On-line optical inspection system and method for wafer backside edge local anomalies

Abstract

The present application relates to a kind of wafer backside edge local anomaly on-line optical detection system and method, semiconductor wafer detection technical field belongs to, including image acquisition module, wafer transmission and rotation module, image processing module, abnormality determination module, alarm and sorting module and data storage module.System is through low angle annular light source cooperation line sweep camera, carries out 360 ° full circle scanning imaging to wafer backside edge;Image processing module extracts chamfer boundary, calculates chamfer radial width and local gray scale feature;Abnormality determination module is based on the double determination logic of width mutation and gray scale deviation, identifies geometric size anomaly and reflection feature anomaly.Can effectively detect the increase of chamfer width and SiO2 film damage caused by grinding wheel error in end face grinding process, realizes non-contact, high-precision on-line full circle detection, improves wafer sorting efficiency and manufacturing yield.

Description

Technical Field

[0001] This invention relates to the field of semiconductor wafer inspection technology, and specifically to an online optical inspection system and method for local anomalies on the back edge of a wafer. Background Technology

[0002] With the continuous development of semiconductor manufacturing processes, wafer diameters are constantly increasing and process nodes are shrinking, leading to increasingly stringent requirements for wafer edge quality. The condition of the wafer edges directly affects the yield of subsequent processes such as photolithography, resist coating, and etching. If there is damage or contamination at the edges, it can easily lead to wafer breakage, uneven resist peeling, or abnormal edge exposure, ultimately resulting in chip scrap.

[0003] In the edge polishing process of wafer manufacturing, it is usually necessary to chamfer the wafer edges to remove sharp corners. However, in actual production, if the polishing equipment has precision deviations or mechanical failures, the polishing wheel is very likely to accidentally contact the SiO2 film layer on the back of the wafer (especially in composite structures containing both a POLY layer and a SiO2 film layer). This accidental contact can not only cause microscopic changes in the radial width of the chamfer on the back of the wafer (such as local widening), but also cause mechanical damage to the SiO2 film layer (such as scratches and film peeling).

[0004] Currently, the main problems with wafer edge detection technologies are as follows: I. Low efficiency and high subjectivity of manual visual inspection: Traditional manual microscope inspection methods are time-consuming and labor-intensive, and are prone to missing minute width changes and film damage, which cannot meet the needs of large-scale mass production.

[0005] II. Blind spots in existing testing equipment: Existing optical testing equipment is mostly focused on pattern defects on the front side of the wafer or macroscopic particle contamination on the back side. It lacks the ability to detect complex defects involving chamfer geometry and film reflection characteristics at the edge of the back side of the wafer (a narrow area of ​​0~5mm).

[0006] Third, single detection methods have significant limitations: conventional edge inspection instruments usually judge defects based on geometric dimensions only, and cannot effectively identify "hidden defects" where the geometric dimensions are not obvious but the film layer is damaged; or they rely solely on appearance imaging, making it difficult to quantitatively assess minute changes in chamfer width. Summary of the Invention

[0007] This invention addresses the shortcomings of existing technologies by providing an online optical inspection system and method for detecting local anomalies at the back edge of wafers. It solves the problem of detecting local anomalies at the back edge (chamfer), particularly the localized increase in chamfer width and film damage caused by accidental contact of the grinding wheel with the SiO2 film during the end-face grinding process. It effectively detects increased chamfer width and SiO2 film damage caused by accidental contact during the end-face grinding process, achieving non-contact, high-precision online full-circumference inspection, thus improving wafer sorting efficiency and manufacturing yield.

[0008] The above-mentioned technical problems of the present invention are mainly solved by the following technical solutions: An online optical inspection system for local anomalies on the back edge of a wafer includes an image acquisition module, a wafer transport and rotation module, an image processing module, an anomaly detection module, an alarm and sorting module, and a data storage module.

[0009] The image acquisition module includes a line scan camera or an area scan camera, and a low-angle ring light source set in conjunction with the camera; the image acquisition module is configured to perform a 360° full-circumference scan of the 0-5mm area on the back edge of the wafer to acquire image information of the back edge.

