Wafer edge corner defect and roundness detection method based on contour fitting

Abstract

The application discloses a wafer edge corner defect and roundness detection method based on contour fitting, and particularly relates to the technical field of wafer edge corner defect and roundness detection, and effectively weakens high-frequency interference components of process microfluctuation and transient vibration in the edge contour through pretreatment and smooth constraint on wafer edge sub-pixel point cloud data; an ideal circle fitting model is constructed and optimized, rotation scanning and conveying vibration factors are introduced into the fitting process, the stability of the reference geometric datum is ensured, reliable identification of the corner defect area and accurate quantification of the corner defect angle and the missing area are realized through systematic analysis of the radial deviation between the real edge points and the fitting circle, the overall geometric quality of the wafer is uniformly represented in the form of roundness deviation index and corner defect position index, the geometric quality index and dynamic environment monitoring data are coupled to correct the sampling position and time synchronization relationship in real time, and systematic errors introduced due to attitude drift, vibration impact and sampling phase dislocation are compensated.

Description

Technical Field

[0001] This invention relates to the field of wafer edge defect and roundness detection technology, and more specifically, to a wafer edge defect and roundness detection method based on contour fitting. Background Technology

[0002] In high-speed automated wafer manufacturing lines, wafer edge chipping and roundness detection typically need to be performed online without stopping the line. During the detection process, the wafer is often under dynamic conditions such as rotational scanning, robotic gripping and releasing, and continuous movement of the conveyor mechanism. Due to the coupling of these multi-source motion states, the wafer inevitably experiences micro-vibrations, attitude drift, and relative position shifts during the detection transients, causing the edge contour acquisition to no longer meet the static geometric consistency assumption. On the one hand, the superposition of rotational scanning and conveyor vibration introduces circumferential angle sampling errors, causing the correspondence between adjacent edge points in the angular domain to shift. On the other hand, the sudden acceleration caused by the gripping and releasing transients leads to phase misalignment in the edge point radius sampling, causing the same physical location to be mapped to inconsistent radial distances at different sampling times. At the same time, because it is difficult to completely synchronize the detection trigger and the motion state, the edge point cloud exhibits uneven sampling density in local areas, with some areas having sparse point sets and others having excessively concentrated points. Under the combined influence of the above factors, the spatial distribution of the wafer edge contour point set no longer satisfies the assumptions of the unified geometric model. This makes the reference circle center and radius established based on contour fitting susceptible to the systematic influence of dynamic disturbances, thereby weakening the stability and reliability of the fitting results. This introduces a risk of deviation in the corner identification and roundness assessment results under high-speed dynamic detection scenarios. Therefore, how to improve the adaptability of contour fitting to transient vibrations and positional drift in the actual working conditions of strong coupling between dynamic handling and online detection remains a key technical problem faced by existing wafer edge corner and roundness detection methods. Summary of the Invention

[0003] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a wafer edge missing corner and roundness detection method based on contour fitting, in order to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: A method for detecting wafer edge defects and roundness based on contour fitting includes the following steps: Acquire wafer edge contour data, and obtain sub-pixel precision point cloud data of the wafer edge through high-precision optical scanning equipment; An ideal circle fitting model is constructed, and circle fitting is performed on the edge data to obtain the ideal center and radius of the wafer. The fitting is then optimized by combining rotation scanning and transport vibration factors. Based on the radial deviation between the optimized ideal circle fitting model and the actual edge points, the missing corner region is identified, the missing corner angle and missing area are calculated, the overall roundness of the wafer is evaluated, and the roundness deviation index and missing corner position index are extracted. By combining the roundness deviation index and the corner defect position index with dynamic environmental monitoring data, the wafer is corrected in real time, and the sampling position is adjusted to be synchronized with the time.

[0005] In a preferred embodiment, the process of constructing an ideal circle fitting model and performing circle fitting on the edge data to obtain the ideal center and radius of the wafer is as follows: The ideal circular model for constructing the edge of a wafer is as follows: ,in, With the center of the circle, The radius is used to input the wafer edge point cloud data into the error minimization function. By analytically minimizing the error, a preliminary fitted circle center is obtained. and radius ; The minimized error function for: ,in For the initial fitting of the circle center , For radius, For the i-th real edge point of the wafer, This represents the total number of edge points.

