A multi-modal detection method for small-scale stacking fault defects in silicon carbide substrates

Abstract

The present application belongs to the technical field of silicon carbide defect detection, and particularly relates to a multi-modal detection method for small-scale stacking fault defects in a silicon carbide substrate, comprising the following steps: performing photoluminescence imaging on a silicon carbide substrate to be detected to obtain a candidate region of small-scale stacking fault defects; performing two-dimensional mapping scanning of micro-photoluminescence spectrum on the candidate region to obtain a spectral intensity mapping image in the candidate region; constructing a multi-modal matching model, and performing spatial registration on the spectral intensity mapping image and the photoluminescence imaging to confirm a single small-scale stacking fault defect by calculating a weighted pixel overlap rate; and performing two-dimensional mapping scanning of micro-photoluminescence spectrum on the single small-scale stacking fault defect to determine the defect morphology and size. The method has the advantages of high sensitivity and high resolution, is suitable for rapid detection and online quality control of defects in a large-size silicon carbide substrate, and has a good industrial application prospect.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of silicon carbide defect detection, and particularly relates to a multi-modal detection method for small-scale stacking fault defects in a silicon carbide substrate. BACKGROUND

[0002] Wide-bandgap semiconductor silicon carbide (SiC) is widely used in high-voltage, high-frequency and high-temperature power devices due to its excellent physical properties such as high breakdown field strength, high thermal conductivity and high chemical stability. 4H silicon carbide (4H-SiC) as the mainstream epitaxial substrate material, its crystal quality directly affects the performance of the device. However, in the preparation process of 4H-SiC substrate, stacking faults (SFs), micropipes, dislocations and other crystal defects are inevitably introduced, which affect the subsequent epitaxial growth quality, device electrical performance and long-term stability. Among them, small-scale stacking fault defects are difficult to accurately identify by current detection technology due to their small size, but such defects may develop into more serious problems in subsequent processes, affecting the reliability of the device. Therefore, it is crucial to develop a high-sensitivity and reliable defect detection technology.

[0003] Currently, methods for detecting crystal defects such as optical microscopy, X-ray diffraction (XRD), scanning electron microscopy (SEM), transmission electron microscopy (TEM) and chemical etching method, although have certain effect in identifying large-size defects, but have significant limitations: optical microscopy and XRD have insufficient spatial resolution, making it difficult to capture microscopic defect details; SEM and TEM detection require complex sample pretreatment, not only the operation process is cumbersome, the equipment cost is high, but also the detection efficiency is low; chemical etching method is a destructive detection method, which will damage the substrate and cannot meet the subsequent epitaxial growth or device preparation requirements. These limitations make it difficult to meet the needs of small-scale stacking fault defects for efficient, non-destructive and accurate detection.

[0004] In recent years, photoluminescence (PL) technology has been widely used in the study of wide-bandgap semiconductor defects due to its non-contact, non-destructive and high sensitivity. Among them, PL imaging can quickly identify the non-uniformity of substrate surface luminescence and assist in analyzing defect distribution. However, the spatial resolution of traditional PL imaging is low, and it is difficult to clearly distinguish the microscopic stacking faults with small size and weak luminescence contrast. SUMMARY

[0005] The application aims to provide a multi-modal detection method for small stacking fault defects in a silicon carbide substrate, so as to overcome the shortcomings of the prior art, realize accurate registration and fusion of PL imaging and micro-PL spectral mapping by introducing a matching model based on affine transformation, greatly improve the consistency of multi-modal data in spatial position, and quickly and reliably identify small stacking fault defects in a 4H-SiC single crystal substrate under non-destructive conditions, which has high spatial resolution and is simple to operate and suitable for industrialized silicon carbide quality control.

[0006] In order to achieve the above-mentioned purpose, the technical scheme of the application is as follows:

[0007] The application provides a detection method for silicon carbide stacking fault defects, comprising the following steps:

[0008] (1) performing photoluminescence imaging on a to-be-detected silicon carbide substrate to obtain a candidate region of small stacking fault defects;

[0009] (2) performing micro-photoluminescence spectral two-dimensional mapping scanning on the candidate region to obtain a spectral intensity mapping image in the candidate region;

[0010] (3) constructing a multi-modal matching model to perform spatial registration on the spectral intensity mapping image and the photoluminescence imaging, and confirming a single small stacking fault defect by calculating a pixel overlap rate;

[0011] (4) performing micro-photoluminescence spectral two-dimensional mapping scanning on the single small stacking fault defect to determine a defect morphology and size.

