Wafer secondary chamfer self-adaptive grinding method and system based on edge topography perception

Abstract

The application discloses a wafer secondary chamfer adaptive grinding method and system based on edge topography perception, and belongs to the technical field of semiconductor wafer processing, aiming to solve the technical problems that the existing secondary chamfer adopts a fixed parameter batch processing mode, which cannot adapt to individual defect differences of wafer edges, and easily causes defect expansion, wafer cracking and yield fluctuation. The method collects wafer full-edge topography images through pre-scanning, extracts edge collapse and crack characteristic parameters, and completes risk grading; according to the grading results, a differentiated grinding strategy is matched, and the corresponding parameter partition grinding is performed according to the circumferential sector; vibration and acoustic emission signals are collected during the processing to monitor the state in real time; after the processing is completed, profile re-inspection is carried out and the results are fed back, forming a quality closed loop. The corresponding system integrates five types of functional units, namely wafer transmission, optical detection, image processing, grinding execution and process monitoring. The application can effectively inhibit edge defect expansion, reduce wafer scrap and consumable loss, and improve edge profile consistency.

Description

Technical Field

[0001] This invention relates to the field of semiconductor wafer processing technology, and in particular to an adaptive grinding method and system for secondary chamfering of wafers based on edge morphology perception. Background Technology

[0002] In the entire semiconductor manufacturing process, the integrity and consistency of the wafer edge morphology directly affect the processing quality of subsequent processes such as thin film deposition, photolithography, and chemical mechanical polishing. It also relates to the wafer's resistance to breakage during handling and transport, and is a key link in ensuring wafer manufacturing yield and device reliability. Among them, secondary chamfering, as an edge finishing process in the front-end wafer processing, is used to correct edge stress and micro-damage introduced by previous processes, and plays a decisive role in the final wafer edge contour accuracy and surface quality.

[0003] Currently, the secondary chamfering process for wafers in the industry generally adopts a batch grinding mode with fixed process parameters. That is, for wafers of the same size and material, uniform process parameters such as grinding wheel speed, feed speed and downward pressure are preset. The entire batch of wafers is chamfered and ground using the same parameters. Conventional secondary chamfering equipment mainly consists of a wafer carrier rotation mechanism, a diamond grinding wheel and a feed drive mechanism. During the processing, the wafer rotates at a uniform speed with the carrier table, and the grinding wheel feeds to the edge of the wafer with preset parameters. Through continuous material removal, the target arc or bevel chamfer profile is formed. Some mass production lines will use offline optical sampling inspection to sample and inspect the edge morphology of the wafers after the batch processing is completed to determine whether the processing quality of the entire batch meets the specifications. Iterative adjustment of processing parameters usually relies on manual experience and batch sampling inspection results.

[0004] The aforementioned existing technical solutions have significant limitations in actual production. Due to individual differences in the stress experienced by wafers during pre-processing and handling, the edge condition of wafers entering the secondary chamfering process is not consistent. Some wafer edges already have micro-defects such as micron-level edge chipping and subsurface hidden cracks. When grinding with uniform high-intensity fixed parameters, the grinding force will cause the original micro-defects to expand rapidly. This can lead to increased edge chipping size or even complete wafer breakage and scrapping. At the same time, the fragments generated by the breakage can easily cause damage to the grinding wheel and contamination of the processing chamber, increasing the cost of processing consumables and equipment maintenance. In addition, the existing processing mode lacks real-time status perception and dynamic parameter adjustment capabilities during the processing process, which is a "blind grinding" processing method. It cannot perform differentiated processing for local defects of a single wafer, nor can it detect and intervene in new defects in time during the processing. Ultimately, this results in poor wafer edge contour consistency and large fluctuations in processing yield, making it difficult to meet the fine quality control requirements of high-precision semiconductor manufacturing.

