Edge detection equipment with alignment function

Abstract

The invention discloses edge detection equipment with an alignment function, and belongs to the field of semiconductor manufacturing equipment.The equipment comprises an edge detection module, a straightening mechanism of the edge detection module comprises a suction cup, a rotating motor, a two-dimensional motion module and an edge searching camera, the two-dimensional motion module is composed of an X-axis guide rail, a Y-axis guide rail, a sliding plate and a motor, and wafer translation and rotating alignment are achieved; the optical assembly comprises a camera, a prism assembly, a transverse movement module and a camera fine adjustment mechanism, the prism assembly captures light rays at the upper edge and the lower edge of a wafer through an upper prism and a lower prism, and the light rays are integrated through a 90-degree turning prism to form a composite image containing multi-part information; wafer alignment, edge detection and identification code reading can be completed synchronously, the detection precision and efficiency are improved, and the problems that in an alignment and detection separation mode in existing semiconductor manufacturing, independent detection equipment needs to be additionally configured, the wafer processing time is long, damage is likely to happen after multiple times of transmission, and the yield is low are solved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of semiconductor manufacturing equipment, and particularly relates to an edge detection device with alignment function, mainly applied to the wafer pretreatment link, which can realize alignment and edge quality detection of wafers in the semiconductor chip production process, adapt to wafer pretreatment requirements before key processes such as photolithography, etching and deposition, and be suitable for precise and efficient processing scenarios of large-scale semiconductor wafer production lines. BACKGROUND

[0002] In the semiconductor chip manufacturing process, wafer transmission and alignment are important pre-processes to ensure the accuracy of subsequent key processes such as photolithography, etching and deposition. The core hub connecting the wafer box and the process equipment is responsible for the preliminary positioning and attitude correction of the wafer taken out from the wafer box, to ensure that the wafer can be sent to the subsequent process equipment stage with precise position and angle, avoiding process failure or product yield reduction caused by wafer position deviation.

[0003] The current mainstream alignment system usually adopts mechanical positioning or optical positioning to realize wafer alignment. The mechanical positioning method contacts the wafer edge through positioning pins, blocks and other mechanical structures, and pushes the wafer to move to the preset reference position to complete the alignment; the optical positioning method uses cameras, laser sensors and other optical devices to collect feature information such as wafer edge or notch, flat edge, etc., and drives the mechanical structure to adjust the wafer attitude after calculating the deviation by algorithm to realize alignment.

[0004] In semiconductor manufacturing, wafer edge quality has a significant impact on subsequent processes and product quality. If the edge has cracks, notches and other problems, it will cause particle contamination, lead to photolithographic pattern misplacement, and even cause the wafer to crack. However, the existing technology needs to separately set up an edge detection device, adopts the "alignment first, then detection" mode, and has defects such as increasing equipment cost and space, prolonging processing time, and increasing transmission risk, so designing a system that can simultaneously complete alignment and edge inspection has become a problem to be solved. SUMMARY

[0005] The purpose of the present application is to provide an edge detection device with alignment function, which solves the problem of "separation of alignment and detection" in the existing semiconductor manufacturing process, which requires additional independent detection equipment, long wafer processing time, multiple transmissions, damage and low yield.

[0006] The first purpose of the present application is to provide an edge detection device with alignment function, which includes an edge detection module, the edge detection module includes a positioning mechanism for carrying wafers and an optical assembly for imaging the edge of the wafer; The positioning mechanism includes: a suction cup for adsorbing and carrying the wafer; A rotary motor, whose output shaft is connected with the chuck, is used to drive the chuck to rotate; A two-dimensional motion module, whose motion platform is fixedly connected with the rotary motor, is used to drive the rotary motor and the chuck to move in a horizontal plane; An edge-seeking camera, whose optical axis is directed to the wafer edge on the chuck; The two-dimensional motion module comprises: An X-axis guide rail fixedly arranged; An X-axis sliding plate slidably arranged on the X-axis guide rail and driven by an X-axis module motor; A Y-axis guide rail fixedly arranged on the X-axis sliding plate, the extending direction of the Y-axis guide rail being perpendicular to the X-axis guide rail; A Y-axis sliding plate slidably arranged on the Y-axis guide rail and driven by a Y-axis module motor; The rotary motor is fixedly mounted on the Y-axis sliding plate; The optical assembly comprises a light source, a camera, a prism assembly, a transverse motion module and a camera fine adjustment mechanism; The light source irradiates the wafer edge region in an oblique incidence manner, achieving dark field illumination or low angle illumination; The transverse motion module is used to drive the image capturing mechanism as a whole to move along a direction parallel to the radial direction of the wafer; The camera fine adjustment mechanism is used to finely adjust the camera in multiple dimensions; The prism assembly comprises an upper prism and a lower prism, the upper prism and the lower prism being correspondingly arranged above and below the wafer edge respectively, and being used to capture light from the upper edge surface and the lower edge surface of the wafer respectively; The prism assembly further comprises a 90-degree turning prism, the 90-degree turning prism being arranged behind the light paths of the upper prism and the lower prism, and being used to turn and guide the light from the light paths of the upper prism and the lower prism to the camera; The device further comprises a code reading mechanism, the code reading mechanism being arranged near the alignment mechanism, and being used to read the identification code of the wafer.

