Image generation model training method and device, image generation method and device

Abstract

The application provides a training method and device of an image generation model, and an image generation method and device. The training method of the image generation model comprises the following steps: obtaining a first SEM image and a training process variation bandwidth image; inputting the first SEM image into a generator of the image generation model, and training the generator with the training process variation bandwidth image as a target image; and iteratively training the generator according to a loss function in the image generation model until the pixel value of the training result image output in a coordinate position and the pixel value of the training process variation bandwidth image in a corresponding coordinate position satisfy a preset threshold condition. By inputting the SEM image under a specific process into the image generation model, the process variation bandwidth image under a process window can be obtained, the change of the pattern on the wafer after exposure is facilitated to be researched, and the photolithography process is continuously improved according to the change.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of semiconductor integrated circuit manufacturing lithography process, and in particular to a method and device for training an image generation model, and a method and device for generating an image. BACKGROUND

[0002] In actual production, the exposure dose and focus value of the lithography machine will have some instability. In addition, due to the mask preparation process, the pattern size on the mask cannot be exactly the same as required. These deviations or instabilities are inevitable, but can be controlled within a certain range. In order to verify the influence of these deviations on the lithography process, the optical proximity correction based on the lithography process window is usually used, and the concept of process variation band (PV Band) is introduced, that is, process deviations are deliberately introduced in the calculation, and the changes of the pattern on the wafer after exposure are observed. As shown in FIG. 1, under a given set of {focus point, exposure dose} conditions, 03 represents the SEM image generated under the standard process, and the PV Band is defined as the area between the outer contour 01 and the inner contour 02. Figure 1

[0003] For the calculation method of the PV Band, the common method at present is to accurately calculate the exposure dose and focus value under different process conditions, and use the metrology machine to obtain the scanning electron microscope (SEM) images under these different process conditions, and then superimpose these SEM images, and measure the process variation band image by extracting the contours after superimposition. This method is very complicated and time-consuming.

[0004] Therefore, in order to improve the efficiency of the semiconductor integrated circuit lithography process, the present application provides a method and device for training an image generation model, and a method and device for generating an image. SUMMARY

[0005] The present application provides a method and device for training an image generation model, and a method and device for generating an image. By inputting the SEM image under the standard process into the image generation model, the process variation band image under the process window can be obtained, which is convenient for studying the changes of the pattern on the wafer after exposure, and continuously improving the lithography process according to the changes.

[0006] In a first aspect, the present application provides a method for training an image generation model, comprising:

[0007] ​obtaining a first SEM image and a training process variation bandwidth image, the first SEM image being an SEM image formed under a specific lithography focus value and exposure dose matrix condition, the specific lithography focus value and exposure dose matrix condition being selected from optimal lithography focus value and exposure dose matrix condition required by semiconductor manufacturing; inputting the first SEM image into a generator of an image generation model, and training the generator with the training process variation bandwidth image as a target image; iteratively training the generator according to a loss function in the image generation model until the first SEM image is input into the trained generator, and pixel values of a training result image output by the generator at coordinate positions satisfy a preset threshold condition with pixel values of the training process variation bandwidth image at corresponding coordinate positions, and then the training of the image generation model is completed.

[0008] The beneficial effects of the present application are that: by establishing a coordinate system on the image and setting the coordinate positions that need to be compared, the calculation amount of the loss function in the comparison process can be reduced. Because in the semiconductor manufacturing process, the lithography process is to copy the image onto the wafer through the exposure technology, and the electric charge in the exposure technology is very small, so the influence of the exposure technology on the lithography process needs to be considered, and the pixel value is more appropriate for comparison, and the pixel value can reflect the average brightness information of a small square on the image; the present application further iteratively trains the generator according to the loss function in the image generation model, so that the pixel values of the training result image output by the generator at the coordinate positions satisfy the preset threshold condition with the pixel values of the training process variation bandwidth image at the corresponding coordinate positions, and the image generation model suitable for the application scenario of the lithography process is obtained at the end of the training.

