Method and system for adjusting electrical characteristics of gallium oxide micro-defects

Abstract

The present application belongs to the technical field of gallium oxide doping regulation, and particularly relates to a gallium oxide micro-defect electrical property regulation method and system, which comprises: collecting a transmission electron microscope (TEM) bright field whole image of gallium oxide material; determining a plurality of defect sensitive areas, and performing stable cluster purification and regional clustering to form defect sensitive area clusters and de-cluster abnormal points; determining a representative area for each defect sensitive area cluster, collecting dark field images and high resolution images of the representative area and identifying defect types, and independently identifying the de-cluster abnormal points; pre-flowing according to the unrepairable leading area based on the representative area identification result; calculating a repair coefficient and recommending a doping parameter for the un-flowed area; the system is used to implement the above method, and the present application can reduce the subsequent image collection quantity and area-by-area identification burden, and improve the gallium oxide micro-defect detection and electrical property regulation efficiency.

Description

Technical Field

[0001] This invention pertains to gallium oxide doping control methods, specifically referring to methods and systems for adjusting the electrical properties of gallium oxide micro-defects. Background Technology

[0002] Gallium oxide (GaO) materials possess wide bandgap, high breakdown electric field, and good thermal stability, making them highly valuable for applications in high-power electronic devices and high-frequency devices. However, GaO materials often contain microscopic defects such as dislocations, stacking faults, grain boundaries, antisites, and impurity clusters. Different types of microscopic defects can adversely affect carrier mobility, conductivity, leakage current, breakdown voltage, and device consistency to varying degrees. Therefore, accurately acquiring information about microscopic defects and implementing targeted adjustments to electrical characteristics has become a crucial issue in the optimization of GaO materials.

[0003] The prior art CN119048455A discloses a method for adjusting the electrical properties of microscopic defects in gallium oxide. First, a bright-field TEM image of the gallium oxide material is acquired, and defect-sensitive regions are determined based on this image. Then, dark-field and high-resolution images are acquired for each defect-sensitive region. The acquired dark-field and high-resolution images are input into a defect recognition model to obtain the defect type for the corresponding region, thereby distinguishing between repairable and unrepairable types. Subsequently, a repair coefficient is calculated for repairable types based on the defect severity, and doping parameters are recommended according to the defect type and repair coefficient.

[0004] Compared to directly traversing the entire material area for high-precision detection, this approach can narrow down the scope of subsequent detailed inspection and improve defect detection efficiency. However, when multiple defect-sensitive areas exist in the sample, it is usually still necessary to acquire dark-field images and high-resolution images for each defect-sensitive area separately, and then distinguish the defect type and determine whether it is repairable or not. As a result, a large number of dark-field images and high-resolution images still need to be acquired subsequently, and type identification needs to be performed on multiple areas one by one, leading to a heavy burden on subsequent image acquisition and identification. Summary of the Invention

[0005] The purpose of this invention is to provide a method for adjusting the electrical properties of gallium oxide microdefects to solve the above-mentioned technical problems, comprising the following steps:

[0006] Acquire a full-field bright-field image of gallium oxide material using transmission electron microscopy (TEM).

[0007] Based on the overall bright-field image, multiple defect-sensitive regions are determined, and stable cluster purification and regional clustering are performed according to the regional correlation between the defect-sensitive regions to form at least one defect-sensitive region cluster and obtain de-clustering anomalies.

[0008] For each cluster of defect-sensitive regions, a representative region is determined. Dark field images and high-resolution images of the representative regions are acquired, and the dark field images and high-resolution images are input into a defect type recognition model to obtain the defect type of the representative region.

[0009] Anomaly identification is performed independently for each of the aforementioned de-clustering anomalies;

[0010] Based on the identification result that the representative area belongs to the unrepairable type, the unrepairable dominant index is determined, and when the unrepairable dominant index meets the preset conditions, the corresponding defect-sensitive area cluster is pre-diverted into an unrepairable restricted cluster.

