An EUV photoresist morphology and roughness regulation method and device based on automatic parameter optimization and related products

Abstract

The application provides an EUV photoresist topography and roughness control method and device based on automatic parameter optimization, related products, and relates to extreme ultraviolet lithography technology and semiconductor manufacturing field. The method comprises constructing a random exposure model, simulating secondary electron random walk and scattering effect, and establishing a near neighbor condensation crosslinking mechanism based on lattice topological constraints; a discrete random development model is constructed based on the Gillese ratio algorithm, and the competitive reaction between the developing solution and the metal oxide core is simulated; the quantum yield, the electron blur length, and the number of development critical sites are used as optimization variables, a fitness function is constructed based on the three-dimensional topographic feature parameters of the photoresist, and the target parameter combination is calculated and output by an optimization algorithm. The application can realize fine control of photoresist topography and effective suppression of roughness, and improve the process window and manufacturing yield of extreme ultraviolet lithography.

Description

Technical Field

[0001] This application relates to the fields of extreme ultraviolet lithography technology and semiconductor manufacturing, and in particular to a method, apparatus and related products for controlling the morphology and roughness of EUV photoresist based on automated parameter optimization. Background Technology

[0002] As integrated circuit manufacturing processes continue to advance to 3nm and below, EUV (Extreme Ultraviolet) lithography has become a core patterning technology in advanced processes. Among these, MOR (Metal Oxide Resist) has become the core photoresist material system in the EUV field. Currently, for EUV photoresist morphology control, the industry mostly uses continuous dielectric models to simulate the lithography process and adjusts material formulations and process parameters through manual trial and error to optimize key indicators such as LER (Line Edge Roughness) and LWR (Line Width Roughness) of the photoresist, ensuring the quality of the lithographic pattern.

[0003] Existing photoresist morphology control suffers from significant technical shortcomings. On the one hand, traditional continuous medium models, based on the assumption of a homogeneous medium in the photoresist, cannot accurately describe the microscopic morphology fluctuations caused by photon shot noise, secondary electron random walk, and the randomness of crosslinking reactions in EUV. They also struggle to accurately simulate core chemical reactions in metal oxide photoresists, such as topologically constrained condensation crosslinking, ligand substitution, and oxygen bridge bond breaking, thus failing to reveal the root cause of roughness at the microscopic level. On the other hand, photoresist morphology control involves the coupled optimization of multiple parameters, such as quantum yield, electron blur length, and the number of critical development sites. Traditional manual trial-and-error methods are extremely inefficient and struggle to lock in the globally optimal combination of process parameters. Furthermore, the lack of systematic model calibration methods can easily lead to the deterioration of photoresist edge roughness, severely impacting the electrical performance and production yield of devices under advanced processes.

[0004] Therefore, how to effectively control the morphology of photoresist has become an urgent technical problem to be solved. Summary of the Invention

[0005] In view of the above problems, this application is made to provide a method, apparatus, and related products for controlling the morphology and roughness of EUV photoresist based on automated parameter optimization, which overcomes or at least partially solves the above problems. The technical solution is as follows: In a first aspect, a method for controlling the morphology and roughness of EUV photoresist based on automated parameter optimization is provided, the method comprising: Initialize the numerical range of the optimization variables and generate an initial parameter population containing at least one candidate parameter combination, each candidate parameter combination including the values ​​of quantum yield, electronic blur length and number of development critical sites; Each candidate parameter combination in the initial parameter population is input into the random exposure model and the discrete random development model for forward simulation to generate the three-dimensional morphology of the photoresist corresponding to each candidate parameter combination. Extract the morphological feature parameters of each photoresist three-dimensional morphology, construct a fitness function based on the morphological feature parameters, calculate the fitness value of each candidate parameter combination, and determine whether the fitness value meets the preset convergence condition; When the fitness value does not meet the preset convergence condition, the numerical range of the optimization variable is updated using the optimization algorithm, a new generation of parameter population is generated, and the forward simulation, morphological feature parameter extraction and fitness value calculation steps are repeated until the preset convergence condition is met. When the fitness value meets the preset convergence condition, the corresponding candidate parameter combination is output as the target parameter combination.

[0006] In one possible implementation, before inputting each candidate parameter combination from the initial parameter population into the random exposure model and the discrete random development model for forward simulation to generate the photoresist 3D morphology corresponding to each candidate parameter combination, the method further includes: A random exposure model was constructed to simulate the random walk and scattering effects of secondary electrons, and a nearest-neighbor condensation crosslinking mechanism was established based on lattice topological constraints. A discrete random development model is constructed based on the Gillespie algorithm to simulate the competitive reaction between developer molecules and the core surface sites of metal oxides.

[0007] In one possible implementation, the construction of the random exposure model, which simulates the random walk and scattering effects of secondary electrons, includes: The photon absorption process is modeled as a random process following a Poisson distribution, generating photoelectrons; The photoelectron generation process is modeled as a random event defined by quantum yield, which excites secondary electrons. The Gaussian convolution algorithm is used to simulate the random walk and scattering effect of secondary electrons, and the spatial diffusion range of secondary electrons is limited by the electron fuzzy length parameter. Secondary electrons are used to activate the ligand dissociation on the surface of the metal oxide core, generating active sites. This activates the metal oxide core after ligand dissociation, and the active sites participate in the oxygen bridge bond formation reaction in the nearest-neighbor condensation crosslinking mechanism.

[0008] In one possible implementation, the establishment of the nearest-neighbor condensation crosslinking mechanism based on lattice topological constraints includes: The photoresist space is divided into the first discrete lattice unit; In a discrete lattice unit, the effective range of the crosslinking reaction is limited to a randomly selected six-neighbor area; Identify the first metal oxide core that is in an active state and carries an active site, and determine the activation state of the nearest neighbor second metal oxide core within the six-neighbor range of the first metal oxide core. When both metals are in an active state, oxygen bridge bonds are created between the first metal oxide core and the second oxygen metal oxide core, and the corresponding active sites are consumed to obtain the second discrete lattice unit.

[0009] One possible implementation of the discrete random imaging model based on the Gillespie algorithm includes: The input layer for constructing the discrete random development model is the second discrete lattice unit. The input parameters include the ligand binding site state on the surface of each metal oxide core in the second discrete lattice unit, and the oxygen bridge bond connection state generated between each two metal oxide cores through the nearest neighbor condensation crosslinking mechanism. Construct a discrete random development model reaction system, define the competing reactions included in the simulated development process, and the competing reactions include at least one of ligand substitution reaction, solvent reverse substitution reaction, and oxygen bridge bond breaking reaction; Based on the Gillespie algorithm, a dynamic evolution layer of the discrete random development model is constructed. Rules for calculating the probability of competitive reactions, the calculation of reaction time step, the reaction execution rules, and the dynamic update rules of the metal oxide core state are set to execute competitive reactions in the development process and update the ligand binding site state and oxygen bridge bond connection state of the metal oxide core in the second discrete lattice unit to obtain the third discrete lattice unit. An output decision layer of a discrete random development model is constructed, a dissolution and stripping decision rule for metal oxide cores is set, the dissolution and stripping result of metal oxide cores in the third discrete lattice unit is determined based on the dissolution and stripping decision rule, the distribution state of metal oxide cores in the third discrete lattice unit is updated, and the three-dimensional morphology of the developed photoresist is generated.

[0010] In one possible implementation, the topographic feature parameters include line edge roughness, line width roughness, and simulation critical size, and the fitness function satisfies: F=w1×LER+w2×LWR+w3×|CD_target-CD_sim| Where w1 is the first weighting coefficient, LER is the line edge roughness, w2 is the second weighting coefficient, LWR is the line width roughness, w3 is the third weighting coefficient, CD_target is the preset target critical size, and CD_sim is the simulation critical size.

