound defects in traditional detection. It can accurately identify severe morphological defects formed by the superposition of lateral offset abnormality and axial continuous offset abnormality in TSV, improve the accuracy and pertinence of TSV defect detection, ensure that the detection results can truly and accurately reflect the actual quality status of TSV, and improve the identification accuracy of TSV defects.

[0107] Optionally, steps 40441 to 40444 include: Step 40441: Calculate the extreme difference based on the lateral offset component values ​​corresponding to each depth position in the composite deviation region and the two consecutive depth positions before and after it, to obtain the local lateral fluctuation amplitude.

[0108] Optionally, for each depth position in the composite deviation region, the defect detection device extracts the lateral offset component value corresponding to the depth position and the two consecutive depth positions before and after it. That is, each target depth position corresponds to the lateral offset component value of five consecutive depth positions (the two previous depth positions, the current depth position, and the two next depth positions), ensuring that the lateral offset fluctuation characteristics of the depth position and its surrounding area can be fully captured, avoiding the omission of fluctuation characteristics caused by focusing only on a single depth position.

[0109] Furthermore, the defect detection device performs extreme value difference calculation on the five lateral offset component values ​​corresponding to each target depth position to obtain the local lateral fluctuation amplitude. The extreme value difference calculation refers to first selecting the maximum and minimum values ​​from the five lateral offset component values, and then subtracting the minimum value from the selected maximum value. The difference obtained is the local lateral fluctuation amplitude corresponding to the target depth position. The local lateral fluctuation amplitude refers to the maximum fluctuation range of the lateral offset component values ​​at a certain depth position and its surrounding area within the composite deviation area. It can intuitively reflect the intensity of the lateral offset fluctuation in this area. The larger the fluctuation amplitude, the more unstable the lateral offset in this area is, and the higher the possibility of morphological anomalies.

[0110] Step 40442: Based on the curvature radius and local lateral fluctuation amplitude of the preset spatial path in the corresponding depth interval of the composite deviation region, a geometric tolerance comparison is performed. If the local lateral fluctuation amplitude is greater than one-tenth of the curvature radius, the composite deviation region is determined as the morphological distortion interval.

[0111] Optionally, the radius of curvature refers to the measure of the degree of curvature of the preset spatial path within the corresponding depth range of the composite deviation region. The larger the radius of curvature, the smoother the preset spatial path is within the depth range. The smaller the radius of curvature, the more pronounced the curvature of the preset spatial path is within the depth range. This radius of curvature is calculated based on the design parameters of the preset spatial path and serves as a benchmark for judging whether the lateral fluctuation exceeds the geometric tolerance.

[0112] Furthermore, the defect detection device calculates one-tenth of the radius of curvature of the preset spatial path in the depth range corresponding to the composite deviation area. The calculation process in this embodiment of the invention is as follows: divide the value of the radius of curvature by 10, and the resulting value is one-tenth of the radius of curvature. This value serves as the critical value for judging whether the amplitude of local lateral fluctuation exceeds the geometric tolerance, reflecting the maximum lateral fluctuation range allowed by the preset spatial path in the depth range.

[0113] Furthermore, the defect detection device compares the local lateral fluctuation amplitude corresponding to each depth position with one-tenth of the radius of curvature of the preset spatial path in the corresponding interval at that depth position using geometric tolerance. Geometric tolerance comparison refers to comparing the local lateral fluctuation amplitude with the critical value to determine whether the local lateral fluctuation exceeds the fluctuation range allowed by the preset spatial path.

[0114] Furthermore, if the comparison result shows that the amplitude of the local lateral fluctuation is greater than one-tenth of the radius of curvature, it indicates that the local lateral fluctuation corresponding to the depth position exceeds the geometric tolerance allowed by the preset spatial path, and the lateral offset fluctuation in this area has exceeded the normal range. The defect detection device then determines the composite deviation area where the depth position is located as the morphological distortion interval. The morphological distortion interval refers to the depth interval where the lateral fluctuation of the through silicon via exceeds the geometric tolerance and there is obvious morphological abnormality.

[0115] Step 40443: If the cumulative value of axial offset increases or decreases monotonically within the morphology distortion range, then the detection result of the through-silicon via defect is determined to be a tilting anomaly caused by etching depth control failure.

[0116] Optionally, the defect detection device extracts the cumulative axial offset values ​​from the cumulative axial offset value sequence, corresponding to the cumulative axial offset values ​​at all depth positions within the morphological distortion interval, and arranges them sequentially from the top depth to the bottom depth to obtain a subset of the cumulative axial offset values ​​within the morphological distortion interval. This subset can reflect the changing trend of the cumulative axial offset values ​​within the morphological distortion interval.