[0012] In the application, under room temperature conditions, a large-area photoluminescence (PL) imaging is performed on a to-be-detected silicon carbide substrate by using an ultraviolet excitation light source, a linear region with local photoluminescence intensity attenuation is identified as a candidate region of small stacking fault defects, micro-PL spectral two-dimensional mapping scanning is performed on the candidate region to obtain a two-dimensional PL intensity distribution map, a multi-modal matching model based on two-dimensional affine transformation is introduced to perform spatial registration on the PL imaging image and the micro-PL spectral two-dimensional mapping image, the registration accuracy is evaluated by calculating a pixel overlap rate, and accurate positioning of a single small defect is completed; finally, according to the two-dimensional PL intensity distribution map and characteristic peak position information under a smaller scanning step, the morphology characteristics and size information of a single small stacking fault defect are determined, and multi-modal non-destructive detection of small stacking fault defects in a silicon carbide substrate is realized.

[0013] In some other embodiments, in step (1), the silicon carbide substrate is a 4H-SiC substrate, and the small stacking fault defect is a continuous linear defect with an aspect ratio > 5:1 and a length in the range of 0.2-1 mm;

[0014] To further improve the accuracy of identifying small stacking fault defects, transmission electron microscopy can be used to verify the microstructure of candidate regions for small stacking fault defects initially identified by photoluminescence imaging. By observing the atomic-scale morphology of the defects, the accuracy of the photoluminescence detection results can be verified.

[0015] In some other embodiments, in step (1), the light source used for photoluminescence imaging is an ultraviolet laser source, and a red high-pass filter is configured in the signal receiving path. Specifically, the wavelength of the ultraviolet laser source is 313 nm or 320 nm; the cutoff wavelength of the red high-pass filter is 660 nm or 700 nm (only red light signals with wavelengths ≥660 nm or ≥700 nm are allowed to pass through).

[0016] In some other embodiments, in step (2), the light source for the two-dimensional mapping scan of the microscopic photoluminescence spectrum is an ultraviolet excitation source with a spatial resolution >1 μm, a emission spectrum coverage range of 360-600 nm, a scan step size of ≤5 μm, and an integration time ≥0.01 s. Specifically, the characteristic emission wavelength is 426 nm.

[0017] In some other implementations, in step (3), the multimodal image matching model is a two-dimensional affine transformation multimodal matching model, and the least squares method is used to estimate the model parameters.

[0018] Specifically, the endpoints of large-sized rod-shaped SFs are preferentially selected as feature points. If the sample has no large-sized defects, laser markings on the wafer edge can be manually used as feature points. This is because the endpoints of large-sized rod-shaped SFs are clearly visible in both PL imaging and micro-PL imaging systems. The pixel coordinates of each pair of feature points in the two images are recorded, where the pixel coordinates of the PL imaging image are (…). The pixel coordinates of the microscopic PL spectral image are ( ), =1, 2, 3; Substituting the pixel coordinates of the feature points into a two-dimensional affine transformation multimodal matching model, pixel-level spatial alignment is achieved through feature point calibration. The vector p is then solved using the least squares method to complete spatial registration.

[0019] The mathematical expression for the multimodal matching model of two-dimensional affine transformation is:

[0020] ,

[0021] in,( ) represents the pixel coordinates of the PL-imported image. () represents the pixel coordinates of the microscopic PL spectral mapping image. =1, 2, 3; , , ， is a linear transformation coefficient, is a translation coefficient.

[0022] The mathematical expression for solving the vector p by the least square method is:

[0023]

[0024] wherein, is a parameter vector to be solved, A and are respectively composed of the feature point coordinates of the source image and the target image.

[0025] The obtained optimal parameters are used to perform geometric transformation on the PL imaging image, so that the PL imaging image and the microscopic PL spectral mapping image are one-to-one corresponding in spatial position.

[0026] In some other embodiments, in step (3), the pixel overlap rate is calculated as follows: based on the binarization processing, the photoluminescence imaging image and the microscopic photoluminescence spectral mapping image are aligned, the corresponding binarized pixel matrix is obtained and the intersection pixel number is calculated; based on the total pixel number of the candidate region and the intersection pixel number, the pixel overlap rate is calculated.