[0011] Secondly, an automated parameter optimization-based device for controlling the morphology and roughness of EUV photoresist is provided, the device comprising: The first model building unit is used to build a random exposure model, which simulates the random walk and scattering effect of secondary electrons and establishes a nearest-neighbor condensation crosslinking mechanism based on lattice topological constraints. The second model building unit is used to build a discrete random development model based on the Gillespie algorithm, which simulates the competitive reaction between developer molecules and the core surface sites of metal oxides. The parameter initialization unit is used to initialize the numerical range of the optimization variables and generate an initial parameter population containing at least one candidate parameter combination. Each candidate parameter combination includes the values ​​of quantum yield, electronic blur length, and number of development critical sites. The forward simulation unit is used to input each candidate parameter combination in the initial parameter population into the random exposure model and the discrete random development model for forward simulation, and generate the three-dimensional morphology of the photoresist corresponding to each candidate parameter combination. The fitness calculation unit is used to extract the morphological feature parameters of each photoresist three-dimensional morphology, construct a fitness function based on the morphological feature parameters, calculate the fitness value of each candidate parameter combination, and determine whether the fitness value meets the preset convergence condition. The optimization and update unit is used to update the numerical range of the optimization variables using an optimization algorithm when the fitness value does not meet the preset convergence condition, generate a new generation of parameter population, and repeatedly execute the forward simulation, morphological feature parameter extraction and fitness value calculation steps until the preset convergence condition is met. The parameter output unit is used to output the corresponding candidate parameter combination as the target parameter combination when the fitness value meets the preset convergence condition.

[0012] Thirdly, an electronic device is provided, comprising a processor and a memory, wherein the memory stores a computer program, and the processor is configured to run the computer program to perform the EUV photoresist morphology and roughness control method based on automated parameter optimization as described in any of the preceding claims.

[0013] Fourthly, a storage medium is provided, the storage medium storing a computer program, wherein the computer program is configured to execute the EUV photoresist morphology and roughness control method based on automated parameter optimization as described above when running.

[0014] Fifthly, a computer program product is provided, including a computer program configured to execute the method for controlling the morphology and roughness of EUV photoresist based on automated parameter optimization as described above during runtime.

[0015] Using the above technical solutions, the embodiments of this application provide a method, apparatus, and related products for controlling the morphology and roughness of EUV photoresist based on automated parameter optimization. The method includes constructing a random exposure model to simulate the random walk and scattering effects of secondary electrons, and establishing a nearest-neighbor condensation crosslinking mechanism based on lattice topological constraints; constructing a discrete random development model based on the Gillespie algorithm to simulate the competitive reaction between the developer and the metal oxide core; using quantum yield, electron blur length, and the number of critical development sites as optimization variables, constructing a fitness function based on the three-dimensional morphology characteristic parameters of the photoresist, and iteratively calculating and outputting the target parameter combination through an optimization algorithm, thereby achieving refined control of the photoresist morphology and effective suppression of roughness, improving the process window and manufacturing yield of extreme ultraviolet lithography. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the description of the embodiments of this application will be briefly introduced below.

[0017] Figure 1 The flowchart of the EUV photoresist morphology and roughness control method based on automated parameter optimization provided in the embodiments of this application is shown. Figure 2a This paper shows an example of the light intensity distribution within the photoresist of the EUV photoresist morphology and roughness control method based on automated parameter optimization provided in an embodiment of this application. Figure 2b This paper shows an example of the random distribution of photon absorption in the EUV photoresist morphology and roughness control method based on automated parameter optimization provided in an embodiment of this application. Figure 2c This paper shows an example diagram of the secondary electron space density distribution of the EUV photoresist morphology and roughness control method based on automated parameter optimization provided in an embodiment of this application. Figure 3a The power spectral density analysis curve of the first candidate parameter combination provided by the EUV photoresist morphology and roughness control method based on automated parameter optimization in the embodiments of this application is shown. Figure 3b This shows a power spectral density analysis curve formed by the second candidate parameter combination of the EUV photoresist morphology and roughness control method based on automated parameter optimization provided in the embodiments of this application; Figure 4a The diagram shows an example of the three-dimensional distribution of the crosslinking network of the EUV photoresist morphology and roughness control method based on automated parameter optimization provided in this application embodiment at an exposure dose of 20.0 mJ / cm². Figure 4bThe diagram shows an example of the three-dimensional distribution of the crosslinking network of the EUV photoresist morphology and roughness control method based on automated parameter optimization provided in this application embodiment at an exposure dose of 40.0 mJ / cm². Figure 4c The diagram shows an example of the three-dimensional distribution of the crosslinking network of the EUV photoresist morphology and roughness control method based on automated parameter optimization provided in this application embodiment at an exposure dose of 90.0 mJ / cm². Figure 4d The diagram shows an example of the three-dimensional distribution of the maximum continuous crosslinked network under an exposure dose of 90.0 mJ / cm², based on the automated parameter optimization method for controlling the morphology and roughness of EUV photoresist provided in this application embodiment. Figure 5 This paper shows an example diagram of the spatial distribution of oxygen bridge bonds in the EUV photoresist morphology and roughness control method based on automated parameter optimization provided in an embodiment of this application. Figure 6a The image shows a three-dimensional example of the initial bond distribution of the metal oxide core before development using the EUV photoresist morphology and roughness control method based on automated parameter optimization provided in this application embodiment. Figure 6b This paper shows an example diagram of the three-dimensional distribution of the development wavefront under the first development time for the EUV photoresist morphology and roughness control method based on automated parameter optimization provided in the embodiments of this application. Figure 6c This paper shows an example diagram of the three-dimensional distribution of the development wavefront under the second development time for the EUV photoresist morphology and roughness control method based on automated parameter optimization provided in the embodiments of this application. Figure 7a This paper shows an example of the three-dimensional morphology of EUV photoresist after development, based on the automated parameter optimization method for controlling the morphology and roughness of EUV photoresist provided in the embodiments of this application. Figure 7b This paper shows an example of the three-dimensional morphology projection of the EUV photoresist morphology and roughness control method based on automated parameter optimization provided in the embodiments of this application after development. Figure 8a This paper shows an example of the three-dimensional morphology of the photoresist formed by the first candidate parameter combination using the EUV photoresist morphology and roughness control method based on automated parameter optimization provided in an embodiment of this application. Figure 8b This paper shows an example of the three-dimensional morphology of the photoresist formed by the first candidate parameter combination using the EUV photoresist morphology and roughness control method based on automated parameter optimization provided in an embodiment of this application. Figure 9 This paper shows a structural diagram of an EUV photoresist morphology and roughness control device based on automated parameter optimization provided in an embodiment of this application. Figure 10 A structural diagram of an electronic device provided in an embodiment of this application is shown. Detailed Implementation

[0018] Exemplary embodiments of the present application will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this application will be thorough and complete, and will fully convey the scope of the present application to those skilled in the art.

[0019] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such use can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the term "comprising" and its variations should be interpreted as open-ended terms meaning "including but not limited to."

[0020] The following explains some concepts and terms involved in the method, apparatus and related products for controlling EUV photoresist morphology and roughness based on automated parameter optimization provided in the embodiments of this application.

[0021] 1. Line edge roughness (LER) is a parameter describing the degree of microscopic fluctuation in the horizontal deviation of the photoresist pattern edge from an ideal straight line. LER mainly originates from photon shot noise, random diffusion of secondary electrons, and the non-uniformity of chemical reactions within the photoresist during extreme ultraviolet (EUV) lithography. The magnitude of LER directly affects the electrical performance of transistors and chip manufacturing yield. It is typically quantitatively characterized using methods such as root mean square deviation (RMSD) after obtaining the pattern edge contour using a scanning electron microscope.

[0022] 2. Linewidth roughness (LWR) is a parameter describing the random fluctuations in the linewidth of a photoresist pattern along the line direction, reflecting the cumulative effect of roughness on both sides of the pattern's edges. LWR and LER are closely related in physical mechanism, both originating from random effects during the photolithography process. LWR has a more direct impact on device performance and is one of the key indicators for evaluating the stability of photolithography processes.

