Electronic apparatus supporting manufacturing of semiconductor device and operating method of the electronic apparatus

In semiconductor device manufacturing, the processor is used to align the edges and corners of the layout image and the captured image, and the machine learning is used to generate the modified layout image, the problem of low alignment accuracy in semiconductor device manufacturing is solved, and manufacturing efficiency is improved. Reduced costs.

CN119008438AInactive Publication Date: 2024-11-22SAMSUNG ELECTRONICS CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410011796.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Priority Date
2023-05-16
Filing Date
2024-01-02
Publication Date
2024-11-22
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

In the manufacturing process of semiconductor devices, due to the increase in process complexity, the occurrence of defects, and the prior art is difficult to quickly and accurately align for generating modified layout images, thereby affecting manufacturing efficiency and cost.

Method used

By receiving and shooting the layout images on the processor, highlighting and aligning edges and corners, using the machine learning module to train to generate modified layout images, improving alignment accuracy and reducing manufacturing time and cost.

Benefits of technology

A more accurate and fast alignment process is achieved, improving machine learning accuracy in semiconductor device manufacturing, and reducing manufacturing time and cost.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119008438A_ABST
    Figure CN119008438A_ABST
Patent Text Reader

Abstract

Disclosed is an operation method of an electronic device including a processor and supporting manufacturing of a semiconductor device. The operating method includes: receiving, at a processor, a layout image for manufacturing a semiconductor device and a captured image generated by capturing an actually manufactured semiconductor device; aligning, at the processor, the layout image and the captured image based on a result of highlighting edges and corners of the layout image and the captured image; and at the processor, performing learning based on the aligned layout image and the aligned captured image such that a first modified layout image is generated from the layout image, and the semiconductor device is manufactured based on a second modified layout image generated from the layout image.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims the priority of Korean Patent Application No. 10-2023-0063174 filed in the Korean Intellectual Property Office on May 16, 2023, the disclosure of which is incorporated herein by reference in its entirety. Technical Field

[0003] Embodiments of the present disclosure described herein relate to an electronic device, and more particularly, to an electronic device that supports manufacturing of a semiconductor device through more accurate learning and an operating method of the electronic device. Background Art

[0004] Semiconductor devices are manufactured through various processes. With the development of semiconductor design technology, the number of processes used to manufacture semiconductor devices continues to increase, and the complexity of each process continues to increase. As the number and complexity of semiconductor processes increase, various defects may occur in the process of manufacturing semiconductor devices.

[0005] In order to prevent the occurrence of various defects, a method of modifying a layout designed for manufacturing a semiconductor device and manufacturing the semiconductor device by using the modified layout may be used. The process of generating the modified layout may be performed empirically based on the characteristics of the material and process used to manufacture the semiconductor device.

[0006] In order to automate empirical layout modification, machine learning can be applied to modify the layout. Machine learning requires a large amount of source information to learn. In addition, in order to perform learning to more accurately generate a modified layout, a large amount of source information needs to be aligned. Summary of the invention

[0007] Embodiments of the present disclosure provide an electronic device capable of reducing manufacturing time and cost of a semiconductor device by accurately and quickly aligning source information for machine learning to support the manufacture of the semiconductor device and an operating method of the electronic device.

[0008] According to an embodiment, an operating method of an electronic device, which includes a processor and is configured to support the manufacture of a semiconductor device, includes: at the processor, receiving a layout image for manufacturing a semiconductor device and a captured image generated by photographing an actually manufactured semiconductor device; at the processor, in response to determining that there are two or more different orientations in the layout image, highlighting the edges and corners of the layout image and the captured image; at the processor, aligning the layout image and the captured image based on the result of highlighting the edges and corners of the layout image and the captured image; and at the processor, performing learning based on the aligned layout image and the aligned captured image, so that a first modified layout image is generated based on the layout image, and the semiconductor device is manufactured based on a second modified layout image, and the second modified layout image is generated based on the layout image based on the learning.

[0009] According to an embodiment, an operating method of an electronic device, the electronic device including a processor and configured to support the manufacture of a semiconductor device, the operating method including: at the processor, receiving a layout image for manufacturing a semiconductor device; and at the processor, generating a modified layout image by modifying the layout image using a modification module based on machine learning. The modification module based on machine learning is trained to: generate a modified layout image according to the layout image based on multiple layout images and multiple captured images of multiple semiconductor devices, and the multiple semiconductor devices are actually manufactured based on the multiple modified layout images generated according to the multiple layout images. The modification module based on machine learning is trained to: highlight and align the edges and corners of the multiple layout images and the multiple captured images, and generate a modified layout image according to the layout image based on the aligned layout image and the aligned captured images.

[0010] According to an embodiment, an electronic device for manufacturing a semiconductor device includes a processor and a memory, the memory including a plurality of layout images for manufacturing a plurality of semiconductor devices, and a plurality of captured images captured after manufacturing the plurality of semiconductor devices. The processor performs learning by aligning the plurality of layout images and the plurality of captured images by highlighting the edges and corners of the plurality of layout images and the plurality of captured images, so that the processor is trained to generate a modified layout image according to the layout image based on the aligned layout images and the aligned captured images. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The above and other objects and features of the present disclosure will become apparent by describing in detail embodiments of the present disclosure with reference to the accompanying drawings.