[0117] Furthermore, the defect detection device analyzes the changing trend of the subset of cumulative axial offset values ​​to determine whether the cumulative axial offset values ​​show a monotonically increasing or monotonically decreasing trend within the morphological distortion interval. A monotonically increasing trend means that within the morphological distortion interval, the cumulative axial offset value corresponding to each depth position from top to bottom is greater than the cumulative axial offset value corresponding to the previous depth position, and there is no decrease. A monotonically decreasing trend means that within the morphological distortion interval, the cumulative axial offset value corresponding to each depth position from top to bottom is less than the cumulative axial offset value corresponding to the previous depth position, and there is no increase.

[0118] If the judgment result is that the cumulative value of axial offset increases or decreases monotonically within the morphology distortion range, it indicates that the cumulative value of axial offset of the through-silicon via shows a continuous unidirectional change in the morphology abnormal area. This change characteristic is consistent with the offset law caused by etching depth control failure. The defect detection device then determines that the defect detection result of the through-silicon via is a tilting anomaly caused by etching depth control failure. The tilting anomaly refers to the through-silicon via having a continuous unidirectional offset of its axis along the depth direction due to improper etching depth control, forming a tilted morphology defect.

[0119] Step 40444: If the cumulative value of axial offset changes less than the preset axial stability threshold within the morphology distortion range and the value of lateral offset component fluctuates alternately, then the detection result of the through-silicon via defect is determined to be a serpentine anomaly caused by uneven stress release on the sidewall.

[0120] Optionally, the preset axial stability threshold is a pre-set critical value used to determine whether the change range of the cumulative axial offset is within a stable range. This threshold is determined based on the design standards and production allowable deviations of the through-silicon via. When the change range of the cumulative axial offset is less than this threshold, it indicates that the cumulative axial offset is in a stable state and there is no obvious unidirectional change trend.

[0121] Optionally, the defect detection device extracts a subset of the cumulative axial offset values ​​within the morphological distortion range and calculates the variation range of the subset. The calculation process for the variation range is as follows: the maximum and minimum values ​​in the subset of cumulative axial offset values ​​are selected, and the difference is obtained by subtracting the minimum value from the maximum value. This difference is the variation range of the cumulative axial offset value within the morphological distortion range, which is used to determine whether the cumulative axial offset value is stable.

[0122] Furthermore, the defect detection device extracts the lateral offset component values ​​corresponding to all depth positions within the morphological distortion range, analyzes their variation characteristics sequentially according to depth, and determines whether the lateral offset component values ​​exhibit an alternating fluctuation trend. The alternating fluctuation trend refers to the lateral offset component values ​​exhibiting a cyclical change state of "increase-decrease-increase-decrease" within the morphological distortion range, without a continuous increase or decrease pattern, reflecting that the lateral offset is unstable and fluctuates back and forth within this range.

[0123] Furthermore, if the judgment result is that the cumulative value of axial offset within the morphology distortion range is less than the preset axial stability threshold, and the lateral offset component value shows an alternating fluctuation trend, it indicates that the axial offset of the through-silicon via in the morphology abnormal area does not have a significant unidirectional change, but the lateral offset fluctuates back and forth. This change characteristic is consistent with the offset law caused by uneven release of sidewall stress. The defect detection device then determines that the defect detection result of the through-silicon via is a serpentine anomaly caused by uneven release of sidewall stress. The serpentine anomaly refers to the through-silicon via fluctuating back and forth along the depth direction due to uneven release of sidewall stress, forming a morphological defect similar to a serpentine motion.

[0124] This invention combines abnormal area location, fluctuation feature analysis, and defect type judgment, effectively solving the problem that traditional detection methods can only determine the existence of defects but cannot accurately distinguish the defect type. It refines the precision of TSV defect detection, improves the accuracy and practicality of TSV defect detection, avoids missed detections, misjudgments, and misjudgments of defect types, and ensures that the detection results not only reflect whether defects exist, but also clarify the cause and type of defects, thereby improving the accuracy of TSV defect identification.

[0125] Furthermore, the TSV via defect detection device in 3D stacked packaging provided by the present invention will be described below. The TSV via defect detection device in 3D stacked packaging described below can be referred to in correspondence with the TSV via defect detection method in 3D stacked packaging described above.