[0007] The second invention of the present application aims to provide an edge detection method with alignment function, comprising the following steps: S1 wafer loading and initial fixing: the wafer is transferred and placed on the chuck of the alignment mechanism by a robot, and then vacuum suction is started to fix the wafer; S2 wafer precise centering step: the rotary motor is controlled to drive the chuck to rotate, while the edge-seeking camera is used to capture the image of the wafer edge, the pose deviation of the wafer center is calculated based on the image, and the two-dimensional motion module is driven to move to correct the deviation; S3 identity recognition step: the wafer is controlled to rotate to a preset angle, so that the identification code area of the wafer edge is aligned with the code reader, the code reader is triggered to perform image acquisition and optical character recognition (OCR) on the identification code, and the unique serial number of the wafer is obtained; S4 optical path self-calibration step: a fixed reference mark arranged in the optical path of the prism assembly is imaged by the camera pair, a compensation signal is generated based on the deviation of the current image pose of the reference mark from the standard pose, and the camera is physically adjusted or the subsequently collected wafer image is compensated by software coordinate transformation according to the compensation signal; S5 detection distance adjustment: according to the specification information of the detected wafer, the corresponding preset parameter is called, and the preset parameter is used to drive the transverse motion module to move to adjust the object distance between the imaging mechanism and the wafer edge to the best imaging position; S6 edge imaging step: driving the wafer to rotate, and guiding the light of the multiple surfaces of the wafer edge in the prism assembly in the optical assembly to the camera for imaging to obtain a composite image of the wafer edge; S7 defect analysis step: processing the composite image, segmenting the sub-image regions corresponding to the wafer front chamfer, edge vertical surface, circumferential surface, back vertical surface and back chamfer from the single composite image, and performing defect analysis respectively; S8 discharging and data archiving: making a pass / fail judgment on the wafer according to the defect analysis result; vacuum releasing, and the robot taking away and sorting the wafer, and saving the detection data.

[0008] Further, the S6 edge imaging step includes an S6 adaptive sampling process: in the wafer rotation scanning process, the collected images are analyzed in real time; when no defect risk feature is identified in the continuous area, a first sampling mode is used for detection; when a defect risk feature is identified, a second sampling mode is switched to for detecting the current risk area, and the sampling density of the second sampling mode is higher than that of the first sampling mode.

[0009] Compared with the prior art, the wafer clamping mechanism provided by the application has the following remarkable and beneficial technical effects: (1) The application realizes the integration of alignment and detection, integrates an optical detection assembly in the alignment system, does not need to additionally configure a separate edge detection device, synchronously completes alignment and edge inspection, reduces the procurement cost of semiconductor production line equipment, reduces the space occupation of the clean room, solves the problems of high cost and large occupation of traditional separate equipment, helps compact layout of the production line, and improves the space utilization.

[0010] (2) The wafer does not need to be additionally transmitted between the alignment and detection devices, the pretreatment cycle is greatly shortened, the production rhythm is accelerated, the large-scale production demand is met, the total processing time of alignment and edge inspection is shortened compared with the "first alignment, then detection" mode, the semiconductor production efficiency is improved, and the efficiency bottleneck of the traditional process is broken through.

[0011] (3) The present application reduces the number of wafer transmission times, reduces the frequency of contact with the transmission mechanism, reduces friction particle pollution, avoids scratches and cracks caused by multiple transmissions, reduces damage rate in the pretreatment stage, improves product yield, and the optical detection assembly is matched with a high-resolution camera and a deep learning algorithm, the defect recognition is accurate, the closed-loop control precision of the alignment mechanism is high, the system is reliable and stable, and a foundation is laid for subsequent processes. BRIEF DESCRIPTION OF DRAWINGS

[0012] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments or prior art description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0013] Fig. 1 The structural diagram of the edge detection device with alignment function of the embodiment of the present application; Fig. 2 The front view of the edge detection device with alignment function of the embodiment of the present application; Fig. 3 The side view of the edge detection device with alignment function of the embodiment of the present application; Fig. 4 The top view of the edge detection device with alignment function of the embodiment of the present application.

