A critical dimension measurement method, device, apparatus and storage medium

Abstract

Embodiments of the present application provide a key size measurement method, device and equipment, and a storage medium, and relate to the technical field of communication. The method comprises: obtaining a scatter diagram of an object to be measured; the scatter diagram of the object to be measured is formed after X-rays are scattered by the object to be measured; extracting a characteristic signal of the object to be measured from the scatter diagram of the object to be measured; determining a key size of the object to be measured based on the characteristic signal of the object to be measured through a target model; and wherein the target model is obtained by training an initial model through simulation samples and then training the initial model again through measurement samples. In the embodiments of the present application, the initial model is trained through simulation samples, and a basic model with good generalization ability can be obtained. Then, the initial model is trained again using the object to be measured, and a target model is obtained, which can effectively improve the prediction deviation problem caused by random interference of the measurement system.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method, apparatus, device, and storage medium for measuring key dimensions. Background Technology

[0002] Small-angle X-rays are generally used to measure the critical dimension (CD) of semiconductor devices. X-rays are scattered after passing through the object under test, forming a scattering map containing structural information of the object. By analyzing the distribution characteristics of the scattering map, the critical dimension of the object under test can be measured.

[0003] The conventional method for determining critical dimensions is as follows: First, the data acquisition system collects small-angle X-ray scattering field signals of the object under test. Then, a set of CDs is used to characterize the physical model of the object. Maxwell's equations or approximate physical formulas are used to solve for the simulated scattering field distribution corresponding to a set of CD_i structures. Finally, the difference between the simulated scattering field and the collected experimental scattering field signal of the sample is calculated. An optimization algorithm is used to update the CD_i structure, and the minimum difference between the simulated and experimental scattering field signals is found. The CD corresponding to this minimum value is the critical dimension of the object under test.

[0004] However, as semiconductor structures become increasingly complex and their dimensions become smaller, conventional methods for measuring the critical dimensions of semiconductors are too slow to meet industrial needs. Summary of the Invention

[0005] This application provides a method, apparatus, device, and storage medium for measuring critical dimensions, which improves the speed and accuracy of critical dimension measurement.

[0006] In a first aspect, embodiments of this application provide a method for measuring a critical dimension, including:

[0007] Obtain a scattering pattern of the object to be measured; the scattering pattern of the object to be measured is formed after X-rays are scattered by the object to be measured;

[0008] Extract the feature signal of the object to be measured from the scattering map of the object to be measured;

[0009] Based on the feature signals of the object to be measured, the key dimensions of the object to be measured are determined by a target model; wherein, the target model is obtained by training an initial model with simulation samples, and then training the initial model again with measurement samples.

[0010] In this embodiment, an initial model is trained using simulated samples to obtain a basic model with good generalization ability, which can effectively avoid overfitting. The initial model is then trained again using the object to be measured to obtain the target model, making the target model more stable and accurate, and effectively improving the prediction bias caused by random interference in the measurement system.

[0011] Optionally, the scattering pattern of the object to be measured is formed after X-rays are scattered by the object to be measured, including:

[0012] The scattering pattern of the object to be measured is formed by X-rays being scattered by the object at different azimuth angles and different incident angles, or the scattering pattern of the object to be measured is formed by X-rays being scattered by the object at the same azimuth angle and different incident angles.

[0013] In this embodiment, if the structure of the object to be measured is irregular, by irradiating the object with X-rays at different azimuth angles and different incident angles, the scattering patterns of the object at various angles can be obtained, thus making the acquisition of the key dimensions of the object more accurate. If the structure of the object to be measured is relatively regular, only the scattering patterns at different incident angles at the same azimuth angle can be acquired, avoiding the large workload caused by acquiring scattering patterns and improving the efficiency of scattering pattern acquisition.

[0014] Optionally, the target model is obtained by training an initial model using simulation samples, and then retraining the initial model using measurement samples, including:

[0015] For any simulation sample, the feature signal of the simulation sample is used as input and the key dimensions of the simulation sample are used as output to train an initial model.

[0016] For any measurement sample, the feature signal of the measurement sample is input into the initial model to obtain the predicted key size of the measurement sample; based on the predicted key size, the predicted feature signal corresponding to the predicted key size is determined; the initial model is adjusted by the loss value between the feature signal of the measurement sample and the predicted feature signal until the training termination condition is met.

[0017] In this embodiment, the key dimensions are obtained through feature signals. When training the model, it is not necessary to rely on label values, which can overcome the problem that the key dimensions are difficult to obtain.