[0009] Optionally, the method further comprises: obtaining a second SEM image and a verification process variation bandwidth image, the second SEM image being an SEM image formed under a specific lithography focus value and exposure dose matrix condition, the specific lithography focus value and exposure dose matrix condition being selected from optimal lithography focus value and exposure dose matrix condition required by semiconductor manufacturing; inputting the second SEM image into a generator of an image generation model, and outputting a verification result image; when the pixel values of the verification result image at the coordinate positions satisfy a preset error condition with the pixel values of the verification process variation bandwidth image at the corresponding coordinate positions, the image generation model verification is successful. The beneficial effects of the present application are that: by judging whether the pixel values of the verification result image at the coordinate positions satisfy the preset error condition with the pixel values of the verification process variation bandwidth image at the corresponding coordinate positions, it is determined whether the image generation model can be put into production and used, and the accuracy of the image generation model in generating the required image is also improved.

[0010] Optionally, the image generation model is trained iteratively on the generator according to the loss function in the image generation model until the first SEM image is input into the trained generator, and the pixel value of the output training result image at the coordinate position and the pixel value of the training process variation bandwidth image at the corresponding coordinate position satisfy a preset threshold condition, and the image generation model training is completed, including: determining a preset auxiliary loss function corresponding to the first SEM image at the coordinate position according to the pixel value of the training result image at the coordinate position and the pixel value of the training process variation bandwidth image at the corresponding coordinate position, and a preset loss function threshold, wherein the auxiliary loss function represents a minimized mean square error loss function; obtaining a final loss function of the generator according to the optimization loss function of the image generation model, the auxiliary loss function and the weight of the auxiliary loss function; and training the generator iteratively according to the final loss function until the first SEM image is input into the trained generator, and the pixel value of the output training result image at the coordinate position and the pixel value of the training process variation bandwidth image at the corresponding coordinate position satisfy a preset threshold condition, and the image generation model training is completed. The beneficial effect is that by comparing the pixel value of the output training result image at the coordinate position and the pixel value of the training process variation bandwidth image at the corresponding coordinate position, the judgment efficiency of the image generation model can be improved more efficiently and accurately, and the minimized mean square error loss function can be used to evaluate the change degree of the pixel value at the corresponding coordinate position, so as to judge whether the pixel value of the output training result image at the coordinate position and the pixel value of the training process variation bandwidth image at the corresponding coordinate position satisfy the preset threshold condition, and when the threshold condition is satisfied, the image generation model training is successful.

[0011] In a second aspect, the present application provides a process variation bandwidth image generation method, including: obtaining a standard SEM image, the standard SEM image being an SEM image formed under a specific lithography focus value and exposure dose matrix condition, and the specific lithography focus value and exposure dose matrix condition being selected from the optimal lithography focus value and exposure dose matrix condition required for semiconductor manufacturing; inputting the standard SEM image into a generator of an image generation model, and outputting a result image, wherein the image generation model is trained according to the training method of the image generation model in the first aspect described above, and the result image is a process variation bandwidth image of the standard SEM image. The beneficial effect is that by inputting an SEM image formed under the optimal exposure process condition into the generator of the image generation model, the corresponding process variation bandwidth image can be obtained, without the need to superimpose and align all SEM images under the process window, and then measure and calculate the process variation bandwidth image manually. The method of the present application is simple and fast, and improves the working efficiency of the lithography process.

[0012] In a third aspect, the present application provides a device for training an image generation model, which comprises modules / units for implementing the method of any possible design of the first aspect. These modules / units can be implemented by hardware, or by hardware executing corresponding software.

[0013] In a fourth aspect, the present application provides a device for generating a process variation bandwidth image, which comprises modules / units for implementing the method of any possible design of the second aspect. These modules / units can be implemented by hardware, or by hardware executing corresponding software.

[0014] In a fifth aspect, the present application provides an electronic device, which comprises a processor and a memory. The memory is configured to store one or more programs; and when the one or more programs stored in the memory are executed by the processor, the electronic device is enabled to implement the method of any possible design of the first aspect and the second aspect.

