Silicon carbide crystal cutting method, apparatus, and storage medium

Abstract

The application relates to the technical field of semiconductor material processing, in particular to a silicon carbide crystal cutting method and device and a storage medium, the method comprising the following steps: acquiring a tomographic image of a crystal; constructing a three-dimensional reconstruction model based on the tomographic image; identifying different types of defects in the tomographic image by using a trained neural network to obtain a defect category; creating a digital twin management model matched with the crystal based on the defect category and the three-dimensional reconstruction model; performing simulation pre-cutting on the digital twin management model to obtain a simulation pre-cutting result; and determining the cutting parameters of the crystal based on the simulation pre-cutting result. The method can more efficiently cut the crystal and reduce the waste of crystal materials by constructing the cutting parameters of the crystal through relatively accurate simulation pre-cutting results.

Description

Technical Field

[0001] The present invention relates to the field of semiconductor material processing technology, and in particular to a silicon carbide crystal cutting method, device and storage medium in the field of semiconductor material processing technology. Background Art

[0002] Silicon carbide (SiC) represents the third generation of wide bandgap (WBG) semiconductor materials, following the first generation of elemental semiconductors and the second generation of compound semiconductors. SiC possesses incomparable properties to traditional Si materials, such as a wide bandgap, a high critical breakdown field, high thermal conductivity, and a high carrier saturation drift velocity. These characteristics make SiC highly promising for applications in high-temperature, high-frequency, high-power, optoelectronics, and radiation-resistant fields.

[0003] In the related art, the mainstream cutting methods for silicon carbide crystals include wire cutting and dual laser processing. These two methods usually involve cutting silicon carbide crystals according to a certain thickness (for example, 350 microns (μm)), and then performing yield inspection on the cut wafers. Generally speaking, the yield is between 30-50%, which means that 50-70% of the cut wafers are unusable. This is mainly due to the presence of micron-level defects such as polytypes, microtubes, and carbon inclusions in silicon carbide wafers. These defects are fatal to subsequent devices and cannot be repaired by subsequent epitaxial processes. Therefore, the cutting method in the related art is blind and inefficient, resulting in problems such as waste of cutting time and consumables. Summary of the Invention

[0004] The object of the present invention is to provide a method, device and storage medium for cutting silicon carbide crystals. The technical solutions adopted are as follows:

[0005] In a first aspect, an embodiment of the present invention provides a method for cutting a silicon carbide crystal, the method comprising:

[0006] Obtaining cross-sectional images of the interior of the crystal;

[0007] constructing a three-dimensional reconstruction model based on the tomographic image;

[0008] Using a trained neural network to identify different types of defects in the tomographic image to obtain defect categories; wherein the trained neural network is trained based on the tomographic image and the initial defect categories of the tomographic image;

[0009] Creating a digital twin management model for the crystal matching based on the defect category and the three-dimensional reconstruction model;

[0010] Performing simulation pre-cutting on the digital twin management model to obtain simulation pre-cutting results;

[0011] Based on the simulated pre-cutting result, cutting parameters of the crystal are determined.

[0012] In a second aspect, an embodiment of the present invention provides a silicon carbide crystal cutting device, the device comprising:

[0013] A tomographic image acquisition module, used for acquiring tomographic images inside the crystal;

[0014] A crystal three-dimensional reconstruction module, used for constructing a three-dimensional reconstruction model based on the tomographic image;

[0015] a defect category recognition module, configured to use a trained neural network to identify different types of defects in the tomographic image and obtain defect categories; wherein the trained neural network is trained based on the tomographic image and the initial defect category of the tomographic image;

[0016] a digital twin management model creation module, configured to create a digital twin management model for the crystal matching based on the defect category and the three-dimensional reconstruction model;

[0017] A simulation pre-cutting module is used to perform simulation pre-cutting on the digital twin management model to obtain simulation pre-cutting results;

[0018] A cutting parameter determination module is used to determine the cutting parameters of the crystal based on the simulated pre-cutting result.

[0019] In a third aspect, a computer program product is provided, comprising: a computer program code, which, when executed on a computer, enables the computer to execute the method in the first aspect or any one of the possible implementations of the first aspect.

[0020] In a fourth aspect, a computer-readable storage medium is provided, which stores a computer program code. When the computer program code runs on a computer, the computer executes the method in the first aspect or any possible implementation method described in the first aspect.

