Method for optimizing a reticle map and related products

Abstract

The application provides a mask pattern optimization method and related products. The optimization method includes obtaining a target pattern of a photolithography process and a mask pattern to be optimized; simulating the photolithography process based on the mask pattern to obtain a simulation pattern; obtaining a difference degree of the simulation pattern and the target pattern, and obtaining a back propagation gradient for optimizing the mask pattern based on the difference degree; correcting the back propagation gradient using the transmittance of each part of the mask pattern; using the corrected back propagation gradient to iteratively optimize the transmittance of each part of the mask pattern with the minimum difference degree as the optimization target until the iteration stopping condition is met to obtain a target mask pattern. The method realizes fast and stable continuous transmittance promotion to the vicinity of the binary, the obtained target mask pattern is closer to the actual manufactured binary mask pattern, and the performance deviation between the actual manufactured mask pattern and the expected target is reduced.

Description

Technical Field

[0001] This invention relates to the field of integrated circuit technology, and in particular to a method for optimizing mask layouts, a computer-readable storage medium, a computer program product, and a computer device. Background Technology

[0002] As semiconductor manufacturing processes shrink to the nanometer or even sub-nanometer level, the optical proximity effect in photolithography becomes increasingly pronounced, causing severe distortion between the optical spatial image projected onto the wafer and the design layout. Reverse lithography is commonly used to eliminate this optical proximity effect. Through complex mathematical iterative optimization algorithms, reverse lithography calculates the geometry of the mask from the desired target pattern, making it one of the key technologies for achieving high-precision pattern transfer.

[0003] In the reverse lithography process, a crucial step is generating and optimizing sub-resolution auxiliary patterns near the main pattern on the photomask. These sub-resolution auxiliary patterns themselves do not image on the wafer, but they can effectively improve the imaging quality of the main pattern and the process window by altering the interference of the surrounding light field.

[0004] However, in the iterative optimization process of reverse lithography, the mask is typically used as a continuous transmission mask capable of producing continuous transmittance variations. In reality, the masks used for manufacturing are strictly binary (0 and 1), meaning that any region exists in only two physical states: "completely transparent" (e.g., quartz regions) and "completely opaque" (e.g., chromium layer regions). How to quickly and stably advance the continuous transmittance of the mask to near binary values ​​during the iterative optimization process remains a pressing problem to be solved. Summary of the Invention

[0005] One object of the present invention is to provide a method for optimizing a mask pattern, a computer-readable storage medium, a computer program product, and a computer device, so as to rapidly and stably advance the continuous transmittance to near binary during the iterative optimization process of the mask pattern, thereby reducing the performance deviation between the actually manufactured mask pattern and the expected target.

[0006] Specifically, according to one aspect of the present invention, the present invention provides a method for optimizing a mask layout, comprising: Obtain the target pattern and the mask pattern to be optimized for the photolithography process, wherein the mask pattern includes a continuous transmission mask pattern; Based on the mask pattern, the photolithography process is simulated to obtain a simulated pattern; The difference between the simulation pattern and the target pattern is obtained, and the backpropagation gradient for optimizing the mask pattern is obtained based on the difference. The reverse propagation gradient is corrected using the transmittance at various points on the mask pattern; With the minimum difference as the optimization objective, the transmittance at each point of the mask pattern is iteratively optimized using the modified backpropagation gradient until the iteration stopping condition is met, thus obtaining the target mask pattern.

[0007] Optionally, the step of simulating the photolithography process based on the mask pattern to obtain a simulated pattern includes: Obtain the S-shaped function used for nonlinear transformation of transmittance; The S-shaped function is used to convert the transmittance at various points on the mask pattern; The photolithography process is simulated using the converted mask pattern to obtain the simulated pattern.

[0008] Optionally, correcting the reverse propagation gradient using the transmittance at various points on the mask pattern includes: Obtain the slope of the transmittance at various points on the transformed mask pattern on the S-curve function; The reverse propagation gradient with a transmittance slope less than a preset lower threshold is enhanced and corrected, while the reverse propagation gradient with a transmittance slope greater than a preset upper threshold is weakened and corrected.

[0009] Optionally, obtaining the S-shaped function for nonlinear binary transformation of transmittance includes: Obtain the initial sigmoid function; Based on the transmittance distribution at various locations in the mask pattern, the values ​​of preset parameters for the S-shaped function are determined. These preset parameters are used to control the width and position of the transition region of the S-shaped function. The transition region refers to the region where the slope is greater than a set slope threshold.

