Wafer defect detection method and wafer defect detection system

Abstract

The application provides a wafer defect detection method and system based on multi-Die inter-image similarity matching, the wafer defect detection method comprises the following steps: establishing a Die position index table, scanning to obtain original optical images of the Die, defining a to-be-tested image and a reference image, defining a neighborhood range size, constructing a candidate reference image set, constructing a similarity measurement set, defining the number of reference images in a Job, obtaining a final reference object, outputting a defect detection result of the Job, traversing all the Die on the wafer and outputting a complete defect detection result. The application dynamically evaluates multiple Die scanned on a straight line, adaptively selects optimal reference images to construct a detection Job, provides higher-quality reference images for differential calculation, reduces noise signals caused by non-defect differences on a differential image, improves the signal-to-noise ratio of real defects in a scanning machine, the accuracy and stability of detection, and facilitates popularization and application in the field of semiconductor optical detection.

Description

Technical Field

[0001] This invention belongs to the field of semiconductor inspection and relates to a wafer defect detection method and system based on similarity matching between multiple die images. In particular, it relates to a method for adaptively constructing a reference image detection group (Job) in the wafer defect detection process to improve the accuracy and stability of differential detection results. Background Technology

[0002] In wafer defect detection, current mainstream scanning equipment employs a line-by-line scanning strategy, typically using a die-to-die comparison approach. The die-to-die method constructs a job by comparing the target die in the same row with adjacent reference dies during scanning. Through image processing, differential calculation, and threshold filtering, it identifies potential defects within the job. This method is suitable for ideal conditions where the target die and adjacent reference dies on the wafer exhibit high consistency in brightness and structure, and the same area on adjacent reference dies is defect-free. Therefore, the reference dies serve as effective references, and the differential calculation results accurately reflect the pattern differences between the compared dies, thus efficiently detecting true defects.

[0003] In actual production or inspection processes, due to factors such as wafer manufacturing processes or equipment conditions, non-defect differences often exist between dies in the same row on a wafer, such as uneven brightness, illumination shift, focal length differences, or pattern distortion. This leads to certain differences in grayscale distribution and texture features between adjacent dies. When using traditional die-to-die comparison strategies, these non-defect differences severely interfere with the quality of the difference map, resulting in a large amount of noise signals. Noise signals significantly affect the signal-to-noise ratio calculation of true defects, causing a large number of noise signals to be misclassified as defects while detecting true defects, thus generating a large number of nuisance defects and seriously affecting the accuracy and reliability of inspection equipment. Furthermore, although signal noise can be processed at the subsequent algorithm level using image processing and thresholding to obtain an effective distribution of true defects, when the noise signal is strong or the true defect signal is weak, image processing methods can cause the true defect signal to be covered by the distribution of noise signals. In this case, thresholding methods are unlikely to truly achieve the purpose of filtering noise and retaining true defects.

[0004] The die-to-die method constructs jobs based on adjacent positions. However, in actual production testing, adjacent dies are not necessarily the optimal reference objects. Therefore, the strategy of always selecting dies to construct jobs is difficult to adaptively handle the aforementioned anomalies. Summary of the Invention

[0005] To overcome the shortcomings of the prior art, the present invention aims to provide a wafer defect detection method and a wafer defect detection system based on similarity matching between multiple die images, so as to solve the problems existing in the above-mentioned prior art.

[0006] Design Principle: This wafer defect detection method is based on similarity matching among multiple die images. It selects the optimal reference object among multiple dies to form a Job, reducing the difference between the reference image and the image to be tested within the Job, thereby improving the accuracy and stability of the detection algorithm. The overall design scheme is as follows.

[0007] A wafer defect detection method based on similarity matching between multiple die images, the wafer defect detection method includes the following steps.

[0008] S1. Obtain wafer design layout data, define the position distribution of dies on the wafer, and establish an aligned die position index table.

[0009] S2. Based on the Die position index table, perform the scanning step to obtain the original optical images of all dies along the same straight line according to the Die position index table.

[0010] S3. Based on the original optical image, define the original optical image to be detected in the current detection process as the image to be tested, and the remaining original optical images on the same straight line as the reference image.

[0011] S4. Get the size of the user-defined neighborhood range.

[0012] S5. Based on the image to be tested, the reference image, and the size of the neighborhood range, construct a corresponding set of candidate reference images.

