Wafer defect detection method and device

Abstract

The invention provides a wafer defect detection method and device, and belongs to the technical field of semiconductor manufacturing. The wafer defect detection method comprises the steps that defect boundary extraction is carried out on a wafer to be judged, and morphology information of defects of the wafer is determined, and the morphology information comprises the number of breakpoints, defect distribution density and the length and angle of each section of defect; generating a defect density matrix comprising a plurality of feature vectors according to the morphology information, wherein the defect density matrix is used for representing the density change and breakpoint distribution condition of defects; evaluating the defect density matrix by using a defect judgment model to obtain a risk score of the wafer; and comparing the risk score of the wafer with a preset risk threshold value, and judging the wafer of which the risk score is greater than the risk threshold value as the wafer with discontinuous Special pattern defects. According to the method and the device, the wafer with discontinuous Special pattern defects can be identified.

Description

Technical Field

[0001] The present invention relates to the field of semiconductor manufacturing technology, and in particular to a wafer defect detection method and device. Background Art

[0002] With the continuous improvement of semiconductor wafer manufacturing technology, higher precision requirements are placed on the detection and interception of defect morphology on the wafer surface. In the semiconductor wafer manufacturing process, complex morphological defects may appear on the wafer surface, such as discontinuous special patterns. These defects often present intermittent distribution forms, and due to the discontinuity, complex length and density characteristics of such special patterns, it is difficult to quantify and define and train models in actual production.

[0003] The currently widely used ADC (Automatic Defect Classification) model mainly relies on the existing defect morphology in the sample library for training and judgment. However, since discontinuous special pattern defect samples are difficult to collect in actual production, and their distribution is complex and difficult to quantify, the traditional ADC model cannot accurately identify and effectively intercept such defects when judging defects, resulting in missed detection or misjudgment. Summary of the invention

[0004] The present invention provides a wafer defect detection method and device, which can identify wafers with discontinuous special pattern defects.

[0005] In order to achieve the above object, the technical solution adopted by the embodiment of the present invention is:

[0006] A wafer defect detection method, comprising:

[0007] Extracting defect boundaries of the wafer to be judged, and determining morphological information of the defects of the wafer, wherein the morphological information includes the number of breakpoints, defect distribution density, and the length and angle of each defect segment;

[0008] Generating a defect density matrix including a plurality of characteristic vectors according to the morphology information, wherein the defect density matrix is ​​used to characterize density changes and breakpoint distribution of defects;

[0009] Using a defect determination model to evaluate the defect density matrix to obtain a risk score of the wafer;

[0010] The risk score of the wafer is compared with a preset risk threshold, and a wafer having a risk score greater than the risk threshold is determined to be a wafer having a discontinuous special pattern defect.

[0011] In some embodiments, before extracting defect boundaries of the wafer to be judged, the method further includes:

[0012] The ADC model is used to screen the wafers to be inspected, and the wafers to be judged are obtained after the abnormal wafers are intercepted.

[0013] In some embodiments, there are multiple defect judgment models, and evaluating the defect density matrix using the defect judgment model to obtain the risk score of the wafer includes:

[0014] Inputting the defect density matrix into a plurality of defect determination models respectively to obtain a plurality of detection results respectively;

[0015] The multiple detection results are fused according to the weights of the multiple defect judgment models to obtain the risk score of the wafer.

[0016] In some embodiments, the method further comprises:

[0017] Matching the production process parameters of the wafer to be determined with the production process parameters of the first wafer in the historical database;

[0018] The risk threshold is adjusted according to the matching degree between the production process parameters of the wafer and the production process parameters of the first wafer, where the first wafer is determined to have a discontinuous special pattern defect.

[0019] In some embodiments, adjusting the risk threshold according to the matching degree between the production process parameters of the wafer and the production process parameters of the first wafer includes:

[0020] If the matching degree between the production process parameters of the wafer and the production process parameters of the first wafer is greater than or equal to a preset first matching degree, lowering the risk threshold;

[0021] If the matching degree between the production process parameters of the wafer and the production process parameters of the first wafer is less than or equal to a preset second matching degree, increasing the risk threshold;

[0022] The second matching degree is smaller than the first matching degree.