[0023] 3. Critical Dimension (CD) is a key parameter in integrated circuit manufacturing that characterizes the size of patterned features, typically referring to the minimum width or spacing of lines. In photolithography, the uniformity of CD determines the consistency of device performance and is one of the core optimization goals for controlling the morphology of photoresist.

[0024] 4. Power Spectral Density (PSD) is a tool used to analyze the frequency domain characteristics of photoresist edge roughness. By decomposing roughness into the power distribution of different spatial frequency components, it reveals the contribution of each frequency component to morphology fluctuations. PSD analysis can effectively distinguish roughness components caused by different physical mechanisms (such as photon shot noise, material inhomogeneity, etc.), providing a frequency domain dimension of evaluation for photoresist morphology control.

[0025] 5. Genetic Algorithm (GA) is a heuristic optimization algorithm that simulates natural selection and genetic mechanisms. GA maps optimization variables to chromosomes through encoding, and uses genetic operations such as selection, crossover, and mutation to iteratively evolve the population, efficiently searching for the global optimum in the parameter space. It is suitable for complex optimization problems involving multiple coupled parameters in photoresist applications.

[0026] 6. Bayesian Optimization (BO) is a global optimization algorithm based on a probabilistic surrogate model. It fits the objective function through models such as Gaussian processes, balances exploration and utilization using the acquisition function, and intelligently selects the next set of evaluation points during the iteration process. It is suitable for photoresist forward simulation optimization scenarios where the objective function evaluation cost is high.

[0027] 7. Particle Swarm Optimization (PSO) is a swarm intelligence optimization algorithm that simulates the foraging behavior of bird flocks. It initializes a set of random particles, each representing a candidate parameter combination. The particles follow their own optimal position and the group's optimal position in the search space to update their velocity and move their position. It iteratively converges to the global optimum and is suitable for fast optimization in continuous parameter spaces.

[0028] 8. Simulated Annealing (SA) is a stochastic optimization algorithm that simulates the annealing process of metals. By introducing the Metropolis acceptance criterion, it accepts inferior solutions with a certain probability during the search process, effectively avoiding getting trapped in local optima. As the number of iterations increases, the acceptance probability gradually decreases, eventually converging to the global optimum. It is suitable for complex nonlinear optimization problems.

[0029] 9. Quantum yield is a parameter describing the average number of electrons excited per absorbed photon, and it is a key indicator affecting the sensitivity and roughness of photoresist. Quantum yield reflects the efficiency of the photoresist material in converting incident photon energy into secondary electrons. In extreme ultraviolet lithography, a higher quantum yield results in more secondary electrons excited per unit dose, leading to higher photoresist sensitivity. Simultaneously, quantum yield is negatively correlated with line edge roughness; improving quantum yield helps suppress morphology fluctuations caused by random effects, making it an important optimization variable in photoresist formulation design.

[0030] 10. Electron Blur Length is a parameter describing the effective spatial range of secondary electrons diffusing outward from the photon absorption point and activating ligand dissociation. It affects exposure resolution and line edge roughness. Secondary electrons undergo a random walk process in the photoresist, and their diffusion distance determines the spatial distribution range of energy deposition. The longer the electron blur length, the wider the range of influence of secondary electrons, which may lead to non-specific exposure in adjacent areas, increasing line edge roughness and reducing resolution.

[0031] 11. Critical Site Number for Development (CSD) is the threshold number of sites on the surface of a metal oxide core that can be replaced by solvent. When this threshold is exceeded and all oxygen bridges are broken, dissolution and stripping are considered complete. CSD reflects the minimum degree of chemical modification required for the metal oxide core to dissolve from the cross-linked network. It directly affects the morphology and roughness characteristics of the photoresist after development and is negatively correlated with line edge roughness. That is, the higher the CSD, the more difficult it is for the core to dissolve, which may lead to incomplete development or increased roughness.

[0032] 12. The fitness function is used in automated optimization to evaluate the quality of parameter combinations. In this method, it is defined as the weighted sum of line edge roughness, linewidth roughness, and critical size error, guiding the optimization algorithm to search for the optimal combination of process parameters. By adjusting the weight coefficients of each index, the importance of different morphological features can be flexibly balanced, achieving precise control over the three-dimensional morphology of the photoresist.

[0033] 13. Spectral Coverage is a metric used for model calibration. It measures the extent to which the test pattern covers the frequency domain, ensuring that the model has good response capabilities to different frequency characteristics.

[0034] 14. The metal oxide core is a nanoscale structural unit in photoresist composed of metal atoms and oxygen bridges, and it is also the basic building block of the cross-linked network. Metal oxide cores are interconnected through oxygen bridges to form a three-dimensional cross-linked network, which determines the mechanical properties and solubility characteristics of the photoresist. During exposure, ligands on the core surface are dissociated by secondary electrons to generate active sites; during development, the ligands on the core surface are replaced by solvent, and the oxygen bridges break, causing the core to peel off from the network, affecting the final photoresist morphology.

[0035] 15. Ligand substitution reaction is a chemical reaction in which solvent molecules replace the original ligands on the surface of the metal oxide core during the development process. It is one of the prerequisites for the dissolution of the metal oxide core.

[0036] 16. Oxygen bridge breaking is a reaction in which the oxygen bridges connecting the metal oxide cores are broken during the development process, causing the metal oxide cores to peel off from the crosslinked network.

[0037] 17. The process window is the range of process parameters that can be varied while ensuring the quality of the photolithography pattern. It reflects the stability and robustness of the process.

[0038] As mentioned in the background section, extreme ultraviolet (EUV) lithography has become a core patterning technology in advanced manufacturing processes. Metal oxide (MOD) photoresists have become the core photoresist material system. However, with the continuous shrinking of feature sizes, photon shot noise, the inhomogeneity of material microstructure, and the stochastic effects caused by random chemical reactions have become increasingly significant. These directly lead to the deterioration of photoresist line edge roughness and linewidth roughness, severely affecting the electrical performance and yield of devices. Therefore, how to effectively control the photoresist morphology and suppress edge roughness has become a key problem that urgently needs to be solved in the field of EUV lithography.

[0039] The relevant technical solutions mainly fall into the following categories: 1. Continuous Medium Model Approach: This approach constructs a photoresist model based on the continuous medium assumption, treating the photoresist as a homogeneous medium and simulating the exposure and development processes by solving optical and diffusion equations. This approach assumes that photon absorption and secondary electron diffusion within the photoresist are continuous and deterministic, neglecting microscopic randomness. However, in EUV lithography, photon absorption events are rare, and secondary electron diffusion paths are random. The continuous medium model cannot accurately describe the microscopic morphology fluctuations caused by photon shot noise and random walks of secondary electrons, resulting in insufficient roughness prediction accuracy.

[0040] 2. Manual Trial and Error Optimization: In photolithography process development, photoresist material parameters (such as quantum yield and electron blur length) and process parameters (such as development time and developer concentration) are adjusted through experimental design or experience. Exposure experiments and morphology checks are repeatedly conducted to screen for optimal parameter combinations. This approach relies on engineers' accumulated experience, requires numerous experimental iterations, is time-consuming and costly, and struggles to guarantee finding the globally optimal solution, making it unsuitable for the rapid iteration demands of process development.

[0041] 3. Empirical Calibration Scheme: This scheme calibrates the parameters of the photoresist model based on a small number of test patterns or standard measurement data. It typically focuses only on matching critical dimensions and ignores the verification of morphological characteristics at different spatial frequencies. This scheme lacks a comprehensive correction mechanism for model parameters in both the frequency and spatial domains. The calibrated model has insufficient generalization ability under different pattern structures or process conditions, making it difficult to accurately predict the morphological characteristics of complex patterns.

[0042] It is evident that the shortcomings of existing technologies make it difficult to control the morphology of photoresist and effectively suppress edge roughness, thus limiting the expansion of the EUV lithography process window and the improvement of product yield.

