Method for regulating the electrical properties of gallium oxide micro-defects

Abstract

The present invention relates to the technical field of gallium oxide material optimization, and more particularly to a method and system for adjusting the electrical properties of microscopic defects in gallium oxide. The method comprises obtaining an overall structural image of the gallium oxide material using a brightfield imaging mode, determining defect-sensitive areas based on the overall structural image, obtaining darkfield images and high-resolution images of the inspection area, and inputting the acquired darkfield and high-resolution images into a pre-built defect type recognition model to determine the defect type in the inspection area. The present invention eliminates the need to use a microscope to traverse the entire gallium oxide material, greatly improving its defect detection efficiency. Furthermore, by determining defect-sensitive areas, then acquiring darkfield and high-resolution images of the defect-sensitive areas and inputting them into a trained, pre-built defect type recognition model for defect identification, the method eliminates the need for manual defect type determination, further improving the efficiency and stability of the defect type determination.

Description

Technical Field

[0001] The present invention relates to the technical field of gallium oxide material optimization, and in particular to a method for regulating the electrical properties of gallium oxide micro-defects. Background Art

[0002] Gallium oxide (GAO) is an important candidate material for next-generation high-power, high-frequency electronic devices due to its wide bandgap, excellent thermal stability, and high breakdown electric field. However, microscopic defects in GAO materials can significantly affect their electrical properties, thereby affecting the performance and stability of devices.

[0003] Chinese patent publication number CN1 17347304A discloses a method for detecting atomic-level defects in gallium oxide crystals. The method uses a scattering scanning near-field optical microscope system to detect atomic-level defects in gallium oxide crystals. The scattering scanning near-field optical microscope system collects optical signals of gallium oxide crystal defects through the scattering enhancement effect of the probe tip for imaging. The scattering scanning near-field optical microscope system includes an atomic force microscope system and an optical system, wherein the light source of the optical system is a tunable quantum cascade laser and / or a carbon dioxide laser, most preferably a tunable quantum cascade laser. The atomic force microscope system includes a probe and a laser.

[0004] Existing methods for detecting microscopic defects in gallium oxide materials are generally performed by using a microscope to capture images for observation. That is, the entire gallium oxide material needs to be observed through a microscope for microscopic defect inspection. Because the system needs to gradually and slowly traverse the entire gallium oxide material area, the inspection efficiency is slow and it is also prone to omissions. Secondly, manual observation is generally used for determination, resulting in a high degree of misjudgment, and therefore cannot meet the needs of users. Summary of the Invention

[0005] In order to solve the above problems, the present invention provides a method for adjusting the electrical characteristics of gallium oxide micro-defects.

[0006] The present invention adopts the following technical solution, a method for adjusting the electrical properties of gallium oxide micro-defects, comprising:

[0007] Acquire a TEM image of the gallium oxide material, wherein the TEM image is a bright field image, that is, an overall structural image of the gallium oxide material obtained using a bright field imaging mode, and determine the defect-sensitive area based on the overall structural image;

[0008] Acquire an area in the gallium oxide material corresponding to a defect-sensitive area, mark it as an inspection area, and acquire a dark field image and a high-resolution image of the inspection area;

[0009] Input the acquired dark field image and high-resolution image into a pre-built defect type recognition model to obtain the defect type of the inspection area, and determine the defect type based on the defect type, where the defect type includes repairable type and non-repairable type;

[0010] According to the repairable type, the defect type and the defect degree corresponding to the defect type are obtained, and the defect repair coefficient is calculated;

[0011] Input the collected defect types and defect repair coefficients into a pre-built doping parameter recommendation model to obtain a doping parameter recommendation set label, and then obtain a doping parameter recommendation set corresponding to the doping parameter recommendation set label;

[0012] According to the parameters in the recommended doping parameter set, the gallium oxide material is doped and regulated to optimize the electrical properties of the gallium oxide material.

[0013] As a further description of the above technical solution: the method for acquiring a TEM image of a gallium oxide material includes:

[0014] Step q1: Mount the gallium oxide material on the TEM sample holder to fix the gallium oxide material;

[0015] Step q2: Insert the sample holder into the TEM sample chamber and evacuate to 10 -6 Pa;

[0016] Step q3: Set the accelerating voltage to 200 kV and adjust the objective lens focus and beam intensity;

[0017] Step q4: Use bright field imaging mode to obtain the overall structural image of the gallium oxide material.