[0006] In a preferred embodiment, the fitting optimization process combining rotational scanning and transport vibration factors is as follows: after obtaining the wafer edge point cloud and completing the initial circle fitting, dynamic reliability weights are introduced to the edge points to obtain the minimized error function after fitting optimization. ; The minimized error function after fitting optimization for: ,in is the reliability weight of the i-th wafer's true edge point.

[0007] In a preferred embodiment, the process of identifying the missing corner region and calculating the missing corner angle and missing area based on the radial deviation between the optimized ideal circle fitting model and the actual edge point is as follows: After obtaining the optimized fitted circle center and radius Then, based on each real edge point Calculate its radial deviation from the fitted circle. The radial deviation sequence is obtained: ; According to the radial deviation sequence Identify missing corner areas: When the radial deviation value within a continuous area... The deviation is consistently less than the preset threshold. At that time, it was determined that there was a missing corner in the area, and the missing corner interval was marked as... ,in and These are the indices of the starting and ending points of the missing corner, respectively. The starting and ending points of the missing corner section can be determined by the following conditions: ; The angle of the missing corner is obtained based on the start and end angles of the missing corner area. : ,in This represents the total number of edge points; simultaneously, the area of ​​the missing corner is obtained based on the corner angle. : .

[0008] In a preferred embodiment, the process of evaluating the overall roundness of the wafer and extracting the roundness deviation index and the corner defect location index is as follows: The overall roundness evaluation index of the wafer includes the maximum radial deviation. and root mean square deviation : ; The roundness deviation index is calculated by combining the maximum radial deviation and the root mean square deviation. : ; The missing corner position index is calculated by combining the missing corner angle and the missing corner area. : ,in This represents the total area of ​​the wafer.

[0009] In a preferred embodiment, the process of combining the roundness deviation index and the corner defect position index with dynamic environmental monitoring data to perform real-time correction of the wafer and adjusting the sampling position and time synchronization is as follows: Correction coefficients are defined based on the roundness deviation index and the corner missing position index. ; The correction coefficient It can be represented as: ,in , These are the preset proportional coefficients for the roundness deviation index and the missing corner position index, respectively. , All are greater than 0, and satisfy ; Acquire dynamic environmental data of the wafer's environment, including the wafer's rotation speed. Vibration acceleration of the conveyor belt ; The correction factor is calculated by combining dynamic environmental monitoring data with each timestamp. Corresponding wafer position correction value and time correction value The actual sampling location is adjusted in real time. ; During the actual sampling process, the wafer sampling position and time synchronization are adjusted, and the sensor's sampling frequency and scanning position are reconfigured based on the corrected position and time synchronization information. ,in The corrected scan position. The scan position before correction. The corrected sampling time. The sampling time before correction.

[0010] The technical effects and advantages of this invention are as follows: 1. This invention effectively reduces the high-frequency interference components of process micro-fluctuations and transient vibrations in the edge contour by preprocessing and smoothing the sub-pixel point cloud data of the wafer edge. It constructs and optimizes an ideal circle fitting model, introducing rotation scanning and transport vibration factors into the fitting process, so that the center and radius of the fitted circle are no longer significantly affected by local sampling misalignment or uneven point cloud density, thereby ensuring the stability of the reference geometric benchmark. Through systematic analysis of the radial deviation between the real edge point and the fitted circle, it achieves reliable identification of the missing corner area and accurate quantification of the missing corner angle and missing area. It also uniformly represents the overall geometric quality of the wafer in the form of roundness deviation index and missing corner position index. It couples the geometric quality index with dynamic environmental monitoring data, and effectively compensates for the systematic errors introduced by attitude drift, vibration impact and sampling phase misalignment by real-time correction of the sampling position and time synchronization relationship. Thus, it can maintain the reliability and consistency of the missing corner identification and roundness evaluation results under continuous operation and high-speed dynamic detection conditions, and significantly improve the adaptability of the wafer edge detection method based on contour fitting to the actual dynamic working conditions of production, and reduce the risk of false detection and missed detection. Attached Figure Description

