Electronic device and operation method thereof

Abstract

The invention provides an electronic device and an operation method thereof. The electronic device includes: a memory configured to store instructions; and a processing circuit configured to execute instructions stored in the memory to cause the electronic device to acquire a first input image based on an image of a design pattern to be formed on the wafer, transform the first input image into a second input image based on a factor that does not affect a process of manufacturing the semiconductor device, and training a process simulation model using the paired first and second input images.

Description

Technical Field

[0001] The present disclosure relates to an electronic device and a method of operating the same. Background Art

[0002] In an exposure process or an etching process of manufacturing a semiconductor device, for example, due to diffraction effects or process effects, a designed pattern and an actual pattern formed on a wafer may be different from each other. Optical proximity correction (OPC) techniques and process proximity correction (PPC) techniques are being developed to predict this error and correct a mask pattern for a designed pattern to be formed on a wafer. Accordingly, there is a need for a simulation model that can accurately simulate a corresponding process. Summary of the Invention

[0003] Some example embodiments of the present disclosure provide an electronic device capable of reducing or preventing overfitting of a process simulation model and improving robustness of the process simulation model.

[0004] According to an example embodiment, an electronic device may include: a memory configured to store instructions; and a processing circuit configured to execute the instructions stored in the memory to cause the electronic device to obtain a first input image based on an image of a designed pattern to be formed on a wafer, transform the first input image into a second input image based on factors that do not affect a process of manufacturing a semiconductor device, and train a process simulation model using paired first and second input images.

[0005] According to an example embodiment, an electronic device may include: a memory configured to store instructions; and a processing circuit configured to execute the instructions stored in the memory to cause the electronic device to obtain a first input image based on an image of a designed pattern to be formed on a wafer, transform the first input image into a second input image based on factors that do not affect a process of manufacturing a semiconductor device, obtain paired output images by inputting the paired first and second input images into a process simulation model, and train the process simulation model to reduce a value of a first loss function based on a difference between one of the paired output images and an image of an actual pattern formed on a wafer, and reduce a value of a second loss function based on a difference between the paired output images.

[0006] According to an example embodiment, a method of operating an electronic device may include: obtaining a first input image based on an image of a designed pattern to be formed on a wafer; transforming the first input image into a second input image based on factors that do not affect a process of manufacturing a semiconductor device; and training a process simulation model using the paired first and second input images.

[0007] As described above, according to some example embodiments of the present disclosure, overfitting of a process simulation model can be alleviated or prevented, and the robustness of the process simulation model can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 is a block diagram of an electronic device according to an example embodiment.

[0009] Figure 2 is a flowchart showing a method of operating an electronic device according to an example embodiment.

[0010] Figure 3 is a diagram showing a method of an electronic device training a process simulation model according to an example embodiment.

[0011] Figure 4 is a diagram showing a method of an electronic device transforming an input image according to an example embodiment.

[0012] Figure 5 is a diagram showing a method of an electronic device inverse-transforming an output image according to an example embodiment.

[0013] Figure 6 is a diagram showing a method of an electronic device training a process simulation model according to an example embodiment.

[0014] Figure 7 is a diagram showing an example of a computer device implementing an electronic device according to an example embodiment. DETAILED DESCRIPTION

[0015] Hereinafter, some example embodiments of the present disclosure will be described in detail with reference to the accompanying drawings so that those skilled in the art can easily practice the present disclosure. However, the present disclosure can be implemented in various different forms and is not limited to the example embodiments described herein.

[0016] Parts irrelevant to the description are omitted to clearly describe the present disclosure, and throughout the specification, the same or similar components are denoted by the same reference numerals.

[0017] In addition, the size and thickness of each component shown in the drawings are arbitrarily shown for ease of explanation, and thus, the present disclosure is not necessarily limited to what is shown in the drawings. The thickness is exaggerated in the drawings to clearly show several layers and regions. In addition, for ease of explanation, the thickness of some layers and regions is exaggerated in the drawings.

[0018] In addition, when an element such as a layer, film, region, or substrate is referred to as "on another element" or "above another element", the element can be "directly on the other element" or there can be a third element therebetween. On the other hand, when an element is referred to as "directly on the other element", there is no third element therebetween. In addition, when an element is referred to as "on a reference element" or "above a reference element", the element can be disposed on or below the reference element and may not necessarily be "on the reference element" or "above the reference element" in the opposite direction of gravity.

[0019] In addition, throughout the specification, unless otherwise described, "including" any component will be understood to imply including other elements without excluding other elements.

[0020] In addition, throughout the specification, the expression "on a plane" may indicate a situation of observing an object from the top, and the expression "in a cross-section" may indicate a situation of observing a cross-section taken along a vertical direction of the object from the side.

[0021] In addition, terms such as "part", "~er / or", "module", etc. described in the specification may indicate a unit that processes at least one function or operation, and may be implemented by hardware, software, or a combination of hardware and software. In addition, except for "part", "~er / or", or "module" that needs to be implemented by specific hardware, a plurality of "parts", a plurality of "~er / ors", or a plurality of "modules" may also be integrated into at least one module and implemented by at least one processor.