[0015] In a sixth aspect, the present application provides a storage medium, which stores a program. When the program is executed by a processor, the method of any possible design of the first aspect and the second aspect is implemented. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 A schematic diagram of a process variation bandwidth image;

[0017] Figure 2 A flowchart of a method for training an image generation model according to an embodiment of the present application;

[0018] Figure 3 A schematic diagram of training an image generation model according to an embodiment of the present application;

[0019] Figure 4 A flowchart of a method for generating a process variation bandwidth image according to an embodiment of the present application;

[0020] Figure 5 A schematic diagram of a device for training an image generation model according to an embodiment of the present application;

[0021] Figure 6 A schematic diagram of a device for generating a process variation bandwidth image according to an embodiment of the present application;

[0022] Figure 7 A schematic diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0023] Before the embodiments of the present application are described in detail, the following first explains some terms in the embodiments of the present application, so as to facilitate understanding by those skilled in the art.

[0024] 1) Photolithography process window

[0025] The photolithography process is one of the most core processes in the integrated circuit manufacturing technology, and the photolithography process window refers to the range of exposure dose and focus value that meet the line width requirement.

[0026] 2) Optimal photolithography focus value and exposure dose matrix condition

[0027] There are many methods that can be used to determine the optimal photolithography focus value and exposure dose matrix condition in the photolithography process. At present, the method for determining the optimal photolithography focus value and exposure dose matrix condition is mainly to analyze the Poisson curve formed by the data of the SEM image collected by the scanning electron microscope and the imaging quality of the SEM image, and to determine the optimal photolithography focus value and exposure dose matrix condition according to the analysis result. The specific photolithography focus value and exposure dose matrix condition mentioned in the present application is selected from the optimal photolithography focus value and exposure dose matrix condition required for semiconductor manufacturing.

[0028] The technical solutions in the embodiments of the present application will be described below in combination with the drawings in the embodiments of the present application. In the description of the embodiments of the present application, the terms used in the following embodiments are only for the purpose of describing the specific embodiments and are not intended to be limiting on the present application. As used in the specification and the appended claims of the present application, the singular expressions "a", "the", "said", "the", and "this" are intended to also include expressions such as "one or more", unless there is clear indication to the contrary in the context. It should also be understood that in the following embodiments of the present application, "at least one", "one or more" means one or more than two (including two). The term "and / or" is used to describe the association relationship of the associated objects, which means that there can be three kinds of relationships; for example, A and / or B can represent the following three cases: A exists alone, A and B exist together, and B exists alone, where A and B can be singular or plural. The character " / " generally represents a "or" relationship between the associated objects before and after it.

[0029] Reference in the specification to "one embodiment" or "some embodiments" means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the application. The appearances of the phrase "in one embodiment" or "in some embodiments" in various places in the specification are not necessarily all referring to the same embodiment, although it can. The terms "including," "comprising," "having" and variations thereof are meant to encompass the items listed thereafter and equivalents thereof as well as additional items. The terms "connected," "coupled," and variations thereof are meant to encompass the items listed thereafter, as well as implementations in which the items are directly connected, and implementations in which the items are indirectly connected, through one or more intermediaries. The terms "first," "second," and variations thereof merely mean different instances and do not imply a relative importance or a chronological order.

[0030] In the present application, the word "exemplary" or "for example" is used to mean serving as an example, instance, or illustration. Any embodiment or design described in this application as "exemplary" or "for example" should not be construed as necessarily preferred or advantageous over other embodiments or designs. Rather, use of terms such as "exemplary" or "for example" is merely intended to present concepts in a concrete manner.

[0031] The present application provides a method for training an image generation model, as shown in Figure 2 The method steps include:

[0032] S201, obtaining a first SEM image and a training process variation bandwidth image.

[0033] The first SEM image is an SEM image formed under a specific lithography focus value and exposure dose matrix condition, and the specific lithography focus value and exposure dose matrix condition is selected from the optimal lithography focus value and exposure dose matrix condition required for semiconductor manufacturing.

[0034] The lithography focus value and exposure dose matrix condition is a matrix condition composed of the focus value and exposure dose of the set process window and the focus value and exposure dose within the variation range thereof, and the variation range can be ±10%.

[0035] S202, inputting the first SEM image into a generator of an image generation model, and training with the training process variation bandwidth image as a target image.

[0036] In this step, as shown in Figure 3 The first SEM image 1 is input into an encoder 301 in the generator of the image generation model, and through a hidden space 302 and a decoder 303, a training result image 2 is generated and input into a discriminator 304.