[0010] The wafer transfer and rotation module includes an automatic feeding robot and a rotating stage. The rotating stage is used to carry and drive the wafer to rotate at a uniform speed to cooperate with the image acquisition module for full-circumference scanning.

[0011] The image processing module is connected to the image acquisition module and has a built-in edge extraction algorithm, width calculation unit and reflection feature analysis unit. The image processing module is configured to receive the image information, extract the chamfer start boundary and end boundary, calculate the chamfer radial width at each angle position, and calculate the local grayscale mean.

[0012] The anomaly detection module is connected to the image processing module and is configured to perform logical judgment on the radial width of the chamfer and the local grayscale mean based on preset width thresholds and grayscale thresholds. If the anomaly conditions are met, an anomaly signal is output.

[0013] The alarm and sorting module is connected to the anomaly determination module and is configured to receive the anomaly signal, perform audible and visual alarms, automatic marking, and unload non-conforming wafers; and the data storage module is configured to save anomaly images, full-circumference width distribution curves, and determination logs.

[0014] Preferably, the incident angle of the low-angle ring light source is in the range of 45° to 70°, which is used to emphasize the wafer edge contour and enhance the reflective characteristics of the SiO2 film surface during imaging.

[0015] Preferably, the resolution of the line scan camera is ≥5μm / pixel, and the positioning accuracy of the rotating stage is ≤0.1°.

[0016] Preferably, the anomaly determination logic in the anomaly determination module is "OR" logic, that is, when the radial width of the chamfer is determined to be abnormal or the reflection feature is determined to be abnormal, the anomaly signal is triggered.

[0017] Preferably, the criteria for determining abnormal chamfer radial width are: calculate the mean value μ of the chamfer radial width across the entire circumference, and if the width W at any position satisfies the relationship: |W-μ| / μ>15%; or the rate of change of chamfer radial width at adjacent angular positions>20% / °.

[0018] Preferably, the criteria for determining the abnormal reflection characteristics are as follows: calculate the mean gray value and standard deviation σ of the entire circumference; if the deviation between the mean gray value of a local location and the mean gray value of the entire circumference is >3σ, then it is determined to be an abnormal reflection.

[0019] An online optical inspection method for local anomalies at the back edge of a wafer includes the following steps: Step S1: Transfer the wafer to be tested to the rotating stage and perform edge positioning.

[0020] Step S2: Start the image acquisition module and low-angle ring light source, control the wafer to rotate 360° at a constant speed, and acquire the full circumference image of the back edge of the wafer.

[0021] Step S3: Preprocess the acquired image, including noise reduction and contrast enhancement.

[0022] Step S4: Extract the chamfer start boundary and end boundary using an edge detection algorithm.

[0023] Step S5: Calculate the radial width of the chamfer at each angle position, generate the full-circumference width distribution curve, and calculate the local grayscale features of the image at the same time.

[0024] Step S6: Determine anomalies based on width and grayscale thresholds.

[0025] Step S7: If a width or reflection abnormality is detected, record the angular coordinates of the abnormal location, trigger an audible and visual alarm, automatically unload the wafer into the non-conforming area, and save the abnormal image; if no abnormality is triggered, the wafer is deemed qualified and flows into the next process.

[0026] Preferably, step S6 specifically includes: Calculate the mean value μ of the radial width of the full circumference chamfer.

[0027] Determine if there are width anomalies: If the width W at any position satisfies |W-μ| / μ>15%, or the width change rate of adjacent angles>20% / °, then it is determined to be a width anomaly.

[0028] To determine if there are any reflection anomalies: Calculate the mean and standard deviation σ of the grayscale value around the entire circumference. If the deviation between the local grayscale mean and the mean grayscale value around the entire circumference is greater than 3σ, then it is determined to be a reflection anomaly.

[0029] Preferably, the method is applicable to wafer structures containing a POLY layer and a SiO2 film layer on top of the POLY layer, and is used to detect localized increases in chamfer width and film damage caused by the grinding wheel accidentally touching the back SiO2 film surface during the end face grinding process.

