A staged quantization method and related apparatus for collaborative optimization of light source curve masks
By adopting a phased quantization-based light source curve mask co-optimization method, the imaging accuracy and computational efficiency problems of light source-mask co-optimization in photolithography technology are solved. This method achieves smooth convergence from grayscale masks to binary manufacturable masks, improving the efficiency of multi-pattern joint optimization and the manufacturability of masks.
Patent Information
- Application Number
- CN202611486368.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-22
- Publication Date
- 2026-11-10
AI Technical Summary
Existing photolithography technologies suffer from low imaging accuracy, poor mask manufacturability, and low computational efficiency in light source-mask co-optimization, especially in scenarios where multiple masks share a light source, making effective co-optimization difficult to achieve.
A staged quantization-based collaborative optimization method for light source curve masks is adopted. By constructing a differentiable optimization link for single light source and multiple masks, and combining an adaptive quantization regularization function and a staged activation strategy, quantization constraints are gradually introduced to achieve smooth convergence of grayscale masks to binary manufacturable masks.
While ensuring imaging fidelity and process robustness, the computational efficiency of multi-pattern joint optimization is improved, avoiding the degradation of main feature edges and the loss of higher-order SRAF, and the output mask directly meets manufacturing requirements.
Smart Images

Figure CN123067033A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of computational lithography technology in integrated circuit manufacturing, and in particular to a method and apparatus for staged quantization of light source curve mask collaborative optimization. Background Technology
[0002] As integrated circuit process nodes continue to evolve downwards, advanced nodes at 28nm and below place stringent demands on lithography imaging quality, process windows, and mask manufacturability. Light source-mask co-optimization (SMO) is a core resolution enhancement technology in computational lithography. By simultaneously optimizing the illumination source and mask pattern, it suppresses optical proximity effects and widens the process variation window. Curved masks offer advantages over traditional polygonal masks in terms of high imaging accuracy and good pattern adaptability. However, curved mask patterns directly generated by reverse lithography algorithms are complex, and grayscale continuous transmittance masks cannot be directly used for mask fabrication, requiring discrete binarization processing.
[0003] In standard cell library engineering applications, a large number of cell layouts need to share the same lighting source for joint optimization. Existing technologies have significant shortcomings: First, traditional fixed threshold quantization methods apply uniform constraints to the entire mask region, failing to distinguish between the characteristics of the main feature region and sub-resolution auxiliary graphics (SRAF), easily causing degradation of the main feature edges or the direct loss of small-sized high-order SRAFs; Second, directly applying strong binary constraints to continuous grayscale masks instantly destroys the continuous imaging degrees of freedom, reducing imaging fidelity and making the optimization process prone to oscillations and difficult to converge; Third, traditional serial architectures for multi-mask shared light source optimization tasks have computational costs that increase linearly with the number of layouts, resulting in huge computational overhead and difficulties in engineering implementation.
[0004] Currently, the mainstream method for lithographic mask optimization is the level-set algorithm. This method relies on solving partial differential equations (PDEs) to complete boundary evolution and achieve mask design, and the algorithm's iteration step size is strictly constrained by the Courant condition (CFL). In contrast, illumination optimization uses pixel-level gradient calculation logic. The computational mechanisms of the two methods differ significantly, making effective collaborative optimization impossible. Furthermore, this method suffers from problems such as easy image degradation, low edge control accuracy, and low computational efficiency, making it difficult to adapt to the high-efficiency optimization requirements of advanced nodes with multiple masks sharing a light source. Therefore, there is an urgent need for a light source-mask collaborative optimization scheme that balances imaging accuracy, mask manufacturability, process robustness, and computational efficiency. Summary of the Invention
[0005] The purpose of this application is to provide a staged quantization method and related apparatus for collaborative optimization of light source curve masks, which can achieve smooth convergence of grayscale masks to binary manufacturable masks while ensuring imaging fidelity and process robustness, and improve the computational efficiency of multi-pattern joint optimization.
[0006] To achieve the above objectives, this application provides the following solution: Firstly, this application provides a staged quantization method for collaborative optimization of light source curve masks, comprising the following steps: Multiple sets of target layouts to be optimized are obtained, and continuous parametric modeling of light sources and masks is performed to construct a differentiable optimization link for a single light source and multiple masks.
[0007] Based on the differentiable optimization link, the first stage of continuous spatial optimization is performed without applying mask binary manufacturing constraints. The imaging error is used as the loss function to jointly optimize the shared light source and multiple sets of masks to obtain a grayscale continuous transmittance mask.
[0008] Based on the transmittance distribution and graphic spatial position of the grayscale continuous transmittance mask, the grayscale continuous transmittance mask is divided into at least two types of functional regions, including the main feature region and the SRAF region.