[0021] The present invention has the following beneficial effects: a three-dimensional reconstruction model of the crystal is constructed through a tomographic image inside the crystal; and different types of defects in the tomographic image are identified through a trained neural network to obtain defect categories, and the trained neural network is trained based on the tomographic image and the initial defect category of the tomographic image; in this way, the trained neural network can more accurately identify the defect categories included in the crystal. Afterwards, a digital twin management model for the crystal matching is created based on the defect category and the three-dimensional reconstruction model; in this way, the defect category and the three-dimensional reconstruction model are combined to create a digital twin management model containing defect spatial distribution information. Finally, a simulated pre-cutting result is obtained by performing simulated pre-cutting on the digital twin management model; and based on the simulated pre-cutting result, the cutting parameters of the crystal are determined. In this way, the cutting parameters of the crystal are constructed through relatively accurate simulated pre-cutting results, so that the crystal can be cut more efficiently and the waste of crystal consumables can be reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0023] Figure 1 This is a schematic diagram of a process for implementing a silicon carbide crystal cutting method provided by an embodiment of the present invention;

[0024] Figure 2 This is another schematic diagram of a process for implementing a silicon carbide crystal cutting method provided by an embodiment of the present invention;

[0025] Figure 3 is a schematic diagram of the category distribution of identifying crystal defects provided by some embodiments of the related art;

[0026] Figure 4 1 is a schematic diagram of an implementation framework of a silicon carbide crystal cutting method provided by an embodiment of the present invention;

[0027] Figure 5 This is another schematic diagram of a process for implementing a silicon carbide crystal cutting method provided by an embodiment of the present invention;

[0028] Figure 6 This is another schematic diagram of a process for implementing a silicon carbide crystal cutting method provided by an embodiment of the present invention;

[0029] Figure 7 This is a schematic diagram of an application scenario of a silicon carbide crystal cutting method provided by an embodiment of the present invention;

[0030] Figure 8 This is a schematic diagram of the structure of a silicon carbide crystal cutting device provided by an embodiment of the present invention;

[0031] Figure 9 It is a structural diagram of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0032] To further illustrate the technical means and effectiveness of the present invention in achieving its intended objectives, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail a method for cutting a silicon carbide crystal according to the present invention, including its specific implementation, structure, features, and effectiveness. In the following description, references to "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0033] In the description of the embodiments of the present invention, unless otherwise specified, " / " means or, for example, A / B can mean A or B: "and / or" in the text is only a description of the association relationship of associated objects, indicating that there can be three relationships, for example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, in the description of the embodiments of the present invention, "multiple" refers to two or more than two.

[0034] In the following, the terms "first" and "second" are used for descriptive purposes only and should not be understood to imply or suggest relative importance or implicitly indicate the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features.

[0035] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0036] The following describes in detail a specific scheme of a silicon carbide crystal cutting method provided by the present invention in conjunction with the accompanying drawings. The three-dimensional structure of the silicon carbide crystal is obtained by a non-destructive characterization method and plays a planning and guiding role in the subsequent crystal cutting process. Figure 1 , which shows a schematic diagram of an implementation process of a silicon carbide crystal cutting method provided by one embodiment of the present invention, the method comprising:

[0037] 101, obtaining a tomographic image inside the crystal.

[0038] Here, the crystal may be a silicon carbide crystal. In some possible implementations, a scanned image is obtained by scanning and imaging the interior of the crystal; for example, the silicon carbide crystal is placed on a sample stage for micro-computed tomography (Micro-CT) scanning imaging, and a scanned image of the silicon carbide crystal is obtained by rotating the sample stage 360°. Subsequently, the scanned image is subjected to noise reduction processing to obtain the tomographic image. For example, the scanned image is filtered to make the obtained tomographic image clearer, thereby improving the image quality of the tomographic image.

[0039] 102. Construct a three-dimensional reconstruction model based on the tomographic image.

[0040] Here, after scanning and obtaining a cross-sectional image of the crystal interior, the grayscale values ​​and contrast of different cross-sectional images are analyzed, and then the grayscale values ​​and contrast are adjusted to construct a three-dimensional reconstruction model of the crystal.

