Method for detecting pit point defects of silicon carbide substrate polished wafer

Abstract

The invention provides a method for detecting pit defects of a polished wafer of a silicon carbide substrate. The method comprises the steps of feeding, mechanical conveying, edge searching, OCR recognition, conveying to a detection table, first-time reflection light source detection and transmission light source detection, second-time reflection light source detection and data analysis. The first light source reflection detection comprises the following steps: irradiating the upper surface of the silicon carbide substrate polished wafer by using a reflection light source, and collecting a reflection light signal and a scattering light signal; the transmission light source detection comprises the following steps of: enabling transmission light to enter from the lower surface of the polished surface of the silicon carbide substrate through a polarizer and exit from the upper surface of the polished surface through a polarization analyzer, and collecting a transmission light signal; and second-time reflected light source detection: shielding the lower surface of the silicon carbide substrate polished wafer so that the light cannot penetrate through the lower surface, then irradiating the upper surface of the silicon carbide substrate polished wafer by using the reflected light source, and collecting a reflected light signal and a scattered light signal. The detection method provided by the invention can improve the precision, efficiency and reliability of pit defect detection.

Description

Technical Field

[0001] The invention belongs to the technical field of semiconductors, and in particular relates to a method for detecting pit defects of a polishing sheet of a silicon carbide substrate. Background Art

[0002] Silicon carbide is widely used in electronic devices, new energy vehicles, aerospace systems and other fields due to its excellent physical and chemical stability, and the application market of silicon carbide substrate polishing sheets has been greatly expanded. However, during the production process of silicon carbide substrate polishing sheets, due to the influence of various factors, various defects such as pits may appear on the surface. These defects not only affect the performance and quality of the product, but also may cause safety hazards in the use of devices. Therefore, higher requirements are put forward for the pit detection technology of surface defects of silicon carbide substrate polishing sheets.

[0003] At present, it is rare to detect surface pits on bare wafers without patterns in the market, especially on the surface defects of polished silicon carbide substrates. Since the polished silicon carbide substrate is transparent or semi-transparent, the surface is atomically smooth, and the smallest pit size is only 3-4um, taking a conventional 6-inch silicon carbide as an example, it is necessary to detect the surface defects of about 1.77*10 10 um 2 , roughly estimated to be 17.7 billion square microns. Considering the requirements of industrial production for test capacity, it generally takes 3-5 minutes to complete the test, and the upper and lower surfaces of the wafer are tested simultaneously, which requires a coordinated match between test efficiency and accuracy, and is highly difficult. At present, no commercial group in the industry has been able to complete the upper and lower surface pit test within 3 minutes at the same time, and ensure its accuracy, repeatability, and stability, which poses a severe test to the product circulation cycle of polished silicon carbide bare wafers.

[0004] Moreover, pits are easily affected by dirt particles on the surface, and it is difficult to distinguish whether they are pits or dirt. Since the silicon carbide substrate polishing sheet is a transparent sheet and the hydrophilicity and hydrophobicity of the surface are different from those of silicon sheets, the surface is more likely to absorb pollutants, affecting the identification and judgment of pits.

[0005] The current mainstream detection method for surface defects in the processing of silicon carbide substrate polishing sheets is manual combined with microscope detection. The detection of surface pit defects mainly relies on manual visual measurement for grade screening and diversion. This method is inefficient and easily affected by uncontrollable factors, resulting in inaccurate detection results. For sizes or morphologies that cannot be estimated manually, it is necessary to use an optical microscope again to confirm the type and size of the defect. During this period, it is necessary to continuously adjust the position, accuracy, clarity, magnification of the objective lens and eyepiece, etc. The operation is complicated, the efficiency is low, and the accuracy is low. At the same time, the silicon carbide substrate polishing sheet will be damaged due to manual operation during the process. Summary of the invention

[0006] The purpose of the present invention is to provide a method for detecting pit defects of a silicon carbide substrate polishing sheet. The detection method of the present invention can improve the accuracy, efficiency and reliability of pit defect detection of a silicon carbide substrate polishing sheet.