[0009] Based on the functional regions, an adaptive quantization regularization function is constructed. The adaptive quantization regularization function has the minimum penalty value at the mask transmittance positions of 0 and 1, and has an adjustable peak position in the transmittance range of 0 to 1. Different regularization function parameters are configured for different functional regions to achieve partitioned differentiated quantization constraints.
[0010] Based on the adaptive quantization regularization function, a total objective function is constructed, and a phased activation strategy is set. The quantization regularization constraints corresponding to each functional region are activated in stages, following the order from the main feature region to the SRAF region. The total objective function integrates the imaging performance evaluation term and the phased activated regional quantization regularization term.
[0011] The overall objective function is solved iteratively, and the shared light source parameters, each set of mask parameters, and each region quantization regularization parameter are updated jointly. In each iteration, the lithography simulation and gradient calculation corresponding to each set of masks are executed in parallel, and the shared light source is updated according to the gradient obtained by multi-mask fusion. After the iteration converges, the light source-curve mask co-optimization result is output. The light source-curve mask co-optimization result includes: the optimal shared light source distribution and the manufacturable binary curve mask corresponding to each of the target layouts to be optimized.
[0012] Secondly, this application provides a computer 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 steps of the staged quantization light source curve mask collaborative optimization method described above.
[0013] Thirdly, this application provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the staged quantization light source curve mask collaborative optimization method described above.
[0014] Fourthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the staged quantization light source curve mask collaborative optimization method described above.
[0015] According to the specific embodiments provided in this application, this application has the following technical effects: This application provides a staged quantization-based light source curve mask collaborative optimization method and related apparatus. First, by constructing a differentiable optimization link for a single light source and multiple masks, combined with continuous space joint optimization without imposing binary manufacturing constraints in the first stage, multi-pattern collaborative optimization can be fully utilized to obtain a grayscale continuous transmittance mask, ensuring the effectiveness of early imaging optimization. Second, the mask is divided into different functional regions: the main feature region and the SRAF region. By constructing an adaptive quantization regularization function and introducing a regionalized adaptive quantization regularization term, a differentiated bias is applied to different regions through a movable peak quantization function. The main feature region uses a conservative bias to protect the edges, while the SRAF region uses a flexible bias to delay polarization, thereby improving the manufacturability of the binary mask. Furthermore, by adopting a staged activation strategy, the quantization regularization constraints are activated stage by stage from the main feature region to the peripheral SRAF region. Combined with the overall objective function of the fused imaging performance evaluation term and the regional quantization regularization term, discretization constraints can be gradually introduced, effectively avoiding the degradation of the main feature edges or the loss of higher-order SRAF caused by a one-time strong constraint, achieving a smooth transition from continuous space to binary space. Furthermore, during the iterative solution process, the lithography simulation and gradient calculation of each mask group are executed in parallel. The shared light source is updated based on the gradient obtained by multi-mask fusion, which can adapt to the optimization scenarios of multiple target layouts and improve the computational efficiency of multi-layout joint optimization. Finally, after iterative convergence, the optimal shared light source distribution and manufacturable binarized curve masks are output. The output masks do not require additional discretization processing and directly meet the binarization requirements of mask manufacturing. Under the premise of ensuring imaging fidelity and process robustness, the grayscale mask can be smoothly converged to the binary manufacturable mask, improving the computational efficiency of multi-layout joint optimization. Attached Figure Description
[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the 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.
