Methods, apparatus, storage media and electronic devices for collaborative optimization of mask pattern quality and efficiency

Abstract

This application discloses a method, apparatus, storage medium, and electronic device for collaborative optimization of mask image quality and efficiency. The method includes: acquiring a mask image to be optimized; extracting a state feature vector of the current target image region for the mask image to be optimized; inputting the state feature vector into a collaborative optimization strategy model to obtain a current optimization action; optimizing the current target image region according to the current optimization action, and if the optimization termination condition is not met, returning to the step of extracting the state feature vector of the current target image region until the optimization termination condition is met, thus obtaining a target optimized image; and merging all target optimized images into a target mask image when all target image regions of the mask image to be optimized are optimized. This application can improve the writing efficiency of mask images while ensuring imaging quality.

Description

Technical Field

[0001] This application relates to the field of photolithography technology, specifically to a method, apparatus, storage medium, and electronic device for co-optimizing mask pattern quality and efficiency. Background Technology

[0002] As semiconductor process nodes continue to evolve towards 5nm, 3nm, and even more advanced processes, the physical limits of photolithography face increasingly severe challenges. Optical Proximity Correction (OPC) technology has become a crucial step in ensuring the accuracy of pattern imaging at the nanoscale. However, to compensate for pattern distortions caused by diffraction, interference, and other effects during photolithography (i.e., optical proximity effects), traditional OPC algorithms introduce a large number of complex auxiliary patterns and jagged edge modifications onto the original design pattern. Although this method significantly improves image fidelity, it inevitably and substantially increases the geometric complexity of the mask pattern.

[0003] This results in an exponential increase in the number of exposures and writing time required during the mask writing process in mask manufacturing, making mask manufacturing cost and cycle time a key bottleneck restricting the research and development and mass production of advanced chips.

[0004] Therefore, how to improve the writing efficiency of mask patterns while ensuring imaging quality has become a pressing technical problem in the current semiconductor manufacturing field. Summary of the Invention

[0005] This application provides a method, apparatus, storage medium, and electronic device for co-optimizing mask pattern quality and efficiency, which can improve the writing efficiency of mask patterns while ensuring imaging quality.

[0006] In a first aspect, embodiments of this application provide a method for collaborative optimization of mask image quality and efficiency, including:

[0007] Obtain the mask image to be optimized;

[0008] For the current target graphic region of the mask graphic to be optimized, extract the state feature vector of the current target graphic region;

[0009] The state feature vector is input into the collaborative optimization strategy model to obtain the current optimization action;

[0010] The current target graphic region is optimized according to the current optimization action. If the optimization termination condition is not met, the process returns to the step of extracting the state feature vector of the current target graphic region until the optimization termination condition is met, and the target optimized graphic is obtained.

[0011] When all target graphic regions of the mask graphic to be optimized are optimized, all the optimized target graphics are merged into a target mask graphic.

[0012] In the mask image quality and efficiency co-optimization method provided in this application embodiment, the step of extracting the state feature vector of the current target image region includes:

[0013] Rapid photolithography simulation is performed on the current target graphic region to obtain local optical simulation features and imaging quality evaluation features;

[0014] Perform geometric analysis on the current target graphic region to generate geometric features;

[0015] The local optical simulation features, the imaging quality assessment features, and the geometric features are concatenated into a state feature vector.

[0016] The mask pattern quality and efficiency co-optimization method provided in this application embodiment further includes, before obtaining the mask pattern to be optimized:

[0017] Construct a collaborative optimization strategy model.

[0018] In the mask image quality and efficiency co-optimization method provided in this application embodiment, the construction of the co-optimization strategy model includes:

[0019] A training environment and a training set are constructed. The training environment includes a photolithography simulation module and a write efficiency evaluation module, and the training set contains diverse mask pattern samples.

[0020] Initialize the policy network model;

[0021] The policy network model is iteratively trained based on the training set and the training environment until the policy network model converges, thereby generating a collaborative optimization policy model.

[0022] In the mask image quality and efficiency co-optimization method provided in this application embodiment, the step of iteratively training the policy network model based on the training set and the training environment until the policy network model converges includes:

[0023] For the current training mask image in the training set, extract the state feature vector of the current training mask image;

[0024] The state feature vector is input into the policy network model to generate training actions;

[0025] The training action is used in the training environment to optimize the current training mask image and to calculate the imaging quality index and write efficiency index.

[0026] Based on the imaging quality index and the writing efficiency index, the parameters of the policy network model are updated using a near-end policy optimization algorithm, and the process returns to the step of extracting the state feature vector of the current training mask image until the policy network model converges.

