A photolithography simulation method and apparatus
By using a neural network to fit the lithography simulation method, the intensity distribution of the light source and the mask transfer matrix are simulated, which solves the problem of inaccurate lithography simulation under high numerical aperture and realizes fast and accurate vector lithography simulation.
Patent Information
- Application Number
- CN202510630638.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-15
- Publication Date
- 2026-11-17
AI Technical Summary
Existing lithography simulation methods cannot accurately consider polarization modes at high numerical apertures, resulting in inaccurate imaging and high computational cost, making it impossible to quickly perform large-size mask and large-scale lithography imaging simulation.
A vector lithography simulation method based on neural networks is adopted. By simulating the continuous intensity distribution of the light source and fitting the mask transmission matrix, polarization mode and light intensity distribution are considered, and complex domain convolutional neural networks are used to accelerate the simulation calculation.
It achieves accurate photolithography simulation at high numerical apertures, with a 54-fold increase in computation speed, enabling rapid simulation of large-size masks and large-scale photolithography imaging.
Smart Images

Figure CN123113246A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of chip technology, and in particular to a photolithography simulation method and apparatus. Background Technology
[0002] Fast and accurate forward lithography simulation methods are fundamental to guiding physical verification tasks such as hotspot detection and resolution enhancement techniques such as optical proximity correction. The most accurate lithography simulation methods are rigorous numerical solutions based on Maxwell's equations. While these methods guarantee high accuracy, their computational complexity and time consumption make them unsuitable for large-size masks and large-scale lithography imaging simulations. Inspired by the tremendous success of deep learning in fields such as computer vision, some progress has been made in accelerating simulations by fitting optical and chemical processes using deep neural networks. However, current methods all rely on scalar approaches to fit the process from mask to spatial image using machine learning models. Scalar models represent the various parts of a projection optics system using scalar functions without considering vector factors such as incident angle and polarization mode. They offer advantages such as model simplicity and computational convenience, and can achieve relatively accurate results when the aperture (NA) of the projection lens is less than 0.6. However, scalar models only handle light intensity and ignore vector characteristics (i.e., polarization mode), resulting in scalar theory being somewhat ineffective for imaging at advanced nodes with wavelengths of 193 nm and high numerical apertures (above 0.6), thus failing to achieve accurate physical simulations.
[0003] Therefore, there is an urgent need for a solution to achieve accurate and fast vector lithography simulation. Summary of the Invention
[0004] This application provides a lithography simulation method and apparatus for achieving accurate and rapid vector lithography simulation.
[0005] In a first aspect, this application provides a photolithography simulation method, which includes: simulating a light source with a continuous intensity distribution during the photolithography process; for each light source point of the light source, accelerating the simulation calculation when the light source point passes through the mask by fitting the transmission matrix of the mask through a neural network, and obtaining the mask near field of the light source point; and obtaining a simulated spatial image of the light source in the photolithography simulation image plane based on the mask near field of each light source point of the light source.
[0006] In one possible design, the light source simulating the continuous intensity distribution during the photolithography process includes: simulating the continuous intensity distribution in the inner diameter region of the annular light source based on a first smoothing kernel, and simulating the continuous intensity distribution in the outer diameter region of the annular light source based on a second smoothing kernel, to obtain a radially symmetrical annular light source with a continuous intensity distribution; each smoothing kernel includes a transition function.
[0007] In one possible design, a radially symmetrical annular light source with continuously distributed intensity is obtained by simulating the continuous intensity distribution in the outer diameter region of the annular light source using a first smooth kernel and simulating the continuous intensity distribution in the inner diameter region of the annular light source using a second smooth kernel. This includes: the radially symmetrical annular light source with continuously distributed intensity is represented by the following formula:
[0008]
[0009] in, The first smooth kernel characterizes the outer diameter coherence coefficient of the annular light source; The second smooth kernel characterizes the inner diameter coherence coefficient of the annular light source; smooth kernel
[0010] In one possible design, the light source simulating the continuous intensity distribution during the photolithography process includes: simulating the continuous intensity distribution in the outer diameter region of the annular light source based on a first smoothing kernel, simulating the continuous intensity distribution in the inner diameter region of the annular light source based on a second smoothing kernel, and obtaining an annular light source with displacement and continuous intensity distribution based on the power offset and geometric offset of the light source; any smoothing kernel includes a transition function.
[0011] In one possible design, a ring-shaped light source with displacement and continuous intensity distribution is obtained based on a first smooth kernel simulating the continuous intensity distribution in the outer diameter region of the ring-shaped light source, a second smooth kernel simulating the continuous intensity distribution in the inner diameter region of the ring-shaped light source, and the power offset and geometric offset of the light source. This includes:
[0012] The annular light source with displacement and continuous intensity distribution is represented by the following formula:
[0013]
[0014] in, The first smooth kernel characterizes the outer diameter coherence coefficient of the annular light source; The second smooth kernel characterizes the inner diameter coherence coefficient of the annular light source; α(θ) * ) represents the power displacement correction term; γ(θ) * ) and β(θ * ) represents the correction term for geometric displacement; smooth kernel
[0015] In one possible design, the simulation calculation of the light source point passing through the mask is accelerated by fitting the mask's transfer matrix through a neural network to obtain the mask's near field of the light source point. This includes: for each element in the mask's transfer matrix, representing each element as the sum of multiple terms in a Fourier series expansion; fitting the fundamental transfer matrix of each region on the mask corresponding to each element through a neural network; and for the fundamental transfer matrix of each region on the mask, obtaining the mask's near field corresponding to the fundamental transfer matrix of each region on the mask based on the fundamental transfer matrix and the polarization direction.