[0010] Optionally, the step of simulating the photolithography process using the converted mask pattern to obtain a simulated pattern includes: The converted mask layout is rasterized to obtain a mask image; Using a defined optical model, the optical spatial image of the mask image in the exposure process is calculated; Using a predefined photoresist model, the photoresist image of the optical spatial image in the development process is calculated as the simulation pattern.

[0011] Optionally, the S-shaped function includes a logistic function, a hyperbolic tangent function, or a piecewise linear function used to fit an S-shaped curve.

[0012] Optionally, the iteration stopping condition includes: The difference between the simulated pattern and the target pattern is less than a set difference threshold; and / or The rate of change of the backpropagation gradient obtained by continuously setting the number of iterations is less than the set rate of change threshold.

[0013] According to another aspect of the present invention, a computer-readable storage medium is also provided, on which a computer program is stored, wherein the computer program, when executed by a processor, implements the steps of the mask layout optimization method described above.

[0014] According to another aspect of the present invention, a computer program product is also provided, comprising a computer program that, when executed by a processor, implements the steps of the mask layout optimization method described above.

[0015] According to another aspect of the present invention, a computer device is also provided, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the mask layout optimization method described above.

[0016] The mask pattern optimization method of this invention, during the iterative optimization process, corrects the backpropagation gradient by using the transmittance at various points on the mask pattern, and then uses the corrected backpropagation gradient to modify and update the mask pattern, thereby achieving a rapid and stable advancement of continuous transmittance to near binary values. The final target mask pattern obtained by this optimization method has its transmittance concentrated around binary values ​​(0 and 1), making the target mask pattern closer to the actual manufactured binary mask, and reducing the performance deviation between the actual manufactured mask pattern and the expected target.

[0017] The above and other objects, advantages and features of the present invention will become more apparent to those skilled in the art from the following detailed description of specific embodiments of the invention in conjunction with the accompanying drawings. Attached Figure Description

[0018] The following sections will describe some specific embodiments of the invention in detail by way of example and not limitation, with reference to the accompanying drawings. The same reference numerals in the drawings denote the same or similar parts or portions. Those skilled in the art should understand that these drawings are not necessarily drawn to scale. In the drawings: Figure 1 This is a flowchart illustrating a method for optimizing a mask layout according to an embodiment of the present invention; Figure 2 This is a schematic flowchart illustrating the optimization method for converting a mask layout according to an embodiment of the present invention. Figure 3 This is a flowchart illustrating the optimization method for obtaining the sigmoid function according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the process of performing photolithography process simulation according to an embodiment of the present invention; Figure 5This is a flowchart illustrating the process of correcting the backpropagation gradient according to an embodiment of the present invention. Figure 6 This is a schematic diagram of a computer program product according to an embodiment of the present invention; Figure 7 This is a schematic diagram of a computer-readable storage medium according to an embodiment of the present invention; and Figure 8 This is a schematic diagram of a computer device according to an embodiment of the present invention. Detailed Implementation

[0019] Sub-Resolution Assistant Feature (SRAF) is a sub-resolution pattern inserted into the surrounding area of ​​a sparsely patterned mask to reduce process variations caused by different pattern densities in integrated circuit layouts. It improves depth of focus and process window uniformity. SRAFs are smaller than the imaging resolution of the lithography system and are typically thin, elongated rectangular lines parallel to the main mask pattern. While SRAFs themselves do not form a lithographic pattern during exposure, they do influence the intensity distribution of the lithographic imaging of the nearby main pattern.

[0020] In the Inverse Lithography (ILT) process, a crucial step is generating and optimizing sub-resolution auxiliary patterns near the main pattern on the mask. The core challenge in optimizing SRAF lies in simultaneously satisfying the dual constraints of optical imaging performance and mask fabrication feasibility. Specifically, it involves balancing the use of a continuous transmission mask (CTM) with continuously varying transmittance during optimization with the use of a strictly binary mask during fabrication.

[0021] Currently, the relevant technologies mainly employ two approaches. One is model-based continuous optimization and post-processing binarization, and the other is real-time binarization optimization based on the level set method. A brief explanation follows.