[0017] Figure 1 An application environment diagram of a staged quantization light source curve mask collaborative optimization method provided in an embodiment of this application; Figure 2 A flowchart illustrating a staged quantization method for collaborative optimization of light source curve masks is provided in one embodiment of this application. Figure 3 A schematic diagram illustrating the principle and process of staged light source-curve mask collaborative optimization provided in an embodiment of this application; Figure 4 Various layout diagrams provided for an embodiment of this application, wherein, Figure 4 (a) in the diagram is the initial light source diagram; Figure 4 (b) in the diagram represents the I-shaped target map; Figure 4 (c) in the image represents a multi-vertical target map; Figure 4 (d) in the diagram represents a cross-shaped target layout; Figure 4 (e) in the image represents a multi-bar target map; Figure 4 (f) in the diagram represents the target layout of both blocks; Figure 5 Different parameters provided for one embodiment of this application The curve of the adaptive quantization regularization function; Figure 6 This application provides an optimized main graphic and a schematic diagram of the division of continuous binary mask regions of each SRAF level for one embodiment of the present application. Figure 6 (a) in the diagram is a schematic diagram of the continuous mask region division after the optimization of the I-shaped layout; Figure 6 (b) in the diagram is a schematic diagram of the continuous mask region division after the optimization of the multi-vertical strip layout; Figure 6 (c) in the diagram is a schematic diagram of a cross-shaped layout; Figure 6 (d) in the diagram is a schematic diagram of the continuous mask region division after the multi-bar layout optimization; Figure 6 (e) in the diagram is a schematic diagram of the continuous mask region division after the optimization of the two block layouts; Figure 7 This is a schematic diagram of the optimized mask and simulated wafer imaging results provided in an embodiment of this application, wherein... Figure 7 (a) in the diagram is the optimized lighting source diagram; Figure 7 (b) in the image is the grayscale mask of the CTM optimization result corresponding to the I-shaped layout; Figure 7 (c) in the image is the grayscale mask of the CTM optimization result corresponding to the multi-vertical strip layout; Figure 7 (d) in the image represents the grayscale mask of the CTM optimization result corresponding to the cross-shaped layout. Figure 7 (e) in the image is the grayscale mask of the CTM optimization result corresponding to the multi-bar layout; Figure 7(f) in the figure represents the grayscale mask of the CTM optimization result corresponding to the two block layouts; Figure 8 A schematic diagram of a manufacturable binary curve mask output after iterative convergence, provided in an embodiment of this application; wherein, Figure 8 (a) in the diagram is the lighting source diagram obtained by the method optimized in this application. Figure 8 In the diagram, (b) is the binarized curve mask corresponding to the I-shaped layout. Figure 8 In the diagram, (c) represents the binarized curve mask corresponding to the multi-vertical-strip layout. Figure 8 In the diagram, (d) represents the binarized curve mask corresponding to the cross-shaped layout. Figure 8 In the diagram, (e) represents the binarized curve mask corresponding to the multi-bar layout. Figure 8 In the diagram, (f) represents the binarized curve mask corresponding to the two block layouts; Figure 9 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation
[0018] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0019] The purpose of this application is to provide a staged quantization method and related apparatus for co-optimization of light source curves and masks, which can be applied to resolution enhancement processing of deep ultraviolet lithography (DUV) and extreme ultraviolet lithography (EUV), and is adapted to the light source-mask co-optimization scenario of standard cell library layouts of advanced process nodes. It can achieve smooth convergence of grayscale masks to binary manufacturable masks while ensuring imaging fidelity and process robustness, and improve the computational efficiency of multi-layout joint optimization.
[0020] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.
[0021] The staged quantization light source curve mask collaborative optimization method provided in this application embodiment can be applied to, for example... Figure 1In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be set up independently, integrated into server 104, or placed in the cloud or on another server. Terminal 102 can send multiple sets of target layouts to be optimized to server 104. After receiving the multiple sets of target layouts, server 104 performs continuous parametric modeling of the light source and mask for each set of target layouts, constructing a differentiable optimization link for a single light source and multiple masks. Based on the differentiable optimization link, it performs a first-stage continuous spatial optimization without applying binary mask manufacturing constraints, using imaging error as a loss function to jointly optimize the shared light source and multiple masks, obtaining a grayscale continuous transmittance mask. According to the transmittance distribution and graphic spatial position of the grayscale continuous transmittance mask, the grayscale continuous transmittance mask is divided into at least two functional regions, including a main feature region and an SRAF region. Based on the functional regions, an adaptive quantization regularization function is constructed. The adaptive quantization regularization function has the minimum penalty value at the mask transmittance positions of 0 and 1, and is adjustable within the 0~1 transmittance range. At the peak position, differentiated regularization function parameters are configured for different functional regions to achieve partitioned differentiated quantization constraints. Based on the adaptive quantization regularization function, a total objective function is constructed, and a phased activation strategy is set. The quantization regularization constraints corresponding to each functional region are activated in stages according to the order from the main feature region to the SRAF region. The total objective function integrates imaging performance evaluation terms and phased activated regional quantization regularization terms. The total objective function is iteratively solved, and the shared light source parameters, each group of mask parameters, and each region quantization regularization parameter are jointly updated. In each iteration, the lithography simulation and gradient calculation corresponding to each group of masks are executed in parallel, and the shared light source is updated according to the gradient obtained by multi-mask fusion. After the iteration converges, the light source-curve mask co-optimization result is output. The light source-curve mask co-optimization result includes: the optimal shared light source distribution and the manufacturable binary curve mask corresponding to each of the target layouts to be optimized. The server 104 can feed back the obtained light source-curve mask co-optimization result to the terminal 102. Furthermore, in some embodiments, the phased quantization of the light source curve mask collaborative optimization method can also be implemented separately by the server 104 or the terminal 102. For example, the terminal 102 can directly perform phased light source-curve mask collaborative optimization processing on multiple sets of target layouts to be optimized, or the server 104 can obtain multiple sets of target layouts to be optimized from the data storage system and perform phased light source-curve mask collaborative optimization processing on multiple sets of target layouts to be optimized.
