Optical scattering method based on inverse problem solving algorithm and mixed measurement method for batch samples

Abstract

The invention relates to the technical field of optical measurement, in particular to an optical scattering method based on an inverse problem solving algorithm and a mixed measurement method for batch samples thereof, and the method comprises the following steps: carrying out optical measurement on samples to obtain actually measured optical signals; obtaining prior structure information of the sample; constructing an initial model; analyzing the initial model by utilizing strict coupling wave analysis to generate a simulation optical signal; judging whether the difference between the generated simulation optical signal and the actually measured optical signal meets a set requirement or not; if yes, outputting a measurement result; and if not, gradient calculation is carried out on each parameter in the initial model, gradients of all the parameters are obtained by utilizing an automatic differential algorithm, then the initial model is optimized by utilizing the gradients of the parameters to obtain an optimized model, and the step of analyzing by utilizing the strict coupling waves is repeatedly executed until a set requirement is met. According to the method, more detailed contour optimization can be provided, and the calculation efficiency can be improved.

Description

Technical Field

[0001] This invention relates to the field of optical measurement technology, and specifically to an optical scattering method based on an inverse problem solving algorithm and its mixed measurement method for batch samples. Background Technology

[0002] Optical scattering, also known as optical critical dimension measurement, is widely used for characterizing periodic micro / nanostructures and materials due to its speed, non-contact nature, and low cost. Optical scattering obtains the morphology of micro / nanostructures by measuring the scattering signals of light and the micro / nanostructure and solving the optical inverse scattering problem. Specifically, it includes two steps: optical measurement and inverse problem solving. Optical measurement involves using appropriate optical signal measurement devices to measure the corresponding optical signals; inverse problem solving involves a combination of numerical simulation and algorithms.

[0003] Due to the complexity of inverse problems, prior conditions are often used to reduce the complexity during the solution process. Traditional optical scattering methods almost always use strict contour constraints when solving inverse problems. For example, when measuring target gratings, rectangles or trapezoids are used to parameterize the structure, thus pre-defining the structure as a specific geometric shape and simplifying the reconstructed topography into a few shape parameters (e.g., height and width for a rectangle). This approach simplifies the inverse problem and improves the solution efficiency, but it also leads to the drawback of oversimplifying the structure. Although the complexity of the structure can be increased by adding more details to the pre-defined structure, such as adding Bézier curves to represent the curvature of some structures, this essentially follows the traditional measurement mode. To represent more details, shape parameters can only be continuously added, without truly optimizing the topography contour. However, if the contour of the structure is to be directly optimized, the dimensionality of the solution space in the inverse problem will increase significantly. Existing algorithms such as library search and gradient descent will consume a lot of time and computational costs, making them difficult to apply. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide an optical scattering method based on an inverse problem solving algorithm and a mixed measurement method for batch samples. This solves the problem that existing strict contour-constrained inverse problem solving algorithms require the addition of detailed shape parameters for characterization in order to overcome the drawback of oversimplification, which leads to a significant increase in the dimension of the solution space, high time and computational costs, and difficulty in application.

[0005] The technical solution to achieve the above objectives is:

[0006] This invention provides an optical scattering method based on an inverse problem solving algorithm, comprising the following steps:

[0007] Optical measurements are performed on the micro / nano structure to be measured to obtain the measured optical signal;

[0008] Obtain prior structural information related to the micro / nano structure to be measured;

[0009] An initial model is constructed based on the prior structural information related to the micro / nano structure to be measured.

[0010] Rigorous coupled-wave analysis was used to analyze the initial model to generate simulated optical signals;

[0011] Determine whether the difference between the generated simulated optical signal and the measured optical signal meets the set requirements;

[0012] If so, the contour of the model corresponding to the simulated optical signal is output as the measurement result;

[0013] If not, perform gradient calculations on each parameter in the initial model, use an automatic differentiation algorithm to obtain the gradients of all parameters, and then use the gradients of the parameters to optimize the initial model to obtain the optimized model. Repeat the steps of using rigorous coupled-wave analysis until it is determined that the difference between the generated simulated optical signal and the measured optical signal meets the set requirements.

