Determination method and device of photoetching process window, electronic device and storage medium

Abstract

The invention relates to a determination method and device of a photoetching process window, an electronic device and a storage medium, and the determination method of the photoetching process window comprises the following steps: carrying out pattern segmentation operation on an obtained original photoetching image to generate a plurality of standard pixel images; inputting the plurality of standard pixel images into the trained target pattern extraction model to carry out photoetching pattern extraction processing, and outputting a required predicted photoetching pattern; wherein the target pattern extraction model is generated by training an initial generative adversarial network based on preset random noise data; and carrying out photoetching process window analysis processing based on the predicted photoetching pattern and the original photoetching image, and determining the photoetching process window. According to the method and the device, the problem of poor accuracy and reliability of an analysis result caused by limited sample data in photoetching process window analysis is solved, the demand on the quantity of original data is remarkably reduced, and the robustness and the accuracy of a photoetching process window analysis algorithm are improved.

Description

Technical Field

[0001] The present application relates to the field of semiconductor manufacturing processes, and in particular to a method, device, electronic device and storage medium for determining a photolithography process window. Background Art

[0002] The photolithography process is an indispensable part of the semiconductor integrated circuit manufacturing process. Its purpose is to accurately transfer the pattern on the mask to the silicon wafer to form a fine circuit structure. As an important parameter in the photolithography process, the process window is crucial to ensuring production stability and product quality.

[0003] In terms of related technologies, traditional lithography process window analysis relies on large amounts of high-quality data for model training and verification. However, in actual production environments, due to factors such as high data acquisition costs, complex processes, and time constraints, it is often difficult to obtain a sufficient amount of lithography image data. This small sample data set problem greatly limits the accuracy and reliability of process window analysis, thereby affecting the quality and production efficiency of semiconductor products.

[0004] Currently, no effective solution has been proposed for how to solve the problem of poor accuracy and reliability of analysis results due to limited sample data in related technologies in lithography process window analysis. Summary of the invention

[0005] The embodiments of the present application provide a method, device, system, electronic device and storage medium for determining a lithography process window, so as to at least solve the problem in the related art of how to solve the poor accuracy and reliability of analysis results due to limited sample data in lithography process window analysis.

[0006] In a first aspect, an embodiment of the present application provides a method for determining a lithography process window, comprising:

[0007] Performing a pattern segmentation operation on the acquired original photolithography image to generate a plurality of standard pixel images;

[0008] Inputting the plurality of standard pixel images into the trained target pattern extraction model for lithography pattern extraction processing, and outputting the required predicted lithography pattern; wherein the target pattern extraction model is generated by training an initial generative adversarial network based on preset random noise data;

[0009] A photolithography process window analysis process is performed based on the predicted photolithography pattern and the original photolithography image to determine the photolithography process window.

[0010] In some embodiments, the step of performing a pattern segmentation operation on the acquired original photolithography image to generate a plurality of standard pixel images includes:

[0011] Based on the pattern type of the original photolithography image, performing a pattern segmentation operation on the original photolithography image to obtain a plurality of single feature images;

[0012] A preprocessing operation is performed on the plurality of single feature images to generate a plurality of standard pixel images.

[0013] In some embodiments, the preprocessing operation is performed on the plurality of single feature images to generate the plurality of standard pixel images, including:

[0014] The plurality of single feature images are rotated and Gaussian blurred to generate a plurality of standard pixel images.

[0015] In some embodiments, the pattern segmentation operation is performed on the original lithography image based on the pattern type of the original lithography image to obtain a plurality of single feature images, including:

[0016] If the pattern type is a line-intensive type or a space-intensive type, based on the position of the original photolithography image and the key dimension data of the original photolithography image, the original photolithography image is segmented to obtain a plurality of the single feature images;

[0017] If the pattern type is a line sparse type or a space sparse type, performing normalized segmentation processing on the photolithography pattern to obtain a plurality of the single feature images;

[0018] If the pattern type is a hole-dense type or a diameter-dense type, based on the image side length of the original photolithography image and the distribution positions of the holes in the original photolithography image, the photolithography pattern is segmented to obtain a plurality of the single feature images;

[0019] If the pattern type is a hole sparse type or a diameter sparse type, the photolithography pattern is normalized and segmented to obtain a plurality of the single feature images.

[0020] In some embodiments, the generative adversarial network includes a discriminator and a generator; the method further includes:

[0021] Acquire a training photolithography image, and perform a pattern segmentation operation on the training photolithography image to generate a plurality of training segmentation images;

[0022] Inputting the random noise data into the generator in the generative adversarial network, and outputting a predicted training image;

[0023] Input the training segmented image and the predicted training image into the discriminator in the generative adversarial network, and output a probability prediction result;

[0024] Based on the probability prediction results, the discriminator and the generator in the generative adversarial network are continuously optimized through iteration, and finally a trained target pattern extraction model is generated.

[0025] In some embodiments, the step of inputting the training segmented image and the predicted training image into the discriminator in the generative adversarial network and outputting a probability prediction result includes:

[0026] Inputting the training segmented image into the discriminator, using the discriminator to extract features of the training segmented image to obtain image feature data, and determining fixed network parameters for the discriminator based on the image feature data;

[0027] The predicted training image is input into a discriminator with the fixed network parameters for convolution calculation processing, and the probability prediction result is output.