[0126] Optional, refer to Figure 2 , Figure 2 This is a schematic diagram of the TSV via defect detection device in 3D stacked packaging provided by the present invention. The TSV via defect detection device in 3D stacked packaging includes: The optical image acquisition module 210 is used to locate the opening center position of multiple through-silicon vias on the top surface of the three-dimensional stacked package sample based on a first optical image of the top surface of the sample, and to acquire a second optical image corresponding to each through-silicon via on the bottom surface of the sample based on the opening center position. The axis offset module 220 is used to determine the emission center position of each through-silicon via on the bottom surface based on the second optical image, and to calculate the axis offset vector of each through-silicon via in the vertical direction based on the opening center position and the emission center position. The spatial trajectory construction module 230 is used to extract the sidewall contour line of each through-silicon via along the depth direction based on the third optical image of the sidewall region of the three-dimensional stacked package sample, and to construct the target spatial trajectory of each through-silicon via based on the axis offset vector and the sidewall contour line. The defect detection module 240 is used to determine whether each through-silicon via has an abnormal shape that deviates from the preset spatial path based on the target spatial trajectory, and to obtain the through-silicon via defect detection results.

[0127] The embodiments of the present invention effectively avoid the problems of missed detection and misjudgment caused by the inability to capture three-dimensional morphology, thereby improving the identification accuracy of TSV through-hole defects.

[0128] Please see Figure 3 , Figure 3 An embodiment diagram of an electronic device provided in accordance with the present invention. For example... Figure 3 As shown, an embodiment of the present invention provides an electronic device 300, including a memory 310, a processor 320, and a computer program 311 stored in the memory 310 and executable on the processor 320. When the processor 320 executes the computer program 311, it performs the following steps: Based on the first optical image of the top surface of the three-dimensional stacked package sample, the center position of the opening of multiple through silicon vias on the top surface is located, and based on the center position of the opening, a second optical image corresponding to each through silicon via is obtained on the bottom surface of the three-dimensional stacked package sample. The exit center position of each through-silicon via on the bottom surface is determined based on the second optical image, and the axial offset vector of each through-silicon via in the vertical direction is calculated based on the opening center position and the exit center position. Based on the third optical image of the sidewall region of the three-dimensional stacked package sample, the sidewall contour line of each through-silicon via along the depth direction is extracted, and the target spatial trajectory of each through-silicon via is constructed based on the axis offset vector and the sidewall contour line. Based on the target spatial trajectory, determine whether each through-silicon via has any morphological anomalies that deviate from the preset spatial path, and obtain the through-silicon via defect detection results.

[0129] Please see Figure 4 , Figure 4An embodiment diagram of a computer-readable storage medium provided in accordance with an embodiment of the present invention is shown. Figure 4 As shown, this embodiment provides a computer-readable storage medium 400 on which a computer program 311 is stored. When the computer program 311 is executed by a processor, it performs the following steps: Based on the first optical image of the top surface of the three-dimensional stacked package sample, the center position of the opening of multiple through silicon vias on the top surface is located, and based on the center position of the opening, a second optical image corresponding to each through silicon via is obtained on the bottom surface of the three-dimensional stacked package sample. The exit center position of each through-silicon via on the bottom surface is determined based on the second optical image, and the axial offset vector of each through-silicon via in the vertical direction is calculated based on the opening center position and the exit center position. Based on the third optical image of the sidewall region of the three-dimensional stacked package sample, the sidewall contour line of each through-silicon via along the depth direction is extracted, and the target spatial trajectory of each through-silicon via is constructed based on the axis offset vector and the sidewall contour line. Based on the target spatial trajectory, determine whether each through-silicon via has any morphological anomalies that deviate from the preset spatial path, and obtain the through-silicon via defect detection results.

[0130] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to perform the TSV via defect detection method in 3D stacked packaging provided by the above methods. The method includes: Based on the first optical image of the top surface of the three-dimensional stacked package sample, the center position of the opening of multiple through silicon vias on the top surface is located, and based on the center position of the opening, a second optical image corresponding to each through silicon via is obtained on the bottom surface of the three-dimensional stacked package sample. The exit center position of each through-silicon via on the bottom surface is determined based on the second optical image, and the axial offset vector of each through-silicon via in the vertical direction is calculated based on the opening center position and the exit center position. Based on the third optical image of the sidewall region of the three-dimensional stacked package sample, the sidewall contour line of each through-silicon via along the depth direction is extracted, and the target spatial trajectory of each through-silicon via is constructed based on the axis offset vector and the sidewall contour line. Based on the target spatial trajectory, determine whether each through-silicon via has any morphological anomalies that deviate from the preset spatial path, and obtain the through-silicon via defect detection results.