[0016] In one possible design, based on the mask near-field of each light source point of the light source, a simulated spatial image of the light source in the lithographic simulation image plane is obtained, including: determining a first component and a second component of the simulated spatial image of the light source in the lithographic simulation image plane based on the sum of multiple terms of the Fourier series expansion corresponding to each element and Abbe imaging theory; the first component depends on the incident angle; the second component does not depend on the incident angle; the second component is obtained based on the mask near-field corresponding to the fundamental transfer matrix of each region on the mask.
[0017] In one possible design, the sum of multiple terms in the Fourier series expansion corresponding to each element is u(s,k). s ) is represented as:
[0018]
[0019] The first component C of the simulated spatial image of the light source on the lithographic simulation image plane l,mn :
[0020]
[0021] The second component I of the simulated spatial image of the light source on the photolithographic simulation plane l,mn (s′):
[0022]
[0023] in,
[0024] In one possible design, the underlying transfer matrix of each region on the mask corresponding to each element is fitted by a neural network, including: flattening the underlying transfer matrix to obtain a one-dimensional vector and then inputting it into a convolutional network through position encoding; based on the optical features of the light source point at each position on the mask, inputting them into the convolutional network through position encoding; and using a 1×1 convolutional layer repeated N times to extract optical features at different depths to obtain the fitted underlying transfer matrix of each element.
[0025] In one possible design, based on the mask near field of each light source point of the light source, a simulated spatial image of the light source in photolithography simulation is obtained, including: for the k-th layer of the photoresist, based on the mask near field of each light source point of the light source, and the refractive index and height of the k-th layer and the layers preceding the k-th layer of the photoresist, a simulated spatial image of the light source in the k-th layer of the photoresist is obtained.
[0026] This application provides a lithography simulation method. The vectorized lithography imaging in this method considers not only the polarization mode of the light source under off-axis illumination but also the reflection and refraction of light intensity distribution in the photoresist, and generates spatial images at different heights using the transfer matrix method. Compared to traditional scalarized lithography imaging, the calculation of vectorized lithography imaging is more complex. Although the computational complexity is higher, vectorized lithography imaging can provide more accurate simulation results. Furthermore, this application designs a complex-domain convolutional neural network to predict the mask transfer matrix, achieving fast and accurate vectorized lithography simulation through simplified operations of Fourier series expansion and neural network fitting.
[0027] Secondly, embodiments of this application provide a photolithography simulation apparatus, the apparatus comprising:
[0028] The processing module is used to simulate a light source with a continuous intensity distribution during the photolithography process. For each light source point, the simulation calculation of the light source point passing through the mask is accelerated by fitting the transmission matrix of the mask through a neural network to obtain the mask near field of the light source point. Based on the mask near field of each light source point, the simulated spatial image of the light source in the photolithography simulation image plane is obtained.
[0029] In one possible design, the processing module is further configured to simulate the continuous intensity distribution in the inner diameter region of the annular light source based on a first smoothing kernel, and to simulate the continuous intensity distribution in the outer diameter region of the annular light source based on a second smoothing kernel, thereby obtaining a radially symmetrical annular light source with a continuous intensity distribution; each smoothing kernel includes a transition function.
[0030] In one possible design, the processing module is further configured to represent the radially symmetrical and continuously intensity-distributed annular light source by the following formula:
[0031]
[0032] in, The first smooth kernel characterizes the outer diameter coherence coefficient of the annular light source; The second smooth kernel characterizes the inner diameter coherence coefficient of the annular light source; smooth kernel
[0033] In one possible design, the processing module is further configured to simulate the continuous intensity distribution in the outer diameter region of the annular light source based on a first smoothing kernel, simulate the continuous intensity distribution in the inner diameter region of the annular light source based on a second smoothing kernel, and obtain an annular light source with displacement and continuous intensity distribution based on the power offset and geometric offset of the light source; any smoothing kernel includes a transition function.
[0034] In one possible design, the processing module is further configured to represent the annular light source with displacement and continuous intensity distribution using the following formula:
[0035]
[0036] in, The first smooth kernel characterizes the outer diameter coherence coefficient of the annular light source; The second smooth kernel characterizes the inner diameter coherence coefficient of the annular light source; α(θ) * ) represents the power displacement correction term; γ(θ) * ) and β(θ * ) represents the correction term for geometric displacement; smooth kernel
[0037] In one possible design, the processing module is further configured to, for each element in the transmission matrix of the mask, represent each element as the sum of multiple terms in a Fourier series expansion; fit the fundamental transmission matrix of each region on the mask corresponding to each element using a neural network; and, for the fundamental transmission matrix of each region on the mask, obtain the mask near field corresponding to the fundamental transmission matrix of each region on the mask based on the fundamental transmission matrix and the polarization direction.
