Method and device for detecting photolithography process window

Abstract

The present invention provides a method and device for detecting a lithography process window. The method comprises: inputting a first lithography spatial feature map and sequentially inputting lithography conditions in a first lithography condition set into a trained neural network model to generate a test image corresponding to the lithography conditions; extracting the lithography contour in the test image to form a contour image; and aggregating all the obtained contour images into a contour image set; obtaining an OPC image corresponding to the first lithography spatial feature map, and calculating the error between the contour image and the OPC image to determine whether the contour image set meets a first preset condition. When the first preset condition is met, the detection result of the process window corresponding to the OPC image is qualified. The present invention calculates the error of the current lithography process window and optimizes the current lithography process window to meet the requirements of actual production.

Description

Technical Field

[0001] The present invention relates to the technical field of semiconductor integrated circuit manufacturing processes, and in particular to a method and device for detecting a photolithography process window. Background Art

[0002] As the complexity of integrated circuits increases, the feature size becomes smaller and smaller. When the feature size of the integrated circuit approaches the system limit of the lithography machine exposure, that is, when the feature size is close to or smaller than the lithography light source, the layout manufactured on the silicon wafer will show obvious distortion. This phenomenon is called the optical proximity effect. In order to deal with the optical proximity effect, resolution enhancement technology has been proposed. Among them, optical proximity correction has become the most important technology. The so-called optical proximity correction (OPC) refers to the correction of the pattern on the photomask to compensate for the deformation of the lithography pattern caused by diffraction.

[0003] Furthermore, in the photolithography process of semiconductor manufacturing, it is necessary to set photolithography conditions to obtain the optimal pattern of components, circuits, etc. on the wafer. If the photolithography conditions are incorrect, the resulting pattern will differ from the ideal state, affecting the performance of the resulting components and circuits.

[0004] The resulting image errors primarily come from uneven lens focal planes, inaccurate wafer surface topography estimation, and residual wafer surface topography after dynamic scanning correction. As process nodes advance, error margins decrease. Therefore, it's necessary to check the lithography process window size of the post-OPC mask data to determine whether it meets actual production requirements.

[0005] Existing technologies reduce errors through measurement and calculation, but cannot completely eliminate errors or estimate the current error redundancy.

[0006] Therefore, the present invention proposes a method and device for detecting a photolithography process window, which are used to calculate the error of the current photolithography process window and optimize the current photolithography process window to meet the requirements of actual production. Summary of the Invention

[0007] The present invention provides a method and device for detecting a photolithography process window, which are used to calculate the error of the current photolithography process window and optimize the current photolithography process window to meet the requirements of actual production.

[0008] In a first aspect, the present invention provides a method for detecting a lithography process window, comprising: obtaining a first lithography space feature map and a first lithography condition set, wherein the first lithography condition set includes at least one lithography condition; inputting the first lithography space feature map into a trained neural network model, and inputting the lithography conditions in the first lithography condition set in sequence, to generate an image to be tested corresponding to the first lithography space feature map and the lithography conditions, extracting the lithography contour in the image to be tested to form a contour image, and aggregating all the obtained contour images into a contour image set; obtaining an OPC image corresponding to the first lithography space feature map, and calculating the error between each contour image in the contour image set and the OPC image, to determine whether the contour image set meets a first preset condition. When the contour image set meets the first preset condition, the detection result of the process window corresponding to the OPC image is qualified.

[0009] The beneficial effect is that the contour image set obtained by the trained neural network model can be considered to be the corresponding contour image set of the mask obtained under the standard process window. Therefore, by comparing the OPC image obtained in actual production with the images in the contour image set and calculating the error, it can be determined whether the current process window is standard or meets production requirements.

[0010] Optionally, the calculation of the error between each contour image in the contour image set and the OPC image to determine whether the contour image set meets a first preset condition includes: calculating the distance difference between all pixels of each contour image in the contour image set and the corresponding pixels in the OPC image, and obtaining the set of contour images whose distance differences are all less than a first preset threshold as the preferred image set; and determining whether the preferred image set meets the set requirements of the process window. When the preferred image set meets the set requirements, the contour image set meets the first preset condition. This has the beneficial effect that, in actual production needs, it is not necessary for all images in the contour image set to meet the requirements; as long as the preferred image set meets the set requirements of the process window, it is sufficient.