[0014] A further improvement of the optical scattering method based on the inverse problem solving algorithm of this invention lies in the step of using rigorous coupled-wave analysis to analyze the initial model to generate simulated optical signals, which includes:

[0015] The initial model is divided into multiple micro-layers;

[0016] Obtain parameter information for each micro-layer;

[0017] Rigorous coupled-wave analysis was used to analyze the electromagnetic field distribution in each micro-layer to generate simulated optical signals.

[0018] A further improvement of the optical scattering method based on the inverse problem solving algorithm in this invention lies in setting an error function and using an optimizer to optimize the parameters when optimizing the initial model.

[0019] The error function is set as follows:

[0020]

[0021] In Equation 1, X 2 Let k represent the chi-square loss, N represent the total number of samples, and w represent the total number of samples. k y represents the weight of each sample. k f represents the experimentally measured value of the optical signal of the k-th sample. k(x) represents the numerical simulation value of the optical signal of the k-th sample, and x represents the set of parameters to be measured;

[0022] The result of optimizing the parameters using the optimizer is expressed as follows:

[0023]

[0024] In Equation 2, Ω represents the solution space consisting of all parameters that satisfy the physical meaning.

[0025] The further improvement of the optical scattering method based on the inverse problem solving algorithm of this invention lies in the fact that the process of generating simulated optical signals using rigorous coupling analysis can be represented by the following equations 3 and 4:

[0026]

[0027] In equations 3 and 4, a i b represents the superposition coefficient of the forward propagation components of the electric field of a plane wave of a certain order at the interface of the i-th micro-layer. i T represents the superposition coefficient of the backpropagation components of the electric field of a certain order plane wave at the interface of the i-th micro-layer. i Let c be the transmission matrix, representing the boundary conditions that must be satisfied at the interface between two adjacent micro-layers for electromagnetic wave propagation. i d represents the superposition coefficient of the forward propagation components of the electric field of a plane wave of a certain order at the interface of the i-th micro-layer. i Let represent the superposition coefficient of the backpropagation components of the electric field of a certain order plane wave at the interface of the i-th micro-layer. This represents the propagation phase of the i-th micro-layer.

[0028] A further improvement of the optical scattering method based on the inverse problem-solving algorithm of this invention lies in the use of an automatic differentiation algorithm to obtain the gradients of all parameters, which can be expressed by the following equations 5 and 6:

[0029]

[0030] In equations 5 and 6, T represents the mathematical operation of finding partial derivatives. i T This represents the mathematical operation of transpose.

[0031] A further improvement of the optical scattering method based on the inverse problem solving algorithm of this invention is that the prior structural information related to the micro / nano structure to be measured obtained includes the contour information of the micro / nano structure to be measured, and / or the size information of the micro / nano structure to be measured and / or the material information of the micro / nano structure to be measured.

[0032] This invention also provides a method for mixed measurement of batch samples using an optical scattering method based on an inverse problem solving algorithm, comprising the following steps:

[0033] The optical scattering method based on the inverse problem solving algorithm was used to measure each sample in the batch.

[0034] The steps for obtaining prior structural information related to the sample during the measurement process include:

[0035] A sample is selected from the batch of samples, and the selected sample is measured using a transmission electron microscope to obtain the transmission electron microscope structure of the sample.

[0036] The selected samples are measured to obtain the corresponding nominal structure;

[0037] The mapping relationship between the nominal structure and the transmission electron microscope structure can be obtained by using thin plate spline interpolation;

[0038] The samples other than the selected samples in the batch are measured to obtain other nominal structures. The estimated transmission structure corresponding to the other nominal structures is calculated by using the mapping relationship between the nominal structures and the transmission electron microscope structures.

[0039] An edge detection algorithm is used to extract the shape state from the transmission electron microscope structure and the estimated transmission mechanism as the prior structural information of the corresponding sample.