[0012] Figure 1 A semiconductor manufacturing system in accordance with at least one embodiment of the present disclosure is shown.

[0013] Figure 2A block diagram of an electronic device according to at least one embodiment of the present disclosure is shown.

[0014] Figure 3 An operating method of an electronic device according to at least one embodiment of the present disclosure is shown.

[0015] Figure 4 A learning module in accordance with at least one embodiment of the present disclosure is shown.

[0016] Figure 5 An example is shown in which the determination module determines whether to highlight edges and corners.

[0017] Figure 6 An example of a layout image and a captured image including a pattern having one orientation is shown.

[0018] Figure 7 An example of a layout image and a captured image including patterns having two orientations is shown.

[0019] Figure 8 An example of a method in which the first weighting module performs a first weighting operation is shown.

[0020] Fig. 9 An example of a first layout image and a first captured image generated when the first weighting module performs a first weighting operation is shown.

[0021] Fig.10 An example of a method in which the second weighting module performs the second weighting operation is shown.

[0022] Fig.11 An example of a second layout image and a second captured image generated when the second weighting module performs a second weighting operation is shown.

[0023] Fig.12 An example of a method in which a filtering module selectively performs filtering is shown.

[0024] Fig.13 An example of a third layout image and a third photographed image generated when the filtering module performs filtering is shown.

[0025] Fig.14 An example of a method in which the alignment module performs an alignment operation is shown.

[0026] Fig.15 An example of a method in which a semiconductor manufacturing system manufactures a semiconductor device is shown. DETAILED DESCRIPTION

[0027] Hereinafter, embodiments of the present invention will be described in detail and clearly so that those skilled in the art can easily implement the present invention.

[0028] Figure 1 A semiconductor manufacturing system 10 is shown in accordance with at least one embodiment of the present disclosure. Figure 1 , the semiconductor manufacturing system 10 may include a layout generation module 11, a learning module 12, a modification module 13, a manufacturing device 14, an imaging device 15, a database 16, and a defect detection module 17. The semiconductor manufacturing system 10 may manufacture semiconductor devices by using a wafer WAF.

[0029] The layout generation module 11 is configured to generate a layout image LO. For example, the layout generation module 11 may generate and / or receive design information based on a circuit. The layout generation module 11 may generate the layout image LO by placing standard cells in the layout image LO based on the design information. Alternatively, after placing the standard cells, the layout generation module 11 may generate the layout image LO by modifying the standard cells and / or by placing special cells that are not included in the standard cells under the control of the user. Thus, for example, the layout image LO generated by the layout generation module 11 may be a new layout image LO for manufacturing a new semiconductor device.

[0030] The learning module 12 is configured to perform the learning LRN of the modification module 13. For example, the learning module 12 can generate and update the modification module 13 based on machine learning. In at least one embodiment, the learning module 12 can include a trainable structure such as an artificial neural network, a decision tree, a support vector machine, a Bayesian network, a genetic algorithm, a convolutional neural network (CNN), a generative adversarial network (GAN), a recurrent neural network (RNN), a stacked deep neural network (S-DNN), a restricted Boltzmann machine (RBM), a fully convolutional network, a long short-term memory (LSTM) network, a classification network, etc. For example, the learning module 12 can perform the learning LRN based on various machine learning algorithms such as a neural network and / or a generative adversarial network (GAN).

[0031] The learning module 12 is configured to receive a layout image LO (or a modified layout image MLO) and a captured image IMG from the database 16 as training inputs. The layout image LO may be an image of an initial layout for manufacturing a semiconductor device; and the captured image IMG may be an image generated by capturing a semiconductor device manufactured based on the layout image LO (or the modified layout image MLO). For example, the learning module 12 may perform the learning LRN of the modification module 13 based on a front image (e.g., a layout image LO or a modified layout image MLO for manufacturing a semiconductor device) and a rear image (e.g., a captured image IMG after manufacturing a semiconductor device) of a manufactured semiconductor device. Therefore, the learning module 12 may be trained to output a learning LRN based on, for example, the difference between the layout image LO and the captured image IMG.

[0032] The modification module 13 may be trained by the learning module 12. The modification module 13 may receive a layout image LO for manufacturing a new semiconductor device from the layout generation module 11. In at least one embodiment, the modification module 13 may be trained to generate a modified layout image MLO according to the layout image LO. In at least one embodiment, the training of the modification module 13 is based on the learning LRN of the learning module 12.

[0033] For example, the modification module 13 may be trained to generate a modified layout image MLO based on the layout image LO based on various factors caused in the process of manufacturing the semiconductor device. For example, in at least one embodiment, the modification module 13 may be trained to generate a modified layout image MLO based on at least process proximity correction (PPC) and / or optical proximity correction (OPC).

[0034] For example, in a process of generating a photomask for manufacturing a semiconductor device, optical proximity correction may be performed to correct distortion caused in a photoresist pattern due to various factors, including characteristics of a light source, characteristics of a photoresist, a positional relationship between the light source and a pattern formed in the photoresist, etc. Process proximity correction may be used to correct distortion caused during a process (e.g., an etching process) due to various factors, including characteristics of a material used to perform the process, characteristics of a material to which the process is applied, characteristics of a photoresist pattern, etc.