[0022] The model-based continuous optimization and post-processing binarization process is as follows: First, the mask is initialized based on the target layout. Then, iterative optimization is performed in the continuous transmittance space (0≤T≤1) using gradient descent to minimize the mean square error between the simulated and target light intensities, generating a freeform SRAF region with continuous grayscale. Finally, polygonal SRAFs are extracted through global / local threshold binarization and Boolean operations. Although this method can achieve the theoretically optimal light intensity distribution, it suffers from a continuous-discrete mismatch: the optimization process assumes continuously adjustable transmittance, but the actual mask only supports binary states (T=0 or 1), and post-processing binarization simplifies the continuous gradient region to a single threshold cut. This operation introduces irreversible quantization error—abrupt changes in the gradient light field interference characteristics at the SRAF edges, leading to a significant increase in the mean square error (MSE) between the simulated and target light intensities. Experimental results show that quantization error increases wafer linewidth roughness (LWR), expands critical dimension uniformity (CDU) deviation, and ultimately results in excessive depth-of-focus reduction. This defect is essentially a disconnect between the optimization model and physical manufacturing.

[0023] Real-time binarization optimization based on the level set method mainly uses the level set function φ(x,y) to represent the mask state (φ>0 for transparent areas, φ<0 for shading areas, and φ=0 for boundaries), maintaining the characteristics of the binary mask in real time during optimization iterations. While this method avoids post-processing mismatch, it introduces manufacturing bottlenecks due to the free evolution of boundaries. In pursuit of extreme optical performance, the algorithm drives the evolution of the φ function without constraints, resulting in high-frequency jagged edges, micrometer-scale islands, and fragmented topology at the SRAF boundaries. Such irregular geometry dramatically increases the complexity of the mask data: the number of boundary segments that need to be processed during electron beam writing increases, and the writing path planning time is prolonged. Simultaneously, enhanced edge scattering reduces the optical detection signal-to-noise ratio, lowers the accuracy of defect identification, and increases the frequency of repair iterations. This defect stems from the singular optimization objective (focusing only on optical performance).

[0024] The purpose of the mask pattern optimization method in this embodiment is to quickly and stably advance the continuous transmittance to near binary during the iterative optimization process of the mask pattern, thereby reducing the performance deviation between the actual manufactured mask pattern and the expected target.

[0025] Figure 1 This is a flowchart illustrating a method for optimizing a mask layout according to an embodiment of the present invention. The method generally includes: S100: Obtain the target pattern and the mask pattern to be optimized for the photolithography process. The mask pattern includes a continuous transmission mask pattern. S200, based on the mask pattern, simulates the photolithography process to obtain a simulated pattern; S300, obtain the difference between the simulated pattern and the target pattern; S400: Determine whether the difference between the simulated pattern and the target pattern meets the iteration stopping condition; if not, proceed to S500; if so, proceed to S800. S500, with the minimum difference as the optimization objective, yields the backpropagation gradient used to optimize the mask layout; S600 uses the transmittance at various points on the mask pattern to correct the reverse-propagated gradient; S700, update the transmittance at various points on the mask pattern using the corrected reverse propagation gradient; return to execute S200; S800, stop iteration, and obtain the target mask pattern.

[0026] The target pattern in the photolithography process can be an optical spatial image (AI), which is the light intensity distribution of the image formed by the mask pattern on the ideal image plane of the optical lithography system. Correspondingly, in step S200, the optimization method is used to simulate the exposure process to obtain a simulated optical spatial image of the mask pattern to be optimized.

[0027] The target pattern in the photolithography process can also be a photoresist image (RI), which is the outline image formed on the photoresist after the mask pattern is exposed and developed. Correspondingly, in step S200, the optimization method is used to simulate the exposure and development process to obtain a simulated photoresist image of the mask pattern to be optimized.

[0028] The target pattern in the photolithography process can also be an etched image, that is, the outline image formed on the wafer after the mask pattern is exposed, developed, and etched. Correspondingly, in step S200, the optimization method is used to simulate the exposure, development, and etching processes to obtain a simulated etched image of the mask pattern to be optimized.

[0029] In this embodiment, the mask pattern to be optimized includes a continuous transmission mask pattern, where the values ​​at various points represent transmittance values, which can be between 0 and 1. The initial mask pattern to be optimized can be generated based on the target pattern using reverse lithography. For example, a uniformly distributed initial mask pattern to be optimized can be generated based on the target pattern, wherein the main pattern area of ​​the mask can be set to an intermediate transmittance value (e.g., 0.5), and the background area can be set to the same value (e.g., 0, 0.1, etc.). This provides a stable starting point for subsequent optimization, avoiding initial deviations from affecting convergence during the optimization process.

