Super-Resolution Lithography Optical Proximity Effect Correction Method Based on Unet Deep Convolutional Network

Abstract

The present disclosure provides a super-resolution lithography optical proximity effect correction method based on a Unet deep convolutional network, including: performing optical proximity effect correction on a designed pattern to obtain a known data pair composed of designed pattern data and corresponding optimized mask pattern data, and splitting it into a training set and a validation set; inputting the designed pattern data in the training set into the Unet network in batches to obtain predicted pattern data related to training; calculating a training loss value and updating the weight parameters of the Unet network; inputting the designed pattern data in the validation set into the updated Unet network to obtain predicted pattern data related to validation; calculating the training loss value and the IOU value; determining whether the training loss value and the IOU value meet corresponding preset conditions or whether the current number of training times reaches a preset number of training times; if so, using the current Unet network with updated weight parameters as the trained Unet network; inputting the designed pattern to be optimized into the trained Unet network to obtain predicted corrected mask pattern data.

Description

Technical Field

[0001] The present disclosure relates to the field of integrated circuit technologies, and particularly to a super-resolution lithography optical proximity effect correction method, system, electronic device, computer-readable storage medium, and program product based on a Unet deep convolutional network. Background Art

[0002] In the integrated circuit projection lithography process, the pattern on the mask is reduced and projected onto the photoresist through the exposure system. When the pattern size in the light-transmitting area of the mask is very small, most of the energy of the light wave passing through the mask will be concentrated in the high-frequency region. However, due to the low-pass filtering effect formed by the diffraction limitation of the exposure system, a lot of high-frequency information is filtered out, resulting in a blurred image pattern on the photoresist. The phenomenon of image result distortion caused by the interference and diffraction effects of light waves is called the optical proximity effect (OPE), and this effect will affect the chip manufacturing yield. Optical proximity correction (OPC) aims to compensate for the influence brought by the optical proximity effect by modifying the pattern on the mask.

[0003] As the lithography technology node becomes smaller and smaller, modern projection lithography technology faces the disadvantages of a complex process system and a sharp increase in cost. In super-resolution lithography technology, the evanescent wave participates in imaging, which has the advantages of high throughput, high resolution, high aspect ratio, high fidelity, and one-step exposure. It is a potential technology to replace the complex and expensive modern projection lithography technology. In super-resolution lithography technology, the rule-based optical proximity correction method (MB-OPC) has a limited scope of application, while the model-based optical proximity correction method (RB-OPC) and inverse lithography technology (ILT) have the drawback of large computational complexity and cannot be applied to large-scale lithography design layouts.

[0004] Therefore, in super-resolution lithography technology, there is an urgent need for a highly efficient, fast, and widely applicable optical proximity effect correction method for correcting the mask pattern to correct the distortion in the lithography imaging result and improve the yield of integrated circuit products. Summary of the Invention

[0005] (I) Technical Problems to be Solved

[0006] In view of the above problems, the present disclosure provides a super-resolution lithography optical proximity effect correction method, system, electronic device, computer-readable storage medium, and program product based on a Unet deep convolutional network, which are used to solve technical problems such as the difficulty of the traditional method in achieving efficient, fast, and wide-applicability optical proximity effect correction.

[0007] (II) Technical Solutions

[0008] The first aspect of the present disclosure provides a super-resolution lithography optical proximity effect correction method based on a Unet deep convolutional network, including: S1, performing optical proximity effect correction on the design pattern and performing pixelization processing to obtain a known data pair composed of design pattern data and corresponding optimized mask pattern data; splitting the known data pair into a training set and a validation set; S2, inputting the design pattern data in the training set into the Unet deep convolutional network in batches to obtain prediction pattern data related to training; S3, calculating a first training loss value for each batch according to the prediction pattern data related to training and the corresponding optimized mask pattern data, and updating the weight parameters of the Unet deep convolutional network according to the first training loss value; S4, inputting the design pattern data in the validation set into the updated Unet deep convolutional network to obtain prediction pattern data related to validation; S5, calculating a second training loss value and an intersection over union value according to the prediction pattern data related to validation and the corresponding optimized mask pattern data; S6, determining whether the second training loss value and the intersection over union value meet the corresponding preset conditions or whether the current number of training times reaches the preset number of training times; if not, repeating S2-S6 for training; if so, the Unet deep convolutional network with the currently updated weight parameters is the trained Unet deep convolutional network; S7, inputting the design pattern to be optimized into the trained Unet deep convolutional network to obtain predicted corrected mask pattern data, and completing the super-resolution lithography optical proximity effect correction.

[0009] According to an embodiment of the present disclosure, the optical proximity effect correction of the design pattern in S1 includes: performing optical proximity effect correction on the design pattern to obtain an optimized mask pattern; performing pixelization processing on the design pattern and the optimized mask pattern respectively to obtain design pattern data and corresponding optimized mask pattern data, and the design pattern data and the corresponding optimized mask pattern data form a known data pair; wherein, the design pattern to be optimized in S7 is the whole or most of the design patterns of the large-size mask layout to be optimized, and the design pattern in S1 is a small part of the design patterns of the large-size mask layout to be optimized.

[0010] According to an embodiment of the present disclosure, the method for performing optical proximity effect correction in S1 includes any one of a rule-based optical proximity effect correction method, a model-based optical proximity effect correction method, and inverse lithography technology.

[0011] According to an embodiment of the present disclosure, after S1, it further includes: performing data augmentation processing on known data pairs; wherein, the data augmentation processing includes one or more of vertical flipping, horizontal flipping, and rotation operations.

[0012] According to an embodiment of the present disclosure, before S2, it further includes: S20, constructing a Unet deep convolutional network, and the Unet deep convolutional network includes an encoding end, a decoding end, and a prediction network; wherein, the encoding end and the decoding end do not contain fully connected layers.

[0013] According to an embodiment of the present disclosure, S3 includes: for each batch, calculating a first training loss value according to the prediction graphic data related to training and the corresponding optimized mask graphic data through the following formula :

[0014]

[0015] wherein, is the prediction graphic data related to training, is the corresponding optimized mask graphic data, MSE represents the mean square error, and n is the value of the batch size; backpropagating the first training loss value to obtain the gradient of the weight parameters of the Unet deep convolutional network; updating the weight parameters of the Unet deep convolutional network according to the gradient.

[0016] According to an embodiment of the present disclosure, S5 includes: calculating the intersection over union value IoU through the following formula:

[0017]

[0018] wherein, k + 1 represents the number of matrix element value classifications; represents the number of matrix elements that should have been classified as the i value but were predicted as j, represents the number of matrix elements that should have been classified as the i value and were predicted as i, represents the number of matrix elements that should have been classified as the j value but were predicted as i.