[0027] Specifically, the binarization processing: an adaptive threshold method (the PL imaging threshold value is 70% of the average intensity of the defect-free area, and the microscopic PL threshold value is 3 times the noise value of the 426 nm peak intensity) is adopted, the defect pixels are marked as 1, and the remaining pixels are marked as 0, so as to align the two images.

[0028] After the alignment of the two images, the two images are compared pixel by pixel, and the intersection pixel number is obtained N overlap :

[0029]

[0030] wherein, is the serial number of a single pixel point in the candidate region (corresponding to each pixel of “one by one selection”, ensuring that all pixels are covered without repetition or omission); is the binarization result of the th pixel in the PL imaging image (1 for defect pixel, 0 for non-defect pixel), is the binarization result of the th pixel in the microscopic PL spectral image, I imaging and I spectraThe binary pixel matrix of PL imaging image and micro-PL spectral mapping image are multiplied and summed, and the essence is to count the total number of pixels that are determined as defects in both images.

[0031] Let the total number of pixels in the candidate region be N candidate The pixel overlap rate is According to the following formula:

[0032]

[0033] Wherein, The total number of intersection pixels that are determined as defects in both images, N candidate The total number of pixels in the candidate region.

[0034] If R ≥ 85%, it can be preliminarily confirmed that the candidate region contains small SF defects.

[0035] In some other embodiments, in step (3), the accurate identification of a single small stacking fault defect can also be achieved by calculating the weighted pixel overlap rate, which specifically includes: assigning a weight to the pixel point according to its position in the defect region, and obtaining the weighted pixel overlap rate by weighted calculation.

[0036] Specifically, considering that different factors contribute unevenly to the overlap rate, such as image resolution difference, optical distortion, noise, background change, and defect morphology difference, a further calculation method of weighted pixel overlap rate is proposed:

[0037]

[0038] Wherein, The serial number of the single pixel point in the candidate region (corresponding to each pixel of "one by one selection", ensuring that all pixels are covered without omission). = 1 indicates that the pixel is determined as a defect in both images, otherwise 0 (i.e. is the matching result of the th pixel, 1 for matching and 0 for not matching); The weight value of the th pixel (the weight of the defect center pixel is set to 1.0, the weight of the edge pixel is set to 0.5, and the weight of the non-defect region pixel is 0).

[0039] If the calculated ≥ 85%, it can be further confirmed that the region is a single small stacking fault defect.

[0040] In step (4), the shape of the single small stacking fault defect is an irregular polygon;

[0041] The scanning step length of the micro photoluminescence spectrum two-dimensional mapping scanning is ≤2 μm, and the luminescence spectrum coverage range is 360-600 nm.

[0042] The beneficial effects of the present application are:

[0043] (1) The detection method based on PL imaging and micro-PL spectrum fusion analysis provided by the present application can identify small stacking fault defects in a 4H-SiC substrate at room temperature without damage and with high resolution, in combination with a two-dimensional affine transformation matching model. The method locates the position in space by identifying the local intensity attenuation area, and accurately extracts the optical characteristics (center wavelength, number, morphology, and size) of the defect by using two-dimensional spectrum scanning. Compared with traditional technologies, the present application has the significant advantages of non-contact, high sensitivity, and high spatial resolution, and is particularly good at rapid screening of sub-millimeter level defects. The method can be directly applied to industrial large-size silicon carbide wafers, adapt to the online quality control process, serve the defect grading, yield prediction, and substrate process optimization, and significantly improve the detection efficiency and reliability.

[0044] (2) The present application uses an ultraviolet excitation light source to perform large-area photoluminescence (PL) imaging on the silicon carbide substrate to be measured at room temperature, identifies the linear area with local intensity attenuation as the candidate area of the small stacking fault defect, performs two-dimensional mapping scanning of the micro-PL spectrum on the candidate area to obtain a two-dimensional PL intensity distribution map, realizes accurate positioning of a single small defect in combination with the PL imaging result, extracts the luminescence intensity and luminescence peak position, confirms the defect characteristics with a center wavelength of about 426 nm, determines the morphology characteristics and size information of a single small stacking fault defect according to the two-dimensional PL intensity distribution map and peak characteristics with a smaller scanning step length, and realizes accurate identification. The method has the advantages of high sensitivity and high resolution, is suitable for rapid detection and online quality control of large-size silicon carbide substrate defects, and has good industrial application prospects. BRIEF DESCRIPTION OF DRAWINGS

[0045] The drawings accompanying the specification of the present application form a part thereof and serve to provide further understanding of the present application, the illustrative embodiments of the present application and its description serve to explain the present application, and do not constitute an improper limitation of the present application.