[0013] S6. Based on the candidate reference image set, calculate the similarity metric between the image to be tested and the reference image respectively, and construct a similarity metric set.

[0014] S7. Get the number of reference images in the user-defined Job.

[0015] S8. Based on the similarity metric set, the candidate reference image set, and the number of reference images in the Job, obtain one or more of the most similar reference images as the final reference objects.

[0016] S9. Based on the final reference object, construct the detection job, perform image processing and differential calculation, and output the defect detection results of the job.

[0017] S10. Traverse all dies on the wafer, repeat the wafer scanning and job construction process until the defect detection results of all dies are output, and finally output the complete defect detection results of the wafer.

[0018] Furthermore, the process of establishing a die location index table based on wafer design layout data includes: S11, defining the position distribution of dies on the wafer to be inspected in the scanning machine software program based on wafer design layout data; S12, obtaining the wafer coordinate system based on wafer design layout data; S13, establishing a numbered die location index table based on die location distribution and wafer coordinate system.

[0019] Preferably, when defining the size of the neighborhood range, the size of any neighborhood range is not less than 2.

[0020] Preferably, when defining the size of the neighborhood range, the size of any neighborhood range shall not exceed half of the maximum number of dies in the row where the current die is located.

[0021] Furthermore, the calculation process of the similarity measure set includes: S61, calculating the similarity measure between the image to be tested and each candidate reference image based on the image to be tested and the candidate reference image set; the similarity measure adopts one or more of the following: gray-level histogram correlation coefficient, structural similarity index or normalized cross-correlation coefficient; S62, constructing the similarity measure set based on the calculated one or more similarity measures.

[0022] Preferably, when defining the number of reference images within a Job, the number of reference images within a Job does not exceed 2.

[0023] Furthermore, the defect detection result calculation process of the detection job includes: S91, based on the detection job, performing sub-pixel level registration on the image to be tested and the reference image to obtain the registered image to be tested and the reference image; S92, based on the registered image to be tested and the reference image, performing difference operation to generate a difference map; S93, based on the difference map, performing a user-defined threshold strategy to output the defect detection result of the job.

[0024] Preferably, in the user-defined threshold strategy, the threshold is defined as T, and the final defect determination is defined as... : In the formula, Indicating the defect arbitration process, Mark potential defects on the difference map. Defined as: Among them, difference operations , This represents the image to be tested corresponding to the Die with index number k. This represents the set of reference images after subpixel alignment.

[0025] Furthermore, the method supports real-time online detection, meaning that data is collected and matched simultaneously during the scanning process, and similarity calculation and difference analysis can be started without waiting for the entire row to be scanned.

[0026] The present invention also provides a wafer defect detection system, including a wafer stage, an image acquisition module and a processor. The system controls the movement of the wafer stage, the scanning and acquisition by the image acquisition module and the image processing and defect result output by the processor based on the aforementioned wafer defect detection method.

[0027] Compared with the prior art, the beneficial effects of the present invention are as follows: The wafer defect detection method of this application dynamically evaluates multiple dies scanned under the same straight line, adaptively selects the optimal reference image to construct the detection job, provides a higher quality reference image for subsequent job differential calculation, effectively reduces the noise signal caused by non-defect differences on the differential image, improves the signal-to-noise ratio of real defects in the scanning machine, improves the accuracy and stability of wafer detection, and facilitates its application in the fields of semiconductor optical inspection and image processing. Attached Figure Description

[0028] Figure 1 This is a flowchart of the wafer defect detection method based on similarity matching between multiple die images according to the present invention.

[0029] Figure 2 This is a schematic diagram of a job constructed using a lateral scan of the wafer and a local row layout.

[0030] Figure 3 This is a diagram illustrating defect arbitration. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, 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.

[0032] It should be understood that the terms "system," "device," "unit," and / or "module" as used in this specification are a method of distinguishing different components, elements, parts, sections, or assemblies at different levels. However, if other terms can achieve the same purpose, they may be replaced by other expressions.

[0033] Flowcharts are used in this specification to illustrate the operations performed by the system according to embodiments of this specification. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, the steps can be processed in reverse order or simultaneously. Furthermore, other operations can be added to these processes, or one or more steps can be removed from them.