[0038] In one possible design, the processing module is further configured to determine a first and a second component of the simulated spatial image of the light source on the lithographic simulation image plane based on the sum of multiple terms of the Fourier series expansion corresponding to each element and Abbe imaging theory; the first component depends on the incident angle; the second component does not depend on the incident angle; the second component is obtained based on the mask near field corresponding to the fundamental transfer matrix of each region on the mask.
[0039] In one possible design, the sum of multiple terms in the Fourier series expansion corresponding to each element is u(s,k). s ) is represented as:
[0040]
[0041] The first component C of the simulated spatial image of the light source on the lithographic simulation image plane l,mn :
[0042]
[0043] The second component I of the simulated spatial image of the light source on the photolithographic simulation plane l,mn (s′):
[0044]
[0045] in,
[0046] In one possible design, the processing module is further configured to flatten the basic transfer matrix to obtain a one-dimensional vector and then input it into a convolutional network through position encoding; based on the optical features of the light source point at each position of the mask, input them into the convolutional network through position encoding; and use 1×1 convolutional layers repeated N times to extract optical features at different depths to obtain the basic transfer matrix of each element after fitting.
[0047] In one possible design, the processing module is further configured to obtain a simulated spatial image of the light source in the k-th layer of the photoresist, based on the mask near-field of each light source point of the light source, and the refractive index and height of the k-th and preceding layers of the photoresist.
[0048] Thirdly, embodiments of this application also provide a computing device, including:
[0049] Memory, used to store program instructions;
[0050] A processor is configured to invoke program instructions stored in the memory and execute the method described in any possible design of the first aspect, according to the obtained program instructions.
[0051] Fourthly, embodiments of this application also provide a computer-readable storage medium storing computer-readable instructions that, when read and executed by a computer, cause the method described in any possible design of the first aspect to be implemented.
[0052] Fifthly, embodiments of this application also provide a computer program product, which includes computer program code that, when run on a computer, causes the computer to perform the method described in any possible design of the first aspect. Attached Figure Description
[0053] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0054] Figure 1A schematic flowchart of a photolithography simulation method provided in an embodiment of this application;
[0055] Figure 2 A schematic diagram of a photolithography simulation process provided for an embodiment of this application;
[0056] Figure 3 A schematic diagram of a simulated light source provided in an embodiment of this application;
[0057] Figure 4 A flowchart illustrating a method for determining the near-field of a light source point mask, provided in an embodiment of this application;
[0058] Figure 5 A schematic diagram illustrating the imaging of a spatial domain image at different altitudes, provided as an embodiment of this application;
[0059] Figure 6 This is a schematic diagram of the structure of a computing device provided in an embodiment of this application. Detailed Implementation
[0060] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail 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 in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0061] In the embodiments of this application, "multiple" refers to two or more. Terms such as "first" and "second" are used only for descriptive purposes and should not be construed as indicating or implying relative importance or order.
[0062] Fast and accurate forward lithography simulation methods are fundamental for guiding physical verification tasks such as hotspot detection and resolution enhancement techniques such as optical proximity correction. Forward lithography can be divided into optical and chemical processes. The optical process refers to the process where light emitted from a light source of a certain shape propels the pattern on the mask through lenses and other optical devices to form a spatial image. The chemical process refers to the process where chemical materials such as photoresist undergo chemical reactions (including thick baking and exposure) under the light intensity distribution of the spatial image, ultimately projecting the pattern onto the substrate to obtain ADI (Resist). In a lithography system, the intensity distribution of the spatial image, i.e., the optical process, can be calculated using the Abbe or Hopkins methods. Both of these optical imaging methods have scalar-based and vector-based computational paradigms. The chemical process is typically modeled using a hard thresholding model (CTR model).