[0023] In some embodiments, after determining that a wafer having a risk score greater than the risk threshold is a wafer having a discontinuous special pattern defect, the method further includes:

[0024] The wafer determined to have a discontinuous special pattern defect is used as sample data of the defect determination model, and data amplification processing is performed on the sample data to generate a plurality of discontinuous special pattern defect samples with different distribution characteristics, and the sample database of the defect determination model is updated;

[0025] The defect determination model is trained using the updated sample database.

[0026] In some embodiments, extracting defect boundaries of a wafer to be determined and determining morphological information of defects of the wafer includes:

[0027] A variety of filters are used to extract the discontinuous special pattern defect boundary of the wafer, and the curve morphology of the defect is obtained by the least square method and the feature point fitting algorithm to determine the length starting point and end point of the defect, as well as the contour and breakpoint distribution of the discontinuous special pattern defect.

[0028] An embodiment of the present invention further provides a wafer defect detection device, comprising:

[0029] An extraction module is used to extract defect boundaries of the wafer to be judged and determine the morphological information of the defects of the wafer, wherein the morphological information includes the number of breakpoints, defect distribution density, length and angle of each defect segment;

[0030] A generation module, used to generate a defect density matrix including a plurality of characteristic vectors according to the morphology information, wherein the defect density matrix is ​​used to characterize the density change and breakpoint distribution of the defects;

[0031] An evaluation module, used to evaluate the defect density matrix using a defect determination model to obtain a risk score of the wafer;

[0032] The determination module is used to compare the risk score of the wafer with a preset risk threshold, and determine the wafer with a risk score greater than the risk threshold as a wafer with a discontinuous special pattern defect.

[0033] An embodiment of the present invention further provides an electronic device, comprising: a processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps of the wafer defect detection method as described above.

[0034] An embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the wafer defect detection method as described above are implemented.

[0035] An embodiment of the present invention further provides a computer program product, including computer instructions, which, when executed by a processor, implement the steps of the wafer defect detection method as described above.

[0036] The beneficial effects of the present invention are:

[0037] In this embodiment, by introducing a defect density matrix and multiple feature vectors, a more detailed description of the special defect morphology can be made, so that the defect judgment model can accurately distinguish the discontinuous special pattern defect types with different density and length characteristics, and solve the problem of ADC model omission and over-judgment caused by insufficient discontinuous special pattern defect samples in the existing wafer surface defect interception process, ensuring the effective identification and interception of discontinuous special pattern defects on the wafer surface, while improving the accuracy and efficiency of detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 A schematic diagram showing a flow chart of a wafer defect detection method according to an embodiment of the present invention;

[0039] Figure 2 A schematic diagram showing the generation of discontinuous special defect samples with different distribution characteristics according to an embodiment of the present invention;

[0040] Figure 3 A schematic diagram showing the structure of a wafer defect detection device according to an embodiment of the present invention;

[0041] Figure 4 A schematic diagram showing the structure of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solution and advantages of the embodiment of the present invention clearer, the technical solution of the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings of the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all of the embodiments. Based on the described embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field belong to the scope of protection of the present invention.

[0043] When faced with discontinuous special pattern defects, the traditional ADC model has the following technical bottlenecks:

[0044] 1. Insufficient samples: Discontinuous special pattern samples are difficult to collect in actual production. The small number of samples leads to insufficient generalization ability of model training and poor judgment effect.

[0045] 2. Difficulty in quantifying morphology: Due to the discontinuous distribution of special pattern defects, traditional morphological parameters (such as length and width) cannot effectively describe their overall characteristics, which makes the model prone to incorrect judgment when facing changes in defect distribution morphology.

[0046] 3. Lack of dynamic interception strategy: Existing ADC models usually use fixed thresholds for defect judgment, which cannot adapt to the changes in defect morphology caused by process fluctuations during wafer production, resulting in the inability to effectively intercept potential risk wafers in actual production.

[0047] 4. Errors caused by human intervention: Traditional ADC models rely on human intervention for sample screening and feature labeling during sample labeling and model training, which can easily introduce human errors and reduce the overall accuracy and stability of the judgment.