[0030] Next, the photolithography process can be simulated directly using the mask pattern, or the transmittance at various points on the mask pattern can be constrained and transformed before photolithography process simulation to obtain the simulation pattern for this round. This can be achieved using a rigorous photolithography simulation model based on physical and chemical reactions, or a photolithography simulation model based on deep learning.

[0031] Next, the difference between the simulated pattern and the target pattern is calculated. For example, the edge placement error (EPE) between the simulated pattern and the target pattern is used as the difference between each optimization round.

[0032] Next, determine if the difference meets the iteration stopping condition. The iteration stopping condition may include whether the difference has converged, whether the difference is less than a set difference threshold, or whether the number of iterations has reached a set maximum. If the iteration stopping condition is met, stop the iteration, and use the mask pattern with the smallest difference in all iterations as the target mask pattern.

[0033] If the iteration stopping condition is not met, the mask pattern needs to be modified and updated to reduce the difference. Related techniques, such as model-based continuous optimization and post-processing binarization schemes, typically generate a loss function based on the difference, then calculate the backpropagation gradient of the loss value with the minimum loss value as the optimization objective, and use the backpropagation gradient to modify and update the transmittance at various points on the mask pattern. While this scheme can quickly reduce the difference between the simulated pattern and the target pattern, the optimized mask pattern still maintains continuous transmittance. Therefore, post-processing binarization is required after optimization, leading to a continuous-discrete mismatch.

[0034] In this embodiment, when the reverse propagation gradient is propagated to the mask pattern, the reverse propagation gradient is corrected based on the transmittance at various points on the mask pattern. Then, the corrected reverse propagation gradient is used to modify and update the mask pattern, thereby achieving a rapid and stable advancement of continuous transmittance to near binary values. The reverse propagation gradient correction method of this embodiment will be described in detail below.

[0035] To rapidly advance continuous transmittance to near binary values, the following backpropagation gradient correction method can be used: On the one hand, for regions where transmittance is close to a binary state (0 and 1), the backpropagation gradient is weakened, thereby reducing the modification magnitude and maintaining its proximity to a binary state. This avoids repeated fluctuations in regions already close to binary during optimization, improving the stability of the optimization process. On the other hand, for transitional regions where transmittance is in the middle value (within a preset range around 0.5), the backpropagation gradient is strengthened, increasing the modification and update magnitude, thereby reinforcing the faster convergence of transmittance in the intermediate transitional region towards binary values.

[0036] For example, multiple thresholds A, B, C, and D can be set, where A > B > C > D, to weaken the backpropagation gradient in near-binary regions where the transmittance is greater than threshold A or less than threshold D, and to enhance the backpropagation gradient in intermediate regions where the transmittance is less than threshold B and greater than threshold C.

[0037] For example, nonlinear transformation functions (such as logistic functions, hyperbolic tangent functions, etc.) can be used to transform the transmittance at various points on the mask pattern. The slope of each transmittance value on the nonlinear transformation function is then used as a correction coefficient to enhance or weaken the backpropagation gradient. Taking the logistic function as an example, its slope (i.e., derivative) follows a logistic distribution. Its distribution characteristic is that the slope is smaller in regions closer to binary (0 and 1), and larger in regions closer to the intermediate value. By using the slope of the logistic function at various points as correction coefficients to adjust the backpropagation gradient, it is possible to continuously weaken the backpropagation gradient in regions where transmittance is close to binary and enhance the backpropagation gradient in regions where transmittance is close to the intermediate value. Through multiple iterations, this allows for the rapid advancement of continuous transmittance to near the binary level.

[0038] After updating the transmittance of each part of the mask pattern using the corrected backpropagation gradient, the updated mask pattern is simulated in the next iteration to obtain the simulation pattern. Then, steps such as obtaining the difference between the simulation pattern and the target pattern are performed until the iteration stops when the iteration stopping condition is met.

[0039] The mask pattern optimization method in this embodiment, during the iterative optimization process, corrects the backpropagation gradient by using the transmittance at various points on the mask pattern, and then uses the corrected backpropagation gradient to modify and update the mask pattern. This achieves rapid and stable advancement of continuous transmittance to near binary values. The transmittance at various points on the target mask pattern obtained by this optimization method is concentrated near binary values ​​(0 and 1), thereby making the target mask pattern closer to the actual manufactured binary mask and reducing the performance deviation between the actual manufactured mask pattern and the expected target.