[0028] In some embodiments, the probability prediction result includes the similarity between the training switching image and the predicted training image; based on the probability prediction result, the discriminator and the generator in the generative adversarial network are continuously and iteratively optimized to finally generate the trained target pattern extraction model, including:

[0029] The similarity is compared with a preset confidence threshold of the similarity to obtain a similarity comparison result, and the generative adversarial network is iteratively optimized based on the similarity comparison result to generate a trained target pattern extraction model.

[0030] In a second aspect, an embodiment of the present application provides a device for determining a lithography process window, including:

[0031] A preprocessing module is used to perform a pattern segmentation operation on the acquired original photolithography image to generate a plurality of standard pixel images;

[0032] A data enhancement module, used for inputting the plurality of standard pixel images into a trained target pattern extraction model for photolithography pattern extraction processing, and outputting the required predicted photolithography pattern; wherein the target pattern extraction model is generated by training an initial generative adversarial network based on preset random noise data;

[0033] The photolithography analysis module is used to perform photolithography process window analysis based on the predicted photolithography pattern and the original photolithography image to determine the photolithography process window.

[0034] In a third aspect, an embodiment of the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for determining the lithography process window as described in the first aspect above is implemented.

[0035] In a fourth aspect, an embodiment of the present application provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for determining the lithography process window as described in the first aspect above.

[0036] Compared with the related art, the method, device, electronic device and storage medium for determining the lithography process window provided in the embodiments of the present application, by collecting and preprocessing the photoresist pattern measurement image data, constructing a target pattern extraction model for training, and generating synthetic data that is highly similar to the real image, thereby expanding and balancing the data set, and solving the problem of how to solve the poor accuracy and reliability of the analysis results due to limited sample data in the lithography process window analysis, significantly reducing the demand for the amount of original data, and improving the robustness and accuracy of the lithography process window analysis algorithm.

[0037] Details of one or more embodiments of the present application are set forth in the following drawings and description to make other features, objects, and advantages of the present application more readily apparent. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0039] Figure 1 It is a hardware structure block diagram of a terminal of the method for determining the photolithography process window according to an embodiment of the present invention;

[0040] Figure 2 is a flow chart of a method for determining a photolithography process window according to an embodiment of the present application;

[0041] Figure 3 It is a schematic diagram of the structure of a deep convolutional generative adversarial network according to a method for determining a lithography process window in a preferred embodiment of the present application;

[0042] Figure 4 It is a comparison diagram of a synthetic photoresist image and a real image generated during the model training process according to the preferred embodiment of the present application;

[0043] Figure 5 is a graph showing the relationship between the amount of original data and the recognition accuracy of synthetic images according to a preferred embodiment of the present application;

[0044] Figure 6is a schematic diagram of a first segmentation method for a dense pattern of line or space type according to a preferred embodiment of the present application;

[0045] Figure 7 is a segmented image according to the first segmentation method for dense patterns of line or space type in a preferred embodiment of the present application;

[0046] Figure 8 is a schematic diagram of a second cutting method for a dense pattern of line or space type according to a preferred embodiment of the present application;

[0047] Fig. 9 is a schematic diagram of a third cutting method for a dense pattern of line or space type according to a preferred embodiment of the present application;

[0048] Fig.10 is a schematic diagram of a sparse pattern of a line or space type according to a preferred embodiment of the present application;

[0049] Fig.11 is a schematic diagram of a dense pattern of diameters or hole types according to a preferred embodiment of the present application;

[0050] Fig.12 is a cut image of a dense pattern of diameter or hole type according to a preferred embodiment of the present application;

[0051] Fig.13 is a schematic diagram of a sparse pattern of diameter or hole type according to a preferred embodiment of the present application;

[0052] Fig.14 It is a structural block diagram of a device for determining a lithography process window according to an embodiment of the present application. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is described and illustrated below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. Based on the embodiments provided in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application. In addition, it can also be understood that although the efforts made in this development process may be complex and lengthy, for ordinary technicians in the field related to the contents disclosed in the present application, some changes such as design, manufacturing or production based on the technical contents disclosed in the present application are only conventional technical means, and should not be understood as insufficient contents disclosed in the present application.

[0054] Reference to "embodiments" in this application means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those of ordinary skill in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0055] Unless otherwise defined, the technical terms or scientific terms involved in this application should be understood by people with ordinary skills in the technical field to which this application belongs. The words "one", "a", "a", "the" and the like involved in this application do not indicate a quantitative limitation, and may represent the singular or plural. The terms "include", "comprise", "have" and any of their variations involved in this application are intended to cover non-exclusive inclusions; for example, a process, method, system, product or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units that are not listed, or may also include other steps or units inherent to these processes, methods, products or devices. The words "connect", "connected", "coupled" and the like involved in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The "multiple" involved in this application refers to greater than or equal to two. "And / or" describes the association relationship of associated objects, indicating that there can be three relationships, for example, "A and / or B" can represent: A exists alone, A and B exist at the same time, and B exists alone. The terms "first", "second", "third" and the like involved in the present application are merely used to distinguish similar objects and do not represent a specific ordering of the objects.