[0048] Therefore, there is an urgent need for a wafer defect detection solution to improve the sample diversity and model judgment capabilities of discontinuous special pattern defects, to achieve accurate quantification and real-time interception of complex defect morphologies on the wafer surface, and thus ensure the overall quality stability of wafer shipments.

[0049] The present invention provides a wafer defect detection method and device, which can identify wafers with discontinuous special pattern defects.

[0050] The embodiment of the present invention provides a wafer defect detection method, such as Figure 1 As shown, including:

[0051] Step 101: extracting defect boundaries of a wafer to be determined, and determining morphological information of defects of the wafer, wherein the morphological information includes the number of breakpoints, defect distribution density, and the length and angle of each defect segment;

[0052] In this embodiment, multiple filters are used to extract the discontinuous special pattern defect boundaries of the wafer, and the curve morphology of the defect is obtained by the least squares method and feature point fitting algorithm to determine the length starting point and end point of the defect, as well as the contour and breakpoint distribution of the discontinuous special pattern defect.

[0053] In this embodiment, methods such as image segmentation, edge detection and feature quantization can be used to process the defective area suspected to be a discontinuous special pattern, and extract the microscopic morphological features of the discontinuous special defect.

[0054] In some embodiments, before extracting defect boundaries of the wafer to be judged, the method further includes:

[0055] The ADC model is used to screen the wafers to be inspected, and the wafers to be judged are obtained after the abnormal wafers are intercepted.

[0056] This embodiment can use the traditional ADC model to perform preliminary screening on all wafers, quickly detect wafer surface defects with high significance, and directly intercept wafers with obvious abnormalities, reducing the pressure of subsequent sample labeling and model training.

[0057] Step 102: generating a defect density matrix including a plurality of characteristic vectors according to the morphology information, wherein the defect density matrix is ​​used to characterize density changes and breakpoint distribution of defects;

[0058] In this embodiment, a multi-dimensional defect density matrix can be established based on the morphology information to quantitatively describe the morphology of special defects. The defect density matrix can describe the sparseness or density of discontinuous special defects. The defect density matrix includes multiple feature vectors, which can quantitatively describe the morphological characteristics of discontinuous special pattern defects. The defect density matrix can describe the density changes and breakpoint distribution of discontinuous special pattern defects by performing sparsity analysis on the pixel distribution of each segment. The defect density matrix can reflect the overall morphology of discontinuous special defects, and evaluate the discontinuous special defects segment by segment, providing a basis for subsequent dynamic interception strategies.

[0059] Step 103: Evaluate the defect density matrix using a defect determination model to obtain a risk score of the wafer;

[0060] In some embodiments, in order to improve the accuracy of detection, the number of the defect judgment models may be multiple, and the use of the defect judgment model to evaluate the defect density matrix to obtain the risk score of the wafer includes:

[0061] Inputting the defect density matrix into a plurality of defect determination models respectively to obtain a plurality of detection results respectively;

[0062] The multiple detection results are fused according to the weights of the multiple defect judgment models to obtain the risk score of the wafer.

[0063] Multiple defect judgment models can be trained independently, and the weight corresponding to each defect judgment model can be set according to the actual situation, and the detection results corresponding to multiple defect judgment models can be weighted and fused to obtain the final risk score. Among them, multiple defect judgment models can be random forest models, deep learning neural network models, etc.

[0064] Step 104: Compare the risk score of the wafer with a preset risk threshold, and determine the wafer with a risk score greater than the risk threshold as a wafer with a discontinuous special pattern defect.

[0065] Specifically, a wafer with a risk score greater than the risk threshold is determined as a wafer with a discontinuous Specialpattern defect; and a wafer with a risk score not greater than the risk threshold is determined as a wafer without a discontinuous Specialpattern defect.

[0066] In this embodiment, after a wafer is determined to be a wafer with a discontinuous special pattern defect, an alarm may be automatically triggered, and the wafer may be intercepted to prevent unqualified products from entering subsequent process links.