[0040] The beneficial effects of the optical scattering method based on the inverse problem solving algorithm and its mixed measurement method for batch samples in this invention are as follows:

[0041] This invention combines rigorous coupled-wave analysis and automatic differentiation algorithms to propose an inverse problem-solving algorithm suitable for optical scattering methods. It provides an algorithmic foundation for optimizing a large number of parameters in optical scattering methods. The automatic differentiation algorithm is used to calculate the gradient of structural parameters. The gradients of all parameters are obtained through a single forward simulation, realizing an optical scattering method with relaxed contour constraints. It can provide more detailed contour optimization and improve computational efficiency. Attached Figure Description

[0042] Figure 1 This is a schematic diagram of the inverse problem solving algorithm in the optical scattering method based on the inverse problem solving algorithm of this invention.

[0043] Figure 2 This is a flowchart of the optical scattering method based on the inverse problem solving algorithm of the present invention.

[0044] Figure 3 This is a schematic diagram of the simulation experiment of the optical scattering method based on the inverse problem solving algorithm of the present invention.

[0045] Figure 4 for Figure 3 The simulated spectrum of the structure is shown.

[0046] Figure 5 This is a schematic diagram illustrating the principle of obtaining prior structural information related to samples in the mixed measurement method of batch samples using an optical scattering method based on an inverse problem solving algorithm, as described in this invention. Detailed Implementation

[0047] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0048] See Figure 2 This invention provides an optical scattering method based on an inverse problem-solving algorithm and its hybrid measurement method for batch samples, applicable to the measurement of micro / nano structures such as chips, integrated circuits, optical waveguides, grating structures, and superlenses in AR glasses. The inverse problem-solving algorithm combines rigorous coupled-wave analysis and automatic differentiation algorithms, obtaining the gradients of all contour parameters in a single forward simulation, suitable for the measurement requirements of optical scattering methods. This algorithm allows for open optimization of the parameters of each micro-layer; given a sufficient number of micro-layers, their parameters can describe any shape. By incorporating prior information, this algorithm achieves optimization results with higher degrees of freedom and more structural details than traditional strict contour constraints, while also possessing computational efficiency. The following description, in conjunction with the accompanying drawings, illustrates the optical scattering method based on the inverse problem-solving algorithm and its hybrid measurement method for batch samples.

[0049] The optical scattering method based on the inverse problem solving algorithm of the present invention includes the following steps:

[0050] S11, perform optical measurements on the micro / nano structure to be measured to obtain the measured optical signal;

[0051] S12, Obtain prior structural information related to the micro / nano structure to be measured;

[0052] S13, Based on the acquired prior structural information related to the micro / nano structure to be measured, construct an initial model;

[0053] S14, using rigorous coupled-wave analysis to analyze the initial model to generate simulated optical signals;

[0054] S15, determine whether the difference between the generated simulated optical signal and the measured optical signal meets the set requirements;

[0055] If so, the contour of the model corresponding to the simulated optical signal is output as the measurement result;

[0056] If not, perform gradient calculations on each parameter in the initial model, use an automatic differentiation algorithm to obtain the gradients of all parameters, and then use the gradients of the parameters to optimize the initial model to obtain the optimized model. Repeat the steps of using strict coupled-wave analysis (i.e., repeat step S14) until it is determined that the difference between the generated simulated optical signal and the measured optical signal meets the set requirements, and output the corresponding measurement results.

[0057] In the optical scattering method of the present invention, the optical measurement in step S11 can be performed by a suitable optical signal measurement device. This optical signal measurement device is a prior art device that can project an incident light beam onto a micro-nano structure. The incident light beam is received after being diffracted by the micro-nano structure, and the received diffraction signal is the measured optical signal.

[0058] In one specific embodiment of the present invention, in step S12 of the optical scattering method, the prior structural information related to the micro / nano structure to be measured includes the contour information and / or the size information and / or the material information of the micro / nano structure to be measured. The size information of the micro / nano structure to be measured can be obtained through actual measurement using measuring tools, or it can be estimated. This size information serves as a general reference, intended to provide a basis for model construction. The contour information of the micro / nano structure to be measured can be a general description of the contour, such as a rough description of the contour, and may also include periodic information. Theoretically, the contour of the micro / nano structure to be measured can be of any shape. The inverse problem solving algorithm of the present invention does not require precise prior structural information; it only needs to be able to roughly characterize the micro / nano structure to be measured. Of course, if the provided prior structural information is precise contour information and / or size information, the measurement accuracy of the optical scattering method of the present invention can be improved.