[0011] For regions that are not pre-shunted, a repair coefficient is calculated based on the repairable type, and the defect type and the repair coefficient are input into the doping parameter recommendation model to obtain the doping parameter recommendation result.

[0012] Furthermore, the representative region is the joint region corresponding to the connecting position, confluence position, or through position within the defect-sensitive region cluster.

[0013] Furthermore, the regional association relationship is determined based on the boundary distance, grayscale difference, boundary direction difference, and connection zone stability between defect-sensitive regions.

[0014] Furthermore, the region clustering includes: constructing a region graph with each of the defect-sensitive regions as nodes and the region connection relationships that satisfy the stable connection conditions as edges, and determining the connected subgraphs in the region graph as the corresponding defect-sensitive region clusters.

[0015] Furthermore, the stable cluster purification includes: weakening or severing unstable associations formed by weak connections, pseudo-bridging, or local noise, and retaining the abnormal regions that failed to stably merge into any of the defect-sensitive region clusters as the declustering anomalies.

[0016] Furthermore, the de-clustering anomalies include boundary anomalies, pseudo-bridging points, and independent anomalies.

[0017] Furthermore, the unrepairable types include grain boundary defects and impurity cluster defects, and the unrepairable dominant index is determined based on the identification results of the representative region belonging to grain boundary defects and impurity cluster defects.

[0018] Furthermore, the repairable types include dislocation defects, stacking fault defects, and inversion defects, and the repair coefficient is determined based on the dislocation defect density, stacking fault defect density, and inversion defect concentration.

[0019] Furthermore, the recommended doping parameters include at least one of the following: doping element type, doping concentration, annealing temperature, and annealing time.

[0020] On the other hand, the present invention also provides a system for adjusting the electrical properties of gallium oxide microdefects, comprising:

[0021] Bright-field image acquisition module, used to acquire overall bright-field images of gallium oxide materials using transmission electron microscopy (TEM);

[0022] The region cluster construction module is used to determine multiple defect-sensitive regions based on the overall bright-field image, and to perform stable cluster purification and region clustering according to the regional correlation between the defect-sensitive regions to form at least one defect-sensitive region cluster and obtain de-clustering anomalies.

[0023] The representative area determination module is used to determine representative areas for each of the aforementioned defect-sensitive region clusters;

[0024] The second image acquisition module is used to acquire dark field images and high-resolution images of the representative area;

[0025] The defect identification module is used to input the dark field image and the high-resolution image into the defect type identification model to obtain the defect type of the representative area, and to independently perform anomaly identification on the de-clustering anomaly points;

[0026] The pre-shunting module is used to determine the irreparable dominant index based on the identification result that the representative area belongs to the irreparable type, and when the irreparable dominant index meets the preset conditions, pre-shunt the corresponding defect-sensitive area cluster into an irreparable restricted cluster.

[0027] The data processing module is used to calculate the repair coefficient based on the repairable type for areas that have not been pre-shunted, and input the defect type and the repair coefficient into the doping parameter recommendation model to obtain the doping parameter recommendation result.

[0028] The beneficial effects achieved by the present invention using the above solution are as follows:

[0029] (1) This invention first performs stable cluster purification and region clustering on multiple defect-sensitive regions based on the overall bright-field TEM image to form defect-sensitive region clusters, and then determines representative regions for each defect-sensitive region cluster. Next, dark-field images and high-resolution images are acquired for defect identification in the representative regions. Compared with the method of acquiring dark-field images and high-resolution images of each defect-sensitive region individually and identifying them one by one, this invention can reduce the number of subsequent detailed inspection images acquired and the burden of region-by-region identification, thereby improving the processing efficiency of gallium oxide micro-defect detection and subsequent electrical property adjustment.

[0030] (2) The present invention also determines the irreparable dominant index based on the identification result of the representative area belonging to the irreparable type, and when the preset conditions are met, the corresponding defect sensitive area cluster is pre-diverted into the irreparable restricted cluster, so that the irreparable restricted cluster does not enter the repair coefficient calculation. Thus, it can avoid the irreparable type dominant areas such as grain boundary defects and impurity cluster defects from continuing to participate in the repair coefficient calculation and doping parameter recommendation process for the repairable type, thereby reducing the subsequent doping adjustment offset and improving the pertinence and reliability of the electrical property adjustment of gallium oxide materials. Attached Figure Description

[0031] Figure 1 This is a schematic diagram of the electrical characteristic adjustment system architecture of the present invention.