[0040] In some embodiments of the optimization method of the present invention, such as Figure 2 As shown, based on the mask pattern, the photolithography process is simulated to obtain the simulated pattern, including: S211, obtain the S-shaped function used for nonlinear transformation of transmittance; S213 uses an S-shaped function to convert the transmittance at various points on the mask pattern; S215, use the converted mask pattern to simulate the photolithography process and obtain the simulated pattern.

[0041] As mentioned above, the transmittance in a continuous transmission mask pattern is continuously distributed, and there may be transmittance values ​​that deviate significantly from binary (0 and 1). Directly performing photolithography process simulation can easily lead to a continuous-discrete mismatch. To solve this problem, this embodiment uses a sigmoid function to transform the transmittance at various points on the mask pattern before performing photolithography process simulation, bringing the transmittance values ​​closer to binary to more closely resemble the actual manufactured binary mask. Performing photolithography process simulation after this transformation results in simulation results that more realistically reflect the actual imaging effect of the binary mask, thus providing accurate feedback for subsequent optimization.

[0042] The sigmoid function can be the logistic function, the hyperbolic tangent function (tanh), or a piecewise linear function used to fit an S-shaped curve.

[0043] Logistic stethoscopes are of the form The logistic function is a function of x, where x is the transmittance of the mask pattern before nonlinear transformation, and M is the transmittance after nonlinear transformation. The logistic function has the following characteristics: as x approaches positive infinity, the value of M approaches 1, and as x approaches negative infinity, the value of M approaches 0; it is smooth and differentiable everywhere; its derivative is always non-negative, and it has one and only one inflection point; especially due to the exponential function e -x Due to its characteristics, the value of M will quickly approach the binary endpoints.

[0044] The hyperbolic tangent function has an output range of [-1, 1], therefore additional offset and scaling are needed to constrain the range to the [0, 1] interval. Using the hyperbolic tangent function to transform the transmittance of a mask pattern has the advantage of better gradient symmetry. The disadvantage is the need for two additional linear transformation steps, slightly complicating the computation. Furthermore, under the same steepness parameters, the hyperbolic tangent function has a narrower transition region, which may lead to slightly poorer convergence stability of the optimization algorithm.

[0045] The piecewise linear function used to fit the S-curve replaces the S-curve with a piecewise linear function (such as a "trapezoidal" transition). This approach is computationally more efficient (requiring only conditional judgments). However, because it is not differentiable, a subgradient approximation must be used in backpropagation, which may cause oscillations in the optimization process and affect the convergence speed of the optimization algorithm.

[0046] It should be noted that using the sigmoid function to transform the transmittance of the mask pattern will inevitably affect the simulation results, and thus affect the backpropagation gradient. Taking the sigmoid function as the logistic function as an example, x is the transmittance value of the mask pattern before the nonlinear transformation, and M is the transmittance value after the nonlinear transformation, where M is used for lithography process simulation. Therefore, when calculating the backpropagation gradient, according to the chain rule, the gradient of the loss function L with respect to x is: The derivative of M(x) is, for example: Therefore, the backpropagation gradient calculated from the loss function L is: Observing the above formula, when the mask transmittance M of a certain region is close to 0 or 1, it indicates that the region is close to a binary state, and M(1-M)≈0. That is, the gradient of this region will be significantly reduced. This can avoid repeated fluctuations in regions that are already close to binary during the optimization process, thus improving the stability of the optimization process. When the mask transmittance M of a certain region is close to 0.5, it indicates that the region is still in an intermediate grayscale transition state. At this time, M(1-M) reaches its maximum value, that is, the gradient of this region is significantly enhanced, and the optimization algorithm will update these intermediate value regions more strongly, pushing them to converge towards the binary state of 0 or 1. The transmittance conversion and backpropagation gradient correction method in this embodiment only requires the addition of standard function call operations, with extremely low computational overhead, and can be seamlessly integrated into mainstream computational lithography software platforms. It can also be applied to existing advanced lithography processes without special hardware support.

[0047] In some embodiments of the optimization method of the present invention, such as Figure 3 As shown, the S-shaped function used for nonlinear binary transformation of transmittance is obtained, including: S231, obtain the initial sigmoid function; S233, based on the transmittance distribution at various locations on the mask pattern, determines the values ​​of preset parameters for the S-shaped function. The preset parameters are used to control the width and position of the transition region of the S-shaped function. The transition region refers to the region where the slope is greater than the set slope threshold.