[0022] The terminal 102 can be, but is not limited to, various desktop computers, laptops, smartphones, tablets, and IoT devices. The server 104 can be implemented using a standalone server or a server cluster consisting of multiple servers, or it can be a cloud server.
[0023] In one exemplary embodiment, such as Figure 2 As shown, a staged quantization method for collaborative optimization of light source curve masks is provided. This method is executed by a computer device, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, the method is applied to... Figure 1 Taking server 104 as an example, the following steps are included: S1: Obtain multiple sets of target layouts to be optimized, and perform continuous parametric modeling of the light source and mask to construct a differentiable optimization link for a single light source and multiple masks.
[0024] S2: Based on the differentiable optimization link, perform the first stage of continuous spatial optimization, without applying mask binary manufacturing constraints, and use imaging error as the loss function to jointly optimize the shared light source and multiple sets of masks to obtain a grayscale continuous transmittance mask.
[0025] S3: Based on the transmittance distribution and graphic spatial position of the grayscale continuous transmittance mask, the grayscale continuous transmittance mask is divided into at least two types of functional regions, including the main feature region and the SRAF region.
[0026] S4: Based on the functional regions, construct an adaptive quantization regularization function; the adaptive quantization regularization function has the minimum penalty value at the mask transmittance positions of 0 and 1, and has an adjustable peak position in the transmittance range of 0 to 1. Different regularization function parameters are configured for different functional regions to achieve partitioned differentiated quantization constraints.
[0027] S5: Based on the adaptive quantization regularization function, construct the overall objective function and set a phased activation strategy. Activate the quantization regularization constraints corresponding to each functional region in stages, following the order from the main feature region to the SRAF region. The overall objective function integrates the imaging performance evaluation term and the phased activated regional quantization regularization term.
[0028] S6: Iteratively solve the overall objective function and jointly update the shared light source parameters, each group of mask parameters, and each region quantization regularization parameter. In each iteration, the lithography simulation and gradient calculation corresponding to each group of masks are executed in parallel, and the shared light source is updated according to the gradient obtained by multi-mask fusion. After the iteration converges, the light source-curve mask co-optimization result is output. The light source-curve mask co-optimization result includes: the optimal shared light source distribution and the manufacturable binary curve mask corresponding to each of the target layouts to be optimized.
[0029] Implementing steps S1 to S6 above can achieve smooth convergence of grayscale masks to binary manufacturable masks while ensuring imaging fidelity and process robustness. It also improves the computational efficiency of multi-pattern joint optimization and solves the defects of existing light source-mask co-optimization methods, such as continuous mask binary discretization easily destroying imaging quality, main features or SRAF patterns easily failing, and large computational overhead in multi-mask shared light source scenes.
[0030] As an optional implementation, in step S3, the SRAF region includes a first-order SRAF region, a second-order SRAF region, and a third-order SRAF region. The distance between the SRAF regions of different orders and the main feature region increases from near to far, the mask transmittance decreases step by step, and the pattern size decreases step by step.
[0031] As an optional implementation, in step S6, when iteratively solving the overall objective function, a master-slave parallel architecture is adopted. The master-slave parallel architecture includes a Master node and several Worker nodes. The Worker nodes are used to perform lithography simulation and mask gradient calculation for each layout in parallel. The Master node is used to collect the light source gradients uploaded by all the Worker nodes, merge them, update the globally shared light source parameters, and then distribute the updated light source parameters to each Worker node.
[0032] As an optional implementation, in step S6, the criteria for determining the convergence of the iteration include: at least one of the following indicators reaches a preset threshold: Pattern Error (PE), Edge Placement Error (EPE), and Process Variation Band (PVBand); or the number of iterations reaches a preset threshold.
[0033] As an optional implementation, in step S6, the output manufacturable binarized curve mask can be directly used as input data for mask fabrication; the optimal shared light source distribution can be used to configure the lithography machine illumination system; the two can also be used together for lithography simulation verification, standard cell library resolution enhancement processing, simulation condition input for optical proximity correction, and layout design rule evaluation scenarios.
[0034] In an exemplary embodiment, to illustrate the technical solutions provided in the embodiments of this application in detail, the following are provided: Figure 3 This paper presents a staged quantization-based light source curve mask co-optimization method. Specifically, it is a staged light source-curve mask co-optimization algorithm based on the discretization of continuous transmittance of the mask. By designing a staged progressive activation strategy, quantization constraints are applied layer by layer from the main feature region to the peripheral SRAF region, achieving a smooth transition from continuous space to binary space. An adaptive threshold quantization regularization function with movable peak values is designed to support regionally differentiated threshold adjustments. By fusing PE, EPE, PVBand, and regional quantization regularization terms, multi-objective joint optimization is achieved, supporting parallel acceleration across multiple masks and further improving computational efficiency. This method solves the problems of low computational efficiency, rigid quantization constraints, easy loss of SRAF, and decreased imaging accuracy in traditional light source mask co-optimization, achieving high-precision, high-robustness, and high-speed co-optimization under multi-mask shared light source conditions. The specific implementation steps of this method are as follows: Step 101: Read multiple sets of target layouts to be optimized, use cosine mapping to continuously parameterize the light source and mask, complete the continuous parameterization representation of the light source and mask, and establish an end-to-end differentiable optimization link for single light source and multiple masks.