[0059] In one specific embodiment of the present invention, in step S13 of the optical scattering method of the present invention, an initial model is constructed, which can be achieved by numerical simulation.

[0060] In one specific embodiment of the present invention, the step of analyzing the initial model using rigorous coupled-wave analysis to generate simulated optical signals includes:

[0061] The initial model is divided into multiple micro-layers;

[0062] Obtain parameter information for each micro-layer;

[0063] Rigorous coupled-wave analysis was used to analyze the electromagnetic field distribution in each micro-layer to generate simulated optical signals.

[0064] When dividing the initial model into micro-layers, the number of micro-layers should be as large as possible, ranging from 50 to 70 layers, or even more. If the number of micro-layers is sufficient, the parameters of the micro-layers can be used to describe any shape. The number of micro-layers needs to be determined in conjunction with prior information. This number can be set manually; typically, for slowly changing contours, 50 micro-layers are sufficient, while for steeply changing contours, the number of micro-layers can be appropriately increased.

[0065] Furthermore, the process of generating simulated optical signals using rigorous coupling analysis can be represented by the following equations 3 and 4:

[0066]

[0067] In equations 3 and 4, a i b represents the superposition coefficient of the forward propagation components of the electric field of a plane wave of a certain order at the interface of the i-th micro-layer. i T represents the superposition coefficient of the backpropagation components of the electric field of a certain order plane wave at the interface of the i-th micro-layer. i Let c be the transmission matrix, representing the boundary conditions that must be satisfied at the interface between two adjacent micro-layers for electromagnetic wave propagation. i d represents the superposition coefficient of the forward propagation components of the electric field of a plane wave of a certain order at the interface of the i-th micro-layer. i Let represent the superposition coefficient of the backpropagation components of the electric field of a certain order plane wave at the interface of the i-th micro-layer. Let represent the propagation phase of the i-th micro-layer, j represent the imaginary sign, and e represent the mathematical operation exp.

[0068] Furthermore, using an automatic differentiation algorithm, the gradients of all parameters can be obtained as follows:

[0069] Equations 5 and 6 represent:

[0070]

[0071] In equations 5 and 6, T represents the mathematical operation of finding partial derivatives. i T This represents the mathematical operation of transpose.

[0072] When using automatic differentiation algorithms to calculate the gradients of the parameters in the initial model, the gradients are calculated for the parameters of each micro-layer of the initial model. These micro-layer parameters may include the width, thickness, refractive index, etc. To simplify the problem, the material information of the micro-layers is usually prior, and the measured parameters typically only contain structural information. Specifically, the measured parameters are a set, including the thickness, width, and center position of each micro-layer, and these parameters correspond to the structural profile.

[0073] like Figure 1 The diagram shows the principle of the inverse problem solving algorithm of this invention. The model structure is divided into i micro-layers. The propagation of electromagnetic waves from layer (i-1) to layer i is the forward process, and the propagation from layer i to layer (i-1) is the reverse process. The forward process is described by equations 3 and 4 above, and the reverse process is described by equations 5 and 6 above. Equations 5 and 6 are designed to obtain the gradient matrices of all micro-layers through an automatic differentiation algorithm. This invention's rigorous coupled-wave analysis analytically constructs the forward process, allowing gradient calculation of the structural parameters (i.e., the parameters of the micro-layers) using the chain rule. By using the automatic differentiation algorithm, the gradients of all parameters can be obtained in a single forward simulation.