[0027] In the mask image quality and efficiency co-optimization method provided in this application embodiment, the step of updating the parameters of the policy network model based on the imaging quality index and the writing efficiency index using a near-end policy optimization algorithm includes:

[0028] Calculate the instantaneous reward value based on the imaging quality index and the writing efficiency index;

[0029] Based on the instant reward value, the parameters of the policy network model are updated using a near-end policy optimization algorithm.

[0030] In the mask image quality and efficiency co-optimization method provided in this application embodiment, the calculation formula for the instant reward value is as follows:

[0031] R=α·ΔQuality + β·ΔEfficiency

[0032] Where ΔQuality is the imaging quality index, ΔEfficiency is the writing efficiency index, and α and β are dynamically adjusted weighting coefficients.

[0033] Secondly, embodiments of this application provide a mask pattern quality and efficiency co-optimization device, comprising:

[0034] The acquisition unit is used to acquire the mask image to be optimized.

[0035] The extraction unit is used to extract the state feature vector of the current target graphic region for the current target graphic region of the mask graphic to be optimized;

[0036] The input unit is used to input the state feature vector into the collaborative optimization strategy model to obtain the current optimization action;

[0037] An optimization unit is used to optimize the current target graphic region according to the current optimization action, and if the optimization termination condition is not met, return to the step of extracting the state feature vector of the current target graphic region until the optimization termination condition is met, so as to obtain the target optimized graphic.

[0038] The output unit is used to merge all the target optimized graphics into a target mask graphic when all target graphic regions of the mask graphic to be optimized have been optimized.

[0039] Thirdly, this application provides a storage medium storing a plurality of instructions adapted for loading by a processor to execute the mask graphics quality and efficiency co-optimization method described in any of the preceding claims.

[0040] Fourthly, this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the mask graphic quality and efficiency co-optimization method described in any of the above claims.

[0041] In summary, the mask pattern quality and efficiency collaborative optimization method provided in this application includes: acquiring a mask pattern to be optimized; extracting a state feature vector of the current target pattern region of the mask pattern to be optimized; inputting the state feature vector into a collaborative optimization strategy model to obtain a current optimization action; optimizing the current target pattern region according to the current optimization action, and if the optimization termination condition is not met, returning to the step of extracting the state feature vector of the current target pattern region until the optimization termination condition is met to obtain a target optimized pattern; and merging all target optimized patterns into a target mask pattern when all target pattern regions of the mask pattern to be optimized are optimized. This application embodiment can improve the writing efficiency of mask patterns while ensuring imaging quality. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 This is a schematic diagram illustrating an application scenario of the mask graphic quality and efficiency collaborative optimization method provided in the embodiments of this application.

[0044] Figure 2 This is a flowchart illustrating the mask pattern quality and efficiency collaborative optimization method provided in the embodiments of this application.

[0045] Figure 3 This is a flowchart illustrating the collaborative optimization strategy model construction method provided in this application embodiment.

[0046] Figure 4 This is a schematic diagram of the structure of the mask pattern quality and efficiency co-optimization device provided in the embodiments of this application.

[0047] Figure 5 This is a schematic diagram of the structure of the collaborative optimization strategy model construction device provided in the embodiments of this application.

[0048] Figure 6 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0049] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0050] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, components, features, and elements with the same names in different embodiments of this application may have the same meaning or different meanings, the specific meaning of which must be determined by its interpretation in that specific embodiment or further in conjunction with the context of that specific embodiment.

[0051] It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0052] In the following description, the use of suffixes such as "module," "part," or "unit" to denote elements is solely for the purpose of illustration and has no specific meaning in itself. Therefore, "module," "part," or "unit" may be used interchangeably.

[0053] In the description of this application, it should be noted that the terms "upper," "lower," "left," "right," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application. In addition, terms such as "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0054] To compensate for pattern distortion (i.e., optical proximity effect) caused by diffraction and interference during photolithography, traditional OPC algorithms introduce numerous complex auxiliary patterns and jagged edge modifications onto the original design pattern. While this method significantly improves image fidelity, it inevitably and substantially increases the geometric complexity of the mask pattern.

[0055] This results in an exponential increase in the number of exposures and writing time required during the mask writing process in mask manufacturing, making mask manufacturing cost and cycle time a key bottleneck restricting the research and development and mass production of advanced chips.

[0056] Therefore, how to improve the writing efficiency of mask patterns while ensuring imaging quality has become a pressing technical problem in the current semiconductor manufacturing field.