[0046] Figure 1 It is a schematic diagram of the composition of the PL imaging system used in embodiment 1 of the present application;

[0047] Figure 2 It is a distribution map of small stacking fault defects detected by the PL imaging system in embodiment 1 of the present application;

[0048] Figure 3PL imaging result map of any candidate region in Example 1 of the present application;

[0049] Figure 4 Typical PL spectrum map of small SF region and SF-free region in Example 1 of the present application;

[0050] Figure 5 High-resolution microscopic PL spectrum two-dimensional mapping of any three typical small SF defects in Example 1 of the present application, wherein the three typical small SF defects are marked as A, B and C respectively. DETAILED DESCRIPTION

[0051] Those skilled in the art will understand that the following examples are only for illustration of the present application and should not be regarded as limiting the scope of the present application. If specific conditions are not specified in the examples, they are carried out under conventional conditions or according to the manufacturer's recommendations.

[0052] Explanation of terms:

[0053] Photoluminescence: abbreviated as PL, is a process in which a material absorbs the energy of a photon after being excited by an external light source (usually ultraviolet light), causes electron transition, and then releases energy in the form of emitted photons.

[0054] Stacking fault: abbreviated as SF, is a common two-dimensional lattice defect in crystal structure, which refers to the abnormal arrangement of atoms in the normal stacking sequence along a certain crystal plane (such as close-packed plane), and this abnormality is limited to a certain thin layer (usually several atomic layers thick), without changing the overall structure type of the crystal.

[0055] Traditional PL imaging is limited by the optical diffraction limit, making it difficult to distinguish small stacking faults, resulting in missed detection of small defects. Although microscopic PL spectrum can break through the diffraction limit, its point-by-point scanning mode makes the detection of large-size wafers in the whole area time-consuming, which cannot meet the industrial online efficiency requirements. At the same time, there are inherent contradictions between the two types of technology: (1) the differences in field of view and coordinate system between PL imaging (fast, low resolution) and microscopic PL spectrum (slow, high resolution) make spatial registration difficult due to sample tilt and optical distortion, affecting the accuracy of defect positioning; (2) the lack of physical correlation model between light emission intensity decay (PL imaging feature) and specific spectral peak / linewidth (microscopic PL feature) hinders the joint analysis of multi-modal defects; (3) the weak light emission decay of small stacking faults is easily overwhelmed by the background noise (impurity light emission, surface scattering) of the substrate, resulting in a significant reduction in signal-to-noise ratio.

[0056] The present application provides a multi-modal detection method for small stacking fault defects in silicon carbide substrates, which can be applied to multi-modal non-destructive detection of small stacking fault defects in silicon carbide substrates, and specifically includes the following steps:

[0057] (1) photoluminescence imaging is performed on the to-be-tested silicon carbide substrate to obtain a candidate region of small-scale stacking fault defects;

[0058] (2) microscopic photoluminescence spectrum two-dimensional mapping scanning is performed on the candidate region to obtain a spectrum intensity mapping image in the candidate region;

[0059] (3) a multi-modal matching model is constructed to perform spatial registration on the spectrum intensity mapping image and the photoluminescence imaging, and a single small-scale stacking fault defect is confirmed by calculating a pixel overlap rate;

[0060] (4) microscopic photoluminescence spectrum two-dimensional mapping scanning is performed on the single small-scale stacking fault defect to determine the defect morphology and size.

[0061] The scheme of the present application is described below in combination with specific embodiments.

[0062] Embodiment 1

[0063] The present embodiment provides a multi-modal detection method for small-scale stacking fault defects in a silicon carbide substrate, which specifically comprises the following steps:

[0064] S1, photoluminescence imaging is performed on the to-be-tested silicon carbide substrate to obtain a candidate region of small-scale stacking fault defects;

[0065] S2, microscopic photoluminescence spectrum two-dimensional mapping scanning is performed on the candidate region to obtain a spectrum intensity mapping image in the candidate region;

[0066] S3, a multi-modal matching model is constructed to perform spatial registration on the spectrum intensity mapping image and the photoluminescence imaging, and a single small-scale stacking fault defect is confirmed by calculating a pixel overlap rate;

[0067] S4, microscopic photoluminescence spectrum two-dimensional mapping scanning is performed on the single small-scale stacking fault defect to determine the defect morphology and size.