[0011] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings; Figure 1 This is a flowchart of a method according to an embodiment of 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] Example: Figure 1 This invention presents a method for detecting wafer edge defects and roundness based on contour fitting, comprising the following steps: Acquire wafer edge contour data, obtain sub-pixel precision point cloud data of wafer edge through high-precision optical scanning equipment, and preprocess it to remove noise and outliers, and perform smoothing to reduce the impact of small fluctuations introduced by the process. An ideal circle fitting model is constructed, and circle fitting is performed on the edge data to obtain the ideal center and radius of the wafer. The fitting is then optimized by combining rotation scanning and conveying vibration factors to improve the fitting stability. Based on the radial deviation between the optimized ideal circle fitting model and the actual edge points, the missing corner region is identified, the missing corner angle and missing area are calculated, the overall roundness of the wafer is evaluated, and the roundness deviation index and missing corner position index are extracted. By combining the roundness deviation index and the corner defect position index with dynamic environmental monitoring data, the wafer is corrected in real time, and the sampling position and time are synchronized to ensure the accuracy and stability of detection in a high-speed dynamic environment.

[0014] In this embodiment of the invention, the process of acquiring wafer edge contour data, obtaining sub-pixel precision point cloud data of the wafer edge through a high-precision optical scanning device, and preprocessing the data to remove noise and outliers, and performing smoothing to reduce the impact of minor fluctuations introduced by the process, is as follows: An edge data acquisition interface is set on the input side of the wafer edge defect and roundness detection system to simultaneously acquire high-precision optical scanning data and dynamic environmental influence parameters of the wafer edge before detection. The dynamic environmental parameters include at least the vibration effects on the transmission path and the mechanical gripping acceleration, and the data from different sources are aligned by time and label to ensure the consistency and traceability of various parameters in the subsequent contour fitting process. The acquired wafer edge data undergoes preliminary preprocessing, including noise removal, outlier filtering, and edge quality detection. A sub-pixel precision edge extraction algorithm is used to convert the original scanned image into two-dimensional edge point cloud data and map the edge points to the real geometric trajectory of the wafer edge. Outliers that do not conform to geometric rules are removed to ensure a smooth edge point cloud. When smoothing edge points, a Gaussian filtering method is used to smooth the acquired edge points to reduce the influence of small fluctuations caused by the process and filter out high-frequency vibration noise caused by changes in the scanning environment or irregularities on the wafer surface, thereby improving the accuracy and stability of subsequent contour fitting.

[0015] In this embodiment of the invention, an ideal circle fitting model is constructed, and circle fitting is performed on the edge data to obtain the ideal center and radius of the wafer. The fitting is then optimized by combining rotation scanning and conveying vibration factors to improve the fitting stability. The ideal circular model for constructing the edge of a wafer is as follows: ,in, With the center of the circle, The radius is used to input the wafer edge point cloud data into the error minimization function. By analytically minimizing the error, a preliminary fitted circle center is obtained. and radius ; The minimized error function for: ,in For the initial fitting of the circle center , For radius, For the i-th real edge point of the wafer, This represents the total number of edge points; After obtaining the wafer edge point cloud and completing the initial circle fitting, to reduce the interference of rotation scanning and transport vibration on the fitting results during high-speed online inspection, rotation scanning state parameters and transport vibration characteristic parameters are introduced to optimize the fitting process. Dynamic reliability weights are introduced for the edge points to obtain the minimized error function after fitting optimization. Edge points collected during moments of large fluctuations in rotational speed or sudden changes in vibration acceleration are assigned lower weights, while edge points collected within relatively stable motion ranges are assigned higher weights, thereby suppressing the influence of dynamic disturbance points on the estimation of the center and radius of the circle. For example, the first Reliability weights of edge points It can be represented as: ,in For the first The rotational angular acceleration corresponding to the edge point acquisition time, To transmit the amplitude of vibration acceleration; The minimized error function after fitting optimization for: .