[0019] According to an embodiment of the present disclosure, the corresponding preset conditions in S6 include: the second training loss value is less than or equal to the first preset threshold and the intersection over union value is greater than or equal to the second preset threshold.

[0020] The second aspect of the present disclosure provides a super-resolution lithography method, including: obtaining predicted corrected mask pattern data by using the super-resolution lithography optical proximity effect correction method based on the Unet deep convolutional network as described above, outputting the predicted corrected mask pattern, and performing super-resolution lithography according to the predicted corrected mask pattern. The third aspect of the present disclosure provides a super-resolution lithography optical proximity effect correction system based on the Unet deep convolutional network, including: an optical proximity effect correction module for performing optical proximity effect correction on a design pattern and performing pixelization processing to obtain a known data pair composed of design pattern data and corresponding optimized mask pattern data; splitting the known data pair into a training set and a validation set; a training set processing module for batch-inputting the design pattern data in the training set into the Unet deep convolutional network to obtain training-related predicted pattern data; a first calculation module for calculating a first training loss value for each batch according to the training-related predicted pattern data and the corresponding optimized mask pattern data, and updating the weight parameters of the Unet deep convolutional network according to the first training loss value; a validation set processing module for inputting the design pattern data in the validation set into the updated Unet deep convolutional network to obtain validation-related predicted pattern data; a second calculation module for calculating a second training loss value and a union intersection value according to the validation-related predicted pattern data and the corresponding optimized mask pattern data; a judgment module for judging whether the second training loss value and the union intersection value meet corresponding preset conditions or whether the current training times reach the preset training times; if not, repeating the training; if so, the Unet deep convolutional network with the current updated weight parameters is the trained Unet deep convolutional network; a prediction module for inputting the design pattern to be optimized into the trained Unet deep convolutional network to obtain predicted corrected mask pattern data, and completing the super-resolution lithography optical proximity effect correction.

[0021] The fourth aspect of the present disclosure provides an electronic device, including: a processor; a memory storing a computer-executable program, which when executed by the processor causes the processor to execute the super-resolution lithography optical proximity effect correction method based on the Unet deep convolutional network as described above.

[0022] The fifth aspect of the present disclosure provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the super-resolution lithography optical proximity effect correction method based on the Unet deep convolutional network as described above.

[0023] The sixth aspect of the present disclosure provides a computer program product, including a computer program, which when executed by a processor, implements the super-resolution lithography optical proximity effect correction method based on the Unet deep convolutional network as described above.

[0024] (III) Beneficial effects

[0025] A super-resolution lithography optical proximity effect correction method, system, electronic device, computer-readable storage medium, and program product provided by the present disclosure train a Unet deep convolutional network to accurately fit the mapping relationship between the design pattern and the optimized mask pattern. The trained Unet deep convolutional network can efficiently perform optical proximity effect correction prediction on the input lithography layout pattern to obtain the corrected mask pattern. This method can significantly improve the correction efficiency while maintaining high accuracy, and it can correct large-size layout patterns without being limited by the size of the design pattern. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] To more fully understand the present disclosure and its advantages, reference will now be made to the following description taken in conjunction with the accompanying drawings, in which:

[0027] Figure 1 Schematically shows a flowchart of a super-resolution lithography optical proximity effect correction method based on a Unet deep convolutional network according to an embodiment of the present disclosure;

[0028] Figure 2 Schematically shows a schematic diagram of data augmentation results of a design pattern data in a training set according to an embodiment of the present disclosure;

[0029] Figure 3 Schematically shows a schematic diagram of a super-resolution lithography structure used according to an embodiment of the present disclosure;

[0030] Figure 4 Schematically shows a schematic diagram of a Unet deep convolutional network structure according to an embodiment of the present disclosure;

[0031] Figure 5 Schematically shows a schematic diagram of a Resnet network structure in the encoding end of a Unet deep convolutional network according to an embodiment of the present disclosure;

[0032] Figure 6 Schematically shows a schematic diagram of a residual structure in a Unet deep convolutional network according to an embodiment of the present disclosure;

[0033] Figure 7 Schematically shows a schematic diagram of a sub-convolutional kernel structure in a Resnet network structure according to an embodiment of the present disclosure;

[0034] Figure 8 Schematically shows a schematic diagram of the results before and after prediction correction of one of the design pattern data in a validation set under the action of a Unet deep convolutional network according to an embodiment of the present disclosure;

[0035] Figure 9Schematically shows the schematic diagrams of the results before and after the prediction correction of the design pattern outside the known data pair according to the embodiments of the present disclosure;

[0036] Figure 10 Schematically shows the schematic diagrams of the results before and after the prediction correction of the large-size layout pattern according to the embodiments of the present disclosure;

[0037] Figure 11 Schematically shows the IoU change curve and the loss function curve according to the embodiments of the present disclosure;

[0038] Figure 12 Schematically shows the block diagram of the super-resolution lithography optical proximity effect correction system based on the Unet deep convolutional network according to the embodiments of the present disclosure; and

[0039] Figure 13 Schematically shows the block diagram of the electronic device suitable for implementing the method described above according to the embodiments of the present disclosure. Detailed implementation manners

[0040] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present disclosure. In the following detailed description, for the sake of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, obviously, one or more embodiments can also be implemented without these specific details. In addition, in the following description, descriptions of well-known structures and technologies are omitted to avoid unnecessarily confusing the concepts of the present disclosure.

[0041] The terms used herein are merely for describing specific embodiments and are not intended to limit the present disclosure. The terms "including", "comprising", etc. used herein indicate the presence of the described features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0042] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0043] Some block diagrams and / or flowcharts are shown in the accompanying drawings. It should be understood that some blocks in the block diagrams and / or flowcharts, or combinations thereof, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when executed by the processor, these instructions can create a device for implementing the functions / operations illustrated in these block diagrams and / or flowcharts. The technology of the present disclosure can be implemented in the form of hardware and / or software (including firmware, microcode, etc.). Additionally, the technology of the present disclosure can take the form of a computer program product on a computer-readable storage medium storing instructions, which can be used by or in conjunction with an instruction execution system.

[0044] The UNet deep convolutional network is an image semantic segmentation network proposed by foreign researchers in 2015. This network adopts the encoder-decoder idea and has a relatively simple structure, which can be applied to the training of various datasets. In recent years, researchers at home and abroad have achieved many research results using the UNet network in image classification, image detection, etc.