[0007] The present invention provides a method for detecting pit defects of a silicon carbide substrate polishing sheet, including material loading, mechanical transmission, edge finding, OCR recognition, transmission to a detection table, first detection, second detection and data analysis;

[0008] The first detection includes a first reflected light source detection and a transmitted light source detection;

[0009] The first light source reflection detection comprises the following steps:

[0010] Using a reflected light source to illuminate the upper surface of the silicon carbide substrate polishing sheet, and collecting reflected light signals and scattered light signals;

[0011] The transmission light source detection comprises the following steps:

[0012] The transmitted light emitted by the transmission light source is incident from the lower surface of the polished surface of the silicon carbide substrate through a polarizer, and is emitted from the upper surface of the polished surface of the silicon carbide substrate and then passes through an analyzer, and the transmitted light signal passing through the analyzer is collected;

[0013] The second detection includes a second reflected light source detection;

[0014] The second reflected light source detection comprises the following steps:

[0015] The lower surface of the silicon carbide substrate polishing sheet is blocked to prevent light from penetrating, and then a reflected light source is used to illuminate the upper surface of the silicon carbide substrate polishing sheet to collect reflected light signals and scattered light signals.

[0016] Preferably, the positions of the protrusions and / or pits on the surface of the silicon carbide substrate polishing sheet are determined according to the missing positions of the reflected light signal obtained by the first light source reflection detection.

[0017] Preferably, data analysis is performed on the scattered light signal obtained by the first light source reflection detection to identify the definition and division of the pit edge, obtain the size and area of ​​the pit, and distinguish whether the surface of the silicon carbide substrate polishing sheet is a protrusion or a pit based on the size and area of ​​the pit.

[0018] Preferably, through data analysis, the roundness, grayscale area, number of neighborhood low / high gradient difference pixels, grayscale average and axis ratio of the pit are obtained, and then the size and area of ​​the pit are calculated.

[0019] Preferably, data analysis is performed on the transmitted light signal obtained by the transmitted light source detection to determine the pits caused by stress.

[0020] Preferably, data analysis is performed based on the difference between the reflected light signal obtained by the second reflected light source detection and the reflected light signal obtained by the first reflected light source detection to distinguish whether the pit is located on the upper surface or the lower surface of the silicon carbide substrate polishing sheet.

[0021] Preferably, the data analysis of the scattered light signal includes image processing and AI algorithm.

[0022] Preferably, the edges of the pits on the upper surface of the silicon carbide substrate polishing sheet have bright white areas, and the edges of the pits on the lower surface of the silicon carbide substrate polishing sheet have no bright white areas.

[0023] Preferably, the power of the light source is 10 to 100 W, the focusing depth is 3 to 10 μm within the upper and lower surfaces of the silicon carbide substrate polishing sheet, and the objective lens magnification is 2 to 8 times.

[0024] Preferably, through data analysis, the screening conditions for pits are set as follows: pit roundness > 0.2-0.5, grayscale area > 200-350, number of neighborhood low / high gradient difference pixels 40-100, grayscale average > 60-100, axis ratio > 2-8.

[0025] The invention provides a method for detecting pit defects of a silicon carbide substrate polishing sheet, comprising material loading, mechanical transmission, edge finding, OCR recognition, transmission to a detection table, a first detection, a second detection and data analysis; the first detection comprises a first reflected light source detection and a transmitted light source detection; the first light source reflected detection comprises the following steps: using a reflected light source to irradiate the upper surface of the silicon carbide substrate polishing sheet, and collecting reflected light signals and scattered light signals; the transmitted light source detection comprises the following steps: using a transmitted light source to emit transmitted light through a polarizer to be incident from the lower surface of the silicon carbide substrate polishing surface, and then emitting from the upper surface of the silicon carbide substrate polishing surface and passing through an analyzer, and collecting the transmitted light signal passing through the analyzer; the second detection comprises a second reflected light source detection; the second reflected light source detection comprises the following steps: shielding the lower surface of the silicon carbide substrate polishing sheet so that light cannot penetrate, and then using a reflected light source to irradiate the upper surface of the silicon carbide substrate polishing sheet, and collecting reflected light signals and scattered light signals.

[0026] Compared with the prior art, this application has the following advantages:

[0027] (1) Pits are divided into those generated during the manufacturing process and those generated by the packaging. They are visually inspected as weak / strong bright spots under high-intensity lighting. The surface is generally not smooth and has different response signals to the reflection of light. However, compared with other surface defects, the pits on the surface are not obvious and the signals are relatively weak. The present invention can distinguish other defect interferences caused by stress and non-stress by combining transmission polarization.