[0016] In this embodiment of the invention, based on the radial deviation between the optimized ideal circle fitting model and the actual edge points, the missing corner region is identified, the missing corner angle and missing area are calculated, and the overall roundness of the wafer is evaluated. The process of extracting the roundness deviation index and the missing corner position index is as follows: After obtaining the optimized fitted circle center and radius Then, based on each real edge point Calculate its radial deviation from the fitted circle. The radial deviation sequence is obtained: ; It should be noted that the radial deviation describes the magnitude of the deviation between each edge point and the fitted circle. Through the deviation sequence, a set of points that deviate significantly from the fitted circle can be identified, and the regions corresponding to these point sets are the missing corner regions. According to the radial deviation sequence Identify missing corner areas: When the radial deviation value within a continuous area... The deviation is consistently less than the preset threshold. At that time, it was determined that there was a missing corner in the area, and the missing corner interval was marked as... ,in and These are the indices of the starting and ending points of the missing corner, respectively. The starting and ending points of the missing corner section can be determined by the following conditions: ; The angle of the missing corner is obtained based on the start and end angles of the missing corner area. : ,in This represents the total number of edge points; simultaneously, the area of ​​the missing corner is obtained based on the corner angle. : ; The overall roundness evaluation index of the wafer includes the maximum radial deviation. and root mean square deviation : ; The roundness deviation index is calculated by combining the maximum radial deviation and the root mean square deviation. : ; The missing corner position index is calculated by combining the missing corner angle and the missing corner area. : ,in This represents the total area of ​​the wafer.

[0017] In this embodiment of the invention, the process of combining the roundness deviation index and the corner defect position index with dynamic environmental monitoring data to perform real-time correction on the wafer, adjusting the sampling position and time synchronization, and ensuring the detection accuracy and stability in a high-speed dynamic environment is as follows: Correction coefficients are defined based on the roundness deviation index and the corner missing position index. The correction coefficient is used to dynamically adjust the sampling position of the wafer and synchronize it with the time. For example, the correction coefficient It can be represented as: ,in , These are the preset proportional coefficients for the roundness deviation index and the missing corner position index, respectively. , All are greater than 0, and satisfy ; It should be noted that, , The settings should be tailored to the specific circumstances. For example, an expert-empowered approach could be adopted, where experts in relevant fields are invited to determine the pre-defined proportions for each indicator through professional opinion surveys and comprehensive evaluations. , The initial value can be 0.5, 0.5; Acquire dynamic environmental data of the wafer's environment, including but not limited to the wafer's rotation speed. Vibration acceleration of the conveyor belt ; The correction factor is calculated by combining dynamic environmental monitoring data with each timestamp. Corresponding wafer position correction value and time correction value The actual sampling location is adjusted in real time. ; During actual sampling, the real-time performance and accuracy of edge data acquisition are ensured by adjusting the wafer sampling position and time synchronization. Based on the corrected position and time synchronization information, the sensor's sampling frequency and scanning position are reconfigured. ,in The corrected scan position. The scan position before correction. The corrected sampling time. The sampling time before correction.

[0018] This invention effectively reduces the high-frequency interference components of process micro-fluctuations and transient vibrations in the edge contour by preprocessing and smoothing the sub-pixel point cloud data of the wafer edge. It constructs and optimizes an ideal circle fitting model, introducing rotational scanning and transport vibration factors into the fitting process. This ensures that the center and radius of the fitted circle are no longer significantly affected by local sampling misalignment or uneven point cloud density, thus guaranteeing the stability of the reference geometric benchmark. Through systematic analysis of the radial deviation between the real edge points and the fitted circle, it achieves reliable identification of missing corner regions and accurate quantification of missing corner angles and areas. The overall geometric quality of the wafer is uniformly characterized in the form of a roundness deviation index and a missing corner position index. The geometric quality indicators are coupled with dynamic environmental monitoring data. By real-time correction of the sampling position and time synchronization relationship, it effectively compensates for systematic errors introduced by attitude drift, vibration impact, and sampling phase misalignment. Therefore, it can maintain the reliability and consistency of missing corner identification and roundness evaluation results even under continuous, high-speed dynamic detection conditions, significantly improving the adaptability of the contour fitting-based wafer edge detection method to actual production dynamic conditions and reducing the risk of false detection and missed detection.

