CN-122023255-A - Semiconductor device defect detection method and device and electronic equipment

Abstract

The invention provides a method and a device for detecting defects of a semiconductor device and electronic equipment, and belongs to the technical field of defect detection of semiconductor devices. The method comprises the steps of obtaining an image to be detected of a semiconductor device, inputting the image to be detected into a defect detection model to obtain a defect detection result output by the defect detection model, wherein the defect detection model is obtained by training a sample defect data set constructed based on a plurality of generated defect images of the sample semiconductor device and defect labels corresponding to the generated defect images, the sample defect data set is obtained by performing data enhancement on a condition generation countermeasure network based on noise, a plurality of condition vectors and a real defect image of the sample semiconductor device, and the condition vectors comprise defect condition parameters representing defect types of the sample semiconductor device. The invention is used for solving the problem of lower accuracy and reliability in defect detection by the existing method.

Inventors

• CHEN YANNING
• LIU FANG
• WU BO
• ZHAO DONGYAN
• CHEN YINING
• WANG YUPING
• DENG YONGFENG
• LUO ZONGLAN

Assignees

• 北京智芯微电子科技有限公司
• 浙江大学

Dates

Publication Date

20260512

Application Date

20251226

Claims (15)

1. 1. A method for detecting defects in a semiconductor device, comprising: Acquiring an image to be detected of the semiconductor device; inputting the image to be detected into a defect detection model to obtain a defect detection result output by the defect detection model; The defect detection model is obtained by training a sample defect data set constructed based on a plurality of generated defect images of a sample semiconductor device and defect labels corresponding to the generated defect images, wherein the sample defect data set is obtained by performing data enhancement on a countermeasure network based on noise, a plurality of condition vectors and a real defect image of the sample semiconductor device by utilizing condition generation, and the condition vectors comprise defect condition parameters representing defect types of the sample semiconductor device.
2. 2. The method for detecting a defect of a semiconductor device according to claim 1, wherein, the sample defect dataset is trained by the following steps: Acquiring a real background image of the noise and the sample semiconductor device; Performing first data enhancement based on the noise and the real background image by using an initial generation countermeasure network to obtain a sample background data set, wherein the sample background data set comprises a plurality of sample background images of a sample semiconductor device; And performing second data enhancement on the basis of the plurality of sample background images, the plurality of condition vectors and the real defect image of the sample semiconductor device by using the condition generation countermeasure network to obtain the sample defect data set.
3. 3. The method for detecting defects of a semiconductor device according to claim 2, wherein the initial generation countermeasure network includes a first generator and a first discriminator, wherein the performing the first data enhancement based on the noise and the true background image by using the initial generation countermeasure network to obtain a sample background data set includes: repeatedly inputting the noise to the first generator until the set times are reached, and obtaining the plurality of sample background images output by the first generator; The first generator is trained based on the noise, the real background image and the first discriminator, and the first discriminator is used for distinguishing the real background image and the false background image of the sample semiconductor device.
4. 4. The method according to claim 2, wherein the condition generating countermeasure network includes a second generator and a second discriminator, wherein the performing second data enhancement using the condition generating countermeasure network based on the plurality of sample background images, the plurality of condition vectors, and the actual defect image of the sample semiconductor device to obtain the sample defect data set includes: the following steps are repeatedly executed until the generated defect images corresponding to all the condition vectors are obtained: inputting the sample background image and the first condition vector into the second generator to obtain a generated defect image corresponding to the first condition vector output by the second generator; The first condition vector is any one of the plurality of condition vectors, the second generator is trained based on the sample background image, the real defect image, the plurality of condition vectors and the second discriminator, and the second discriminator is used for distinguishing the real defect image and the false defect image corresponding to the condition vector.
5. 5. The method according to claim 1, wherein after inputting the image to be inspected to a defect inspection model to obtain a defect inspection result output by the defect inspection model, further comprising: inputting the first defect detection result into a reinforcement learning model under the condition that the occurrence frequency of the first defect detection result exceeds a set threshold value, and obtaining a process parameter adjustment strategy output by the reinforcement learning model; The reinforcement learning model is trained based on a state space, an action space and a reward function corresponding to the sample defect detection result.
6. 6. The method according to any one of claims 1 to 5, wherein the condition vector includes a defect morphology, a defect position, or the condition vector includes a defect morphology, a defect position, and a defect size parameter.
7. 7. A semiconductor device defect detecting apparatus, comprising: the acquisition module is used for acquiring an image to be detected of the semiconductor device; The detection module is used for inputting the image to be detected into a defect detection model to obtain a defect detection result output by the defect detection model; The defect detection model is obtained by training a sample defect data set constructed based on a plurality of generated defect images of a sample semiconductor device and defect labels corresponding to the generated defect images, wherein the sample defect data set is obtained by performing data enhancement on a countermeasure network based on noise, a plurality of condition vectors and a real defect image of the sample semiconductor device by utilizing condition generation, and the condition vectors comprise defect condition parameters representing defect types of the sample semiconductor device.
8. 8. The semiconductor device defect inspection apparatus of claim 7, wherein the sample defect dataset is trained by: Acquiring a real background image of the noise and the sample semiconductor device; Performing first data enhancement based on the noise and the real background image by using an initial generation countermeasure network to obtain a sample background data set, wherein the sample background data set comprises a plurality of sample background images of a sample semiconductor device; And performing second data enhancement on the basis of the plurality of sample background images, the plurality of condition vectors and the real defect image of the sample semiconductor device by using the condition generation countermeasure network to obtain the sample defect data set.
9. 9. The apparatus according to claim 8, wherein the initial generation countermeasure network includes a first generator and a first discriminator, wherein the performing the first data enhancement based on the noise and the true background image with the initial generation countermeasure network to obtain a sample background data set includes: repeatedly inputting the noise to the first generator until the set times are reached, and obtaining the plurality of sample background images output by the first generator; The first generator is trained based on the noise, the real background image and the first discriminator, and the first discriminator is used for distinguishing the real background image and the false background image of the sample semiconductor device.
10. 10. The semiconductor device defect detection apparatus of claim 8, wherein the condition generation countermeasure network includes a second generator and a second arbiter, wherein the performing a second data enhancement using the condition generation countermeasure network based on the plurality of sample background images, a plurality of condition vectors, and a true defect image of the sample semiconductor device, to obtain the sample defect dataset, comprises: the following steps are repeatedly executed until the generated defect images corresponding to all the condition vectors are obtained: inputting the sample background image and the first condition vector into the second generator to obtain a generated defect image corresponding to the first condition vector output by the second generator; The first condition vector is any one of the plurality of condition vectors, the second generator is trained based on the sample background image, the real defect image, the plurality of condition vectors and the second discriminator, and the second discriminator is used for distinguishing the real defect image and the false defect image corresponding to the condition vector.
11. 11. The apparatus according to claim 7, wherein after inputting the image to be inspected to a defect inspection model to obtain a defect inspection result outputted by the defect inspection model, further comprising: inputting the first defect detection result into a reinforcement learning model under the condition that the occurrence frequency of the first defect detection result exceeds a set threshold value, and obtaining a process parameter adjustment strategy output by the reinforcement learning model; The reinforcement learning model is trained based on a state space, an action space and a reward function corresponding to the sample defect detection result.
12. 12. The apparatus according to any one of claims 7 to 11, wherein the condition vector includes a defect morphology, a defect position, or the condition vector includes a defect morphology, a defect position, and a defect size parameter.
13. 13. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the semiconductor device defect detection method of any one of claims 1 to 6 when the program is executed by the processor.
14. 14. A machine readable storage medium having stored thereon a computer program, which when executed by a processor implements the semiconductor device defect detection method of any of claims 1 to 6.
15. 15. A computer program product comprising a computer program, characterized in that the computer program, when executed by a processor, implements the semiconductor device defect detection method of any of claims 1 to 6.