[0041] In some possible implementations, the above step 102 can be performed by Figure 2 The steps shown achieve:

[0042] 201 : Determine the grayscale value and contrast of the tomographic image.

[0043] Here, after the tomographic image is obtained, the grayscale value and contrast of the tomographic image are read by a computer.

[0044] 202 : Based on the ideal crystal region, adjust the grayscale value and contrast of the tomographic image to obtain an adjusted image.

[0045] Here, the ideal crystal region is a defect-free crystal region under a predetermined ideal state. By analyzing the grayscale value and contrast of the ideal crystal region, the grayscale value and contrast of the tomographic image are adjusted accordingly to obtain an adjusted image corresponding to the tomographic image, thereby improving the image difference between the defective region and the ideal crystal region in the tomographic image.

[0046] 203 : Perform three-dimensional crystal reconstruction based on the adjusted image to obtain the three-dimensional reconstructed model.

[0047] Here, a 3D crystal reconstruction method is used. Based on the adjusted image, the crystal is reconstructed in 3D, resulting in a 3D reconstructed model that includes image information from different sections of the crystal. By adjusting the grayscale and contrast of the section image based on the ideal crystal region, the image difference between the defective region and the ideal crystal region can be eliminated, allowing the reconstructed 3D model to more clearly reflect the defective region of the crystal.

[0048] 103, a trained neural network is used to identify different types of defects in the tomographic image and obtain defect categories.

[0049] The trained neural network is obtained by training based on the tomographic image and the initial defect category of the tomographic image.

[0050] In some possible implementations, a trained neural network can be obtained through the following process:

[0051] First, initial defect recognition is performed on the tomographic image to obtain an initial defect category.

[0052] Here, based on the tomographic images, we conducted experiments on the defects inside the crystal using morphological characterization techniques (Raman spectrum, steady-state photoluminescence spectroscopy, time-resolved photoluminescence spectroscopy, and confocal microscopy). We found that there are three types of initial defects at the micron level inside the crystal, such as Figure 3 As shown, there are three types of initial defects at the micron level inside the crystal, including: Figure 3 (a) The polytype shown (including: Figure 3 4H-SiC, 6H-SiC and 15R-SiC in (a), Figure 3 (b) shows the microtubules and Figure 3 (c) The carbon wrapping shown; Figure 3 In the figure, the horizontal axis represents the Raman shift (cm -1 ), the ordinate represents the relative intensity (au).

[0053] Secondly, the trained neural network is obtained based on the initial defect category and the tomographic image.

[0054] Here, the trained neural network is a residual network (Resnet-50) architecture; Figure 4 As shown, from Figure 4 (a) It can be seen that the tomographic image database of a 6-inch silicon carbide ingot characterized by micro-CT technology is combined with traditional optical characterization technology and input into a trained neural network (i.e., neural network Resnet-50) to identify the defect category of the input image through artificial intelligence. The network architecture of the trained neural network is shown in the figure below. Figure 4As shown in (b), the database is first loaded and preprocessed. For example, the input image is preprocessed, such as digitizing and normalizing, to obtain image data with a data size of 224*224*3. Next, feature extraction is performed on the input image data to obtain feature maps. The feature maps are then classified using a fully connected layer to obtain the classification results (output), which may include: single crystal, carbon inclusion, micropipe, polytype, or none. In this way, by using the tomographic image of the defect as the input to the trained neural network, it is possible to automatically and accurately judge and identify different types of defects, and achieve higher defect recognition accuracy.

[0055] 104 . Create a digital twin management model for the crystal matching based on the defect category and the three-dimensional reconstruction model.

[0056] Here, a digital twin management model containing information on the spatial distribution of defects is established using a framework based on 3D reconstructed crystal information and artificial intelligence (convolutional neural networks). This model includes micron-level defects within the crystal, such as polymorphs, microtubules, and carbon inclusions. By combining the defect categories output by the trained neural network with the 3D reconstructed model, a digital twin management model containing information on the spatial distribution of defects is created. The scale of this digital twin management model matches the crystal size.

[0057] In some possible implementations, the above step 104 may be implemented by the following steps 141 to 143 (not shown):

[0058] 141. Determine the flat edge of the crystal positioning edge and the crystal growth surface based on the three-dimensional reconstruction model.

[0059] Here, after the three-dimensional reconstruction model is constructed, the flat side of the positioning side and the crystal growth surface of the actual crystal are analyzed through the three-dimensional reconstruction model.