[0057] Based on this, embodiments of this application provide a mask pattern quality and efficiency collaborative optimization method, apparatus, storage medium, and electronic device. Specifically, the mask pattern quality and efficiency collaborative optimization apparatus can be integrated into an electronic device, which can be a server or a terminal, etc. The terminal can include mobile phones, wearable smart devices, tablets, laptops, and personal computers (PCs), etc., and other computers and auxiliary devices. The server can be a single server or a server cluster composed of multiple servers, and can be a physical server or a virtual server.

[0058] For example, such as Figure 1 As shown, the electronic device can acquire the mask pattern to be optimized; for the current target pattern region of the mask pattern to be optimized, extract the state feature vector of the current target pattern region; input the state feature vector into the collaborative optimization strategy model to obtain the current optimization action; optimize the current target pattern region according to the current optimization action, and if the optimization termination condition is not met, return to the step of extracting the state feature vector of the current target pattern region until the optimization termination condition is met, and obtain the target optimized pattern; when all target pattern regions of the mask pattern to be optimized are optimized, merge all target optimized patterns into the target mask pattern.

[0059] The technical solutions shown in this application will be described in detail below through specific embodiments. It should be noted that the order of description of the following embodiments is not intended to limit the priority of the embodiments.

[0060] Please see Figure 2 , Figure 2 This is a flowchart illustrating the mask pattern quality and efficiency co-optimization method provided in this application embodiment. The specific flow of this mask pattern quality and efficiency co-optimization method can be as follows:

[0061] 101. Obtain the mask image to be optimized.

[0062] The mask pattern to be optimized can be read from local or network storage devices, such as directly connected hard drives, network-attached storage, or storage area networks, and can be read from layout files pre-generated and stored by the design department. This mask pattern is typically stored in standard formats such as GDSII or OASIS.

[0063] The mask pattern to be optimized can be a design pattern that has been initially laid out but has not yet undergone OPC, or it can be a complex pattern that has already undergone traditional OPC processing but needs further optimization of write efficiency.

[0064] 102. For the current target graphic region of the mask graphic to be optimized, extract the state feature vector of the current target graphic region.

[0065] In this embodiment, to achieve refined optimization, the entire mask pattern is typically divided into multiple processing regions (such as hierarchical blocks or fixed-size sliding windows). Then, for the target region of the mask pattern currently being processed, features can be extracted to construct its digital representation, i.e., a state feature vector.

[0066] Specifically, rapid photolithographic simulation can be performed on the current target graphic region to obtain local optical simulation features and imaging quality assessment features; geometric analysis can be performed on the current target graphic region to generate geometric features; and the local optical simulation features, imaging quality assessment features, and geometric features can be concatenated into a state feature vector.

[0067] In some embodiments, a lightweight lithography simulation model (such as a convolution-based fast optical model) can be invoked to simulate the current region and extract local optical simulation features (e.g., light intensity distribution of the spatial image, image log slope ILS) and imaging quality evaluation features (e.g., edge placement error EPE of the current pattern, local uniformity of the critical dimension CD).

[0068] The geometric features include, but are not limited to, the total edge length of the polygon, the edge curvature, the minimum spacing between adjacent polygons, and the area-to-perimeter ratio of the polygon. In some embodiments, graphics complexity features that are strongly correlated with write efficiency may also be included, such as the number of polygon vertices and the degree of fragmentation of the graphics (number of small polygons).

[0069] It is understandable that, since local optical simulation features, imaging quality assessment features, and geometric features may have different dimensions (such as nanometers, intensity units, dimensionless ratios) and numerical ranges, direct stitching may lead to some features being overemphasized or ignored in subsequent processes.

[0070] Therefore, before stitching, the three features—local optical simulation features, imaging quality assessment features, and geometric features—can be normalized separately; then, these three features are stitched together in terms of dimensions to form a state feature vector.

[0071] 103. Input the state feature vector into the collaborative optimization strategy model to obtain the current optimization action.

[0072] The collaborative optimization strategy model is a deep neural network model (e.g., a policy network model) pre-trained through reinforcement learning. In this embodiment, the collaborative optimization strategy model can be directly loaded.

[0073] After the state feature vector is input into the collaborative optimization strategy model, the collaborative optimization strategy model can perform forward propagation calculations and output a current optimization action based on the complex mapping relationship learned internally about how to balance imaging quality and writing efficiency.

[0074] The optimization action is typically defined in a discrete or continuous action space, directly corresponding to executable modification instructions for the geometric data of the graphic. For example, the optimization action may include: "fine-tuning the specified edge to the left by 0.5 nanometers", "merging two auxiliary graphics with a spacing smaller than a preset value", "deleting a specific small-sized auxiliary graphic", or "keeping the current graphic unchanged".