[0019] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0020] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0021] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for detecting wafer edge defects and roundness based on contour fitting, characterized in that: Includes the following steps: Acquire wafer edge contour data, and obtain sub-pixel precision point cloud data of the wafer edge through high-precision optical scanning equipment; An ideal circle fitting model is constructed, and circle fitting is performed on the edge data to obtain the ideal center and radius of the wafer. The fitting is then optimized by combining rotation scanning and transport vibration factors. Based on the radial deviation between the optimized ideal circle fitting model and the actual edge points, the missing corner region is identified, the missing corner angle and missing area are calculated, the overall roundness of the wafer is evaluated, and the roundness deviation index and missing corner position index are extracted. By combining the roundness deviation index and the corner defect position index with dynamic environmental monitoring data, the wafer is corrected in real time, and the sampling position is adjusted to be synchronized with the time.

2. The wafer edge missing corner and roundness detection method based on contour fitting according to claim 1, characterized in that: The process of constructing an ideal circle fitting model and performing circle fitting on the edge data to obtain the ideal center and radius of the wafer is as follows: The ideal circular model for constructing the edge of a wafer is as follows: ,in, With the center of the circle, The radius is used to input the wafer edge point cloud data into the error minimization function. By analytically minimizing the error, a preliminary fitted circle center is obtained. and radius ; The minimized error function for: ,in For the initial fitting of the circle center , For radius, For the i-th real edge point of the wafer, This represents the total number of edge points.

3. The wafer edge missing corner and roundness detection method based on contour fitting according to claim 2, characterized in that: The process of fitting and optimizing by combining rotational scanning and transport vibration factors is as follows: After obtaining the wafer edge point cloud and completing the initial circle fitting, dynamic reliability weights are introduced to the edge points to obtain the minimized error function after fitting optimization. ; The minimized error function after fitting optimization for: ,in is the reliability weight of the i-th wafer's true edge point.

4. The wafer edge missing corner and roundness detection method based on contour fitting according to claim 3, characterized in that: Based on the radial deviation between the optimized ideal circle fitting model and the actual edge points, the process of identifying the missing corner region and calculating the missing corner angle and area is as follows: After obtaining the optimized fitted circle center and radius Then, based on each real edge point Calculate its radial deviation from the fitted circle. The radial deviation sequence is obtained: ; According to the radial deviation sequence Identify missing corner areas: When the radial deviation value within a continuous area... The deviation is consistently less than the preset threshold. At that time, it was determined that there was a missing corner in the area, and the missing corner interval was marked as... ,in and These are the indices of the starting and ending points of the missing corner, respectively. The starting and ending points of the missing corner section can be determined by the following conditions: ; The angle of the missing corner is obtained based on the start and end angles of the missing corner area. : ,in This represents the total number of edge points; simultaneously, the area of ​​the missing corner is obtained based on the corner angle. : .

5. The wafer edge missing corner and roundness detection method based on contour fitting according to claim 4, characterized in that: The process of evaluating the overall roundness of a wafer and extracting the roundness deviation index and the corner defect location index is as follows: The overall roundness evaluation index of the wafer includes the maximum radial deviation. and root mean square deviation : ; The roundness deviation index is calculated by combining the maximum radial deviation and the root mean square deviation. : ; The missing corner position index is calculated by combining the missing corner angle and the missing corner area. : ,in This represents the total area of ​​the wafer.

6. The wafer edge missing corner and roundness detection method based on contour fitting according to claim 5, characterized in that: The process of combining the roundness deviation index and the corner defect position index with dynamic environmental monitoring data to perform real-time correction on the wafer, adjusting the sampling position and time synchronization, and ensuring the accuracy and stability of detection in a high-speed dynamic environment is as follows: Correction coefficients are defined based on the roundness deviation index and the corner missing position index. ; The correction coefficient It can be represented as: ,in , These are the preset proportional coefficients for the roundness deviation index and the missing corner position index, respectively. , All are greater than 0, and satisfy ; Acquire dynamic environmental data of the wafer's environment, including the wafer's rotation speed. Vibration acceleration of the conveyor belt ; The correction factor is calculated by combining dynamic environmental monitoring data with each timestamp. Corresponding wafer position correction value and time correction value The actual sampling location is adjusted in real time. ; During the actual sampling process, the wafer sampling position and time synchronization are adjusted, and the sensor's sampling frequency and scanning position are reconfigured based on the corrected position and time synchronization information. ,in The corrected scan position. The scan position before correction. The corrected sampling time. The sampling time before correction.

Images