[0060] 142. Use the flat edge of the crystal positioning edge as the reference positioning edge and the crystal growth surface as the reference surface to construct an initial digital twin management model.

[0061] Here, the digital twin management model of the crystal is constructed using the flat edge of the actual crystal positioning edge as the reference positioning edge, and the crystal growth surface as the reference surface. The size of the digital twin management model is the maximum cutting diameter. In this way, by using the flat edge of the actual crystal positioning edge as the reference positioning edge and the crystal growth surface as the reference surface, a digital twin management model of the crystal is constructed, resulting in a digital twin management model with a 1:1 ratio to the actual crystal size.

[0062] 143. Based on the defect category, transfer learning is performed on the initial digital twin management model to obtain a digital twin management model that can characterize the spatial distribution of defects in the crystal.

[0063] Here, after constructing a digital twin management model that matches the actual crystal size 1:1, the defect categories are embedded into the digital twin management model through transfer learning. This allows the resulting digital twin management model to include the spatial distribution of defects in the crystal, allowing the digital twin management model to more accurately represent the actual crystal. In this way, by adding the more accurate defect categories output by the trained neural network to the three-dimensional reconstruction model through transfer learning, a digital twin management model that includes the spatial distribution of defects in the crystal can be more accurately constructed, allowing pre-cutting to be performed using the digital twin management model, thereby establishing efficient and accurate cutting parameters.

[0064] 105 , performing simulation pre-cutting on the digital twin management model to obtain simulation pre-cutting results.

[0065] Here, the digital twin management model is simulated and pre-cut layer by layer to obtain the simulation pre-cutting results.

[0066] In some possible implementations, the above step 105 can be performed by Figure 5 The steps shown achieve:

[0067] 501. Based on a preset defect density threshold, determine an available area in the digital twin management model.

[0068] Here, the preset defect density threshold can be a custom setting, for example, the preset defect density threshold is 30 per square centimeter (cm -2 Since the size of the digital twin management model is 1:1 with the size of the actual crystal, it is possible to select an area in the digital twin management model where the defect density is less than a preset defect density threshold, i.e., obtain a usable area.

[0069] 502 , performing simulated pre-cutting on the available area to obtain the simulated pre-cutting result.

[0070] Here, simulated pre-cutting is performed layer by layer within the usable area according to a certain layer thickness, generating simulated pre-cutting results. This simulated pre-cutting result includes the pre-cut wafers within each layer of the usable area. By performing simulated pre-cutting of the usable area within the digital twin management model according to a preset defect density threshold, a relatively accurate simulated pre-cutting result can be obtained.

[0071] In some possible implementations, the digital twin management model is pre-cut layer by layer according to a certain layer thickness to obtain a simulated pre-cutting result. That is, the above step 502 can be implemented by the following steps 521 and 522 (not shown):

[0072] 521 , determining each cutting layer in the available area based on a preset layer thickness.

[0073] Here, the preset layer thickness can be custom set, for example, the preset layer thickness is 350 micrometers (μm). According to the layer thickness of 350 μm, each cutting layer in the available area is determined so that the available area can be cut layer by layer to obtain wafers of different sizes.

[0074] 522 , performing simulated pre-cutting on the available area layer by layer according to the cutting layers and the preset cutting size to obtain the simulated pre-cutting result.

[0075] Here, the digital twin management network model is simulated and pre-cut layer by layer (layer thickness is 350μm), and the available area (defect density <30cm -2 ) According to the conventional wafer size, wafers are cut into 2-inch, 4-inch, and 6-inch wafers to obtain simulated pre-cut results. In this way, according to the preset layer thickness and preset cutting size, the available area is simulated pre-cut layer by layer, thus obtaining simulated pre-cut results that are more in line with the actual effect.

[0076] 106. Determine cutting parameters of the crystal based on the simulated pre-cutting result.

[0077] Here, the wafer cutting value score is calculated by simulating the pre-cutting results. When the wafer cutting value score reaches the maximum value, the cutting parameters of the crystal are constructed to achieve a better cutting effect.

[0078] In some possible implementations, the above step 106 can be performed by Figure 6 The steps shown achieve:

[0079] 601 : Based on the defect density of the simulated pre-cutting result, classify the simulated pre-cutting result into different wafer grades.