[0014] Explanation of reference signs: 5.1 - alignment mechanism; 5.2 - code reading mechanism; 5.3 - transverse motion module; 5.4 - adapter plate; 5.5 - upper prism; 5.6 - lower prism; 5.7 - light source; 5.8 - 90-degree turning prism; 5.9 - camera; 5.10 - camera fine adjustment mechanism; 5.11 - edge finding camera. DETAILED DESCRIPTION

[0015] In the following description, a large number of specific details are given in order to provide a more thorough understanding of the present application. However, it is obvious to those skilled in the art that the present application can be implemented without one or more of these details. In other examples, some technical features known in the art are not described in order not to obscure the present application.

[0016] In order to thoroughly understand the present application, detailed steps and detailed structures will be proposed in the following description in order to explain the technical solutions of the present application. The preferred embodiments of the present application are described in detail as follows, however, in addition to these detailed descriptions, the present application can have other implementation manners.

[0017] Embodiment 1 As Figs. 1 to 4The core innovation of the edge detection device with alignment function of the present application is a highly integrated edge detection module. The edge detection module deeply integrates high-precision mechanical centering mechanism, reliable identity recognition system and an innovative single-camera multi-surface optical imaging technology, and introduces intelligent self-calibration and adaptive detection algorithm, thereby reducing the complexity, cost and maintenance difficulty of the system while ensuring high detection precision.

[0018] The positioning mechanism 5.1 is the starting point of the entire detection process, and its positioning accuracy directly determines the accuracy and repeatability of the subsequent edge imaging and defect detection. The mechanism is not simply a wafer carrier, but a complex system integrating precise mechanical movement, machine vision and real-time control technology. The positioning mechanism 5.1 includes a chuck, a rotary motor, a two-dimensional motion module and an edge finding camera 5.11.

[0019] The chuck is a direct carrier of the wafer. Its top surface is usually distributed with complex vacuum suction channels. When the wafer is placed on the chuck, the vacuum system is started to generate negative pressure through these small holes to firmly and smoothly adsorb the wafer on the surface of the chuck. The body of the chuck is usually made of anodized aluminum alloy or stainless steel, etc. with good rigidity and high thermal stability to ensure that its deformation is minimal under stress and environmental temperature changes. The theoretical rotation center of the chuck, i.e. its geometric center, is the reference for subsequent centering operations.

[0020] The rotary motor is a power source for driving the chuck and wafer rotation. In a preferred embodiment of the present application, the rotary motor preferably adopts a torque motor. Compared with the traditional "servo motor + speed reducer" scheme, the torque motor adopts a direct drive method, and the rotor is directly connected to the main shaft of the chuck, eliminating intermediate transmission links such as gears and belts, thereby bringing multiple advantages: it can achieve high motion control precision and positioning accuracy without reverse backlash; it can maintain stable rotation at very low speed, effectively avoiding the crawling phenomenon, which is crucial for the edge finding camera 5.11 to obtain clear continuous images; its high rigidity and low vibration characteristics result from direct drive to avoid elastic deformation of the transmission chain, thereby ensuring the stability of the imaging system; at the same time, due to the absence of mechanical contact and wear, high cleanliness is achieved, no wear particles are generated, meeting the cleanliness requirements of semiconductor manufacturing. In addition, the rotation angle of the motor is fed back to the control system in real time by a high-precision encoder, thereby realizing accurate speed and position control.

[0021] The two-dimensional motion module precisely adjusts the position of the chuck and wafer in the horizontal plane (i.e., XY plane) to compensate for the centering deviation caused by the initial placement of the wafer. It is essentially a high-precision cross slide. The bottom layer is the fixed X-axis guide rail, which is firmly installed on the base plate of the device through a bracket. The X-axis slide plate is seated on the X-axis guide rail through a high-precision linear slide. The X-axis module motor drives a ball screw through a coupling, or converts rotary motion into linear motion of the X-axis slide plate along the X-axis guide rail. Above the X-axis slide plate, the Y-axis guide rail is fixedly installed, with its extension direction strictly perpendicular to the X-axis. The Y-axis slide plate is also installed on the Y-axis guide rail through a slide block, and is driven by the Y-axis module motor to move along the Y-axis. The rotary motor is fixedly installed on the Y-axis slide plate. By controlling the coordinated motion of the X-axis module motor and the Y-axis module motor, the rotary motor, the chuck, and the wafer thereon can be precisely moved in any trajectory in the XY plane. The precision level of the guide rail and the screw, and the control resolution of the motor together determine the final positioning accuracy of the centering system, which is usually in the micron level or even sub-micron level.

[0022] The edge finding camera 5.11 is used to non-contact measure the position of the wafer edge, which is a high-resolution digital camera equipped with a telecentric lens to ensure that the imaging has no perspective error. The edge finding camera 5.11 is precisely installed so that its optical axis is perpendicular to the tangent direction of the wafer edge, and is focused on the edge region of the wafer. Its field of view size can cover the wafer edge and still capture the edge image when there is the maximum expected deviation in the center.