[0075] 104. Optimize the current target graphic region according to the current optimization action, and if the optimization termination condition is not met, return to the step of extracting the state feature vector of the current target graphic region until the optimization termination condition is met, and obtain the target optimized graphic.

[0076] Specifically, by analyzing the current optimization action, the geometric data of the current target graphic region can be optimized accordingly.

[0077] In this embodiment of the application, the optimization termination condition can be any one of the following: strategy termination condition, iteration termination condition, and convergence termination condition.

[0078] The strategy termination condition refers to the collaborative optimization strategy model outputting a "hold" instruction. This indicates that the collaborative optimization strategy model believes it has approached the optimal equilibrium point. The iteration termination condition refers to the optimization loop count for the current region reaching a preset maximum value (e.g., 20 times) to prevent infinite loops. The convergence termination condition refers to the improvement in both imaging quality and write efficiency metrics falling below preset thresholds in several consecutive iterations, indicating that the optimization has stabilized.

[0079] Understandably, if any of the above optimization termination conditions are not met, the process can return to the step of extracting the state feature vector of the current target graphic region based on the optimized new graphic, thus initiating the next "perception-decision-update" loop. This loop will continue until the optimization termination condition is met, at which point the current target graphic region is considered to have been optimized, and the optimized target graphic can be output.

[0080] 105. When all target graphic regions of the mask graphic to be optimized have been optimized, merge all optimized target graphics into a target mask graphic.

[0081] The phrase "all target graphic regions of the mask graphic to be optimized have been optimized" means that steps 101-104 are repeated until all target graphic regions of the mask graphic to be optimized are traversed and optimized.

[0082] Once all target regions of the mask pattern to be optimized have been optimized, they can be stitched and integrated according to their original positions to resolve potential pattern consistency issues at region boundaries (such as edge alignment and pattern overlap handling), ultimately generating a complete target mask pattern that has undergone quality and efficiency co-optimization. This target mask pattern can be directly used in mask writing devices, theoretically guaranteeing imaging quality comparable to post-processing optimization schemes, while significantly reducing pattern complexity and writing time costs.

[0083] In summary, the mask image quality and efficiency collaborative optimization method provided in this application includes: acquiring the mask image to be optimized; extracting the state feature vector of the current target image region for the mask image to be optimized; inputting the state feature vector into the collaborative optimization strategy model to obtain the current optimization action; optimizing the current target image region according to the current optimization action, and if the optimization termination condition is not met, returning to the step of extracting the state feature vector of the current target image region until the optimization termination condition is met to obtain the target optimized image; when all target image regions of the mask image to be optimized are optimized, merging all target optimized images into the target mask image. This application embodiment, by extracting state feature vectors including local optical simulation features, imaging quality evaluation features, and geometric features, can ensure that each optimization decision is made under the premise of perceiving the current imaging state. Then, it utilizes the pre-trained collaborative optimization strategy model to output targeted optimization actions, and through iterative loops, gradually converges the mask image to a state with better writing efficiency while satisfying imaging quality constraints. This application's embodiments shift the consideration of writing efficiency to the pattern generation stage, avoiding the generation of unnecessary complex structures from the source. This significantly reduces the geometric complexity of the mask pattern while strictly ensuring imaging quality, ultimately achieving the goal of improving writing efficiency and reducing manufacturing costs.

[0084] It is understandable that before executing steps 101-105 in the above embodiments, it is necessary to complete the construction of the collaborative optimization strategy model. The essence of constructing the collaborative optimization strategy model is a reinforcement learning training phase, such as... Figure 3 As shown, Figure 3 This is a flowchart illustrating the collaborative optimization strategy model construction method provided in this application embodiment. The specific process of constructing the collaborative optimization strategy model construction method can be as follows:

[0085] 201. Construct a training environment and training set. The training environment includes a lithography simulation module and a write efficiency evaluation module. The training set contains a variety of mask pattern samples.

[0086] This training environment can simulate a new mask pattern after performing optimization actions on the mask pattern, and call the two modules mentioned above to calculate the precise changes in imaging quality indicators and write efficiency indicators.

[0087] The lithography simulation module simulates the imaging process of a mask pattern in a lithography machine. It receives the input mask pattern and, based on preset process conditions (such as wavelength, numerical aperture, and illumination method), calculates and outputs its spatial image or resist image on the silicon wafer surface using a partially coherent imaging model or a more precise vector imaging model. The fidelity of this lithography simulation module directly determines the reliability of the "imaging quality" knowledge learned by the collaborative optimization strategy model.