[0080] Here, each wafer is divided into different wafer grades according to the defect density of each wafer in the simulation pre-cutting results. Figure 7 As shown, according to the micropipe distribution histogram 701 of the wafer in the simulation pre-cutting result, the defect density is divided into P level (11cm -2 )、R-level (15cm -2 ) and D-level (30cm -2 );from Figure 7 It can be seen that the micropipe distribution histogram 701 includes two grades of wafers, namely, 2-inch R-grade wafers 71 and 2-inch D-grade wafers 72 .

[0081] 602 : Determine a wafer cutting value score corresponding to each cutting layer based on the number of each wafer grade in the different wafer grades.

[0082] Here, the wafer dicing value score S of the simulated pre-dicing results is calculated: S = (a1×b1+a2×b2+a3×b3+a4×b4+a5×b5+a6×b6+a7×b7+a8×b8+a9×b9) / (a7+a8+a9)×b7; where a1 is the number of 2-inch P-grade wafers, a2 is the number of 2-inch R-grade wafers, a3 is the number of 2-inch D-grade wafers, a4 is the number of 4-inch P-grade wafers, a5 is the number of 4-inch R-grade wafers, a6 is the number of 4-inch D-grade wafers, a7 is the number of 6-inch P-grade wafers, a8 is the number of 6-inch R-grade wafers, and a9 is the number of 6-inch D-grade wafers. The letters a and b represent different dicing layers.

[0083] 603 : Construct cutting parameters of the crystal based on the wafer cutting value score.

[0084] Here, by analyzing the number of wafer grades in different wafer grades, the wafer cutting value score corresponding to each cutting layer is calculated, so that more efficient and accurate cutting parameters can be constructed through the wafer cutting value score.

[0085] In some possible implementations, the above step 603 may be implemented by the following steps 631 to 633 (not shown):

[0086] 631. When the wafer cutting value score reaches a maximum value, determine an optimal solution for wafer cutting planning.

[0087] 632. Determine a cutting radius of the crystal based on the optimal solution.

[0088] 633. Determine the cutting radius and the simulated pre-cutting position information in the digital twin management model as the cutting parameters.

[0089] In this way, when the wafer cutting value score reaches the maximum value, by obtaining the optimal solution of the wafer cutting plan and constructing the cutting parameters of the crystal, the optimal cutting of the crystal can be achieved, thereby significantly improving the effect of crystal cutting.

[0090] In some possible implementations, after the cutting parameters are established, the crystal can be cut layer by layer through the following process:

[0091] First, a candidate wafer image corresponding to the optimal solution is determined.

[0092] Here, after calculating the optimal solution for wafer cutting planning, the optimal solution can represent the size of the cut wafer and the chip grade of the wafer. The candidate wafer image corresponding to the optimal solution is an image with the size and chip grade.

[0093] Secondly, based on the candidate wafer image, the image of the wafer to be cut in the crystal is adjusted so that the image of the wafer to be cut coincides with the candidate wafer image.

[0094] Here, the wafer state in the candidate wafer image, that is, the optimal solution, can be used to characterize the size of the cut wafer and the chip grade of the wafer, and the image of the wafer to be cut in the crystal is adjusted so that the two images overlap.

[0095] Again, based on the simulated pre-cutting position information in the digital twin management model, the target cutting position is determined in the image of the wafer to be cut.

[0096] Here, since the size of the digital twin management model is 1:1 with the size of the actual crystal, after the image of the wafer to be cut coincides with the image of the candidate wafer, the target cutting position can be determined in the image of the wafer to be cut according to the cutting position in the candidate wafer image.

[0097] Finally, the crystal is cut layer by layer based on the cutting radius and the target cutting position.

[0098] Here, according to the cutting radius, the target cutting position in the crystal is cut layer by layer, thereby achieving efficient cutting of the crystal. In the digital twin management model, the flat edge of the actual silicon carbide crystal is set as the positioning reference edge of the digital twin management model, the (0001) plane is set as the reference plane, and the real crystal is cut according to the spatial position XYZ value (i.e., the target cutting position) and the cutting radius of the simulated pre-cutting knife in the digital twin management model. In this way, the image of the wafer to be cut is adjusted according to the candidate wafer image corresponding to the optimal solution so that the two graphics overlap, and the size of the digital twin management model is 1:1 with the size of the actual crystal. Therefore, after the two graphics overlap, the target cutting position can be accurately determined in the digital twin management model according to the position of the candidate wafer image; thus, according to the cutting radius, the actual crystal is cut layer by layer at the target cutting position, which can not only cut a wafer with a higher chip grade, but also improve the efficiency of crystal cutting.