[0011] Further optionally, the detection method of the lithography process window further includes: obtaining a second lithography space feature atlas set and a second lithography condition set, the second lithography space feature atlas set including at least one lithography space feature map, and the second lithography condition set including at least one lithography condition; obtaining an SEM image of any one of the lithography space feature maps in the second lithography space feature map set under any one of the lithography conditions in the second lithography condition set, and aggregating all the obtained SEM images into an SEM image set, and dividing the SEM image set into a training set and a validation set in proportion; sequentially inputting the SEM images in the training set corresponding to the SEM images in the second lithography space feature map set. The lithography space feature map in the space feature map set and the SEM images in the training set correspond to the lithography conditions in the second lithography condition set, and an image training neural network model is generated with the corresponding SEM images in the training set as targets; and the neural network model is verified using the SEM images in the validation set, the SEM images in the validation set corresponding to the lithography space feature map in the second lithography space feature map set, and the SEM images in the validation set corresponding to the lithography conditions in the second lithography condition set, until the neural network model meets the verification conditions and the training of the neural network model is completed.

[0012] Further optionally, the process of satisfying the verification conditions of the neural network model includes: the difference between the pixel values ​​of the pixel points corresponding to the contour lines on the image in the test set and the SEM image in the verification set does not exceed a second preset threshold; the test set is a set of images generated by the neural network model based on the lithography space feature map in the second lithography space feature map set corresponding to the SEM image in the verification set, and the lithography conditions in the second lithography condition set corresponding to the SEM image in the verification set.

[0013] Further optionally, the neural network model includes: an input layer, a hidden layer and an output layer connected in sequence; the input layer sequentially obtains the SEM images in the verification set corresponding to the lithography space feature map in the second lithography space feature map set, and the SEM images in the verification set corresponding to the lithography conditions in the second lithography condition set; the hidden layer generates the image in the test set based on the SEM images in the verification set corresponding to the lithography space feature map in the second lithography space feature map set, and the SEM images in the verification set corresponding to the lithography conditions in the second lithography condition set; the output layer outputs the image in the test set.

[0014] Further optionally, the neural network model is a deep convolutional neural network model.

[0015] The second invention provides a detection device for a lithography process window, including: an acquisition module, a generation module, an extraction module and a judgment module; the acquisition module includes a first acquisition unit and a second acquisition unit, the first acquisition unit is used to acquire the first lithography space feature map, the second acquisition unit is used to acquire the first lithography condition set, the first lithography condition set includes at least one lithography condition; the generation module includes an input unit and a test image generation unit, the input unit is used to input the first lithography space feature map and the lithography conditions in the first lithography condition set in sequence into the neural network model after training, the test image generation unit is used to generate the first lithography space feature map corresponding to the lithography condition. The image to be tested; the extraction module includes a contour extraction unit and a summarizing unit, the contour extraction unit is used to extract the lithography contour in the image to be tested to form a contour image, and the summarizing unit is used to summarize all the obtained contour images into a contour image set; the judgment module includes an OPC image acquisition unit, a calculation unit and a detection unit, the OPC image acquisition unit is used to obtain the OPC image corresponding to the first lithography space feature map, the calculation unit is used to calculate the error between the contour image and the OPC image, and the detection unit is used to judge whether the contour image set meets a first preset condition. When the contour image set meets the first preset condition, the detection result of the process window corresponding to the OPC image is qualified.

[0016] Further optionally, the calculation unit is used to calculate the error between the contour image and the OPC image, and the detection unit is used to determine whether the contour image set meets a first preset condition, including: the calculation unit is used to calculate the distance difference between all pixel points of the contour image and the corresponding pixel points in the OPC image, and obtain the set of contour images whose distance differences are all less than a first preset threshold as the preferred image set; the detection unit is used to determine whether the preferred image set meets the set requirements of the process window, and when the preferred image set meets the set requirements, the contour image set meets the first preset condition.