[0035] The manufacturing equipment 14 may receive the modified layout image MLO from the modification module 13. The manufacturing equipment 14 may apply and / or modify the process PRC to the wafer WAF based on the modified layout image MLO. For example, the process PRC may include an etching process, a deposition process, a growth process, a planarization process, and the like. For example, the process PRC may be added, modified, and / or omitted based on the modified layout image MLO. In at least one embodiment, for example, the angle of the light source and / or the pattern may be adjusted, the duration of the process may be modified (e.g., reduced and / or increased), the critical dimension (CD) may be modified, the shape of the features in the photomask may be adjusted, and the like. When the process PRC is applied to the wafer WAF, a semiconductor device may be formed in the wafer WAF.

[0036] The imaging device 15 is configured to capture the wafer WAF (see Figure 1 In at least one example embodiment, the imaging device 15 may include an electron microscope, such as a scanning electron microscope (SEM).

[0037] The database 16 is configured to receive the layout image LO from the layout generating module 11 and receive the captured image IMG of the semiconductor device manufactured based on the layout image LO from the imaging device 15. The database 16 is configured to store and manage a pair of the layout image LO and the corresponding captured image IMG.

[0038] The defect detection module 17 is configured to receive the layout image LO and the corresponding captured image IMG from the database 16, and detect defects of the semiconductor device by comparing the layout image LO and the captured image IMG. For example, the defect detection module 17 can detect defects of the semiconductor device by comparing the front image (e.g., layout image LO) and the rear image (e.g., captured image IMG) of the semiconductor device. In at least one embodiment, the database 16 and / or the learning module 12 receives data representing defects detected by the defect detection module 17, so that the defect data can be included in the learning LRN.

[0039] The learning module 12 may also receive the layout image LO and the photographed image IMG used for manufacturing the semiconductor device from the database 16 and may perform learning.

[0040] In at least one embodiment, the layout generation module 11, the learning module 12, the modification module 13 and the defect detection module 17 may be implemented using software executable by a processor, a processor designed to perform related functions, or a combination of hardware and software designed to perform related functions.

[0041] Figure 2 FIG. 1 is a block diagram of an electronic device 100 according to at least one embodiment of the present disclosure. Figure 2 , the electronic device 100 may include a processor 110 , a random access memory 120 , a device driver 130 , a storage device 140 , a modem 150 , and a user interface 160 .

[0042] The processor 110 may include, for example, at least one general-purpose processor, such as a central processing unit (CPU) 111 and / or an application processor (AP) 112. In addition, the processor 110 may also include at least one dedicated processor, such as a neural processing unit (NPU) 113, a neuromorphic processor (NP) 114, and / or a graphics processing unit (GPU) 115. In at least one embodiment, the processor 110 may include two or more homogeneous processors.

[0043] At least one of the processors 110 may be used to train the module 200 or execute the trained module 200. At least one of the processors 110 may train or execute the module 200 based on various data or information. For example, the module 200 may be implemented in the form of instructions (or codes) executed by at least one of the processors 110. In this case, the at least one processor may load the instructions (or codes) of the module 200 to the random access memory 120.

[0044] For another example, at least one (or at least another) processor in the processors 110 may be manufactured to implement the module 200. For example, at least one processor may be a dedicated processor implemented in hardware based on the module 200 generated through learning of the module 200.

[0045] For another example, at least one (or at least another) processor in the processor 110 may be manufactured to implement various machine learning or deep learning modules. At least one processor may implement the module 200 by receiving information corresponding to the module 200 (eg, instructions or codes).

[0046] The random access memory 120 may be used as a working memory of the processor 110 and / or may be used as a main memory or a system memory of the electronic device 100. The random access memory 120 may include, for example, a volatile memory (such as a dynamic random access memory or a static random access memory) and / or a nonvolatile memory (such as a phase change random access memory, a ferroelectric random access memory, a magnetic random access memory, or a resistive random access memory).

[0047] The device driver 130 may control the following peripheral devices (e.g., storage device 140, modem 150, and user interface 160) according to a request of the processor 110. The storage device 140 may include a fixed storage device (such as a hard disk drive or a solid state drive) and / or a removable storage device (such as an external hard disk drive, an external solid state drive, or a removable memory card).

[0048] The modem 150 is configured to provide remote communication with an external device. The modem 150 can perform wired or wireless communication with an external device. The modem 150 can communicate with an external device based on at least one of various communication schemes such as Ethernet, wireless fidelity (Wi-Fi), long term evolution (LTE), fifth generation (5G) mobile communication, etc.

[0049] The user interface 160 may receive information from the user and / or may provide information to the user. The user interface 160 may include at least one user output interface (such as a display 161 or a speaker 162) and at least one user input interface (such as a mouse 163, a keyboard 164 or a touch input device 165).

[0050] The instructions (or codes) of the module 200 may be received through the modem 150 and may be stored in the storage device 140. The instructions (or codes) of the module 200 may be stored in a removable storage device, and the removable storage device may be connected to the electronic device 100. The instructions (or codes) of the module 200 may be loaded from the storage device 140 to the random access memory 120 and executed.