[0099] In an embodiment of the present invention, a three-dimensional reconstruction model of the crystal is constructed through a tomographic image inside the crystal; and different types of defects in the tomographic image are identified through a trained neural network to obtain defect categories. The trained neural network is trained based on the tomographic image and the initial defect category of the tomographic image; in this way, the trained neural network can more accurately identify the defect categories included in the crystal. Afterwards, a digital twin management model for the crystal matching is created based on the defect category and the three-dimensional reconstruction model; in this way, the defect category and the three-dimensional reconstruction model are combined to create a digital twin management model containing defect spatial distribution information. Finally, a simulated pre-cutting result is obtained by performing simulated pre-cutting on the digital twin management model; and based on the simulated pre-cutting result, the cutting parameters of the crystal are determined. In this way, the cutting parameters of the crystal are constructed through more accurate simulated pre-cutting results, so that the crystal can be cut more efficiently and the waste of crystal consumables can be reduced.

[0100] The embodiment of the present invention provides a silicon carbide crystal cutting device, please refer to Figure 8 , which shows a schematic structural diagram of a silicon carbide crystal cutting device provided by one embodiment of the present invention. The device 800 includes:

[0101] A tomographic image acquisition module 801 is used to acquire a tomographic image inside the crystal;

[0102] A crystal three-dimensional reconstruction module 802 is used to construct a three-dimensional reconstruction model based on the tomographic image;

[0103] a defect category recognition module 803 configured to use a trained neural network to identify different types of defects in the tomographic image and obtain defect categories; wherein the trained neural network is trained based on the tomographic image and the initial defect categories of the tomographic image;

[0104] A digital twin management model creation module 804 is configured to create a digital twin management model for the crystal matching based on the defect category and the three-dimensional reconstruction model;

[0105] A simulation pre-cutting module 805 is used to perform simulation pre-cutting on the digital twin management model to obtain a simulation pre-cutting result;

[0106] The cutting parameter determination module 806 is configured to determine the cutting parameters of the crystal based on the simulated pre-cutting result.

[0107] In some possible implementations, the tomographic image acquisition module 801 is further configured to scan and image the interior of the crystal to obtain a scanned image; and perform noise reduction processing on the scanned image to obtain the tomographic image.

[0108] In some possible implementations, the crystal three-dimensional reconstruction module 802 is also used to determine the grayscale value and contrast of the tomographic image; based on the ideal crystal area, the grayscale value and contrast of the tomographic image are adjusted to obtain an adjusted image; based on the adjusted image, the crystal three-dimensional reconstruction is performed to obtain the three-dimensional reconstruction model.

[0109] In some possible implementations, the defect category recognition module 803 is further configured to perform initial defect recognition on the tomographic image to obtain an initial defect category; and obtain the trained neural network based on the initial defect category and the tomographic image.

[0110] In some possible implementations, the digital twin management model creation module 804 is also used to determine the flat edge of the crystal positioning edge and the crystal growth surface based on the three-dimensional reconstruction model; use the flat edge of the crystal positioning edge as the reference positioning edge and the crystal growth surface as the reference surface to construct an initial digital twin management model; based on the defect category, transfer learning is performed on the initial digital twin management model to obtain a digital twin management model that can characterize the spatial distribution of defects in the crystal.

[0111] In some possible implementations, the simulation pre-cutting module 805 is further used to determine the available area in the digital twin management model based on a preset defect density threshold; perform simulation pre-cutting on the available area to obtain the simulation pre-cutting result.

[0112] In some possible implementations, the simulated pre-cutting module 805 is further used to determine each cutting layer in the available area based on a preset layer thickness; and perform simulated pre-cutting on the available area layer by layer according to each cutting layer and the preset cutting size to obtain the simulated pre-cutting result.

[0113] In some possible implementations, the cutting parameter determination module 806 is also used to divide the simulated pre-cutting results into different chip grades based on the defect density of the simulated pre-cutting results; determine the wafer cutting value score corresponding to each cutting layer based on the number of each chip grade in the different chip grades; and construct the cutting parameters of the crystal based on the wafer cutting value score.