[0088] The write efficiency evaluation module quantifies the manufacturing complexity and cost of mask patterns. It receives the mask pattern and, based on the working principle of a specific mask writing device (such as an electron beam direct writing device), estimates the number of exposure points (Shot Count) or the total write time required to write the mask pattern using a pattern segmentation algorithm (decomposing complex polygons into basic patterns that the writing device can process, such as trapezoids) and exposure dose calculation rules. This write efficiency evaluation module provides an objective metric for "write efficiency."

[0089] The diverse mask pattern samples should cover different design layers (such as transistor gate layers, metal interconnect layers, contact hole layers, etc.), different pattern features (such as dense lines, isolated lines, line ends, contact hole arrays, two-dimensional complex patterns, etc.), and different process nodes (such as covering the layout features of multiple advanced process nodes (such as 7nm, 5nm, 3nm) within the target application range as much as possible), thereby ensuring that the trained collaborative optimization strategy model has strong generalization ability and can adapt to various real chip design scenarios.

[0090] 202. Initialize the policy network model.

[0091] The policy network model is a deep neural network whose function is to map the input state feature vector to the optimal (or optimal probability distribution) action.

[0092] In practical implementation, different network architectures can be selected according to the type of action space (discrete or continuous), for example:

[0093] For discrete actions (such as "merging graphics A and B" or "deleting feature C"), an architecture based on a multilayer perceptron (MLP) or convolutional neural network (CNN) combined with a softmax output layer can be used.

[0094] For continuous actions (such as "fine-tuning the edge by Δx nm"), an architecture with Gaussian distributed parameters (mean and variance) can be selected.

[0095] It should be noted that the input layer dimension of the policy network model must match the dimension of the state feature vector, and the output layer dimension must match the dimension of the action space.

[0096] In some embodiments, standard neural network parameter initialization methods (such as Xavier initialization and He initialization) can be used to randomly initialize all weights and biases of the policy network model.

[0097] 203. Iteratively train the policy network model based on the training set and training environment until the policy network model converges, generating a collaboratively optimized policy model.

[0098] For the current training mask image in this training set, the following steps can be repeated:

[0099] First, the state feature vector of the current training mask pattern is extracted. It should be noted that this extraction method is completely consistent with the application stage (step 102) to ensure consistency between training and application. Then, the state feature vector is input into the policy network model, which performs forward propagation to generate one (or a set of) training actions. The training actions are then used in the training environment to optimize the current training mask pattern, and the lithography simulation module and write efficiency evaluation module calculate the imaging quality index and write efficiency index.

[0100] Finally, based on the imaging quality and write efficiency metrics, the parameters of the policy network model are updated using the proximal policy optimization algorithm. The process then returns to extracting the state feature vector of the current training mask image, continuing until the policy network model converges. Specifically, an immediate reward value can be calculated based on the imaging quality and write efficiency metrics; then, based on this immediate reward value, the parameters of the policy network model are updated using the proximal policy optimization (PPO) algorithm.

[0101] In this embodiment, the imaging quality index ΔQuality and the write efficiency index ΔEfficiency can be substituted into the composite reward function R=α·ΔQuality + β·ΔEfficiency to calculate the immediate reward value for this decision.

[0102] Here, α and β are dynamically adjustable weight coefficients. In the early stages of training, a larger α and a smaller β can be set to guide the policy network model to prioritize learning how to improve image quality and lay a solid foundation. As the training epochs increase, the weight of β is gradually increased to encourage the policy network model to explore ways to improve writing efficiency while ensuring quality, thereby achieving "co-optimization".

[0103] In some embodiments, the step "updating the parameters of the policy network model based on the immediate reward value using the proximal policy optimization algorithm" can specifically be: using the PPO algorithm, using the currently collected interaction data (including state feature vectors, training actions, immediate reward values, and optimized training mask graphics), calculating the policy gradient, and updating the parameters of the policy network model.

[0104] The PPO algorithm effectively controls the magnitude of each parameter update by introducing stabilization techniques such as pruning, thus achieving more stable and efficient learning. After each update, the decision-making ability of the policy network model is improved.

[0105] Understandably, multiple rounds of training are performed on the entire training set by repeating the above iterative loop. The overall performance of the current policy model is periodically evaluated on an independent validation set (e.g., assessing the average image quality compliance rate and average write efficiency improvement rate of the optimized mask pattern). When the performance metrics on the validation set no longer significantly improve or reach a preset performance threshold over several consecutive evaluation periods, the model is considered to have converged. At this point, the final policy network parameters and structure can be saved, resulting in a collaboratively optimized policy model ready for practical application.