[0005] The existing secondary chamfering process uses a batch grinding mode with uniform and fixed parameters. The parameter setting relies on manual experience and batch sampling results, and the adjustment has a significant lag. The edge condition of the wafers entering the process varies from person to person. Some micro-defects inherent in some wafers are prone to expansion under high-intensity grinding, causing wafer breakage and damage to processing consumables. There is no real-time status sensing means during the processing, and new defects cannot be detected and dealt with in a timely manner. Offline sampling after the batch is completed cannot intervene in the processing process. It does not have the ability to process localized differences, and the consistency of edge contours is difficult to guarantee. The processing yield fluctuates significantly and cannot meet the fine control requirements of high-precision semiconductor manufacturing. Summary of the Invention

[0006] To address the technical problems existing in the prior art, embodiments of the present invention provide a wafer secondary chamfering adaptive polishing method and system based on edge morphology perception. The technical solution is as follows: An adaptive grinding method for wafer secondary chamfering based on edge morphology awareness includes the following steps: S1. Wafer loading and edge full-circumference optical pre-scan: The wafer to be subjected to secondary chamfering is transferred to the pre-processing station of the processing equipment. The non-contact optical inspection module is activated. High-resolution line scan imaging combined with dark field detection mode is used to perform a 360° full-circumference scan on the top edge, bottom edge and chamfered side of the wafer, and collect the morphological image data of the wafer's full-circumference edge. S2. Edge defect identification and processing risk classification: The acquired edge morphology image is transmitted to the image processing unit to extract the geometric feature parameters of edge chipping and hidden crack defects in the image. Based on the preset defect judgment threshold system, the corresponding circumferential regions of the wafer edge are divided into three risk levels: no defects or slight defects, moderate defects, and severe defects. S3. Adaptive grinding strategy is dynamically generated. Based on the risk classification results of each region at the edge of the wafer, an appropriate combination of grinding process parameters is matched. Three differentiated processing strategies are generated for edge regions with different risk levels: standard grinding, flexible trimming, skipping or light polishing. S4. Real-time monitoring of the segmented grinding and processing process: According to the generated grinding strategy, the grinding head is controlled to perform grinding processing with corresponding parameters on multiple sectors pre-divided on the edge of the wafer. At the same time, the vibration signal and acoustic emission signal of the grinding wheel are collected in real time throughout the grinding process to monitor the processing status. S5. Post-grinding contour re-inspection and quality closed-loop verification: After all grinding processes are completed, the optical inspection module is restarted to perform a second full-circumference scan of the wafer edge to verify whether the final chamfer contour meets the target specifications. The verification results are then fed back to the process control unit.

[0007] As a further aspect of the present invention, in step S1, a high-resolution line scan camera is used in conjunction with a ring-shaped tilting light source to complete the imaging acquisition. For micron-level hidden cracks, a dark field detection mode is used to capture the scattered light generated at the edge defect to form an imaging image of a dark background and a bright defect. During the scanning process, the wafer is controlled to rotate at a constant speed, and the line scanning camera's row acquisition frequency enables continuous circumferential acquisition. The scanning range covers the corner areas of the top edge, bottom edge, and all beveled areas of the side chamfers of the wafer, forming a continuous edge morphology acquisition sequence along the circumference.

[0008] As a further aspect of the present invention, in step S2, the extracted defect geometric feature parameters include the depth and width parameters of the chipped edge, as well as the length and orientation parameters of the crack, and the risk classification is performed according to a preset multi-dimensional threshold standard. Among them, no defects or minor defects correspond to defect sizes far below the safety threshold, moderate defects correspond to defect sizes close to the safety threshold, and severe defects correspond to defect sizes exceeding the safety threshold. The classification results are bound one-to-one with the circumferential position of the wafer edge, and each defect is marked with its corresponding circumferential angle and location.

[0009] As a further aspect of the present invention, in step S3, standard grinding parameters are used to process areas with no defects or minor defects. For areas with moderate defects, a flexible dressing strategy is adopted, which reduces the grinding wheel speed, feed rate and downward contact force accordingly. For areas with severe defects, skip processing or perform only very low-speed, light polishing. The parameter combinations for the three types of processing strategies are all pre-stored in the process parameter library. The system directly calls the matching parameter combination based on the risk classification result and the corresponding circumferential position.