[0028] (2) In order to better distinguish the pits on the upper and lower surfaces, the present invention first performs bright field reflection, dark field, and transmission polarization detection and then performs two scans of the upper surface. During the second scan, the light source can only reflect but not transmit. Based on the signal difference of the two morphologies, data filtering and processing are performed to roughly distinguish whether the location is on the upper surface or the lower surface.

[0029] (3) Increase the image processing and AI algorithm optimization mode. There are obvious differences in the imaging of the front and back images. The grayscale contrast difference in the pit area on the front is large and the signal is strong. By collecting and analyzing the scattered light signal, the image is enhanced using computer vision technology to form a morphology that is easy to distinguish between the upper and lower surfaces. At the same time, increase the database of deep learning, continuously increase the sample capacity, continue training, form a wisdom library, and continuously improve the accuracy of defect classification. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0031] Figure 1 It is a schematic structural diagram of a device for detecting pit defects of a silicon carbide substrate polishing sheet according to the present invention;

[0032] Figure 2 It is a schematic diagram of reflected light detection and transmitted light detection in the pit defect detection of the silicon carbide substrate polishing sheet of the present invention;

[0033] Figure 1-2 In the figure, 1 is a loading device (Load port), 2 is a pre-alignment device, 3 is an OCR recognition device, 4 is a manipulator, 5 is a mechanical track, 6 is a suction cup, 7 is a transmitted light source, 8 is a reflected light source, 9 is a polarizer, 10 is an analyzer, and 11 is a camera;

[0034] Figure 3 This is a rendering of the front pit morphology detected by the camera in Example 1 of the present invention;

[0035] Figure 4 This is a rendering of the back surface pit morphology detected by the camera in Example 1 of the present invention. DETAILED DESCRIPTION

[0036] The present invention provides a method for detecting pit defects of a silicon carbide substrate polishing sheet, including material loading, mechanical transmission, edge finding, OCR recognition, transmission to a detection table, first detection, second detection and data analysis;

[0037] The first detection includes a first reflected light source detection and a transmitted light source detection;

[0038] The first light source reflection detection comprises the following steps:

[0039] Using a reflected light source to illuminate the upper surface of the silicon carbide substrate polishing sheet, and collecting reflected light signals and scattered light signals;

[0040] The transmission light source detection comprises the following steps:

[0041] The transmitted light emitted by the transmission light source is incident from the lower surface of the polished surface of the silicon carbide substrate through a polarizer, and is emitted from the upper surface of the polished surface of the silicon carbide substrate and then passes through an analyzer, and the transmitted light signal passing through the analyzer is collected;

[0042] The second detection includes a second reflected light source detection;

[0043] The second reflected light source detection comprises the following steps:

[0044] The lower surface of the silicon carbide substrate polishing sheet is blocked to prevent light from penetrating, and then a reflected light source is used to illuminate the upper surface of the silicon carbide substrate polishing sheet to collect reflected light signals and scattered light signals.

[0045] The present invention uses Figure 1 and Figure 2 The device shown detects surface pit defects on silicon carbide substrate polishing sheets. An open cassette loaded with silicon carbide substrate polishing sheets is loaded onto the loading device (Loadport) of the machine. The limit block indication sensor shows that the material position is within a reasonable range. A robotic arm with a mapping sensor scans the number and position of the wafers from bottom to top, and matches them with the information entered during loading. When the matching degree is confirmed to be 100%, the robotic arm transports the silicon carbide substrate polishing sheets in a single piece from bottom to top. The wafers are first loaded into the prealigner position. The OCR recognizes the ID code of the polishing sheet, and the robotic finger transfers the polishing sheet to the wafer carrier through a mechanical track.

[0046] In the present invention, a polarizer is arranged above the stage, a CCD camera is arranged above the polarizer, the stage is hollow inside, or the part in contact with the sample is transparent to allow transmitted light to pass through; a polarizer is arranged below the stage, a transmitted light source is arranged below the polarizer.

[0047] In order to better distinguish the pits on the upper surface and the lower surface, the present invention performs two tests. The first test includes bright field reflection, dark field scattering and transmission polarization detection, and the second test only includes bright field reflection and dark field scattering of the upper surface, but not transmission.