[0032] Figure 2 This is a flowchart of the method for adjusting the electrical properties of gallium oxide microdefects according to the present invention.

[0033] Figure 3 This is a schematic diagram of the formation of defect-sensitive region clusters and the determination of representative regions according to the present invention. The left side shows multiple abnormal regions in the overall bright field image, and the right side shows defect-sensitive region cluster A, defect-sensitive region cluster B, and de-clustering anomalies formed after stable cluster purification and region clustering. Detailed Implementation

[0034] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0035] like Figure 1-3 As shown, the present invention provides a method and system for adjusting the electrical properties of gallium oxide micro-defects, which is used to reduce the number of subsequent dark field images and high-resolution images acquired when there are multiple defect-sensitive regions in gallium oxide materials, and to reduce the interference of the unrepairable type dominant region on the subsequent doping parameter recommendation results.

[0036] like Figure 2 , 3 As shown, the method for adjusting the electrical properties of gallium oxide microdefects provided by the present invention first obtains a bright-field overall image of the gallium oxide material to be adjusted using a transmission electron microscope (TEM), and then extracts multiple bright-field anomalous regions based on the overall bright-field image.

[0037] To improve the consistency of subsequent defect-sensitive region identification, grayscale normalization, background brightness correction, and noise suppression can be performed on the overall bright-field image first. Then, based on local grayscale anomalies, texture abrupt changes, edge aggregation degree, and regional morphological anomalies, multiple candidate anomaly regions can be extracted from the overall bright-field image, and the candidate anomaly regions that meet the preset anomaly conditions can be identified as defect-sensitive regions.

[0038] Extract the center coordinates, average gray value, main direction of boundary, area, and connection information with neighboring areas for each defect-sensitive region.

[0039] For any two defect-sensitive areas and Constructing regional correlation strength Regional correlation strength Calculate using the following formula:

[0040]

[0041] in, Indicates the area With the region The closest distance between the boundaries and Representing regions With the region Average gray value, and Representing regions With the region The principal direction angle of the boundary, This represents the stability coefficient of the connecting zone between two regions. , , , These are the weighting coefficients. and These are the normalized parameters.

[0042] when Greater than or equal to the preset stable connection threshold At that time, the area was determined. With the region There is a stable connection between them; when Less than the preset stable connection threshold At that time, the area was determined. With the region The connections between them are unstable. Based on this determination, stable cluster purification is performed on defect-sensitive regions to weaken or sever unstable associations formed by weak connections, pseudo-bridging, or localized sporadic noise.

[0043] Construct all defect-sensitive areas into a region map , where the set of nodes For each defect-sensitive region, the edge set Corresponding to satisfy A stable connection relationship.

[0044] Connected subgraphs are extracted from the region map, and each connected subgraph corresponds to a defect-sensitive region cluster. Abnormal regions that fail to be incorporated into any stable connected subgraph, or are stripped during the purification process of stable clusters, are retained as de-clustered anomalies. The resulting defect-sensitive region clusters are not ordinary image grouping results, but rather structural objects used to characterize the continuous distribution relationship of the same anomalous structure.

[0045] Preferably, the de-clustering anomalies include boundary anomalies, pseudo-bridging points, and independent anomalies. Boundary anomalies are located at the edges of the anomalous structure; pseudo-bridging points are weak anomalies that temporarily connect two originally independent regions; and independent anomalies are isolated anomalous regions that lack a stable and continuous relationship with the surrounding regions. Retaining de-clustering anomalies helps prevent anomalous information from being directly discarded during the stabilization cluster purification process.