[0017] Further optionally, the detection method of the lithography process window further includes: a third acquisition unit, a fourth acquisition unit, a fifth acquisition unit, a division unit, a training module, and a verification module; the third acquisition unit is used to acquire a second lithography space feature atlas set, the second lithography space feature atlas set includes at least one lithography space feature map, the fourth acquisition unit is used to acquire a second lithography condition set, the second lithography condition set includes at least one lithography condition; the fifth acquisition unit is used to acquire an SEM image of any one of the lithography space feature maps in the second lithography space feature map set under any one of the lithography conditions in the second lithography condition set, and the division unit is used to aggregate all the obtained SEM images into an SEM image set, and divide the SEM image set into training modules in proportion. set and a verification set; the training module is used to sequentially input the SEM images in the training set corresponding to the lithography space feature maps in the second lithography space feature map set, and the SEM images in the training set corresponding to the lithography conditions in the second lithography condition set, and generate an image training neural network model with the corresponding SEM images in the training set as the target; and the verification module is used to use the SEM images in the verification set, the SEM images in the verification set corresponding to the lithography space feature maps in the second lithography space feature map set, and the SEM images in the verification set corresponding to the lithography conditions in the second lithography condition set to verify the neural network model until the neural network model meets the verification conditions and the training of the neural network model is completed.

[0018] Further optionally, the verification module includes a difference calculation unit, which is used to calculate the difference between the pixel values ​​of the pixel points corresponding to the contour lines on the image in the test set and the SEM image in the verification set, and the verification module is used to determine that the neural network model meets the verification conditions when the difference does not exceed a second preset threshold; the test set is a set of images generated by the neural network model based on the lithography space feature map in the second lithography space feature map set corresponding to the SEM image in the verification set, and the lithography conditions in the second lithography condition set corresponding to the SEM image in the verification set.

[0019] The beneficial effects of the second aspect can refer to the description of the beneficial effects of the relevant content of the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 A flow chart of a method for detecting a photolithography process window provided by the present invention;

[0021] Figure 2 This is a schematic diagram of an embodiment of a detection device for a photolithography process window provided by the present invention. DETAILED DESCRIPTION

[0022] The technical solutions in the embodiments of the present application are described below in conjunction with the drawings in the embodiments of the present application. Among them, in the description of the embodiments of the present application, the terms used in the following embodiments are only for the purpose of describing specific embodiments, and are not intended to be used as limitations on the present application. As used in the specification and claims of the present application, the singular expressions "a", "the", "above", "the" and "this" are intended to also include expressions such as "one or more", unless there is a clear contrary indication in the context. It should also be understood that in the following embodiments of the present application, "at least one", "one or more" refer to one or more (including two). The term "and / or" is used to describe the association relationship of associated objects, indicating that three relationships can exist; for example, A and / or B can represent: the existence of A alone, the existence of A and B at the same time, and the existence of B alone, where A and B can be singular or plural. The character " / " generally indicates that the associated objects before and after are in an "or" relationship.

[0023] References to "one embodiment" or "some embodiments" etc. described in this specification mean that the specific features, structures or characteristics described in conjunction with the embodiment are included in one or more embodiments of the present application. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. appearing in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in another way. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in another way. The term "connected" includes direct and indirect connections, unless otherwise stated. "First" and "second" are used for descriptive purposes only and are not to be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated.

[0024] In the embodiments of this application, words such as "exemplarily" or "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "exemplarily" or "for example" in the embodiments of this application should not be interpreted as being preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplarily" or "for example" is intended to present the relevant concepts in a concrete manner.

[0025] In order to calculate the focus error of the current photolithography process window and optimize the current photolithography process window to meet the requirements of actual production, the present invention proposes a method and device for detecting the photolithography process window.

[0026] The embodiment of the present application provides a method for detecting a photolithography process window, the flow chart of which is as follows: Figure 1 As shown, the specific steps include:

[0027] S101 , obtaining a first lithography spatial characteristic map and a first lithography condition set, where the first lithography condition set includes at least one lithography condition.

[0028] In this step, the first lithography spatial characteristic map corresponds to the pattern on the mask required for the lithography process, and is used to characterize the spatial characteristics of the pattern on the corresponding mask. Optionally, the first lithography condition set includes a lithography condition group with one factor as a variable and the other factors as constants. Exemplarily, the lithography conditions include a focus value and an exposure dose. The lithography condition group uses the exposure dose as a fixed value and the focus value as a variable to generate a corresponding image to be tested. Optionally, the lithography condition group can also use the exposure dose as a variable and the focus value as a fixed value to generate a corresponding image to be tested.

[0029] S102, input the first lithography spatial feature map and the lithography conditions in the first lithography condition set in sequence into the trained neural network model to generate an image to be tested corresponding to the first lithography spatial feature map and the lithography conditions, extract the lithography contour in the image to be tested to form a contour image, and aggregate all the obtained contour images into a contour image set.

[0030] In this step, it is more preferred to input the first lithography spatial feature map into the trained neural network model, and input each lithography condition in the first lithography condition set in sequence to provide more images to be tested.