[0048] In the present invention, the bright field reflection and dark field scattering in the first detection are realized by irradiating the upper surface of the silicon carbide substrate polishing sheet with a reflective light source. When the incident light from the reflective light source irradiates the upper surface of the silicon carbide substrate polishing sheet without any obstruction above or below, and there are no pit defects, the light will be mirror-reflected at the same angle as the incident angle; if there are pits on the surface of the wafer, the reflected light will deviate from the normal mirror reflection angle, causing the corresponding light source receiver to fail to receive the signal at this point, so it can be determined that there are rough protrusions / pits on the surface of the silicon carbide substrate polishing sheet at this point.

[0049] Since there may be interference factors such as electrostatically adsorbed ions on the surface of the silicon carbide substrate polishing sheet, the present invention further determines whether the defect at the position where the reflected light signal is not received is a protrusion or a pit through dark field scattering. Under normal circumstances, the protruding particles on the surface of the polishing sheet are mostly electrostatically adsorbed particles, corrosion residues, etc., and these particles are mostly of high roundness. The present invention performs data analysis on the scattered light signal obtained by the first light source reflection detection, and combines the AI ​​algorithm to calculate the roundness or major axis, minor axis ratio, and area of ​​the above-mentioned defect to distinguish whether the above-mentioned detected defect is a protrusion or a pit. For example, defects with a side length greater than 12 and an area greater than 100 are set as pits.

[0050] In order to prevent interference from surface particles with low roundness, the present invention adds the distinction of transmission polarization in the first detection, that is, transmission light source detection. The transmission light source detection specifically includes:

[0051] Turn on the transmission light source, and allow the transmitted light to pass through the polarizer, the silicon carbide substrate polishing sheet and the analyzer in sequence, and enter the light signal receiver to obtain the transmitted light signal.

[0052] Because the presence of surface pits is generally caused by wrapped materials or processing damage, both of which are generally caused by the existence of internal stress in the silicon carbide crystal. When light is irradiated on the silicon carbide polished film through a polarizer, linear polarization is converted into elliptical polarization, and then converted into linear polarization with a certain angle change through a polarizer, it is determined that the pits are caused by stress.

[0053] In the present invention, the transmitted light source detection and the first reflected light source detection are performed separately to avoid mutual interference between the transmitted light and the scattered light. The present invention does not make special requirements on the order of the transmitted light source detection and the first reflected light source detection.

[0054] The above-mentioned bright field reflection is used to determine the approximate situation of the surface defects (such as the position and quantity), and then the scattered light signals of dark field scattering are used to screen out the protrusions with more obvious features from the above-mentioned defects. Further, the pit defects caused by stress are screened out in combination with the transmitted polarization. For the screened pit defects caused by stress, the present invention identifies the definition and division of the pit edge according to the strength of the scattered light signals of dark field scattering at the screened pit defects caused by stress, so as to accurately determine the size and area of ​​the pit. The calculation method of the size / area is defined by a combination of multiple rules, including but not limited to roundness, grayscale area, number of pixels with low / high gradient difference in the neighborhood, grayscale average, axis ratio, etc.

[0055] After the above-mentioned first detection, the pit point data caused by stress on the surface of the silicon carbide substrate polishing sheet can be obtained. The pit point data includes both the pit points on the upper surface of the silicon carbide substrate polishing sheet and the pit points in the lower surface of the silicon carbide substrate polishing sheet. In order to further distinguish the pit points on the upper and lower surfaces of the silicon carbide substrate polishing sheet, the present invention performs a second detection. The second detection only performs reflected light detection. The specific operation is: the lower surface of the silicon carbide substrate polishing sheet is blocked to prevent light from penetrating, and then a reflected light source is used to irradiate the upper surface of the silicon carbide substrate polishing sheet to collect reflected light signals and scattered light signals.

[0056] In the present invention, the second reflected light detection only detects the pits on the upper surface of the silicon carbide substrate polishing sheet. The result of the second reflected light detection is compared with the pit information of the first detection. According to the signal difference of the two morphologies, data filtering and processing are performed to distinguish whether the position of the pit is on the upper surface or the lower surface.

[0057] According to the detection method of the present invention, there is a bright white area at the edge of the pit defect on the front side, but there is no bright white area at the edge of the pit defect on the back side.