[0048] This embodiment modifies the S-shaped function to intentionally control the width and position of the transition region during the transmittance conversion process, making it compatible with the transmittance distribution across the mask pattern. This achieves stable and controllable mask constraint conversion, thereby avoiding output saturation issues, improving embedding discrimination, and ultimately enhancing the simulation and subsequent optimization effects of the lithography process.

[0049] Taking the sigmoid function as an example of the logistic function, it can be expressed as: The preset parameter α is the logistic growth rate or the steepness of the curve, and the preset parameter θ is the x-value of the midpoint of the S-curve. For example, the slope threshold can be set to 1 / 2, 1 / 3, etc., of the maximum slope of M(x) (when x=θ).

[0050] α is mainly used to control the width of the transition region. When α is small, the S-curve is flatter, and the region where the mask transmittance transitions from 0 to 1 is wider, making the optimization process smoother; when α is large, the S-curve is steeper, the transition region becomes narrower, and the mask transmittance is closer to the hard binary state.

[0051] θ is primarily used to control the position of the transition region. When x = θ, the converted mask transmittance satisfies M = 0.5. Therefore, θ corresponds to the center position of the transition region. Adjusting θ can shift the entire transition region towards the transparent or opaque region, thereby controlling the boundary position of the mask transmittance. For example, when θ = 0.5, the system uses x = 0.5 as the binarization center; when θ is appropriately increased, a larger x is needed to make the mask transmittance approach 1; when θ is appropriately decreased, the mask transmittance converges to 1 more easily.

[0052] In practical applications, the appropriate values ​​for preset parameters such as α and θ can be dynamically determined based on the transmittance distribution across the mask pattern, mask mesh size, target pattern complexity, and optimization stability. For example, α can be set between 10 and 50, and θ between 0.45 and 0.55. This parameter range can be determined experimentally to ensure that the conversion process is neither too gradual, leading to insufficient binarization, nor too steep, resulting in unstable gradients.

[0053] For example, when the transmittance distribution across the mask pattern is relatively uniform, the value of α can be appropriately reduced to expand the transition region and prevent distortion of the simulation results. After multiple iterations of optimization, if the transmittance distribution across the mask pattern has become biased towards a binary region, the transition region can be narrowed to allow more mask transmittance to approach a hard binary state. For instance, in each iteration, the transmittance distribution characteristics of the main graphic region in the mask pattern can be obtained, and the value of θ can be determined based on these characteristics, thereby controlling the convergence method of the mask transmittance.

[0054] In some embodiments of the optimization method of the present invention, such as Figure 4 As shown, the photolithography process is simulated using the converted mask pattern to obtain the simulated pattern, including: S251, the converted mask layout is rasterized to obtain the mask image; S253, using a set optical model, calculates the optical spatial image of the mask image in the exposure process; S255 uses a pre-defined photoresist model to calculate the photoresist image of the optical spatial image in the developing process, as a simulation pattern.

[0055] In this embodiment, the simulated photolithography process is the exposure and development process. Specifically, firstly, the mask pattern after S-curve transformation is rasterized into a mask image supported by an optical model, and the optical model outputs an optical spatial image. The optical model can be a rigorous photolithography simulation model based on physical and chemical reactions, a photolithography simulation model based on deep learning, etc. Next, the optical spatial image is input into a photoresist model, and the photoresist model outputs a photoresist image. The photoresist model can also be a rigorous photolithography simulation model based on physical and chemical reactions, a photolithography simulation model based on deep learning, etc.

[0056] In some embodiments, the simulation model combines the functions of an optical model and a photoresist model. When in use, a mask image is input into the simulation model, and the model can directly output a photoresist image. The simulation model can be a rigorous photoresist simulation model based on physical and chemical reactions, a photoresist simulation model based on deep learning, etc.

[0057] In some embodiments of the optimization method of the present invention, such as Figure 5 As shown, the backpropagation gradient is corrected using the transmittance at various points on the mask pattern, including: S611, obtain the slope of the transmittance at various points on the S-shaped function of the converted mask pattern; S613, enhance the backpropagation gradient with a transmittance slope less than a preset lower threshold, and weaken the backpropagation gradient with a transmittance slope greater than a preset upper threshold.

[0058] Let's continue using the sigmoid function as an example of the logistic function. The sigmoid curve of the logistic function has a lower slope near the binary region. Using it to correct the backpropagation gradient may cause the gradient vanishing problem, which is not conducive to correcting the mask transmittance in the binary region. Furthermore, the slope of the sigmoid curve is higher in the middle region. Using it to correct the backpropagation gradient may result in an excessively large update amplitude.