[0031] S103, obtaining the OPC image corresponding to the first lithography space feature map, and calculating the error between each contour image in the contour image set and the OPC image to determine whether the contour image set meets the first preset condition. When the contour image set meets the first preset condition, the detection result of the process window corresponding to the OPC image is qualified.

[0032] In this embodiment, the contour image set obtained by the trained neural network model can be considered to be the corresponding reticle contour image set obtained under a standard process window. Therefore, by comparing the OPC images obtained in actual production with the images in the contour image set and calculating the error, it can be determined whether the current process window is standard or meets production requirements.

[0033] In one possible embodiment, the error between each contour image in the contour image set and the OPC image is calculated to determine whether the contour image set meets the first preset condition, including: calculating the distance difference between all pixel points of each contour image in the contour image set and the corresponding pixel points in the OPC image, and obtaining the set of contour images whose distance differences are all less than a first preset threshold as the preferred image set; optionally, the distance difference is set according to actual process requirements. Determine whether the preferred image set meets the set requirements of the process window. When the preferred image set meets the set requirements, the contour image set meets the first preset condition. In this embodiment, in actual production requirements, it is not necessary for all images in the contour image set to meet the requirements, as long as the preferred image set meets the set requirements of the process window.

[0034] In another possible embodiment, the detection method of the lithography process window further includes: obtaining a second lithography spatial feature atlas set and a second lithography condition set, wherein the second lithography spatial feature atlas set includes at least one lithography spatial feature map, and the second lithography condition set includes at least one lithography condition; obtaining an SEM image of any one of the lithography spatial feature maps in the second lithography spatial feature map set under any one of the lithography conditions in the second lithography condition set. Optionally, an SEM image of each of the lithography spatial feature maps in the second lithography spatial feature map set under each of the lithography conditions in the second lithography condition set is obtained, and all the obtained SEM images are aggregated into an SEM image set. The SEM image set is divided into a training set and a validation set in proportion; optionally, the SEM image set includes at least 10 SEM images. Optionally, the ratio of the number of images in the training set and the validation set is 7:3. The SEM images in the training set corresponding to the lithography spatial feature maps in the second lithography spatial feature map set and the lithography conditions in the second lithography condition set corresponding to the SEM images in the training set are sequentially input, and an image training neural network model is generated with the corresponding SEM images in the training set as targets; and the neural network model is verified using the SEM images in the validation set, the lithography spatial feature maps in the second lithography spatial feature map set corresponding to the SEM images in the validation set, and the lithography conditions in the second lithography condition set corresponding to the SEM images in the validation set, until the neural network model meets the verification conditions and the training of the neural network model is completed. Optionally, if the neural network model does not meet the verification conditions, the number of images in the second lithography spatial feature map set and the lithography conditions in the second lithography condition set can be appropriately increased, or the images in the training set and the validation set can be exchanged.

[0035] In another possible embodiment, the neural network model satisfies the verification condition, including: the difference between the pixel values ​​of the pixel points corresponding to the contour lines on the image in the test set and the SEM image in the verification set does not exceed a second preset threshold; the test set is a set of images generated by the neural network model based on the lithography spatial feature map in the second lithography spatial feature map set corresponding to the SEM image in the verification set, and the lithography conditions in the second lithography condition set corresponding to the SEM image in the verification set. Because pixel values ​​can reflect the average brightness information of a small square on an image, these images can be compared by comparing the pixel values ​​between corresponding pixel points on different images, and by establishing a coordinate system and calculating the difference between the pixel values ​​of the partial images that can represent the image information, the training efficiency of the neural network model can be improved. 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 image obtained by the scan can be saved. The image size depends on the specific situation. In this example, it is 512*512 pixels. Because the coordinates are fixed, the position of each pixel is also fixed. By selecting the pixel values ​​of the pixels at the coordinate positions within the appropriate range for comparison, the comparison results can be obtained more efficiently. Optionally, the mean square error between the pixel values ​​of the pixels corresponding to the contour lines on the image in the test set and the SEM image in the verification set does not exceed 10^-4. And for different process nodes, the preset conditions can be different.