[0058] The invention provides a method for detecting pit defects of a silicon carbide substrate polishing sheet, comprising material loading, mechanical transmission, edge finding, OCR recognition, transmission to a detection table, a first detection, a second detection and data analysis; the first detection comprises a first reflected light source detection and a transmitted light source detection; the first light source reflected detection comprises the following steps: using a reflected light source to irradiate the upper surface of the silicon carbide substrate polishing sheet, and collecting reflected light signals and scattered light signals; the transmitted light source detection comprises the following steps: using a transmitted light source to emit transmitted light through a polarizer to be incident from the lower surface of the silicon carbide substrate polishing surface, and then emitting from the upper surface of the silicon carbide substrate polishing surface and passing through an analyzer, and collecting the transmitted light signal passing through the analyzer; the second detection comprises a second reflected light source detection; the second reflected light source detection comprises the following steps: shielding the lower surface of the silicon carbide substrate polishing sheet so that light cannot penetrate, and then using a reflected light source to irradiate the upper surface of the silicon carbide substrate polishing sheet, and collecting reflected light signals and scattered light signals.

[0059] Compared with the prior art, this application has the following advantages:

[0060] (1) Pits are divided into those generated during the manufacturing process and those generated by the packaging. They are visually inspected as weak / strong bright spots under high-intensity lighting. The surface is generally not smooth and has different response signals to the reflection of light. However, compared with other surface defects, the pits on the surface are not obvious and the signals are relatively weak. The present invention can distinguish other defect interferences caused by stress and non-stress by combining transmission polarization.

[0061] (2) In order to better distinguish the pits on the upper and lower surfaces, the present invention first performs bright field reflection, dark field, and transmission polarization detection and then performs two scans of the upper surface. During the second scan, the light source can only reflect but not transmit. Based on the signal difference of the two morphologies, data filtering and processing are performed to roughly distinguish whether the location is on the upper surface or the lower surface.

[0062] (3) Increase the image processing and AI algorithm optimization mode. There are obvious differences in the imaging of the front and back images. The grayscale contrast difference in the pit area on the front is large and the signal is strong. By collecting and analyzing the scattered light signal, the image is enhanced using computer vision technology to form a morphological effect map that is easy to distinguish between the upper and lower surfaces. At the same time, increase the database of deep learning, continuously increase the sample capacity, continue training, form a wisdom library, and continuously improve the accuracy of defect classification.

[0063] In order to further illustrate the present invention, a method for detecting pit defects of a silicon carbide substrate polishing sheet provided by the present invention is described in detail below in conjunction with an embodiment, but it should not be construed as limiting the protection scope of the present invention.

[0064] Example 1

[0065] 25 open cassettes loaded with silicon carbide substrate polishing wafers are manually placed in the loadport limit block of the inspection machine. The positioning sensor detects that the materials are in place and performs mapping operations. The robot receives the signal and transports the silicon carbide substrate polishing wafers from top to bottom to the Flat / Notch positioning identification. The robot then uses the robot to transfer the wafers to the OCR to identify the ID number. After completing this action, the wafers are transferred to the test carrier and irradiated with a 60W white light source. The automatic focusing depth is 5um within the upper and lower surfaces, the objective lens magnification is 5 times, and the resolution is 2um.

[0066] Scanning is done in a line scanning manner, and the objective lens moves in an S-shaped position to take pictures continuously. Two inspections are performed. The first is a reflected light upper surface scan, and the transmitted light is irradiated to the back of the wafer; the second is a reflected light upper surface scan alone, and the back suction cup blocks the light and prevents it from penetrating. After the two scans, the results of bright field reflection, transmitted polarization, and dark field scattering are analyzed and merged. The size, morphology, and position of the surface pits are analyzed by algorithm. The pit roundness is set to > 0.3, the grayscale area is > 320, the number of neighborhood low / high gradient difference pixels is 50, the grayscale average is > 70, and the axis ratio (aspect ratio) is > 6. Through conditional screening, the pits on the upper and lower surfaces are finally distinguished. There is a bright white area at the edge of the front pit defect (such as Figure 3 As shown), there is no bright white area at the edge of the pit defect on the back (as shown Figure 4 shown).

[0067] Example 2

[0068] The open cassette with 1-25 silicon carbide substrate polishing wafers is loaded onto the Loadport stage of the machine. The limit block indicator sensor shows that the material position is within a reasonable range. The robotic arm with a mapping sensor scans the number and position of the wafers from bottom to top, and matches them with the information entered during loading. When the matching degree is confirmed to be 100%, the robotic arm transfers the silicon carbide substrate polishing wafers in a single piece from bottom to top. The wafers are first loaded into the Prealigner position, and the OCR recognizes the ID code of the polishing wafer. The polishing wafer is transferred to the wafer stage by the robotic finger. The 30W white light source is automatically pre-calibrated, and then further precisely focused. The focusing depth is 7um on the upper and lower surfaces of the polishing wafer, the objective lens is magnified 2 times, and the resolution is 5um.