[0030] Preferably, the chamfer start boundary extracted in step S4 is the boundary line between the back plane of the wafer and the chamfer surface, and the end boundary is the boundary line between the chamfer surface and the edge vertex.

[0031] The present invention can achieve the following effects: This invention provides an online optical inspection system and method for detecting local anomalies at the edge of the back side of a wafer, which has the following advantages compared with the prior art: I. Achieving Dual Composite Detection of Geometric Dimensions and Surface Reflection Features. Breaking the limitations of traditional single-detection methods, this system simultaneously extracts the radial width of the chamfer (geometric dimension) and local grayscale features (optical dimension) through an image processing module. For typical defects caused by accidental contact of the grinding wheel during the end-face grinding process (widened chamfer and damaged film), the system employs an "OR" logic judgment mechanism. An alarm is triggered whenever there is a sudden change in geometric dimensions (such as abnormal width) or a change in surface reflectivity (such as grayscale anomalies caused by film damage). This dual judgment mechanism significantly improves the accuracy of defect identification and avoids missed detections.

[0032] II. Low-angle ring light source imaging is employed to highlight edge defect features. A low-angle ring light source with an incident angle of 45°–70° is selected. According to optical principles, low-angle incident light can effectively emphasize the contours and micro-undulations of an object's surface. For the chamfered contours of the wafer's back edge, this light source can clearly delineate the boundary; for the minute damage on the SiO2 film surface, this light source can enhance its diffuse reflection characteristics, making the defects appear with high contrast in the image, thus ensuring image quality and providing a reliable data foundation for subsequent algorithm processing.

[0033] III. Quantitative Judgment Standards for High Detection Accuracy and Consistency. Unlike subjective judgments based on manual visual inspection, this invention establishes strict quantitative judgment standards. In width anomaly judgment, dual indicators of relative deviation (15%) and rate of change (20% / °) are introduced, enabling the identification of both overall deviations and the capture of local abrupt changes. In reflectance anomaly judgment, statistical principles (3σ criterion) are utilized to accurately eliminate abnormal grayscale points. Combined with a high-resolution line scan camera (≥5μm / pixel) and a high-precision rotating stage (≤0.1°), the system achieves micron-level high-precision detection, ensuring the consistency and traceability of the detection results.

[0034] IV. Fully Automated Online Inspection Significantly Improves Production Efficiency. The system integrates automatic feeding, rotary scanning, data processing, audible and visual alarms, and automatic sorting functions, achieving full automation of the "feeding-inspection-sorting" process. The system can complete a full circumference scan during wafer rotation (360°) without downtime, significantly shortening the inspection cycle and meeting the semiconductor production line's demand for high-throughput, real-time online inspection, effectively reducing production costs. Detailed Implementation

[0035] The technical solution of the invention will be further described in detail below through examples.

[0036] Example: An online optical inspection system for localized anomalies on the back edge of a wafer is mainly used in the semiconductor manufacturing process for non-contact online inspection of localized anomalies on the back edge of a wafer caused by accidental contact by the grinding wheel, etc. The system specifically includes the following modules: I. Image Acquisition Module. The image acquisition module is the "eyes" of the system, primarily responsible for acquiring high-resolution images of the back edge of the wafer. This module includes a line scan camera (or area scan camera) and a low-angle ring light source.

[0037] Camera configuration: A high-resolution line scan camera with a resolution of ≥5μm / pixel is preferred to ensure the ability to capture minute edge defects or film changes. The camera is mounted obliquely above the wafer edge, with a field of view covering an area 0-5mm inward from the back edge of the wafer.

[0038] Light source configuration: A low-angle ring light source is crucial for image quality. The incident angle of the light source is set between 45° and 70°. Low-angle illumination (dark-field illumination principle) is used, with light entering the camera after being reflected from the chamfered surface of the wafer edge. For smooth, mirror-like areas, most of the light is reflected out of the field of view, resulting in a darker image; while for areas with damage, particles, or changes in chamfer contours, light undergoes diffuse reflection into the lens, resulting in a brighter image. This design effectively emphasizes the wafer edge contour and enhances the reflective characteristics of the SiO2 film, improving the contrast between defects and the background.