[0045] The optical proximity effect correction method based on the Unet deep convolutional network is an efficient and widely applicable mask pattern correction method. This method trains the Unet deep convolutional network to learn the mapping rule from the designed pattern to the optimized mask pattern, and quickly and efficiently generates the optimized mask pattern. Through this correction method, the lithographic imaging pattern can be made as close as possible to the designed pattern, improving the fidelity of the lithography result of the near-field super-resolution lithography system.

[0046] In the present disclosure, for the sake of convenience, only the designed pattern, the optimized mask pattern, and the predicted corrected mask pattern are referred to as patterns, while the results obtained from the calculation process and the imaging process in the optical proximity effect correction are all referred to as data. It can be understood that the data in the process can all correspondingly output the corresponding patterns.

[0047] The present disclosure provides a super-resolution lithography optical proximity effect correction method based on the Unet deep convolutional network, as Figure 1As shown in the figure, it includes: S1, performing optical proximity effect correction on the design pattern and pixelizing it to obtain a known data pair composed of the design pattern data and the corresponding optimized mask pattern data; splitting the known data pair into a training set and a validation set; S2, inputting the design pattern data in the training set into the Unet deep convolutional network in batches to obtain prediction pattern data related to training; S3, for each batch, calculating a first training loss value based on the prediction pattern data related to training and the corresponding optimized mask pattern data, and updating the weight parameters of the Unet deep convolutional network according to the first training loss value; S4, inputting the design pattern data in the validation set into the updated Unet deep convolutional network to obtain prediction pattern data related to validation; S5, calculating a second training loss value and an intersection over union value based on the prediction pattern data related to validation and the corresponding optimized mask pattern data; S6, determining whether the second training loss value and the intersection over union value meet the corresponding preset conditions or whether the current number of training times reaches the preset number of training times; if not, repeating S2 - S6 for training; if so, taking the Unet deep convolutional network with the current updated weight parameters as the trained Unet deep convolutional network; S7, inputting the design pattern to be optimized into the trained Unet deep convolutional network to obtain the predicted and corrected mask pattern data, and completing the super-resolution lithography optical proximity effect correction.

[0048] In this method, first, optical proximity effect correction needs to be performed on the design pattern to obtain the corresponding optimized mask pattern. After converting the design pattern and the optimized mask pattern into matrix data, a known data pair is formed, and the known data pair is split into a training set and a validation set. In each round of training, the design pattern data in the training set is input into the built Unet deep convolutional network to obtain prediction pattern data related to training. The first training loss value is calculated using the prediction pattern data related to training and the corresponding optimized mask pattern data in the training set, and the weight parameters of the Unet deep convolutional network are updated according to the first training loss value. The design pattern data in the validation set is input into the Unet network with updated weight parameters to obtain prediction pattern data related to validation. The second training loss value and the intersection over union value are calculated using the optimized mask pattern data related to validation and the corresponding prediction pattern data in the validation set. After multiple rounds of training, when the second training loss value is less than or equal to the first preset threshold and the intersection over union value is greater than or equal to the second preset threshold or reaches the predetermined number of training rounds, the training ends, and the weight parameters of the Unet deep convolutional network are saved to obtain the trained Unet deep convolutional network. Then, the layout pattern other than the design pattern is converted into layout pattern data and input into the trained Unet deep convolutional network to obtain the predicted and corrected mask pattern data, and the predicted and corrected mask pattern data is converted into the predicted and corrected mask pattern.

[0049] The method of the present disclosure trains a Unet deep convolutional network in a supervised mode to find the mapping rule from the design pattern to the optimized mask pattern. According to this rule, the correction efficiency can be significantly improved while maintaining a high accuracy. In addition, this method can be implemented for correcting large-size layout patterns without being limited by the size of the design pattern. It should be noted that a large-size layout pattern refers to a layout pattern in which both dimensions of the layout pattern data are larger than those of the design pattern data. For example, if the dimension of the design pattern data is (210, 210), the dimension of the large-size layout pattern data is (1024, 1024).

[0050] Based on the above embodiments, the optical proximity effect correction of the design pattern in S1 includes: obtaining an optimized mask pattern by performing optical proximity effect correction on the design pattern; respectively performing pixelization processing on the design pattern and the optimized mask pattern to obtain design pattern data and corresponding optimized mask pattern data, and the design pattern data and the corresponding optimized mask pattern data form a known data pair; wherein, the design pattern to be optimized in S7 is the whole or most of the design patterns of the large-size mask layout to be optimized, and the design pattern in S1 is a small part of the design patterns of the large-size mask layout to be optimized.

[0051] Select some patterns from the lithography design layout to be optimized as the design patterns, perform optical proximity effect correction on these design patterns, and obtain an optimized mask pattern after correction. Perform pixelization processing on both the design pattern and the optimized mask pattern to obtain design pattern data and optimized mask pattern data respectively. The design pattern data and the optimized mask pattern data are paired to form a known data pair, and the known data pair is split into a training set and a validation set according to a certain quantity ratio, where the quantity ratio refers to the ratio of the number of data pairs in the training set to the number of data pairs in the validation set. For example, the quantity ratio is 7:3, that is, the ratio of the number of known data pairs in the training set to the number of known data pairs in the validation set is 7:3. This quantity ratio can also be determined according to the actual situation. Usually, the number of known data pairs in the training set should be more than the number of known data pairs in the validation set.

[0052] Based on the above embodiments, the method for performing optical proximity effect correction in S1 includes any one of a rule-based optical proximity effect correction method, a model-based optical proximity effect correction method, and inverse lithography technology.

[0053] Obtaining the corresponding optimized mask pattern through the above optical proximity effect correction method is beneficial for the subsequent Unet deep convolutional network to learn the mapping rule from the design pattern to the optimized mask pattern.

[0054] Based on the above embodiments, after S1, it further includes: performing data augmentation processing on the known data pair; wherein, the data augmentation processing includes one or more of vertical flipping, horizontal flipping, and rotation operations.

[0055] Perform data augmentation on known data pairs, which can increase the number of samples and the diversity of sample data, thereby enhancing the generalization ability of the Unet deep convolutional network, enabling the Unet deep convolutional network to better find the mapping rule from the designed graphic data to the optimized mask graphic data.

[0056] The vertical flipping operation can be expressed as

[0057] The horizontal flipping operation can be expressed as

[0058] The rotation operation can be expressed as

[0059] where is the known data pair, N is the dimension size of the mask matrix, is the rotation angle.

[0060] Based on the above embodiments, before S2, it further includes: S20, constructing a Unet deep convolutional network, which includes an encoding end, a decoding end, and a prediction network; among them, the encoding end and the decoding end do not contain fully connected layers.