[0022] In the specification, "transmit" or "provide" may include not only direct transmission or provision, but also indirect transmission or provision through another device or using a bypass path.

[0023] Unless explicitly expressed such as "one" or "single", singular terms in the specification may be interpreted as singular or plural.

[0024] Hereinafter, reference Figure 1 is made to describe an electronic device according to an exemplary embodiment.

[0025] An electronic device according to an exemplary embodiment can train an image-based deep learning model that simulates a process of manufacturing a semiconductor device by processing a wafer (e.g., an exposure process or an etching process), and predict a process result based on the trained model. For example, in an exposure process or an etching process, a mask pattern and a pattern formed on a wafer can be different from each other. That is, a design pattern and an actual pattern of a wafer can be different from each other, so a process simulation model that can accurately predict an actual pattern based on a design pattern is required. The process simulation model can be used to predict and correct process defects.

[0026] Figure 1 is a block diagram of an electronic device according to an exemplary embodiment.

[0027] Referring Figure 1 , the electronic device 100 according to an exemplary embodiment may include a processor 110 and a memory 120. In some exemplary embodiments, the electronic device 100 may further include another component (e.g., a communication circuit).

[0028] The processor 110 may be operably connected to the memory 120. The processor 110 may run instructions stored in the memory 120. The processor 110 may run instructions stored in the memory 120 to allow the electronic device 100 to perform the operations described below. Operations described below as being performed by the processor 110 may be performed by the processor 110 and / or at least one other component of the electronic device 100 connected to the processor 110. Thus, it can be understood that the operations are performed by the electronic device 100.

[0029] The processor 110 may obtain a first input image based on an image of a design pattern to be formed on a wafer. The image of the design pattern may be an image in which a design engineer of a semiconductor device designs a circuit pattern to be formed on a wafer. The design pattern may include, for example, an exposure mask pattern or an etch mask pattern. The image of the design pattern may be stored in the memory 120. For example, the image of the design pattern may be received by the electronic device 100 from an external electronic device and stored in the memory 120.

[0030] The processor 110 may obtain a first input image by enhancing the image of the design pattern. For example, the image of the design pattern may be an image corresponding to one wafer. The processor 110 may randomly crop some regions of the image of the design pattern to a specified size. The processor 110 may obtain a plurality of cropped images from the image of one image of the design pattern. The processor 110 may rotate or flip each cropped image to generate a plurality of enhanced images from one cropped image. The plurality of enhanced images generated from one image of the design pattern may be used as input data to train a process simulation model. The processor 110 may obtain a plurality of first input images by enhancing one image of the design pattern. The processor 110 may obtain a plurality of first input images corresponding to each image of a plurality of design patterns.

[0031] As described above, the electronic device 100 may increase the amount of training data by obtaining a plurality of first input images from one image of the design pattern, thereby improving the performance of the process simulation model as a training target.

[0032] Processor 110 may transform a first input image into a second input image based on factors that do not affect the process. The factors that do not affect the process may include at least one of translation, rotation, or flipping of the design pattern. For example, in a specific process, the result (result 1) of the process performed when the design pattern is moved in the x-axis direction and / or the y-axis direction may be substantially the same as the result (result 2) of the process performed when the design pattern is not moved. Here, the fact that the process results are substantially the same as each other may indicate that result 2 matches result 1 when result 2 is translated in the same manner as the design pattern is translated. For example, in a specific process, the result (result 1) of the process performed when the design pattern is rotated by 90°, 180°, and / or 270° may be substantially the same as the result (result 2) of the process performed when the design pattern is not rotated. Here, the fact that the process results are substantially the same as each other may indicate that result 2 matches result 1 when result 2 is rotated in the same manner as the design pattern is rotated. For example, in a specific process, the result (result 1) of the process performed when the design pattern is flipped horizontally and / or vertically may be substantially the same as the result (result 2) of the process performed when the design pattern is not flipped. Here, the fact that the process results are substantially the same as each other may indicate that result 2 matches result 1 when result 2 is flipped in the same manner as the design pattern is flipped. When the process results of the untransformed design pattern and the transformed design pattern are substantially the same as each other, even if the design pattern is transformed by translation, rotation, and / or flipping, the transformation of this design pattern may be a factor that does not affect the process.

[0033] The factors that do not affect the process may be different for each process. For example, in a specific process, translation or rotation of the design pattern may not affect the process, while flipping may affect the process. Again, for example, in a specific process, a 180° rotation may not affect the process, while a 90° rotation or a 270° rotation may affect the process. These are only examples, and the factors that do not affect the process may be different for each process.

[0034] The processor 110 may obtain a second input image by performing at least one transformation of translation, rotation, or flipping on a pattern included in the first input image. The translation may include, for example, the movement of the pattern in the x-axis direction and / or the y-axis direction, and is not limited thereto. The rotation may include, for example, a rotation of 90°, a rotation of 180°, and / or a rotation of 270°, and is not limited thereto, and may include rotations of various other angles. The flipping may include, for example, horizontal flipping and / or vertical flipping, and is not limited thereto. The processor 110 may obtain the second input image by performing at least one transformation corresponding to (one or more) factors that do not affect the process among the above various types of transformations. For example, in a specific process, when the x-axis movement, y-axis movement, and 180° rotation of the design pattern do not affect the process, the processor 110 may obtain a second input image including the pattern in the first input image that has been moved by x1 in the x-axis direction, moved by y1 in the y-axis direction, and rotated by 180°.