[0043] To address the aforementioned technical problems, embodiments of this application provide a method for controlling the morphology and roughness of EUV photoresist based on automated parameter optimization, such as... Figure 1 As shown, the method for controlling the morphology and roughness of EUV photoresist based on automated parameter optimization may include the following steps S101 to S105: Step S101: Initialize the numerical range of the optimization variables and generate an initial parameter population containing at least one candidate parameter combination, where each candidate parameter combination includes the values ​​of quantum yield, electronic blur length, and number of development critical sites.

[0044] In one possible implementation, when initializing the numerical range of the optimization variables, corresponding upper and lower thresholds can be set for quantum yield, electron blur length, and number of development critical sites, respectively. The numerical range of quantum yield can be set according to the photon absorption characteristics of the photoresist material, the numerical range of electron blur length can be set according to the diffusion characteristics of secondary electrons, and the numerical range of number of development critical sites can be set according to the total number of surface sites of the metal oxide core.

[0045] In another possible implementation, the aforementioned initial parameter population is the set of initial parameters for iteration. The number of candidate parameter combinations contained in the initial parameter population can be flexibly adjusted according to computing resources and parameter space dimensions. This embodiment does not impose any restrictions on this.

[0046] In one example, the quantum yield can be set to a range of 0.1 to 10, the electron blur length to a range of 0.5 nm to 5 nm, and the number of development critical sites to a range of 2 to 10. The initial parameter population is set to 50 groups. Fifty candidate parameter combinations are generated within the range of each optimization variable by random sampling. Each candidate parameter combination includes the values ​​of quantum yield, electron blur length, and number of development critical sites, thereby completing the generation of the initial parameter population.

[0047] In one possible implementation, the initial parameter population can be generated using Latin Hypercube Sampling (LHS) instead of random generation, ensuring a uniform distribution of the initial parameter population in the parameter space and improving the global search capability of the optimization algorithm.

[0048] In another possible implementation, the above-mentioned optimization variables may also include other process parameters such as developer polarity factor, exposure dose, and development time, in order to expand the optimization dimensions and adapt to different photoresist materials and process scenarios.

[0049] Step S102: Input each candidate parameter combination in the initial parameter population into the random exposure model and the discrete random development model for forward simulation to generate the three-dimensional morphology of the photoresist corresponding to each candidate parameter combination.

[0050] In one possible implementation, the aforementioned forward simulation simulates the complete photolithography process from extreme ultraviolet exposure to development. The input of the forward simulation is a combination of candidate parameters, and the output is the three-dimensional morphology of the photoresist corresponding to that set of parameters.

[0051] In one example, such as Figure 2a As shown, the input to the random exposure model is the bulk light intensity distribution within the photoresist during extreme ultraviolet lithography. Figure 2a In this diagram, X, Y, and Z represent the X-axis, Y-axis, and Z-axis of a three-dimensional coordinate system, respectively. The XY Bulk Image (z=20.0nm) represents the light intensity distribution of the photoresist phase at a depth of 20.0 nanometers within the plane formed by the X and Y axes, where nm (nanometer) is a unit of measurement. The XZ Bulk Image (Cross-section) represents the light intensity distribution of the photoresist phase within the cross-section formed by the X and Z axes. Figure 2a In this context, Width (nm) represents the width in nanometers, and Depth (nm) represents the depth in nanometers. The light intensity values ​​gradually change from 0 to 1.0, corresponding to the exposed and unexposed areas of the lithographic target lines, providing a spatial distribution reference for subsequent photon absorption and electron excitation processes.

[0052] In another example, such as Figure 2bAs shown, when running the random exposure model, the photon absorption process is modeled as a random process following a Poisson distribution, resulting in random distribution maps of photon absorption in the XY plane and XZ cross section within the photoresist. Here, XYPhotons represents the random distribution map of photon absorption in the plane formed by the X and Y axes, and XZ Photons represents the random distribution map of photon absorption in the cross section formed by the X and Z axes. Figure 2b The discretely distributed points represent the spatial location and quantity of photons absorbed by the photoresist, intuitively demonstrating the random effect of photon shot noise in extreme ultraviolet lithography, and providing a basis for the subsequent excitation of photoelectrons and secondary electrons.

[0053] In another example, such as Figure 2c As shown, based on the input electron fuzzy length parameter, the Gaussian convolution algorithm is used to simulate the random walk and scattering effect of secondary electrons, and the spatial density distribution of secondary electrons in the XY plane and XZ cross section in the photoresist is obtained. Here, XY Electrons represents the spatial density distribution of secondary electrons in the plane formed by the X-axis and Y-axis, and XZ Electrons represents the spatial density distribution of secondary electrons in the cross section formed by the X-axis and Z-axis. Figure 2c The color change from blue to red represents the change in secondary electron density from low to high, clearly showing the spatial range of secondary electron diffusion from the photon absorption point, and intuitively demonstrating the regulatory effect of the electron fuzzy length parameter on the effective range of secondary electron-activated ligand dissociation.

[0054] In one possible implementation, for multiple sets of candidate parameter combinations in the initial parameter population, multiple sets of forward simulations can be run simultaneously using parallel computing to improve simulation efficiency and shorten the optimization cycle.

[0055] In one example, for 50 candidate parameter combinations in the initial parameter population, an independent computing thread is assigned to each parameter combination. Each parameter combination is simultaneously input into the random exposure model and the discrete random development model, and 50 Monte Carlo simulations are run in parallel. Each simulation completes the entire process simulation from exposure to development, and finally outputs 50 sets of photoresist three-dimensional morphologies that correspond one-to-one with the candidate parameter combinations.

[0056] One possible implementation is to use a continuous medium model combined with random perturbation correction to replace the forward simulation process. Based on the continuous medium model, Gaussian noise or random field perturbation is introduced to approximate photon shot noise and secondary electron diffusion effects. This improves computational efficiency while ensuring the accuracy of the basic simulation, and is suitable for fast process window scanning scenarios.

[0057] Another possible implementation is to use CGMD (Coarse-Grained Molecular Dynamics) to replace the forward simulation process. By simulating the cross-linking and dissolution process between metal oxide cores through coarse-grained molecular dynamics, it can replace the discrete lattice model and is suitable for more detailed photoresist material design and analysis scenarios.

[0058] Step S103: Extract the morphological feature parameters of each photoresist three-dimensional morphology, construct a fitness function based on the morphological feature parameters, calculate the fitness value of each candidate parameter combination, and determine whether the fitness value meets the preset convergence condition.

[0059] In one possible implementation, the aforementioned topographic feature parameters include at least line edge roughness, line width roughness, and simulation critical size. Line edge roughness measures the degree of fluctuation at the edge of the photoresist line, line width roughness measures the degree of variation in line width, and simulation critical size measures the absolute accuracy of the line width.

[0060] In one example, such as Figure 3a As shown, based on the three-dimensional morphology of the photoresist after development for the first candidate parameter combination, the power spectral density analysis of the line edge roughness can be completed, and the corresponding analysis curve can be obtained. Among them, LER PSD Analysis represents the power spectral density analysis of the line edge roughness. This power spectral density analysis curve can quantitatively reveal the influence of different frequency components on morphology fluctuations. Data point index represents the data point index, which is used to identify the sequence number of the sampled data points in the power spectral density analysis. Figure 3a The small fluctuation amplitude of the medium curve corresponds to good edge roughness of the photoresist lines. The power spectral density curve matching degree can be incorporated into the morphological feature parameters for quantitative evaluation of the target optimization.

[0061] In one example, such as Figure 3b As shown, the three-dimensional morphology of the photoresist after development was analyzed based on the second candidate parameter combination with poor morphology. The analysis curve obtained showed large fluctuations and a high proportion of high-frequency components, corresponding to poor edge roughness of the photoresist lines.

[0062] In one possible implementation, the fitness function described above is constructed as a weighted sum of line edge roughness, line width roughness, and critical size error, i.e. F=w1×LER+w2×LWR+w3×|CD_target-CD_sim| Where w1 is the first weighting coefficient, LER is the line edge roughness, w2 is the second weighting coefficient, LWR is the line width roughness, w3 is the third weighting coefficient, CD_target is the preset target critical size, and CD_sim is the simulation critical size.