[0039] II. Wafer Transfer and Rotation Module. This module is used to automate wafer loading, unloading, and rotation scanning operations, including an automatic loading robot and a rotary table.

[0040] Rotary stage: Used to support the wafer and has high-precision rotation control capability, with a positioning accuracy of ≤0.1°. During the inspection process, the rotary stage drives the wafer to rotate 360° at a constant speed, and works in conjunction with a fixed line scan camera to perform a full-circumference scan.

[0041] Automatic loading robot: Used to transfer wafers to be tested from the wafer transfer box or the previous process to the rotary table, and after testing, to sort and unload the wafers.

[0042] III. Image Processing Module. The image processing module, connected to the image acquisition module, is the core "brain" of the system. It includes a built-in edge extraction algorithm, width calculation unit, and reflection feature analysis unit.

[0043] Edge extraction: Receive image information captured by the camera and extract the starting boundary (the boundary line between the back plane of the wafer and the chamfered surface) and the ending boundary (the boundary line between the chamfered surface and the edge vertex) of the chamfer using edge detection algorithms (such as the Canny operator or the Sobel operator).

[0044] Feature calculation: Based on the extracted boundaries, calculate the radial width of the chamfer at each angular position; at the same time, calculate the local gray-scale mean of the image as analysis data for reflection features.

[0045] IV. Anomaly Detection Module. The anomaly detection module is connected to the image processing module and is configured to perform logical judgments based on preset width and grayscale thresholds.

[0046] Judgment Logic: An "OR" logic mechanism is used. That is, an abnormality signal is triggered whenever the radial width of the chamfer or the reflection characteristics are found to be abnormal. This logic can cover changes in geometric dimensions (such as bumps or excessively wide chamfers) and changes in surface condition (such as film scratches or contamination), avoiding missed detections.

[0047] V. Alarm and Sorting Module. This module connects to the anomaly detection module and is used to perform physical actions. When an anomaly signal is received, it triggers an audible and visual alarm to alert the operator; controls the marking machine to automatically mark the abnormal wafer; and controls the robotic arm to unload the wafer into the non-conforming area.

[0048] VI. Data Storage Module. Used to store key data during the inspection process, including abnormal images, full-circumference width distribution curves, judgment logs, etc., to facilitate subsequent quality traceability and process improvement.

[0049] An online optical inspection method for local anomalies at the edge of the back side of a wafer, comprising the following steps: Step S1: Loading and Positioning. An automated loading robot picks up the wafer to be tested and transfers it to the center of the rotating stage. The system performs preliminary position correction on the wafer using edge sensors or visual positioning to ensure that the rotation center coincides with the wafer's geometric center.

[0050] Step S2: Full-circle image scanning, starting the image acquisition module and low-angle ring light source, controlling the rotating stage to drive the wafer to rotate 360° at a uniform speed. The camera continuously scans the 0~5mm area of ​​the back edge of the wafer to obtain complete back edge unfolded image information.

[0051] Step S3: Image preprocessing. The acquired raw image is preprocessed. First, Gaussian filtering is performed to remove noise interference from the sensor; then contrast enhancement (such as histogram equalization) is performed to strengthen the difference between edge contours and background, improving the robustness of subsequent algorithms.

[0052] Step S4: Boundary extraction and analysis of image grayscale gradient using edge detection algorithms. The boundary line between the back face of the wafer and the chamfered surface is defined as the starting boundary, and the boundary line between the chamfered surface and the edge vertex is defined as the ending boundary. Sub-pixel edge localization technology is used to accurately locate the coordinates of the two boundary lines.

[0053] Step S5: Feature quantization calculation.

[0054] Width calculation: Calculate the pixel distance between the starting and ending boundaries at each angle position, and convert it into the actual physical size (μm) according to the camera calibration parameters to generate a 360° chamfered radial width distribution curve.

[0055] Gray-scale feature calculation: Within the chamfered region, calculate the local gray-scale mean at each angular position.