[0056] The method embodiment provided in this embodiment can be executed in a terminal, a computer or a similar computing device. Taking running on a terminal as an example, Figure 1 FIG. 1 is a hardware structure block diagram of a terminal of a method for determining a photolithography process window according to an embodiment of the present invention. Figure 1 As shown, the terminal may include one or more ( Figure 1 Only one is shown in the figure) a processor 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data. Optionally, the terminal may also include a transmission device 106 and an input / output device 108 for communication functions. It can be understood by those skilled in the art that Figure 1 The structure shown is for illustration only and does not limit the structure of the above terminal. Figure 1 More or fewer components as shown, or with Figure 1Different configurations are shown.

[0057] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the method for determining the lithography process window in the embodiment of the present invention. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, to implement the above method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0058] The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (Radio Frequency, referred to as RF) module, which is used to communicate with the Internet wirelessly.

[0059] This embodiment provides a method for determining a photolithography process window. Figure 2 is a flow chart of a method for determining a photolithography process window according to an embodiment of the present application, such as Figure 2 As shown, the process includes the following steps:

[0060] Step S201 , performing a pattern segmentation operation on the acquired original photolithography image to generate a plurality of standard pixel images.

[0061] Among them, this step preprocesses the original lithography image for subsequent input into the deep learning model. Specifically, the lithography image can be segmented according to the physical meaning and characteristics of the lithography image. For example, for dense patterns of line or space type, the original image can be segmented into multiple small images containing only a single feature according to the width and position of the pitch and critical dimension bar (CDbar). The segmentation method should ensure that each small image can accurately reflect the specific features in the original image. This step can greatly expand the amount of data and reduce the adverse effects of excessive data classification through the segmentation operation; the preprocessing operation can make the image data more standardized, which helps to improve the training efficiency and accuracy of the deep learning model.

[0062] Step S202, input multiple standard pixel images into the trained target pattern extraction model for lithography pattern extraction processing, and output the required predicted lithography pattern; wherein the target pattern extraction model is generated by training the initial generative adversarial network based on preset random noise data.

[0063] Among them, the target pattern extraction model is the DCGAN (Deep Convolutional GenerativeAdversarial Networks) model. This step uses the trained DCGAN model to extract the lithography pattern. Specifically, the generated standard pixel image is input into the DCGAN generator. The generator will generate a computer-generated image similar to the real lithography pattern based on the input random noise vector. These synthetic images predict the lithography pattern and are output to the subsequent analysis steps. This step can generate a large number of high-quality synthetic images through the DCGAN model, significantly reducing the demand for the amount of raw data, and improving the efficiency and accuracy of lithography pattern extraction, providing strong support for subsequent lithography process window analysis.

[0064] Step S203 , performing photolithography process window analysis and processing based on the predicted photolithography pattern and the original photolithography image to determine the photolithography process window.

[0065] Among them, this step combines the predicted lithography pattern with the original lithography image to further train the advanced lithography process window analysis tool based on deep learning. Through data enhancement and data set expansion, the lithography process window analysis model can better identify and predict the characteristics of abnormal exposure, development or measurement process in the photoresist pattern measurement image. The final model will determine the lithography process window based on these characteristics, that is, the exposure energy and focus value range that can meet the critical dimension (CD) and overlay requirements. This step improves the robustness and accuracy of the lithography process window analysis tool through data enhancement and expansion, reduces manual intervention, and improves the efficiency and consistency of the analysis process.

[0066] Through the above steps, the present application performs a pattern segmentation operation on the acquired original lithography image, and based on the pattern type of the lithography image (such as line-dense, space-dense, hole-dense, etc.), the original image is segmented into multiple small images containing only a single feature, and then converted into a standard pixel image suitable for input into DCGAN. This process not only expands the amount of data, but also improves the consistency of the data through standardization. Then, the multiple standard pixel images are input into the trained target pattern extraction model for lithography pattern extraction processing. The target pattern extraction model is based on preset random noise data and is generated by training the initial DCGAN. DCGAN consists of two neural networks: a generator and a discriminator. Composition, through optimizing adversarial training, the generator can gradually generate images similar to the real photolithography pattern measurement images, these generated images have similar features to the real data, can effectively supplement the deficiencies of small sample data sets, and balance various types of data; finally, based on the predicted photolithography pattern and the original photolithography image, the photolithography process window analysis is processed, and the computer-synthesized photoresist pattern measurement image data generated by DCGAN is combined with the original data for further training of advanced photolithography process window analysis tools based on deep learning. Through data enhancement and data set expansion, the model can better identify and predict the features in the photoresist pattern measurement image, thereby accurately determining the photolithography process window. Compared with the related art, this application significantly expands the data set through DCGAN technology and improves the balance of the data. Using the data-enhanced data set for training, the deep learning model is more fully trained and can more accurately extract the features of the photoresist pattern. Therefore, while reducing the demand for raw data, this application maintains a performance level comparable to that when only raw data is used, thereby effectively solving the problem of poor accuracy and reliability of the analysis results due to limited sample data.

[0067] In some embodiments, a pattern segmentation operation is performed on the acquired original photolithography image to generate a plurality of standard pixel images, including:

[0068] Based on the pattern type of the original photolithography image, a pattern segmentation operation is performed on the original photolithography image to obtain a plurality of single feature images;

[0069] Preprocessing operations are performed on multiple single feature images to generate multiple standard pixel images.