[0059] In this embodiment, a lower threshold (e.g., 0.1, 0.01) and an upper threshold for the slope can be set in the early rounds of iterative optimization, and a real-time gradient magnitude monitoring module can be set to monitor the slope. When the slope is less than the preset lower threshold, scaling or residual connection is automatically triggered to enhance and correct it, thereby preventing gradient vanishing and ensuring efficient convergence of the optimization. When the slope is greater than the preset upper threshold, it is weakened and corrected to avoid excessive update amplitude and ensure the stability of the optimization process.

[0060] In some embodiments of the optimization method of the present invention, the iteration stopping condition is: The difference between the simulated pattern and the target pattern is less than the set difference threshold, or The rate of change of the backpropagation gradient obtained by continuously setting the iteration rounds is less than the set rate of change threshold.

[0061] This embodiment uses a dual-conditional convergence determination based on the difference threshold and the gradient change rate to obtain better optimization results and save computing resources.

[0062] In some embodiments of the optimization method of the present invention, the iteration stopping condition further includes: The ratio of the near-binary region to the intermediate region of transmittance in the mask pattern is greater than or equal to a preset ratio.

[0063] The near-binary region refers to the area where the transmittance is greater than or equal to the upper limit of transmittance, or less than or equal to the lower limit of transmittance. The upper limit of transmittance can be, for example, 0.90 or 0.95, and the lower limit can be, for example, 0.10 or 0.05. The intermediate region refers to the area where the transmittance is less than the upper limit of transmittance but greater than the lower limit of transmittance.

[0064] If the area ratio of the near-binary region to the median region reaches the preset ratio, it indicates that the transmittance in the mask image is highly concentrated around 0 or 1, which is close to the actual manufactured binary mask. If the area ratio of the near-binary region to the median region does not reach the preset ratio, it indicates that the relative area of ​​the median region in the mask image is still large, and the gap from the actual manufactured binary mask is still large, requiring further iterative optimization.

[0065] The flowchart provided in this embodiment is not intended to indicate that the operations of the method will be performed in any particular order, or that all operations of the method are included in every case. Furthermore, the method may include additional operations. Within the scope of the technical concept provided by the method in this embodiment, additional variations can be made to the above method.

[0066] It should be understood that in some embodiments, the components may be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods may be implemented using software or firmware stored in memory and executed by a suitable instruction execution system.

[0067] This invention also provides a computer program product 10, a computer-readable storage medium 20, and a computer device 30. Figure 6 This is a schematic diagram of a computer program product 10 according to an embodiment of the present invention. Figure 7 This is a schematic diagram of a computer-readable storage medium 20 according to an embodiment of the present invention. Figure 8This is a schematic diagram of a computer device 30 according to an embodiment of the present invention. The computer program product 10 includes a computer program 11, which, when executed by the processor 32, implements the steps of the mask layout optimization method described above. A computer-readable storage medium 20 stores the computer program 11 thereon, which, when executed by the processor 32, implements the steps of the mask layout optimization method described above. The computer device 30 may include a memory 31, a processor 32, and the computer program 11 stored in the memory 31 and running on the processor 32.

[0068] The computer program 11 used to perform the operations of this invention may be assembly instructions, Instruction Set Architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, integrated circuit configuration data, or source code or object code written in any combination of one or more programming languages ​​and procedural programming languages. The computer program 11 may execute entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the latter case, the remote computer may be connected to the user's computer via any type of network, including a Local Area Network (LAN) or Wide Area Network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, to perform aspects of this invention, electronic circuits, including, for example, programmable logic circuits, Field-Programmable Gate Arrays (FPGAs), or Programmable Logic Arrays (PLAs), may execute computer-readable program instructions to personalize the electronic circuits by utilizing state information from computer-readable program instructions.

[0069] For the purposes of this embodiment, computer program product 10 is a related product that includes computer program 11.

[0070] For the purposes of this embodiment, the computer-readable storage medium 20 is a tangible device capable of holding and storing a computer program 11. It can be any device capable of containing, storing, communicating, propagating, or transmitting the computer program 11 for use by or in conjunction with an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of the computer-readable storage medium 20 include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable optical disc read-only memory (CD-ROM), digital versatile disc (DVD), memory stick, floppy disk, mechanical encoding device, and any suitable combination thereof.