[0037] S203, iteratively training the generator according to a loss function in the image generation model until the first SEM image is input into the trained generator, and a pixel value of the output training result image at a coordinate position and a pixel value of the training process variation bandwidth image at a corresponding coordinate position satisfy a preset threshold condition, and the image generation model training is completed.

[0038] A coordinate system is established on the image, and a coordinate position that needs to be compared is set, so as to reduce the calculation amount of the loss function in the comparison process. Because in the semiconductor manufacturing process, the photolithography process is to copy the image to the wafer through the exposure technology, and the electric charge in the exposure technology is very small, the influence of the exposure technology on the photolithography process needs to be considered, and it is more appropriate to compare the pixel values, and the pixel values can reflect the average brightness information of a small square on the image. In this step, the discriminator 304 judges whether the pixel value of the output training result image 2 at the coordinate position and the pixel value of the training process variation bandwidth image 3 at the corresponding coordinate position satisfy the preset threshold condition, and if yes, the training result image is output; otherwise, the training is re-performed, and a new training result image is obtained.

[0039] In this embodiment, a scanning electron microscope can be used to perform lateral scanning at different coordinates of the wafer after photolithography, and the SEM images obtained by scanning are saved. The image size is determined according to the specific situation, and in this example, it is 512*512 pixels. Because the coordinates are fixed, the positions of each pixel are also fixed. By selecting the pixel values of the pixels at the coordinate positions in the appropriate range for comparison, the comparison result can be obtained more efficiently.

[0040] In a possible embodiment, as shown in FIG. 4, the training method of the image generation model can further include the following steps: Figure 2

[0041] S204, obtaining a second SEM image and a verification process variation bandwidth image.

[0042] The second SEM image is an SEM image formed under a specific photolithography focus value and an exposure dose matrix condition. The specific photolithography focus value and the exposure dose matrix condition are selected from the optimal photolithography focus value and the exposure dose matrix condition required for semiconductor manufacturing.

[0043] S205, inputting the second SEM image into the generator of the image generation model, and outputting a verification result image.

[0044] ​S206, when the pixel value of the verification result image at the coordinate position and the pixel value of the verification process variation bandwidth image at the corresponding coordinate position satisfy a preset error condition, the image generation model verification is successful.

[0045] In this step, if the deviation of the pixel value of the verification result image at the coordinate position and the pixel value of the verification process variation bandwidth image at the corresponding coordinate position is too large, the training of the image generation model is re-performed, and if the re-training still fails to achieve accuracy, the data preparation of the training sample image is re-performed, and data enhancement is appropriately performed.

[0046] In another possible embodiment, the image generation model is trained according to the loss function in the image generation model until the pixel value of the training result image at the coordinate position and the pixel value of the training process variation bandwidth image at the corresponding coordinate position satisfy a preset threshold condition, and the image generation model training is completed, including: determining a preset auxiliary loss function corresponding to the first SEM image at the coordinate position according to the pixel value of the training result image at the coordinate position and the pixel value of the training process variation bandwidth image at the corresponding coordinate position, and a preset loss function threshold, the auxiliary loss function represents a minimum mean square error loss function; obtaining a final loss function of the generator according to the optimization loss function of the image generation model, the auxiliary loss function and the weight of the auxiliary loss function; the generator is iteratively trained according to the final loss function until the first SEM image is input into the trained generator, and the pixel value of the output training result image at the coordinate position and the pixel value of the training process variation bandwidth image at the corresponding coordinate position satisfy a preset threshold condition, and the image generation model training is completed.

[0047] In this embodiment, by comparing the pixel value of the output training result image at the coordinate position and the pixel value of the training process variation bandwidth image at the corresponding coordinate position, the judgment efficiency of the image generation model can be more efficiently and accurately improved, and the minimum mean square error loss function can be used to evaluate the change degree of the pixel value at the corresponding coordinate position, so as to judge whether the pixel value of the output training result image at the coordinate position and the pixel value of the training process variation bandwidth image at the corresponding coordinate position satisfy the preset threshold condition, when the threshold condition is satisfied, the image generation model training is successful.