[0070] Among them, the pattern type of the original lithography image is first identified, which may include line-dense type, space-dense type, line-sparse type, space-sparse type, hole-dense type, and diameter-dense type, etc. Different pattern segmentation operations are performed on the original lithography image according to the pattern type to obtain multiple single feature images, and then the single feature images obtained by segmentation are preprocessed. Through precise pattern segmentation and preprocessing operations, this embodiment can significantly expand the data set required for lithography process window analysis, especially for those categories with a small number, thereby reducing the adverse effects of imbalance of small sample data sets on training deep learning networks; the automated pattern segmentation and preprocessing process reduces manual intervention, improves the efficiency and consistency of data set processing, and replaces manual naked eye judgment with computer unified standard judgment, ensuring the accuracy and reliability of data processing.

[0071] In some embodiments, a preprocessing operation is performed on a plurality of single feature images to generate a plurality of standard pixel images, including:

[0072] Multiple single feature images are rotated and Gaussian blurred to generate multiple standard pixel images.

[0073] Among them, each single feature image is rotated at a certain angle (such as 0°, 90°, 180°, 270°, etc.) to simulate the lithography pattern at different angles. The rotation operation can maintain the integrity of the image content and increase the diversity of the data set; a Gaussian blur filter is applied to the rotated image, and the degree of blur is controlled by adjusting the blur parameter (such as standard deviation). Gaussian blur processing can simulate the image blur phenomenon that may exist in the actual lithography process and improve the model's recognition ability of blurred images; the image after rotation and Gaussian blur processing is adjusted to a standard size suitable for input deep convolutional generative adversarial network (DCGAN) (such as 64×64 pixels). The size adjustment can ensure that all input images have consistent dimensions, which is convenient for model processing. This embodiment can generate a large number of images with different feature combinations through rotation and Gaussian blur processing, thereby significantly enhancing the diversity and richness of the data set, helping the model to learn more robust and generalized feature representations and improve the generalization ability of the model; using pre-processed standard pixel images for training can make the model better adapt to the diversity of actual lithography patterns, which helps to improve the accuracy and efficiency of the model in lithography process window analysis.

[0074] In some embodiments, based on the pattern type of the original lithography image, a pattern segmentation operation is performed on the original lithography image to obtain a plurality of single feature images, including:

[0075] If the pattern type is a line-intensive type or a space-intensive type, based on the position of the original photolithography image and the key dimension data of the original photolithography image, the original photolithography image is segmented to obtain a plurality of single feature images;

[0076] If the pattern type is a line sparse type or a space sparse type, the photolithography pattern is normalized and segmented to obtain a plurality of single feature images;

[0077] If the pattern type is a hole-dense type or a diameter-dense type, the photolithography pattern is segmented based on the image side length of the original photolithography image and the distribution positions of the holes in the original photolithography image to obtain a plurality of single feature images;

[0078] If the pattern type is a hole sparse type or a diameter sparse type, the photolithography pattern is normalized and segmented to obtain a plurality of single feature images.

[0079] Among them, for line-dense or space-dense patterns, the segmentation operation will take into account the specific location information of the lithography image and the data of key dimensions (such as line width, spacing, etc.), and accurately segment the original image according to the density of the pattern, the distribution of lines or spaces, and the specific values ​​of the key dimensions to obtain multiple small images containing only a single feature; for line-sparse or space-sparse patterns, due to the large spacing between pattern elements, a normalized segmentation method is used to process the original lithography image to normalize it to ensure that each segmented small image has consistent size and features, which is convenient for subsequent processing and analysis. For hole-dense or diameter-dense patterns, the segmentation operation will take into account the overall size of the image and the specific distribution position of the holes, and accurately segment the original image according to the density and distribution law of the holes to obtain multiple small images containing a single hole feature; similar to line-sparse or space-sparse patterns, hole-sparse or diameter-sparse patterns are also processed using a normalized segmentation method to ensure that each segmented small image has consistent size and features, which is convenient for subsequent data enhancement and deep learning model training.

[0080] This embodiment improves data quality and diversity: through precise segmentation operations, multiple standard pixel images containing single features can be obtained. These images have higher quality and diversity, which help to improve the training effect of the deep learning model; the automated segmentation method greatly reduces the degree of manual participation and improves the efficiency and consistency of data processing; using a precisely segmented and preprocessed data set for training can enhance the robustness and accuracy of the deep learning model for lithography process window analysis, and specific segmentation methods are proposed for different types of lithography patterns, making this method more widely applicable.

[0081] In some embodiments, the generative adversarial network includes a discriminator and a generator; the method further includes:

[0082] Acquire a training lithography image, and perform a pattern segmentation operation on the training lithography image to generate a plurality of training segmentation images;

[0083] Input random noise data into the generator in the generative adversarial network and output the predicted training image;

[0084] Input the training segmentation image and the predicted training image into the discriminator in the generative adversarial network and output the probability prediction result;

[0085] Based on the probability prediction results, the discriminator and generator in the generative adversarial network are continuously optimized through iteration, and finally a trained target pattern extraction model is generated.