[0036] In one possible embodiment, the neural network model includes: an input layer, a hidden layer, and an output layer connected in sequence; the input layer sequentially obtains the lithography spatial feature maps corresponding to the SEM images in the validation set in the second lithography spatial feature map set, and the lithography conditions corresponding to the SEM images in the validation set in the second lithography condition set; the hidden layer generates images in the test set based on the lithography spatial feature maps corresponding to the SEM images in the validation set in the second lithography spatial feature map set, and the lithography conditions corresponding to the SEM images in the validation set in the second lithography condition set; the output layer outputs the images in the test set. The hidden layer includes several convolution-pooling-activation layers.

[0037] In another possible embodiment, the neural network model is a deep convolutional neural network model. In this step, the deep convolutional neural network model does not need to be based on the entire contour information of the object to be imaged, but only needs to be based on a partial feature map for training and recognition, thereby improving work efficiency.

[0038] In order to implement the detection method of the photolithography process window provided in any of the above embodiments, the present invention provides a detection device for the photolithography process window, such as Figure 2 As shown, the detection device includes: an acquisition module 201, a generation module 202, an extraction module 203 and a judgment module 204.

[0039] The acquisition module 201 includes a first acquisition unit 2011 and a second acquisition unit 2012, wherein the first acquisition unit 2011 is used to acquire the first lithography space feature map, and the second acquisition unit 2012 is used to acquire the first lithography condition set, wherein the first lithography condition set includes at least one lithography condition; the generation module 202 includes an input unit 2021 and a test image generation unit 2022, wherein the input unit 2021 is used to input the first lithography space feature map and the lithography conditions in the first lithography condition set in sequence into the trained neural network model, and the test image generation unit 2022 is used to generate a test image corresponding to the first lithography space feature map and the lithography conditions in the lithography condition set; the extraction module 203 includes a contour extraction unit 2031 and a collection unit 2032. The total unit 2032, the contour extraction unit 2031 is used to extract the lithography contour in the image to be tested to form a contour image, and the summarization unit 2032 is used to summarize all the obtained contour images into a contour image set; the judgment module 204 includes an OPC image acquisition unit 2041, a calculation unit 2042 and a detection unit 2043, the OPC image acquisition unit 2041 is used to obtain the OPC image corresponding to the first lithography spatial feature map, the calculation unit 2042 is used to calculate the error between the contour image and the OPC image, and the detection unit 2043 is used to determine whether the contour image set meets a first preset condition. When the contour image set meets the first preset condition, the detection result of the process window corresponding to the OPC image is qualified. Optionally, the calculation unit 2041 is used to respectively calculate the error between each contour image in the contour image set and the OPC image.

[0040] In a possible embodiment, the calculation unit is used to calculate the error between the contour image and the OPC image, and the detection unit is used to determine whether the contour image set meets the first preset condition, including: the calculation unit is used to calculate the distance difference between all pixel points of the contour image and the corresponding pixel points in the OPC image, and obtain the set of contour images whose distance differences are all less than a first preset threshold as the preferred image set. Optionally, the distance difference is set according to actual process requirements; the detection unit is used to determine whether the preferred image set meets the setting requirements of the process window. When the preferred image set meets the setting requirements, the contour image set meets the first preset condition.

[0041] In another possible embodiment, the detection method of the lithography process window further includes: a third acquisition unit, a fourth acquisition unit, a fifth acquisition unit, a division unit, a training module, and a verification module; the third acquisition unit is used to acquire a second lithography space feature atlas set, the second lithography space feature atlas set includes at least one lithography space feature map, the fourth acquisition unit is used to acquire a second lithography condition set, the second lithography condition set includes at least one lithography condition; the fifth acquisition unit is used to acquire an SEM image of any one of the lithography space feature maps in the second lithography space feature map set under any one of the lithography conditions in the second lithography condition set, and the division unit is used to aggregate all the obtained SEM images into an SEM image set, and divide the SEM image set into A training set and a validation set; the training module is used to sequentially input the SEM images in the training set corresponding to the lithography space feature maps in the second lithography space feature map set, and the SEM images in the training set corresponding to the lithography conditions in the second lithography condition set, and generate an image training neural network model with the corresponding SEM images in the training set as the target; and the validation module is used to verify the neural network model using the SEM images in the validation set, the SEM images in the validation set corresponding to the lithography space feature maps in the second lithography space feature map set, and the SEM images in the validation set corresponding to the lithography conditions in the second lithography condition set, until the neural network model meets the validation conditions and the training of the neural network model is completed.

