CN-122023292-A - Method and device for determining wafer bonding defect

CN122023292ACN 122023292 ACN122023292 ACN 122023292ACN-122023292-A

Abstract

The application provides a method and a device for determining bonding defects of wafers, wherein the method comprises the steps of obtaining an initial scanning image after bonding of the wafers, carrying out contour calibration segmentation on the initial scanning image based on wafer layout data to generate single-wafer images, carrying out image feature recognition on the single-wafer images to determine defect positioning information of bubble defects under a local coordinate system of the wafers, respectively calling first defect data of an upper wafer before bonding and second defect data of a lower wafer before bonding, which have a physical corresponding relation with the single-wafer images, based on a wafer bonding mapping relation, determining spatial distance parameters of defect positioning information and defect points in the first defect data and the second defect data, carrying out quantization calculation on the spatial distance parameters to respectively generate upper-layer influence scores and lower-layer influence scores, and determining source types of the bubble defects based on the upper-layer influence scores and the lower-layer influence scores. The application can effectively improve the accuracy of defect root cause analysis.

Inventors

LIU WENHUA
XIE DONG
HU ZHIPENG
DU HAO

Assignees

湖北星辰技术有限公司

Dates

Publication Date: 20260512
Application Date: 20260115

Claims (10)

1. A method for determining a wafer bonding defect, the method comprising: acquiring an initial scanning image after wafer bonding, and carrying out contour calibration segmentation on the initial scanning image based on wafer layout data to generate a single-grain image; based on a grain bonding mapping relation, respectively calling first defect data of an upper wafer before bonding and second defect data of a lower wafer before bonding, wherein the first defect data and the second defect data have a physical correspondence relation with the single-grain image; Determining space interval parameters of the defect positioning information and each defect point in the first defect data and the second defect data respectively; performing quantitative calculation on the space interval parameters to respectively generate an upper-layer influence score and a lower-layer influence score; determining a source category of the bubble defect based on the upper layer impact score and the lower layer impact score.
2. The method of claim 1, wherein contour calibration segmentation of the initial scan image to generate a single die image comprises: performing high-pass filtering treatment on the gray level image to obtain a filtered image; executing differential operation of the gray level image and the filtering image to obtain a differential operation result, and determining a physical contour line of the crystal grain based on the differential operation result; Determining a theoretical contour line based on the wafer layout data, and performing position matching on the physical contour line and the theoretical contour line to obtain a matching result; determining the offset of the physical contour line relative to the theoretical contour line based on the matching result, and correcting a cutting path based on the offset to obtain a corrected cutting path; And based on the corrected cutting path, cutting out the initial scanning image to generate the single-crystal grain image.
3. The method according to claim 2, wherein performing the difference operation between the gray scale image and the filtered image results in a difference operation result, comprising: subtracting the gray value of the pixel point at the corresponding position in the filtering image from the gray value of each pixel point in the gray image to obtain the difference operation result; the determining the physical contour line of the crystal grain based on the differential operation result comprises the following steps: And generating a difference image after noise interference is eliminated by using the difference operation result, performing binarization processing on the difference image, and extracting the physical contour line of the crystal grain from the image after binarization processing.
4. The method of claim 1, wherein the performing image feature recognition on the single die image to determine defect localization information of the bubble defect in a die local coordinate system comprises: Calculating the background pixel ratio of the single-grain image, and selecting a single-grain image with the background pixel ratio lower than a preset proportion as an effective grain image; Scanning the effective grain image by adopting a self-adaptive contrast detection algorithm, identifying a pixel gray scale mutation region from the effective grain image, and extracting an edge closed path of the mutation region; And determining the geometric center of the edge closed path, and mapping the geometric center to the local coordinate system of the crystal grain to obtain the defect positioning information.
5. The method of claim 1, wherein the quantitatively calculating is performed by a stress field attenuation model, the stress field attenuation model including a preset effective influence distance threshold, and wherein the quantitatively calculating the spatial separation parameter generates an upper influence score and a lower influence score, respectively, comprising: Substituting the spatial interval parameter corresponding to the first defect data into the stress field attenuation model, calculating to obtain a first influence score, and determining the first influence score as the upper layer influence score; substituting the spatial interval parameter corresponding to the second defect data into the stress field attenuation model, calculating to obtain a second influence score, and determining the second influence score as the lower layer influence score; The calculation logic of the stress field attenuation model meets the requirements that when the substituted space interval parameter is larger than the effective influence distance threshold, the output influence score is zero, and when the substituted space interval parameter is smaller than or equal to the effective influence distance threshold, the nonlinear mapping operation based on the power function logic is executed, and the influence score which increases with the decrease of the substituted space interval parameter is output.
6. The method of claim 5, wherein performing a power function logic based nonlinear mapping operation outputs an impact score that increases as the substituted spatial pitch parameter decreases, comprising: Calculating the ratio of the space interval parameter to the effective influence distance threshold; Performing power operation based on a first nonlinear coefficient on the ratio to obtain an intermediate ratio, and calculating a difference value between a unit numerical value and the intermediate ratio to obtain a basic retention factor; And performing power operation based on a second nonlinear coefficient on the basic retention factor to obtain the influence score, wherein the first nonlinear coefficient and the second nonlinear coefficient are used for fitting the stress attenuation curvature of the material.
7. The method of claim 5, wherein the first defect data or the second defect data comprises a plurality of defect points; The determining the first impact score as the upper layer impact score comprises: Independently calculating the corresponding first influence scores for each defect point in the first defect data, executing accumulation operation on all the first influence scores, and taking an accumulation result as the upper-layer influence score; the determining the second impact score as the lower layer impact score comprises: and independently calculating the corresponding second influence scores for each defect point in the second defect data, executing accumulation operation on all the second influence scores, and taking the accumulation result as the lower influence score.
8. The method of claim 1, wherein the determining the source category of the bubble defect based on the upper layer impact score and the lower layer impact score comprises: if the lower-layer influence score is larger than a preset effective judgment score and the difference value between the lower-layer influence score and the upper-layer influence score exceeds a preset amplitude, judging that the source category of the bubble defect is lower-layer wafer dominant; And if the upper layer influence score is larger than the effective judgment score and the difference value between the upper layer influence score and the lower layer influence score exceeds the preset amplitude, judging that the source category of the bubble defect is the upper layer wafer dominant.
9. The method of claim 8, wherein the determining the source category of the bubble defect further comprises: if the upper layer influence score and the lower layer influence score are both larger than the effective judgment score and the difference value does not exceed the preset amplitude, judging that the source type of the bubble defect is double-side wafer collaborative triggering; And if the upper layer influence score and the lower layer influence score are smaller than the effective judgment score, judging that the source type of the bubble defect is caused by external process environment factors.
10. A wafer bonding defect determining apparatus, the apparatus comprising: The identification module is used for acquiring an initial scanning image after wafer bonding, carrying out contour calibration segmentation on the initial scanning image based on wafer layout data to generate a single-grain image; The transferring module is used for respectively transferring first defect data of an upper wafer before bonding and second defect data of a lower wafer before bonding, which have a physical corresponding relation with the single-crystal-grain image, based on the crystal grain bonding mapping relation; The calculation module is used for determining the space interval parameters of the defect positioning information and each defect point in the first defect data and the second defect data respectively; And the determining module is used for determining the source category of the bubble defect based on the upper layer influence score and the lower layer influence score.