[0046] Preferably, a representative region is determined for each defect-sensitive region cluster. To ensure that the representative region can characterize the dominant defect type of the entire cluster, rather than just local edge anomalies, each candidate region within the cluster... Calculate joint score Key point scoring Calculate using the following formula:

[0047]

[0048] in, Indicates candidate region The connectivity within the cluster of defect-sensitive regions. Indicates candidate region betweenness centrality, Indicates candidate region The degree of convergence, Indicates candidate region The extent to which the original cluster splits after removal , , , These are the weighting coefficients.

[0049] Preferably, the candidate region with the highest key point score is determined as the representative region, that is:

[0050]

[0051] in, Indicates the first A cluster of defect-sensitive regions. The representative region determined in this way preferably corresponds to the key point region corresponding to the connection position, confluence position, or through position within the cluster. Therefore, compared with random point selection or edge selection, it can better reflect the dominant defect nature of the entire cluster. The representative region can be the key point region corresponding to the connection position within the cluster, the key point region corresponding to the confluence position of multiple abnormal sub-regions, or the key point region corresponding to the through position of abnormal continuous paths.

[0052] Preferably, dark-field images and high-resolution images are acquired only for each representative region, and the acquired dark-field images and high-resolution images are input into the defect type identification model to output a probability vector indicating whether the representative region belongs to a certain type of defect:

[0053]

[0054] in, , , , , These represent the identification probabilities of the representative region belonging to dislocation defects, stacking fault defects, grain boundary defects, antisite defects, and impurity cluster defects, respectively.

[0055] Preferably, anomaly identification is performed independently for each detached anomaly, without inheriting the identification results from neighboring defect-sensitive regions. The anomaly identification model inputs the local image corresponding to a single detached anomaly and outputs its defect category probability vector:

[0056]

[0057] When the following formula is true, the declustering anomaly is treated as an independent anomaly object and participates in subsequent result recording:

[0058]

[0059] when When this happens, the out-of-cluster anomaly is marked as a low-confidence anomaly. This method avoids directly ignoring boundary anomalies, pseudo-bridging points, or independent anomalies, thereby improving the completeness and reliability of the overall identification results.

[0060] Preferably, an irreparable dominant index is constructed based on the identification results of the representative areas. Irreparable dominant index Calculate using the following formula:

[0061]

[0062] in, and For weighting coefficients. When Greater than or equal to the preset unrepairable threshold When the representative area is determined to be dominated by an unrepairable type of defect-sensitive region cluster, this cluster is pre-diverted into an unrepairable restricted cluster before entering the repair coefficient calculation; when Less than the preset unrepairable threshold If the condition is met, the cluster is allowed to proceed to the subsequent repair coefficient calculation step. Unrepairable types include grain boundary defects and impurity cluster defects.

[0063] Pre-shunting is not merely for reducing computational steps, but also to prevent regions dominated by grain boundary defects or impurity cluster defects from continuing to participate in the repair coefficient calculation for repairable types, thereby reducing the deviation of subsequent doping parameter recommendations. When a defect-sensitive region cluster has been identified as being dominated by an unrepairable type, including it in the repair coefficient calculation chain based on the repairable type could easily cause the output doping parameters to deviate from the truly suitable adjustment direction. Therefore, it is pre-shunted as an unrepairable restricted cluster and prevented from entering the repair coefficient calculation step.

[0064] Preferably, for regions not pre-shunted, a repair coefficient is calculated based on the repairable type. Repairable types include dislocation defects, stacking fault defects, and inversion defects. The dislocation defect density is extracted for each. Stacking fault density and the concentration of inversion defects After normalization, we get:

[0065]

[0066] Repair coefficient Calculate using the following formula:

[0067]

[0068] in, , , Let be the weighting coefficient, and satisfy:

[0069]

[0070] Repair coefficient Used to characterize the degree of repair need or adjustment priority of the target region to improve its electrical properties through subsequent doping.