[0018] As a further description of the above technical solution: the method for determining defect-sensitive areas based on the overall structural image includes:

[0019] Step P1: grayscale processing is performed on the collected overall structural image to obtain the grayscale value of each pixel block in the overall structural image and mark it as a real-time grayscale value;

[0020] Step P2: Obtain image data of a normal gallium oxide material, perform grayscale processing, obtain the grayscale value of each pixel block in the standard image data, and mark it as a standard grayscale value;

[0021] Step P3: Compare and analyze the real-time grayscale value of each pixel block in the overall structural image with the standard grayscale value of the same position in the standard image data to determine whether to mark the pixel block as a defect;

[0022] Step P4: Extract pixel blocks with defect marks and record them as defective pixel blocks. Randomly select a defective pixel block as the target pixel block, preset a spacing threshold, obtain the distance between other defective pixel blocks and the target pixel block, and when the distance does not exceed the spacing threshold, use the obtained defective pixel block as the target pixel block again, and obtain the distance between other pixel blocks except the target pixel block and the target pixel block again. Traverse all defective pixel blocks and mark the area formed by all target pixel blocks as a defect-sensitive area.

[0023] Step P5: Repeat step P4 for the remaining defective pixel blocks until all defective pixel blocks are divided into defect-sensitive areas, thereby forming N defect-sensitive areas, where N≥1.

[0024] As a further description of the above technical solution: the method for obtaining a dark field image and a high-resolution image of the inspection area includes:

[0025] After using the bright field imaging mode to obtain the overall structural image of the gallium oxide material, the position and focus of the gallium oxide material are adjusted, and then the dark field imaging mode and high-resolution imaging mode are used to collect images respectively, thereby obtaining dark field images and high-resolution images of defect-sensitive areas.

[0026] As a further description of the above technical solution: the training method of the defect type recognition model includes:

[0027] Dark field images and high-resolution images of gallium oxide materials with different defect types are collected, and labels are set for the defect types of the dark field images and high-resolution images, that is, the labels are defect types. The dark field images and high-resolution images and the labels corresponding to the dark field images and high-resolution images constitute a set of training data. X sets of training data constitute a sample set, where X is an integer greater than 1. The sample set is divided into a training set and a test set. The dark field images and high-resolution images in the training set are used as inputs of a defect type recognition model, and the labels corresponding to the dark field images and high-resolution images in the training set are used as outputs of the defect type recognition model. The defect type recognition model is trained to obtain an initial defect type recognition model, and minimizing the sum of prediction errors is used as a training objective. The initial defect type recognition model is evaluated using a test set, and the initial defect type recognition model when the sum of prediction errors reaches convergence is used as the constructed defect type recognition model.

[0028] The defect type identification model is a deep neural network model.

[0029] As a further description of the above technical solution: the types of defects include: dislocation defects, stacking fault defects, grain boundary defects, antisite defects and impurity cluster defects;

[0030] Among them, the repairable types include dislocation defects, stacking fault defects and antilocation defects;

[0031] Unrepairable types include grain boundary defects and impurity cluster defects.

[0032] As a further description of the above technical solution: the calculation formula of the defect repair coefficient is:

[0033] QXxf=β1×WCqx+β2×CCqx+β3×FWqx;

[0034] Where QXxf is the defect repair coefficient, WCqx is the dislocation defect density, CCqx is the stacking fault defect density, FWqx is the antisite defect concentration, and β1, β2, and β3 are weight factors, all greater than 0.

[0035] As a further description of the above technical solution: the recommended set of doping parameters is {(a1, b1, c1), (a2, b2, c2), ..., (a n , b n , c n )},a n is the doping element corresponding to the set label n, b n is the doping concentration corresponding to the set label n, c n It is the doping method corresponding to the set label n.

[0036] As a further description of the above technical solution: the training method of the doping parameter recommendation model includes:

[0037] Set the recommended doping parameter sets with corresponding numbers in advance;

[0038] The defect types and defect repair coefficients are converted into a corresponding set of feature vectors, and each set of feature vectors is used as the input of the doping parameter recommendation model. The doping parameter recommendation model uses a set of doping parameter recommendation set numbers corresponding to each set of defect types and defect repair coefficients as output, and a set of doping parameter recommendation set numbers actually corresponding to each set of defect types and defect repair coefficients as prediction targets. Minimizing the loss function value of the doping parameter recommendation model is used as the training target. When the loss function value of the doping parameter recommendation model is less than or equal to the preset target loss value, training is stopped. The doping parameter recommendation model is one of the models such as support vector machine regression, random forest regression or neural network regression.