[0086] Among them, a certain number of lithography images are selected from the existing lithography process data set as training lithography images, and the corresponding segmentation method is applied according to the pattern type of the lithography image. The segmented image is used as the training segmented image for subsequent GAN training. The generator receives a random noise vector as input, and gradually converts the random noise vector into a two-dimensional computer-synthesized photoresist pattern measurement image through the deconvolution layer. These generated images (i.e., predicted training images) have similar features to the real lithography images. The training segmented images and predicted training images are input to the discriminator in the generative adversarial network. The discriminator is a standard convolutional neural network used to distinguish whether the input image is a real training segmented image or a predicted training image generated by the generator. The discriminator distinguishes and judges the training segmented image and the predicted training image and outputs a probability prediction result, that is, the probability that the image is considered to be a real image. During the training process, the generator tries to generate realistic images to deceive the discriminator, while the discriminator strives to identify the true and false images. By optimizing adversarial training, the network parameters of the generator and the discriminator are gradually adjusted to improve the quality of the images generated by the generator. After multiple iterations, the generator can generate high-quality images similar to the real data set. The training process ends when the images generated by the generator are realistic enough and the discriminator has difficulty distinguishing between true and false. At this point, the resulting generator is a trained target pattern extraction model that can be used to generate synthetic images similar to real lithography images.

[0087] This embodiment uses GAN technology to generate a large amount of high-quality synthetic data using a small amount of real data, thereby reducing dependence on a large amount of original data. The generated synthetic data can balance the quantity of various types of data and reduce the adverse effects of imbalanced small sample data sets on deep learning network training. GAN technology can automatically generate highly similar synthetic images, reduce the workload of manual intervention and data annotation, and improve the efficiency of data collection and processing; and using a data-enhanced data set for training, the deep learning network can more accurately extract the features of the photoresist pattern and improve the accuracy and robustness of the lithography process window analysis tool.

[0088] In some embodiments, the training segmented image and the predicted training image are input into a discriminator in a generative adversarial network, and a probability prediction result is output, including:

[0089] Input the training segmented image into the discriminator, use the discriminator to extract features of the training segmented image to obtain image feature data, and determine fixed network parameters for the discriminator based on the image feature data;

[0090] The predicted training image is input into the discriminator with fixed network parameters for convolution calculation and outputs the probability prediction result.

[0091] Among them, multiple training segmented images obtained through pattern segmentation operation are input into the discriminator, which is a standard convolutional neural network with multiple convolutional layers for feature extraction. The discriminator performs convolution operation on the input training segmented images to extract image feature data. Through the training process, the discriminator gradually learns how to distinguish the features of real images from generated images. When the feature extraction of the training segmented images by the discriminator reaches a stable state, the network parameters of the discriminator are fixed so that they can be used to judge the authenticity of the generated images later. Random noise data is input into the generator in the generative adversarial network to generate predicted training images, and these predicted training images are input into the discriminator with fixed network parameters for convolution calculation processing. The discriminator judges the predicted training images based on the learned features and outputs the probability prediction result, that is, the probability that the image is considered to be a real image.

[0092] In this embodiment, the discriminator extracts features of the training segmented images, and the discriminator can learn the key features of the real image, which helps the discriminator to more accurately identify the authenticity of the image when subsequently judging the generated image, thereby improving the image recognition ability; during the training process, the generator attempts to generate realistic images to deceive the discriminator, and the discriminator provides feedback to the generator by outputting probability prediction results. The generator continuously adjusts its generation strategy based on the feedback from the discriminator, thereby gradually improving the quality of its generated images; after fixing the network parameters of the discriminator, the amount of calculation in the training process can be reduced, which helps to accelerate the training process and improve training efficiency. Moreover, through the continuous game and optimization between the discriminator and the generator, the model finally generated can better adapt to various image data, which helps to enhance the robustness of the model and make it have better performance in practical applications.

[0093] In some embodiments, the probability prediction result includes the similarity between the training switching image and the predicted training image; based on the probability prediction result, the discriminator and the generator in the generative adversarial network are continuously optimized in a loop, and finally a trained target pattern extraction model is generated, including:

[0094] The similarity is compared with the preset confidence threshold of the similarity to obtain a similarity comparison result, and the generative adversarial network is iteratively optimized based on the similarity comparison result to generate a trained target pattern extraction model.

[0095] Among them, during the training process, the discriminator will discriminate the input training segmentation image and the predicted training image generated by the generator, and output the similarity between them. This similarity is usually a probability value, indicating the possibility that the input image is considered by the discriminator to be a real image (i.e., similar to the image in the training data set). In order to evaluate the quality of the image generated by the generator, it is necessary to preset a confidence threshold of similarity. This threshold is used to judge whether the generated image is realistic enough and whether it can be regarded as valid data for subsequent iterative optimization. The similarity output by the discriminator is compared with the preset confidence threshold to obtain a similarity comparison result. If the similarity of the generated image is higher than the confidence threshold, it means that the performance of the generator is good and can be further fine-tuned to pursue higher image quality; if the similarity of the generated image is lower than the confidence threshold, the parameters of the generator need to be adjusted to improve its ability to generate images. This comparison and adjustment process will be iterated continuously until the generator can generate a synthetic image that is highly similar to the real image. After multiple iterative optimizations, the generator will be able to stably generate high-quality synthetic images. At this point, the trained generator can be used as a target pattern extraction model for subsequent lithography process window analysis tasks. This embodiment can significantly improve the quality of the image generated by the generator by comparing the similarity with a preset confidence threshold and iteratively optimizing the generative adversarial network, so that the generated synthetic image is more realistic and can better reflect the characteristics of the real photoresist pattern.