[0063] In one possible implementation, the aforementioned preset convergence conditions may include any one or more combinations of the following: line edge roughness is lower than a first preset threshold, fitness value change is lower than a second preset threshold, and the number of iterations reaches a preset maximum. For example, convergence can be determined when the line edge roughness is less than 2.0 nanometers, or when the fitness value change of the parameter population for three consecutive generations is less than 0.01.

[0064] In one possible implementation, the aforementioned morphological feature parameters may also include power spectral density curve matching degree, which is used to measure the degree of matching between the simulated morphology and the target morphology in the frequency domain, thereby improving the control capability of the morphology in the frequency domain.

[0065] In one possible implementation, the fitness function can also incorporate CD uniformity as an optimization objective to evaluate the dimensional consistency of photoresist lines throughout the exposure field, making it suitable for optimization scenarios in large-scale mass production processes.

[0066] Step S104: When the fitness value does not meet the preset convergence condition, the numerical range of the optimization variable is updated using the optimization algorithm to generate a new generation of parameter population, and the forward simulation, morphological feature parameter extraction and fitness value calculation steps are repeated until the preset convergence condition is met.

[0067] In one possible implementation, the optimization algorithm is an intelligent algorithm that automatically searches for the optimal combination of parameters in the parameter space, and the new generation parameter population is a set of parameters generated by updating the previous generation population through the optimization algorithm based on the fitness evaluation results of the previous generation population.

[0068] In another possible implementation, when updating the numerical range of the optimization variables, the numerical range corresponding to high-fitness parameters can be narrowed and the numerical range corresponding to low-fitness parameters can be expanded based on the fitness value distribution of the previous generation of parameter population, thus guiding the optimization algorithm to search towards the global optimal solution.

[0069] In one possible implementation, the optimization algorithm described above is a genetic algorithm. When the fitness value does not meet the preset convergence condition, selection, crossover, and mutation operations are performed: based on the fitness value of each candidate parameter combination in the current parameter population, a candidate parameter combination with higher fitness is selected as the parent generation; a crossover operation is performed on the parent parameters to generate offspring parameter combinations; and random mutation is performed on the offspring parameters to introduce new parameter changes and generate a new generation of parameter population.

[0070] In another possible implementation, the genetic algorithm described above can employ roulette wheel selection or tournament selection for the selection operation; single-point crossover or multi-point crossover for the crossover operation; and Gaussian mutation or uniform mutation for the mutation operation.

[0071] In one example, after fitness calculation, none of the 50 candidate parameter combinations in the initial parameter population meet the preset convergence condition. At this point, a genetic algorithm is used to update the optimization variables. First, a tournament selection is used to select the top 20 high-quality parameter combinations with the best fitness values. Then, a single-point crossover operation is performed on the selected parameter combinations to generate 30 new parameter combinations. Finally, a random mutation operation with a 5% probability is performed on all 50 parameter combinations to generate a new generation of 50 candidate parameter combinations. The new generation of parameter population is then re-input into the model, and the steps of forward simulation, morphological feature extraction, and fitness calculation are repeated until a parameter combination that meets the preset convergence condition appears in a certain generation of the population.

[0072] In one possible implementation, the above optimization algorithm can employ Bayesian optimization. Bayesian optimization constructs a Gaussian process surrogate model, predicts the fitness distribution of unsampled regions based on historical simulation results, and uses a sampling function to select the next optimal sampling point, thereby reducing the number of forward simulations while ensuring optimization accuracy.

[0073] In one possible implementation, the optimization algorithm can be Bayesian optimization instead of genetic algorithm. This is suitable for parameter optimization scenarios with high dimensionality and small sample size. By constructing a Gaussian process surrogate model to predict the fitness value of parameter combinations, the number of forward simulations can be greatly reduced, thus improving optimization efficiency.

[0074] In another possible implementation, the optimization can use the Particle Swarm Optimization (PSO) algorithm or the Simulated Annealing (SA) algorithm instead of the Genetic Algorithm. This is suitable for scenarios with small continuous parameter spaces or limited computational resources, and can also achieve the updating of optimization variables and the generation of a new generation of parameter populations.

[0075] Step S105: When the fitness value meets the preset convergence condition, output the corresponding candidate parameter combination as the target parameter combination.

[0076] In one possible implementation, the above-mentioned target parameter combination is a globally optimal parameter combination obtained through automated iterative optimization, which can be directly used to guide the material formulation design of metal oxide photoresists and the adjustment of extreme ultraviolet lithography process parameters.

[0077] In another possible implementation, when there are multiple candidate parameter combinations that satisfy the preset convergence conditions in a certain generation of parameter population, the candidate parameter combination with the smallest fitness value can be selected as the target parameter combination output.

[0078] In one possible implementation, after outputting the target parameter combination, the frequency domain characteristics of the photoresist three-dimensional morphology corresponding to the target parameter combination can be verified by power spectral density analysis technology to confirm the suppression effect of edge roughness. At the same time, the corresponding photolithography process window can be determined based on the target parameter combination.

[0079] In one example, when the LER corresponding to the tenth candidate parameter combination in the third-generation parameter population is 1.8nm, which satisfies the preset convergence condition that the LER is less than 2.0nm, and the tenth candidate parameter combination is the parameter combination with the smallest fitness value in the third-generation parameter population, then the tenth parameter combination can be output as the target parameter combination. The quantum yield, electronic blur length, and number of development critical sites contained in this set of parameters are the optimal parameters for achieving control of photoresist morphology and roughness.

[0080] In one possible implementation, the output target parameter combination can adopt the Pareto optimal solution output method of multi-objective optimization. When the optimization variables include multiple dimensions such as roughness, sensitivity, and process window width, multiple sets of Pareto optimal parameter combinations can be output for different process scenarios.

[0081] In another possible implementation, after outputting the target parameter combination, the random exposure model and the discrete random development model can be calibrated using a method based on spectral coverage. Based on the frequency domain coverage relationship between the test patterns, key pattern combinations containing dose anchoring patterns, general patterns, and special patterns are selected to calibrate the model parameters of the random exposure model and the discrete random development model, thereby further improving the accuracy of the model simulation.

[0082] This application embodiment provides a possible implementation method. Before step S102 above, which inputs each candidate parameter combination in the initial parameter population into the random exposure model and the discrete random development model for forward simulation to generate the photoresist three-dimensional morphology corresponding to each candidate parameter combination, it may further include the following steps A1 to A2: Step A1: Construct a random exposure model to simulate the random walk and scattering effects of secondary electrons, and establish a nearest-neighbor condensation crosslinking mechanism based on lattice topological constraints. Step A2: Construct a discrete random development model based on the Gillespie algorithm to simulate the competitive reaction between developer molecules and the core surface sites of metal oxides.

[0083] This application provides a possible implementation method. Step A1 above constructs a random exposure model to simulate the random walk and scattering effects of secondary electrons. Specifically, it may include the following steps: The photon absorption process is modeled as a random process following a Poisson distribution, generating photoelectrons; The photoelectron generation process is modeled as a random event defined by quantum yield, which excites secondary electrons. The Gaussian convolution algorithm is used to simulate the random walk and scattering effect of secondary electrons, and the spatial diffusion range of secondary electrons is limited by the electron fuzzy length parameter. Secondary electrons are used to activate the ligand dissociation on the surface of the metal oxide core, generating active sites. This activates the metal oxide core after ligand dissociation, and the active sites participate in the oxygen bridge bond formation reaction in the nearest-neighbor condensation crosslinking mechanism.

[0084] This embodiment constructs a random exposure model to transform the macroscopic exposure energy distribution into the microscopic spatial distribution of chemically active sites, providing accurate state input for subsequent crosslinking reactions based on lattice topological constraints.