[0023] The centering process is as follows: the wafer is transferred by an upstream robot or automated rail and initially placed on the chuck of the centering mechanism 5.1. The vacuum system is activated to fix the wafer. At this time, the geometric center of the wafer is usually offset from the theoretical rotation center of the chuck by several hundred microns or even more. The control system controls the rotation motor to drive the wafer to rotate at a constant and low speed. At the same time, the edge finding camera 5.11 continuously captures images of the wafer edge at a fixed frequency. The image processing unit runs the edge extraction algorithm in real time to accurately locate the position of the wafer edge in the camera coordinate system from each frame of image. Since the wafer is rotating, the motion trajectory of a point on the edge of the wafer in the field of view of the camera 5.9 should theoretically be a perfect circle. If the wafer center is offset, the actual center of this trajectory circle will deviate from the image center. By collecting the coordinates of the edge points in multiple images during rotation, at least 3 non-collinear points, and using curve fitting algorithms such as least squares, the center coordinates (Cx, Cy) and radius R of the trajectory circle can be accurately calculated. The offset (Δx, Δy) of this center coordinates (Cx, Cy) relative to the image center (0, 0) is converted through the camera 5.9 calibration parameters to obtain the XY direction deviation of the wafer geometric center relative to the chuck rotation center in the device coordinate system. The control system uses the calculated deviation (Δx, Δy) as a command to drive the X-axis module motor and Y-axis module motor to move the corresponding distance. The two-dimensional motion module moves the chuck and the wafer as a whole, so that the geometric center of the wafer coincides with the rotation center of the chuck. After centering is completed, the wafer can be rotated again for a full circle, and the edge finding camera 5.11 verifies whether the edge trajectory is a circle with the image center as the center to ensure that the centering accuracy meets the requirements.

[0024] After centering is completed, or when the wafer is rotated to a specific angle, the code reading mechanism 5.2 begins to work, and its task is to read the unique identification code, usually an alphanumeric sequence, engraved on the wafer edge at a specific position. The code reading mechanism 5.2 preferably uses an OCR optical character recognition reader. The reader usually integrates a specific wavelength illumination source and a small high-resolution camera. The illumination source illuminates the identification code area at an appropriate angle and brightness to avoid strong reflection. After the camera captures a clear character image, it performs image preprocessing, character segmentation and recognition through the built-in OCR algorithm, and finally converts the image into a sequence number in text format. The recognized sequence number is uploaded to the upper manufacturing execution system or local database, and is bound with all image data and defect information generated by the current edge detection. This ensures the traceability of the entire production process, and once a problem is found in the subsequent process, the edge detection results of the specific wafer can be quickly and accurately traced back.

[0025] After the wafer is neutralized and read, the edge of the core enters the edge optical detection link. The optical assembly is responsible for imaging the wafer edge, mainly including light source 5.7, camera 5.9, prism assembly, transverse motion module 5.3 and camera fine adjustment mechanism 5.10.

[0026] The transverse motion module 5.3 is a high-precision electric linear module, and the motion axis is accurately adjusted to be parallel to the radial direction of the wafer. The entire imaging mechanism is installed on the slider of the module through a solid adapter plate 5.4. By controlling the module motion through the program, the imaging mechanism can adapt to the detection needs of wafers of different diameters, and the object distance can be finely adjusted, so that the wafer edge is always within the best depth of field of the camera 5.9, and the clearest image is obtained.

[0027] In order to realize high-quality imaging, the present application sets a light source 5.7 near the edge of the wafer, which is preferably an LED light source with adjustable brightness and uniform light emission, and the emitted light is accurately directed to the edge region of the wafer; the preferred illumination method is dark field illumination or low angle illumination, that is, the light line illuminates the edge at a large inclination angle, so that the light reflected by the smooth surface cannot enter the camera, and the defects (such as edge collapse, cracks, contamination) will scatter the light into the camera 5.9, thereby forming a bright defect image on a dark background, which improves the signal-to-noise ratio and contrast of the defect, which is the premise of obtaining a clear and high-contrast composite image., and the unique prism combination together constitutes the optical path system for realizing single-camera multi-surface imaging.

[0028] After the equipment is assembled or maintained, the pose of the camera 5.9 needs to be finely calibrated through the camera fine adjustment mechanism 5.10. The present application preferably adopts a six-dimensional adjustment frame. The mechanism provides six degrees of freedom for fine adjustment: translation along the X, Y and Z axes, and rotation around the X, Y and Z axes. By precisely adjusting the screws, the position and angle of the camera 5.9 can be finely adjusted to ensure that the camera sensor plane is perpendicular to the optical path, and the optical axis coincides with the optical axis of the entire prism system. This is the basis for obtaining distortion-free and high-contrast images.

[0029] The edge region of the wafer to be detected usually includes five surfaces: the top front chamfer, the front edge vertical surface, the outermost circumferential surface, the back edge vertical surface and the bottom back chamfer. Traditionally, 3-5 sets of independent cameras and optical systems are required to complete the detection work, and the optical assembly of the present application realizes the ingenious folding and synthesis of the optical path through its unique prism combination, which can be completed with only one set of camera and one set of prism.