[0018] Optionally, determining the predicted feature signal corresponding to the predicted key size based on the predicted key size includes:

[0019] The predicted key dimensions are input into the prediction model to obtain the predicted feature signals corresponding to the predicted key dimensions; the prediction model is trained by taking the key dimensions of the simulation samples as input and the feature signals of the simulation samples as output.

[0020] In this embodiment of the application, by training the prediction model, feature signals can be obtained quickly, which can effectively improve the training speed of the model.

[0021] Optionally, the number of simulated samples is much larger than the number of measured samples.

[0022] Optionally, the feature signal of the simulated sample is obtained based on the distribution characteristics of the scattered signal in the scattering map of the simulated sample and the correlation between the key dimensions of the simulated sample and the scattering map of the simulated sample.

[0023] Optionally, the feature signal of the measured sample is obtained by inputting the scattering map of the measured sample into the feature extraction model; the feature extraction model is obtained by training the model with the scattering map of the simulated sample as input and the feature signal of the simulated sample as output.

[0024] Optionally, the feature extraction model is a model that can perform feature extraction on scattering maps under different azimuth angles and different incident angles, or there are multiple feature extraction models, each of which can perform feature extraction on scattering maps under a single azimuth angle and a single incident angle.

[0025] Secondly, embodiments of this application provide a measuring device for a critical dimension, comprising:

[0026] The acquisition module is used to acquire the scattering pattern of the object to be measured; the scattering pattern of the object to be measured is formed by X-rays being scattered by the object at different azimuth angles and different incident angles;

[0027] An extraction module is used to extract the feature signals of the object to be measured from the scattering map of the object to be measured;

[0028] The determination module is used to determine the key dimensions of the object to be measured based on the feature signals of the object to be measured through a target model; wherein the target model is obtained by training an initial model with simulation samples and then retraining the initial model with measurement samples.

[0029] Optionally, the determining module is specifically used for:

[0030] For any simulation sample, the feature signal of the simulation sample is used as input and the key dimensions of the simulation sample are used as output to train an initial model.

[0031] For any measurement sample, the feature signal of the measurement sample is input into the initial model to obtain the predicted key size of the measurement sample; based on the predicted key size, the predicted feature signal corresponding to the predicted key size is determined; the initial model is adjusted by the loss value between the feature signal of the measurement sample and the predicted feature signal until the training termination condition is met.

[0032] Optionally, the determining module is specifically used for:

[0033] The predicted key dimensions are input into the prediction model to obtain the predicted feature signals corresponding to the predicted key dimensions; the prediction model is trained by taking the key dimensions of the simulation samples as input and the feature signals of the simulation samples as output.

[0034] Optionally, the number of simulated samples is much larger than the number of measured samples.

[0035] Optionally, the feature signal of the simulated sample is obtained based on the distribution characteristics of the scattered signal in the scattering map of the simulated sample and the correlation between the key dimensions of the simulated sample and the scattering map of the simulated sample.

[0036] Optionally, the feature signal of the measured sample is obtained by inputting the scattering map of the measured sample into the feature extraction model; the feature extraction model is obtained by training the model with the scattering map and key dimensions of the simulated sample as input and the feature signal of the simulated sample as output.

[0037] Optionally, the feature extraction model is a model that can perform feature extraction on scattering maps under different azimuth angles and different incident angles, or there are multiple feature extraction models, each of which can perform feature extraction on scattering maps under a single azimuth angle and a single incident angle.

[0038] Thirdly, embodiments of this application provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of any of the methods described above.

[0039] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program executable by a computer device, which, when run on the computer device, causes the computer device to perform the steps of any of the methods described above. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of the present invention, 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 the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 A system architecture diagram provided for an embodiment of this application;

[0042] Figure 2 A flowchart illustrating a method for measuring a critical dimension provided in this application embodiment. Figure 1 ;

[0043] Figure 3 A top view and a side view of a certain object to be measured provided in an embodiment of this application;

[0044] Figure 4 The scattering patterns of the object to be measured obtained under different incident angles and azimuth angles provided in the embodiments of this application;

[0045] Figure 5 This application provides a schematic diagram of a process for retraining an initial model, as illustrated in an embodiment. Figure 1 ;

[0046] Figure 6 This is a feature signal distribution map of the object to be measured provided in an embodiment of this application;

[0047] Figure 7 The key dimensions of the object to be measured in the x and y directions provided by the embodiments of this application;

[0048] Figure 8 A flowchart illustrating a method for measuring a critical dimension provided in this application embodiment. Figure 2 ;

[0049] Figure 9 A measuring device for a critical dimension provided in an embodiment of this application;

[0050] Figure 10 A computer device provided in an embodiment of this application. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0052] To facilitate understanding of this solution, its application scenarios are described below.