[0035] When multiple transformations correspond to factors that do not affect the process, the processor 110 may perform only some of these factors. For example, in a specific process, when the x-axis movement, y-axis movement, and 180° rotation of the design pattern do not affect the process, the processor 110 may obtain a second input image including the pattern in the first input image that has been rotated by 180°.

[0036] The processor 110 may train a process simulation model by using a pair of the first input image and the second input image. The process simulation model is a deep learning model that simulates the process based on the image and may predict an actual pattern formed on the wafer by inputting an image of a design pattern to be formed on the wafer, and output an image of the predicted actual pattern. The processor 110 may obtain a pair of first output images and second output images by inputting the pair of the first input image and the second input image into the process simulation model, where the first output image corresponds to the first input image and the second output image corresponds to the second input image.

[0037] The processor 110 may train the process simulation model based on the difference between the first output image and the image of the actual pattern formed on the wafer. The image of the actual pattern formed on the wafer may be an image obtained by measuring the actual pattern formed on the wafer as a result of performing a process based on the design pattern corresponding to the first input image. The image obtained by measuring the actual pattern may be stored in the memory 120. For example, the image obtained by measuring the actual pattern may be received by the electronic device 100 from an external electronic device (e.g., a measuring instrument) and stored in the memory 120. As described above, the first input image may be an image obtained by enhancing the image of the design pattern. Therefore, the image of the actual pattern to be compared may be an enhanced image of the image obtained by measuring the actual pattern. For example, the image of the actual pattern may be an image obtained by cropping, rotating, and / or flipping the image obtained by measuring the actual pattern.

[0038] The processor 110 may train the process simulation model to reduce the value of the first loss function based on the difference between the first output image and the image of the actual pattern.

[0039] As described above, the process result of the design pattern corresponding to the second input image may be substantially the same as the process result of the design pattern corresponding to the first input image. The process result may correspond to the output data of the process simulation model, and thus the first output image corresponding to the first input image and the second output image corresponding to the second input image may be substantially the same as each other. Therefore, the processor 110 may train the process simulation model based on the difference between the first output image and the second output image in addition to the difference between the first output image and the image of the actual pattern.

[0040] The second output image is the output data corresponding to the second input image obtained by transforming the first input image. Therefore, the first output image and the second output image may also have the same relationship as the transformation relationship between the first input image and the second input image. According to an exemplary embodiment, the processor 110 may perform an inverse transformation on the second output image. The processor 110 may perform an inverse transformation on the second output image by reversing the transformation from the first input image to the second input image. The processor 110 may compare the first output image with the inversely transformed second output image. However, the exemplary embodiment is not limited thereto. In some exemplary embodiments, the processor 110 may obtain a transformed first output image by performing the same transformation on the first output image as the transformation from the first input image to the second input image. Here, the processor 110 may compare the transformed first output image with the second output image.

[0041] According to an example embodiment, the processor 110 may train a process simulation model based on the difference between the first output image and the image of the actual pattern and the difference between the first output image and the second output image after inverse transformation. The processor 110 may train the process simulation model to reduce the value of a first loss function based on the difference between the first output image and the image of the actual pattern, and to reduce the value of a second loss function based on the difference between the first output image and the second output image after inverse transformation.

[0042] For example, the processor 110 may train the process simulation model to reduce the sum of the value of the first loss function and the value of the second loss function. For example, the processor 110 may train the process simulation model to minimize the sum of the value of the first loss function and the value of the second loss function. Training the process simulation model may indicate adjusting its parameters to reduce the value of the loss function of the process simulation model.

[0043] In addition to the above-mentioned first loss function and second loss function, the loss function for training the process simulation model may further include a loss function related to another loss. The processor 110 may calculate the sum of multiple loss functions by multiplying multiple loss functions by ratios for each loss function. For example, the processor 110 may calculate the sum of the first loss function and the second loss function by adding the value obtained by multiplying the value of the first loss function by the ratio corresponding to the first loss function and the value obtained by multiplying the value of the second loss function by the ratio corresponding to the second loss function.

[0044] The processor 110 may generate multiple mini-batches by randomly sampling pairs of input images. The processor 110 may generate multiple mini-batches by randomly extracting a specified number of pairs of input images multiple times from multiple pairs of input images, where each pair of input images includes a first input image and a second input image transformed from the first input image. The processor 110 may train the process simulation model by using multiple mini-batches. The processor 110 may train the process simulation model in units of mini-batches. For example, the processor 110 may extract n pairs of input images m times from all pairs of input images and generate m mini-batches of size n. The processor 110 may train the process simulation model in units of mini-batches and update the process simulation model whenever it finishes learning for each mini-batch. The processor 110 may train the process simulation model in units of mini-batches, thereby allowing the process simulation model to have improved learning robustness by learning more general patterns, having an improved learning speed, and having a reduced noise impact.