[0068] In step S1, under room temperature conditions, a photoluminescence (PL) imaging system is used to perform large-area imaging on the to-be-tested silicon carbide substrate in the full-frame range, and the imaging image is analyzed by relying on the image processing function of the photoluminescence (PL) imaging system to obtain a candidate region of small-scale stacking fault defects.

[0069] The composition of the adopted PL imaging system is shown in Figure 1 from top to bottom, including an imaging camera, a red light high-pass filter and a substrate placement platform, and an ultraviolet laser light source irradiating the substrate placement platform. The wavelength of the ultraviolet laser light source is 313 nm or 320 nm; the cutoff wavelength of the red light high-pass filter is 660 nm or 700 nm (only allowing red light signals with a wavelength of ≥660 nm or ≥700 nm to pass).

[0070] The working principle of the PL imaging system is as follows: the ultraviolet laser light source irradiates the silicon carbide substrate to be measured, the silicon carbide material is excited to generate a PL signal, the generated PL signal is filtered through a red high-pass filter provided in a signal receiving path, the short-wave interference signal in the background is effectively removed, and thus the contrast and signal-to-noise ratio of the image are improved. The PL signal filtered and processed is finally received and recorded by an imaging camera to form a PL imaging image representing defects.

[0071] Specifically, the obtaining process of the candidate region of the small stacking fault defect is as follows:

[0072] (1) Under room temperature conditions, the PL imaging system is used to perform large-area imaging on the whole range of the silicon carbide substrate to be measured; the PL imaging system analyzes and processes the large-area imaging according to the image processing function provided by the system, that is, the light emission intensity of each region of the substrate is intuitively reflected by the difference between light and dark, that is, the region with strong light emission intensity appears bright, and the region with local light emission intensity decay appears dim.

[0073] The system takes the light emission intensity of the defect-free region on the surface of the substrate as a reference, automatically marks the region obviously deviating from the reference (too bright or too dark) as a light emission intensity abnormal region as a preliminary screened defect. By analyzing and processing the large-area imaging, the dim region is obtained.

[0074] (2) For the dim region screened out, a linear region with an aspect ratio > 5:1, a length in the range of 0.2-1 mm and continuous without obvious breakpoints is screened out; at the same time, point-shaped dark spots formed by surface contamination, irregular block-shaped dark areas caused by scratches and other non-linear interference are automatically excluded, and continuous dark lines highly matched with the local light emission decay characteristics caused by small stacking fault lattice defects are retained. Through multiple imaging verification of the same substrate, if the imaging results of the above linear region are consistent, it can be confirmed that it is a candidate region of the small stacking fault defect, which is marked and recorded by the system. The distribution map of the small stacking fault defect detected by the PL imaging system is shown in Figure 2 .

[0075] (3) The small stacking fault defect candidate region detected by the PL imaging system is further verified by a spherical aberration correction scanning transmission electron microscope, and the defect is attributed to a stacking fault.

[0076] The micro-scale linear stacking fault obtained in Figure 2 is defined as a small stacking fault (small SF). In the defect distribution map obtained by PL imaging, a black dot is used to mark the position of each detected small SF to facilitate subsequent analysis and screening. In addition, the local region containing the small SF is preliminarily defined as a candidate region.

[0077] In the present application,Figure 2 The distribution of only small stacking fault defects is given, which does not mean that other defects cannot be identified and classified. Figure 3 Any defect candidate region screened by the PL imaging system. In the present application, the silicon carbide substrate to be tested is a 4H-SiC substrate material, and there is no special limitation on its specific source. Commercially available products known to those skilled in the art can be selected.

[0078] In step S2, the candidate region determined in S1 is subjected to microscopic PL spectral two-dimensional mapping scanning to obtain a spectral intensity mapping image in the region. The specific process is as follows: the silicon carbide substrate to be tested is fixed on a precision displacement platform of a microscopic PL spectrometer, and is observed and precisely positioned to the candidate region by an optical microscope. The scanning range (covering the candidate region and its surrounding normal range) is set to ensure complete inclusion of the target defect and the surrounding reference region.