[0063] The most accurate lithography simulation methods are rigorous numerical solutions based on Maxwell's equations (such as the finite element method). While these methods guarantee high accuracy, their computational complexity and time-consuming nature make them unsuitable for large-scale masks and large-scale lithography imaging simulations. Inspired by the tremendous success of deep learning in fields such as computer vision, some progress has been made in accelerating simulations by fitting optical and chemical processes using deep neural networks. Two implementation paths exist: one is to fit the entire process (optical + chemical processes) using deep neural networks such as generative adversarial networks (cGANs); the other is to consider the Fourier transform in the optical process, using only neural networks (such as multilayer perceptrons) to fit the optical parameters, while employing a hard-threshold model for the chemical process. Because the second approach considers the physical processes of the lithography system more meticulously, rather than solely relying on the powerful fitting ability of neural networks, it theoretically yields better results. These machine learning-based approaches all establish a black box from the mask to the spatial image in a scalar manner, using machine learning models to fit this optical process. The scalar model represents the various parts of the projection optical system using scalar functions without considering vector factors such as incident angle and polarization mode. Scalar imaging theory offers advantages such as simple models and convenient calculations, achieving relatively accurate results when the aperture (NA) of the projection lens is less than 0.6. However, scalar models only handle light intensity, neglecting vector characteristics (i.e., polarization modes). This renders scalar theory somewhat ineffective for imaging at advanced nodes with wavelengths up to 193 nm and high numerical apertures (above 0.6), where vector characteristics have a greater impact on imaging performance. Vector imaging utilizes polarized light propagation and incorporates additional higher-order parameters to represent light sources in various scenarios, enabling accurate physical simulations. Therefore, this application proposes a vector-based fast lithography simulation method, including a vector lithography white-box model considering different light sources and an optically perceptive neural network gray-box lithography model based on this white-box model. Based on the established vector lithography white-box model, a specially designed neural network is used to fit the calculation of the transfer matrix, achieving accurate and fast lithography simulation.
[0064] Figure 1 This is a flowchart illustrating a photolithography simulation method provided in an embodiment of this application, as shown below. Figure 1 As shown, the method includes:
[0065] Step 101: Simulate a light source with a continuous intensity distribution during the photolithography process.
[0066] refer to Figure 2 The lithography simulation flow shown in this embodiment considers a more realistic light source during simulation, using a smoothing kernel to fit the continuous intensity distribution of the actual light source. Each smoothing kernel includes a transition function. For example, a smoothing function with a smoothness degree of t can be provided using arctan, and the smoothing kernel is represented as...
[0067] Smooth kernel K t (x) In light source modeling, in one possible implementation, a light source with a continuous intensity distribution during the photolithography process can be simulated by: simulating the continuous intensity distribution in the inner diameter region of the annular light source based on a first smooth kernel, and simulating the continuous intensity distribution in the outer diameter region of the annular light source based on a second smooth kernel, thereby obtaining a radially symmetrical annular light source with a continuous intensity distribution, wherein any smooth kernel includes a transition function.
[0068] refer to Figure 3 , Figure 3 (a) is a ring-shaped light source without considering the intensity of the light source. Figure 3 (b) is a simulated radially symmetrical annular light source with a continuous intensity distribution. Figure 3 (b) Simulated light source relative to Figure 3 (a) A more realistic light source.
[0069] For example, a radially symmetrical annular light source J with a continuous intensity distribution can be represented by Equation 1. a :
[0070]
[0071] in, The first smooth kernel characterizes the outer diameter coherence coefficient of the annular light source; The second smooth kernel characterizes the inner diameter coherence coefficient of the annular light source; smooth kernel
[0072] In one possible implementation, considering the spatial variation of the light source and various distorted displacements, a light source with a continuous intensity distribution during the photolithography process can be simulated as follows: the continuous intensity distribution in the outer diameter region of the annular light source is simulated based on a first smoothing kernel, the continuous intensity distribution in the inner diameter region of the annular light source is simulated based on a second smoothing kernel, and an annular light source with displacement and continuous intensity distribution is obtained based on the power offset and geometric offset of the light source, wherein any smoothing kernel includes a transition function.
[0073] refer to Figure 3 , Figure 3 (c) and Figure 3 (d) is a simulated ring-shaped light source with displacement and continuous intensity distribution. Figure 3 (c) and Figure 3 (d) Simulated light source relative to Figure 3 (a) and Figure 3 (b) Closer to real light sources.
[0074] For example, a ring-shaped light source with displacement and continuous intensity distribution can be represented by Equation 2.
[0075] in, The first smooth kernel characterizes the outer diameter coherence coefficient of the annular light source; The second smooth kernel characterizes the inner diameter coherence coefficient of the annular light source; (θ,r) to (θ) in polar coordinates * ,r * ), equivalent to the displacement from (x,y) to (x-x0,y-y0) in Cartesian coordinates; α(θ) * ) represents the power displacement correction term; γ(θ) * ) and β(θ * ) represents the correction term for geometric displacement; smooth kernel
[0076] Step 102: For each light source point, the simulation calculation of the light source point passing through the mask is accelerated by fitting the transmission matrix of the mask through a neural network, so as to obtain the mask near field of the light source point.
[0077] In one possible implementation, it can be done as follows: Figure 4 The method shown accelerates the simulation calculation of the light source point passing through the mask by fitting the mask's transfer matrix through a neural network, thus obtaining the mask's near field of the light source point:
[0078] Step 401: For each element in the transmission matrix of the mask, represent each element as the sum of multiple terms in the Fourier series expansion.
[0079] Considering the most important polarization characteristic in vector lithography, e is defined as ⊥ and e || Representing the directions of longitudinal polarization (TM) and transverse polarization (TE), respectively, local coordinates (e ⊥ ,e || The relationship between (x, y, z) and global coordinates is shown in Formula 3:
[0080]
[0081] in, k=[α,β,γ] T E represents the cosine component of the coordinates (x, y, z). x E y and E z These are the scalar components of the vector electric field E in the three directions. ⊥ and E || e || and e⊥ The directional component.