[0061] Initialize the weight parameters of the encoding end, the decoding end, and the prediction network. The weight parameters of the encoding end, the decoding end, and the prediction network together constitute the weight parameters of the Unet deep convolutional network, where the weight parameters refer to all variables in the network that can be updated according to the gradient. Different from the conventional Unet deep convolutional network, in order to be applicable to input data of different dimensions, the encoding end and the decoding end of the Unet deep convolutional network constructed here do not contain fully connected layers. Divide the designed graphic data and the optimized mask graphic data in the training set into multiple batches. For each batch, input the designed graphic data in the training set into the Unet deep convolutional network, which passes through the encoding end and the decoding end of the Unet deep convolutional network in sequence, and finally passes through the prediction network to output the prediction graphic data related to training.

[0062] Based on the above embodiments, S3 includes: for each batch, calculate the first training loss value according to the prediction graphic data related to training and the corresponding optimized mask graphic data through the following formula :

[0063]

[0064] where is the prediction graphic data related to training, is the corresponding optimized mask graphic data, MSE represents the mean square error, and n is the value of the batch size; backpropagate the first training loss value to obtain the gradient of the weight parameters of the Unet deep convolutional network; update the weight parameters of the Unet deep convolutional network according to the gradient.

[0065] For each batch, the predicted graphic data related to its training and the optimized mask graphic data corresponding to this batch perform the above mean square error (MSE) operation to obtain the first training loss value; backpropagate the obtained first training loss value, solve to obtain the gradient of the weight parameters of the Unet deep convolutional network, and use the gradient descent method to update the weight parameters of the Unet deep convolutional network.

[0066] Based on the above embodiments, S5 includes: calculating the intersection over union value IoU through the following formula:

[0067]

[0068] where k + 1 represents the number of matrix element value classifications; represents the number of matrix elements that should have been classified as value i but were predicted as j, represents the number of matrix elements that should have been classified as value i and were predicted as i, represents the number of matrix elements that should have been classified as value j but were predicted as i.

[0069] Input the design graphic data in the validation set into the Unet deep convolutional network with updated weight parameters to obtain the predicted graphic data related to validation. Perform a loss operation on the predicted graphic data related to validation and the corresponding optimized mask graphic data in the validation set to obtain the training loss value, and at the same time perform an intersection over union (IoU) operation to obtain the IoU value.

[0070] When the IoU value is less than the second preset threshold or the second training loss value is greater than the first preset threshold, loop to execute steps S2 - S6. When the IoU value is greater than or equal to the second preset threshold and the second training loss value is less than or equal to the first preset threshold, or when the number of repeated training reaches the preset number of training times, exit the loop. Save the weight parameters of the current Unet deep convolutional network, and the Unet deep convolutional network after updating the weight parameters is the trained Unet deep convolutional network.

[0071] Select layout patterns outside the designed pattern. After converting the layout patterns into layout pattern data, input them into the trained Unet deep convolutional network to obtain the predicted and corrected mask pattern data, thus completing the super-resolution lithography optical proximity effect correction. Further, convert the predicted and corrected mask pattern data into the predicted and corrected mask pattern. The size of the layout patterns in this step may be inconsistent with the size of the designed patterns in step S1.

[0072] The present disclosure also provides a super-resolution lithography method, including: obtaining the predicted and corrected mask pattern data by using the above-mentioned super-resolution lithography optical proximity effect correction method based on the Unet deep convolutional network, outputting the predicted and corrected mask pattern, and performing super-resolution lithography according to the predicted and corrected mask pattern.

[0073] That is, based on the above-mentioned super-resolution lithography optical proximity effect correction method based on the Unet deep convolutional network, output the predicted and corrected mask pattern, and perform super-resolution lithography according to the predicted and corrected mask pattern. Operations S1 to S7 are the same as the foregoing operations and will not be elaborated here.

[0074] The present disclosure uses the Unet deep convolutional network to perform optical proximity effect correction. The Unet deep convolutional network consists of an encoding end, a decoding end, and a prediction network. The encoding end includes a residual module structure. This Unet deep convolutional network structure can not only avoid the problems of gradient disappearance or gradient explosion during training, but also use deeper convolutional layers to improve the network fitting ability, thereby better fitting the mapping rule between the designed pattern and the optimized mask pattern, and improving the accuracy of predicting and correcting the mask pattern. In addition, to adapt to the optical proximity effect correction of larger-sized lithography design layouts, the fully connected layers are removed from both the encoding end and the decoding end of the Unet deep convolutional network in the present disclosure. The input and output graphic data dimensions are not restricted, and the optical proximity effect correction can be performed on larger-sized layout patterns, greatly improving the correction efficiency.

[0075] The following further illustrates the present disclosure through specific embodiments. The above-mentioned super-resolution lithography optical proximity effect correction method based on the Unet deep convolutional network is specifically described in the following embodiments. However, the following embodiments are only used to illustrate the present disclosure, and the scope of the present disclosure is not limited thereto.

[0076] Specifically, as Figure 1 shown, the method of this embodiment includes the following steps:

[0077] Step S01: Select some design patterns from the layout to be corrected. Obtain the optimized mask patterns after optical proximity effect correction through the optical proximity effect correction method. Binarize the design patterns and the optimized mask patterns and store them in matrix form to form design pattern data and optimized mask pattern data respectively. Then pair the two to form known data pairs, and split the known data pairs into a training set and a validation set according to a quantity ratio of 7:3. This is equivalent to step S1 above.

[0078] In this step, the pixel sizes of the design patterns and the optimized mask patterns are the same, and the matrix dimension sizes of the design pattern data and the optimized mask pattern data are both 256×256. Figure 2 This is the data enhancement result of a design pattern in the training set. Among them, 201 is the design pattern before data enhancement, 202 is the design pattern after vertical mirror operation, 203 is the design pattern after horizontal mirror operation, and 204 is the design pattern after rotation operation. The same operations are performed on the optimized mask pattern corresponding to the design pattern data to form data pairs.

[0079] The optical proximity effect correction method used in this embodiment is inverse lithography technology, and the lithography model in this embodiment is a super-resolution lithography model. Figure 3 This is a schematic diagram of the super-resolution lithography structure used in S01. The super-resolution lithography structure 301 includes a mask substrate (SiO2), a mask pattern layer (Cr), an air spacer layer (Air), a metal layer (Ag), a photoresist layer (Pr), a metal reflection layer (Ag), and a substrate (SiO2). In this embodiment, the mask thickness is set to 50 nm, the air spacer is 20 nm, the upper metal layer of the photoresist is 20 nm, the lower metal layer is 70 nm, and the photoresist thickness is 40 nm.