[0096] The embodiments of the present application are described and illustrated below through preferred embodiments.

[0097] Step 1: Collection, preprocessing and annotation of photoresist pattern measurement image data.

[0098] (1) Create an automatic measurement program (recipe) in the critical dimension scanning electron microscope, set the IDW and IDP parameters of the relevant recipe, and place the wafer to be measured. The machine will execute the steps set in the recipe, measure the photoresist pattern image and its corresponding critical dimension (CD) of the input wafer, and then collect the photoresist pattern image data.

[0099] (2) The collected original image of the photoresist pattern is segmented according to the corresponding method to obtain a small image of 120×120 pixels containing only a single feature. These small images are preprocessed by rotating and adding Gaussian blur, and then converted to a standard size of 64×64 pixels suitable for input into DCGAN through an algorithm.

[0100] (3) According to the classification criteria applicable to the lithography process window analysis, each image is manually labeled with categories 0, 1, and 2, and the image files are moved to the corresponding locations.

[0101] Step 2: Build a deep convolutional generative adversarial network (DCGAN).

[0102] Figure 3 is a schematic diagram of the structure of a deep convolutional generative adversarial network according to a method for determining a lithography process window in a preferred embodiment of the present application, such as Figure 3 As shown, DCGAN consists of two neural networks: a generator and a discriminator.

[0103] (1) Generator: The input of the generator is a 1-dimensional 100-bit deep random vector (encoded by the random function in the code, by randomly generating 100 random number sequences and transposing them into a 1-dimensional 100-bit vector). Through three deconvolution layers, the random noise vector is gradually converted into a two-dimensional computer-generated photoresist pattern measurement image. The deconvolution layer will gradually decode the noise vector data into image data until a 64×64 pixel image similar to the real photolithography pattern measurement image is generated. The above similarity discrimination criteria are obtained by training the discriminator.

[0104] (2) Discriminator: The discriminator is a standard convolutional neural network with three convolutional layers. Its input is a 64×64 pixel image. When training the discriminator, a certain type of image in the labeled data set is input, and the input image is feature extracted through three convolutional layers, and the corresponding network parameters are fixed. When using the discriminator for judgment, the image generated by the generator is input into the discriminator. The convolutional neural network with fixed parameters through training will perform convolution operations on the generated image and output whether the image is considered to have a high similarity with the image in the data set (real image) or a low similarity (fake image). DCGAN uses the discriminator to compete with the generator, and through optimized adversarial training, the quality of the image generated by the generator is gradually improved.

[0105] Step 3, DCGAN training process.

[0106] (1) During the training process, the generator will try to generate realistic images in an attempt to deceive the discriminator, while the discriminator will try to identify whether the image is real or fake based on the features extracted during training. Figure 4 is a comparison diagram of a synthetic photoresist image and a real image generated during the model training process according to a preferred embodiment of the present application, such as Figure 4 As shown, as training progresses, the generator is gradually able to generate high-quality images similar to the real dataset.

[0107] (2) The specific training parameters are as follows: the input image size is 64×64 pixels, the batch size is 8, the random noise vector depth is 100, the training is 300 epochs, and a checkpoint is set to prevent too many epochs.

[0108] The model uses the Adam optimizer with a learning rate of 0.0002 (each time the weight is updated, the weight is adjusted by 0.0002 based on the calculated gradient), and a β1 parameter of 0.5 (used to calculate the decay rate of the first-order moment estimate (i.e., the exponential moving average of the gradient), which controls the exponential decay rate of past gradients and is used to determine the weight of the current gradient when updating parameters).

[0109] (3) During the training process, the generator and the discriminator continue to compete until the images generated by the generator are visually highly similar to the real images. By regularly checking the performance of the discriminator, the quality of the images generated by the generator is ensured to gradually improve.

[0110] Step 4: Generate synthetic data and expand the dataset.

[0111] (1) Using the trained DCGAN model, 1000 sets of 1-dimensional 100-bit random noise vectors were input to generate 1000 computer-generated photoresist pattern measurement image data. These images have similar features to real data, which can effectively supplement the deficiencies of small sample data sets and balance various types of data.

[0112] (2) The discriminator is trained by inputting different numbers of original photoresist pattern measurement images (such as 1, 10, 100, 1000 and 10000), and 10000 computer-generated photoresist pattern measurement images are generated using the DCGAN trained under the above conditions. Figure 5 is a curve diagram showing the relationship between the amount of original data and the recognition accuracy of the synthetic image according to the preferred embodiment of the present application. Figure 5 As shown, these synthetic images were tested for similarity using the existing advanced lithography process window analysis tool based on deep learning. The results showed that when 1,000 real images were used for training, the similarity between the generated images and the real images reached 73.1%; when 10,000 real images were used, the similarity increased to 85.2%.

[0113] Step 5, applied to lithography process window analysis.

[0114] (1) Combine the computer-generated photoresist pattern measurement image data generated by DCGAN with the original photoresist pattern measurement image data to further train the advanced photolithography process window analysis tool based on deep learning. Through data enhancement and data set expansion, the model can better identify and predict the features of abnormal exposure, development or measurement process in the photoresist pattern measurement image, and accurately determine the photolithography process window.