[0079] More specifically, after starting the scanning of the microscopic PL spectrometer, the precision displacement platform drives the silicon carbide substrate to be tested to move point by point in the X-Y plane at a set step size. Upon reaching each test point, the excitation light is focused on the point and PL signals are excited, and the spectral detector synchronously collects the complete PL spectrum of the point. After scanning is completed, by comparing the PL spectra of the defect-free region and the candidate region, the characteristic emission wavelength of the defect region is determined to be 426 nm. The spectral peak intensity at 426 nm and nearby bands is subjected to integral processing, and the integral time is greater than or equal to 0.01 s. The integral result is associated with the spatial coordinates of each test point to generate a spectral intensity mapping image that can directly reflect the distribution of the defect emission intensity, as shown in FIG. 4. It can be seen that the characteristic emission wavelength of the small SF is 426 nm, which provides basic data support for subsequent small SF defect identification.

[0080] The adopted microscopic PL spectrometer is equipped with a 320 nm ultraviolet excitation light source, the spatial scanning resolution is better than 1 μm, and the emission spectrum detection range is 360-600 nm (which can cover the main defect-related emission peaks in SiC); the scanning step size is set to be less than or equal to 5 μm.

[0081] In step S3, due to the differences in resolution and field of view range between PL imaging and microscopic PL imaging technology (PL imaging is large-area low-resolution, and microscopic PL is small-area high-resolution), direct comparison will cause spatial offset. In order to realize accurate positioning of the defect, the spectral intensity mapping image obtained in S2 needs to be spatially registered with the PL imaging result obtained in S1. Through a multi-modal matching model of two-dimensional affine transformation, geometric transformation is performed to realize spatial registration of the spectral intensity mapping image and the photoluminescence imaging; and the single small stacking fault defect is further confirmed by calculating the pixel overlap rate or the weighted pixel overlap rate. The specific steps include the following steps;

[0082] (1) Preferentially select the end point of large-size rod-shaped SF as the feature point, and if the sample has no large-size defects, manually mark the laser mark at the edge of the wafer as the feature point. This is because the end point of large-size rod-shaped SF is clearly visible in both PL imaging and microscopic PL imaging systems.

[0083] (2) Record the pixel coordinates of each pair of feature points in the two images, where the pixel coordinates of the PL imaging image are (x1, y1), and the pixel coordinates of the microscopic PL spectral mapping image are (x2, y2), i = 1, 2, 3. =1, 2, 3; Substitute the feature point pixel coordinates into the multi-modal matching model of two-dimensional affine transformation to achieve pixel-level spatial alignment through feature point calibration. And solve the vector p by least squares method to complete spatial registration.

[0084] The mathematical expression of the multi-modal matching model of two-dimensional affine transformation is:

[0085] ,

[0086] Where (x1, y1) is the pixel coordinate of the PL imaging image, (x2, y2) is the pixel coordinate of the microscopic PL spectral mapping image, =1, 2, 3; , , ， is the linear transformation coefficient, is the translation coefficient.

[0087] The mathematical expression of solving the vector p by least squares method is:

[0088]

[0089] Where, is the parameter vector to be solved, A and are composed of the feature point coordinates of the source image and the target image, respectively.

[0090] Use the obtained vector p to perform geometric transformation on the PL imaging image, so that it is one-to-one corresponding with the microscopic PL spectral mapping image in spatial position.

[0091] (3) After spatial registration, introduce pixel overlap rate to quantitatively evaluate the registration accuracy of the two images, the specific process is as follows:

[0092] ​​​​Binaryzation: The adaptive threshold method was used (PL imaging threshold value was 70% of the average intensity of the defect-free area, and the micro-PL threshold value was 3 times the noise value of the 426 nm peak intensity). The defect pixels were marked as 1, and the rest of the pixels were marked as 0, realizing the alignment of the two images.

[0093] After the alignment of the two images, the number of intersection pixels was obtained by comparing the two images pixel by pixel N overlap :

[0094]

[0095] wherein, is the serial number of a single pixel point in the candidate area (corresponding to each pixel of "one by one selection", ensuring that all pixels are covered without repetition or omission); is the binaryzation result of the th pixel in the PL imaging image (1 for defect pixel, 0 for non-defect pixel), is the binaryzation result of the th pixel in the micro-PL spectral image, I imaging and I spectra are the binaryzation pixel matrices of the PL imaging image and the micro-PL spectral mapping image respectively, and the sum of the multiplication of the two is the essence of counting the total number of pixels that are judged as defects in both images.