[0035] Step 102: Employing a Continuous Tone Mask (CTM) strategy, using PE as the loss function, a first-stage continuous spatial joint optimization is performed on the shared light source and multiple masks. In this stage, discretization manufacturing constraints are not applied; instead, the continuous transmittance degree of freedom is fully utilized to improve imaging fidelity, ensuring the consistency and stability of multi-mask collaborative optimization. Ultimately, a grayscale continuous transmittance mask containing higher-order SRAFs within the 0-1 range is generated. The loss function... The formula is as follows: (1).
[0036] in, For loss function, The target total number of territories, Indicates the first The graphic error corresponding to each target map Indicates a shared light source With corresponding mask The wafer pattern obtained through simulation under the influence of the action. For the first The mask corresponding to the target layout. Indicates the first A shared light source Indicates the first One target territory, It is the square of the L2 norm.
[0037] Step 103: Based on the grayscale continuous transmittance mask obtained in Step 102, perform statistical analysis on the mask transmittance distribution. Combining transmittance intensity, spatial connectivity, and distance from the target pattern boundary, divide the grayscale continuous transmittance mask into two categories and four regions: main feature region, first-order SRAF region, second-order SRAF region, and third-order SRAF region. The main feature region corresponds to the high-transmittance main pattern region directly involved in wafer imaging; the first-order, second-order, and third-order SRAF regions correspond to auxiliary imaging regions located around the main feature, with progressively decreasing transmittance and smaller size. The main feature region has high transmittance, clear outline, and directly determines the accuracy of key dimensions. The first-order, second-order, and third-order SRAF regions have progressively decreasing transmittance, finer structures, and are sensitive to quantization perturbations. This partitioning allows for differentiated constraints, avoiding degradation of the main feature edges or loss of higher-order SRAF due to uniform quantization.
[0038] Step 104: Design an adaptive quantization regularization function with a movable peak value. This function takes a minimum value of 0 at both 0 and 1, and has a unique controllable maximum value within the 0-1 interval. The minimum penalty point of this adaptive quantization regularization function is anchored at the physical values of 0 and 1. A nonlinear mapping function is used to controllably shift the penalty center of gravity in the intermediate transmittance region towards the target extreme point, thereby preserving more physical degrees of freedom for the weak SRAF region in the initial stage of quantization. This is achieved through parameters... (Parameters controlling the quantization bias position) The penalty center is flexibly adjusted to achieve differentiated quantization bias in each region; the adaptive quantization regularization function first transforms the mask transmittance through a nonlinear mapping function, and then applies polarization constraints, so that the maximum penalty position can be adjusted according to the parameters. Instead of being fixed at the intermediate transmittance point, the system moves flexibly, employing a conservative bias to protect the edges of the main feature region and a flexible bias to delay polarization in the SRAF region, preventing premature collapse of tiny auxiliary patterns. The adaptive quantization regularization function is as follows: (2).
[0039] in, For adaptive quantization regularization function, For intermediate mapping functions, To control the parameters of the quantization bias position, This indicates the mask transmittance.
[0040] intermediate mapping function Represented as: (3).
[0041] Step 105: Integrate the regionalized quantization regularization term, PE, EPE, and PVBand into the cost function. Set a phased activation strategy, activating the main feature region, first-order SRAF region, second-order SRAF region, and third-order SRAF region sequentially in four stages. Apply quantization constraints layer by layer outward from the main feature region to avoid loss oscillations and optimization collapse caused by a one-time strong constraint, achieving smooth discretization convergence. Overall objective function. for: (4).
[0042] in, Let be the overall objective function. The target total number of territories, , , They represent the first The target layout corresponds to the graphic error, edge placement error, and process variation zone. , , and These are the weighting coefficients for graphic error, edge placement error, process variation band, and quantization regularization term, respectively. The current active region number. For the first Each functional area For the first The quantization regularization term corresponding to each functional area For the first The mask corresponding to the target layout. Indicates the first Regular control parameters for each functional area.
[0043] Step 106: A multi-stage optimizer is used to jointly iteratively update the shared light source parameters, each mask parameter, and the region quantization parameters. In each iteration, the lithography simulation and gradient calculation corresponding to each mask can be executed in parallel, and the gradient of each mask is used to update the corresponding mask parameters; the shared light source gradient is updated uniformly by fusing the light source gradients of multiple mask samples.