[0074] The rigorous coupled-wave analysis (or rigorous coupled-wave theory) in the inverse problem solving algorithm of this invention is a method for solving Maxwell's equations using plane wave expansion. Due to its advantages such as good convergence and ease of programming implementation, it is a widely used solution method in optical scattering measurements. Specifically, this rigorous coupled-wave analysis first divides the model structure into layers, and rigorously solves the eigenvalue equations in each layer by performing Fourier transforms on the field and material dielectric functions. Then, a transmission matrix is ​​constructed based on the boundary conditions to obtain the transmission relationship of electromagnetic waves between layers, and finally, the optical signal (i.e., simulated optical signal) of the interaction between the electromagnetic wave and the sample is solved. The inverse problem solving algorithm of this invention also incorporates an automatic differentiation algorithm. Automatic differentiation is a technique widely used in machine learning that enables backpropagation of gradients in neural networks without manual calculation. Its core idea is to store the calculation process as a computational graph and then use the chain rule to calculate the gradient. Since the solution process of the interaction signal between light and sample in rigorous coupled-wave analysis is analytical, the gradient of the structural parameters can be calculated using the chain rule. Then, combined with the automatic differentiation algorithm, the gradients of all parameters can be obtained through a single forward simulation. The inverse problem solving algorithm of this invention allows for open optimization of the parameters of each micro-layer. As long as the number of micro-layers is sufficient, the parameters of these micro-layers can be used to describe any shape. Furthermore, this inverse problem solving algorithm can incorporate more detailed prior information, resulting in optimization results with higher degrees of freedom and more structural detail than those with strict contour constraints.

[0075] In one specific embodiment of the present invention, when optimizing the initial model, an error function is set, and an optimizer is used to optimize the parameters.

[0076] The error function is set as follows:

[0077]

[0078] In Equation 1, X 2 Let k represent the chi-square loss, N represent the total number of samples, and w represent the total number of samples. k y represents the weight of each sample. k f represents the experimentally measured value of the optical signal of the k-th sample. k (x) represents the numerical simulation value of the optical signal of the k-th sample, and x represents the set of parameters to be measured;

[0079] The result of optimizing the parameters using the optimizer is expressed as follows:

[0080]

[0081] In Equation 2, Ω represents the solution space consisting of all parameters that satisfy the physical meaning.

[0082] like Figure 3 As shown, the feasibility of the inverse problem solving algorithm of the present invention will be verified through a model experiment. Figure 3 The first structure in ( Figure 3 The leftmost structure is the target structure, which is a model established using numerical simulation. The target structure is set as a one-dimensional tilted grating, with an added curvature at the top to increase the complexity of the structure. The base and top layer materials are set as monocrystalline silicon and amorphous silicon, respectively, and the middle layer material is set as silicon dioxide.

[0083] An initial model is established based on the prior structural information of the target structure. The structure of this initial model is shown below. Figure 3The initial structure, located in the middle, can be manually input as prior structural information, and then a numerical simulation is used to establish the initial structural model. The initial and target structures are divided into multiple micro-layers, and the contour of the target structure can be described by the positions of these micro-layers. During the optimization of the initial structure, the relative difference between the positions of each micro-layer in the initial structure and the positions of the micro-layers in the target structure is kept below a certain threshold to constrain the contour. For example, 70 micro-layers are formed, the substrate material is set to monocrystalline silicon, and the materials for the first 20 layers and the last 50 layers are set to amorphous silicon and silicon dioxide, respectively. For the first 30 layers, the threshold for the difference in relative position between the optimized model structure and the target structure is set to 1 nanometer. For the last 40 layers, after each optimization, the positions on both sides of the micro-layers in the optimized model structure are linearly fitted to ensure that the edges of the micro-layers fall on the fitted line. In actual measurement, the period of the structure can be directly solved analytically using the grating equation (as shown in Equation 7), and the period is usually not set as an optimization parameter.

[0084]

[0085] In Equation 7, θ i θ represents the angle of incidence. m λ represents the diffraction angle, m represents the diffraction order, λ represents the incident wavelength, and Λ represents the grating period.

[0086] The initial structure has the same period and material settings as the target structure. Optimization is performed using the Adam optimizer with a learning rate of 1e. -5 Choosing the error function represented by Equation 1 above, and setting the target signal to be within the wavelength range of 200-1000 nanometers when the incident angle is 10°, 30°, and 50°, - Polarized mirror reflection spectrum. The optimized structure can be represented by Equation 2 above. The optimized structure is as follows. Figure 3 The third structure (i.e., the result structure) is shown in the diagram. This result structure is highly consistent with the target structure, as shown in the diagram. Figure 4 As shown, it demonstrates Figure 3 The simulated spectra of the target structure, initial structure, and result structure were obtained. The GOF value of the initial spectrum was 0.90, while the optimized spectrum reached a GOF value of 0.99, thus verifying the feasibility of the inverse problem solving algorithm of the present invention.