[0039] A system for regulating the electrical properties of gallium oxide micro-defects, which is used to implement the method for regulating the electrical properties of gallium oxide micro-defects, comprises:

[0040] A first image acquisition module is configured to acquire a TEM image of the gallium oxide material, wherein the TEM image is a bright field image, that is, an overall structural image of the gallium oxide material acquired using a bright field imaging mode;

[0041] Region division module, which determines defect-sensitive areas based on the overall structural image;

[0042] The second image acquisition module acquires an area in the gallium oxide material corresponding to the defect-sensitive area, marks it as an inspection area, and acquires a dark field image and a high-resolution image of the inspection area;

[0043] The defect recognition module inputs the acquired dark field image and high-resolution image into a pre-built defect type recognition model to obtain the defect type of the inspection area and determine the defect type based on the defect type, where the defect type includes repairable type and non-repairable type;

[0044] The data analysis module obtains the defect type and the corresponding defect degree according to the repairable type, and calculates and generates the defect repair coefficient;

[0045] The data processing module inputs the collected defect types and defect repair coefficients into a pre-built doping parameter recommendation model to obtain a doping parameter recommendation set label, and then obtains the doping parameter recommendation set corresponding to the doping parameter recommendation set label. Based on the parameters in the doping parameter recommendation set, the gallium oxide material is doped and regulated to optimize the electrical properties of the gallium oxide material.

[0046] Beneficial effects

[0047] The present invention provides a method for adjusting the electrical properties of gallium oxide micro-defects. First, an overall structural image of the gallium oxide material is acquired using a bright-field imaging mode. Based on the overall structural image, defect-sensitive areas are determined. Then, dark-field images and high-resolution images are collected from the defect-sensitive areas. Defect types are identified based on a trained, pre-built defect type identification model. This method eliminates the need to use a microscope to traverse the entire gallium oxide material, greatly improving defect detection efficiency. Furthermore, by determining the defect-sensitive areas, then collecting dark-field images and high-resolution images of the defect-sensitive areas and inputting them into the trained, pre-built defect type identification model for defect identification, no manual judgment of defect types is required, further improving the efficiency and stability of the judgment and enhancing the intelligence of defect detection.

[0048] Furthermore, by obtaining the defect type, determining the defect type based on the defect type, obtaining the defect type and the defect degree corresponding to the defect type according to the repairable type, calculating and generating the defect repair coefficient, the collected defect type and defect repair coefficient are input into the pre-built doping parameter recommendation model to obtain the doping parameter recommendation set label, and then obtaining the doping parameter recommendation set corresponding to the doping parameter recommendation set label. According to the parameters in the doping parameter recommendation set, the gallium oxide material is doped and regulated, that is, intelligent predictive doping regulation is realized. By introducing suitable doping elements, the energy level changes caused by the defects are corrected, and the electrical properties of the gallium oxide material are optimized. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The present invention will be further explained below in conjunction with the accompanying drawings and examples:

[0050] Figure 1 Schematic diagram of the process of adjusting the electrical characteristics of gallium oxide micro-defects provided by the embodiment of the present invention Figure 1 ;

[0051] Figure 2 Schematic diagram of the process of adjusting the electrical characteristics of gallium oxide micro-defects provided by the embodiment of the present invention Figure 2 ;

[0052] Figure 3 A flow chart of a method for determining sensitive areas in a method for adjusting electrical properties of gallium oxide micro-defects provided in an embodiment of the present invention;

[0053] Figure 4 A schematic diagram of the structure of a system for regulating the electrical properties of gallium oxide micro-defects provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0054] In order to make the technical means, creative features, objectives and effects of the present invention easier to understand, the present invention is further described below with reference to specific diagrams. It should be noted that the embodiments and features in the embodiments of this application can be combined with each other unless they conflict.

[0055] Example 1

[0056] See also Figure 1 and Figure 2 The embodiment of the present invention provides a technical solution: a method for adjusting the electrical characteristics of microscopic defects in gallium oxide.

[0057] Acquire a TEM image of the gallium oxide material, wherein the TEM image is a bright field image, that is, an overall structural image of the gallium oxide material obtained using a bright field imaging mode, and determine the defect-sensitive area based on the overall structural image;

[0058] It should be noted that TEM images are transmission electron microscope images, which can provide nanometer-level resolution and directly observe microscopic defects in gallium oxide materials.