[0115] (2) The actual experimental results show that by using the DCGAN technology to expand the data set and perform data enhancement, if the similarity is 70% as the confidence threshold, the existing deep learning-based advanced lithography process window analysis tool reduces the demand for raw photoresist pattern measurement image data by about 13 times, and the performance of the lithography process window analysis tool is equivalent to that when only the raw data is used.

[0116] In addition, the above-mentioned segmentation method for the original image is as follows:

[0117] ① For dense patterns of line or space type, the measurement is divided into the cases where the left and right ends of the image are within the pitch and the cases where they are within the critical dimension bar (CD bar).

[0118] The situation where the left and right ends of the measured image are within the spacing is divided into the situation where the center of the measured image is within the spacing and the situation where the center of the measured image is within the critical dimension line.

[0119] i. For the case where the center of the measured image is within the spacing, Figure 6 is a schematic diagram of a first segmentation method for a dense pattern of line or space type according to a preferred embodiment of the present application, such as Figure 6 As shown, the measured image is a square image with a side length (pixel) of v, the key dimension line width value corresponds to pixel c, the spacing width value corresponds to pixel p, and there are n key dimension lines on each side of the center spacing, where n is a positive integer. Then, the original image is cut into 2n segments along the x direction starting from the upper left corner (point A), and into [v / (c+p)] segments along the y direction. Figure 7 is a segmented image according to the first segmentation method for dense patterns of line or space type in the preferred embodiment of the present application, such as Figure 7 As shown, the segmented image is a square image of (c+p) pixels.

[0120] ii. When the center of the measured image is within the critical dimension line, Figure 8 is a schematic diagram of a second segmentation method for a dense pattern of line or space type according to a preferred embodiment of the present application, such as Figure 8 As shown, it is known that the measured image is a square image with a side length (pixel) of v, the key dimension line width value corresponds to pixel c, the spacing width value corresponds to pixel p, and there are n key dimension lines on each side of the central key dimension line, where n is a positive integer. The original image is cut into (2n+1) segments along the x direction starting from the upper left corner (point A), and into [v / (c+p)] segments along the y direction, and the cut image is a square image of (c+p) pixels.

[0121] When the left and right ends of the measured image are within the critical dimension lines, Fig. 9is a schematic diagram of a third segmentation method for a dense pattern of line or space type according to a preferred embodiment of the present application, such as Fig. 9 As shown, it is known that the measured image is a square image with a side length (pixel) of v, the key dimension line width value corresponds to pixel c, and the spacing width value corresponds to pixel p. It is only necessary to move the starting point A of the segmentation when the center of the measured image is within the spacing from the upper left corner to the right to the first segment spacing, that is, the x-coordinates of all segmentation points are added with a bias of (c+p / 2), and the rest of the segmentation methods are consistent with the two cases where the left and right ends of the measured image are within the spacing.

[0122] ②For sparse patterns of line or space type, Fig.10 is a schematic diagram of a sparse pattern of line or space type according to a preferred embodiment of the present application, such as Fig.10 As shown, it is known that the measured image is a square image with a side length (pixel) of v, and the key dimension line width value corresponds to a pixel of c, and normalization processing is performed directly.

[0123] ③For dense patterns of diameter or hole type, Fig.11 is a schematic diagram of a dense pattern of diameters or hole types according to a preferred embodiment of the present application, such as Fig.11 As shown, the known measured image is a square image with a side length (pixel) of v, where each row has h holes and each column has t holes. The original image is evenly divided into h segments along the x direction and t segments along the y direction starting from the upper left corner (point A). Fig.12 is a segmented image of a dense pattern of diameter or hole type according to a preferred embodiment of the present application, such as Fig.12 As shown, the segmented image is a rectangle with a length and a width (or a width and a length) of v / h pixels and v / t pixels respectively.

[0124] ④For sparse patterns of diameter or hole type, Fig.13 is a schematic diagram of a sparse pattern of diameter or hole type according to a preferred embodiment of the present application, such as Fig.13 As shown, it is known that the measured image is a square image with a side length (pixel) of v, and normalization is performed directly.

[0125] This embodiment also provides a device for determining a lithography process window, which is used to implement the above-mentioned embodiments and preferred implementation modes, and will not be repeated hereafter. As used below, the terms "module", "unit", "subunit", etc. may be a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0126] Fig.14 is a structural block diagram of a device for determining a lithography process window according to an embodiment of the present application, such as Fig.14As shown, the device comprises:

[0127] The pre-processing module 10 is used to perform a pattern segmentation operation on the acquired original photolithography image to generate a plurality of 1 standard pixel images.

[0128] The data enhancement module 20 is used to input multiple standard pixel images into the trained target pattern extraction model for lithography pattern extraction processing, and output the required predicted lithography pattern; wherein the target pattern extraction model is generated by training the initial generative adversarial network based on preset random noise data.

[0129] The photolithography analysis module 30 is used to perform photolithography process window analysis based on the predicted photolithography pattern and the original photolithography image to determine the photolithography process window.

[0130] It should be noted that the above modules can be functional modules or program modules, and can be implemented by software or hardware. For modules implemented by hardware, the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0131] This embodiment further provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0132] Optionally, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0133] Optionally, in this embodiment, the processor may be configured to perform the following steps through a computer program:

[0134] S1, performing a pattern segmentation operation on the acquired original photolithography image to generate a plurality of standard pixel images.