[0010] As a further aspect of the present invention, in step S4, the vibration signal and acoustic emission signal collected in real time are compared with a preset normal processing threshold. If an abnormal peak value exceeding the threshold is detected, it is determined that a new edge chipping defect may occur at the current processing position. The system immediately pauses the current grinding operation, controls the optical inspection module to perform a local re-inspection of the abnormal location, readjusts the grinding parameters of the subsequent sectors based on the re-inspection results, and then continues processing. The abnormal location and the re-inspection results are recorded synchronously in the processing log of the corresponding wafer.

[0011] As a further aspect of the present invention, in step S5, the second full-circumference scan adopts the same imaging configuration and detection accuracy as the pre-scan, and the verification content includes the arc shape of the chamfer contour, the edge surface quality, and the state of residual defects. If the verification results meet the target chamfer specifications, the processing is deemed qualified. If there are areas that do not meet the standards, a corresponding re-grinding strategy is generated, and the corresponding areas are re-grinded and then verified again. The verification results are simultaneously archived in the processing file of the corresponding wafer.

[0012] As a further aspect of the present invention, before performing the grinding process, the wafer edge is pre-divided into several equal processing sectors along the circumference, and each sector corresponds to an independent risk classification result and grinding parameter combination. The grinding head automatically switches the corresponding process parameters when processing different sectors. The number of sectors is adjusted according to the wafer size and processing accuracy requirements. The start and end angles of each sector correspond one-to-one with the circumferential position of the defect classification.

[0013] An adaptive polishing system for wafer secondary chamfering based on edge morphology awareness, the system comprising: The wafer transfer and loading module is used to sequentially transfer the wafers to be processed to the pre-processing station, the grinding station and the re-inspection station to complete the wafer process flow. The edge optical inspection module is equipped with a high-resolution line scan camera and a ring-shaped tilted light source, and supports dark field inspection mode for 360° full-circumference scanning and topographic image acquisition of wafer edges. The image processing and risk classification module is used to receive edge morphology images and extract defect feature parameters, and to determine the risk level of the edge region based on a preset threshold. The adaptive grinding execution module, which includes a grinding head and a drive control unit, is used to perform segmented differentiated grinding processes based on risk classification results. The process monitoring and closed-loop feedback module is equipped with vibration sensors and acoustic emission sensors for real-time monitoring of the grinding process and contour re-inspection after grinding.

[0014] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: In this invention, the circumferential morphology of the wafer edge and micro-defect identification are completed before the secondary chamfering process. This allows for precise capture of minute edge damage and extraction of corresponding geometric features. Processing risk levels are classified according to the severity of defects, and differentiated grinding parameters are matched for different levels of circumferential regions. Processing actions of corresponding intensity are executed according to the sectors divided by the circumference. During the processing, the running signals are collected simultaneously to monitor the status and dynamically adjust the strategy. After processing, contour re-inspection and result feedback are carried out, which effectively suppresses the expansion of the original defects, reduces the probability of wafer scrap, reduces the consumption of processing consumables, ensures the consistency of edge contours, and achieves refined quality control throughout the entire process. Attached Figure Description

[0015] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0016] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0017] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0018] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, their intended meanings are consistent. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, their intended meanings are consistent.

[0019] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0020] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0021] Please see Figure 1 This invention relates to the field of semiconductor wafer processing technology, specifically to an adaptive grinding method and system for secondary chamfering of wafers based on edge morphology perception. It is mainly applied to the secondary chamfering and finishing process of 5-8 inch wafers, used to eliminate edge micro-defects generated in the front-end wafer processing, while avoiding defect propagation problems caused by fixed grinding parameters. The technical solution of this invention is described in detail below through several embodiments. All parameters involved in the embodiments are actual production data under mass production scenarios. Those skilled in the art can adjust the parameters within the framework of the principles of this invention according to actual processing needs, and the corresponding adjustment schemes all fall within the protection scope of this invention.

[0022] Example 1: This example is mainly used to complete the loading and transfer of the wafer to be processed and the acquisition of full-circumference images of the edge morphology. Specifically, it includes transferring the wafer to be processed for secondary chamfering to the pre-processing station of the processing equipment, activating the non-contact optical detection module, and using high-resolution line scan imaging combined with dark field detection mode to perform a 360° full-circumference scan on the top edge, bottom edge and chamfered side of the wafer to acquire the morphology image data of the full-circumference edge of the wafer.