[0045] The processor 110 may predict a process result based on a trained process simulation model. The processor 110 may obtain an output image corresponding to an actual pattern predicted to be formed on a wafer when a process is performed based on a design pattern by inputting an input image obtained from an image of the design pattern into the trained process simulation model.

[0046] According to an example embodiment, the electronic device 100 may mitigate or prevent overfitting of the process simulation model by training the process simulation model based on a first input image and further based on a second input image obtained by transforming the first input image based on factors that do not affect the process.

[0047] According to an example embodiment, the electronic device 100 may input a pair of the first input image and the second input image into the process simulation model, and thus train the process simulation model based on a difference between a first output image corresponding to the first input image and an image of the actual pattern and further based on a difference between a second output image corresponding to the second input image and the first output image.

[0048] According to an example embodiment, the electronic device 100 may input the first input image and the second input image into the process simulation model respectively, and thus train the process simulation model based on a difference between a first output image corresponding to the first input image and an image of the actual pattern, and improve the robustness of the process simulation model as compared with a comparative example of training the process simulation model based on a difference between a second output image corresponding to the second input image and an image of the actual pattern.

[0049] The following description refers to Figure 2 a method of operating an electronic device according to an example embodiment.

[0050] Figure 2 is a flowchart showing a method of operating an electronic device according to an example embodiment. The following operations may be performed by Figure 1 the electronic device 100.

[0051] In operation 210, the electronic device may obtain a first input image based on an image of a design pattern to be formed on a wafer. The electronic device may obtain the first input image by enhancing the image of the design pattern. For example, the electronic device may obtain the first input image by cropping, rotating, and / or flipping the image of the design pattern.

[0052] In operation 220, the electronic device may transform the first input image into a second input image based on factors that do not affect the process. The factors that do not affect the process may include at least one of translation, rotation, or flipping of the design pattern. The factors that do not affect the process may be different for each process. The electronic device may obtain the second input image by performing at least one transformation of translation, rotation, or flipping corresponding to the factors that do not affect the process on the pattern included in the first input image.

[0053] In operation 230, the electronic device may train a process simulation model by using a pair of the first input image and the second input image. The process simulation model is a deep learning model based on images and may receive an image corresponding to a design pattern and output an image corresponding to the predicted actual pattern. The electronic device may obtain a pair of first output images and second output images by inputting the pair of the first input image and the second input image into the process simulation model, where the first output image corresponds to the first input image and the second output image corresponds to the second input image.

[0054] The electronic device may perform an inverse transformation on the second output image. The electronic device may perform an inverse transformation on the second output image by reversing the transformation from the first input image to the second input image.

[0055] The electronic device may obtain an image of the actual pattern by enhancing an image obtained by measuring a wafer on which a process is performed based on a design pattern. The electronic device may obtain an image of the actual pattern by enhancing the image obtained by measuring the wafer in the same manner as the enhancement method for obtaining the first input image from the image of the design pattern.

[0056] The electronic device may train the process simulation model based on the difference between the first output image and the image of the actual pattern and the difference between the first output image and the second output image after the inverse transformation. The electronic device may train the process simulation model to reduce the value of a first loss function based on the difference between the first output image and the image of the actual pattern and the value of a second loss function based on the difference between the first output image and the second output image after the inverse transformation. For example, the electronic device may train the process simulation model to minimize the sum of the value of the first loss function and the value of the second loss function. For example, the electronic device may train the process simulation model to minimize the sum of the value obtained by multiplying the value of the first loss function by a ratio corresponding to the first loss function and the value obtained by multiplying the value of the second loss function by a ratio corresponding to the second loss function.

[0057] The electronic device can generate multiple mini - batches by randomly sampling pairs of input images. The electronic device can generate a mini - batch by randomly extracting a specified number of pairs of input images from multiple pairs of input images, each of which includes a first input image and a second input image. The electronic device can perform random sampling multiple times to generate multiple mini - batches.

[0058] The electronic device can train a process simulation model by using the multiple generated mini - batches. The electronic device can train the process simulation model in units of mini - batches. Whenever the model completes its learning for each mini - batch, the electronic device can update the parameters of the process simulation model.

[0059] In operation 240, the aforementioned process simulation model can be used to fabricate a mask to correct a mask pattern. In operation 250, the mask can be used to fabricate a semiconductor chip.

[0060] The following description refers to Figure 3 illustrate a method for an electronic device to train a process simulation model according to an exemplary embodiment.

[0061] Figure 3 is a diagram showing a method for an electronic device to train a process simulation model according to an exemplary embodiment. The electronic device according to the exemplary embodiment can be Figure 1 the electronic device 100.

[0062] Referring to Figure 3 , the electronic device can obtain a first input image 320 by enhancing an image 310 of a design pattern to be formed on a wafer. For convenience, Figure 3 the image 310 of the design pattern in can be an image of the design pattern on an entire wafer and indicates the area to be cropped to obtain the first input image 320. For example, the electronic device can obtain the first input image 320 by cropping the area at a random position in the image 310 of the design pattern to a specified size and rotating it - 90° (e.g., performing a 90° counter - clockwise rotation).