[0114] In some possible implementations, the cutting parameter determination module 806 is also used to determine the optimal solution for wafer cutting planning when the wafer cutting value score reaches a maximum value; determine the cutting radius of the crystal based on the optimal solution; and determine the cutting radius and the simulated pre-cutting position information in the digital twin management model as the cutting parameters.

[0115] In some possible implementations, the cutting parameter determination module 806 is also used to determine a candidate wafer image corresponding to the optimal solution; based on the candidate wafer image, the image of the wafer to be cut in the crystal is adjusted so that the image of the wafer to be cut coincides with the candidate wafer image; based on the simulated pre-cutting position information in the digital twin management model, the target cutting position is determined in the image of the wafer to be cut; based on the cutting radius and the target cutting position, the crystal is cut layer by layer.

[0116] Optionally, the transmission medium can be a wired link (for example, but not limited to, coaxial cable, optical fiber and digital subscriber line (DSL)) or a wireless link (for example, but not limited to, wireless Fidelity (WIFI), Bluetooth and mobile device network). It should be noted that the device provided in the above embodiment is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device can be divided into different functional modules to complete all or part of the functions described above. In addition, the method embodiments provided in the above embodiments belong to the same concept. The specific implementation process is detailed in the method embodiments, which will not be repeated here.

[0117] Figure 9 FIG. 1 is a schematic diagram of the structure of a computer device provided by an embodiment of the present invention. For example, Figure 9As shown, the computer device 900 includes: a memory 901, a processor 902, and a computer program 903 stored in the memory 901 and running on the processor 902, wherein when the processor 902 executes the computer program 903, the computer device can execute any one of the silicon carbide crystal cutting methods described above.

[0118] In addition, an embodiment of the present invention also protects a system, which may include a memory and a processor, wherein an executable program code is stored in the memory, and the processor is used to call and execute the executable program code to perform a silicon carbide crystal cutting method provided by an embodiment of the present invention. This embodiment can divide the system into functional modules according to the above method example. For example, it can correspond to each functional module, or two or more functions can be integrated into one processing module, and the above integrated module can be implemented in the form of hardware. It should be noted that the division of modules in this embodiment is schematic and is only a logical function division. There may be other division methods in actual implementation. It should be noted that all relevant contents of each step involved in the above method embodiment can be referred to the functional description of the corresponding functional module, and will not be repeated here.

[0119] It should be understood that the device provided in this embodiment is used to perform the above-mentioned silicon carbide crystal cutting method, and therefore can achieve the same effect as the above-mentioned implementation method. In the case of an integrated unit, the device may include a processing module and a storage module. Among them, when the device is applied to a device, the processing module can be used to control and manage the actions of the device. The storage module can be used to support the device to execute mutual program codes, etc. Among them, the processing module can be a processor or a controller, which can implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of the present invention. The processor can also be a combination that implements a computing function, such as a combination of one or more microprocessors, a combination of digital signal processing (DSP) and a microprocessor, etc., and the storage module can be a memory.

[0120] In addition, the device provided in an embodiment of the present invention may be specifically a chip, component, or module. The chip may include a connected processor and memory; wherein the memory is used to store instructions. When the processor calls and executes the instructions, the chip can execute a silicon carbide crystal cutting method provided in the above embodiment. This embodiment also provides a computer-readable storage medium, which stores computer program code. When the computer program code is executed on a computer, it causes the computer to execute the above-mentioned related method steps to implement a silicon carbide crystal cutting method provided in the above embodiment.

[0121] This embodiment also provides a computer program product, which, when executed on a computer, causes the computer to execute the above-mentioned steps to implement a method for cutting a silicon carbide crystal provided in the above embodiment. The device, computer-readable storage medium, computer program product, or chip provided in this embodiment are all used to execute the corresponding method provided above. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding method provided above, and will not be repeated here. Through the description of the above embodiments, those skilled in the art can understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example. In actual application, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In the embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of modules or units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not performed. On the other hand, the mutual coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interface, device or unit, which may be electrical, mechanical or other forms.