[0042] In a possible embodiment, the verification module includes a difference calculation unit, which is used to calculate the difference between the pixel values ​​of the pixel points corresponding to the contour lines on the image in the test set and the SEM image in the verification set, and the verification module is used to determine that the neural network model meets the verification conditions when the difference does not exceed a second preset threshold; the test set is a set of images generated by the neural network model based on the lithography space feature map in the second lithography space feature map set corresponding to the SEM image in the verification set, and the lithography conditions in the second lithography condition set corresponding to the SEM image in the verification set.

[0043] In this step, because the pixel value can reflect the average brightness information of a small square on the image, the images can be compared by the pixel values ​​between the corresponding pixel points on different images, and the difference between the pixel values ​​of the partial images that can represent the image information can be calculated by establishing a coordinate system, which can improve the training efficiency of the neural network model. In this embodiment, a scanning electron microscope can be used to perform a horizontal scan at different coordinates of the wafer after lithography, and the SEM image obtained by the scan can be saved. The image size depends on the specific situation. In this example, it is 512*512 pixels. Because the coordinates are fixed, the position of each pixel is also fixed. By selecting the pixel values ​​of the pixels at the coordinate positions within the appropriate range for comparison, the comparison results can be obtained more efficiently. Optionally, the mean square error between the pixel values ​​of the pixels corresponding to the contour lines on the image in the test set and the SEM image in the verification set does not exceed 10^-4. And for different process nodes, the preset conditions can be different.

[0044] The above description is merely a specific implementation of the embodiments of the present application, but the scope of protection of the embodiments of the present application is not limited thereto. Any changes or substitutions within the technical scope disclosed in the embodiments of the present application shall be included in the scope of protection of the embodiments of the present application. Therefore, the scope of protection of the embodiments of the present application shall be based on the scope of protection of the claims.

Claims

1. A method for detecting a photolithography process window, characterized in that: include: Obtaining a first lithography spatial feature map and a first lithography condition set, wherein the first lithography condition set includes at least one lithography condition; inputting the first lithography spatial feature map and the lithography conditions in the first lithography condition set into a trained neural network model to generate an image to be tested corresponding to the first lithography spatial feature map and the lithography condition; extracting a lithography profile from the image to be tested to form a profile image; and aggregating all the obtained profile images into a profile image set; Obtaining an OPC image corresponding to the first lithography spatial feature map, and calculating an error between each contour image in the contour image set and the OPC image to determine whether the contour image set meets a first preset condition, and when the contour image set meets the first preset condition, the detection result of the process window corresponding to the OPC image is qualified; Calculating the error between each contour image in the contour image set and the OPC image to determine whether the contour image set meets a first preset condition includes: Calculating distance differences between all pixel points of each contour image in the contour image set and corresponding pixel points in the OPC image, and obtaining a set of contour images whose distance differences are all less than a first preset threshold as a preferred image set; It is determined whether the preferred image set meets the set requirements of the process window. When the preferred image set meets the set requirements, the contour image set meets the first preset condition.

2. The method for detecting a photolithography process window according to claim 1, wherein: Also includes: Obtaining a second lithography spatial feature atlas set and a second lithography condition set, where the second lithography spatial feature atlas set includes at least one lithography spatial feature map, and the second lithography condition set includes at least one lithography condition; obtaining an SEM image of any one of the lithography spatial feature maps in the second lithography spatial feature map set under any one of the lithography conditions in the second lithography condition set, aggregating all the obtained SEM images into an SEM image set, and dividing the SEM image set into a training set and a validation set in proportion; sequentially inputting the lithography spatial feature maps in the second lithography spatial feature map set corresponding to the SEM images in the training set, and the lithography conditions in the second lithography condition set corresponding to the SEM images in the training set, and generating an image training neural network model with the corresponding SEM images in the training set as targets; And using the SEM images in the verification set, the lithography space feature maps in the second lithography space feature map set corresponding to the SEM images in the verification set, and the lithography conditions in the second lithography condition set corresponding to the SEM images in the verification set, the neural network model is verified until the neural network model meets the verification conditions and the training of the neural network model is completed.

3. The method for detecting a photolithography process window according to claim 2, wherein: The step of: until the neural network model satisfies the verification condition includes: The difference between the pixel values ​​of the pixels corresponding to the contour lines on the image in the test set and the SEM image in the verification set does not exceed a second preset threshold; The test set is a set of images generated by the neural network model based on the lithography space feature maps in the second lithography space feature map set corresponding to the SEM images in the verification set, and the lithography conditions in the second lithography condition set corresponding to the SEM images in the verification set.