[0059] Methods for acquiring TEM images of gallium oxide materials include:

[0060] Step q1: Mount the gallium oxide material on the TEM sample holder to fix the gallium oxide material;

[0061] Step q2: Insert the sample holder into the TEM sample chamber and evacuate to 10 -6 Pa;

[0062] Step q3: Set the accelerating voltage to 200 kV and adjust the objective lens focus and beam intensity;

[0063] Step q4: Use bright field imaging mode to obtain the overall structural image of the gallium oxide material.

[0064] Acquire an area in the gallium oxide material corresponding to a defect-sensitive area, mark it as an inspection area, and acquire a dark field image and a high-resolution image of the inspection area;

[0065] Methods for acquiring darkfield and high-resolution images of the inspection area include:

[0066] After using the bright field imaging mode to obtain the overall structural image of the gallium oxide material, the position and focus of the gallium oxide material are adjusted, and then the dark field imaging mode and high-resolution imaging mode are used to collect images respectively, thereby obtaining dark field images and high-resolution images of defect-sensitive areas.

[0067] It should be noted that the dark field imaging mode is used to obtain dark field images, which facilitates the observation of defects and grain boundaries in the sample, enhances contrast, and allows for detailed analysis of dislocations and stacking faults. The high-resolution imaging mode is used to obtain high-resolution images, which can observe lattice structure and atomic-level defects.

[0068] Input the acquired dark field image and high-resolution image into a pre-built defect type recognition model to obtain the defect type of the inspection area, and determine the defect type based on the defect type, where the defect type includes repairable type and non-repairable type;

[0069] The training methods for the defect type recognition model include:

[0070] Dark field images and high-resolution images of gallium oxide materials with different defect types are collected, and labels are set for the defect types of the dark field images and high-resolution images, that is, the labels are defect types. The dark field images and high-resolution images and the labels corresponding to the dark field images and high-resolution images constitute a set of training data. X sets of training data constitute a sample set, where X is an integer greater than 1. The sample set is divided into a training set and a test set. The dark field images and high-resolution images in the training set are used as inputs of a defect type recognition model, and the labels corresponding to the dark field images and high-resolution images in the training set are used as outputs of the defect type recognition model. The defect type recognition model is trained to obtain an initial defect type recognition model, and minimizing the sum of prediction errors is used as a training objective. The initial defect type recognition model is evaluated using a test set, and the initial defect type recognition model when the sum of prediction errors reaches convergence is used as the constructed defect type recognition model.

[0071] The defect type identification model is a deep neural network model.

[0072] It should be noted that in the machine learning model, the calculation formula for prediction error is: k =(a k -w k ) 2 , where k is the number of the feature data (i.e., extracted image), Z k is the prediction error, a k is the predicted state value corresponding to the kth set of feature data. The state value is the defect type. w k is the actual state value corresponding to the kth group of training data.

[0073] Specifically, an overall structural image of the gallium oxide material is acquired using a brightfield imaging mode, and defect-sensitive areas are determined based on the overall structural image. Darkfield images and high-resolution images are then collected of the defect-sensitive areas, and defect type identification is performed based on a trained pre-built defect type identification model. This method of defect identification eliminates the need for a microscope system to traverse the entire gallium oxide material, thereby greatly improving defect detection efficiency. Furthermore, by determining the defect-sensitive areas, and then collecting darkfield images and high-resolution images of the defect-sensitive areas and inputting them into a trained pre-built defect type identification model for defect identification, there is no need for manual judgment of the defect type, further improving the efficiency and stability of the judgment and enhancing the intelligence of defect detection.

[0074] The defect types include: dislocation defects, stacking fault defects, grain boundary defects, antisite defects and impurity cluster defects;

[0075] Among them, the repairable types include dislocation defects, stacking fault defects and antilocation defects;

[0076] Unrepairable types include grain boundary defects and impurity cluster defects.

[0077] It should be noted that for dislocation defects, stacking fault defects and antisite defects, the generation of defects can be reduced by controlling the doping concentration and type, and high-temperature treatment can promote the migration and rearrangement of atoms or defects, thereby reducing the defect density;

[0078] Grain boundary defects are inherent interfaces between grains of grain boundary materials. Their formation and existence are part of the material's microstructure and usually cannot be treated or eliminated. Impurity clusters are formed by the aggregation of foreign elements in the material. These defects usually exist stably and are difficult to completely eliminate through annealing or other treatment methods.