[0048] Based on the image generation model obtained by the image generation model training method mentioned in the above embodiments, the present embodiment provides a process variation bandwidth image generation method, as shown in Figure 4 The steps are as follows:

[0049] S401, acquire a standard SEM image.

[0050] The standard SEM image is an SEM image formed under a specific photolithography focus value and exposure dose matrix condition, the specific photolithography focus value and exposure dose matrix condition being selected from optimal photolithography focus value and exposure dose matrix conditions required for semiconductor manufacturing.

[0051] S402, input the standard SEM image into a generator of an image generation model, and output a result image.

[0052] The image generation model is trained according to the training method of the image generation model in any one of the above embodiments, and the result image is a process variation bandwidth image of the standard SEM image.

[0053] An embodiment of the present application provides a training device 500 of an image generation model, as shown in the figure, the device 500 comprises: Figure 5

[0054] An acquisition module 501 is configured to acquire a first SEM image and a training process variation bandwidth image, the first SEM image being the SEM image formed under a specific photolithography focus value and exposure dose matrix condition, the specific photolithography focus value and exposure dose matrix condition being selected from optimal photolithography focus value and exposure dose matrix conditions required for semiconductor manufacturing.

[0055] The first SEM image and the training process variation bandwidth image can be called in a database.

[0056] An input module 502 is configured to input the first SEM image into a generator of an image generation model, and train the generator with the training process variation bandwidth image as a target image.

[0057] A training module 503 is configured to iteratively train the generator according to a loss function in the image generation model, until the first SEM image is input into the trained generator, and a pixel value of a training result image output by the trained generator at a coordinate position satisfies a preset threshold condition with a pixel value of the training process variation bandwidth image at a corresponding coordinate position, then the training of the image generation model is completed.

[0058] The embodiment mainly relies on a neural network algorithm to train the generator in the image generation model.

[0059] ​In a possible embodiment, the training apparatus 500 further comprises a verification module 504, configured to: acquire a second SEM image and a verification process variation bandwidth image, the second SEM image being an SEM image formed under a specific lithography focus value and an exposure dose matrix condition, the specific lithography focus value and the exposure dose matrix condition being selected from optimal lithography focus values and exposure dose matrix conditions required for semiconductor manufacturing; input the second SEM image into a generator of the image generation model, and output a verification result image; and when a pixel value of the verification result image at a coordinate position and a pixel value of the verification process variation bandwidth image at a corresponding coordinate position satisfy a preset error condition, the image generation model is verified to be successful.

[0060] In another possible embodiment, the training apparatus iteratively trains the generator according to a loss function in the image generation model until a pixel value of a training result image output by the generator after the first SEM image is input into the generator and a pixel value of the training process variation bandwidth image at a corresponding coordinate position satisfy a preset threshold condition, and the image generation model is trained to be complete, and the training apparatus is further configured to: determine a preset auxiliary loss function corresponding to the first SEM image at a coordinate position according to the pixel value of the training result image at the coordinate position, the pixel value of the training process variation bandwidth image at the corresponding coordinate position, and a preset loss function threshold, the auxiliary loss function representing a minimized mean square error loss function.

[0061] The final loss function of the generator is obtained according to the optimized loss function of the image generation model, the auxiliary loss function, and a weight of the auxiliary loss function.

[0062] The generator is iteratively trained according to the final loss function until a pixel value of a training result image output by the generator after the first SEM image is input into the generator and a pixel value of the training process variation bandwidth image at a corresponding coordinate position satisfy a preset threshold condition, and the image generation model is trained to be complete.

[0063] Embodiments of the present application provide a process variation bandwidth image generation apparatus 600, as shown in Figure 6 The apparatus 600 comprises:

[0064] An acquisition module 601 is configured to acquire a standard SEM image, the standard SEM image being an SEM image formed under a specific lithography focus value and an exposure dose matrix condition, the specific lithography focus value and the exposure dose matrix condition being selected from optimal lithography focus values and exposure dose matrix conditions required for semiconductor manufacturing.

[0065] The input module 602 is used to input the standard SEM image into the generator of the image generation model and output the result image. The image generation model is trained according to the training method of the image generation model as described in any of the above embodiments, and the result image is the process variation bandwidth image of the standard SEM image.