[0080] Step S02: Perform data enhancement processing on the known data pairs in step S01. The data enhancement means in this embodiment are: vertical mirror, horizontal mirror, and rotation. After the data enhancement processing in this step, the known data pairs become enhanced known data pairs, and the number of data pairs will increase significantly. In this embodiment, the total number of data pairs increases to 4 times that before data enhancement. This is equivalent to step S20 above.

[0081] Step S03: Construct a Unet deep convolutional network. The encoding end and the decoding end of the Unet deep convolutional network constructed in this embodiment do not contain fully connected layers. Divide the design pattern data and the optimized mask pattern data in the training set into multiple batches. For each batch, input the design pattern data in the training set into the Unet deep convolutional network, which passes through the encoding end and the decoding end of the Unet deep convolutional network in sequence, and finally outputs prediction pattern data related to training through the prediction network. This is equivalent to step S2 above.

[0082] In this embodiment, the Unet deep convolutional network is divided into three parts: an encoding end, a decoding end, and a prediction network. The encoding end consists of a Resnet network structure, which can perform four downsamplings and perform zero-padding operations during the downsampling process. The decoding end consists of a transposed convolution network structure, which performs upsampling in the form of transposed convolution. Each upsampling in the decoding end combines the channels of the downsampled feature maps of the Resnet network structure at the same level for common convolution to achieve global feature prediction. The prediction network of the Unet deep convolutional network classifies the element values of each matrix into 0-1 through a normalization function.

[0083] In this embodiment, when training the Unet deep convolutional network, the classification of the element values of the matrix of the network is set to 2. The learning rate adjustment method of the network is set to adam optimized gradient descent. The number of batch processes for each iteration of the network is set to 10. The maximum number of iterations of the network is set to 1000. The iteration accuracy of the network is set to 0.0001.

[0084] In this embodiment, adding a batch normalization layer after each convolutional layer enables the final result to converge faster. Performing zero-padding operations during each convolutional operation can make the network layers clearer, so that the Unet deep convolutional network retains all features when shearing the feature map matrix, reduces the feature loss of downsampling, retains more information, and improves the fitting effect.

[0085] Step S04: For each batch, perform a loss operation on the prediction graphic data related to training and the optimized mask graphic data of this batch to obtain the first training loss value. Then, obtain the gradient of the weight parameters of the Unet deep convolutional network through backpropagation of the first training loss value, and update the weight parameters of the Unet deep convolutional network according to this gradient. This is equivalent to step S3 above.

[0086] Step S05: Execute step S03 on the design graphic data in the validation set to obtain the prediction graphic data related to validation. Perform a loss operation on the prediction graphic data related to validation and the corresponding optimized mask graphic data in the validation set to obtain the second training loss value, and at the same time perform an intersection over union operation to obtain the IoU value. In this embodiment, k = 1 is set. This is equivalent to steps S4~S5 above.

[0087] Step S06: Repeat steps S03 - S05. The number of repetitions is called the number of training epochs. As the number of training epochs increases, the second training loss value in S05 tends to decrease; the IoU value tends to increase. The curve of the training loss value changing with the number of training epochs is called the loss function curve, and the curve of the IoU value changing with the number of training epochs is called the IoU change curve. In this embodiment, the IoU value is monitored in real time during each round of training. When the IoU value is continuously greater than or equal to 90% within the most recent 100 epochs and the second training loss value is continuously less than or equal to 0.15 within these 100 epochs, the network training is terminated; the network training can also be terminated after the number of training epochs reaches the target number of epochs. Since there are fluctuations in the IoU value and the second training loss value during the training process, taking the results of 100 consecutive epochs can reduce the impact of fluctuations on the network output and make the judgment more accurate.

[0088] Step S07: Save the weight parameter values of the Unet deep convolutional network. The Unet deep convolutional network with the updated parameter values is the trained Unet deep convolutional network. This is equivalent to step S6 above.

[0089] Step S08: Select the layout graphics outside the designed graphics. After converting the layout graphics into layout graphic data, input it into the trained Unet deep convolutional network to obtain the predicted and corrected mask graphic data, and then convert the predicted and corrected mask graphic data into the predicted and corrected mask graphics. The size of the layout graphics in this step can be inconsistent with the size of the designed graphics in step S01. This is equivalent to step S7 above.

[0090] Figure 4 It is a schematic diagram of the Unet deep convolutional network structure, which is divided into three parts: an encoding end, a decoding end, and a prediction network. The left dashed box is the encoding end, composed of a Resnet network, which can implement downsampling and feature extraction functions; the middle dashed box is the decoding end, which can implement upsampling and enhanced feature extraction; the rightmost is the prediction network.

[0091] In this embodiment, the designed graphic matrix data enters the encoding end of the Unet deep convolutional network in the form of a tensor with dimensions (10, 1, 256, 256). Under the action of 7×7 convolution and max pooling, it becomes a tensor T1 with dimensions (10, 64, 128, 128). After passing through the first feature extraction module, it becomes a tensor T2 with dimensions (10, 128, 64, 64). After passing through the second feature extraction module, it becomes a tensor T3 with dimensions (10, 256, 32, 32). After passing through the third feature extraction module, it becomes a tensor T4 with dimensions (10, 512, 16, 16). After passing through the fourth feature extraction module, it becomes a tensor T5 with dimensions (10, 1024, 8, 8). Tensor T5 first passes through the upsampling module at the decoding end and its form becomes a tensor T6 with dimensions (10, 512, 16, 16). Tensor T6 and tensor T4 are glued together in the second dimension to form a tensor T7 with dimensions (10, 1024, 16, 16); Tensor T7 passes through the upsampling module and its form becomes a tensor T8 with dimensions (10, 256, 32, 32). Tensor T8 and tensor T3 are glued together in the second dimension to form a tensor T9 with dimensions (10, 512, 32, 32); After passing through the upsampling module, it becomes a tensor T 10 , tensor T 10 is glued to tensor T2 to form a tensor T 11 ; Tensor T 11 passes through the upsampling module and becomes a tensor T 12 with dimensions (10, 64, 128, 128), tensor T 12 and tensor T1 are glued together in the second dimension to form a tensor T 13 with dimensions (10, 128, 128, 128); Tensor T 13 passes through upsampling and becomes a tensor T 14 with dimensions (10, 64, 256, 256). Tensor T 14 then passes through a 3×3 convolution in the prediction network to obtain a tensor T 15 with dimensions (10, 2, 256, 256). Finally, after passing through the normalization function, the first channel of the second dimension is taken to obtain the predicted graph, and its tensor form is (10, 1, 256, 256).

[0092] Figure 5 is a schematic diagram of the Resnet network, which contains 101 convolutional kernels. Every 3 of these 101 convolutional kernels form a sub-convolutional kernel. Each sub-convolutional kernel contains a residual structure.