[0071] Preferably, the defect type and repair coefficient are input into the doping parameter recommendation model to establish a mapping relationship between the input features and the target doping parameters. The input feature vector can be represented as:

[0072]

[0073] The output doping parameter vector can be represented as:

[0074]

[0075] in, Indicates the recommended doping concentration. This indicates the recommended dopant element type encoding. Indicates the recommended annealing temperature. The recommended annealing time and doping parameter recommendation model can be expressed as:

[0076]

[0077] Preferably, when using a candidate parameter set screening method, the input feature vector can also be used as a reference. Multiple candidate doping parameter combinations are evaluated, and the parameter combination with the optimal objective function value is selected from the candidate doping parameter set as the recommended result, i.e.:

[0078]

[0079] in, Represents the set of candidate doping parameters. Indicates parameter combination In input features A comprehensive evaluation function under certain conditions. Therefore, based on different defect types and repair coefficients, joint recommendations can be made regarding the type of doping element, doping concentration, and heat treatment conditions to improve the targeted adjustment of the electrical properties of gallium oxide materials.

[0080] The above processing method first performs stable cluster purification based on regional correlation strength to form defect-sensitive region clusters that can characterize the continuous distribution relationship of the same anomalous structure. Then, representative regions are selected from each defect-sensitive region cluster based on keypoint scoring. Pre-shunting is then performed on defect-sensitive region clusters dominated by unrepairable types based on the unrepairable dominance index. Finally, repair coefficients are calculated only for regions not pre-shunted, and recommended doping parameters are output. This reduces the number of dark-field and high-resolution image acquisitions and minimizes the misleading influence of unrepairable dominant regions on subsequent doping adjustments, thereby improving the efficiency and targeting of gallium oxide micro-defect electrical property adjustment.

[0081] like Figure 1 As shown, the present invention also provides a system for adjusting the electrical properties of gallium oxide micro-defects, including a bright-field image acquisition module, a region cluster construction module, a representative region determination module, a second image acquisition module, a defect identification module, a pre-shunting module, and a data processing module.

[0082] The bright-field image acquisition module is used to acquire overall bright-field TEM images of gallium oxide materials;

[0083] The region cluster construction module is used to identify defect-sensitive regions, cluster these regions to form defect-sensitive region clusters, and obtain out-of-cluster anomalies.

[0084] The representative area determination module is used to determine representative areas for each cluster of defect-sensitive regions;

[0085] The second image acquisition module is used to acquire dark field images and high-resolution images of the representative area;

[0086] The defect identification module is used to identify the types of defects in the representative area and to identify de-clustering anomalies;

[0087] The pre-shunting module is used to pre-shunt the defect-sensitive region cluster to which the representative area belongs as an unrepairable restricted cluster when the representative area is identified as being dominated by an unrepairable type.

[0088] The data processing module is used to calculate the repair coefficient for the region that was not pre-splittered and output the recommended doping parameters.

[0089] Preferably, the regional cluster construction module includes a stable cluster purification unit, used to perform unstable connection removal and anomaly point retention processing on bright-field anomalous regions before forming defect-sensitive regional clusters; the representative region determination module includes a keypoint sorting unit, used to determine representative regions based on connectivity, betweenness centrality, convergence, and the degree of cluster splitting influence; the data processing module includes a doping parameter recommendation unit, used to output adjustment parameters for the target gallium oxide material based on defect type and repair coefficient. Thus, the system can realize the steps and functions in the above method implementation.

[0090] The above are merely specific embodiments of the present invention and are not intended to limit the invention. For those skilled in the art, any modifications, equivalent substitutions, or improvements made to the above embodiments without departing from the inventive concept should fall within the protection scope of the present invention. The protection scope of the present invention is defined by the claims, and the specification and drawings can be used to interpret the claims.