[0051] In at least one embodiment, module 200 may include reference to Figure 1 At least one of the described layout generation module 11, learning module 12, modification module 13 and defect detection module 17. The layout generation module 11, learning module 12, modification module 13 and defect detection module 17 can be implemented or executed in different electronic devices. At least two of the layout generation module 11, learning module 12, modification module 13 and defect detection module 17 can be implemented or executed in one electronic device.

[0052] Figure 3 FIG. 1 shows an operation method of the electronic device 100 according to at least one embodiment of the present disclosure. Figure 1 , Figure 2 and Figure 3 In operation S110, the electronic device 100 receives the layout image LO and the photographed image IMG. For example, the learning module 12 of the module 200 executed by the processor 110 of the electronic device 100 may receive the layout image LO and the photographed image IMG from the database 16.

[0053] In operation S120, the electronic device 100 may align the layout image LO and the captured image IMG based on the edges and corners. For example, the learning module 12 of the module 200 executed by the processor 110 of the electronic device 100 may highlight the edges and corners of the patterns of the layout image LO and the captured image IMG to generate a new layout image and a new captured image, and the layout image LO and the captured image IMG may be aligned by using the new layout image and the new captured image.

[0054] In at least one embodiment, the outline of the pattern of the layout image LO or the photographed image IMG may be an edge. The portion where the outlines extending in two or more different directions intersect among the outlines of the pattern of the layout image LO or the photographed image IMG may be a corner.

[0055] The learning module 12 according to at least one embodiment of the present disclosure may improve the alignment accuracy of the layout image LO and the photographed image IMG by highlighting edges and corners and aligning the layout image LO and the photographed image IMG.

[0056] In operation S130, the electronic device 100 may perform machine learning for layout modification. For example, the learning module 12 of the module 200 executed by the processor 110 of the electronic device 100 may perform machine learning based on the aligned layout image LO and the aligned captured image IMG. The learning module 12 may perform machine learning so that the contour of the aligned captured image IMG is closer to the contour of the aligned layout image LO.

[0057] The learning module 12 according to at least one embodiment of the present disclosure performs machine learning based on the layout image LO and the captured image IMG aligned with improved accuracy. Therefore, the accuracy of machine learning is improved and the distortion of the learning direction is prevented.

[0058] Figure 4 FIG. 3 shows a learning module 300 according to at least one embodiment of the present disclosure. In at least one embodiment, the learning module 300 may correspond to Figure 1 Learning Module 12. Figure 1 , Figure 2 and Figure 4 The learning module 300 may include a determination module 310 , a first weighting module 320 , a second weighting module 330 , a filtering module 340 , an alignment module 350 , and a machine learning module 360 ​​.

[0059] The determination module 310 may receive the layout image LO and the captured image IMG (e.g., from the database 16). The determination module 310 may selectively determine whether to highlight the edges and corners for the layout image LO and the captured image IMG. When it is determined that it is not necessary to highlight the edges and corners, the determination module 310 is configured to send the layout image LO and the captured image IMG to the machine learning module 360. The machine learning module 360 ​​may perform machine learning by using the layout image LO and the captured image IMG in which the edges and corners are not highlighted. When it is determined that it is necessary to highlight the edges and corners and / or it is beneficial to highlight the edges and corners, the determination module 310 is configured to send the layout image LO and the captured image IMG to the first weighting module 320.

[0060] For example, when it is determined that edges and corners are to be highlighted, the first weighting module 320 receives the layout image LO and the captured image IMG from the determination module 310. The first weighting module 320 may perform a first weighting operation so that the edges and / or corners of the layout image LO and the captured image IMG are weighted. For example, the first weighting operation may be used to highlight the edges of the patterns of the layout image LO and the captured image IMG. The first weighting module 320 may generate a first layout image LO1 and a first captured image IMG1 by highlighting the edges and / or corners of the layout image LO and the captured image IMG. The first layout image LO1 may be at least partially different from the layout image LO, and the first captured image IMG1 may be at least partially different from the captured image IMG.

[0061] In addition, when it is determined that edges and corners are to be highlighted, the second weighting module 330 receives the first layout image LO1 and the first captured image IMG1 from the first weighting module 320. The second weighting module 330 may perform a second weighting operation so that the edges and / or corners of the first layout image LO1 and the first captured image IMG1 are weighted. For example, the second weighting operation may be used to highlight the corners of the patterns of the first layout image LO1 and the first captured image IMG1. The second weighting module 330 may generate a second layout image LO2 and a second captured image IMG2 by highlighting the edges and / or corners of the first layout image LO1 and the first captured image IMG1. The second layout image LO2 may be at least partially different from the first layout image LO1, and the second captured image IMG2 may be at least partially different from the first captured image IMG1.

[0062] Since the edges are weighted (or highlighted) by the first weighting module 320 and the corners are weighted (or highlighted) by the second weighting module 330, the influence of the edges and corners in the pattern of the second layout image LO2 and the second captured image IMG2 may be increased compared to the influence of the edges and corners in the pattern of the layout image LO and the captured image IMG.

[0063] The filtering module 340 is configured to receive the second layout image LO2 and the second photographed image IMG2 from the second weighting module 330. The filtering module 340 may further increase the influence of edges and corners by selectively performing filtering on the second layout image LO2 and the second photographed image IMG2.