[0093] Figure 6 is a schematic diagram of the residual structure.

[0094] Figure 7Schematic diagram of the composition structure of sub-convolution kernels. Each sub-convolution kernel is composed of convolution kernels with dimensions of (1,1), (3,3), and (1,1), and these convolution kernels are connected by a residual structure.

[0095] Figure 8 Shown is the prediction correction result of one of the design graphic data in the validation set under the action of the Unet deep convolutional network. Among them, 801 represents the design graphic; 802 is the optimized mask graphic obtained by using the inverse lithography technology; 803 is the prediction graphic of the validation set obtained by the trained Unet deep convolutional network, and its IoU value is 85.749%.

[0096] Figure 9 Shown is the prediction correction result of the design graphic other than the known data pair. 901 represents the design graphic; 902 represents the corrected mask graphic obtained by using the inverse lithography technology; 903 represents the predicted corrected mask graphic obtained by the method of the present disclosure. The IoU value between 902 and 903 is 84.673%, and this value is comparable to the result of the validation set, indicating that the method of the present disclosure has good applicability to different design graphics.

[0097] Figure 10 It is the prediction correction result obtained by training the Unet deep convolutional network with known data pairs of dimension 256×256 according to the method of the present disclosure. After completing the training of the Unet deep convolutional network, layout graphic data with a dimension of 2100×2100 is input. Among them, 1001 is the layout graphic with a matrix dimension of 2100×2100, and 1002 is the predicted corrected mask graphic with a dimension of 2100×2100 obtained by inputting 1001 into the trained Unet deep convolutional network.

[0098] Figure 11 It is the IoU change curve and loss function curve obtained in step S06. 1101 is the IoU change curve, the ordinate of which is the IoU value, and the abscissa is the number of training epochs; 1102 is the loss function curve, the ordinate of which is the second training loss value, and the abscissa is the number of training epochs. It can be analyzed from the two figures that the network converges normally.

[0099] Figure 12 Schematically shows a block diagram of a super-resolution lithography optical proximity effect correction system based on the Unet deep convolutional network according to an embodiment of the present disclosure.

[0100] As Figure 12 shown, the Figure 12Schematically shown is a block diagram of a super-resolution lithography optical proximity effect correction system based on a Unet deep convolutional network according to an embodiment of the present disclosure. The correction system 1200 includes: an optical proximity effect correction module 1210, a training set processing module 1220, a first calculation module 1230, a validation set processing module 1240, a second calculation module 1250, a judgment module 1260, and a prediction module 1270.

[0101] The optical proximity effect correction module 1210 is configured to perform optical proximity effect correction on a design pattern and perform pixelization processing to obtain a known data pair composed of design pattern data and corresponding optimized mask pattern data; split the known data pair into a training set and a validation set. According to an embodiment of the present disclosure, the optical proximity effect correction module 1210 can be used, for example, to execute the S1 step described above with reference to Figure 1 which will not be elaborated herein.

[0102] The training set processing module 1220 is configured to batch-input the design pattern data in the training set into the Unet deep convolutional network to obtain prediction pattern data related to training. According to an embodiment of the present disclosure, the training set processing module 1220 can be used, for example, to execute the S2 step described above with reference to Figure 1 which will not be elaborated herein.

[0103] The first calculation module 1230 is configured to calculate a first training loss value for each batch according to the prediction pattern data related to training and the corresponding optimized mask pattern data, and update the weight parameters of the Unet deep convolutional network according to the first training loss value. According to an embodiment of the present disclosure, the first calculation module 1230 can be used, for example, to execute the S3 step described above with reference to Figure 1 which will not be elaborated herein.

[0104] The validation set processing module 1240 is configured to input the design pattern data in the validation set into the updated Unet deep convolutional network to obtain prediction pattern data related to validation. According to an embodiment of the present disclosure, the validation set processing module 1240 can be used, for example, to execute the S4 step described above with reference to Figure 1 which will not be elaborated herein.

[0105] The second calculation module 1250 is configured to calculate a second training loss value and a union intersection value according to the prediction pattern data related to validation and the corresponding optimized mask pattern data. According to an embodiment of the present disclosure, the second calculation module 1250 can be used, for example, to execute the S5 step described above with reference to Figure 1 which will not be elaborated herein.

[0106] A judgment module 1260 is configured to determine whether the second training loss value and the union intersection value meet corresponding preset conditions or whether the current number of training times reaches a preset number of training times; if not, the training is repeated; if so, the Unet deep convolutional network after updating the weight parameters currently is used as the trained Unet deep convolutional network. According to an embodiment of the present disclosure, the judgment module 1260 may be configured to execute, for example, step S6 described above with reference to Figure 1 and will not be elaborated herein.

[0107] A prediction module 1270 is configured to input a design pattern to be optimized into the trained Unet deep convolutional network to obtain predicted and corrected mask pattern data, thereby completing super-resolution lithography optical proximity effect correction. According to an embodiment of the present disclosure, the prediction module 1270 may be configured to execute, for example, step S7 described above with reference to Figure 1 and will not be elaborated herein.

[0108] It should be noted that, according to an embodiment of the present disclosure, any multiple of, or at least part of the functions of, the modules, sub-modules, units, and sub-units may be implemented in one module. Any one or more of the modules, sub-modules, units, and sub-units according to the embodiments of the present disclosure may be split into multiple modules for implementation. Any one or more of the modules, sub-modules, units, and sub-units according to the embodiments of the present disclosure may be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on chip, a system on substrate, a system on package, an application specific integrated circuit (ASIC), or may be implemented by any other reasonable manner of integrating or packaging circuits, or implemented in any one of the three implementation manners of software, hardware, and firmware, or in a suitable combination of any several of them. Alternatively, one or more of the modules, sub-modules, units, and sub-units according to the embodiments of the present disclosure may be at least partially implemented as a computer program module, and when the computer program module is run, the corresponding functions may be executed.

[0109] For example, any combination of the optical proximity effect correction module 1210, the training set processing module 1220, the first calculation module 1230, the validation set processing module 1240, the second calculation module 1250, the determination module 1260, and the prediction module 1270 can be combined and implemented in one module, or any one of them can be split into multiple modules. Alternatively, at least part of the functions of one or more of these modules can be combined with at least part of the functions of other modules and implemented in one module. According to an embodiment of the present disclosure, at least one of the optical proximity effect correction module 1210, the training set processing module 1220, the first calculation module 1230, the validation set processing module 1240, the second calculation module 1250, the determination module 1260, and the prediction module 1270 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on chip, a system on substrate, a system on package, an application specific integrated circuit (ASIC), or can be implemented by any other reasonable means of integrating or packaging circuits, etc., in hardware or firmware, or implemented in any one of the three implementation manners of software, hardware, and firmware, or in an appropriate combination of any several of them. Alternatively, at least one of the optical proximity effect correction module 1210, the training set processing module 1220, the first calculation module 1230, the validation set processing module 1240, the second calculation module 1250, the determination module 1260, and the prediction module 1270 can be at least partially implemented as a computer program module, which can execute corresponding functions when the computer program module is run.