[0087] like Figure 2As shown, the process of the optical scattering method based on the inverse problem solving algorithm of the present invention includes: First, providing a sample, which is a micro / nano structure to be measured; Second, performing optical measurements on the provided sample to obtain optical signals; Third, using the algorithm under the optical scattering system to generate a simulated optical signal corresponding to the initial model, and determining whether the simulated optical signal matches the optical signal obtained by optical measurement or whether the difference is less than a set threshold. If yes, it indicates that the requirements have been met, and the result is output, which is the contour of the model corresponding to the simulated optical signal; if no, the surface has not met the requirements, and then the fourth step is executed to update the parameters. The updated parameters are the parameters of all micro-layers, which may include the width, thickness, refractive index, etc. of the micro-layers. The fifth step involves a relaxed contour constraint, which, compared to traditional strict contour constraints such as setting shape parameters like rectangles or trapezoids, allows for a free micro-layered contour shape that can be arbitrary. Here, an automatic differentiation algorithm is used to obtain the gradients of all parameters. The sixth step involves optimizing the initial model through numerical simulation based on the gradients of all parameters. Then, an algorithm under the optical scattering system is used to generate the corresponding simulated optical signal, which is then judged until the simulated optical signal matches the optical signal obtained by optical measurement or the difference is less than a set threshold.

[0088] The present invention also provides a method for mixed measurement of batch samples using an optical scattering method based on an inverse problem solving algorithm, which is described below.

[0089] The present invention provides a method for measuring the mixture of batch samples using an optical scattering method based on an inverse problem solving algorithm, comprising the following steps:

[0090] The optical scattering method based on the inverse problem solving algorithm described above was used to measure each sample in the batch.

[0091] The steps for obtaining prior structural information related to the sample during the measurement process include:

[0092] A sample is selected from the batch of samples, and the selected sample is measured using a transmission electron microscope to obtain the transmission electron microscope structure of the sample.

[0093] The selected samples are measured to obtain the corresponding nominal structure;

[0094] like Figure 5 As shown, the mapping relationship between the nominal structure and the transmission electron microscope structure is obtained by using thin plate spline interpolation;

[0095] The samples other than the selected samples in the batch are measured to obtain other nominal structures. The corresponding estimated transmission structure is calculated using the mapping relationship between the nominal structure and the transmission electron microscope structure.

[0096] An edge detection algorithm is used to extract the shape state from the transmission electron microscope structure and the estimated transmission mechanism as the prior structural information of the corresponding sample.

[0097] This invention combines transmission electron microscopy (TEM) and optical scattering methods to obtain more accurate and detailed measurement results. However, measuring each sample using TEM would significantly increase costs and time, making it impractical. To address this issue, this invention proposes combining thin-plate spline interpolation to estimate the structure of other samples not examined by TEM, achieving rapid and detailed structural reconstruction of batches of samples at a relatively low cost.

[0098] The beneficial effect of this invention is that it develops an algorithm for solving the inverse problem of optical scattering, providing an algorithmic basis for optimizing a large number of parameters in optical scattering. By combining this algorithm, an optical scattering method with relaxed contour constraints can be realized, providing a new technical means for optical scattering. It can not only provide more detailed contour optimization, but also be compatible with more detailed prior contour information, while having a certain degree of computational efficiency.

[0099] The present invention has been described in detail above with reference to the accompanying drawings and embodiments. Those skilled in the art can make various modifications to the present invention based on the above description. Therefore, certain details in the embodiments should not be construed as limiting the present invention, and the scope of protection of the present invention shall be defined by the appended claims.