Claims

1. A method for adjusting the electrical properties of gallium oxide microdefects, characterized in that, include: Acquire a full-field bright-field image of gallium oxide material using transmission electron microscopy (TEM). Based on the overall bright-field image, multiple defect-sensitive regions are determined, and stable cluster purification and regional clustering are performed according to the regional correlation between the defect-sensitive regions to form at least one defect-sensitive region cluster and obtain de-clustering anomalies. For each cluster of defect-sensitive regions, a representative region is determined. Dark field images and high-resolution images of the representative regions are acquired, and the dark field images and high-resolution images are input into a defect type recognition model to obtain the defect type of the representative region. Anomaly identification is performed independently for each of the aforementioned de-clustering anomalies; Based on the identification result that the representative area belongs to the unrepairable type, the unrepairable dominant index is determined, and when the unrepairable dominant index meets the preset conditions, the corresponding defect-sensitive area cluster is pre-diverted into an unrepairable restricted cluster. For regions that are not pre-shunted, a repair coefficient is calculated based on the repairable type, and the defect type and the repair coefficient are input into the doping parameter recommendation model to obtain the doping parameter recommendation result.

2. The method for adjusting the electrical properties of gallium oxide microdefects according to claim 1, characterized in that, The representative area is the joint area corresponding to the connecting position, confluence position or through position within the defect-sensitive region cluster.

3. The method for adjusting the electrical properties of gallium oxide microdefects according to claim 1, characterized in that, The regional association is determined based on the boundary distance, grayscale difference, boundary direction difference, and connection zone stability between defect-sensitive regions.

4. The method for adjusting the electrical properties of gallium oxide microdefects according to claim 1, characterized in that, The region clustering includes: constructing a region graph with each of the defect-sensitive regions as nodes and the region connection relationships that satisfy the stable connection conditions as edges, and determining the connected subgraphs in the region graph as the corresponding defect-sensitive region clusters.

5. The method for adjusting the electrical properties of gallium oxide microdefects according to claim 1, characterized in that, The stable cluster purification includes: weakening or severing unstable associations formed by weak connections, pseudo-bridging, or local noise, and retaining the abnormal regions that fail to stably merge into any of the defect-sensitive region clusters as the declustering anomalies.

6. The method for adjusting the electrical properties of gallium oxide microdefects according to claim 1, characterized in that, The declustering anomalies include boundary anomalies, pseudo-bridging points, and independent anomalies.

7. The method for adjusting the electrical properties of gallium oxide microdefects according to claim 1, characterized in that, The unrepairable types include grain boundary defects and impurity cluster defects, and the unrepairable dominant index is determined based on the identification results of the representative region belonging to grain boundary defects and impurity cluster defects.

8. The method for adjusting the electrical properties of gallium oxide microdefects according to claim 1, characterized in that, The repairable types include dislocation defects, stacking fault defects, and inversion defects, and the repair coefficient is determined based on the dislocation defect density, stacking fault defect density, and inversion defect concentration.

9. The method for adjusting the electrical properties of gallium oxide microdefects according to claim 1, characterized in that, The recommended doping parameters include at least one of the following: doping element type, doping concentration, annealing temperature, and annealing time.

10. The system for adjusting the electrical properties of gallium oxide microdefects according to any one of claims 1 to 9, characterized in that, include: Bright-field image acquisition module, used to acquire overall bright-field images of gallium oxide materials using transmission electron microscopy (TEM); The region cluster construction module is used to determine multiple defect-sensitive regions based on the overall bright-field image, and to perform stable cluster purification and region clustering according to the regional correlation between the defect-sensitive regions to form at least one defect-sensitive region cluster and obtain de-clustering anomalies. The representative area determination module is used to determine representative areas for each of the aforementioned defect-sensitive region clusters; The second image acquisition module is used to acquire dark field images and high-resolution images of the representative area; The defect identification module is used to input the dark field image and the high-resolution image into the defect type identification model to obtain the defect type of the representative area, and to independently perform anomaly identification on the de-clustering anomaly points; The pre-shunting module is used to determine the irreparable dominant index based on the identification result that the representative area belongs to the irreparable type, and when the irreparable dominant index meets the preset conditions, pre-shunt the corresponding defect-sensitive area cluster into an irreparable restricted cluster. The data processing module is used to calculate the repair coefficient based on the repairable type for areas that have not been pre-shunted, and input the defect type and the repair coefficient into the doping parameter recommendation model to obtain the doping parameter recommendation result.

Images