[0085] This application provides a possible implementation method. Step A1 above establishes a nearest-neighbor condensation crosslinking mechanism based on lattice topological constraints, which may specifically include the following steps: The photoresist space is divided into the first discrete lattice unit; In a discrete lattice unit, the effective range of the crosslinking reaction is limited to a randomly selected six-neighbor area; Identify the first metal oxide core that is in an active state and carries an active site, and determine the activation state of the nearest neighbor second metal oxide core within the six-neighbor range of the first metal oxide core. When both metals are in an active state, oxygen bridge bonds are created between the first metal oxide core and the second oxygen metal oxide core, and the corresponding active sites are consumed to obtain the second discrete lattice unit.

[0086] In one example, such as Figure 4a As shown, based on the aforementioned nearest-neighbor condensation crosslinking mechanism, a three-dimensional distribution map of the crosslinked network within the metal oxide photoresist was obtained at an exposure dose of 20.0 mJ / cm², where Dose represents the exposure dose, referring to the exposure energy received by the photoresist per unit area. Figure 4a Different colors correspond to the size of the metal oxide core aggregates, with blue to red representing a gradual increase in cluster size. At low exposure doses, only local areas of metal oxide cores form aggregates through oxygen bridging bonds, the crosslinking network is scattered, and the overall size of the core aggregates is relatively small, which intuitively reflects the incompleteness of the crosslinking reaction at low doses.

[0087] In another example, such as Figure 4b As shown, after increasing the exposure dose to 40.0 mJ / cm², the corresponding three-dimensional distribution map of the crosslinking network was obtained. Compared with the dose condition of 20.0 mJ / cm², the coverage of the crosslinking network was significantly expanded, and the proportion of large-sized core aggregates was significantly increased. This shows that increasing the exposure dose can effectively promote the occurrence of the nearest-neighbor condensation crosslinking reaction and expand the formation range of the metal oxide crosslinking network.

[0088] In another example, such as Figure 4c As shown, when the exposure dose is increased to 90.0 mJ / cm², a three-dimensional distribution map of the corresponding cross-linking network is obtained. A large-area continuous metal oxide cross-linking network is formed in the photoresist exposure area. Most of the metal oxide cores are connected by oxygen bridges to form large-sized aggregates. Only a small number of scattered uncross-linked cores exist in the network edge area, which reflects the sufficiency of the nearest-neighbor condensation cross-linking reaction under high exposure dose.

[0089] In another example, such as Figure 4d As shown, the three-dimensional distribution map of the largest continuous cross-linked network in the photoresist was extracted under an exposure dose of 90.0 mJ / cm². The Largest Network (Dose 90) represents the largest continuous cross-linked network formed in the photoresist under an exposure dose of 90 mJ / cm². Size: 9516 cores indicates that the largest continuous cross-linked network contains 9516 metal oxide cores, forming a complete and continuous resist network structure, which provides a structural basis for the complete preservation of the photolithographic pattern during the subsequent development process.

[0090] In one example, such as Figure 5 As shown, after processing using the nearest-neighbor condensation crosslinking mechanism of the random exposure model, the spatial distribution map of oxygen bridges in the metal oxide photoresist is obtained. Here, Oxo-bonds Distribution (XY Plane-Sum) represents the total distribution of oxygen bridges in the plane formed by the X and Y axes, and Oxo-bonds Distribution (XZ Cross-section) represents the distribution of oxygen bridges in the cross-section formed by the X and Z axes. Figure 5 The Z (Depth) on the left represents the depth direction, with the unit being nanometers. Figure 5 The darker the color, the more oxygen bridges there are at the corresponding position. This visually presents the planar and depth distribution of the cross-linking network between the metal oxide cores in the exposure area, quantitatively reflecting the spatial occurrence pattern of the nearest-neighbor condensation cross-linking reaction, and providing accurate cross-linking structure input for the subsequent simulation of the discrete random development model.

[0091] In this embodiment, the updated second discrete lattice unit is obtained through the cross-linking rules of topological constraints. This unit records the cross-linking network structure after the formation of oxygen bridge bonds, providing an accurate three-dimensional network state of the photoresist for the input of the subsequent discrete random development model.

[0092] This application provides a possible implementation method. Step A2 above, which constructs a discrete random imaging model based on the Gillespie algorithm, may specifically include the following steps: The input layer for constructing the discrete random development model is the second discrete lattice unit. The input parameters include the ligand binding site state on the surface of each metal oxide core in the second discrete lattice unit, and the oxygen bridge bond connection state generated between each two metal oxide cores through the nearest neighbor condensation crosslinking mechanism. Construct a discrete random development model reaction system, define the competing reactions included in the simulated development process, and the competing reactions include at least one of ligand substitution reaction, solvent reverse substitution reaction, and oxygen bridge bond breaking reaction; Based on the Gillespie algorithm, a dynamic evolution layer of the discrete random development model is constructed. Rules for calculating the probability of competitive reactions, the calculation of reaction time step, the reaction execution rules, and the dynamic update rules of the metal oxide core state are set to execute competitive reactions in the development process and update the ligand binding site state and oxygen bridge bond connection state of the metal oxide core in the second discrete lattice unit to obtain the third discrete lattice unit. An output decision layer of a discrete random development model is constructed, a dissolution and stripping decision rule for metal oxide cores is set, the dissolution and stripping result of metal oxide cores in the third discrete lattice unit is determined based on the dissolution and stripping decision rule, the distribution state of metal oxide cores in the third discrete lattice unit is updated, and the three-dimensional morphology of the developed photoresist is generated.

[0093] In one example, such as Figure 6a As shown, the input to the discrete random development model is a three-dimensional map of the initial bond distribution of the metal oxide core, output by the random exposure model. Figure 6a Different colors represent the number of oxygen bridges connecting a single metal oxide core. The color changes from purple to yellow, corresponding to a gradient of the number of oxygen bridges from 1.0 to 4.0. This fully presents the initial cross-linked network state within the photoresist after exposure and cross-linking, providing initial structural input for simulating the competitive reaction during the development process.

[0094] In another example, such as Figure 6b As shown, based on the Gillespie algorithm, a discrete random development model is run, and a three-dimensional distribution map of the development wavefront is obtained under the first development duration. Figure 6bThe red area represents the cross-linked network of metal oxides that has not been dissolved and stripped, while the blank area represents the photoresist area that has been dissolved and stripped by the developer. This visually illustrates the wavefront progression of the developer eroding the photoresist and dissolving and stripping the metal oxide core after the ligand substitution reaction and oxygen bridge bond breaking reaction occur during the development process.

[0095] In another example, such as Figure 6c As shown, extending the development time to the second development time yielded a corresponding three-dimensional distribution map of the development wavefront. Compared to the first development time, the development wavefront further penetrated into the photoresist, and the undissolved cross-linked network region further shrank. This clearly demonstrates the influence of development time on the dissolution range of the metal oxide core and also verifies the regulatory role of parameters such as the number of critical development sites and development time on the final morphology of the photoresist.

[0096] In another example, such as Figure 7a As shown, after running the discrete random development model to complete the development simulation, a three-dimensional morphology image of the developed photoresist is obtained. Among them, 3D Development Profile (Scatter Plot) represents the three-dimensional development morphology (scatter plot); Z (Height) represents the height direction, and the unit is nanometers; the view color scale is based on Height (Z). Figure 7a The image clearly shows the three-dimensional outline of multiple sets of parallel lithographic lines formed after development.

[0097] In another example, such as Figure 7b As shown, Figure 7a The XY plane view and XZ cross-sectional view corresponding to the lithographic lines are shown in the figure. The XY Plane Projection (Top View) represents the projection of the plane formed by the X-axis and Y-axis (i.e., the top view); the XZ Plane Projection (Side View) represents the projection of the plane formed by the X-axis and Z-axis (i.e., the side view); and the Y Position represents the position of the Y-axis. Figure 7b The fluctuation state of the line edge directly reflects the line edge roughness and line width roughness characteristics of the photoresist, and fully presents the photoresist morphology result of the final output of the forward simulation.