[0030] The upper prism 5.5 is arranged above the wafer edge. Its specific prism angle and coating design enable it to effectively capture scattered or reflected light from the upper edge surface of the wafer, mainly including the front chamfer and part of the edge surface, and refract the light into a specific exit direction. Similarly, the lower prism 5.6 is arranged below the wafer edge and is responsible for capturing light from the lower edge surface, mainly including the back chamfer and part of the edge surface, and refracting the light. Through precise optical design, the upper prism 5.5 and the lower prism 5.6 guide the originally spatially separated light paths they capture to a common, virtual imaging plane, i.e., the information from the upper and lower edges of the wafer is optically "stitched" together. The combined light path from the upper prism 5.5 and the lower prism 5.6 is horizontally propagating. In order to adapt to the usual vertical installation of the camera 5.9, the light path needs to be turned 90 degrees. The 90-degree turning prism 5.8 uses the principle of total reflection to accurately deflect the horizontally incident light by 90 degrees, changing it to propagate in the vertical direction.

[0031] This turning not only changes the direction of the light path, but more importantly, it further integrates the light paths from the upper prism 5.5, the lower prism 5.6, and the outermost circumferential surface and directs them to the only imaging element, the camera 5.9. A suitable imaging lens is installed in front of the camera 5.9. After the above complex light path folding and synthesis, the light information from the five surfaces of the wafer edge finally converges on the sensor target surface of the camera 5.9, forming a single, synthesized image containing all the information of the surfaces to be inspected. In the obtained synthesized image, different edge regions will usually appear as bright or dark bands at different positions in the image. Through image processing algorithms, sub-images corresponding to the front chamfer, edge surface, and circumferential surface can be accurately separated from this image and analyzed for defects, respectively. The present invention uses only one set of cameras and prisms to complete the work that traditionally requires 3-5 independent camera and light path systems, simplifying the mechanical structure, reducing hardware cost, calibration cost, and maintenance cost, while avoiding the synchronization and calibration problems between multiple cameras, significantly improving system reliability and consistency.

[0032] During long-term operation of the device, due to mechanical stress release, environmental temperature fluctuations, or slight vibrations, the relative positions of the optical components may slowly drift by a few microns, leading to a decrease in imaging quality, changes in detection standards, and ultimately affecting the accuracy and consistency of the detection results.

[0033] In some preferred embodiments, to further improve the long-term stability of the device, the present application also integrates a smart control unit, which includes an optical path online automatic compensation system. The system includes a small, shape-known fixed reference mark arranged in the optical path of the prism assembly. The fixed reference mark is arranged in a stable and less movable position in the optical path, such as etched or attached on a certain non-light-passing surface of the 90-degree turning prism 5.8. The system can be set to automatically perform the calibration process at the start of each day, at regular intervals, or after each batch of wafers is detected. In the calibration mode, the camera 5.9 images the reference mark. The image analysis algorithm accurately calculates the center coordinates, rotation angle and other pose parameters of the mark in the current image, and compares them with the "standard pose" stored when the system is initially calibrated, calculates the deviation, including the translation deviation and rotation deviation in the X and Y directions, to determine whether the optical path has deviated.

[0034] The compensation execution unit generates a compensation signal based on the calculated deviation. The compensation can be achieved in two ways: one is active compensation. That is, the camera fine adjustment mechanism 5.10 is driven to perform physical calibration to adjust the pose of the camera 5.9. The second is passive compensation. That is, all subsequent collected wafer edge images are compensated by software coordinate transformation, and the image position is corrected. When active compensation is used, the compensation signal is sent to the camera fine adjustment mechanism 5.10, which adjusts the spatial pose of the camera 5.9 in the opposite direction according to the instructions, thereby offsetting the detected optical path deviation and restoring the image of the reference mark to the standard pose. This method can correct the error from the root. Passive compensation is used in some scenarios where frequent physical adjustment is not convenient or the cost of the mechanism is limited. The system converts the calculated deviation into a coordinate transformation matrix, which is applied to all subsequent collected wafer edge images before defect analysis. This is equivalent to "pulling back" the image to the correct position at the digital level. This method is low in cost and fast, but it will not change the actual optical path.

[0035] Uniform high-resolution and high-frame-rate scanning of the entire circumference of the wafer will generate a large amount of data, which will take a long time to process and affect the overall detection throughput. However, defects on the edge of the wafer are not usually uniformly distributed, and most areas are perfect.