[0110] Figure 13 FIG. schematically shows a block diagram of an electronic device suitable for implementing the method described above according to an embodiment of the present disclosure. Figure 13 The shown electronic device is merely an example and should not impose any limitation on the functions and usage scope of the embodiments of the present disclosure.

[0111] As Figure 13 shown, the electronic device 1300 described in this embodiment includes: a processor 1301, which can perform various appropriate actions and processes according to a program stored in a read only memory (ROM) 1302 or a program loaded from a storage section 1308 into a random access memory (RAM) 1303. The processor 1301 can include, for example, a general microprocessor (such as a CPU), an instruction set processor, and / or a related chipset, and / or a dedicated microprocessor (such as an application specific integrated circuit (ASIC)), etc. The processor 1301 can also include on-board memory for caching purposes. The processor 1301 can include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present disclosure.

[0112] In the RAM 1303, various programs and data required for the operation of the system 1300 are stored. The processor 1301, the ROM 1302, and the RAM 1303 are connected to each other via a bus 1304. The processor 1301 performs various operations of the method flow according to the embodiments of the present disclosure by executing programs in the ROM 1302 and / or the RAM 1303. It should be noted that the programs can also be stored in one or more memories other than the ROM 1302 and the RAM 1303. The processor 1301 can also perform various operations of the method flow according to the embodiments of the present disclosure by executing programs stored in one or more memories.

[0113] According to an embodiment of the present disclosure, the electronic device 1300 may further include an input / output (I / O) interface 1305, and the input / output (I / O) interface 1305 is also connected to the bus 1304. The system 1300 may further include one or more of the following components connected to the I / O interface 1305: an input part 1306 including a keyboard, a mouse, etc.; an output part 1307 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc. and a speaker, etc.; a storage part 1308 including a hard disk, etc.; and a communication part 1309 including a network interface card such as a LAN card, a modem, etc. The communication part 1309 performs communication processing via a network such as the Internet. A drive 1310 is also connected to the I / O interface 1305 as needed. A removable medium 1311, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 1310 as needed so that a computer program read from it can be installed into the storage part 1308 as needed.

[0114] According to an embodiment of the present disclosure, the method flow according to the embodiments of the present disclosure can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable storage medium, and the computer program includes program codes for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via the communication part 1309, and / or installed from the removable medium 1311. When the computer program is executed by the processor 1301, the above-mentioned functions defined in the system according to the embodiments of the present disclosure are executed. According to an embodiment of the present disclosure, the above-described system, device, apparatus, module, unit, etc. can be implemented by computer program modules.

[0115] Embodiments of the present disclosure also provide a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or may exist independently without being assembled into the device / apparatus / system. The above computer-readable storage medium stores one or more programs, and when the above one or more programs are executed, a super-resolution lithography optical proximity effect correction method based on the Unet deep convolutional network according to the embodiments of the present disclosure is implemented.

[0116] According to an embodiment of the present disclosure, the computer-readable storage medium may be a non-volatile computer-readable storage medium, for example, including but not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above. In an embodiment of the present disclosure, the computer-readable storage medium may be any tangible medium that contains or stores a program, and the program can be used by or in combination with an instruction execution system, apparatus, or device. For example, according to an embodiment of the present disclosure, the computer-readable storage medium may include the above-described ROM 1302 and / or RAM 1303 and / or one or more memories other than ROM1302 and RAM 1303.

[0117] Embodiments of the present disclosure also include a computer program product, which includes a computer program that contains program code for executing the method shown in the flowchart. When the computer program product runs in a computer system, the program code is used to cause the computer system to implement the super-resolution lithography optical proximity effect correction method provided by the embodiments of the present disclosure.

[0118] When the computer program is executed by the processor 1301, the above functions defined in the system / apparatus of the embodiments of the present disclosure are executed. According to an embodiment of the present disclosure, the above-described systems, apparatuses, modules, units, etc. may be implemented by computer program modules.

[0119] In one embodiment, the computer program may rely on tangible storage media such as optical storage devices and magnetic storage devices. In another embodiment, the computer program may also be transmitted and distributed in the form of a signal on a network medium, and downloaded and installed through the communication part 1309, and / or installed from the removable medium 1311. The program code included in the computer program may be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0120] In such an embodiment, the computer program can be downloaded and installed from a network through the communication section 1309, and / or installed from the removable medium 1311. When the computer program is executed by the processor 1301, the above-described functions defined in the system of the embodiments of the present disclosure are performed. According to an embodiment of the present disclosure, the above-described system, device, apparatus, module, unit, etc. can be implemented by computer program modules.

[0121] According to an embodiment of the present disclosure, program code for executing the computer program provided in the embodiments of the present disclosure can be written in any combination of one or more programming languages. Specifically, these computing programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages include, but are not limited to, such as Java, C++, Python, the "C" language, or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (e.g., by connecting through the Internet using an Internet service provider).

[0122] It should be noted that in each embodiment of the present disclosure, the various functional modules can be integrated in one processing module, or each module can exist physically separately, or two or more modules can be integrated in one module. The above-mentioned integrated modules can be implemented in the form of hardware or in the form of software functional modules. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present disclosure, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product.

[0123] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagram may represent a module, a segment of a program, or a portion of code that contains one or more executable instructions for implementing the specified logical function. It should also be noted that, in some alternative implementations, the functions noted in the blocks may occur in a different order than that noted in the accompanying drawings. For example, two consecutive blocks shown may actually be executed substantially in parallel, or they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram or flowchart, and combinations of blocks in the block diagram or flowchart, can be implemented by a dedicated hardware-based system that performs the specified functions or operations, or by a combination of dedicated hardware and computer instructions.

[0124] Those skilled in the art will appreciate that the features recited in the various embodiments and / or claims of the present disclosure may be combined and / or combined in various ways, even if such combinations or combinations are not explicitly recited in the present disclosure. In particular, without departing from the spirit and teachings of the present disclosure, the features recited in the various embodiments and / or claims of the present disclosure may be combined and / or combined in various ways. All such combinations and / or combinations fall within the scope of the present disclosure.