[0131] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0132] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0133] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting TSV via defects in 3D stacked packaging, characterized in that, include: Based on a first optical image of the top surface of the three-dimensional stacked package sample, the center positions of the openings of multiple through-silicon vias on the top surface are located, and based on the opening center positions, a second optical image corresponding to each through-silicon via is obtained on the bottom surface of the three-dimensional stacked package sample. The exit center position of each through-silicon via on the bottom surface is determined based on the second optical image, and the axial offset vector of each through-silicon via in the vertical direction is calculated based on the opening center position and the exit center position. Based on the third optical image of the sidewall region of the three-dimensional stacked packaged sample, the sidewall contour line of each through-silicon via along the depth direction is extracted, and the target spatial trajectory of each through-silicon via is constructed based on the axis offset vector and the sidewall contour line. Based on the target spatial trajectory, determine whether each through-silicon via has any morphological anomalies that deviate from the preset spatial path, and obtain the through-silicon via defect detection results.

2. The method for detecting TSV via defects in 3D stacked packaging according to claim 1, characterized in that, The step of constructing the target spatial trajectory for each through-silicon via based on the axis offset vector and the sidewall contour line includes: In a preset three-dimensional space, a straight line segment connecting the opening center position indicated by the axis offset vector and the exit center position is constructed to obtain the initial axial direction. Starting from the top depth along the depth direction, based on the set of projection points of the contour points in the sidewall contour line of the current depth layer on the plane perpendicular to the initial axial direction and the initial axial direction, the first candidate trajectory point of the current depth layer is determined. Based on the first candidate trajectory point of the current depth layer and the second candidate trajectory point of the previous depth layer, the current axial direction is determined, and based on the distribution of the contour points of the current depth layer on the plane perpendicular to the current axial direction, the current principal normal direction is determined. The current subnormal direction is obtained by cross product of the current axial direction and the current principal normal direction. The target spatial trajectory is constructed based on the difference between the first maximum spacing of the contour points of the current depth layer in the current principal normal direction and the second maximum spacing in the current subnormal direction.

3. The method for detecting TSV via defects in 3D stacked packaging according to claim 2, characterized in that, The target spatial trajectory is constructed based on the distance difference between the first maximum distance between the contour points of the current depth layer in the current principal normal direction and the second maximum distance in the current secondary normal direction, including: Based on the spacing difference, the first candidate trajectory point of the current depth layer is shifted by a preset offset distance along the current principal normal direction to the side with smaller spacing to obtain the current corrected trajectory point; The corrected trajectory points of each depth layer from the top depth to the bottom depth are determined in sequence to obtain a preliminary spatial trajectory composed of discrete trajectory points. Based on the spatial vector between the first point of the preliminary spatial trajectory and the center position of the opening, the preliminary spatial trajectory is translated so that the first point after translation coincides with the center position of the opening, thus obtaining the translated spatial trajectory. Based on the positional deviation between the end point of the translated spatial trajectory and the launch center position, the minimum rotation angle around the center position of the opening is calculated so that the rotated end point coincides with the launch center position, and the minimum rotation angle is applied to the translated spatial trajectory to obtain the aligned spatial trajectory point sequence. Based on the aligned spatial trajectory point sequence, adjacent trajectory points are connected sequentially to form polyline segments, and a transition point is inserted at the connection of every two adjacent polyline segments to obtain the target spatial trajectory; the transition point is located on the angle bisector of the plane formed by the two consecutive polyline segments, and the distance to the connecting vertex is equal to one-tenth of the harmonic average of the lengths of the two consecutive segments.

4. The method for detecting TSV via defects in 3D stacked packaging according to claim 2, characterized in that, The steps for determining the first candidate trajectory point in the current depth layer include: The midpoint of the line connecting the projection points of the contour points of the current depth layer with the farthest target projection point is taken as the top local center point. The direction of the line is taken as the initial principal normal direction. The initial axial direction is cross-multiplied with the initial principal normal direction to obtain the initial secondary normal direction. Project each contour point of the current depth layer along the initial principal normal direction and the initial sub-normal direction to obtain the principal normal projection extreme point and the sub-normal projection extreme point; The first candidate trajectory point of the current depth layer is determined based on the average position of the first midpoint between the principal normal projection extrema and the second midpoint between the secondary normal projection extrema.