[0096] The total number of pixels in the candidate area is denoted as N candidate , then the pixel overlap rate is calculated according to the following formula:

[0097]

[0098] wherein, is the total number of intersection pixels that are judged as defects in both images, N candidate is the total number of pixels in the candidate area.

[0099] If R ≥ 85%, the candidate area can be preliminarily confirmed to contain small SF defects.

[0100] Considering that different factors contribute unevenly to the overlap rate, such as image resolution difference, optical distortion, noise, background change, and defect morphology difference, a further calculation method of weighted pixel overlap rate is proposed:

[0101]

[0102] wherein, Similarly, the index of a single pixel within the candidate region (corresponding to each pixel selected one by one, ensuring that all pixels are covered without repetition or omission). =1 indicates that the pixel is judged as a defect in both images; otherwise, it is 0 (i.e., ...). M i It is the first The matching result for each pixel, where 1 indicates a match and 0 indicates a non-match. For the first The weight values ​​of each pixel are set (the weight of the defect center pixel is set to 1.0, the weight of the edge pixel is set to 0.5, and the weight of the non-defect area pixel is 0).

[0103] If the calculation yields If the failure rate is ≥85%, it can be further confirmed that the region is a single small stacking fault defect with a center wavelength of approximately 426 nm, and its spatial position on the substrate can be recorded.

[0104] Using the above method, the PL imaging features and microspectral features of each small SF defect can be obtained simultaneously, enabling precise spatial positioning and confirming that the defect center wavelength is approximately 426 nm.

[0105] After registering the candidate regions of the 4H-SiC substrate to be tested in this embodiment, R was calculated to be 89%. =91%, meeting the accuracy requirements. At this point: the dark line area in the PL imaging completely coincides with the 426 nm strong peak area in the microscopic PL; the deviation mainly comes from the edge distortion of the PL imaging (the deviation at 10 mm from the center is ≤2 pixels), which can be effectively corrected through weighted calculation.

[0106] Using the above method, the PL imaging features and microspectral features of each small SF defect can be obtained simultaneously, achieving precise spatial positioning and confirming that the defect center wavelength is approximately 426 nm.

[0107] In step S4, for the single small stacking fault defect region located in S3, a micro PL spectrometer is used to perform high-precision point-by-point scanning with a smaller scanning step size. By integrating the spectral peak intensity of 426 nm and its vicinity, a micro PL spectral intensity mapping image of the micro-region is obtained. The small stacking fault presents bright contrast in the image. Based on this, its true morphology can be observed and its size parameters such as length and width can be quantified through the image scale.

[0108] Specifically: For individual small SF defects in any three regions, a two-dimensional microscopic photoluminescence (PL) mapping scan was further performed. The microscopic PL spectrometer used was equipped with a 320 nm ultraviolet excitation source, with a spatial scanning resolution better than 1 μm and a emission spectrum detection range of 360-600 nm (covering the emission peaks related to major defects in SiC). The scanning step size was set to ≤2 μm. The PL spectral intensity within the 426±10 nm band was integrated to obtain the cumulative light intensity signal in this band, i.e., the two-dimensional microscopic PL mapping scan result. Figure 5 As shown in the figure, “CCD cts” refers to the charge-coupled device photon count, which represents the photon signal count detected and quantified by the CCD sensor at the corresponding point between the “specific spatial position” and the “specific wavelength within the 426±10nm band” during the scanning process. The value directly reflects the relative intensity of PL emission at that position and wavelength. Figure 5 The numbers 609.3 and -98.5 in graph A are... Figure 5 The numbers 407.3, -102.8, and in graph B are shown. Figure 5 The numbers 371.5 and -102 in Figure C represent the relative luminescence intensity values ​​(unit: CCD cts) of the PL in the 426±10nm band at the corresponding micro-region location. The positive values ​​609.3, 407.3, and 371.5 represent the luminescence signal intensity of the core region of the small stacking fault defect in Figures A, B, and C, respectively. The higher the value, the more significant the luminescence of the defect at that location. The negative values ​​-98.5, -102.8, and -102 represent the luminescence signal intensity of the non-defect region surrounding the defect in the corresponding sub-figure. The difference between the two types of values ​​clearly distinguishes the luminescence characteristics of the defect region and the non-defect region.