[0079] What you need to know is that dislocation defects will reduce carrier mobility and conductivity, increase leakage current and local stress; stacking fault defects will change the band structure, introduce defect states, and affect the capacitance effect and breakdown voltage; grain boundary defects will form interface potential barriers, increase trap states, reduce conductivity and device consistency, and antisite defects will change the local doping concentration, introduce defect states, increase carrier recombination and conductivity changes; impurity cluster defects will cause local conductivity changes, increase leakage current, and affect device stability.

[0080] According to the repairable type, the defect type and the defect degree corresponding to the defect type are obtained, and the defect repair coefficient is calculated and generated.

[0081] The calculation formula of the defect repair coefficient is:

[0082] QXxf=β1×WCqx+β2×CCqx+β3×FWqx;

[0083] Where QXxf is the defect repair coefficient, WCqx is the dislocation defect density, CCqx is the stacking fault defect density, FWqx is the antisite defect concentration, and β1, β2, and β3 are weight factors, all greater than 0.

[0084] It should be noted that the size of the weight factor is a specific numerical value obtained by quantifying each data to facilitate subsequent comparison. The size of the weight factor depends on the number of comprehensive parameters and the preliminary setting of the corresponding weight factor for each set of comprehensive parameters by technical personnel in this field.

[0085] It should be noted that the dislocation defect density is obtained by X-ray diffraction technology, which is an indirect but effective method for obtaining dislocation density. This process mainly includes measuring the rocking curve of the crystal and the full width at half maximum (FWHM) of the analysis peak, and then calculating the dislocation density according to a theoretical model, wherein the theoretical model is to calculate the dislocation density using the Wilson-Hal 1 method.

[0086] The stacking fault density is an indirect but quantitative characterization method that measures the broadening and deformation of the diffraction peak using X-ray diffraction technology, extracts the peak's half-maximum width (FWHM) and symmetry parameters, and calculates the stacking fault density in combination with a theoretical model. This method can effectively evaluate the degree of stacking fault defects in materials.

[0087] Example: Assume that the (111) diffraction peak of gallium oxide crystal is selected for measurement;

[0088] Perform fine angle scanning (e.g., 0.01° per step) within a range of approximately 2θ = 35.7° (the specific peak position is determined by the material), and record the diffraction curve showing the change in diffraction intensity with angle;

[0089] The half-width (FWHM) and symmetry parameters of the peaks are obtained by Gaussian fitting of the diffraction peaks. Assuming that the measured FWHM is 0.2° and the intrinsic peak width of the defect-free crystal is 0.1°,

[0090] Using the Wilson equation:

[0091] Assume that the interlayer distance d = 2.5 × 10 -8 ;

[0092] Calculate the remaining width: β-β0=0.2°-0.1°=0.1°;

[0093] Convert to radians:

[0094] Calculate the stacking fault density:

[0095] The method for obtaining the antisite defect concentration is to determine the antisite defect concentration based on XPS. The specific process is to quantitatively calculate the concentration of antisite defects through high-resolution XPS measurement, background subtraction, peak fitting and chemical shift analysis. Determining the antisite defect concentration by XPS is an existing technology and will not be elaborated on here.

[0096] Input the collected defect types and defect repair coefficients into a pre-built doping parameter recommendation model to obtain a doping parameter recommendation set label, and then obtain a doping parameter recommendation set corresponding to the doping parameter recommendation set label;

[0097] According to the parameters in the recommended doping parameter set, the gallium oxide material is doped and regulated to optimize the electrical properties of the gallium oxide material.

[0098] The recommended set of doping parameters is {(a1, b1, c1), (a2, b2, c2), ..., (a n , b n , c n )},a n is the doping element corresponding to the set label n, b n is the doping concentration corresponding to the set label n, c n is the doping mode corresponding to the set label n;

[0099] It should be noted that the doping elements include:

[0100] Silicon (Si): A commonly used n-type doping element that can effectively increase carrier concentration;

[0101] Tin (Sn): a commonly used n-type doping element with good doping efficiency;

[0102] Antimony (Sb): Although less commonly used, it can also be used as an n-type dopant element in some applications;

[0103] Nitrogen (N): Nitrogen can provide additional free electrons by replacing oxygen, but its doping efficiency may be low;

[0104] Magnesium (Mg): It is considered a potential element for achieving p-type doping, but the doping efficiency and stability are low;

[0105] Copper (Cu): also used to try p-type doping, with certain feasibility;

[0106] Lithium (Li): Holes can be introduced in certain circumstances, but achieving stable p-type conductivity remains challenging.