[0036] In some preferred embodiments, to further improve the detection efficiency of the apparatus, the intelligent control unit of the present application comprises an adaptive sampling detection module. The adaptive sampling detection module is used to optimize the detection efficiency. During the wafer rotation scanning process, the adaptive sampling detection module analyzes the images collected in real time. If no defect features are identified in the continuous area, the camera 5.9 and / or the rotation motor is controlled to enter the "fast scan" mode with a low sampling rate. Once the image analysis algorithm identifies potential defect risk features, such as a sudden change in local contrast of the image, an abnormality in a specific texture pattern, or a cluster of pixels exceeding a preset gray scale threshold, the module immediately triggers a switch to the "fine scan" mode with a high sampling rate to conduct a detailed inspection of the risk area. After the risk area is scanned, the "fast scan" mode is restored.

[0037] When starting detection, the system scans at a high sampling rate to ensure that the finest defects can be captured. The image processing algorithm analyzes the latest continuous image sequence in real time and extracts features. A perfect edge area usually exhibits uniform gray scale or a specific texture pattern. A sudden change in local contrast of the image is caused by a clear light-dark edge at the edge collapse or crack; an abnormality in a specific texture pattern is caused by pollution that can destroy the uniformity of the texture; and a cluster of pixels exceeding a preset gray scale threshold is caused by abnormal reflection or attachments that form abnormal bright or dark spots.

[0038] When a certain length of the circumferential area, such as 90 degrees of rotation angle, is continuously analyzed and no risk features are found, the system determines that the area is "low risk". Subsequently, the adaptive sampling detection module automatically switches to the "fast scan" mode. In this mode, the area can be quickly passed through by reducing the sampling frame rate of the camera 5.9 and / or increasing the wafer rotation speed. As soon as real-time image analysis identifies any risk features, the adaptive sampling detection module will immediately trigger a switch back to the "fine scan" mode. The camera 5.9 is restored to a high frame rate, and the wafer rotation speed is reduced or even paused for a moment to ensure the most sufficient and clear imaging of the suspicious area. The algorithm will conduct intensive sampling and in-depth analysis of this area. To ensure no missed detection, the triggering algorithm of the mode switch will set a certain lead and lag area to ensure safety before entering "fast scan" and still scan a short distance in "fine scan" mode after leaving the risk area. This adaptive sampling strategy can reduce the collection and processing of redundant data and improve the overall detection efficiency without sacrificing defect detection rate.

[0039] Example 2 Based on the above device embodiment, the application also correspondingly provides an edge detection method with alignment function. Through programmed control, the method cooperatively mobilizes each module of the device to realize the full automation process of wafer edge from automatic feeding, precise centering, identity recognition to high-speed and high-precision defect detection. The method embodiment not only covers the basic operation steps, but also integrates the advanced functions of the above intelligent control unit, reflecting the intelligence and efficiency of the method.

[0040] An edge detection method with alignment function is executed by using the edge detection device with alignment function as described above, and the method comprises the following steps: S1 wafer feeding and initial fixation: through an industrial robot or an automatic material conveying system, the wafer to be detected is taken out from an upstream station or a carrier and placed stably in the center area of the suction disc of the centering mechanism 5.1. After the placement is completed, the control system triggers the vacuum generator to generate negative pressure on the suction disc, so that the wafer is firmly adsorbed on the surface of the suction disc, and the initial fixation is completed.

[0041] S2 wafer precise centering step: the control system instructs the rotary motor to drive the suction disc and the wafer to rotate at a preset, low and stable speed. At the same time, the edge finding camera 5.11 is triggered to continuously shoot the wafer edge image passing through its field of view at a high frame rate. The image processing unit processes the collected image sequence in real time. First, the image filtering algorithm is used to reduce noise, and then the edge detection operator is used to accurately extract the position coordinates of the wafer edge in each image at the sub-pixel level. Since the wafer is rotating, the motion trajectory of the edge point of the wafer in the coordinate system of the camera 5.9 should be a circle. Through a curve fitting algorithm such as least squares method, the edge point set extracted from continuous multiple frames is fitted to a circle, and the actual center coordinates (Cx, Cy) and the radius R of the current wafer edge trajectory are calculated. Compare the actual center with the image center, and convert it through the camera calibration parameters to obtain the deviation amount (ΔX, ΔY) of the geometric center of the wafer relative to the rotation center of the suction disc in the device XY coordinate system. The control system takes the calculated deviation amount (ΔX, ΔY) as the motion instruction to drive the X-axis module motor and Y-axis module motor of the two-dimensional motion module to move cooperatively, drive the suction disc and the wafer to move in the horizontal plane, and accurately compensate the center deviation. After the movement is completed, a verification scan is performed to confirm that the deviation has been eliminated, and to ensure that the geometric center of the wafer coincides with the rotation center of the suction disc.