[0064] In at least one embodiment, when there is no filtering target in the second layout image LO2 and the second captured image IMG2, the filtering module 340 is configured to output the second layout image LO2 and the second captured image IMG2. When there is a filtering target in the second layout image LO2 and the second captured image IMG2, the filtering module 340 is configured to generate a third layout image LO3 by filtering the second layout image LO2 and the second captured image IMG2 (see Fig.13 ) and the third captured image IMG3 (see Fig.13 ), and a third layout image LO3 and a third photographed image IMG3 can be output.

[0065] The alignment module 350 may receive the second layout image LO2 and the second captured image IMG2 from the filtering module 340, and / or may receive the third layout image LO3 and the third captured image IMG3 from the filtering module 340. The alignment module 350 may align the received layout image (e.g., the second layout image LO2 or the third layout image LO3) and the received captured image (e.g., the second captured image IMG2 or the third captured image IMG3) based on the received layout image (e.g., the second layout image LO2 or the third layout image LO3) and the received captured image (e.g., the second captured image IMG2 or the third captured image IMG3).

[0066] The alignment module 350 may determine the alignment information AI as a result of aligning the received layout image (e.g., the second layout image LO2 or the third layout image LO3) with the received captured image (e.g., the second captured image IMG2 or the third captured image IMG3). The alignment information AI may include information indicating how much the layout image LO has moved relative to the captured image IMG or information indicating how much the captured image IMG has moved relative to the layout image LO, for example, relative position information. The alignment module 350 may send the alignment information AI to the machine learning module 360.

[0067] The machine learning module 360 ​​may receive the layout image LO and the captured image IMG from the determination module 310. The machine learning module 360 ​​may selectively receive alignment information AI from the alignment module 350. For example, when it is determined that it is not necessary to highlight the edges and corners, the machine learning module 360 ​​may not receive the alignment information AI from the alignment module 350. In the absence of the alignment information AI, the machine learning module 360 ​​may perform machine learning on the modification module 13 to generate a modified layout image MLO based on the layout image LO and the captured image IMG, so that the modified layout image MLO is configured to generate a semiconductor device similar to the layout image LO according to the layout image LO.

[0068] When it is determined that edges and corners need to be highlighted, the machine learning module 360 ​​can align the layout image LO and the captured image IMG based on the alignment information AI. For example, the machine learning module 360 ​​can align the layout image LO and the captured image IMG by moving the layout image LO or the captured image IMG based on the alignment information AI (e.g., based on the relative position information included in the alignment information AI). Based on the aligned layout image LO and the aligned captured image IMG, the machine learning module 360 ​​can perform machine learning on the modification module 13 to generate a modified layout image MLO.

[0069] As described above, the learning module 12 can generate an image highlighting edges and corners based on the layout image LO and the captured image IMG, and can detect alignment information AI based on the image highlighting edges and corners. The learning module 12 can align the layout image LO and the captured image IMG based on the alignment information AI detected based on the highlighted edges and corners, and can perform machine learning. Since machine learning is performed with improved accuracy through the highlighted edges and corners, the learning module 12 can train the modification module 13 more accurately.

[0070] Figure 5 FIG. 3 shows an example in which the determination module 310 determines whether to highlight edges and corners. Figure 4 and Figure 5 , in operation S210, the determination module 310 receives the layout image LO and the photographed image IMG.

[0071] In operation S220, the determination module 310 calculates a histogram of oriented gradients (HOG) based on the layout image LO. For example, the determination module 310 may detect the orientation of the outline of the pattern in the layout image LO. The determination module 310 may calculate the HOG based on the layout image LO, based on the captured image IMG, or based on the layout image LO and the captured image IMG.

[0072] In operation S230, the determination module 310 determines whether the layout image LO includes two or more orientations. When it is determined that the layout image LO includes two or more orientations, in operation S240, the determination module 310 may send the layout image LO and the captured image IMG to the first weighting module 320. That is, the determination module 310 may determine that edges and corners need to be highlighted based on determining that there are two or more orientations.

[0073] When the layout image LO or the photographed image IMG does not include two or more orientations in operation S250 (for example, when the layout image LO or the photographed image IMG includes one orientation or does not include an orientation), the determination module 310 may send the layout image LO and the photographed image IMG to the machine learning module 360. That is, the determination module 310 may determine that it is not necessary to highlight edges and corners.

[0074] Figure 6 An example of a layout image LO and a captured image IMG including a pattern having one orientation is shown. Figure 6 , the pattern (eg, dot pattern) of the layout image LO and the captured image IMG extends along the first direction. The outline of the pattern also extends along the first direction. The outline of the pattern extends only along the first direction and does not extend along the second direction.

[0075] For example, when the patterns of the layout image LO and the photographed image IMG have only one orientation, there may be no corners in the layout image LO and the photographed image IMG. When there may be no corners in the layout image LO and the photographed image IMG, based on the layout image LO and the photographed image IMG, the determination module 310 may determine that it is not necessary to highlight the edges and corners.

[0076] Figure 7 An example of a layout image LO and a captured image IMG including patterns having two orientations is shown. Figure 5 and Figure 7 , the outline of the pattern extending in the first direction and the outline of the pattern extending in the second direction may contact each other at the corner CN. In at least one embodiment, the main direction in which the pattern of the layout image LO and the captured image IMG extends may be the first direction. In the layout image LO and the captured image IMG, the outline extending in a direction different from the main direction of the pattern may be the corner CN. Figure 7 In the layout image LO and the captured image IMG, the outline of the pattern extending along the second direction may be a corner CN.