[0106] In summary, the collaborative optimization strategy model construction method provided in this application utilizes a training environment, enabling the strategy network model to autonomously learn how to collaboratively balance graphics quality improvement and write efficiency enhancement in each graphics optimization decision through interactive trial and error with the training environment, rather than relying on preset geometric rules. In particular, guided by a composite reward function that dynamically adjusts weights (α, β), the strategy network model can gradually transition from "prioritizing quality" to a learning stage of "maximizing write efficiency under graphics quality constraints." Ultimately, the collaborative optimization strategy model obtained through training convergence fundamentally achieves source-level collaborative optimization capabilities. Its decisions directly avoid the generation of complex graphics structures that are difficult to correct later, thus not only solving the problems of target fragmentation and timing lag in post-processing schemes, but also providing powerful, data-driven adaptive optimization capabilities to cope with diverse process technologies and graphics types.

[0107] To facilitate better implementation of the mask pattern quality and efficiency collaborative optimization method provided in this application embodiment, this application embodiment also provides a mask pattern quality and efficiency collaborative optimization device. The meanings of the terms used are the same as in the mask pattern quality and efficiency collaborative optimization method described above, and specific implementation details can be found in the descriptions in the method embodiments.

[0108] Please see Figure 4 , Figure 4 This is a schematic diagram of the mask pattern quality and efficiency collaborative optimization device provided in an embodiment of this application. The mask pattern quality and efficiency collaborative optimization device may include an acquisition unit 301, an extraction unit 302, an input unit 303, an optimization unit 304, and an output unit 305.

[0109] Acquisition unit 301 is used to acquire the mask image to be optimized;

[0110] Extraction unit 302 is used to extract the state feature vector of the current target graphic region for the current target graphic region of the mask graphic to be optimized;

[0111] Input unit 303 is used to input the state feature vector into the collaborative optimization strategy model to obtain the current optimization action;

[0112] The optimization unit 304 is used to optimize the current target graphic region according to the current optimization action, and if the optimization termination condition is not met, return to the step of extracting the state feature vector of the current target graphic region until the optimization termination condition is met, so as to obtain the target optimized graphic.

[0113] The output unit 305 is used to merge all the target optimized graphics into a target mask graphic when all target graphic regions of the mask graphic to be optimized have been optimized.

[0114] For specific implementation methods of each of the above units, please refer to the embodiments of the above-mentioned mask pattern quality and efficiency collaborative optimization method, which will not be repeated here.

[0115] In summary, the mask pattern quality and efficiency collaborative optimization device provided in this application embodiment can acquire the mask pattern to be optimized by the acquisition unit 301; the extraction unit 302 extracts the state feature vector of the current target pattern region of the mask pattern to be optimized; the input unit 303 inputs the state feature vector into the collaborative optimization strategy model to obtain the current optimization action; the optimization unit 304 optimizes the current target pattern region according to the current optimization action, and if the optimization termination condition is not met, returns to the step of extracting the state feature vector of the current target pattern region until the optimization termination condition is met, thus obtaining the target optimized pattern; the output unit 305 merges all target optimized patterns into the target mask pattern when all target pattern regions of the mask pattern to be optimized are optimized. This application embodiment moves the consideration of writing efficiency to the pattern generation stage, avoiding the generation of unnecessary complex structures from the source, thereby significantly reducing the geometric complexity of the mask pattern while strictly ensuring the imaging quality, ultimately achieving the purpose of improving writing efficiency and reducing manufacturing costs.

[0116] To facilitate better implementation of the collaborative optimization strategy model construction method provided in this application, this application also provides a collaborative optimization strategy model construction apparatus. The meanings of the terms used are the same as in the aforementioned collaborative optimization strategy model construction method, and specific implementation details can be found in the descriptions within the method embodiments.

[0117] Please see Figure 5 , Figure 5 This is a schematic diagram of the structure of the collaborative optimization strategy model construction device provided in this application embodiment. The collaborative optimization strategy model construction device may include a construction unit 401, an initialization unit 402, and a training unit 403.

[0118] The construction unit 401 is used to build the training environment and training set. The training environment includes a photolithography simulation module and a write efficiency evaluation module, and the training set contains a variety of mask pattern samples.

[0119] Initialization unit 402 is used to initialize the policy network model;

[0120] Training unit 403 is used to iteratively train the policy network model based on the training set and training environment until the policy network model converges, thereby generating a collaboratively optimized policy model.

[0121] For specific implementation methods of each of the above units, please refer to the embodiments of the above-mentioned mask pattern quality and efficiency collaborative optimization method, which will not be repeated here.