[0044] Step 107: After meeting the preset convergence conditions, output the optimal shared light source distribution and the manufacturable binarized curve mask. This manufacturable binarized curve mask is a binarized mask that meets the requirements of mass production. It does not contain continuous grayscale transmittance values and can be used as the data basis for subsequent mask manufacturing. It can be directly used for mask manufacturing.
[0045] Furthermore, the results of the light source-curve mask co-optimization can be verified and evaluated using PE, EPE, and PVBand. Specifically, after iterative convergence and output of the optimal shared light source distribution and the manufacturable binary curve masks corresponding to each target layout to be optimized, the optimal shared light source distribution and binary curve masks are input into the lithography simulation link to simulate and obtain wafer imaging patterns. The PE, EPE, and PVBand index values for each group of layouts are calculated, and the results of this light source-curve mask co-optimization are verified and evaluated based on these indexes. For example, the optimization effect of this light source-curve mask co-optimization is verified and evaluated based on the relationship between each index value and a preset threshold. Among them, PE is used to evaluate the imaging fidelity between the simulated imaging pattern and the target layout; EPE is used to evaluate the edge position offset level of the lithographic imaging pattern; and PVBand is used to characterize the fluctuation range of the imaging size under process disturbance conditions and evaluate process robustness. It has significant computational efficiency advantages in multi-mask scenarios and is suitable for optimization scenarios of advanced node standard cell libraries.
[0046] This application employs a phased progressive activation strategy, applying quantization constraints layer by layer from the main feature region to the peripheral SRAF region. This effectively avoids the degradation of the main feature edges or the loss of higher-order SRAFs caused by a single strong constraint, achieving a smooth transition from continuous space to binary space. By introducing a regionalized adaptive quantization regularization term and applying differentiated biases to different regions through a movable peak quantization function, the main feature region uses a conservative bias to protect the edges, while the SRAF region uses a flexible bias to delay polarization, improving the manufacturability of the binary mask. Furthermore, by fusing PE, EPE, and PVBand constraints, a good balance is achieved between nominal imaging accuracy, edge alignment accuracy, and process robustness, resulting in overall performance superior to traditional level set algorithms. In addition, this method supports parallel acceleration across multiple masks, achieving a significant speedup in multi-mask scenarios, and is suitable for efficient optimization of large-scale standard cell libraries at advanced nodes.
[0047] The specific principle of the phased light source-curve mask collaborative optimization in this application embodiment is as follows: taking a multi-pattern shared light source as the optimization scenario, firstly, the continuous spatial high-precision imaging optimization is completed through the CTM strategy; then, the transmittance is partitioned and an adaptive movable peak quantization regularization is applied; PE, EPE, and PVBand are modeled in a unified manner with the phased constraints to achieve smooth binarization convergence, and finally, the optimal shared light source distribution and a fabricable binarized curve mask are output, taking into account accuracy, robustness and efficiency.
[0048] This application first reads multiple sets of target layouts to complete the continuous parameterized representation of the light source and mask; then, using a CTM strategy with PE as the loss function, it performs a first-stage continuous spatial joint optimization of the shared light source and multiple masks to generate a grayscale transmittance mask containing higher-order SRAF in the 0~1 interval; then, based on the grayscale mask transmittance distribution, it is dynamically divided into four regions: the main feature region, the first-order SRAF region, the second-order SRAF region, and the third-order SRAF region; then, it designs an adaptive threshold quantization regularization function with movable peak values. This function takes a minimum value of 0 at 0 and 1, and has a unique controllable maximum value point in the 0~1 interval, which is then used to determine the parameter value. The maximum value position is adjusted to achieve differentiated quantization bias in each region. Then, the regionalized quantization regularization term, PE, EPE, and PVBand are integrated into the cost function, and a phased activation strategy is set to apply quantization constraints layer by layer from the main feature region to the outer first-order SRAF region, second-order SRAF region, and third-order SRAF region. A multi-stage optimizer is then used to jointly update the shared light source, each mask parameter, and the regional quantization parameters. Parallel computation of multiple masks is supported to improve computational efficiency. Finally, the optimal shared light source distribution and a fabricable binarized curve mask are output as the final light source-curve mask collaborative optimization result. The accuracy and efficiency advantages of the embodiments of this application are verified through PE, EPE, and PVBand.