[0034] See Figure 1 A wafer defect detection method based on similarity matching between multiple die images includes the following steps.

[0035] Step 1: Define the die coordinate system as a Cartesian coordinate system. Die in coordinate system The width and height of the middle are respectively and Based on the wafer coordinate system P(x,y) and width and height, a die position index table Index is established.

[0036] ;in, The outer frame of the Die rectangle in the coordinate system The coordinates of the top left corner in the middle, It represents the total number of dies on the current wafer.

[0037] Step 2: Regarding the coordinate system in Step 1 and Die position index table See Figure 2 Perform a row-line defect scan operation, sequentially acquiring the original optical images of each die along the same straight line direction, and obtaining the set of position indices L of the die in that row.

[0038] ;in It represents the number of dies contained in the current row, and the original optical image set corresponding to the dies obtained from the scan. .

[0039] Among them, and because the image resolution of a single scan is [missing information] Pixels, therefore each scan only covers a local area within the die, so each image This is a partial scan result of the die.

[0040] Step 3: For the Die position index set in Step 2 Establish a global coordinate system for the wafer The mapping relationship with the local coordinate system within the die is defined as follows: Then the coordinates of any point within the Die satisfy the following relationship.

[0041] ;in .

[0042] Step 4: For the original optical image set from Step 2 Compared with the local coordinate system within the Die described in step three Transform the relationship to extract the original optical image set of the same region within the die. , .

[0043] Step 5: For the set of location images of the same region obtained in Step 4 Define the size of the neighborhood range of user input as ,in This indicates the radius range of candidate reference dies selected to the left and right of the current die. Specifically, if the index of the current target die is... Define the corresponding image as the image to be tested. Construct a set of candidate reference images .

[0044] Where the index exceeds the set range, This invention employs a circular indexing method to supplement the reference image from the other side. .

[0045] , .

[0046] Step Six: For the image to be tested in Step Five With candidate reference image set Calculate the combination of similarity metrics between the image to be tested and the candidate reference image. .

[0047] ;in These are weighting coefficients, satisfying... . These are the gray-scale histogram correlation coefficient, structural similarity index, and normalized cross-correlation coefficient, respectively, and are defined as follows.

[0048] ;

[0049] ;

[0050] .

[0051] in, It is the gray-level histogram of the image at the th The frequency of each gray level, where L is the total number of gray levels. The mean of the image. Standard deviation For covariance, It is the stability constant.

[0052] Step 7: For the test image from Step 5, the candidate reference image set, and the similarity metric value combination obtained in Step 6. ,definition The number of internal reference images is Construct a detection task group .

[0053] ;in, A set of similarity measures Center front The set of reference images with the highest similarity .

[0054] The definition of is: .

[0055] Step 8: Target the detection task group obtained in Step 7. ,right Perform subpixel alignment on the reference image to obtain the aligned image detection group. .

[0056] ;in It is the set of reference images after subpixel alignment. .

[0057] .

[0058] Step 9: Detect the aligned image obtained in Step 8. Perform the difference operation to obtain the difference graph. , .

[0059] Step 10: For the difference image obtained in Step 9, execute the defined thresholding strategy, where the threshold is defined as follows: The final defect determination is defined as follows: .

[0060] ; see also Figure 3 , Indicating the defect arbitration process, This marks potential defects on the difference map. The definition is as follows.

[0061]

[0062] Step 11: For the defect detection results of a single detection task group obtained in Step 10, perform a full wafer-wide traversal and result summary operation to generate a complete wafer defect detection output.

[0063] To improve the detection sensitivity of the scanning equipment, the neighborhood size should not be less than 2 when defining the neighborhood range. This is because if the neighborhood size is 1, the method will degenerate into the original die-to-die method, resulting in inaccurate detection results.

[0064] To reduce the time required for wafer scanning and computation, when constructing the candidate reference image set, since the height of the high-speed camera is insufficient to cover the entire die height, this invention only considers images acquired along the same straight line during line-by-line scanning as candidates. Therefore, after each line-by-line scan, the system in the scanning machine begins to calculate and construct the detection task group for differential analysis. This allows for simultaneous acquisition and matching, eliminating the need to wait for the scanning results of the next line of the die.