[0098] In one example, such as Figure 8a As shown, after completing the forward simulation based on the optimized first candidate parameter combination, a three-dimensional morphology image of the developed photoresist is obtained. Figure 8a The edges of the lithographic lines are smooth, and the overall morphology of the lines is excellent, with no obvious irregular depressions or protrusions, which intuitively demonstrates the good control effect of the photoresist morphology after optimization by this method.

[0099] In another example, such as Figure 8bAs shown, after completing the forward simulation based on the unoptimized second candidate parameter combination, the three-dimensional morphology of the developed photoresist is obtained. Figure 8b The edges of the photolithography lines exhibit obvious irregular fluctuations, depressions, and protrusions, resulting in poor uniformity of the line morphology. This directly demonstrates the direct impact of parameter combinations on the three-dimensional morphology and edge quality of the photoresist.

[0100] This embodiment updates the ligand binding site state and oxygen bridge bond connection state of the metal oxide core in the second discrete lattice unit by executing competitive reactions during the development process, thereby obtaining the third discrete lattice unit. This achieves accurate simulation of the development kinetics process, generates the three-dimensional morphology of the photoresist after development, and accurately simulates the dissolution and stripping behavior of the metal oxide core during the development process at the microscopic level, providing a reliable simulation basis for subsequent morphology feature extraction and parameter optimization.

[0101] The above introduces Figure 1 The embodiments shown have various implementation methods for each stage. It should be noted that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application. In practical applications, all the above possible implementation methods can be arbitrarily combined in a combined manner to form possible embodiments of this application, which will not be described in detail here.

[0102] Based on the methods for controlling EUV photoresist morphology and roughness based on automated parameter optimization provided in the above embodiments, and based on the same inventive concept, this application also provides an apparatus for controlling EUV photoresist morphology and roughness based on automated parameter optimization.

[0103] Figure 9 This is a structural diagram of the EUV photoresist morphology and roughness control device based on automated parameter optimization provided in an embodiment of this application. Figure 9 As shown, the EUV photoresist morphology and roughness control device based on automated parameter optimization may specifically include a parameter initialization unit 210, a forward simulation unit 220, a fitness calculation unit 230, an optimization update unit 240, and a parameter output unit 250.

[0104] The parameter initialization unit 210 is used to initialize the numerical range of the optimization variables and generate an initial parameter population containing at least one candidate parameter combination. Each candidate parameter combination includes the values ​​of quantum yield, electronic blur length and number of development critical sites. The forward simulation unit 220 is used to input each candidate parameter combination in the initial parameter population into the random exposure model and the discrete random development model for forward simulation, and generate the photoresist three-dimensional morphology corresponding to each candidate parameter combination. The fitness calculation unit 230 is used to extract the morphological feature parameters of each photoresist three-dimensional morphology, construct a fitness function based on the morphological feature parameters, calculate the fitness value of each candidate parameter combination, and determine whether the fitness value meets the preset convergence condition. The optimization and update unit 240 is used to update the numerical range of the optimization variables using an optimization algorithm when the fitness value does not meet the preset convergence condition, generate a new generation of parameter population, and repeatedly execute the forward simulation, morphological feature parameter extraction and fitness value calculation steps until the preset convergence condition is met. The parameter output unit 250 is used to output the corresponding candidate parameter combination as the target parameter combination when the fitness value meets the preset convergence condition.

[0105] This application provides a possible implementation, wherein the device further includes: a first model building unit and a second model building unit (not shown in the accompanying drawings), specifically: The first model building unit is used to build a random exposure model, which simulates the random walk and scattering effect of secondary electrons and establishes a nearest-neighbor condensation crosslinking mechanism based on lattice topological constraints. The second model building unit is used to build a discrete random development model based on the Gillespie algorithm, which simulates the competitive reaction between developer molecules and the core surface sites of metal oxides.

[0106] This application embodiment provides a possible implementation, wherein the first model building unit is further configured to: The photon absorption process is modeled as a random process following a Poisson distribution, generating photoelectrons; The photoelectron generation process is modeled as a random event defined by quantum yield, which excites secondary electrons. The Gaussian convolution algorithm is used to simulate the random walk and scattering effect of secondary electrons, and the spatial diffusion range of secondary electrons is limited by the electron fuzzy length parameter. Secondary electrons are used to activate the ligand dissociation on the surface of the metal oxide core, generating active sites. This activates the metal oxide core after ligand dissociation, and the active sites participate in the oxygen bridge bond formation reaction in the nearest-neighbor condensation crosslinking mechanism.

[0107] This application embodiment provides a possible implementation, wherein the first model building unit is further configured to: The photoresist space is divided into the first discrete lattice unit; In a discrete lattice unit, the effective range of the crosslinking reaction is limited to a randomly selected six-neighbor area; Identify the first metal oxide core that is in an active state and carries an active site, and determine the activation state of the nearest neighbor second metal oxide core within the six-neighbor range of the first metal oxide core. When both metals are in an active state, oxygen bridge bonds are created between the first metal oxide core and the second oxygen metal oxide core, and the corresponding active sites are consumed to obtain the second discrete lattice unit.

[0108] This application embodiment provides a possible implementation, wherein the second model building unit is further configured to: The input layer for constructing the discrete random development model is the second discrete lattice unit. The input parameters include the ligand binding site state on the surface of each metal oxide core in the second discrete lattice unit, and the oxygen bridge bond connection state generated between each two metal oxide cores through the nearest neighbor condensation crosslinking mechanism. Construct a discrete random development model reaction system, define the competing reactions included in the simulated development process, and the competing reactions include at least one of ligand substitution reaction, solvent reverse substitution reaction, and oxygen bridge bond breaking reaction; Based on the Gillespie algorithm, a dynamic evolution layer of the discrete random development model is constructed. Rules for calculating the probability of competitive reactions, the calculation of reaction time step, the reaction execution rules, and the dynamic update rules of the metal oxide core state are set to execute competitive reactions in the development process and update the ligand binding site state and oxygen bridge bond connection state of the metal oxide core in the second discrete lattice unit to obtain the third discrete lattice unit. An output decision layer of a discrete random development model is constructed, a dissolution and stripping decision rule for metal oxide cores is set, the dissolution and stripping result of metal oxide cores in the third discrete lattice unit is determined based on the dissolution and stripping decision rule, the distribution state of metal oxide cores in the third discrete lattice unit is updated, and the three-dimensional morphology of the developed photoresist is generated.

[0109] Based on the same inventive concept, embodiments of this application also provide an electronic device, including a processor and a memory, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the EUV photoresist morphology and roughness control method based on automated parameter optimization of any of the above embodiments.

[0110] In an exemplary embodiment, an electronic device is provided, such as Figure 10 As shown, Figure 10 The illustrated electronic device 300 includes a processor 301 and a memory 303. The processor 301 and the memory 303 are connected, for example, via a bus 302. Optionally, the electronic device 300 may also include a transceiver 304. It should be noted that in practical applications, the transceiver 304 is not limited to one type, and the structure of this electronic device 300 does not constitute a limitation on the embodiments of this application.

[0111] Processor 301 may be a CPU (Central Processing Unit), GPU (Graphics Processing Unit), DSP (Digital Signal Processor), ASIC (Application Specific Integrated Circuit), FPGA, or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 301 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0112] Bus 302 may include a pathway for transmitting information between the aforementioned components. Bus 302 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 302 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 10 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0113] The memory 303 may be a ROM (Read Only Memory) or other type of static storage device capable of storing static information and instructions, RAM (Random Access Memory) or other type of dynamic storage device capable of storing information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto.

[0114] The memory 303 stores computer program code that executes the scheme of this application, and its execution is controlled by the processor 301. The processor 301 executes the computer program code stored in the memory 303 to implement the content shown in the foregoing method embodiments.