[0063] The electronic device can obtain a second input image 330 by performing a transformation corresponding to a factor that does not affect the process on the first input image 320. For example, the transformation corresponding to a factor that does not affect the process can include at least one of translation, rotation, or flipping.

[0064] The electronic device may input a pair of first input images 320 and second input images 330 into the process simulation model 300. The process simulation model 300 may output a pair of first output images 340 and second output images 350. The first output image 340 may include the actual pattern predicted to be formed on the wafer when the process is performed based on the pattern included in the first input image 320. The second output image 350 may include the actual pattern predicted to be formed on the wafer when the process is performed based on the pattern included in the second input image 330.

[0065] The electronic device may obtain the second output image 360 after inverse transformation by performing the transformation from the first input image 320 to the second input image 330 on the second output image 350 conversely. For example, the electronic device may obtain the second input image 330 by rotating the first input image 320 by +180° (e.g., performing a 180° clockwise rotation). In this case, the electronic device may obtain the second output image 360 after inverse transformation by rotating the second output image 350 by -180° (e.g., performing a 180° counterclockwise rotation).

[0066] For example, the process may be actually performed based on the design pattern, and the actual pattern formed on the wafer may then be measured by a measuring instrument. Here, the electronic device may obtain the image 380 of the actual pattern by enhancing the measured image or the image 370 to be measured. For convenience, Figure 3 the image 370 to be measured in may be an image of the design pattern formed on an entire wafer and indicate the area to be cut out to obtain the image 380 of the actual pattern. The electronic device may obtain the image 380 of the actual pattern in the same manner as the method of enhancing the image 310 of the design pattern to obtain the first input image 320. For example, the electronic device may cut out the area of the image 370 to be measured corresponding to the area where the image 310 of the design pattern is cut out. For example, the electronic device may obtain the first input image 320 by cutting out a partial area of the image 310 of the design pattern and then rotating it by -90° (e.g., performing a 90° counterclockwise rotation). In this case, the electronic device may obtain the image 380 of the actual pattern by rotating the image corresponding to the cut-out area of the image 370 to be measured by -90° (e.g., performing a 90° counterclockwise rotation).

[0067] The electronic device may calculate the value of the first loss function L1 based on the difference between the first output image 340 and the image 380 of the actual pattern. The electronic device may calculate the value of the second loss function L2 based on the difference between the first output image 340 and the second output image 360 after inverse transformation. The electronic device may train the process simulation model 300 to reduce the values of the first loss function L1 and the second loss function L2. For example, the electronic device may train the process simulation model 300 by updating the parameters of the process simulation model 300 to have parameter values for minimizing the sum of the value of the first loss function L1 and the value of the second loss function L2. For example, the electronic device may update the parameters of the process simulation model 300 to have parameter values for minimizing the sum of the value of the first loss function L1 multiplied by a ratio c1 corresponding to the first loss function L1 and the value of the second loss function L2 multiplied by a ratio c2 corresponding to the second loss function L2.

[0068] The electronic device may obtain a plurality of first input images 320 by enhancing an image 310 of a design pattern, and train the process simulation model 300 by using paired input images and an image 380 of an actual pattern corresponding to the image 310 of the design pattern.

[0069] The electronic device may obtain a plurality of first input images 320 by enhancing each of a plurality of images 310 of a design pattern, and train the process simulation model 300 by using paired input images and a plurality of images 380 of actual patterns respectively corresponding to the plurality of images 310 of the design pattern.

[0070] The following description refers to Figure 4 a method for the electronic device according to an exemplary embodiment to transform an input image.

[0071] Figure 4 a diagram showing a method for the electronic device according to an exemplary embodiment to transform an input image. The electronic device according to an exemplary embodiment may be Figure 1 the electronic device 100.

[0072] Referring to Figure 4 , the electronic device may generate a second input image 420 by transforming a first input image 410. The electronic device may transform the first input image 410 into the second input image 420 based on factors that do not affect the process. For example, the factors that do not affect the process may include at least one of translation, rotation, or flipping of the design pattern. Translation may include, for example, the movement of the pattern in the x-axis direction and / or the y-axis direction, and is not limited thereto. Flipping may include, for example, horizontal flipping or vertical flipping, and is not limited thereto. Rotation may include, for example, a rotation of 90°, a rotation of 180°, or a rotation of 270°, and is not limited thereto.

[0073] The electronic device may perform at least one transformation corresponding to a factor that does not affect the process among multiple transformations. The multiple transformations may include, for example, a first transformation 421-1 that moves a pattern in the x-axis direction or the y-axis direction, a second transformation 422-1 that flips the pattern in the horizontal direction, a third transformation 422-2 that flips the pattern in the vertical direction, a fourth transformation 423-1 that rotates the pattern clockwise by 90°, a fifth transformation 423-2 that rotates the pattern clockwise by 180°, and a sixth transformation 423-3 that rotates the pattern clockwise by 270°, and these transformations are not limited thereto. The electronic device may obtain a second input image 420 by performing at least one transformation corresponding to a factor that does not affect the process among the transformations including the first transformation 421-1 to the sixth transformation 423-3 on the pattern included in the first input image 410.