4. The method for detecting a photolithography process window according to claim 3, wherein: The neural network model includes: an input layer, a hidden layer and an output layer connected in sequence; The input layer sequentially obtains the lithography spatial feature maps in the second lithography spatial feature map set corresponding to the SEM images in the verification set, and the lithography conditions in the second lithography condition set corresponding to the SEM images in the verification set; The hidden layer generates the image in the test set according to the lithography spatial feature map in the second lithography spatial feature map set corresponding to the SEM image in the validation set, and the lithography conditions in the second lithography condition set corresponding to the SEM image in the validation set; The output layer outputs the images in the test set.

5. The method for detecting a photolithography process window according to claim 4, wherein: The neural network model is a deep convolutional neural network model.

6. A detection device for a photolithography process window, characterized in that: include: Acquisition module, generation module, extraction module and judgment module; The acquisition module includes a first acquisition unit and a second acquisition unit, the first acquisition unit is used to acquire a first lithography spatial characteristic map, and the second acquisition unit is used to acquire the first lithography condition set, wherein the first lithography condition set includes at least one lithography condition; The generation module includes an input unit and a test image generation unit, wherein the input unit is used to input the first lithography spatial feature map and the lithography conditions in the first lithography condition set in sequence into the trained neural network model, and the test image generation unit is used to generate a test image corresponding to the first lithography spatial feature map and the lithography conditions; The extraction module includes a contour extraction unit and a summarization unit, wherein the contour extraction unit is used to extract the photolithography contour in the image to be measured to form a contour image, and the summarization unit is used to summarize all the obtained contour images into a contour image set; The judgment module includes an OPC image acquisition unit, a calculation unit, and a detection unit. The OPC image acquisition unit is used to acquire an OPC image corresponding to the first lithography space feature map. The calculation unit is used to calculate the error between the contour image and the OPC image. The detection unit is used to determine whether the contour image set meets a first preset condition. When the contour image set meets the first preset condition, the detection result of the process window corresponding to the OPC image is qualified. The calculation unit is used to calculate the error between the contour image and the OPC image, and the detection unit is used to determine whether the contour image set meets a first preset condition, including: The calculation unit is used to calculate the distance differences between all pixel points of the contour image and corresponding pixel points in the OPC image, and obtain a set of contour images whose distance differences are all less than a first preset threshold as a preferred image set; The detection unit is used to determine whether the preferred image set meets the set requirements of the process window. When the preferred image set meets the set requirements, the contour image set meets the first preset condition.

7. The detection device for the photolithography process window according to claim 6, characterized in that: Also includes: A third acquisition unit, a fourth acquisition unit, a fifth acquisition unit, a division unit, a training module, and a verification module; The third acquisition unit is used to acquire a second lithography spatial feature atlas set, the second lithography spatial feature atlas set including at least one lithography spatial feature map; the fourth acquisition unit is used to acquire a second lithography condition set, the second lithography condition set including at least one lithography condition; the fifth acquisition unit is used to acquire an SEM image of any one of the lithography spatial feature maps in the second lithography spatial feature map set under any one of the lithography conditions in the second lithography condition set; and the division unit is used to aggregate all the obtained SEM images into an SEM image set, and divide the SEM image set into a training set and a validation set in proportion; The training module is configured to sequentially input the lithography spatial feature maps in the second lithography spatial feature map set corresponding to the SEM images in the training set, and the lithography conditions in the second lithography condition set corresponding to the SEM images in the training set, and generate an image training neural network model using the corresponding SEM images in the training set as targets; The verification module is used to verify the neural network model using the SEM images in the verification set, the lithography space feature maps in the second lithography space feature map set corresponding to the SEM images in the verification set, and the lithography conditions in the second lithography condition set corresponding to the SEM images in the verification set, until the neural network model meets the verification conditions and the training of the neural network model is completed.

8. The detection device for the photolithography process window according to claim 7, characterized in that: The verification module includes a difference calculation unit, which is used to calculate the difference between the pixel values ​​of the pixel points corresponding to the contour lines on the image in the test set and the SEM image in the verification set, and the verification module is used to determine that the neural network model meets the verification condition when the difference does not exceed a second preset threshold; The test set is a set of images generated by the neural network model based on the lithography space feature maps in the second lithography space feature map set corresponding to the SEM images in the verification set, and the lithography conditions in the second lithography condition set corresponding to the SEM images in the verification set.

Images