5. The method for detecting TSV via defects in 3D stacked packaging according to claim 1, characterized in that, The steps for determining the detection results of the through-silicon via defects include: Based on the spatial trajectory points arranged in depth order in the target spatial trajectory and the reference points at corresponding depth positions in the preset spatial path, the position deviation is calculated layer by layer to obtain the three-dimensional offset vector between the spatial trajectory point and the reference point at each depth position. Vector projection decomposition is performed based on the 3D offset vector at each depth position and the tangent direction of the preset spatial path at the corresponding depth position to obtain the axial offset component along the tangent direction of the preset spatial path and the lateral offset component perpendicular to the tangent direction. The modulus is calculated based on the lateral offset vector at each depth position to obtain the sequence of lateral offset components. The axial offset components at each depth position and all axial offset components from the top depth to the current depth are summed to obtain the sequence of cumulative axial offset values. Based on the lateral offset component sequence and the axial offset cumulative value sequence, morphological anomaly detection is performed on each through-silicon via to obtain the through-silicon via defect detection results.

6. The method for detecting TSV via defects in 3D stacked packaging according to claim 5, characterized in that, The method of performing morphology anomaly detection on each through-silicon via (TSV) based on the lateral offset component sequence and the axial offset cumulative value sequence to obtain TSV defect detection results includes: The first-order difference of the cumulative axial offset value is calculated based on the cumulative axial offset value sequence and the depth spacing between adjacent depth positions to obtain the cumulative axial offset rate sequence. Based on the lateral offset component sequence, the set of depth locations where the lateral offset component value exceeds the preset lateral anomaly threshold is determined as the lateral over-limit region. Based on the axial offset cumulative change rate sequence, the continuous depth interval where the axial offset cumulative change rate has the same sign and a value greater than zero at three or more consecutive depth locations is determined as the axial continuous offset region. Based on the lateral over-limit region and the axial continuous offset region, a region intersection operation is performed, and the depth interval belonging to the lateral over-limit region and the axial continuous offset region is determined as the composite deviation region. Based on the lateral offset component values ​​corresponding to each depth position and the two consecutive depth positions before and after it in the composite deviation region, morphological anomaly detection is performed to obtain the through-silicon via defect detection results.

7. The method for detecting TSV via defects in 3D stacked packaging according to claim 6, characterized in that, Based on the lateral offset component values ​​corresponding to each depth position and the two consecutive depth positions before and after it in the composite deviation region, morphological anomaly detection is performed to obtain the through-silicon via defect detection results, including: The extreme difference is calculated based on the lateral offset component values ​​corresponding to each depth position and the two consecutive depth positions before and after it in the composite deviation region to obtain the local lateral fluctuation amplitude. Based on the geometric tolerance comparison of the radius of curvature and the amplitude of local lateral fluctuation in the corresponding depth range of the composite deviation region according to the preset spatial path, if the amplitude of local lateral fluctuation is greater than one-tenth of the radius of curvature, the composite deviation region is determined as the morphological distortion range. If the cumulative value of axial offset increases or decreases monotonically within the morphology distortion range, then the detection result of the through-silicon via defect is determined to be a tilting anomaly caused by etching depth control failure. If the cumulative value of axial offset changes less than the preset axial stability threshold within the morphology distortion range and the value of lateral offset component fluctuates alternately, then the detection result of the through-silicon via defect is determined to be a serpentine anomaly caused by uneven stress release on the sidewall.

8. A device for detecting TSV via defects in 3D stacked packaging, characterized in that, Applied to the TSV via defect detection method in 3D stacked packaging as described in any one of claims 1 to 7; The TSV via defect detection device in the 3D stacked packaging includes: An optical image acquisition module is used to locate the opening center position of multiple through-silicon vias on the top surface of the three-dimensional stacked package sample based on a first optical image of the top surface of the sample, and to acquire a second optical image corresponding to each through-silicon via on the bottom surface of the sample based on the opening center position. The axis offset module is used to determine the emission center position of each through-silicon via on the bottom surface based on the second optical image, and to calculate the axis offset vector of each through-silicon via in the vertical direction based on the opening center position and the emission center position. The spatial trajectory construction module is used to extract the sidewall contour line of each through-silicon via along the depth direction based on the third optical image of the sidewall region of the three-dimensional stacked package sample, and to construct the target spatial trajectory of each through-silicon via based on the axis offset vector and the sidewall contour line. The defect detection module is used to determine whether each through-silicon via has an abnormal shape that deviates from the preset spatial path based on the target spatial trajectory, and to obtain the through-silicon via defect detection result.

9. An electronic device, comprising: Memory, used to store computer software programs; A processor for reading and executing the computer software program, characterized in that, when the processor executes the computer software program, it implements the TSV via defect detection method in 3D stacked packaging as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium, wherein a computer software program is stored therein, characterized in that, When the computer software program is executed by the processor, it implements the TSV via defect detection method in 3D stacked packaging as described in any one of claims 1 to 7.

Images