[0056] Step S6: Anomaly Detection. This step uses a dual-criteria approach for judgment, with the specific logic as follows: Chamfer radial width anomaly determination: Calculate the mean value of the radial width of the chamfer around the entire circumference.

[0057] Condition A: If the width WW at any position satisfies the relationship |W-\mu| / \mu > 15\%|W−μ| / μ>15%, it indicates that the chamfer width at that position has changed significantly relative to the average level (such as the chamfer becoming wider due to accidental contact with the grinding wheel), and is judged as an abnormal width.

[0058] Condition B: If the radial width change rate of the chamfer at adjacent angle positions is > 20% / ^circ > 20% / ∘, it indicates that there is a sudden change (such as edge chipping) at that location, and it is judged as a width anomaly.

[0059] If either condition A or B is met, the geometric dimensions at that location are determined to be abnormal.

[0060] Anomaly detection of reflection characteristics: Calculate the mean and standard deviation of grayscale values ​​around the entire circumference, σ.

[0061] If the deviation between the local grayscale mean at any location and the overall grayscale mean is greater than 3σ, it indicates a drastic change in reflectivity at that location. For wafer structures containing both a POLY layer and a SiO2 film, this typically signifies film surface damage or contamination, and is thus classified as an abnormal reflection.

[0062] Step S7: Result processing.

[0063] If an abnormality in width or reflection is detected, the system records the angular coordinates of the abnormal location, triggers an audible and visual alarm, an automatic marking machine marks the defect location, a robotic arm unloads the wafer into the non-conforming area, and simultaneously saves the abnormal image and log to the data storage module.

[0064] If no abnormal conditions are triggered, the wafer is deemed qualified, and the robotic arm transfers it to the next process.

[0065] Application Example: Detection of SiO2 film damage in wafer end-face grinding process. During wafer end-face grinding, the grinding wheel may sometimes deviate from its position, accidentally contacting the SiO2 film layer on the back side of the wafer.

[0066] Phenomenon: This accidental contact will cause a local increase in the radial width of the chamfer on the back side of the wafer, and at the same time damage the smooth surface of the SiO2 film, resulting in enhanced diffuse reflection.

[0067] Detection process: Using the system of this embodiment, a low-angle light source is irradiated onto the damaged area. Due to the damage to the film layer, the intensity of the reflected light changes, and the gray value shows a sudden change (satisfying the condition >3σ > 3σ); at the same time, the chamfer width data shows a local peak (satisfying the condition |W-μ| / μ > 15% |W−μ| / μ > 15%).

[0068] In summary, the online optical inspection system and method for local anomalies on the back edge of wafers can effectively detect the increase in chamfer width and damage to the SiO2 film caused by accidental contact of the grinding wheel during the end face grinding process. It achieves non-contact, high-precision online full-circumference inspection, thereby improving wafer sorting efficiency and manufacturing yield.

[0069] The above description is only a specific embodiment of the present invention, but the structural features of the present invention are not limited thereto. Any changes or modifications made by those skilled in the art within the scope of the present invention are covered by the patent scope of the present invention.

Claims

1. An online optical inspection system for local anomalies on the back edge of a wafer, characterized in that: It includes an image acquisition module, a wafer transfer and rotation module, an image processing module, an anomaly detection module, an alarm and sorting module, and a data storage module; The image acquisition module includes a line scan camera or an area scan camera, and a low-angle ring light source set in conjunction with the camera; the image acquisition module is configured to perform a 360° full-circumference scan of the 0-5mm area on the back edge of the wafer to acquire image information of the back edge; The wafer transfer and rotation module includes an automatic feeding robot and a rotating stage. The rotating stage is used to carry and drive the wafer to rotate at a constant speed, so as to cooperate with the image acquisition module to perform full-circumference scanning. The image processing module is connected to the image acquisition module and has a built-in edge extraction algorithm, width calculation unit, and reflection feature analysis unit. The image processing module is configured to receive the image information, extract the chamfer start boundary and end boundary, calculate the chamfer radial width at each angle position, and calculate the local grayscale mean. The anomaly detection module is connected to the image processing module and is configured to perform logical judgment on the radial width of the chamfer and the local grayscale mean based on preset width threshold and grayscale threshold. If the anomaly condition is met, an anomaly signal is output. The alarm and sorting module is connected to the anomaly determination module and is configured to receive the anomaly signal, perform audible and visual alarms, automatic marking, and unload non-conforming wafers; and the data storage module is configured to save anomaly images, full-circumference width distribution curves, and determination logs.