[0082] Considering each light source point s = (x s ,y s The light wave, whose mask near field E mask In the incident direction, i.e., k′ s =[α s ,β s ,γ s ] T Having k in the direction s =2πk′ s The components of the / λ angle are shown in Formula 4:
[0083]
[0084] in, Let E be the coordinate mapping matrix. ⊥ and E || These represent the components in the x and y directions, respectively. The vector-based mask transfer matrix is: It contains two Jones vectors of size 2×1 [u 11 (s,k s ),u 21 (s;k s )] T and[u 12 (s,k s ),u 22 (s,k s )] T .
[0085] For a given mask near field E mask The sum of multiple terms in the Fourier series expansion corresponding to each element, u(s,k) s As shown in Formula 5:
[0086]
[0087] Where, k0=2π[α 0s ,β 0s ,γ 0s ] T / λ is a constant plane wave vector. Based on the transmission matrix, The changes depend only on s, mn th Fourier series expansion term (α) s -α 0s ) m (β s -β 0s ) n The change depends only on the incident angle ks.
[0088] Step 402: Fit the basic transfer matrix of each region on the mask corresponding to each element using a neural network.
[0089] To achieve rapid computation, this application uses an optically-aware deep convolutional network to fit the fundamental transfer matrix. In one possible implementation, the fundamental transfer matrix of each region on the mask corresponding to each element can be fitted via a neural network as follows: the fundamental transfer matrix is flattened into a one-dimensional vector and then input into the convolutional network through position encoding. Based on the optical features of the light source at each position on the mask, these features are also input into the convolutional network through position encoding. A 1×1 convolutional layer repeated N times is used to extract optical features at different depths, resulting in the fitted fundamental transfer matrix for each element.
[0090] Considering that local features are more important than global features during light propagation, this application employs a complex-based convolutional neural network to fit the fundamental transfer matrix. For example, refer to... Figure 2 This application employs N repetitions of 1×1 convolutional layers and activation functions in the complex domain to extract optical features at different depths and outputs four scalar mask transfer matrices, which are then merged to obtain a vector mask transfer matrix. Preferably, the first step is to... After being flattened to obtain a one-dimensional vector, it is input into a convolutional network through positional encoding. To obtain optical features at different depths, three 1×1 convolutional layers are used to multiply the number of channels by 8, 4, and 1 respectively, and this process is repeated N times. Due to the transfer matrix... Since it is a complex matrix, a complex-based convolution operation is used as shown in Equation 6:
[0091]
[0092] Here, a[n1,n2] and bj[n1,n2] represent the values of the real and imaginary fields at position [n1,n2], respectively. First, rotate h(m1,m2) 180 degrees around its origin, then translate its origin, translate n1 on the m1 axis and n2 on the m2 axis, and then add the product of the two functions to obtain the output a′+b′j at [n1,n2].
[0093] In the above technical solution, while ensuring strict accuracy in three-dimensional vector simulation, the fundamental transfer matrix for calculating the near field of the mask is also considered. Since it cannot be obtained through simple linear operations, we applied to use neural networks to accelerate the calculation of the basic transfer matrix, laying a solid foundation for subsequent rapid multi-target mask optimization.
[0094] Step 403: For the fundamental transfer matrix of each region on the mask, based on the fundamental transfer matrix and the polarization direction, obtain the mask near field corresponding to the fundamental transfer matrix of each region on the mask.
[0095] The fundamental transfer matrix of each region on the mask corresponding to each element is obtained by fitting a neural network. Then, based on the basic transmission matrix And the polarization direction, to obtain the mask near field corresponding to the fundamental transfer matrix of each region on the mask. Among them, E ⊥ and E || e || and e ⊥ directional component, e ⊥ and e || These represent the directions of longitudinal polarization (TM) and transverse polarization (TE), respectively.
[0096] Step 103: Based on the mask near field of each light source point, obtain the simulated spatial image of the light source in the photolithographic simulation image plane.
[0097] In one possible implementation, the simulated spatial image of the light source on the lithographic simulation image plane can be obtained by implementing the mask near field based on each light source point of the light source in the following way: based on the sum of multiple terms of the Fourier series expansion corresponding to each element and Abbe imaging theory, the first and second components of the simulated spatial image of the light source on the lithographic simulation image plane are determined; the first component depends on the incident angle; the second component does not depend on the incident angle; the second component is obtained based on the mask near field corresponding to the fundamental transfer matrix of each region on the mask.