[0065] The adaptive wafer defect detection method based on image similarity of this invention can adaptively select the optimal reference images within the wafer range to form a detection task group and perform dynamic differential detection. This significantly reduces the interference of noise signals caused by non-defect differences on the differential image on the real defect signal, while improving detection sensitivity and signal-to-noise ratio. It is suitable for complex situations such as uneven brightness, illumination shift, focal length differences, or pattern distortion, bringing more accurate and stable results to wafer inspection.

[0066] 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 wafer defect detection method based on similarity matching between multiple die images, characterized in that, The wafer defect detection method includes the following steps: S1. Obtain wafer design layout data, define the position distribution of dies on the wafer, and establish an aligned die position index table. S2. Based on the Die position index table, perform the scanning step to obtain the original optical images of all dies on the same straight line according to the Die position index table; S3. Based on the original optical image, define the original optical image to be detected in the current detection process as the image to be tested, and the remaining original optical images on the same straight line as the reference image; S4. Obtain the size of the user-defined neighborhood range; S5. Based on the image to be tested, the reference image, and the size of the neighborhood range, construct a corresponding set of candidate reference images; S6. Based on the candidate reference image set, calculate the similarity metric between the image to be tested and the reference image respectively, and construct a similarity metric set; S7. Obtain the number of reference images within the user-defined Job; S8. Based on the similarity metric set, the candidate reference image set, and the number of reference images in the Job, obtain one or more of the most similar reference images as the final reference objects; S9. Based on the final reference object, construct the detection job, perform image processing and differential calculation, and output the defect detection results of the job; S10. Traverse all dies on the wafer, repeat the wafer scanning and job construction process until the defect detection results of all dies are output, and finally output the complete defect detection results of the wafer.

2. The wafer defect detection method according to claim 1, characterized in that, The process of building a die location index table based on wafer design layout data includes: S11. Based on the wafer design layout data, define the position distribution of the die on the wafer to be inspected in the software program of the scanning machine; S12. Obtain the wafer coordinate system based on the wafer design layout data; S13. Based on the die location distribution and wafer coordinate system, establish a numbered die location index table.

3. The wafer defect detection method according to claim 1, characterized in that: When defining the size of a neighborhood, the size of any neighborhood range shall not be less than 2.

4. The wafer defect detection method according to claim 1, characterized in that: When defining the size of a neighborhood range, the size of any neighborhood range shall not exceed half of the maximum number of dies in the row where the current die is located.

5. The wafer defect detection method according to claim 1, characterized in that, The calculation process for the similarity measure set includes: S61. Based on the image to be tested and the set of candidate reference images, calculate the similarity measure between the image to be tested and each candidate reference image; the similarity measure can be one or more of the following: gray-level histogram correlation coefficient, structural similarity index, or normalized cross-correlation coefficient. S62. Construct a set of similarity measures based on one or more calculated similarity measures.

6. The wafer defect detection method according to claim 1, characterized in that: When defining the number of reference images within a Job, the number of reference images within a Job shall not exceed 2.

7. The wafer defect detection method according to claim 1, characterized in that: The calculation process for the defect detection results of the job includes: S91. Based on the detection job, perform sub-pixel-level registration between the image to be tested and the reference image to obtain the registered image to be tested and the reference image. S92. Based on the registered image to be tested and the reference image, perform difference operation to generate a difference map; S93. Based on the difference graph, execute the user-defined threshold strategy and output the defect detection results of the Job.

8. The wafer defect detection method according to claim 7, characterized in that: In the user-defined threshold strategy, the threshold is defined as T, and the final defect determination is defined as follows: : In the formula, Indicating the defect arbitration process, Mark potential defects on the difference map. Defined as: Among them, difference operations , This represents the image to be tested corresponding to the Die with index number k. This represents the set of reference images after subpixel alignment.

9. The wafer defect detection method according to claim 1, characterized in that: The method supports real-time online detection, meaning that data is collected and matched simultaneously during the scanning process, and similarity calculation and difference analysis can be started without waiting for the entire row to be scanned.

10. A wafer defect detection system, comprising a wafer motion stage, an image acquisition module, and a processor, characterized in that: The wafer defect detection system controls the movement of the wafer stage, the scanning and acquisition of the image acquisition module, and the image processing and defect result output of the processor based on the wafer defect detection method according to any one of claims 1-9.

Images