[0023] In the specific implementation process, the wafer to be processed is a 6-inch silicon-based polished wafer with a nominal diameter of 150mm and a thickness of 725μm. Before processing, the thin film deposition and chemical mechanical polishing processes have been completed, and the wafer enters the secondary chamfering station for edge finishing.

[0024] The optical inspection module in the preprocessing station is equipped with an 8K resolution line scan camera with a single pixel size of 5μm in the line array direction. The incident light from the ring-shaped tilted light source forms a 30° angle with the chamfered edge of the wafer side.

[0025] For micron-level hidden cracks, a dark field detection mode is used. In this mode, the light source does not directly hit the camera lens, and the intact wafer edge will not reflect light to the camera's photosensitive surface, so the image background appears dark. Defects such as chipped edges and cracks cause light to scatter, and some of the scattered light enters the camera lens, ultimately forming an image with a dark background and a bright defect, thereby improving the sensitivity of identifying tiny defects.

[0026] During the scanning process, the wafer is controlled to rotate at a constant speed of 60 r / min, which means it completes one revolution per second. Combined with the 7200 Hz line acquisition frequency of the line scan camera, continuous circumferential acquisition is achieved. It is calculated that 7200 lines of effective image data can be acquired for each revolution of the wafer, corresponding to a circumferential resolution of 360° ÷ 7200 = 0.05° / line.

[0027] The scanning range fully covers the corner areas of the wafer top edge, the corner areas of the bottom edge, and all the beveled areas of the side chamfers, forming a continuous edge morphology acquisition sequence along the circumference, with no blind spots in the scanning detection.

[0028] Example 2: Based on the edge morphology image acquisition completed in Example 1, this example further realizes the identification of wafer edge defects and the classification of processing risks. Specifically, it includes transmitting the acquired edge morphology image to the image processing unit, extracting the geometric feature parameters of edge chipping and hidden crack defects in the image, and classifying the corresponding circumferential regions of the wafer edge into three risk levels: no defects or slight defects, moderate defects, and severe defects, according to the preset defect judgment threshold system.

[0029] In the specific implementation process, the image processing unit adopts an FPGA hardware acceleration architecture to complete real-time image processing. The complete processing time of a single set of full-circle images does not exceed 2 seconds, which meets the production line cycle requirements.

[0030] The extracted defect geometric feature parameters include the depth and width parameters of the chipped edge, as well as the length and orientation parameters of the crack. The depth parameter refers to the distance the defect extends towards the center of the wafer, the width parameter refers to the extension dimension of the defect along the circumference of the wafer, and the orientation parameter refers to the angle between the crack extension direction and the wafer tangent direction.

[0031] Risk classification is performed based on preset multi-dimensional threshold standards, where no defects or minor defects correspond to chipping depth of less than 5μm, crack length of less than 10μm, and defect size is far below the safety threshold. The moderate defects correspond to chipping depths of 5μm to 15μm and crack lengths of 10μm to 30μm, with the defect size approaching the safety threshold. Severe defects are defined as edge chipping depths greater than 15 μm and crack lengths greater than 30 μm, with defect sizes exceeding safety thresholds. The grading results are linked one-to-one with the circumferential positions of the wafer edges. Each defect is marked with its corresponding circumferential angle and location, and the position marking accuracy for a single defect can reach ±0.1°, providing precise location data for subsequent partitioning processes.

[0032] Example 3: Based on the edge defect risk classification completed in Example 2, this example further realizes the dynamic generation of adaptive grinding strategies. Specifically, it includes matching appropriate grinding process parameter combinations according to the risk classification results of each region of the wafer edge, and generating three differentiated processing strategies for edge regions with different risk levels: standard grinding, flexible trimming, skipping, or light polishing.

[0033] In practice, the equipment's process parameter library pre-stores complete parameter combinations for three types of processing strategies. The system directly calls the matching parameter combination based on the risk classification results and the corresponding circumferential position, without the need for manual input.