[0107] Titanium (Ti): can be used as a doping element to regulate the optical and electrical properties of materials;

[0108] Manganese (Mn): used to regulate the magnetic and electrical properties of materials;

[0109] Cobalt (Co): used to regulate the magnetic and electrical properties of materials.

[0110] Doping methods include:

[0111] Chemical vapor deposition (CVD): Deposition is performed at high temperature using a doping gas source (such as SiH4, SnC14), controlling the gas flow rate and reaction time.

[0112] Physical Vapor Deposition (PVD): Sputtering deposition using doped targets, controlling power, gas pressure, and deposition time.

[0113] Sol-gel method: The doping element is dissolved in a precursor solution, and a thin film is prepared by spin coating or drop coating, followed by heat treatment to form doped gallium oxide.

[0114] The training method of the doping parameter recommendation model includes:

[0115] Set the recommended doping parameter sets with corresponding numbers in advance;

[0116] The defect types and defect repair coefficients are converted into a corresponding set of feature vectors, and each set of feature vectors is used as the input of the doping parameter recommendation model. The doping parameter recommendation model uses a set of doping parameter recommendation set numbers corresponding to each set of defect types and defect repair coefficients as output, and a set of doping parameter recommendation set numbers actually corresponding to each set of defect types and defect repair coefficients as prediction targets. Minimizing the loss function value of the doping parameter recommendation model is used as the training target. When the loss function value of the doping parameter recommendation model is less than or equal to the preset target loss value, training is stopped. The doping parameter recommendation model is one of the models such as support vector machine regression, random forest regression or neural network regression.

[0117] The loss function value of the doping parameter recommendation model is the mean square error.

[0118] By transforming the loss function The model is trained with minimization as the goal, so that the doping parameter recommendation model can better fit the data, thereby improving the performance and accuracy of the model. In the loss function, MSE is the loss function value of the doping parameter recommendation model, x is the eigenvector group number; m is the number of eigenvector groups; y x is the recommended set number of doping parameters corresponding to the xth group of eigenvectors, The recommended set number of the doping parameter corresponding to the x-th group of eigenvectors in real time.

[0119] Specifically, by obtaining the defect type, determining the defect type based on the defect type, obtaining the defect type and the defect degree corresponding to the defect type according to the repairable type, calculating and generating the defect repair coefficient, and inputting the collected defect type and defect repair coefficient into the pre-built doping parameter recommendation model to obtain the doping parameter recommendation set label, and then obtaining the doping parameter recommendation set corresponding to the doping parameter recommendation set label. According to the parameters in the doping parameter recommendation set, the gallium oxide material is doped and regulated, that is, intelligent predictive doping regulation is realized. By introducing suitable doping elements, the energy level changes caused by the defects are corrected, and the electrical properties of the gallium oxide material are optimized.

[0120] Example 2

[0121] See also Figure 3 This embodiment provides a method for determining defect-sensitive areas based on an overall structural image, including:

[0122] Step P1: grayscale processing is performed on the collected overall structural image to obtain the grayscale value of each pixel block in the overall structural image and mark it as a real-time grayscale value;

[0123] Step P2: Obtain image data of a normal gallium oxide material, perform grayscale processing, obtain the grayscale value of each pixel block in the standard image data, and mark it as a standard grayscale value;

[0124] Step P3: Compare and analyze the real-time grayscale value of each pixel block in the overall structural image with the standard grayscale value of the same position in the standard image data to determine whether to mark the pixel block as a defect;

[0125] Step P4: Extract pixel blocks with defect marks and record them as defective pixel blocks. Randomly select a defective pixel block as the target pixel block, preset a spacing threshold, obtain the distance between other defective pixel blocks and the target pixel block, and when the distance does not exceed the spacing threshold, use the obtained defective pixel block as the target pixel block again, and obtain the distance between other pixel blocks except the target pixel block and the target pixel block again. Traverse all defective pixel blocks and mark the area formed by all target pixel blocks as a defect-sensitive area.

[0126] Step P5: Repeat step P4 for the remaining defective pixel blocks until all defective pixel blocks are divided into defect-sensitive areas, thereby forming N defect-sensitive areas, where N≥1.