[0122] In summary, the collaborative optimization strategy model construction device provided in this application constructs a training environment and training set through a construction unit 401. The training environment includes a lithography simulation module and a write efficiency evaluation module, and the training set contains diverse mask pattern samples. An initialization unit 402 initializes the strategy network model. A training unit 403 iteratively trains the strategy network model based on the training set and training environment until the strategy network model converges, generating a collaborative optimization strategy model. This application embodiment fundamentally achieves source-level collaborative optimization capabilities. Its decision-making directly avoids the generation of complex pattern structures that are difficult to correct later. This not only solves the problems of target fragmentation and timing lag in post-processing solutions but also provides powerful, data-driven adaptive optimization capabilities to cope with diverse process technologies and pattern types.

[0123] This application also provides an electronic device that may integrate the mask image quality and efficiency co-optimization device and / or co-optimization strategy model construction device of this application. Figure 6 As shown, it illustrates a structural schematic diagram of the electronic device involved in the embodiments of this application, specifically:

[0124] The electronic device may include components such as a processor 501 with one or more processing cores and a memory 502 with one or more computer-readable storage media. Those skilled in the art will understand that... Figure 6 The electronic device structure shown does not constitute a limitation on the electronic device and may include more or fewer components than shown, or combine certain components, or have different component arrangements. Wherein:

[0125] The processor 501 is the control center of the electronic device. It connects various parts of the electronic device via various interfaces and lines. By running or executing software programs stored in the memory 502 and / or this application, and by calling data stored in the memory 502, it performs various functions and processes data, thereby providing overall monitoring of the electronic device. Optionally, the processor 501 may include one or more processing cores; preferably, the processor 501 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operation of the storage medium, user interface, and application programs, while the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor 501.

[0126] The memory 502 can be used to store software programs and this application. The processor 501 executes various functional applications and data processing by running the software programs and this application stored in the memory 502. The memory 502 may mainly include a program storage area and a data storage area. The program storage area may store applications required for operating the storage medium and at least one function; the data storage area may store data created according to the use of the electronic device. In addition, the memory 502 may include high-speed random access memory and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, the memory 502 may also include a memory controller to provide the processor 501 with access to the memory 502.

[0127] Although not shown, the electronic device may also include a display unit, an input unit, and a power supply, etc., which will not be described in detail here. Specifically, in this embodiment, the processor 501 in the electronic device loads the executable files corresponding to the processes of one or more application programs into the memory 502 according to the following instructions, and the processor 501 runs the application programs stored in the memory 502 to realize various functions, as follows:

[0128] Obtain the mask image to be optimized;

[0129] For the current target graphic region of the mask graphic to be optimized, extract the state feature vector of the current target graphic region;

[0130] The state feature vector is input into the collaborative optimization strategy model to obtain the current optimization action;

[0131] The current target graphic region is optimized based on the current optimization action. If the optimization termination condition is not met, the process returns to the step of extracting the state feature vector of the current target graphic region until the optimization termination condition is met, and the target optimized graphic is obtained.

[0132] When all target graphic regions of the mask graphic to be optimized have been optimized, all optimized target graphics are merged into a target mask graphic.

[0133] Or as follows:

[0134] The training environment and training set are constructed. The training environment includes a lithography simulation module and a write efficiency evaluation module, and the training set contains a variety of mask pattern samples.

[0135] Initialize the policy network model;

[0136] The policy network model is iteratively trained based on the training set and training environment until the policy network model converges, generating a collaboratively optimized policy model.

[0137] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be performed by instructions, or by instructions controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0138] Therefore, embodiments of this application provide a storage medium storing a plurality of instructions that can be loaded by a processor to execute steps in any of the methods provided in embodiments of this application. For example, the instructions can execute the following steps:

[0139] Obtain the mask image to be optimized;

[0140] For the current target graphic region of the mask graphic to be optimized, extract the state feature vector of the current target graphic region;

[0141] The state feature vector is input into the collaborative optimization strategy model to obtain the current optimization action;

[0142] The current target graphic region is optimized according to the current optimization action. If the optimization termination condition is not met, the process returns to the step of extracting the state feature vector of the current target graphic region until the optimization termination condition is met, and the target optimized graphic is obtained.

[0143] When all target graphic regions of the mask graphic to be optimized are optimized, all the optimized target graphics are merged into a target mask graphic.

[0144] Or perform the following steps:

[0145] The training environment and training set are constructed. The training environment includes a lithography simulation module and a write efficiency evaluation module, and the training set contains a variety of mask pattern samples.