[0125] Although the present disclosure has been shown and described with reference to particular exemplary embodiments thereof, those skilled in the art should understand that various changes in form and detail may be made therein without departing from the spirit and scope of the present disclosure as defined by the appended claims and their equivalents. Therefore, the scope of the present disclosure should not be limited to the above embodiments, but should be determined not only by the appended claims but also by the equivalents of the appended claims.

Claims

1. A super-resolution lithography optical proximity effect correction method based on the Unet deep convolutional network, characterized in that, Including: S1. Perform optical proximity effect correction on the design pattern and perform pixelization to obtain a known data pair composed of design pattern data and corresponding optimized mask pattern data; split the known data pair into a training set and a validation set. S2. Input the design pattern data in the training set into the Unet deep convolutional network in batches to obtain predicted pattern data related to training. S3. For each batch, calculate a first training loss value based on the predicted pattern data related to training and the corresponding optimized mask pattern data, and update the weight parameters of the Unet deep convolutional network according to the first training loss value. S4. Input the design pattern data in the validation set into the updated Unet deep convolutional network to obtain predicted pattern data related to validation. S5. Calculate a second training loss value and an intersection over union value based on the predicted pattern data related to validation and the corresponding optimized mask pattern data. S6. Determine whether the second training loss value and the intersection over union value meet corresponding preset conditions or whether the current training times reach the preset training times. If not, repeat S2 - S6 for training; if so, the Unet deep convolutional network with the currently updated weight parameters is the trained Unet deep convolutional network. S7. Input the design pattern to be optimized into the trained Unet deep convolutional network to obtain predicted and corrected mask pattern data, and complete the super-resolution lithography optical proximity effect correction.

2. The super-resolution lithography optical proximity effect correction method based on the Unet deep convolutional network according to claim 1, wherein The optical proximity effect correction of the design pattern in S1 includes: Performing optical proximity effect correction on the design pattern to obtain an optimized mask pattern. Performing pixelization on the design pattern and the optimized mask pattern respectively to obtain design pattern data and corresponding optimized mask pattern data, and the design pattern data and the corresponding optimized mask pattern data form a known data pair. Among them, the design pattern to be optimized in S7 is the whole or most of the design patterns of the large-size mask layout to be optimized, and the design pattern in S1 is a small part of the design patterns of the large-size mask layout to be optimized.

3. The super-resolution lithography optical proximity effect correction method based on the Unet deep convolutional network according to claim 1, characterized in that The method for performing optical proximity effect correction in S1 includes any one of a rule-based optical proximity effect correction method, a model-based optical proximity effect correction method, and inverse lithography technology.

4. The super-resolution lithography optical proximity effect correction method based on the Unet deep convolutional network according to claim 1, characterized in that After S1, it further includes: Performing data augmentation processing on the known data pair; wherein, the data augmentation processing includes one or more of vertical flipping, horizontal flipping, and rotation operations.

5. The super-resolution lithography optical proximity effect correction method based on the Unet deep convolutional network according to claim 1, characterized in that Before S2, it further includes: S20. Construct a Unet deep convolutional network, and the Unet deep convolutional network includes an encoding end, a decoding end, and a prediction network; among them, the encoding end and the decoding end do not contain fully connected layers.

6. The super-resolution lithography optical proximity effect correction method based on the Unet deep convolutional network according to claim 1, wherein S3 includes: For each batch, calculate the first training loss value according to the prediction graphic data related to the training and the corresponding optimized mask graphic data through the following formula :[[]]END]] Among them, is the prediction graphic data related to training, is the corresponding optimized mask graphic data, MSE represents the mean square error, and n is the value of the batch size; Backpropagate the first training loss value to obtain the gradient of the weight parameters of the Unet deep convolutional network. Update the weight parameters of the Unet deep convolutional network according to the gradient.

7. The super-resolution lithography optical proximity effect correction method based on the Unet deep convolutional network according to claim 1, characterized in that S5 includes: Calculate the intersection over union value IoU through the following formula: Among them, k + 1 represents the number of matrix element value classifications; represents the number of matrix elements that should have been classified as the value i but were predicted as j, represents the number of matrix elements that should have been classified as the value i and were predicted as i, represents the number of matrix elements that should have been classified as the value j but were predicted as i.

8. The super-resolution lithography optical proximity effect correction method based on the Unet deep convolutional network according to claim 1, characterized in that The corresponding preset conditions in S6 include: The second training loss value is less than or equal to a first preset threshold and the union intersection value is greater than or equal to a second preset threshold.

9. A super-resolution lithography method, characterized in that, Including: Using the super-resolution lithography optical proximity effect correction method based on the Unet deep convolutional network according to any one of claims 1 to 8 to obtain predicted corrected mask pattern data, outputting the predicted corrected mask pattern, and performing super-resolution lithography according to the predicted corrected mask pattern.

10. A super-resolution lithography optical proximity effect correction system based on the Unet deep convolutional network, characterized in that, Including: An optical proximity effect correction module for performing optical proximity effect correction on a design pattern and performing pixelization processing to obtain a known data pair composed of design pattern data and corresponding optimized mask pattern data, and splitting the known data pair into a training set and a validation set; A training set processing module for batch-inputting the design pattern data in the training set into the Unet deep convolutional network to obtain training-related predicted pattern data; A first calculation module for calculating a first training loss value according to the training-related predicted pattern data and the corresponding optimized mask pattern data for each batch, and updating the weight parameters of the Unet deep convolutional network according to the first training loss value; A validation set processing module for inputting the design pattern data in the validation set into the updated Unet deep convolutional network to obtain validation-related predicted pattern data; A second calculation module for calculating a second training loss value and a union intersection value according to the validation-related predicted pattern data and the corresponding optimized mask pattern data; A judgment module for judging whether the second training loss value and the union intersection value meet corresponding preset conditions or whether the current training times reach a preset training times; if not, repeat the training; if so, the Unet deep convolutional network with the current updated weight parameters is the trained Unet deep convolutional network; A prediction module for inputting the design pattern to be optimized into the trained Unet deep convolutional network to obtain predicted corrected mask pattern data, and completing the super-resolution lithography optical proximity effect correction.

11. An electronic device, including: A processor; A memory storing a computer-executable program, which when executed by the processor causes the processor to execute the super-resolution lithography optical proximity effect correction method based on the Unet deep convolutional network according to any one of claims 1 to 8.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the super-resolution lithography optical proximity effect correction method based on the Unet deep convolutional network according to any one of claims 1 to 8.

13. A computer program product including a computer program, which when executed by a processor implements the super-resolution lithography optical proximity effect correction method based on the Unet deep convolutional network according to any one of claims 1 to 8.

Images