Claims

1. An optical scattering method based on an inverse problem-solving algorithm, characterized in that, Includes the following steps: Optical measurements are performed on the micro / nano structure to be measured to obtain the measured optical signal; Obtain prior structural information related to the micro / nano structure to be measured; An initial model is constructed based on the prior structural information related to the micro / nano structure to be measured. Rigorous coupled-wave analysis was used to analyze the initial model to generate simulated optical signals; Determine whether the difference between the generated simulated optical signal and the measured optical signal meets the set requirements; If so, the contour of the model corresponding to the simulated optical signal is output as the measurement result; If not, perform gradient calculations on each parameter in the initial model, use an automatic differentiation algorithm to obtain the gradients of all parameters, and then use the gradients of the parameters to optimize the initial model to obtain the optimized model. Repeat the steps of using rigorous coupled-wave analysis until it is determined that the difference between the generated simulated optical signal and the measured optical signal meets the set requirements.

2. The optical scattering method based on the inverse problem solving algorithm as described in claim 1, characterized in that, The steps for generating simulated optical signals by analyzing the initial model using rigorous coupled-wave analysis include: The initial model is divided into multiple micro-layers; Obtain parameter information for each micro-layer; Rigorous coupled-wave analysis was used to analyze the electromagnetic field distribution in each micro-layer to generate simulated optical signals.

3. The optical scattering method based on the inverse problem solving algorithm as described in claim 1 or 2, characterized in that, When optimizing the initial model, an error function is set, and an optimizer is used to optimize the parameters. The error function is set as follows: In Equation 1, X 2 Let k represent the chi-square loss, N represent the total number of samples, and w represent the total number of samples. k y represents the weight of each sample. k f represents the experimentally measured value of the optical signal of the k-th sample. k (x) represents the numerical simulation value of the optical signal of the k-th sample, and x represents the set of parameters to be measured; The result of optimizing the parameters using the optimizer is expressed as follows: In Equation 2, Ω represents the solution space consisting of all parameters that satisfy the physical meaning.

4. The optical scattering method based on the inverse problem solving algorithm as described in claim 1 or 2, characterized in that, The process of generating simulated optical signals using rigorous coupling analysis can be represented by the following equations 3 and 4: In equations 3 and 4, a i b represents the superposition coefficient of the forward propagation components of the electric field of a plane wave of a certain order at the interface of the i-th micro-layer. i T represents the superposition coefficient of the backpropagation components of the electric field of a certain order plane wave at the interface of the i-th micro-layer. i Let c be the transmission matrix, representing the boundary conditions that must be satisfied at the interface between two adjacent micro-layers for electromagnetic wave propagation. i d represents the superposition coefficient of the forward propagation components of the electric field of a plane wave of a certain order at the interface of the i-th micro-layer. i Let represent the superposition coefficient of the backpropagation components of the electric field of a certain order plane wave at the interface of the i-th micro-layer. This represents the propagation phase of the i-th micro-layer.

5. The optical scattering method based on the inverse problem solving algorithm as described in claim 4, characterized in that, Using an automatic differentiation algorithm, the gradients of all parameters can be obtained, and can be expressed by the following equations 5 and 6: In equations 5 and 6, This represents the mathematical operation of finding partial derivatives. This represents the mathematical operation of transpose.

6. The optical scattering method based on the inverse problem solving algorithm as described in claim 2, characterized in that, The prior structural information acquired related to the micro / nano structure to be measured includes the contour information of the micro / nano structure to be measured, and / or the size information of the micro / nano structure to be measured, and / or the material information of the micro / nano structure to be measured.

7. A method for mixed measurement of batch samples using an optical scattering method based on an inverse problem solving algorithm, characterized in that, Includes the following steps: The optical scattering method based on the inverse problem solving algorithm as described in any one of claims 1 to 6 is used to measure each sample in a batch of samples; The steps for obtaining prior structural information related to the sample during the measurement process include: A sample is selected from the batch of samples, and the selected sample is measured using a transmission electron microscope to obtain the transmission electron microscope structure of the sample. The selected samples are measured to obtain the corresponding nominal structure; The mapping relationship between the nominal structure and the transmission electron microscope structure can be obtained by using thin plate spline interpolation; The samples other than the selected samples in the batch are measured to obtain other nominal structures. The estimated transmission structure corresponding to the other nominal structures is calculated by using the mapping relationship between the nominal structures and the transmission electron microscope structures. An edge detection algorithm is used to extract the shape state from the transmission electron microscope structure and the estimated transmission mechanism as the prior structural information of the corresponding sample.

Images