[0069] Using line scanning, the objective lens scans line by line in an S shape to take pictures. Two tests are performed, the first is a reflected light upper surface scan, and the transmitted light is irradiated to the back of the wafer; the second is a reflected light upper surface scan alone, and the back suction cup is blocked and the light cannot penetrate. According to the results of the two scans, data integration and analysis are performed, and the wisdom library, AI algorithm, etc. are used to form a rendering of the true morphology of the defects on the upper and lower surfaces. The roundness of the pits is set to > 0.4, the grayscale area is > 200, the number of neighborhood low / high gradient difference pixels is 80, the grayscale average is > 90, and the axis ratio (aspect ratio) is > 3. Through conditional screening, the pits on the upper and lower surfaces are finally distinguished. The test data results are further compressed and stored, and the detection is completed.

[0070] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for detecting pit defects of a polished silicon carbide substrate, comprising loading, mechanical transmission, edge finding, OCR recognition, transmission to a testing station, first detection, second detection and data analysis; The first detection includes a first reflected light source detection and a transmitted light source detection; The first light source reflection detection comprises the following steps: Using a reflected light source to illuminate the upper surface of the silicon carbide substrate polishing piece, and collecting reflected light signals and scattered light signals; The transmission light source detection comprises the following steps: The transmitted light emitted by the transmission light source is incident from the lower surface of the polished surface of the silicon carbide substrate through a polarizer, and is emitted from the upper surface of the polished surface of the silicon carbide substrate and then passes through an analyzer, and the transmitted light signal passing through the analyzer is collected; The second detection includes a second reflected light source detection; The second reflected light source detection comprises the following steps: The lower surface of the silicon carbide substrate polishing sheet is blocked to prevent light from penetrating, and then a reflected light source is used to illuminate the upper surface of the silicon carbide substrate polishing sheet to collect reflected light signals and scattered light signals.

2. The detection method according to claim 1, characterized in that: The positions of the protrusions and / or pits on the surface of the silicon carbide substrate polishing sheet are determined according to the missing positions of the reflected light signals obtained by the first light source reflection detection.

3. The detection method according to claim 2, characterized in that: Data analysis is performed on the scattered light signal obtained from the first light source reflection detection to identify the definition and division of the pit edge, obtain the size and area of ​​the pit, and distinguish whether the surface of the silicon carbide substrate polishing sheet is a protrusion or a pit based on the size and area of ​​the pit.

4. The detection method according to claim 3, characterized in that: Through data analysis, the roundness, grayscale area, number of neighborhood low / high gradient difference pixels, grayscale average and axis ratio of the pit are obtained, and then the size and area of ​​the pit are calculated.

5. The detection method according to claim 1, characterized in that: Data analysis is performed on the transmitted light signal obtained by the transmitted light source detection to determine the pits caused by stress.

6. The detection method according to claim 1, characterized in that: Data analysis is performed based on the difference between the reflected light signal obtained by the second reflected light source detection and the reflected light signal obtained by the first reflected light source detection to distinguish whether the pit is located on the upper surface or the lower surface of the silicon carbide substrate polishing sheet.

7. The detection method according to claim 3, characterized in that: The data analysis of the scattered light signal includes image processing and AI algorithm.

8. The detection method according to claim 7, characterized in that: The edges of the pits on the upper surface of the silicon carbide substrate polishing sheet have bright white areas, while the edges of the pits on the lower surface of the silicon carbide substrate polishing sheet do not have bright white areas.

9. The detection method according to claim 1, characterized in that: The power of the light source is 10 to 100 W, the focusing depth is 3 to 10 μm within the upper and lower surfaces of the silicon carbide substrate polishing sheet, and the objective lens magnification is 2 to 8 times.

10. The detection method according to claim 1, characterized in that: Through data analysis, the screening conditions for pits are set as follows: pit roundness > 0.2-0.5, grayscale area > 200-350, number of neighborhood low / high gradient difference pixels 40-100, grayscale average > 60-100, and axis ratio > 2-8.

Images