[0074] For example, the factors that do not affect the process may include the movement of the pattern in the x-axis direction or the y-axis direction, a clockwise rotation of 270°, and a horizontal flip. The electronic device may obtain the second input image 420 from the first input image 410 by performing the first transformation 421-1, the second transformation 422-1, and the sixth transformation 423-3. For example, the electronic device may obtain a second input image 420 including a pattern that has been moved -5 in the x-axis direction and +1 in the y-axis direction, flipped horizontally, and rotated clockwise by 270° among those included in the first input image 410.

[0075] The factors that do not affect the process may be different for each process. The method by which the electronic device transforms the first input image 410 into the second input image 420 may depend on the factors that do not affect the process.

[0076] The following description refers to Figure 5 a method for the electronic device according to an exemplary embodiment to perform an inverse transformation on an output image.

[0077] Figure 5 FIG. is a diagram showing a method for the electronic device according to an exemplary embodiment to perform an inverse transformation on an output image. The electronic device according to an exemplary embodiment may be Figure 1 the electronic device 100. Figure 5 The first input image 410 and the second input image 420 shown may correspond to the first input image 410 and the second input image 420 shown, respectively. Figure 4 respectively.

[0078] Refer to Figure 5, the electronic device can obtain a second input image 420 by transforming a first input image 410. The transformation from the first input image 410 to the second input image 420 can be referred to as "T". For example, "T" can be that the pattern is moved -5 in the x-axis direction and +1 in the y-axis direction, is horizontally flipped, and is rotated 270° clockwise. The electronic device can obtain a second input image 420 including the pattern that is moved -5 in the x-axis direction and +1 in the y-axis direction, is horizontally flipped, and is rotated 270° clockwise in the first input image 410.

[0079] The electronic device can obtain a second output image 430 by inputting the second input image 420 into the process simulation model 300. The electronic device can obtain an inverse-transformed second output image 440 by performing an inverse transformation on the second output image 430. For example, the electronic device can reverse the transformation from the first input image 410 to the second input image 420 on the second output image 430. That is, the inverse transformation on the second output image 430 can be referred to as "T -1 ". For example, "T -1 " can be that the pattern is rotated 270° counterclockwise, is horizontally flipped, and is moved +5 in the x-axis direction and -1 in the y-axis direction. The electronic device can obtain an inverse-transformed second output image 440 including the pattern that is rotated 270° counterclockwise, is horizontally flipped, and is moved +5 in the x-axis direction and -1 in the y-axis direction in the second output image 430.

[0080] The following description refers to Figure 6 to describe a method for an electronic device to train a process simulation model according to an exemplary embodiment.

[0081] Figure 6 is a diagram showing a method for an electronic device to train a process simulation model according to an exemplary embodiment. The electronic device according to the exemplary embodiment can be Figure 1 the electronic device 100.

[0082] Referring to Figure 6 , the electronic device can randomly Figure 4The paired first input images 410 and second input images 420 are sampled to generate multiple mini - batches, where the second input images 420 are transformed from the first input images 410. For example, the electronic device can generate m mini - batches of size n. For example, the electronic device can extract n first input images 410 from multiple first input images 410, and extract n second input images 420 corresponding to each of the n first input images 410 extracted from multiple second input images 420. Hereinafter, the n extracted first input images 410 can be referred to as the first input mini - batch 410mb, and the corresponding n extracted second input images 420 can be referred to as the second input mini - batch 420mb.

[0083] The following description illustrates a method for the electronic device to train a process simulation model for one mini - batch.

[0084] The electronic device can generate an input mini - batch 400mb that is a pair of the first input mini - batch 410mb and the second input mini - batch 420mb. The input mini - batch 400mb can include a total of 2n images. The electronic device can obtain an output mini - batch 401mb by inputting the input mini - batch 400mb into the process simulation model 300. The output mini - batch 401mb can include a first output mini - batch 450mb and a second output mini - batch 430mb. The first output mini - batch 450mb can correspond to the first input mini - batch 410mb and include n first output images. The second output mini - batch 430mb can correspond to the second input mini - batch 420mb and include n second output images.

[0085] The electronic device can obtain an inverse - transformed second output mini - batch 440mb by performing an inverse transformation on the second output images of the second output mini - batch 430mb. The inverse - transformed second output mini - batch 440mb can include n inverse - transformed second output images. The description has referred to Figure 5 The method for the electronic device to perform an inverse transformation on the second output image has been described in detail, so its redundant description is omitted.

[0086] The electronic device can generate an actual mini - batch 460mb including n images of actual patterns respectively corresponding to the n extracted first input images 410. The images of the actual patterns included in the actual mini - batch 460mb can be n images obtained by enhancing the measured images of the actual patterns formed on the wafer after performing a process on the design pattern. The electronic device can obtain n images of actual patterns respectively corresponding to the n extracted first input images 410 by enhancing the measured images in the same way as the enhancement method for extracting the n first input images 410 from the images of the design pattern.

[0087] The electronic device can obtain the value of the first loss function L1 by comparing the first output mini-batch 450mb with the actual mini-batch 460mb. The electronic device can obtain the value of the first loss function L1 based on the difference between the n first output images included in the first output mini-batch 450mb and the n images of the actual patterns included in the corresponding actual mini-batch 460mb.