[0067] In this embodiment, by introducing a defect density matrix and multiple feature vectors, a more detailed description of the special defect morphology can be made, so that the defect judgment model can accurately distinguish the discontinuous special pattern defect types with different density and length characteristics, and solve the problem of ADC model omission and over-judgment caused by insufficient discontinuous special pattern defect samples in the existing wafer surface defect interception process, ensuring the effective identification and interception of discontinuous special pattern defects on the wafer surface, while improving the accuracy and efficiency of detection.

[0068] In actual production, since the surface morphology of the wafer will be affected by process fluctuations, a fixed risk threshold cannot effectively adapt to morphological changes. In this embodiment, the risk threshold can be dynamically updated to ensure that a high interception accuracy rate can be maintained under different process conditions. The method also includes:

[0069] Matching the production process parameters of the wafer to be determined with the production process parameters of the first wafer in the historical database;

[0070] The risk threshold is adjusted according to the matching degree between the production process parameters of the wafer and the production process parameters of the first wafer, where the first wafer is determined to have a discontinuous special pattern defect.

[0071] Specifically, adjusting the risk threshold according to the matching degree between the production process parameters of the wafer and the production process parameters of the first wafer includes:

[0072] If the matching degree between the production process parameters of the wafer and the production process parameters of the first wafer is greater than or equal to a preset first matching degree, lowering the risk threshold;

[0073] If the matching degree between the production process parameters of the wafer and the production process parameters of the first wafer is less than or equal to a preset second matching degree, increasing the risk threshold;

[0074] The second matching degree is smaller than the first matching degree.

[0075] In order to solve the problem of small number of discontinuous special pattern defect samples, in this embodiment, after determining that a wafer having a risk score greater than the risk threshold is a wafer having a discontinuous special pattern defect, the method further includes:

[0076] The wafer determined to have a discontinuous special pattern defect is used as sample data of the defect determination model, and data amplification processing is performed on the sample data to generate a plurality of discontinuous special pattern defect samples with different distribution characteristics, and the sample database of the defect determination model is updated;

[0077] The defect determination model is trained using the updated sample database.

[0078] This embodiment performs data amplification processing on limited discontinuous special pattern defect data, and uses image data amplification methods such as geometric transformation (such as rotation, scaling, translation) and random cropping to generate discontinuous special pattern defect samples with various distribution characteristics, such as Figure 2 As shown, the appearance characteristics of defect morphology can be changed, the diversity of sample sets can be improved, and the problem of insufficient sample collection in the actual production process can be compensated.

[0079] This embodiment can also dynamically update the sample library and model parameters based on the newly collected defect samples in production. By introducing an automated data cleaning and preprocessing module, new samples are automatically classified and labeled, and merged with the original sample set, and the ADC model is regularly updated and retrained online. This can significantly improve the real-time and adaptability of the system, effectively reduce the errors caused by manual intervention, further improve the accurate recognition rate of discontinuous special pattern defects, and enhance the interception capability of discontinuous special pattern defects.

[0080] This embodiment reduces the risk of over-judgment and under-judgment of traditional models when the sample size is insufficient, and improves the overall interception capability of risky wafers; it uses image data amplification methods to generate a variety of defect samples, and introduces a variety of machine learning algorithms to train the amplified samples, which can complete the optimization and update of the model in a relatively short time, thereby greatly shortening the model training cycle, effectively avoiding the inconsistency caused by traditional manual labeling and training models, and improving the stability of defect detection results. In addition, since this embodiment uses automated data processing and algorithm fusion technology in the sample amplification, model training and judgment process, it greatly reduces the reliance on manual labeling and sample selection, avoids the introduction of human errors, and improves the consistency and stability of model judgment.

[0081] The embodiment of the present invention also provides a wafer defect detection device, such as Figure 3 As shown, including:

[0082] The extraction module 21 is used to extract the defect boundary of the wafer to be judged, and determine the morphological information of the defect of the wafer, wherein the morphological information includes the number of breakpoints, the defect distribution density, and the length and angle of each defect segment;

[0083] A generating module 22, used to generate a defect density matrix including a plurality of characteristic vectors according to the morphology information, wherein the defect density matrix is ​​used to characterize the density variation and breakpoint distribution of defects;

[0084] An evaluation module 23, used to evaluate the defect density matrix using a defect determination model to obtain a risk score of the wafer;

[0085] The determination module 24 is used to compare the risk score of the wafer with a preset risk threshold, and determine the wafer with a risk score greater than the risk threshold as a wafer with a discontinuous special pattern defect.