Description

Semiconductor device defect detection method and device and electronic equipment Technical Field The present invention relates to the field of semiconductor device defect detection technology, and in particular, to a semiconductor device defect detection method, a semiconductor device defect detection apparatus, an electronic device, a machine-readable storage medium, and a computer program product. Background With the development of semiconductor technology, electronic devices made of semiconductor devices play an indispensable role in daily life. In the prior art, defect detection during the fabrication of semiconductor devices (e.g., LDMOS, LATERALLY DIFFUSED METAL OXIDE SEMICONDUCTOR) has relied primarily on conventional optical detection and human inspection methods. These methods are inefficient and have poor accuracy, and cannot meet the strict requirements of modern high density integrated circuits for defect detection accuracy and speed. Therefore, defect detection using a deep learning model appears in the existing method. Although the deep learning model is excellent in defect detection, the efficiency and accuracy of defect detection can be improved to some extent. But its performance is highly dependent on large-scale, high-quality training data. However, in semiconductor manufacturing, defect data tends to be scarce and unbalanced, and the amount of data for part of the defect types is extremely small, resulting in difficulty in efficient learning and recognition of the defects by a model. This data imbalance limits the generalization ability of the deep learning model, affecting the accuracy and reliability of defect detection. Disclosure of Invention The embodiment of the invention aims to provide a method, a device and electronic equipment for detecting defects of a semiconductor device, which are used for solving the problem that the accuracy and the reliability are lower when the existing method is used for detecting the defects. In order to achieve the above object, an embodiment of the present invention provides a method for detecting a defect of a semiconductor device, including: Acquiring an image to be detected of the semiconductor device; inputting the image to be detected into a defect detection model to obtain a defect detection result output by the defect detection model; The defect detection model is obtained by training a sample defect data set constructed based on a plurality of generated defect images of a sample semiconductor device and defect labels corresponding to the generated defect images, wherein the sample defect data set is obtained by performing data enhancement on a countermeasure network based on noise, a plurality of condition vectors and a real defect image of the sample semiconductor device by utilizing condition generation, and the condition vectors comprise defect condition parameters representing defect types of the sample semiconductor device. Optionally, the sample defect dataset is trained by: Acquiring a real background image of the noise and the sample semiconductor device; Performing first data enhancement based on the noise and the real background image by using an initial generation countermeasure network to obtain a sample background data set, wherein the sample background data set comprises a plurality of sample background images of a sample semiconductor device; And performing second data enhancement on the basis of the plurality of sample background images, the plurality of condition vectors and the real defect image of the sample semiconductor device by using the condition generation countermeasure network to obtain the sample defect data set. Optionally, the initial generation countermeasure network includes a first generator and a first arbiter, and the performing the first data enhancement based on the noise and the real background image by using the initial generation countermeasure network to obtain a sample background data set includes: repeatedly inputting the noise to the first generator until the set times are reached, and obtaining the plurality of sample background images output by the first generator; The first generator is trained based on the noise, the real background image and the first discriminator, and the first discriminator is used for distinguishing the real background image and the false background image of the sample semiconductor device. Optionally, the condition generating countermeasure network includes a second generator and a second arbiter, and the generating countermeasure network using the condition performs a second data enhancement based on the plurality of sample background images, the plurality of condition vectors, and the real defect image of the sample semiconductor device to obtain the sample defect data set, including: the following steps are repeatedly executed until the generated defect images corresponding to all the condition vectors are obtained: inputting the sample background image and the first c