[0127] Example 3

[0128] See also Figure 4 This embodiment provides a system for adjusting the electrical properties of gallium oxide micro-defects, which is used to implement the method for adjusting the electrical properties of gallium oxide micro-defects, including:

[0129] A first image acquisition module is configured to acquire a TEM image of the gallium oxide material, wherein the TEM image is a bright field image, that is, an overall structural image of the gallium oxide material acquired using a bright field imaging mode;

[0130] Region division module, which determines defect-sensitive areas based on the overall structural image;

[0131] The second image acquisition module acquires an area in the gallium oxide material corresponding to the defect-sensitive area, marks it as an inspection area, and acquires a dark field image and a high-resolution image of the inspection area;

[0132] The defect recognition module inputs the acquired dark field image and high-resolution image into a pre-built defect type recognition model to obtain the defect type of the inspection area and determine the defect type based on the defect type, where the defect type includes repairable type and non-repairable type;

[0133] The data analysis module obtains the defect type and the corresponding defect degree according to the repairable type, and calculates and generates the defect repair coefficient;

[0134] The data processing module inputs the collected defect types and defect repair coefficients into a pre-built doping parameter recommendation model to obtain a doping parameter recommendation set label, and then obtains the doping parameter recommendation set corresponding to the doping parameter recommendation set label. Based on the parameters in the doping parameter recommendation set, the gallium oxide material is doped and regulated to optimize the electrical properties of the gallium oxide material.

[0135] The basic principles, main features, and advantages of the present invention are shown and described above. Those skilled in the art should understand that the present invention is not limited to the above-described embodiments. The above-described embodiments and descriptions are merely illustrative of the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for regulating the electrical properties of gallium oxide micro-defects, characterized in that: include: Acquire a TEM image of the gallium oxide material, wherein the TEM image is a bright field image, that is, an overall structural image of the gallium oxide material obtained using a bright field imaging mode, and determine the defect-sensitive area based on the overall structural image; Acquire an area in the gallium oxide material corresponding to a defect-sensitive area, mark it as an inspection area, and acquire a dark field image and a high-resolution image of the inspection area; The method for obtaining a dark field image and a high-resolution image of an inspection area comprises: After obtaining an overall structural image of the gallium oxide material using the bright-field imaging mode, the position and focus of the gallium oxide material are adjusted. Then, dark-field imaging mode and high-resolution imaging mode are used to acquire images, respectively, thereby obtaining dark-field images and high-resolution images of defect-sensitive areas. Input the acquired dark field image and high-resolution image into a pre-built defect type recognition model to obtain the defect type of the inspection area, and determine the defect type based on the defect type, where the defect type includes repairable type and non-repairable type; According to the repairable type, the defect type and the defect degree corresponding to the defect type are obtained, and the defect repair coefficient is calculated; The collected defect types and defect repair coefficients are input into a pre-built doping parameter recommendation model to obtain a doping parameter recommendation set label. Then, a doping parameter recommendation set corresponding to the doping parameter recommendation set label is obtained. Based on the parameters in the doping parameter recommendation set, the gallium oxide material is doped and regulated to optimize the electrical properties of the gallium oxide material. The recommended set of doping parameters is: , The collection label is The corresponding doping element is The collection label is The corresponding doping concentration is The collection label is The corresponding doping method; The training method of the doping parameter recommendation model includes: Set the recommended doping parameter sets with corresponding numbers in advance; The defect types and defect repair coefficients are converted into a corresponding set of feature vectors, and each set of feature vectors is used as the input of the doping parameter recommendation model. The doping parameter recommendation model uses a set of doping parameter recommendation set numbers corresponding to each set of defect types and defect repair coefficients as output, a set of doping parameter recommendation set numbers actually corresponding to each set of defect types and defect repair coefficients as prediction targets, and minimizing the loss function value of the doping parameter recommendation model as the training target. Training is stopped when the loss function value of the doping parameter recommendation model is less than or equal to the preset target loss value.

2. The method for adjusting the electrical characteristics of gallium oxide micro-defects according to claim 1, characterized in that: The method for collecting a TEM image of a gallium oxide material comprises: Step q1: Mount the gallium oxide material on the TEM sample holder to fix the gallium oxide material; Step q2: Insert the sample holder into the TEM sample chamber and evacuate to Pa; Step q3: Set the accelerating voltage to 200 kV and adjust the objective lens focus and beam intensity; Step q4: Use bright field imaging mode to obtain the overall structural image of the gallium oxide material.