[0122] It should be noted that the above-mentioned order of the embodiments of the present invention is for description only and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or may be advantageous. The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other. Each embodiment focuses on the differences from other embodiments. The above content is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or replacements within the technical scope disclosed by the present invention, which should be covered within the scope of protection of the present invention.

Claims

1. A method for cutting silicon carbide crystals, characterized in that: The silicon carbide crystal cutting method comprises: Obtaining cross-sectional images of the interior of the crystal; determining the grayscale value and contrast of the tomographic image; Adjusting the grayscale value and contrast of the tomographic image based on the ideal crystal region to obtain an adjusted image; Performing three-dimensional crystal reconstruction based on the adjusted image to obtain a three-dimensional reconstruction model; Using a trained neural network to identify different types of defects in the tomographic image to obtain defect categories; wherein the trained neural network is trained based on the tomographic image and the initial defect categories of the tomographic image; Determining the flat edge of the crystal positioning edge and the crystal growth surface based on the three-dimensional reconstruction model; The flat edge of the crystal positioning edge is used as the reference positioning edge, and the crystal growth surface is used as the reference surface to construct an initial digital twin management model; Based on the defect category, transfer learning is performed on the initial digital twin management model to obtain a digital twin management model capable of characterizing the spatial distribution of defects in the crystal; Based on a preset defect density threshold, determining an available area in the digital twin management model; Performing simulated pre-cutting on the available area to obtain a simulated pre-cutting result; Based on the simulated pre-cutting result, cutting parameters of the crystal are determined.

2. A method for cutting silicon carbide crystals according to claim 1, characterized in that: The obtaining of a tomographic image inside the crystal comprises: Scanning and imaging the interior of the crystal to obtain a scanned image; The scanned image is subjected to noise reduction processing to obtain the tomographic image.

3. A method for cutting silicon carbide crystals according to claim 1, characterized in that: The method further comprises: performing initial defect recognition on the tomographic image to obtain an initial defect category; The trained neural network is obtained based on the initial defect category and the tomographic image.

4. A method for cutting silicon carbide crystals according to claim 1, characterized in that: The performing simulated pre-cutting on the available area to obtain the simulated pre-cutting result includes: determining each cutting layer in the available area based on a preset layer thickness; According to the cutting layers and the preset cutting size, the available area is simulated pre-cut layer by layer to obtain the simulated pre-cutting result.

5. The method for cutting a silicon carbide crystal according to claim 1, wherein: The step of determining the cutting parameters of the crystal based on the simulated pre-cutting result includes: Based on the defect density of the simulated pre-cutting result, the simulated pre-cutting result is divided into different wafer grades; determining a wafer cutting value score corresponding to each cutting layer based on the number of each wafer grade in the different wafer grades; Based on the wafer cutting value score, cutting parameters of the crystal are constructed.

6. A silicon carbide crystal cutting device, characterized in that: Applied to the silicon carbide crystal cutting method according to claim 1, the apparatus comprises: A tomographic image acquisition module, used for acquiring tomographic images inside the crystal; a crystal three-dimensional reconstruction module, configured to determine the grayscale value and contrast of the tomographic image; adjust the grayscale value and contrast of the tomographic image based on an ideal crystal region to obtain an adjusted image; and perform three-dimensional crystal reconstruction based on the adjusted image to obtain a three-dimensional reconstructed model; a defect category recognition module, configured to use a trained neural network to identify different types of defects in the tomographic image and obtain defect categories; wherein the trained neural network is trained based on the tomographic image and the initial defect category of the tomographic image; a digital twin management model creation module, configured to determine, based on the three-dimensional reconstruction model, the flat edge of the crystal positioning edge and the crystal growth surface; construct an initial digital twin management model using the flat edge of the crystal positioning edge as the reference positioning edge and the crystal growth surface as the reference surface; and perform transfer learning on the initial digital twin management model based on the defect category to obtain a digital twin management model capable of characterizing the spatial distribution of defects in the crystal; A simulation pre-cutting module is used to determine a usable area in the digital twin management model based on a preset defect density threshold; perform simulation pre-cutting on the usable area to obtain a simulation pre-cutting result; A cutting parameter determination module is used to determine the cutting parameters of the crystal based on the simulated pre-cutting result.

7. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer program code, which, when executed on a computer, causes the computer to execute the silicon carbide crystal cutting method according to any one of claims 1 to 5 .

Images