2. The online optical inspection system for local anomalies at the back edge of a wafer according to claim 1, characterized in that: The incident angle of the low-angle ring light source is in the range of 45° to 70°, which is used to emphasize the wafer edge contour and enhance the reflective characteristics of the SiO2 film surface during imaging.

3. The online optical inspection system for local anomalies at the back edge of a wafer according to claim 1, characterized in that: The resolution of the line scan camera is ≥5μm / pixel, and the positioning accuracy of the rotating stage is ≤0.1°.

4. The online optical inspection system for local anomalies at the back edge of a wafer according to claim 1, characterized in that: The anomaly determination module uses "OR" logic, meaning that the anomaly signal is triggered when either the radial width of the chamfer is determined to be abnormal or the reflection feature is determined to be abnormal.

5. The online optical inspection system for local anomalies at the back edge of a wafer according to claim 4, characterized in that: The criteria for determining abnormal chamfer radial width are as follows: calculate the mean value μ of the radial width of the chamfer around the entire circumference. If the width W at any position satisfies the relationship: |W-μ| / μ>15%; or the rate of change of the radial width of the chamfer at adjacent angle positions is >20% / °.

6. The online optical inspection system for local anomalies at the back edge of a wafer according to claim 4, characterized in that: The criteria for determining abnormal reflection characteristics are as follows: calculate the mean gray level and standard deviation σ of the entire circumference. If the deviation between the mean gray level at any location and the mean gray level of the entire circumference is >3σ, then it is determined to be an abnormal reflection.

7. An online optical detection method for local anomalies at the edge of the back side of a wafer according to any one of claims 1 to 6, characterized in that, Includes the following steps: Step S1: Transfer the wafer to be tested to the rotating stage and perform edge positioning; Step S2: Start the image acquisition module and low-angle ring light source, control the wafer to rotate 360° at a constant speed, and acquire the full circumference image of the back edge of the wafer; Step S3: Preprocess the acquired image, including noise reduction and contrast enhancement; Step S4: Extract the start and end boundaries of the chamfer using an edge detection algorithm; Step S5: Calculate the radial width of the chamfer at each angle position, generate the full-circumference width distribution curve, and calculate the local grayscale features of the image at the same time; Step S6: Determine anomalies based on width and grayscale thresholds; Step S7: If a width or reflection abnormality is detected, record the angular coordinates of the abnormal location, trigger an audible and visual alarm, automatically unload the wafer into the non-conforming area, and save the abnormal image; if no abnormality is triggered, the wafer is deemed qualified and flows into the next process.

8. The online optical detection method for local anomalies at the back edge of a wafer according to claim 7, characterized in that, Step S6 specifically includes: calculating the mean value μ of the radial width of the full circumference chamfer; Determine if there are width anomalies: If the width W at any position satisfies |W-μ| / μ>15%, or the width change rate of adjacent angles>20% / °, then it is determined to be a width anomaly; To determine if there are any reflection anomalies: Calculate the mean and standard deviation σ of the grayscale value around the entire circumference. If the deviation between the local grayscale mean and the mean grayscale value around the entire circumference is greater than 3σ, then it is determined to be a reflection anomaly.

9. The online optical detection method for local anomalies at the back edge of a wafer according to claim 7, characterized in that: The method is applicable to wafer structures containing a POLY layer and a SiO2 film layer on top of the POLY layer, and is used to detect localized increases in chamfer width and film damage caused by the grinding wheel accidentally touching the back SiO2 film surface during the end face grinding process.

10. The online optical detection method for local anomalies at the back edge of a wafer according to claim 7, characterized in that: The chamfer start boundary extracted in step S4 is the boundary line between the back plane of the wafer and the chamfer surface, and the end boundary is the boundary line between the chamfer surface and the edge vertex.