[0086] In this embodiment, by introducing a defect density matrix and multiple feature vectors, a more detailed description of the special defect morphology can be made, so that the defect judgment model can accurately distinguish the discontinuous special pattern defect types with different density and length characteristics, and solve the problem of ADC model omission and over-judgment caused by insufficient discontinuous special pattern defect samples in the existing wafer surface defect interception process, ensuring the effective identification and interception of discontinuous special pattern defects on the wafer surface, while improving the accuracy and efficiency of detection.

[0087] In some embodiments, the device further comprises:

[0088] The screening module is used to screen the wafers to be detected by using the ADC model, and obtain the wafers to be determined after intercepting abnormal wafers.

[0089] In some embodiments, there are multiple defect judgment models, and the evaluation module 23 is specifically used to input the defect density matrix into multiple defect judgment models respectively to obtain multiple detection results respectively; the multiple detection results are merged according to the weights of the multiple defect judgment models to obtain the risk score of the wafer.

[0090] In some embodiments, the device further comprises:

[0091] An adjustment module is used to match the production process parameters of the wafer to be determined with the production process parameters of the first wafer in the historical database; adjust the risk threshold according to the matching degree between the production process parameters of the wafer and the production process parameters of the first wafer, and the first wafer is a wafer determined to have a discontinuous special pattern defect.

[0092] In some embodiments, the adjustment module is specifically used to reduce the risk threshold if the matching degree between the production process parameters of the wafer and the production process parameters of the first wafer is greater than or equal to a preset first matching degree; if the matching degree between the production process parameters of the wafer and the production process parameters of the first wafer is less than or equal to a preset second matching degree, increase the risk threshold;

[0093] The second matching degree is smaller than the first matching degree.

[0094] In some embodiments, the device further comprises:

[0095] An updating module is used to use the wafer determined to have a discontinuous special pattern defect as sample data of the defect judgment model, and perform data amplification processing on the sample data to generate a variety of discontinuous special pattern defect samples with different distribution characteristics, and update the sample database of the defect judgment model; and use the updated sample database to train the defect judgment model.

[0096] In some embodiments, the extraction module 21 is specifically used to extract the discontinuous special pattern defect boundary of the wafer using multiple filters, and obtain the curve morphology of the defect through the least squares method and feature point fitting algorithm, determine the length starting point and end point of the defect, and the contour and breakpoint distribution of the discontinuous special pattern defect.

[0097] Please refer to Figure 4An embodiment of the present invention further provides an electronic device 30, including a processor 31, a memory 32, and a computer program stored in the memory 32 and executable on the processor 31. When the computer program is executed by the processor 31, each process of the above-mentioned wafer defect detection method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it will not be described here.

[0098] The embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, each process of the above-mentioned wafer defect detection method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it is not repeated here. Among them, the computer-readable storage medium includes permanent and non-permanent, removable and non-removable media, and information storage can be realized by any method or technology. The information can be a computer-readable instruction, a data structure, a module of a program, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic tape magnetic disk storage or other magnetic storage terminal device to be detected or any other non-transmission medium, which can be used to store information that can be accessed by the terminal device to be detected. According to the definition in this article, computer-readable storage media does not include temporary computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0099] The present application also provides a computer program product, including computer instructions, which, when executed by a processor, implement the above Figure 1 The various processes of the method embodiment shown can achieve the same technical effect, and will not be described again here to avoid repetition.

[0100] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0101] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for enabling a terminal (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0102] In the various method embodiments of the present disclosure, the serial numbers of the steps cannot be used to limit the sequence of the steps. For ordinary technicians in this field, without paying any creative work, changes to the sequence of the steps are also within the protection scope of the present disclosure.

[0103] It should be noted that each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the embodiments, since they are basically similar to the product embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the product embodiments.

[0104] The above is a preferred embodiment of the present disclosure. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles described in the present disclosure. These improvements and modifications should also be regarded as the scope of protection of the present disclosure.