Description

Method and device for determining wafer bonding defect Technical Field The present application relates to the field of computer technologies, and in particular, to a method and an apparatus for determining a wafer bonding defect. Background In the wafer bonding process of advanced packaging of semiconductors, interface bubble defects are key factors influencing the yield of products, and currently, the industry mainly relies on an acoustic scanning microscope to acquire bonded static images for defect inspection. Although the detection mode can intuitively present the final defect distribution form, only single result data after bonding is finished is provided, and the surface states of wafers at each layer before bonding cannot be directly related, so that engineering personnel are difficult to directly observe the root cause of bubble generation through a single Zhang Saomiao image. In the prior art, a remarkable data fault exists in the aspect of defect tracing, and because an accurate coordinate mapping mechanism is lacking between a scanning image after bonding and wafer surface detection data before bonding, and a scientific model capable of quantitatively evaluating the influence weight of defects before bonding on bubble formation is lacking, it is difficult for engineering personnel to accurately distinguish whether bubbles originate from upper wafer incoming materials, lower wafer incoming materials or bonding process environments. The lack of the physical association and quantitative analysis means makes the defect root cause analysis depend on manual experience for a long time to carry out fuzzy speculation, so that the defect root cause analysis has poor accuracy. Disclosure of Invention The embodiment of the application provides a method, a device, electronic equipment, a computer readable storage medium and a computer program product for determining wafer bonding defects, which can effectively improve the accuracy of defect root cause analysis. The technical scheme of the embodiment of the application is realized as follows: The embodiment of the application provides a method for determining wafer bonding defects, which comprises the following steps: acquiring an initial scanning image after wafer bonding, and carrying out contour calibration segmentation on the initial scanning image based on wafer layout data to generate a single-grain image; based on a grain bonding mapping relation, respectively calling first defect data of an upper wafer before bonding and second defect data of a lower wafer before bonding, wherein the first defect data and the second defect data have a physical correspondence relation with the single-grain image; Determining space interval parameters of the defect positioning information and each defect point in the first defect data and the second defect data respectively; performing quantitative calculation on the space interval parameters to respectively generate an upper-layer influence score and a lower-layer influence score; determining a source category of the bubble defect based on the upper layer impact score and the lower layer impact score. The embodiment of the application provides a device for determining wafer bonding defects, which comprises the following steps: The identification module is used for acquiring an initial scanning image after wafer bonding, carrying out contour calibration segmentation on the initial scanning image based on wafer layout data to generate a single-grain image; The transferring module is used for respectively transferring first defect data of an upper wafer before bonding and second defect data of a lower wafer before bonding, which have a physical corresponding relation with the single-crystal-grain image, based on the crystal grain bonding mapping relation; The calculation module is used for determining the space interval parameters of the defect positioning information and each defect point in the first defect data and the second defect data respectively; And the determining module is used for determining the source category of the bubble defect based on the upper layer influence score and the lower layer influence score. An embodiment of the present application provides an electronic device, including: A memory for storing computer executable instructions or computer programs; And the processor is used for realizing the method for determining the wafer bonding defect when executing the computer executable instructions or the computer programs stored in the memory. The embodiment of the application provides a computer readable storage medium, which stores computer executable instructions or a computer program for realizing the method for determining the wafer bonding defect provided by the embodiment of the application when the processor is caused to execute. Embodiments of the present application provide a computer program product comprising a computer program or computer-executable instructions stored in a computer-readab