[0088] The electronic device can obtain the value of the second loss function L2 by comparing the first output mini-batch 450mb with the second output mini-batch 440mb after inverse transformation. The electronic device can obtain the value of the second loss function L2 based on the difference between the n first output images included in the first output mini-batch 450mb and the n second output images after inverse transformation respectively corresponding to them and included in the second output mini-batch 440mb after inverse transformation.

[0089] The electronic device can adjust the parameters of the process simulation model to reduce the values of the first loss function L1 and the second loss function L2. The electronic device can update the parameters of the process simulation model to the parameters for minimizing the sum of the value of the first loss function L1 and the value of the second loss function L2. For example, the electronic device can update the parameters of the process simulation model 300 to the parameter values for minimizing the sum of the value of the first loss function L1 multiplied by the ratio c1 corresponding to the first loss function L1 and the value of the second loss function L2 multiplied by the ratio c2 corresponding to the second loss function L2.

[0090] Figure 7 FIG. is a diagram showing an example of a computer device implementing an electronic device according to an exemplary embodiment. Figure 1 The electronic device 100 of Figure 7 can be implemented by the computer device 700 shown in

[0091] Referring to Figure 7 , the computer device 700 may include a memory 710, a processor 720, a communication interface 730, and an input / output interface 740.

[0092] The memory 710 is a computer-readable recording medium and may include a random access memory (RAM), a read-only memory (ROM), and a permanent mass storage device such as a disk drive. Additionally, the memory 710 may store an operating system and at least one program code. These software components may be loaded into the memory 710 from a computer-readable recording medium separate from the memory 710. The separate computer-readable recording medium may include a computer-readable recording medium such as a hard disk, a flash memory, an optical disk, or an external hard disk. Additionally, these software components may be loaded into the memory 710 through the communication interface 730.

[0093] The processor 720 can process the instructions of a computer program by performing basic arithmetic, logical, and input / output operations. The instructions can be provided to the processor 720 from the memory 710 or via the communication interface 730.

[0094] The communication interface 730 can provide functions for the computer device 700 to communicate with another device via the network 800. Here, the communication method of the communication interface 730 is not limited and can include short-range wireless communication between devices and communication methods using communication networks (e.g., mobile communication networks, wired Internet, wireless Internet, or broadcast networks) that can be included in the network 800. For example, the network 800 can include at least one arbitrary network among networks such as personal area network (PAN), local area network (LAN), campus area network (CAN), metropolitan area network (MAN), wide area network (WAN), broadband network (BBN), and Internet. Additionally, the network 800 can include at least one arbitrary network having a network topology such as a bus network, star network, ring network, mesh network, star-bus network, tree, or hierarchical network, and is not limited thereto.

[0095] The input / output interface 740 can be used as an interface that can send commands / instructions or data input from a user or the input / output device 750 to another component (or other components) of the computer device 700. Additionally, the input / output interface 740 can output instructions or data received from another component (or other components) of the computer device 700 to the user or the input / output device 750. For example, the input / output device 750 can include input devices such as a microphone, keyboard, or mouse, and the input / output device 750 can include output devices such as a display or speaker.

[0096] The above exemplary embodiments can be implemented in the form of a computer program that can be run by various components on a computer, and such a computer program can be recorded on a computer-readable medium. Here, the medium can include the following items that are specifically configured to store and run program instructions: magnetic media such as hard disks, floppy disks, or magnetic tapes; optical recording media such as compact disc read-only memory (CD-ROM) or digital versatile disc (DVD); magneto-optical media such as floppy disks; or hardware devices such as read-only memory (ROM), random access memory (RAM), or flash memory.

[0097] Unless there is an explicit order or description to the contrary regarding the steps or operations of configuring the method according to the exemplary embodiments, the steps or operations can be performed in an appropriate order. The present disclosure is not necessarily limited by the order of the above steps or operations.

[0098] The use of any examples or example terms (such as etc.) in the specification is only intended to describe the present disclosure in detail and does not limit the scope of the present disclosure. Additionally, those skilled in the art will understand that various modifications, combinations, and changes can be made within the scope of the patent claims or their equivalents.

[0099] Any functional blocks shown and described above in the figures can be implemented by a processing circuit such as hardware including logic circuits, a hardware / software combination such as a processor running software, or a combination thereof. For example, the processing circuit can more specifically include, but is not limited to, a central processing unit (CPU), an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), a system on a chip (SoC), a programmable logic unit, a microprocessor, an application specific integrated circuit (ASIC), etc.

[0100] Although some example embodiments of the present disclosure have been described in detail above, the scope of the present disclosure is not limited thereto and can include various modifications and variations made by those skilled in the art and using the basic concept of the present disclosure as defined in the claims.

Claims

1. An electronic device, the electronic device comprising: a memory configured to store instructions; and a processing circuit configured to run the instructions stored in the memory to cause the electronic device to obtain a first input image based on an image of a design pattern to be formed on a wafer, to transform the first input image into a second input image based on factors that do not affect the process of manufacturing a semiconductor device, and to train a process simulation model using pairs of the first input image and the second input image.