[0098] Specifically, using Abbe imaging theory, we can obtain the point source (x) in the p-th radiation region. s ,y s The electric field distribution contributed is shown in Equation 7:
[0099]
[0100] Where r is the propagation distance of the plane wave, and α, β, γ are the cosine components in the (x, y, z) directions, respectively. The far-field diffraction distribution of the 3D thick mask is related to the incident angle of the plane wave illuminating the mask, as well as parameters such as the structure and materials of the 3D mask; H is the transmittance function of the projection lens, representing the diffraction-limiting effect of the projection lens; F is the Fourier transform operation, p = (x, y, z) are the global coordinates on the mask, and H... p Let be the transmittance function of the projection lens. The near-field E of the mask in each region p mask The p-direction component. Therefore, at the point source (x s ,y s Under illumination, the electric field intensity distribution on the image plane is as follows:
[0101]
[0102] Finally, the electric field intensity distribution of the entire light source on the image plane can be obtained as shown in Equation 9:
[0103]
[0104] Spatial domain images can be generated by I total It is obtained through inverse Fourier transform.
[0105] Combining Formulas 5, 7, 8, and 9 with the inverse Fourier transform, we obtain the simulated spatial image of the light source on the lithographic simulation plane. This simulated spatial image comprises two parts: the first part, C... l,mn As shown in Formula 10:
[0106]
[0107] The second component I of the simulated spatial image of the light source on the photolithographic simulation plane l,mn (s′) is shown in Formula 11:
[0108]
[0109] in,
[0110] Based on the first and second sub-items, a simulated spatial image of the light source in the photolithographic simulation image plane is obtained.
[0111] In one possible implementation, a simulated spatial image of the light source in the photolithography simulation is obtained based on the mask near field of each light source point of the light source, including: for the k-th layer of the photoresist, based on the mask near field of each light source point of the light source, and the refractive index and height of the k-th layer and the layers preceding the k-th layer of the photoresist, to obtain a simulated spatial image of the light source in the k-th layer of the photoresist.
[0112] The spatial intensity distribution of the vector can propagate to the wafer stack, reference Figure 5 This application further considers imaging at different heights of the spatial pattern. For TE polarization, the relationship between the field intensity distribution of any layer and the input field intensity is as follows:
[0113] E I =M1M2…M k E i
[0114] in The refractive index k0 and height h of each layer k Decide.
[0115] This application provides a lithography simulation method. The vectorized lithography imaging in this method considers not only the polarization mode of the light source under off-axis illumination but also the reflection and refraction of light intensity distribution in the photoresist, and generates spatial images at different heights using the transfer matrix method. Compared to traditional scalarized lithography imaging, the calculation of vectorized lithography imaging is more complex. Although the computational complexity is higher, vectorized lithography imaging can provide more accurate simulation results. Furthermore, this application designs a complex-domain convolutional neural network to predict the mask transfer matrix, achieving fast and accurate vectorized lithography simulation through simplified operations of Fourier series expansion and neural network fitting.
[0116] The vector-based fast lithography simulation method described in this application has a runtime of 0.138 s and achieves an average EPE of 4.8 nm in industrial-scale testing. In comparison, commercial software using the same equipment has a runtime of 18.9 s, representing a 54-fold speed improvement.
[0117] Based on the same technical concept, this application provides a photolithography simulation device, which includes: a processing module for simulating a light source with a continuous intensity distribution during the photolithography process; for each light source point, accelerating the simulation calculation when the light source point passes through the mask by fitting the transmission matrix of the mask through a neural network, and obtaining the mask near field of the light source point; and obtaining a simulated spatial image of the light source in the photolithography simulation image plane based on the mask near field of each light source point.
[0118] In one possible design, the processing module is further configured to simulate the continuous intensity distribution in the inner diameter region of the annular light source based on a first smoothing kernel, and to simulate the continuous intensity distribution in the outer diameter region of the annular light source based on a second smoothing kernel, thereby obtaining a radially symmetrical annular light source with a continuous intensity distribution; each smoothing kernel includes a transition function.
[0119] In one possible design, the processing module is further configured to represent the radially symmetrical and continuously intensity-distributed annular light source by the following formula:
[0120]
[0121] in, The first smooth kernel characterizes the outer diameter coherence coefficient of the annular light source; The second smooth kernel characterizes the inner diameter coherence coefficient of the annular light source; smooth kernel
[0122] In one possible design, the processing module is further configured to simulate the continuous intensity distribution in the outer diameter region of the annular light source based on a first smoothing kernel, simulate the continuous intensity distribution in the inner diameter region of the annular light source based on a second smoothing kernel, and obtain an annular light source with displacement and continuous intensity distribution based on the power offset and geometric offset of the light source; any smoothing kernel includes a transition function.
[0123] In one possible design, the processing module is further configured to simulate the continuous intensity distribution in the outer diameter region of the annular light source based on a first smooth kernel, simulate the continuous intensity distribution in the inner diameter region of the annular light source based on a second smooth kernel, and obtain an annular light source with displacement and continuous intensity distribution based on the power offset and geometric offset of the light source, including:
[0124] The annular light source with displacement and continuous intensity distribution is represented by the following formula:
[0125]
[0126] in, The first smooth kernel characterizes the outer diameter coherence coefficient of the annular light source; The second smooth kernel characterizes the inner diameter coherence coefficient of the annular light source; α(θ) * ) represents the power displacement correction term; γ(θ) * ) and β(θ * ) represents the correction term for geometric displacement; smooth kernel
[0127] In one possible design, the processing module is further configured to, for each element in the transmission matrix of the mask, represent each element as the sum of multiple terms in a Fourier series expansion; fit the fundamental transmission matrix of each region on the mask corresponding to each element using a neural network; and, for the fundamental transmission matrix of each region on the mask, obtain the mask near field corresponding to the fundamental transmission matrix of each region on the mask based on the fundamental transmission matrix and the polarization direction.