[0135] S2, inputting multiple standard pixel images into the trained target pattern extraction model for lithography pattern extraction processing, and outputting the required predicted lithography pattern; wherein the target pattern extraction model is generated by training the initial generative adversarial network based on preset random noise data.

[0136] S3, performing photolithography process window analysis and processing based on the predicted photolithography pattern and the original photolithography image to determine the photolithography process window.

[0137] It should be noted that the specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementation modes, and this embodiment will not be described in detail here.

[0138] In addition, in combination with the method for determining the lithography process window in the above embodiment, the embodiment of the present application can provide a storage medium for implementation. The storage medium stores a computer program; when the computer program is executed by a processor, any one of the methods for determining the lithography process window in the above embodiment is implemented.

[0139] Those skilled in the art should understand that the technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0140] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A method for determining a photolithography process window, characterized in that: include: Performing a pattern segmentation operation on the acquired original photolithography image to generate a plurality of standard pixel images; Inputting the plurality of standard pixel images into the trained target pattern extraction model for lithography pattern extraction processing, and outputting the required predicted lithography pattern; wherein the target pattern extraction model is generated by training an initial generative adversarial network based on preset random noise data; A photolithography process window analysis process is performed based on the predicted photolithography pattern and the original photolithography image to determine the photolithography process window.

2. The method for determining the photolithography process window according to claim 1, characterized in that: The pattern segmentation operation is performed on the acquired original photolithography image to generate a plurality of standard pixel images, including: Based on the pattern type of the original photolithography image, performing a pattern segmentation operation on the original photolithography image to obtain a plurality of single feature images; A preprocessing operation is performed on the plurality of single feature images to generate a plurality of standard pixel images.

3. The method for determining the photolithography process window according to claim 2, characterized in that: The preprocessing operation is performed on the plurality of single feature images to generate the plurality of standard pixel images, including: The plurality of single feature images are rotated and Gaussian blurred to generate a plurality of standard pixel images.

4. The method for determining the photolithography process window according to claim 2, characterized in that: The pattern segmentation operation is performed on the original photolithography image based on the pattern type of the original photolithography image to obtain a plurality of single feature images, including: If the pattern type is a line-intensive type or a space-intensive type, based on the position of the original photolithography image and the key dimension data of the original photolithography image, the original photolithography image is segmented to obtain a plurality of the single feature images; If the pattern type is a line sparse type or a space sparse type, performing normalized segmentation processing on the photolithography pattern to obtain a plurality of the single feature images; If the pattern type is a hole-dense type or a diameter-dense type, based on the image side length of the original photolithography image and the distribution positions of the holes in the original photolithography image, the photolithography pattern is segmented to obtain a plurality of the single feature images; If the pattern type is a hole sparse type or a diameter sparse type, the photolithography pattern is normalized and segmented to obtain a plurality of the single feature images.

5. The method for determining the photolithography process window according to claim 1, characterized in that: The generative adversarial network includes a discriminator and a generator; the method further includes: Acquire a training photolithography image, and perform a pattern segmentation operation on the training photolithography image to generate a plurality of training segmentation images; Inputting the random noise data into the generator in the generative adversarial network, and outputting a predicted training image; Input the training segmented image and the predicted training image into the discriminator in the generative adversarial network, and output a probability prediction result; Based on the probability prediction results, the discriminator and the generator in the generative adversarial network are continuously optimized through iteration, and finally a trained target pattern extraction model is generated.

6. The method for determining the photolithography process window according to claim 5, characterized in that: The step of inputting the training segmented image and the predicted training image into the discriminator in the generative adversarial network and outputting a probability prediction result comprises: Inputting the training segmented image into the discriminator, using the discriminator to extract features of the training segmented image to obtain image feature data, and determining fixed network parameters for the discriminator based on the image feature data; The predicted training image is input into a discriminator with the fixed network parameters for convolution calculation processing, and the probability prediction result is output.

7. The method for determining the photolithography process window according to claim 5, characterized in that: The probability prediction result includes the similarity between the training switching image and the predicted training image; based on the probability prediction result, the discriminator and the generator in the generative adversarial network are continuously optimized in a loop iteration to finally generate the trained target pattern extraction model, including: The similarity is compared with a preset confidence threshold of the similarity to obtain a similarity comparison result, and the generative adversarial network is iteratively optimized based on the similarity comparison result to generate a trained target pattern extraction model.

8. A device for determining a photolithography process window, characterized in that: include: A preprocessing module is used to perform a pattern segmentation operation on the acquired original photolithography image to generate a plurality of standard pixel images; A data enhancement module, used for inputting the plurality of standard pixel images into a trained target pattern extraction model for lithography pattern extraction processing, and outputting the required predicted lithography pattern; wherein the target pattern extraction model is generated by training an initial generative adversarial network based on preset random noise data; The photolithography analysis module is used to perform photolithography process window analysis based on the predicted photolithography pattern and the original photolithography image to determine the photolithography process window.

9. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to run the computer program to execute the method for determining the photolithography process window according to any one of claims 1 to 7.

10. A storage medium, characterized in that: The storage medium stores a computer program, wherein the computer program is configured to execute the method for determining the photolithography process window according to any one of claims 1 to 7 when running.

Images