[0077] When the pattern of the layout image LO and the photographed image IMG includes two or more orientations, the pattern of the layout image LO and the photographed image IMG can be determined to include corners and edges. When the layout image LO and the photographed image IMG have corners and edges, based on the layout image LO and the photographed image IMG, the determination module 310 can determine that the edges and corners need to be highlighted.

[0078] Figure 8 An example of a method in which the first weighting module 320 performs the first weighting operation is shown. Figure 4 and Figure 8 , in operation S310, the first weighting module 320 receives the layout image LO and the photographed image IMG.

[0079] In operation S320, the first weighting module 320 may detect edges. For example, the first weighting module 320 may detect edges of patterns of the layout image LO and the photographed image IMG. For example, the first weighting module 320 may detect contours of patterns of the layout image LO and the photographed image IMG as edges.

[0080] In operation S330, the first weighting module 320 removes the area surrounded by the edge. For example, the first weighting module 320 may remove the remaining pixel values ​​except the pixel values ​​of the outline in the pattern of the layout image LO and the photographed image IMG. The first weighting module 320 may highlight the edge by retaining only the edge in the pattern of the layout image LO and the photographed image IMG.

[0081] Fig. 9 2 shows an example of a first layout image LO1 and a first photographed image IMG1 generated when the first weighting module 320 performs the first weighting operation. Figure 7 Compared with the layout image LO and the photographed image IMG, the first weighting module 320 may highlight the edge by removing the pixel values ​​of the dotted regions in the pattern and leaving only the edge.

[0082] Fig.10 An example of a method in which the second weighting module 330 performs the second weighting operation is shown. Figure 4 and Fig.10 In operation S410, the second weighting module 330 receives the first layout image LO1 and the first photographed image IMG1.

[0083] In operation S420, the second weighting module 330 may detect a corner. For example, the second weighting module 330 may detect a corner according to edges of patterns of the first layout image LO1 and the first photographed image IMG1.

[0084] In operation S430, the second weighting module 330 thickens the corner. For example, the second weighting module 330 may highlight the corner by enlarging the thickness of the corner.

[0085] Fig.11 2 shows an example of a second layout image LO2 and a second photographed image IMG2 generated when the second weighting module 330 performs the second weighting operation. Fig. 9 Compared with the first layout image LO1 and the first captured image IMG1, the second weighting module 330 can generate a second layout image LO2 and a second captured image IMG2 that highlight the corner CN by expanding the thickness of the corner CN extending in a direction (e.g., the second direction) different from the main direction of the edge (e.g., the first direction).

[0086] Fig.12 An example of a method in which the filtering module 340 selectively performs filtering is shown. Figure 4 and Fig.12 In operation S510, the filtering module 340 receives the second layout image LO2 and the second photographed image IMG2.

[0087] In operation S520, the filtering module 340 may detect an isolated pattern based on the second layout image LO2 and the second photographed image IMG2. For example, the filtering module 340 may detect, based on the patterns of the second layout image LO2 and the second photographed image IMG2, a pattern in which all edges are included in the second layout image LO2 and the second photographed image IMG2 and are not connected to other patterns as an isolated pattern.

[0088] In operation S530, the filtering module 340 determines whether an isolated pattern is detected. When an isolated pattern is detected, the filtering module 340 removes the non-isolated pattern in operation S540. In operation S550, the filtering module 340 outputs the third layout image LO3 and the third photographed image IMG3 from which the non-isolated pattern is removed to the alignment module 350.

[0089] When the isolated pattern is not detected, the filtering module 340 may output the second layout image LO2 and the second photographed image IMG2 to the alignment module 350 in operation S560.

[0090] That is, when the second layout image LO2 and the second photographed image IMG2 include an isolated pattern, the filtering module 340 may perform filtering. When filtering is performed so that only the isolated pattern is left, the filtering module 340 may generate a third layout image LO3 and a third photographed image IMG3. That is, when the second layout image LO2 and the second photographed image IMG2 do not include an isolated pattern, the filtering module 340 may omit filtering.

[0091] Fig.13 2 shows an example of a third layout image LO3 and a third photographed image IMG3 generated when the filtering module 340 performs filtering. Fig.11 Compared with the second layout image LO2 and the second photographed image IMG2, the filtering module 340 may generate a third layout image LO3 and a third photographed image IMG3 highlighting edges and corners by retaining only isolated patterns.

[0092] Fig.14 An example of a method in which the alignment module 350 performs an alignment operation is shown. Figure 4 and Fig.14 In operation S610, the alignment module 350 receives an image pair. For example, the alignment module 350 may receive an image pair including the second layout image LO2 and the second photographed image IMG2 from the filtering module 340, or may receive an image pair including the third layout image LO3 and the third photographed image IMG3 from the filtering module 340.

[0093] In operation S620, the alignment module 350 performs cross-correlation. For example, the alignment module 350 may perform cross-correlation on the images of the image pair. The alignment module 350 may perform fast Fourier cross-correlation on the images of the image pair.