[0128] In one possible design, the processing module is further configured to determine a first and a second component of the simulated spatial image of the light source on the lithographic simulation image plane based on the sum of multiple terms of the Fourier series expansion corresponding to each element and Abbe imaging theory; the first component depends on the incident angle; the second component does not depend on the incident angle; the second component is obtained based on the mask near field corresponding to the fundamental transfer matrix of each region on the mask.
[0129] In one possible design, the sum of multiple terms in the Fourier series expansion corresponding to each element is u(s,k). s ) is represented as:
[0130]
[0131] The first component C of the simulated spatial image of the light source on the lithographic simulation image plane l,mn :
[0132]
[0133] The second component I of the simulated spatial image of the light source on the photolithographic simulation plane l,mn (s′):
[0134]
[0135] in,
[0136] In one possible design, the processing module is further configured to flatten the basic transfer matrix to obtain a one-dimensional vector and then input it into a convolutional network through position encoding; based on the optical features of the light source point at each position of the mask, input them into the convolutional network through position encoding; and use 1×1 convolutional layers repeated N times to extract optical features at different depths to obtain the basic transfer matrix of each element after fitting.
[0137] In one possible design, the processing module is further configured to obtain a simulated spatial image of the light source in the k-th layer of the photoresist, based on the mask near-field of each light source point of the light source, and the refractive index and height of the k-th and preceding layers of the photoresist.
[0138] Based on the same technical concept, embodiments of this application provide a computing device, such as... Figure 6 As shown, it includes at least one processor 601 and a memory 602 connected to at least one processor. In this embodiment, the specific connection medium between the processor 601 and the memory 602 is not limited. Figure 6 Taking the connection between the processor 601 and the memory 602 via a bus as an example, the bus can be divided into address bus, data bus, control bus, etc.
[0139] In this embodiment of the application, the memory 602 stores instructions that can be executed by at least one processor 601. By executing the instructions stored in the memory 602, at least one processor 601 can execute the photolithography simulation method listed above.
[0140] The processor 601 is the control center of the computing device. It can connect to various parts of the computer device through various interfaces and lines, and perform resource settings by running or executing instructions stored in the memory 602 and calling data stored in the memory 602.
[0141] Optionally, processor 601 may include one or more processing units. Processor 601 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into processor 601. In some embodiments, processor 601 and memory 602 may be implemented on the same chip; in some embodiments, they may be implemented separately on independent chips.
[0142] Processor 601 can be a general-purpose processor, such as a central processing unit (CPU), digital signal processor, application-specific integrated circuit (ASIC), field-programmable gate array (FPGA), or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component, capable of implementing or executing the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly manifested as being executed by a hardware processor, or executed by a combination of hardware and software modules within the processor.
[0143] Memory 602, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules. Memory 602 may include at least one type of storage medium, such as flash memory, hard disk, multimedia card, card-type memory, random access memory (RAM), static random access memory (SRAM), programmable read-only memory (PROM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), magnetic storage, magnetic disk, optical disk, etc. Memory 602 can be any other medium capable of carrying or storing desired program code in the form of instructions or data structures that can be accessed by a computer, but is not limited thereto. In the embodiments of this application, memory 602 can also be a circuit or any other device capable of implementing storage functions for storing program instructions and / or data.
[0144] Based on the same technical concept, embodiments of this application also provide a computer-readable storage medium storing a computer-executable program, which is used to cause a computer to perform any of the photolithography simulation methods listed above.
[0145] Based on the same technical concept, embodiments of this application also provide a computer program product, which includes computer program code. When the computer program code is run on a computer, it causes the computer to perform the photolithography simulation method as listed in any of the above methods.
[0146] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0147] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0148] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0149] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1The steps of the function specified in one or more boxes.
[0150] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.
[0151] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.
Claims
1. A photolithography simulation method, characterized in that, The method includes: A light source with a continuous intensity distribution that simulates the photolithography process; For each light source point of the light source, the simulation calculation of the light source point passing through the mask is accelerated by fitting the transmission matrix of the mask through a neural network, so as to obtain the mask near field of the light source point; Based on the mask near field of each light source point of the light source, a simulated spatial image of the light source in the photolithographic simulation image plane is obtained.
2. The method according to claim 1, characterized in that, Light sources simulating continuous intensity distribution during photolithography include: Based on the first smooth kernel simulating the continuous intensity distribution in the inner diameter region of the annular light source, and based on the second smooth kernel simulating the continuous intensity distribution in the outer diameter region of the annular light source, a radially symmetrical annular light source with a continuous intensity distribution is obtained; each smooth kernel includes a transition function.