[0066] In this embodiment, by inputting a SEM image formed under specific process conditions into the generator of the image generation model, the corresponding process variation bandwidth image can be obtained. It is not necessary to overlay and align all SEM images under the process window and then obtain the process variation bandwidth image by manual measurement and calculation. The method of the present invention is simple and fast, and improves the working efficiency of the photolithography process.

[0067] In other embodiments of this application, an electronic device is disclosed, such as... Figure 7 As shown, the electronic device may include: one or more processors 701; a memory 702; a display 703; one or more application programs (not shown); and one or more programs 704. These devices can be connected via one or more communication buses 705. The one or more programs 704 are stored in the memory 702 and configured to be executed by the one or more processors 701. The one or more programs 704 include instructions that can be used to perform actions such as... Figure 2 or Figure 4 Each step in the corresponding embodiment.

[0068] Through the above description of the embodiments, those skilled in the art will clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device, and unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0069] In the embodiments of this application, the functional units can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0070] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a storage medium. Based on such an understanding, the technical solutions of the embodiments of the present application essentially or in other words the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The software product is stored in a storage medium or a processor executes all or part of the steps of the method of the embodiments of the present application. The foregoing storage medium includes: a flash memory, a mobile hard disk, a read-only memory, a random access memory, a magnetic disk or an optical disk, and various media that can store program codes.

[0071] The above is only a specific implementation of the embodiments of the present application, but the protection scope of the embodiments of the present application is not limited thereto. Any change or replacement within the technical scope disclosed by the embodiments of the present application should be covered within the protection scope of the embodiments of the present application. Therefore, the protection scope of the embodiments of the present application should be subject to the protection scope of the claims.

Claims

1. A method for training an image generation model, characterized in that, The method comprises the following steps: obtaining a first SEM image and a training process variation bandwidth image, wherein the first SEM image is an SEM image formed under a specific photolithography focus value and an exposure dose matrix condition, and the specific photolithography focus value and the exposure dose matrix condition are selected from optimal photolithography focus values and exposure dose matrix conditions required for semiconductor manufacturing; inputting the first SEM image into a generator of an image generation model, and training the generator with the training process variation bandwidth image as a target image; iteratively training the generator according to a loss function in the image generation model until the first SEM image is input into the trained generator, and pixel values of a training result image output by the trained generator at coordinate positions satisfy a preset threshold condition with pixel values of the training process variation bandwidth image at corresponding coordinate positions, and then the image generation model training is completed.

2. The method of claim 1, wherein, The method further comprises the following steps: obtaining a second SEM image and a verification process variation bandwidth image, wherein the second SEM image is an SEM image formed under a specific photolithography focus value and an exposure dose matrix condition, and the specific photolithography focus value and the exposure dose matrix condition are selected from optimal photolithography focus values and exposure dose matrix conditions required for semiconductor manufacturing; inputting the second SEM image into the generator of the image generation model, and outputting a verification result image; when pixel values of the verification result image at coordinate positions satisfy a preset error condition with pixel values of the verification process variation bandwidth image at corresponding coordinate positions, the image generation model verification is successful.

3. The method of claim 1, wherein, The iteratively training the generator according to a loss function in the image generation model until the first SEM image is input into the trained generator, and pixel values of a training result image output by the trained generator at coordinate positions satisfy a preset threshold condition with pixel values of the training process variation bandwidth image at corresponding coordinate positions, and then the image generation model training is completed, comprises the following steps: determining a preset auxiliary loss function corresponding to the first SEM image at a coordinate position according to pixel values of the training result image at the coordinate position, pixel values of the training process variation bandwidth image at the corresponding coordinate position, and a preset loss function threshold, wherein the auxiliary loss function represents a minimized mean square error loss function; obtaining a final loss function of the generator according to an optimization loss function of the image generation model, the auxiliary loss function, and a weight of the auxiliary loss function; iteratively training the generator according to the final loss function until the first SEM image is input into the trained generator, and pixel values of a training result image output by the trained generator at coordinate positions satisfy a preset threshold condition with pixel values of the training process variation bandwidth image at corresponding coordinate positions, and then the image generation model training is completed.