2. The electronic device according to claim 1, wherein the factors that do not affect the process include at least one of translation, rotation, or flipping of the design pattern, and the processing circuit is configured to run the instructions stored in the memory to further cause the electronic device to obtain the second input image by performing at least one of the translation, the rotation, or the flipping on the design pattern included in the first input image.

3. The electronic device according to claim 1, wherein the processing circuit is configured to run the instructions stored in the memory to further cause the electronic device to obtain a pair of first output images and second output images by inputting the pair of the first input image and the second input image into the process simulation model, the first output image corresponding to the first input image, and the second output image corresponding to the second input image.

4. The electronic device according to claim 3, wherein the processing circuit is configured to run the instructions stored in the memory to further cause the electronic device to train the process simulation model based on the difference between the first output image and an image of an actual pattern formed on the wafer.

5. The electronic device according to claim 4, wherein the processing circuit is configured to run the instructions stored in the memory to further cause the electronic device to perform an inverse transformation on the second output image, and to further train the process simulation model based on the difference between the first output image and the inversely transformed second output image, and the inverse transformation is to reverse the transformation from the first input image to the second input image.

6. The electronic device according to claim 4, wherein the processing circuit is configured to run the instructions stored in the memory to further cause the electronic device to obtain the first input image by enhancing the image of the design pattern, and to obtain the image of the actual pattern by enhancing an image obtained by measuring the wafer on which the process has been performed based on the design pattern.

7. The electronic device according to claim 1, wherein the processing circuit is configured to run the instructions stored in the memory to further cause the electronic device to generate a plurality of mini - batches by randomly sampling multiple pairs of the first input image and the second input image, and Use the plurality of mini - batches to train the process simulation model.

8. The electronic device according to claim 1, wherein, The processing circuit is configured to run the instructions stored in the memory to further cause the electronic device to predict a process result based on the trained process simulation model.

9. An electronic device, the electronic device comprising: A memory configured to store instructions; And A processing circuit configured to run the instructions stored in the memory to cause the electronic device To obtain a first input image based on an image of a design pattern to be formed on a wafer, To transform the first input image into a second input image based on factors that do not affect the process of manufacturing a semiconductor device, To obtain a pair of output images by inputting the pair of the first input image and the second input image into a process simulation model, and To train the process simulation model to reduce the value of a first loss function based on the difference between one of the pair of output images and an image of an actual pattern formed on the wafer, and to reduce the value of a second loss function based on the difference between the pair of output images.

10. The electronic device according to claim 9, wherein, The factors that do not affect the process include at least one of translation, rotation, or flipping of the design pattern, and The processing circuit is configured to run the instructions stored in the memory to further cause the electronic device to obtain the second input image by performing at least one of the translation, the rotation, or the flipping on the design pattern included in the first input image.

11. The electronic device according to claim 9, wherein, The processing circuit is configured to run the instructions stored in the memory to further cause the electronic device To perform an inverse transformation on the other of the pair of output images corresponding to the second input image to generate an inverse - transformed output image, and To compare the inverse - transformed output image with one of the pair of output images corresponding to the first input image, and The inverse transformation is to reverse the transformation from the first input image to the second input image.

12. The electronic device according to claim 9, wherein, The processing circuit is configured to run the instructions stored in the memory to further cause the electronic device To generate a plurality of mini - batches by randomly sampling multiple pairs of the first input image and the second input image, and To use the plurality of mini - batches to train the process simulation model.

13. A method for operating an electronic device, the method comprising: Obtaining a first input image based on an image of a design pattern to be formed on a wafer; Transforming the first input image into a second input image based on factors that do not affect the process of manufacturing a semiconductor device; And Using the pair of the first input image and the second input image to train a process simulation model.

14. The method according to claim 13, wherein, The factors that do not affect the process include at least one of translation, rotation, or flipping of the design pattern, and transforming the first input image includes performing at least one of the translation, rotation, or flipping on the design pattern included in the first input image.

15. The method according to claim 13, wherein training the process simulation model includes obtaining a pair of first output images and second output images by inputting the pair of the first input image and the second input image into the process simulation model, the first output image corresponding to the first input image, and the second output image corresponding to the second input image.

16. The method according to claim 15, wherein training the process simulation model includes training the process simulation model based on the difference between the first output image and an image of the actual pattern formed on the wafer.

17. The method according to claim 16, wherein training the process simulation model includes: performing an inverse transformation on the second output image to generate an inversely transformed second output image; and further training the process simulation model based on the difference between the first output image and the inversely transformed second output image, and the inverse transformation is to reverse the transformation from the first input image to the second input image.

18. The method according to claim 16, wherein obtaining the first input image includes enhancing the image of the design pattern, and the method further includes obtaining the image of the actual pattern by enhancing an image obtained by measuring the wafer on which the process has been performed based on the design pattern.

19. The method according to claim 13, wherein training the process simulation model includes: generating a plurality of mini - batches by randomly sampling multiple pairs of the first input image and the second input image, and using the plurality of mini - batches to perform the training.

20. The method according to claim 13, the method further includes: predicting a process result based on the trained process simulation model.

Images