Claims

1. A wafer defect detection method, characterized in that: include: Extracting defect boundaries of the wafer to be judged, and determining morphological information of the defects of the wafer, wherein the morphological information includes the number of breakpoints, defect distribution density, and the length and angle of each defect segment; Generating a defect density matrix including a plurality of characteristic vectors according to the morphology information, wherein the defect density matrix is ​​used to characterize density changes and breakpoint distribution of defects; Using a defect determination model to evaluate the defect density matrix to obtain a risk score of the wafer; The risk score of the wafer is compared with a preset risk threshold, and a wafer having a risk score greater than the risk threshold is determined to be a wafer having a discontinuous Special pattern defect.

2. The wafer defect detection method according to claim 1, characterized in that: Before extracting defect boundaries of the wafer to be judged, the method further includes: The ADC model is used to screen the wafers to be inspected, and the wafers to be judged are obtained after the abnormal wafers are intercepted.

3. The wafer defect detection method according to claim 1, characterized in that: There are multiple defect judgment models, and the defect density matrix is ​​evaluated by using the defect judgment model to obtain the risk score of the wafer, which includes: Inputting the defect density matrix into a plurality of defect determination models respectively to obtain a plurality of detection results respectively; The multiple detection results are fused according to the weights of the multiple defect judgment models to obtain the risk score of the wafer.

4. The wafer defect detection method according to claim 1, characterized in that: The method further comprises: Matching the production process parameters of the wafer to be determined with the production process parameters of the first wafer in the historical database; The risk threshold is adjusted according to the matching degree between the production process parameters of the wafer and the production process parameters of the first wafer, where the first wafer is a wafer determined to have a discontinuous Special pattern defect.

5. The wafer defect detection method according to claim 4, characterized in that: The adjusting the risk threshold according to the matching degree between the production process parameters of the wafer and the production process parameters of the first wafer comprises: If the matching degree between the production process parameters of the wafer and the production process parameters of the first wafer is greater than or equal to a preset first matching degree, lowering the risk threshold; If the matching degree between the production process parameters of the wafer and the production process parameters of the first wafer is less than or equal to a preset second matching degree, increasing the risk threshold; The second matching degree is smaller than the first matching degree.

6. The wafer defect detection method according to claim 1, characterized in that: After determining that the wafer having a risk score greater than the risk threshold is a wafer having a discontinuous special pattern defect, the method further includes: The wafer determined to have a discontinuous special pattern defect is used as sample data of the defect determination model, and data amplification processing is performed on the sample data to generate a plurality of discontinuous special pattern defect samples with different distribution characteristics, and the sample database of the defect determination model is updated; The defect determination model is trained using the updated sample database.

7. The wafer defect detection method according to claim 1, characterized in that: The step of extracting defect boundaries of the wafer to be determined and determining morphological information of defects of the wafer includes: A variety of filters are used to extract the discontinuous special pattern defect boundary of the wafer, and the curve morphology of the defect is obtained by the least square method and the feature point fitting algorithm to determine the length starting point and end point of the defect, as well as the contour and breakpoint distribution of the discontinuous special pattern defect.

8. A wafer defect detection device, characterized in that: include: An extraction module is used to extract defect boundaries of the wafer to be judged and determine the morphological information of the defects of the wafer, wherein the morphological information includes the number of breakpoints, defect distribution density, length and angle of each defect segment; A generation module, used to generate a defect density matrix including a plurality of characteristic vectors according to the morphology information, wherein the defect density matrix is ​​used to characterize the density change and breakpoint distribution of the defects; An evaluation module, used to evaluate the defect density matrix using a defect determination model to obtain a risk score of the wafer; The determination module is used to compare the risk score of the wafer with a preset risk threshold, and determine the wafer with a risk score greater than the risk threshold as a wafer with a discontinuous special pattern defect.

9. An electronic device, characterized in that: include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein when the program is executed by the processor, the steps of the wafer defect detection method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the wafer defect detection method according to any one of claims 1 to 7 are implemented.

11. A computer program product, characterized in that The method comprises computer instructions, which, when executed by a processor, implement the steps of the wafer defect detection method according to any one of claims 1 to 7.

Images