[0115] Among them, electronic devices include, but are not limited to: mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), and in-vehicle terminals (such as in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. Figure 10 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0116] Based on the same inventive concept, this application also provides a storage medium storing a computer program, wherein the computer program is configured to execute the EUV photoresist morphology and roughness control method based on automated parameter optimization of any of the above embodiments when running.

[0117] Based on the same inventive concept, this application also provides a computer program product, including a computer program configured to execute the EUV photoresist morphology and roughness control method based on automated parameter optimization of any of the above embodiments.

[0118] Those skilled in the art will clearly understand that the specific working process of the systems, devices, and modules described above can be referred to the corresponding process in the foregoing method embodiments. For the sake of brevity, it will not be repeated here.

[0119] Those skilled in the art will understand that the technical solution of this application, or all or part of it, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several program instructions to cause an electronic device (e.g., a personal computer, server, or network device) to execute all or part of the steps of the methods described in the embodiments of this application when running the program instructions. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, portable hard drive, read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0120] Alternatively, all or part of the steps of the foregoing method embodiments can be implemented by hardware (such as electronic devices like personal computers, servers, or network devices) associated with program instructions. The program instructions can be stored in a computer-readable storage medium. When the program instructions are executed by the processor of the electronic device, the electronic device executes all or part of the steps of the methods described in the embodiments of this application.

[0121] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that within the spirit and principles of this application, modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein; and these modifications or substitutions do not cause the corresponding technical solutions to leave the protection scope of this application.

Claims

1. A method for controlling the morphology and roughness of EUV photoresist based on automated parameter optimization, characterized in that, The method includes: Initialize the numerical range of the optimization variables and generate an initial parameter population containing at least one candidate parameter combination, each candidate parameter combination including the values ​​of quantum yield, electronic blur length and number of development critical sites; Each candidate parameter combination in the initial parameter population is input into the random exposure model and the discrete random development model for forward simulation to generate the three-dimensional morphology of the photoresist corresponding to each candidate parameter combination. Extract the morphological feature parameters of each photoresist three-dimensional morphology, construct a fitness function based on the morphological feature parameters, calculate the fitness value of each candidate parameter combination, and determine whether the fitness value meets the preset convergence condition; When the fitness value does not meet the preset convergence condition, the numerical range of the optimization variable is updated using the optimization algorithm, a new generation of parameter population is generated, and the forward simulation, morphological feature parameter extraction and fitness value calculation steps are repeated until the preset convergence condition is met. When the fitness value meets the preset convergence condition, the corresponding candidate parameter combination is output as the target parameter combination.

2. The method according to claim 1, characterized in that, Before inputting each candidate parameter combination from the initial parameter population into the random exposure model and the discrete random development model for forward simulation to generate the photoresist 3D morphology corresponding to each candidate parameter combination, the method further includes: A random exposure model was constructed to simulate the random walk and scattering effects of secondary electrons, and a nearest-neighbor condensation crosslinking mechanism was established based on lattice topological constraints. A discrete random development model is constructed based on the Gillespie algorithm to simulate the competitive reaction between developer molecules and the core surface sites of metal oxides.

3. The method according to claim 2, characterized in that, The construction of the random exposure model, which simulates the random walk and scattering effects of secondary electrons, includes: The photon absorption process is modeled as a random process following a Poisson distribution, generating photoelectrons; The photoelectron generation process is modeled as a random event defined by quantum yield, which excites secondary electrons. The Gaussian convolution algorithm is used to simulate the random walk and scattering effect of secondary electrons, and the spatial diffusion range of secondary electrons is limited by the electron fuzzy length parameter. Secondary electrons are used to activate the ligand dissociation on the surface of the metal oxide core, generating active sites. This activates the metal oxide core after ligand dissociation, and the active sites participate in the oxygen bridge bond formation reaction in the nearest-neighbor condensation crosslinking mechanism.

4. The method according to claim 2, characterized in that, The nearest-neighbor condensation crosslinking mechanism based on lattice topological constraints includes: The photoresist space is divided into the first discrete lattice unit; In a discrete lattice unit, the effective range of the crosslinking reaction is limited to a randomly selected six-neighbor area; Identify the first metal oxide core that is in an active state and carries an active site, and determine the activation state of the nearest neighbor second metal oxide core within the six-neighbor range of the first metal oxide core. When both metals are in an active state, oxygen bridge bonds are created between the first metal oxide core and the second oxygen metal oxide core, and the corresponding active sites are consumed to obtain the second discrete lattice unit.

5. The method according to claim 4, characterized in that, The discrete random imaging model constructed based on the Gillespie algorithm includes: The input layer for constructing the discrete random imaging model is the second discrete lattice unit. The input parameters include the ligand binding site state on the surface of each metal oxide core in the second discrete lattice unit, and the oxygen bridge bond connection state generated between each two metal oxide cores through the nearest neighbor condensation crosslinking mechanism. Construct a discrete random development model reaction system, define the competing reactions included in the simulated development process, and the competing reactions include at least one of ligand substitution reaction, solvent reverse substitution reaction, and oxygen bridge bond breaking reaction; Based on the Gillespie algorithm, a dynamic evolution layer of the discrete random development model is constructed. Rules for calculating the probability of competitive reactions, the calculation of reaction time step, the reaction execution rules, and the dynamic update rules of the metal oxide core state are set to execute competitive reactions in the development process and update the ligand binding site state and oxygen bridge bond connection state of the metal oxide core in the second discrete lattice unit to obtain the third discrete lattice unit. An output decision layer of a discrete random development model is constructed, a dissolution and stripping decision rule for metal oxide cores is set, the dissolution and stripping result of metal oxide cores in the third discrete lattice unit is determined based on the dissolution and stripping decision rule, the distribution state of metal oxide cores in the third discrete lattice unit is updated, and the three-dimensional morphology of the developed photoresist is generated.

6. The method according to any one of claims 1 to 5, characterized in that, The morphological feature parameters include line edge roughness, line width roughness, and simulation critical dimension. The fitness function satisfies: F=w1×LER+w2×LWR+w3×|CD_target-CD_sim| Where w1 is the first weighting coefficient, LER is the line edge roughness, w2 is the second weighting coefficient, LWR is the line width roughness, w3 is the third weighting coefficient, CD_target is the preset target critical size, and CD_sim is the simulation critical size.

7. A device for controlling the morphology and roughness of EUV photoresist based on automated parameter optimization, characterized in that, The device includes: The parameter initialization unit is used to initialize the numerical range of the optimization variables and generate an initial parameter population containing at least one candidate parameter combination. Each candidate parameter combination includes the values ​​of quantum yield, electronic blur length, and number of development critical sites. The forward simulation unit is used to input each candidate parameter combination in the initial parameter population into the random exposure model and the discrete random development model for forward simulation, and generate the three-dimensional morphology of the photoresist corresponding to each candidate parameter combination. The fitness calculation unit is used to extract the morphological feature parameters of each photoresist three-dimensional morphology, construct a fitness function based on the morphological feature parameters, calculate the fitness value of each candidate parameter combination, and determine whether the fitness value meets the preset convergence condition. The optimization and update unit is used to update the numerical range of the optimization variables using an optimization algorithm when the fitness value does not meet the preset convergence condition, generate a new generation of parameter population, and repeatedly execute the forward simulation, morphological feature parameter extraction and fitness value calculation steps until the preset convergence condition is met. The parameter output unit is used to output the corresponding candidate parameter combination as the target parameter combination when the fitness value meets the preset convergence condition.

8. An electronic device, characterized in that, The device includes a processor and a memory, wherein the memory stores a computer program, and the processor is configured to run the computer program to perform the EUV photoresist morphology and roughness control method based on automated parameter optimization as described in any one of claims 1 to 6.

9. A storage medium, characterized in that, The storage medium stores a computer program, wherein the computer program is configured to execute, at runtime, the EUV photoresist morphology and roughness control method based on automated parameter optimization as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, The computer program is configured to execute, at runtime, the EUV photoresist morphology and roughness control method based on automated parameter optimization as described in any one of claims 1 to 6.

Images