4. A method of generating a process variation bandwidth image, the method comprising: The method comprises the following steps: obtaining a standard SEM image, wherein the standard SEM image is an SEM image formed under a specific photolithography focus value and an exposure dose matrix condition, and the specific photolithography focus value and the exposure dose matrix condition are selected from optimal photolithography focus values and exposure dose matrix conditions required for semiconductor manufacturing; inputting the standard SEM image into a generator of an image generation model, and outputting a result image, wherein the image generation model is trained according to the training method of the image generation model in any one of claims 1 to 3, and the result image is a process variation bandwidth image of the standard SEM image.

5. An apparatus for training an image generation model, comprising: The method comprises: an acquisition module, configured to acquire a first SEM image and a training process variation bandwidth image, wherein the first SEM image is an SEM image formed under a specific photolithography focus value and exposure dose matrix condition, and the specific photolithography focus value and exposure dose matrix condition is selected from optimal photolithography focus value and exposure dose matrix conditions required for semiconductor manufacturing; an input module, configured to input the first SEM image into a generator of an image generation model, and train the generator with the training process variation bandwidth image as a target image; a training module, configured to iteratively train the generator according to a loss function in the image generation model, until the first SEM image is input into the trained generator, and a pixel value of a training result image output by the trained generator at a coordinate position meets a preset threshold condition with a pixel value of the training process variation bandwidth image at a corresponding coordinate position, so that the training of the image generation model is completed.

6. The apparatus of claim 5, wherein, The device further comprises: a verification device, configured to acquire a second SEM image and a verification process variation bandwidth image, wherein the second SEM image is an SEM image formed under a specific photolithography focus value and exposure dose matrix condition, and the specific photolithography focus value and exposure dose matrix condition is selected from optimal photolithography focus value and exposure dose matrix conditions required for semiconductor manufacturing; input the second SEM image into the generator of the image generation model, and output a verification result image; and when a pixel value of the verification result image at a coordinate position meets a preset error condition with a pixel value of the verification process variation bandwidth image at a corresponding coordinate position, the image generation model is verified successfully.

7. The apparatus of claim 5, wherein, The training device iteratively trains the generator according to a loss function in the image generation model, until the first SEM image is input into the trained generator, and a pixel value of a training result image output by the trained generator at a coordinate position meets a preset threshold condition with a pixel value of the training process variation bandwidth image at a corresponding coordinate position, so that the training of the image generation model is completed, and is further configured to: determine a preset auxiliary loss function corresponding to the coordinate position of the first SEM image according to the pixel value of the training result image at the coordinate position, the pixel value of the training process variation bandwidth image at the corresponding coordinate position, and a preset loss function threshold, wherein the auxiliary loss function represents a minimized mean square error loss function; obtain a final loss function of the generator according to an optimization loss function of the image generation model, the auxiliary loss function, and a weight of the auxiliary loss function; and The generator is iteratively trained according to the final loss function until the first SEM image is input into the trained generator, and the pixel value of the output training result image at the coordinate position and the pixel value of the training process variation bandwidth image at the corresponding coordinate position satisfy a preset threshold condition, and the image generation model training is completed.

8. An apparatus for generating a bandwidth image of process variation, characterized in that, The method comprises: An acquisition module is configured to acquire a standard SEM image, wherein the standard SEM image is an SEM image formed under a specific photolithography focus value and an exposure dose matrix condition, and the specific photolithography focus value and the exposure dose matrix condition are selected from optimal photolithography focus values and exposure dose matrix conditions required for semiconductor manufacturing; An input module is configured to input the standard SEM image into a generator of an image generation model, and output a result image, wherein the image generation model is trained according to the image generation model training method in any one of claims 1 to 3, and the result image is a process variation bandwidth image of the standard SEM image.

9. An electronic device, comprising: A device comprises a memory and a processor, and the memory stores a program capable of running on the processor, and when the program is executed by the processor, the processor implements the image generation model training method in any one of claims 1 to 3 or the process variation bandwidth image generation method in claim 4.

10. A storage medium having stored therein a program, characterized by The program is executed by the processor to implement the image generation model training method in any one of claims 1 to 3 or the process variation bandwidth image generation method in claim 4.

Images