3. The method according to claim 2, characterized in that, Based on the continuous intensity distribution in the outer diameter region of the annular light source simulated using a first smooth kernel, and the continuous intensity distribution in the inner diameter region of the annular light source simulated using a second smooth kernel, a radially symmetrical annular light source with a continuous intensity distribution is obtained, including: The radially symmetrical and continuously intensity-distributed annular light source is represented by the following formula: in, The first smooth kernel characterizes the outer diameter coherence coefficient of the annular light source; The second smooth kernel characterizes the inner diameter coherence coefficient of the annular light source; smooth kernel 4. The method according to claim 1, characterized in that, Light sources simulating continuous intensity distribution during photolithography include: The continuous intensity distribution in the outer diameter region of the annular light source is simulated based on the first smooth kernel, and the continuous intensity distribution in the inner diameter region of the annular light source is simulated based on the second smooth kernel. Based on the power offset and geometric offset of the light source, an annular light source with displacement and continuous intensity distribution is obtained. Each smooth kernel includes a transition function.
5. The method according to claim 4, characterized in that, Based on the continuous intensity distribution in the outer diameter region of the annular light source simulated by the first smooth kernel, and the continuous intensity distribution in the inner diameter region of the annular light source simulated by the second smooth kernel, and based on the power offset and geometric offset of the light source, an annular light source with displacement and continuous intensity distribution is obtained, including: The annular light source with displacement and continuous intensity distribution is represented by the following formula: in, The first smooth kernel characterizes the outer diameter coherence coefficient of the annular light source; The second smooth kernel characterizes the inner diameter coherence coefficient of the annular light source; α(θ) * ) represents the power displacement correction term; γ(θ) * ) and β(θ * ) represents the correction term for geometric displacement; smooth kernel 6. The method according to any one of claims 1 to 5, characterized in that, The simulation calculation of the light source point passing through the mask is accelerated by fitting the transmission matrix of the mask through a neural network, and the near field of the light source point through the mask is obtained, including: For each element in the transmission matrix of the mask, each element is represented as the sum of multiple terms in a Fourier series expansion; The underlying transfer matrix of each region on the mask corresponding to each element is fitted using a neural network; For the fundamental transfer matrix of each region on the mask, the mask near field corresponding to the fundamental transfer matrix of each region on the mask is obtained based on the fundamental transfer matrix and the polarization direction.
7. The method according to claim 6, characterized in that, Based on the mask near field of each light source point of the light source, a simulated spatial image of the light source in the photolithographic simulation image plane is obtained, including: Based on the sum of multiple terms of the Fourier series expansion corresponding to each element and Abbe imaging theory, the first and second components of the simulated spatial image of the light source on the lithographic simulation image plane are determined; the first component depends on the incident angle; the second component does not depend on the incident angle; the second component is obtained based on the mask near field corresponding to the fundamental transfer matrix of each region on the mask.
8. The method according to claim 7, characterized in that, The number of terms of the Fourier series expansion corresponding to each element and u(s, k s ) is denoted as: The first component C of the simulated spatial image of the light source on the lithographic simulation image plane l,mn : The second component I of the simulated spatial image of the light source on the photolithographic simulation plane l,mn (s′): in, 9. The method according to claim 6, characterized in that, The underlying transfer matrix of each region on the mask corresponding to each element is fitted using a neural network, including: The basic transfer matrix is flattened to obtain a one-dimensional vector, which is then input into the convolutional network through position encoding; based on the optical features of the light source point at each position of the mask, the vector is also input into the convolutional network through position encoding. Optical features at different depths are extracted by repeating 1×1 convolutional layers N times, and the basic transfer matrix of each element after fitting is obtained.
10. The method according to any one of claims 1 to 5, characterized in that, Based on the mask near field of each light source point of the light source, a simulated spatial image of the light source in photolithography simulation is obtained, including: For the k-th layer of the photoresist, based on the mask near field of each light source point of the light source, and the refractive index and height of the k-th layer and the layers preceding the k-th layer of the photoresist, a simulated spatial image of the light source in the k-th layer of the photoresist is obtained.
11. A photolithography simulation device, characterized in that, include: The processing module is used to simulate the continuous intensity distribution of the light source during the photolithography process; The processing module is also used to accelerate the simulation calculation of the light source point passing through the mask by fitting the transmission matrix of the mask through a neural network for each light source point of the light source, so as to obtain the mask near field of the light source point. The processing module is also used to obtain a simulated spatial image of the light source in the photolithographic simulation image plane based on the mask near field of each light source point of the light source.
12. A computing device, characterized in that, include: Memory, used to store program instructions; A processor is configured to invoke program instructions stored in the memory and execute the method as described in any one of claims 1 to 8 according to the obtained program instructions.
13. A computer-readable storage medium, characterized in that, Includes computer-readable instructions that, when read and executed by a computer, cause the method as described in any one of claims 1 to 10 to be implemented.
14. A computer program product, characterized in that, The computer program product includes: computer program code, which, when run on a computer, causes the computer to perform the method described in any one of claims 1 to 10.