[0094] In operation S630, the alignment module 350 detects a maximum intensity point. For example, the alignment module 350 may detect a point having a maximum value in the cross-correlation result as the maximum intensity point.

[0095] In operation S640, the alignment module 350 outputs the position information of the maximum intensity point as the alignment information AI. The maximum intensity point may indicate the position where the similarity of the images of the image pair is the highest. Therefore, the maximum intensity point may correspond to the shift information of the images of the image pair.

[0096] The alignment module 350 calculates the alignment information AI based on the image of the image pair highlighting the edges and corners. Therefore, the alignment module 350 can provide more accurate alignment information AI to the machine learning module 360.

[0097] In at least one embodiment, when highlighting of edges and corners is not performed, the machine learning module 360 ​​can calculate the mean square error (MSE) between the layout image LO and the captured image IMG, and can align the layout image LO and the captured image IMG at the point where the MSE is minimized.

[0098] When performing edge and corner highlighting, the machine learning module 360 ​​can align the layout image LO and the captured image IMG by using the alignment information AI. That is, when the alignment information AI is provided, additional calculations (e.g., calculations of MSE) for aligning the layout image LO and the captured image IMG can be omitted.

[0099] Fig.15 An example of a method of manufacturing a semiconductor device by the semiconductor manufacturing system 10 is shown. Figure 1 and Fig.15 In operation S710 , the semiconductor manufacturing system 10 receives a layout image LO. For example, the layout generation module 11 may generate the layout image LO. The modification module 13 may receive the layout image LO from the layout generation module 11 .

[0100] In operation S720, the semiconductor manufacturing system 10 modifies the layout image LO by using the modification module 13 based on machine learning (ML) trained by edge-corner salient. For example, the modification module 13 can generate a modified layout image MLO according to the layout image LO by using the modification module trained by edge-corner salient. The modified layout image MLO can reflect, for example, PPC, OPC, and / or PPC and OPC.

[0101] In operation S730, the semiconductor manufacturing system 10 manufactures a semiconductor device by using the modified layout image MLO. For example, the manufacturing equipment 14 may receive the modified layout image MLO from the modification module 13. The manufacturing equipment 14 may manufacture the semiconductor device by applying the process PRC to the wafer WAF.

[0102] In operation S740, the semiconductor manufacturing system 10 may capture an image of the semiconductor device. For example, the imaging device 15 may generate a captured image IMG by capturing an image of the wafer WAF to which the process PRC is applied.

[0103] In operation S750, the semiconductor manufacturing system 10 may detect defects based on the layout image LO and the captured image IMG. For example, the defect detection module 17 may compare the layout image LO and the captured image IMG, and may determine whether the captured image IMG is consistent with the design of the layout image LO. When the captured image IMG is consistent with the design of the layout image LO, the semiconductor device may be shipped. When the captured image IMG is inconsistent with the design of the layout image LO, the semiconductor device may be determined as a defective product and may not be shipped. In at least one embodiment, the captured image IMG and / or the defect may be used in, for example, unsupervised training of a modification module and / or a learning module based on ML, so that the training of the modification module and / or the learning module based on ML may be updated.

[0104] For example, in at least one embodiment, when a defect is detected as a result of comparing the layout image LO and the captured image IMG, the reference image may be used. Figures 1 to 14 The method of highlighting edges and corners described above. For example, the defect detection module 17 can calculate the alignment information AI by highlighting (e.g., applying a first weighting operation, a second weighting operation, and a filtering operation) the edges and corners of the layout image LO and the captured image IMG and performing cross-correlation on the generated image pair. The defect detection module 17 can align the layout image LO and the captured image IMG based on the alignment information AI, and can detect defects.

[0105] In the above-mentioned embodiments, the components according to the present disclosure are described by using the terms "first", "second", "third", etc. However, the terms "first", "second", "third", etc. may be used to distinguish components from each other and do not limit the present disclosure. For example, the terms "first", "second", "third", etc. do not refer to any form of order or numerical meaning.

[0106] In the above-mentioned embodiments, the components according to the embodiments of the present disclosure are referenced by using functional blocks. Unless otherwise explicitly indicated, functional blocks (including "units", "devices", "modules", etc.) may include processing circuits or be implemented with processing circuits, such as hardware including logic circuits; hardware / software combinations such as processors that execute software; or combinations thereof. For example, the processing circuits may more specifically include, but are not limited to, central processing units (CPUs), arithmetic logic units (ALUs), digital signal processors, microcomputers, systems on chips (SoCs), programmable logic units, microprocessors, application specific integrated circuits (ASICs), integrated circuits, application specific ICs (ASICs), field programmable gate arrays (FPGAs), and complex programmable logic devices (CPLDs), firmware driven in hardware devices, etc. In addition, these blocks may include circuits implemented with semiconductor elements in integrated circuits, or circuits registered as intellectual property (IP).

[0107] According to an embodiment of the present disclosure, source information is automatically aligned based on edges and corners. An electronic device and an operating method of the electronic device are provided, which aligns source information with improved accuracy and speed, thereby improving the accuracy and speed of machine learning for manufacturing semiconductor devices and reducing the time and cost of manufacturing semiconductor devices.