[0034] For areas with no defects or minor defects, standard grinding parameters are used: a grinding wheel speed of 3000 r / min, a radial feed rate of 0.5 mm / s, and a downward contact force of 2 N. For areas with moderate defects, a flexible dressing strategy is adopted, corresponding to a reduction in the grinding wheel speed to 1500 r / min, a feed rate to 0.2 mm / s, and a downward contact force to 0.8 N. By reducing the amount of material removed per unit time, mechanical impact is reduced to suppress defect propagation.

[0035] For areas with severe defects, skip processing or perform only very low-speed, light polishing, with the polishing speed controlled below 800 r / min and the feed rate not exceeding 0.05 mm / s, without applying additional grinding force to the defective area.

[0036] Before performing the grinding process, the wafer edge is divided into 72 equal processing sectors along the circumference. Each sector corresponds to a 5° circumferential range, and each sector corresponds to an independent risk classification result and grinding parameter combination. The number of sectors can be adjusted according to the wafer size and processing accuracy requirements. The start and end angles of each sector correspond one-to-one with the circumferential position of the defect classification.

[0037] Embodiment 4: Based on the differentiated grinding strategy generated in Embodiment 3, this embodiment further realizes segmented grinding and real-time monitoring of the machining process. Specifically, according to the generated grinding strategy, the grinding head is controlled to perform grinding machining with corresponding parameters on a plurality of pre-divided sectors at the edge of the wafer respectively, and vibration signals and acoustic emission signals of the grinding wheel are collected in real time throughout the whole grinding process to monitor the machining state.

[0038] In the specific implementation process, the grinding head is equipped with a diamond grinding wheel, which is driven by a servo motor to realize precise control of rotation speed, feed and pressure. When the grinding head moves to different sectors, it automatically switches the corresponding process parameters to realize differentiated machining at different positions on the edge of the same wafer, without the need to replace the machining batch or adjust the overall parameters of the equipment.

[0039] A vibration acceleration sensor and an acoustic emission sensor are coupled and installed at the spindle base of the grinding wheel. The sampling frequency of both types of sensors is set to 1 MHz, which can capture tiny state changes during the machining process.

[0040] The system compares the vibration signals and acoustic emission signals collected in real time with preset normal machining thresholds, wherein the normal peak threshold of the vibration signals is set to 0.5g, and the normal amplitude threshold of the acoustic emission signals is set to 40dB.

[0041] If an abnormal peak exceeding the threshold is detected, it is determined that a new edge chipping defect may occur at the current machining position. The system immediately pauses the current grinding action, controls the optical detection module to perform local re-inspection on the abnormal position, and the scanning range of the local re-inspection is the circumferential area of 10° each before and after the abnormal position. After the grinding parameters of subsequent sectors are readjusted according to the re-inspection result, the machining is continued. The abnormal position and the re-inspection result are synchronously recorded in the machining log of the corresponding wafer, which is convenient for subsequent traceability and process optimization.

[0042] Embodiment 5: On the basis of completing all grinding machining in Embodiment 4, this embodiment further realizes post-grinding profile re-inspection and quality closed-loop verification. Specifically, after all the grinding processes are completed, the optical detection module is started again to perform a second full-perimeter scan on the wafer edge to verify whether the final chamfer profile meets the target specification requirements, and the verification result is fed back to the process control unit.

[0043] In the specific implementation process, the second full-perimeter scan adopts the same imaging configuration and detection accuracy as the pre-scan to ensure consistent detection standards before and after. The verification content includes the arc shape of the chamfer profile, the edge surface quality and the residual defect state.

[0044] The target chamfer specification of this machining is an R0.3mm arc chamfer, the profile dimension tolerance is ±0.02mm, and the surface roughness requirement is no higher than Ra0.2μm. If the verification result meets the target chamfer specification, the machining is determined to be qualified, and the wafer can flow into the next process; If any areas fail to meet the standards, a corresponding re-grinding strategy is generated. These areas are then re-grinded and verified again. The maximum number of re-grinding attempts is set to two. If the standards are still not met after two attempts, the workpiece is marked as awaiting review and transferred to the manual verification station. All verification results are synchronously archived in the corresponding wafer's processing file and stored in association with the previous process data for subsequent process iteration and optimization.