[0071] Computer device 30 can be, for example, a server, desktop computer, laptop computer, tablet computer, or smartphone. In some examples, computer device 30 can be a cloud computing node. Computer device 30 can be described in the general context of computer system executable instructions (such as program modules) executed by a computer system. Typically, program modules can include routines, programs, object programs, components, logic, data structures, etc., that perform specific tasks or implement specific abstract data types. Computer device 30 can be implemented in a distributed cloud computing environment where tasks are performed by remote processing devices linked through a communication network. In a distributed cloud computing environment, program modules can reside on local or remote computing system storage media, including storage devices.

[0072] Computer device 30 may include a processor 32 adapted to execute stored instructions and a memory 31 that provides temporary storage space for the operation of said instructions during operation. The processor 32 may be a single-core processor, a multi-core processor, a computing cluster, or any other configuration. The memory 31 may include random access memory (RAM), read-only memory, flash memory, or any other suitable storage system.

[0073] Computer device 30 may also include a network adapter / interface and an input / output (I / O) interface. The I / O interface allows external devices that can be connected to the computer device to input and output data. The network adapter / interface provides communication between the computer device and a network, typically represented as a communication network.

[0074] Therefore, those skilled in the art should recognize that although numerous exemplary embodiments of the present invention have been shown and described in detail herein, many other variations or modifications conforming to the principles of the present invention can be directly determined or derived from the disclosure of the present invention without departing from the spirit and scope of the invention. Thus, the scope of the present invention should be understood and construed as covering all such other variations or modifications.

Claims

1. A method for optimizing a mask layout, characterized in that, include: Obtain the target pattern and the mask pattern to be optimized for the photolithography process, wherein the mask pattern includes a continuous transmission mask pattern; Based on the mask pattern, the photolithography process is simulated to obtain a simulated pattern; The difference between the simulation pattern and the target pattern is obtained, and the backpropagation gradient for optimizing the mask pattern is obtained based on the difference. The reverse propagation gradient is corrected using the transmittance at various points on the mask pattern; With the minimum difference as the optimization objective, the transmittance at each point of the mask pattern is iteratively optimized using the modified backpropagation gradient until the iteration stopping condition is met, thus obtaining the target mask pattern.

2. The optimization method according to claim 1, characterized in that, The step of simulating the photolithography process based on the mask pattern to obtain a simulated pattern includes: Obtain the S-shaped function used for nonlinear transformation of transmittance; The S-shaped function is used to convert the transmittance at various points on the mask pattern; The photolithography process is simulated using the converted mask pattern to obtain the simulated pattern.

3. The optimization method according to claim 2, characterized in that, The step of correcting the reverse propagation gradient using the transmittance at various points on the mask pattern includes: Obtain the slope of the transmittance at various points on the transformed mask pattern on the S-curve function; The reverse propagation gradient with a transmittance slope less than a preset lower threshold is enhanced and corrected, while the reverse propagation gradient with a transmittance slope greater than a preset upper threshold is weakened and corrected.

4. The optimization method according to claim 2, characterized in that, The acquisition of the S-shaped function for nonlinear binary transformation of transmittance includes: Obtain the initial sigmoid function; Based on the transmittance distribution at various locations in the mask pattern, the values ​​of preset parameters for the S-shaped function are determined. These preset parameters are used to control the width and position of the transition region of the S-shaped function. The transition region refers to the region where the slope is greater than a set slope threshold.

5. The optimization method according to claim 2, characterized in that, The process of simulating the photolithography process using the converted mask pattern to obtain a simulated pattern includes: The converted mask layout is rasterized to obtain a mask image; Using a defined optical model, the optical spatial image of the mask image in the exposure process is calculated; Using a predefined photoresist model, the photoresist image of the optical spatial image in the development process is calculated as the simulation pattern.

6. The optimization method according to claim 2, characterized in that, The S-shaped function includes the logistic function, the hyperbolic tangent function, or a piecewise linear function used to fit an S-shaped curve.

7. The optimization method according to claim 1, characterized in that, The iteration stopping conditions include: The difference between the simulated pattern and the target pattern is less than a set difference threshold; and / or The rate of change of the backpropagation gradient obtained by continuously setting the number of iterations is less than the set rate of change threshold.

8. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the steps of the method for optimizing a mask layout as described in any one of claims 1 to 7.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method for optimizing the mask layout as described in any one of claims 1 to 7.

10. A computer device, characterized in that, The method includes a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method for optimizing a mask layout according to any one of claims 1 to 7.

Images