[0049] This application embodiment uses a 193nm ArF (argon fluoride) immersion lithography system to conduct a staged light source-curve mask co-optimization experiment. This immersion lithography system uses a ring illumination light source, and the ring illumination input parameter σ... in =0.6, Ring lighting output parameter σ out =0.9, mask size is 256×256, pixel size is 6nm. Optimization was performed on 5 typical layouts, and the results were compared with the traditional level set algorithm. Figure 4 Several schematic diagrams of the layout are shown, among which, Figure 4 (a) in the diagram is the initial light source diagram; Figure 4 (b) in the diagram represents the I-shaped target map; Figure 4 (c) in the image represents a multi-vertical target map; Figure 4 (d) in the diagram represents a cross-shaped target layout; Figure 4 (e) in the image represents a multi-bar target map; Figure 4 (f) in the diagram represents the target layout of both blocks. Figure 5 For different parameters The following graph shows the curves of the adaptive quantization regularization function with different parameters. The ability to control the position of the quantization peak supports regionally differentiated bias.
[0050] Figure 6The optimized main graphic and schematic diagram of the continuous binary mask region division of each order of SRAF are shown. Figure 6 (a) in the diagram is a schematic diagram of the continuous mask region division after the optimization of the I-shaped layout; Figure 6 (b) in the diagram is a schematic diagram of the continuous mask region division after the optimization of the multi-vertical strip layout; Figure 6 (c) in the diagram is a schematic diagram of a cross-shaped layout; Figure 6 (d) in the diagram is a schematic diagram of the continuous mask region division after the multi-bar layout optimization; Figure 6 (e) in the diagram is a schematic diagram of the continuous mask region division after the optimization of the two block layouts. The black part is the optimized main graphic, the white part is the first-order SRAF, the dark gray part is the second-order SRAF, and the light gray part is the third-order SRAF.
[0051] Figure 7 The diagram shows the optimized mask and the simulated wafer imaging results. Figure 7 (a) in the diagram is the optimized lighting source diagram; Figure 7 (b) in the image is the grayscale mask of the CTM optimization result corresponding to the I-shaped layout; Figure 7 (c) in the image is the grayscale mask of the CTM optimization result corresponding to the multi-vertical strip layout; Figure 7 (d) in the image represents the grayscale mask of the CTM optimization result corresponding to the cross-shaped layout. Figure 7 (e) in the image is the grayscale mask of the CTM optimization result corresponding to the multi-bar layout; Figure 7 (f) in the figure is the grayscale mask image of the CTM optimization result corresponding to the two block layouts. It can be seen that the structure is regular and the wafer imaging effect after optimization is good, with complete pattern and high edge alignment accuracy.
[0052] Figure 8 A schematic diagram of the manufacturable binarized curve mask output after iterative convergence is shown; wherein, Figure 8 (a) is the lighting source diagram obtained by the optimized method of the embodiments of this application. Figure 8 In the diagram, (b) is the binarized curve mask corresponding to the I-shaped layout. Figure 8 In the diagram, (c) represents the binarized curve mask corresponding to the multi-vertical-strip layout. Figure 8 In the diagram, (d) represents the binarized curve mask corresponding to the cross-shaped layout. Figure 8 In the diagram, (e) represents the binarized curve mask corresponding to the multi-bar layout. Figure 8In the diagram, (f) represents the binarized curve mask corresponding to both block layouts, demonstrating the strong manufacturability of the fabricated binarized curve mask. In summary, under the conditions of this embodiment, the results show that compared to the traditional level set algorithm, the method of this embodiment can reduce PE, EPE, and PVBand. The optimized mask is a fabricated binarized curve mask, maintaining SRAF integrity, with relatively clear edges of the main graphic, and high geometric fidelity in wafer imaging.
[0053] This application addresses the challenges of discretizing the transmittance function of continuous masks in multi-mask shared light source scenarios, insufficient binary manufacturability, and the potential for image quality degradation due to fixed threshold discretization. It constructs a continuous parameterized model of the shared light source and multiple masks using cosine mapping, achieving joint light source mask optimization. Secondly, the overall objective function comprises graphic error, edge placement error, process variation bands, and a phased progressive activation region quantization regularization term. This quantization regularization term employs a region partitioning strategy and an adaptive movable peak quantization function to apply differentiated dynamic threshold constraints to the main feature region and the SRAF region. Then, a phased activation strategy is used to apply quantization constraints layer by layer from the main feature region to the peripheral SRAF region, guiding the continuous mask to smoothly converge to a discretized manufacturable state. This application improves mask manufacturability and process robustness while maintaining image fidelity, making it suitable for light source-curve mask co-optimization scenarios using advanced node standard cell libraries.
[0054] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 9 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs in the non-volatile storage media to run. The database stores multiple sets of target layouts to be optimized. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When executed by the processor, the computer program implements a staged quantization method for collaborative optimization of light source curve masks.
[0055] Figure 9The structures shown are merely block diagrams of some structures related to the present application and do not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than shown in the figures, or combine certain components, or have different component arrangements. In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.
[0056] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.
[0057] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.