[0042] S3 identity recognition step: after the completion of the middle, the control system controls the rotating motor to rotate the wafer to a preset angle, so that the identification code area engraved on the edge of the wafer is opposite the code reading mechanism 5.2, preferably an OCR code reader. The OCR code reader triggers the illumination light source and takes a picture, identifies the characters in the image through the built-in OCR algorithm, and obtains the unique serial number of the wafer. The serial number is immediately uploaded and bound with all subsequent detection data of this wafer, stored in the database, and the whole process quality traceability is realized.

[0043] S4 optical path self-calibration step: the control system moves the image taking mechanism to the "Home" position away from the wafer through the lateral motion module 5.3, so that the camera field of view is aligned with the fixed reference mark. The camera 5.9 images the reference mark, and the image analysis algorithm calculates the center position, angle and other parameters of the current mark image, and compares them with the "standard pose" of the initial calibration to obtain the optical path deviation. According to the deviation, a compensation signal is generated, and the camera fine adjustment mechanism 5.10 is driven to move, directly adjusting the camera pose and physically correcting the optical path; or the calculated deviation is stored as an image coordinate transformation parameter, and in subsequent detection, each frame of wafer image collected is first subjected to corresponding geometric transformation, and then subjected to defect analysis. In order to cope with the possible small drift of the optical path during long-term operation of the equipment, this step is performed before the start of each batch of wafer detection or at regular intervals.

[0044] S5 detection distance adjustment: the control system calls the preset parameters according to the specifications of the wafer, drives the lateral motion module to move, so that the entire image taking mechanism moves along the radial direction of the wafer, and adjusts to the best imaging object distance suitable for the wafer of this type, ensuring that the edge area is within the best depth of field of the camera lens.

[0045] S6 edge imaging step: drive the wafer to rotate at a constant speed, the upper prism 5.5 and the lower prism 5.6 capture the light rays of the upper and lower edge surfaces of the wafer respectively, and after being synthesized through the 90-degree turning prism 5.8, they are guided to a single camera 5.9, so as to obtain a synthesized image containing information of the wafer front chamfer, edge vertical surface, circumferential surface, back vertical surface and back chamfer on the sensor of the camera 5.9.

[0046] S61 Adaptive Sampling: The adaptive sampling detection module runs in the background in real time. It performs rapid analysis on the latest collected continuous image frames, extracting key features such as local contrast, texture uniformity, and gray scale distribution. Within the adaptive sampling detection module, risk feature criteria are pre-set, such as sudden changes in image local contrast, abnormalities in specific texture patterns, and the appearance of clusters of pixels exceeding the pre-set gray scale threshold. If, within a certain angle of continuous scanning area, the analysis results do not trigger any risk feature criteria, the adaptive sampling detection module determines that the area is "low risk". Subsequently, the system automatically switches to "fast scanning mode", quickly passing through this "clean" area by reducing the sampling frame rate of the camera 5.9 and / or appropriately increasing the wafer rotation speed. This significantly reduces the amount of image data that needs to be processed and stored, improving throughput. As soon as real-time image analysis identifies any potential risk features, the adaptive sampling detection module will immediately issue instructions to instantly switch to "fine scanning mode". In this mode, the system will increase the camera 5.9 sampling frame rate to the highest, and may simultaneously reduce the wafer rotation speed or even temporarily pause, to ensure the most complete and clear imaging of the suspicious area, capturing possible minor defects. When the fine scanning of the risk area is complete, and consecutive frames of image analysis show that the risk features have disappeared, the adaptive sampling detection module will control the system to switch back to "fast scanning mode" and continue efficient detection.

[0047] S7 Defect Analysis Step: First, using prior knowledge such as the relative positions of each edge surface in the composite image, five sub-image regions corresponding to the wafer front chamfer, edge vertical surface, circumference surface, back vertical surface, and back chamfer are accurately segmented from each frame of the composite image through threshold segmentation, contour extraction, and other algorithms. For each sub-image region, a special defect detection algorithm is used for analysis. For example: for edge collapse and cracks, edge detection combined with morphological operations can be used to identify geometrically abnormal protrusions or depressions. For contamination, texture analysis or spot detection algorithms can be used to identify areas inconsistent with the background texture. Detected defects are classified according to their morphology, size, contrast, etc., such as edge collapse, cracks, and contamination, and their circumferential angular position, size, etc. information is recorded and bound to the wafer ID.

[0048] S8 Discharging and Data Archiving: Based on pre-set defect tolerance standards such as maximum allowed defect number, size, and type, the edge detection results of the entire wafer are comprehensively judged, outputting a "pass" or "fail" conclusion. All raw image data, processing results, defect maps, and final determination conclusions are associated with the wafer ID and archived in the database. After the detection process is completed, the vacuum of the chuck is released. The industrial robot removes the detected wafer from the alignment mechanism 5.1 according to the system instructions, and sorts it into different downstream bins or stations according to the determination result, completing the entire automated detection cycle.