[0109] like Figure 5 Figures A, B, and C show that individual small SF defects exhibit polygonal morphology features of varying sizes and irregular boundaries in the PL two-dimensional mapping scan results. This morphological information can serve as an important basis for substrate quality assessment and is of significant importance in improving the efficiency and accuracy of defect identification.

[0110] In summary, the detection method provided in this embodiment, based on the joint analysis of PL imaging and microscopic PL spectroscopy, and by introducing a two-dimensional affine transformation matching model, can achieve non-destructive, high-resolution detection of small stacking fault defects in 4H-SiC substrates at room temperature.

[0111] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A multi-modal detection method for small stacking fault defects in silicon carbide substrates, characterized in that, Includes the following steps: (1) Photoluminescence imaging is performed on the silicon carbide substrate to be tested to obtain candidate regions for small stacking fault defects; (2) Perform two-dimensional mapping scanning of the candidate region under microscopic photoluminescence spectrum to obtain a spectral intensity mapping image; (3) Construct a multimodal matching model to spatially register the spectral intensity mapping image and photoluminescence imaging, and confirm the single small stacking fault defect by calculating the pixel overlap rate; The multimodal image matching model is a two-dimensional affine transformation model. It also includes geometric transformation of photoluminescence imaging based on the least squares method, and spatial position matching of spectral intensity mapping image and photoluminescence imaging. Specifically, the endpoints of large-sized rod-shaped SF are selected as feature points. If the sample has no large-sized defects, the laser markings on the edge of the wafer are manually marked as feature points. The pixel coordinates of each pair of feature points in the two images are recorded. The pixel coordinates of the feature points are substituted into the two-dimensional affine transformation multimodal matching model. Pixel-level spatial alignment is achieved through feature point calibration. The vector p is solved by the least squares method to complete spatial registration. This includes identifying a single small stacking fault defect by calculating a weighted pixel overlap rate; the weighted pixel overlap rate is calculated as follows: different weights are assigned to the pixels according to their spatial position in the defect area, the matched pixels are weighted and summed, and the weighted pixel overlap rate is obtained by combining the total weight. (4) Perform two-dimensional mapping scanning of microscopic photoluminescence spectrum on a single small stacking fault defect to determine the morphology and size of the defect.

2. The multi-mode detection method for small stacking fault defects in silicon carbide substrates according to claim 1, characterized in that, In step (1), the silicon carbide substrate is a 4H-SiC substrate, and the small stacking fault defect is a continuous, unbroken linear defect with an aspect ratio > 5:1 and a length in the range of 0.2-1 mm.

3. The multi-mode detection method for small stacking fault defects in silicon carbide substrates according to claim 1, characterized in that, In step (1), the light source used for photoluminescence imaging is an ultraviolet laser source, and a red high-pass filter is configured in the signal receiving path.

4. The multi-mode detection method for small stacking fault defects in silicon carbide substrates according to claim 1, characterized in that, In step (2), the light source for the two-dimensional mapping scan of the microscopic photoluminescence spectrum is an ultraviolet excitation source with a spatial resolution >1μm.

5. The multi-mode detection method for small stacking fault defects in silicon carbide substrates according to claim 1, characterized in that, In step (2), the emission spectrum coverage range of the two-dimensional mapping scan of microscopic photoluminescence spectroscopy is 360-600nm, the scan step size is ≤5 μm, and the integration time is ≥0.01 s.

6. The multi-mode detection method for small stacking fault defects in a silicon carbide substrate according to claim 1, characterized in that, In step (3), the pixel overlap rate is calculated as follows: the photoluminescence imaging map and the microscopic photoluminescence spectral mapping map are binarized and aligned to obtain the number of intersecting pixels, and the pixel overlap rate is calculated by combining the total number of pixels in the candidate region.

7. The multi-mode detection method for small stacking fault defects in a silicon carbide substrate according to claim 1, characterized in that, In step (4), the scanning step size of the two-dimensional mapping scan of the microscopic photoluminescence spectrum is ≤2 μm, and the emission spectrum coverage range is 360-600 nm.

8. The multi-mode detection method for small stacking fault defects in silicon carbide substrates according to claim 1, characterized in that, In step (4), the shape of the single small stacking fault defect is an irregular polygon.

Images