[0045] Example 6: This example discloses a wafer secondary chamfer adaptive polishing system based on edge morphology perception, used to execute the polishing methods described in the above examples. The system includes a wafer transport and loading module, an edge optical detection module, an image processing and risk classification module, an adaptive polishing execution module, and a process monitoring and closed-loop feedback module.

[0046] In the specific implementation process, the wafer transfer and loading module adopts a structure that combines a vacuum adsorption robotic arm with a four-station turntable to sequentially transfer the wafers to be processed to the pre-processing station, the grinding station and the re-inspection station, completing the wafer process flow. A single batch can carry 25 6-inch wafers, and the station switching time does not exceed 3 seconds, meeting the production line capacity requirements.

[0047] The edge optical inspection module is equipped with a high-resolution line scan camera and a ring-shaped tilting light source, supporting manual or automatic switching between bright-field and dark-field inspection modes. It performs 360° full-circle scanning and topographic image acquisition of the wafer edge. The module's vertical working distance from the wafer edge is 80mm, with a working distance tolerance of ±2mm. The image processing and risk classification module, equipped with an FPGA processing chip and a preset defect feature database, receives edge topographic images and extracts defect feature parameters. Based on preset thresholds, it determines the risk level of the edge region and is responsible for coordinate transformation and sector matching of defect locations. The adaptive grinding execution module includes a diamond grinding head and a servo drive control unit. It performs segmented differentiated grinding processes based on the risk classification results. The radial position control accuracy of the grinding head is ±1μm, and the pressure control accuracy is ±0.1N.

[0048] The process monitoring and closed-loop feedback module is equipped with a vibration sensor and an acoustic emission sensor. The sensors are coupled and installed at the base of the grinding wheel spindle for real-time monitoring of the grinding process and contour re-inspection after grinding, realizing full-link status control and quality closed-loop of the processing process.

[0049] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A wafer secondary chamfering adaptive polishing method based on edge morphology perception, characterized in that, Includes the following steps: S1. Wafer loading and edge full-circumference optical pre-scan: The wafer to be subjected to secondary chamfering is transferred to the pre-processing station of the processing equipment. The non-contact optical inspection module is activated. High-resolution line scan imaging combined with dark field detection mode is used to perform a 360° full-circumference scan on the top edge, bottom edge and chamfered side of the wafer, and collect the morphological image data of the wafer's full-circumference edge. S2. Edge defect identification and processing risk classification: The acquired edge morphology image is transmitted to the image processing unit to extract the geometric feature parameters of edge chipping and hidden crack defects in the image. Based on the preset defect judgment threshold system, the corresponding circumferential regions of the wafer edge are divided into three risk levels: no defects or slight defects, moderate defects, and severe defects. S3. Adaptive grinding strategy is dynamically generated. Based on the risk classification results of each region at the edge of the wafer, an appropriate combination of grinding process parameters is matched. Three differentiated processing strategies are generated for edge regions with different risk levels: standard grinding, flexible trimming, skipping or light polishing. S4. Real-time monitoring of the segmented grinding and processing process: According to the generated grinding strategy, the grinding head is controlled to perform grinding processing with corresponding parameters on multiple sectors pre-divided on the edge of the wafer. At the same time, the vibration signal and acoustic emission signal of the grinding wheel are collected in real time throughout the grinding process to monitor the processing status. S5. Post-grinding contour re-inspection and quality closed-loop verification: After all grinding processes are completed, the optical inspection module is restarted to perform a second full-circumference scan of the wafer edge to verify whether the final chamfer contour meets the target specifications. The verification results are then fed back to the process control unit.

2. The wafer secondary chamfering adaptive polishing method based on edge morphology perception according to claim 1, characterized in that, In step S1, a high-resolution line scan camera is used in conjunction with a ring tilting light source to complete the imaging acquisition. For micron-level hidden cracks, a dark field detection mode is used to capture the scattered light generated at the edge defect to form an imaging image of a dark background and a bright defect. During the scanning process, the wafer is controlled to rotate at a constant speed, and the line scanning camera's row acquisition frequency enables continuous circumferential acquisition. The scanning range covers the corner areas of the top edge, bottom edge, and all beveled areas of the side chamfers of the wafer, forming a continuous edge morphology acquisition sequence along the circumference.