[0146] Initialize the policy network model;

[0147] The policy network model is iteratively trained based on the training set and training environment until the policy network model converges, generating a collaboratively optimized policy model.

[0148] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0149] The storage medium may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0150] Since the instructions stored in the storage medium can execute the steps of any method provided in the embodiments of this application, the beneficial effects that any method provided in the embodiments of this application can achieve can be realized. For details, please refer to the previous embodiments, which will not be repeated here.

[0151] The above provides a detailed description of the mask pattern quality and efficiency collaborative optimization method, apparatus, storage medium, and electronic device provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for collaborative optimization of mask image quality and efficiency, characterized in that, include: A training environment and a training set are constructed. The training environment includes a photolithography simulation module and a write efficiency evaluation module, and the training set contains diverse mask pattern samples. Initialize the policy network model; The policy network model is iteratively trained based on the training set and the training environment until the policy network model converges, thereby generating a collaborative optimization policy model. Obtain the mask image to be optimized; For the current target pattern region of the mask pattern to be optimized, a fast photolithography simulation is performed on the current target pattern region to obtain local optical simulation features and imaging quality evaluation features; Perform geometric analysis on the current target graphic region to generate geometric features; The local optical simulation features, the imaging quality assessment features, and the geometric features are concatenated into a state feature vector; The state feature vector is input into the collaborative optimization strategy model to obtain the current optimization action; The current target graphic region is optimized according to the current optimization action. If the optimization termination condition is not met, the process returns to the step of extracting the state feature vector of the current target graphic region until the optimization termination condition is met, and the target optimized graphic is obtained. When all target graphic regions of the mask graphic to be optimized are optimized, all the optimized target graphics are merged into a target mask graphic.

2. The mask pattern quality and efficiency co-optimization method as described in claim 1, characterized in that, The iterative training of the policy network model based on the training set and the training environment until the policy network model converges includes: For the current training mask image in the training set, extract the state feature vector of the current training mask image; The state feature vector is input into the policy network model to generate training actions; The training action is used in the training environment to optimize the current training mask image and to calculate the imaging quality index and write efficiency index. Based on the imaging quality index and the writing efficiency index, the parameters of the policy network model are updated using a near-end policy optimization algorithm, and the process returns to the step of extracting the state feature vector of the current training mask image until the policy network model converges.

3. The mask pattern quality and efficiency co-optimization method as described in claim 2, characterized in that, The step of updating the parameters of the policy network model based on the imaging quality index and the writing efficiency index using a near-end policy optimization algorithm includes: Calculate the instantaneous reward value based on the imaging quality index and the writing efficiency index; Based on the instant reward value, the parameters of the policy network model are updated using a near-end policy optimization algorithm.

4. The mask pattern quality and efficiency co-optimization method as described in claim 3, characterized in that, The formula for calculating the instant reward value is as follows: R=α·ΔQuality + β·ΔEfficiency Where ΔQuality is the imaging quality index, ΔEfficiency is the writing efficiency index, and α and β are dynamically adjusted weighting coefficients.

5. A mask pattern quality and efficiency co-optimization device, characterized in that, include: A construction unit is used to construct a training environment and a training set. The training environment includes a photolithography simulation module and a write efficiency evaluation module, and the training set contains diverse mask pattern samples. Initialization unit, used to initialize the policy network model; The training unit iteratively trains the policy network model based on the training set and the training environment until the policy network model converges, thereby generating a collaborative optimization policy model. The acquisition unit is used to acquire the mask image to be optimized. The extraction unit is used to perform fast photolithography simulation on the current target graphic region of the mask graphic to be optimized, and obtain local optical simulation features and imaging quality evaluation features. Perform geometric analysis on the current target graphic region to generate geometric features; concatenate the local optical simulation features, the imaging quality assessment features, and the geometric features into a state feature vector; The input unit is used to input the state feature vector into the collaborative optimization strategy model to obtain the current optimization action; An optimization unit is used to optimize the current target graphic region according to the current optimization action, and if the optimization termination condition is not met, return to the step of extracting the state feature vector of the current target graphic region until the optimization termination condition is met, so as to obtain the target optimized graphic. The output unit is used to merge all the target optimized graphics into a target mask graphic when all target graphic regions of the mask graphic to be optimized have been optimized.

6. A storage medium, characterized in that, The storage medium stores multiple instructions, which are adapted for loading by a processor to execute the mask pattern quality and efficiency co-optimization method according to any one of claims 1-4.

7. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, it implements the mask pattern quality and efficiency co-optimization method as described in any one of claims 1-4.

Images