[0049] In summary, the method embodiment fully utilizes the advantages of the equipment through a series of precise control and analysis steps, especially realizes the efficient detection process of single imaging, multi-surface synchronous detection and separate analysis, and can further improve the stability and efficiency through self-calibration and adaptive sampling.

[0050] The preferred embodiments of the present application are described above. It should be understood that the present application is not limited to the specific embodiments described above, nor is it limited to the edge detection device with alignment function. The devices and structures not described in detail should be understood as being implemented in the ordinary way in the art. Any person skilled in the art can make many possible changes and modifications to the technical solutions of the present application or modify them into equivalent embodiments with equivalent changes without departing from the scope of the technical solutions of the present application. Therefore, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present application, which does not depart from the content of the technical solutions of the present application, still belongs to the scope of protection of the technical solutions of the present application.

Claims

1. An edge detection device with alignment function, comprising an edge detection module, the edge detection module comprising a centering mechanism for carrying a wafer and an optical assembly for imaging the edge of the wafer, characterized in that: the centering mechanism comprises: a chuck for adsorbing and carrying the wafer; a rotary motor, the output shaft of which is connected to the chuck, for driving the chuck to rotate; a two-dimensional motion module, the motion platform of which is fixedly connected to the rotary motor, for driving the rotary motor and the chuck to move in a horizontal plane; an edge-finding camera, the optical axis of which is directed to the edge of the wafer on the chuck; the two-dimensional motion module comprises: an X-axis guide rail fixedly arranged; an X-axis slide plate slidably arranged on the X-axis guide rail and driven by an X-axis module motor; a Y-axis guide rail fixedly arranged on the X-axis slide plate, the extension direction of the Y-axis guide rail being perpendicular to the X-axis guide rail; a Y-axis slide plate slidably arranged on the Y-axis guide rail and driven by a Y-axis module motor; wherein the rotary motor is fixedly mounted on the Y-axis slide plate; the optical assembly comprises a light source, a camera, a prism assembly, a transverse motion module, and a camera fine adjustment mechanism; the light source irradiates the edge region of the wafer in an oblique incidence manner, realizing dark field illumination or low angle illumination; the transverse motion module is used for driving the whole imaging mechanism to move along the radial direction of the wafer; the camera fine adjustment mechanism is used for fine adjustment of the camera in multiple dimensions; the prism assembly comprises an upper prism and a lower prism, the upper prism and the lower prism being correspondingly arranged above and below the edge of the wafer, respectively, for capturing light from the upper edge surface and the lower edge surface of the wafer, respectively; the prism assembly further comprises a 90-degree turning prism, the 90-degree turning prism being arranged behind the light paths of the upper prism and the lower prism, for turning and guiding the light from the light paths of the upper prism and the lower prism to the camera; the device further comprises a code reading mechanism, the code reading mechanism being arranged near the centering mechanism, for reading the identification code of the wafer.

2. The edge detection apparatus having an alignment function according to claim 1, characterized by: The rotary motor is a torque motor.

3. The edge detection apparatus having an alignment function according to claim 1, characterized by: The code reading mechanism is an OCR optical character recognition code reader.

4. The edge detection apparatus having an alignment function according to claim 1, characterized by: The device further comprises an intelligent control unit, the intelligent control unit comprising an optical path online automatic compensation system for real-time monitoring and compensation of optical path deviation of the optical assembly.

5. The edge detection device having an alignment function according to claim 4, characterized in that: The optical path online automatic compensation system comprises a fixed reference mark arranged in the light path of the prism assembly; the optical path online automatic compensation system acquires an image of the fixed reference mark by the camera, and generates a compensation signal based on the deviation between the current pose and the standard pose of the image.

6. The edge detection device having an alignment function according to claim 5, characterized in that: The compensation signal is used to drive the camera fine adjustment mechanism to perform physical calibration, or to perform software coordinate transformation compensation on the subsequent acquired wafer image.

7. The edge detection apparatus having an alignment function according to claim 1, characterized by: The device further comprises an intelligent control unit, the intelligent control unit comprising an adaptive sampling detection module, the adaptive sampling detection module dynamically switching the camera sampling frame rate and / or the wafer rotation speed according to the defect feature analysis result of the already acquired image during the wafer edge scanning process.

8. The edge detection device having an alignment function according to claim 1, characterized in that: The camera fine adjustment mechanism is a six-dimensional adjustment frame for adjusting the translation and rotation of the camera in three-dimensional space.

9. The edge detection device having an alignment function according to claim 4, characterized in that: The defect features include image local contrast mutation, specific texture pattern anomaly, or pixel point cluster exceeding the preset gray threshold.

10. The edge detection device having an alignment function according to claim 1, characterized in that: The upper and lower prisms guide the captured light path to the same imaging plane, and after integration through a 90-degree turning prism, a composite image containing wafer front face chamfer, edge facade, circumferential surface, back facade and back chamfer information is formed on the camera sensor.

Images