[0108] While the present disclosure has been described with reference to the embodiments thereof, it will be apparent to those skilled in the art that various changes and modifications can be made thereto without departing from the spirit and scope of the present disclosure as set forth in the appended claims.

Claims

1. A method for operating an electronic device, the electronic device comprising a processor and configured to support the manufacture of a semiconductor device, the method include: At the processor, receiving a layout image for manufacturing the semiconductor device and a captured image generated by capturing a semiconductor device actually manufactured; At the processor, in response to determining that two or more different orientations exist in the layout image, highlighting edges and corners of the layout image and the captured image; At the processor, aligning the layout image and the captured image based on a result of highlighting edges and corners of the layout image and the captured image; as well as at the processor, performing learning based on the aligned layout images and the aligned captured images so that a first modified layout image is generated from the layout images, The semiconductor device is manufactured based on a second modified layout image, and the second modified layout image is generated from the layout image based on the learning.

2. The operating method according to claim 1, further comprising: include: Determine a Histogram of Oriented Gradients (HOG) according to the layout image, The determining that there are two or more different orientations is obtained based on the HOG.

3. The operating method according to claim 2, further comprising: include: Based on determining that only one orientation exists based on the HOG, the layout image and the captured image are aligned without highlighting the edges and the corners.

4. The operating method according to claim 1, in, Highlighting the edge includes removing an area surrounded by the edge and the corner from the layout image and the captured image.

5. The operating method according to claim 1, in, Highlighting the corner includes: enlarging the corner.

6. The operating method according to claim 1, in, The alignment includes: detecting at least one isolated pattern from at least one of the layout image or the captured image; and Remaining patterns other than the detected at least one isolated pattern are removed from the layout image and the captured image.

7. The operating method according to claim 6, in, The aligning further includes maintaining the layout image and the captured image in response to the at least one isolated pattern not being detected.

8. The operating method according to claim 1, in, The alignment includes: generating a second layout image and a second captured image based on edges and corners of the layout image and the captured image; and The layout image and the photographed image are aligned based on edges and corners of the second layout image and the second photographed image.

9. The operating method according to claim 8, in, Aligning the layout image and the captured image based on edges and corners of the second layout image and the second captured image includes: performing cross-correlation on the second layout image and the second captured image; and Based on the result of the cross-correlation, the layout image and the captured image are aligned.

10. The operating method according to claim 1, in, Generating the first modified layout image and the second modified layout image includes applying at least one of process proximity correction (PPC) or optical proximity correction (OPC) to the layout images.

11. The operating method according to claim 1, in, The processor is trained to generate a third modified layout image based on the second layout image, and The third modified layout image is used to manufacture a second semiconductor device.

12. An operating method of an electronic device, the electronic device comprising a processor and configured to support the manufacture of a semiconductor device, the operating method include: At the processor, receiving a layout image for manufacturing the semiconductor device; as well as At the processor, a modified layout image is generated by modifying the layout image using a modification module based on machine learning. The modification module based on machine learning is trained to generate the modified layout image according to the layout image based on multiple layout images and multiple captured images of multiple semiconductor devices, wherein the multiple semiconductor devices are actually manufactured based on the multiple modified layout images generated according to the multiple layout images, and Among them, the machine learning-based modification module is trained to: highlight and align the edges and corners of the multiple layout images and the multiple captured images, and generate the modified layout image according to the layout image based on the aligned layout images and the aligned captured images.

13. The operating method according to claim 12, in, The machine learning based modification module is configured to: generating a plurality of second layout images and a plurality of second captured images from the plurality of layout images and the plurality of captured images based on the edges and the corners, respectively; as well as The plurality of layout images and the plurality of photographed images are aligned based on the plurality of second layout images and the plurality of second photographed images, respectively.

14. The operating method according to claim 13, in, Generating the plurality of second layout images and the plurality of second photographed images includes removing areas surrounded by the edges and the corners from the plurality of layout images and the plurality of photographed images.

15. The operating method according to claim 13, in, Generating the plurality of second layout images and the plurality of second photographed images includes: expanding corners of the plurality of second photographed images.

16. The operating method according to claim 13, in, Generating the multiple second layout images and the multiple second captured images includes: detecting at least one isolated pattern based on the layout image or at least one of the captured images; and removing remaining patterns in the multiple layout images and the multiple captured images except for the at least one isolated pattern.

17. The operating method according to claim 13, in, Aligning the plurality of layout images and the plurality of photographed images is based on results of performing cross-correlation on the plurality of second layout images and the plurality of second photographed images, respectively.

18. The operating method according to claim 12, in, Generating the modified layout image by modifying the layout image using the machine learning based modification module includes generating the modified layout image by applying at least one of process proximity correction (PPC) or optical proximity correction (OPC) to the layout image.

19. The operating method according to claim 12, further comprising: include: The semiconductor device is manufactured using the modified layout image.

20. An electronic device for manufacturing a semiconductor device, include: processor; as well as a memory including a plurality of layout images for manufacturing a plurality of semiconductor devices and a plurality of captured images captured after the plurality of semiconductor devices are manufactured, Wherein, the processor is configured to perform learning by aligning the multiple layout images and the multiple captured images by highlighting the edges and corners of the multiple layout images and the multiple captured images, so that the processor is trained to generate a modified layout image according to the layout image based on the aligned layout images and the aligned captured images.