3. The method for adjusting the electrical characteristics of gallium oxide micro-defects according to claim 1, characterized in that: The method for determining defect-sensitive areas based on the overall structural image includes: Step P1: grayscale processing is performed on the collected overall structural image to obtain the grayscale value of each pixel block in the overall structural image and mark it as a real-time grayscale value; Step P2: Obtain image data of a normal gallium oxide material, perform grayscale processing, obtain the grayscale value of each pixel block in the standard image data, and mark it as a standard grayscale value; Step P3: Compare and analyze the real-time grayscale value of each pixel block in the overall structural image with the standard grayscale value of the same position in the standard image data to determine whether to mark the pixel block as a defect; Step P4: Extract pixel blocks with defect marks and record them as defective pixel blocks. Randomly select a defective pixel block as the target pixel block, preset a spacing threshold, obtain the distance between other defective pixel blocks and the target pixel block, and when the distance does not exceed the spacing threshold, use the obtained defective pixel block as the target pixel block again, and obtain the distance between other pixel blocks except the target pixel block and the target pixel block again. Traverse all defective pixel blocks and mark the area formed by all target pixel blocks as a defect-sensitive area. Step P5: Repeat step P4 for the remaining defective pixel blocks until all defective pixel blocks are divided into defect-sensitive areas, thereby forming N defect-sensitive areas.

4. The method for adjusting the electrical characteristics of gallium oxide micro-defects according to claim 1, wherein: The training method of the defect type recognition model includes: Dark field images and high-resolution images of gallium oxide materials with different defect types are collected, and labels are set for the defect types of the dark field images and high-resolution images, that is, the labels are defect types. The dark field images, high-resolution images, and the labels corresponding to the dark field images and high-resolution images constitute a set of training data. X sets of training data constitute a sample set, where X is an integer greater than 1. The sample set is divided into a training set and a test set. The dark field images and high-resolution images in the training set are used as inputs of a defect type recognition model, and the labels corresponding to the dark field images and high-resolution images in the training set are used as outputs of the defect type recognition model. The defect type recognition model is trained to obtain an initial defect type recognition model, and minimizing the sum of prediction errors is used as a training objective. The initial defect type recognition model is evaluated using a test set, and the initial defect type recognition model when the sum of prediction errors reaches convergence is used as the constructed defect type recognition model. The defect type identification model is a deep neural network model.

5. The method for adjusting the electrical characteristics of gallium oxide micro-defects according to claim 1, characterized in that: The defect types include: dislocation defects, stacking fault defects, grain boundary defects, antisite defects and impurity cluster defects; Among them, the repairable types include dislocation defects, stacking fault defects and antilocation defects; Unrepairable types include grain boundary defects and impurity cluster defects.

6. The method for adjusting the electrical characteristics of gallium oxide micro-defects according to claim 1, characterized in that: The calculation formula of the defect repair coefficient is: ； Where, is the defect repair coefficient, is the dislocation defect density, is the stacking fault density, is the antisite defect concentration, 、 and are weight factors, all greater than 0.

7. A system for regulating the electrical properties of gallium oxide micro-defects, for implementing the method for regulating the electrical properties of gallium oxide micro-defects according to any one of claims 1 to 6, characterized in that: include: A first image acquisition module is configured to acquire a TEM image of the gallium oxide material, wherein the TEM image is a bright field image, that is, an overall structural image of the gallium oxide material acquired using a bright field imaging mode; Region division module, which determines defect-sensitive areas based on the overall structural image; The second image acquisition module acquires an area in the gallium oxide material corresponding to the defect-sensitive area, marks it as an inspection area, and acquires a dark field image and a high-resolution image of the inspection area; The defect recognition module inputs the acquired dark field image and high-resolution image into a pre-built defect type recognition model to obtain the defect type of the inspection area and determine the defect type based on the defect type, where the defect type includes repairable type and non-repairable type; The data analysis module obtains the defect type and the corresponding defect degree according to the repairable type, and calculates and generates the defect repair coefficient; The data processing module inputs the collected defect types and defect repair coefficients into a pre-built doping parameter recommendation model to obtain a doping parameter recommendation set label, and then obtains the doping parameter recommendation set corresponding to the doping parameter recommendation set label. Based on the parameters in the doping parameter recommendation set, the gallium oxide material is doped and regulated to optimize the electrical properties of the gallium oxide material.

Images