[0058] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of related data must comply with relevant regulations and be authorized by the owner of the corresponding device.
[0059] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).
[0060] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.
[0061] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0062] This document uses specific examples 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 methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.
Claims
1. A staged quantization method for collaborative optimization of light source curve masks, characterized in that, include: Multiple sets of target layouts to be optimized are obtained, and continuous parametric modeling of light sources and masks is performed to construct a differentiable optimization link for a single light source and multiple masks. Based on the differentiable optimization link, the first stage of continuous spatial optimization is performed without applying mask binary manufacturing constraints. The imaging error is used as the loss function to jointly optimize the shared light source and multiple sets of masks to obtain a grayscale continuous transmittance mask. Based on the transmittance distribution and graphic spatial position of the grayscale continuous transmittance mask, the grayscale continuous transmittance mask is divided into at least two types of functional regions, including the main feature region and the SRAF region. Based on the functional regions, an adaptive quantization regularization function is constructed. The adaptive quantization regularization function has the minimum penalty value at the mask transmittance positions of 0 and 1, and has an adjustable peak position in the transmittance range of 0 to 1. Different regularization function parameters are configured for different functional regions to achieve partitioned differentiated quantization constraints. Based on the adaptive quantization regularization function, a total objective function is constructed, and a phased activation strategy is set. The quantization regularization constraints corresponding to each functional region are activated in stages, in the order from the main feature region to the SRAF region. The total objective function integrates the imaging performance evaluation term and the phased activated regional quantization regularization term. The overall objective function is solved iteratively, and the shared light source parameters, each set of mask parameters, and each region quantization regularization parameter are updated jointly. In each iteration, the lithography simulation and gradient calculation corresponding to each set of masks are executed in parallel, and the shared light source is updated according to the gradient obtained by multi-mask fusion. After the iteration converges, the light source-curve mask co-optimization result is output. The light source-curve mask co-optimization result includes: the optimal shared light source distribution and the manufacturable binary curve mask corresponding to each of the target layouts to be optimized.
2. The staged quantization light source curve mask collaborative optimization method according to claim 1, characterized in that, The SRAF region includes a first-order SRAF region, a second-order SRAF region, and a third-order SRAF region. The distance between the SRAF regions of different orders and the main feature region increases from near to far. The mask transmittance decreases step by step, and the pattern size decreases step by step.
3. The staged quantization light source curve mask collaborative optimization method according to claim 1, characterized in that, The expression for the adaptive quantization regularization function is: ; ; in, For adaptive quantization regularization function, For intermediate mapping functions, To control the parameters of the quantization bias position, This indicates the mask transmittance.
4. The staged quantization light source curve mask collaborative optimization method according to claim 1, characterized in that, The expression for the overall objective function is: ; in, Let be the overall objective function. The target total number of territories, , , They represent the first The target layout corresponds to the graphic error, edge placement error, and process variation zone. , , and These are the weighting coefficients for graphic error, edge placement error, process variation band, and quantization regularization term, respectively. The current active region number. For the first Each functional area For the first The quantization regularization term corresponding to each functional area For the first The mask corresponding to the target layout. Indicates the first Regular control parameters for each functional area.
5. The staged quantization method for collaborative optimization of light source curve masks according to claim 1, characterized in that, During the first-stage continuous space optimization, a continuous transmission mask strategy is adopted for joint optimization of the first-stage continuous space. The expression of the loss function is as follows: ; in, For loss function, The target total number of territories, Indicates the first The graphic error corresponding to each target map Indicates a shared light source With corresponding mask The wafer pattern obtained through simulation under the influence of the action. For the first The mask corresponding to the target layout. Indicates the first A shared light source Indicates the first One target territory, It is the square of the L2 norm.
6. The staged quantization light source curve mask collaborative optimization method according to claim 1, characterized in that, When iteratively solving the overall objective function, a master-slave parallel architecture is adopted. The master-slave parallel architecture includes a master node and several worker nodes. The worker nodes are used to perform lithography simulation and mask gradient calculation for each layout in parallel. The master node is used to collect the light source gradients uploaded by all the worker nodes, merge them, update the globally shared light source parameters, and then distribute the updated light source parameters to each worker node.
7. The staged quantization light source curve mask collaborative optimization method according to claim 1, characterized in that, The criteria for determining convergence of the iteration include: at least one of the following indicators reaches a preset threshold: graphic error, edge placement error, or process change zone; or the number of iterations reaches a preset iteration number threshold.
8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that the processor executes the computer program to implement the staged quantization light source curve mask collaborative optimization method according to any one of claims 1-7.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the phased quantization light source curve mask collaborative optimization method as described in any one of claims 1-7.
10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the phased quantization light source curve mask collaborative optimization method as described in any one of claims 1-7.