3. The wafer secondary chamfering adaptive polishing method based on edge morphology perception according to claim 1, characterized in that, In step S2, the extracted defect geometric feature parameters include the depth and width of the chipped edge, as well as the length and orientation of the crack. Risk classification is performed based on a preset multi-dimensional threshold standard. Among them, no defects or minor defects correspond to defect sizes far below the safety threshold, moderate defects correspond to defect sizes close to the safety threshold, and severe defects correspond to defect sizes exceeding the safety threshold. The classification results are bound one-to-one with the circumferential position of the wafer edge, and each defect is marked with its corresponding circumferential angle and location.

4. The wafer secondary chamfering adaptive polishing method based on edge morphology perception according to claim 1, characterized in that, In step S3, standard grinding parameters are used to process areas with no defects or minor defects. For areas with moderate defects, a flexible dressing strategy is adopted, which reduces the grinding wheel speed, feed rate and downward contact force accordingly. For areas with severe defects, skip processing or perform only very low-speed, light polishing. The parameter combinations for the three types of processing strategies are all pre-stored in the process parameter library. The system directly calls the matching parameter combination based on the risk classification result and the corresponding circumferential position.

5. The wafer secondary chamfering adaptive polishing method based on edge morphology perception according to claim 1, characterized in that, In step S4, the vibration signal and acoustic emission signal collected in real time are compared with the preset normal processing threshold. If an abnormal peak value exceeding the threshold is detected, it is determined that a new edge chipping defect may occur at the current processing position. The system immediately pauses the current grinding operation, controls the optical inspection module to perform a local re-inspection of the abnormal location, readjusts the grinding parameters of the subsequent sectors based on the re-inspection results, and then continues processing. The abnormal location and the re-inspection results are recorded synchronously in the processing log of the corresponding wafer.

6. The wafer secondary chamfering adaptive polishing method based on edge morphology perception according to claim 1, characterized in that, In step S5, the second full-circumference scan uses the same imaging configuration and detection accuracy as the pre-scan, and the verification content includes the arc shape of the chamfer contour, the edge surface quality, and the state of residual defects. If the verification results meet the target chamfer specifications, the processing is deemed qualified. If there are areas that do not meet the standards, a corresponding re-grinding strategy is generated, and the corresponding areas are re-grinded and then verified again. The verification results are simultaneously archived in the processing file of the corresponding wafer.

7. The wafer secondary chamfering adaptive polishing method based on edge morphology perception according to claim 1, characterized in that, Before performing the grinding process, the wafer edge is divided into several equal processing sectors along the circumference. Each sector corresponds to an independent risk classification result and grinding parameter combination. The grinding head automatically switches the corresponding process parameters when processing different sectors. The number of sectors is adjusted according to the wafer size and processing accuracy requirements. The start and end angles of each sector correspond one-to-one with the circumferential position of the defect classification.

8. A wafer secondary chamfering adaptive polishing system based on edge morphology perception, characterized in that, The system is used in the wafer secondary chamfering adaptive polishing method based on edge topography awareness as described in any one of claims 1-7, the system comprising: The wafer transfer and loading module is used to sequentially transfer the wafers to be processed to the pre-processing station, the grinding station and the re-inspection station to complete the wafer process flow. The edge optical inspection module is equipped with a high-resolution line scan camera and a ring-shaped tilted light source, and supports dark field inspection mode for 360° full-circumference scanning and topographic image acquisition of wafer edges. The image processing and risk classification module is used to receive edge morphology images and extract defect feature parameters, and to determine the risk level of the edge region based on a preset threshold. The adaptive grinding execution module, which includes a grinding head and a drive control unit, is used to perform segmented differentiated grinding processes based on risk classification results. The process monitoring and closed-loop feedback module is equipped with vibration sensors and acoustic emission sensors for real-time monitoring of the grinding process and contour re-inspection after grinding.

Images