[0053] Small-angle X-ray scattering (SAXS) is a critical dimension measurement method that utilizes the scattered X-rays after they pass through an object under test (AUT) to form a scattered field containing structural information of the AUT. The distribution characteristics of this scattered field are analyzed to measure the critical dimensions of the AUT. The method for solving critical dimensions using SAXS mainly includes the following steps: 1. Acquire the small-angle X-ray scattered field signal of the AUT using a data acquisition system. 2. Use a set of critical dimensions to characterize the physical model of the AUT. The AUT is characterized by a set of critical dimensions, which consists of multiple CD_i structures. Maxwell's equations or approximate physical formulas are used to solve for the simulated scattered field distribution corresponding to the CD_i structure. 3. Calculate the difference between the simulated scattered field and the acquired experimental scattered field signal of the AUT. Update the CD_i structure using an optimization algorithm to find the minimum difference between the simulated and experimental scattered field signals. The corresponding CD_min is the critical dimension of the AUT sample.

[0054] As semiconductor devices become increasingly complex and smaller, the dimensions of key dimensions in physical models increase significantly. This not only leads to longer computation times for simulating scattering fields but also increases the number of iterations required for optimization algorithms to find the optimal solution, ultimately resulting in a very slow solution process that cannot meet industrial demands. To overcome these difficulties, existing technologies use machine learning methods to replace parts of the solution process (such as physical simulation calculations), achieving acceleration. However, there is currently no method for directly predicting the structure of a test sample using machine learning. This is mainly due to the following limitations: 1. Measurement data is difficult to obtain, especially the labeled results, making it impossible to conduct large-scale machine learning training. 2. Models trained with limited measurement data are prone to overfitting and have poor generalization ability, failing to meet measurement requirements. 3. While simulation data can be used for training, the discrepancy between measurement and simulation data means that the accuracy of simulation predictions trained solely on simulation data cannot meet requirements. The following describes the key dimension measurement method provided by the embodiments of this application:

[0055] First, simulation samples are obtained through a simulation program. These samples are obtained from different incident and azimuth angles. A scattering map of the simulation samples is then obtained, and feature signals are extracted from the scattering map. The feature signals of the simulation samples are used as input, and the key dimensions of the simulation samples are used as output to train an initial model. Next, the feature signals of the measured samples are used as input to the initial model to obtain the predicted key dimensions of the measured samples. Finally, the predicted key dimensions of the measured samples are input into the physical or predictive model to obtain the predicted feature signals of the measured samples. The loss value between the predicted feature signals and the feature signals of the measured samples is compared. If the loss value is greater than or equal to a preset threshold, the initial model is adjusted; if the loss value is less than the preset threshold, the initial model is saved.

[0056] refer to Figure 1 The diagram shown illustrates a system architecture provided in an embodiment of this application, including a terminal device 101 and a server 102. The terminal device 101 is pre-installed with business applications, which may be client applications, web applications, mini-program applications, etc. The terminal device 101 may be a smartphone, tablet computer, laptop computer, desktop computer, smart home appliance, smart voice interaction device, smart vehicle device, etc., but is not limited to these.

[0057] Server 102 is used to implement initial model training and target model training. The initial and target models respond to the developer's operations, initiating training and storing the trained models in a database. Server 102 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms. Terminal device 101 and server 102 can be directly or indirectly connected via wired or wireless communication; this application does not impose any restrictions.

[0058] based on Figure 1 The system architecture diagram provided in this application embodiment shows a flowchart of a method for measuring a critical dimension, such as... Figure 2 As shown, the process of this method is executed by a computer device, which can be... Figure 1 The server 102 shown includes the following steps:

[0059] Step 201: Obtain the scattering pattern of the object to be measured; the scattering pattern of the object to be measured is formed after X-rays are scattered by the object to be measured.

[0060] Specifically, the object to be measured is an object whose critical dimensions need to be measured. As the structure and size of semiconductor devices become increasingly complex and smaller, measuring the critical dimensions of semiconductor devices becomes increasingly difficult. Therefore, this application focuses on measuring the critical dimensions of semiconductor devices, but is not limited to semiconductor devices. After irradiating the object to be measured with X-rays, a scattering pattern of the object is obtained. For example... Figure 3 The image shows a top view and a side view of a semiconductor device.

[0061] In some embodiments, the scattering pattern of the object to be measured is formed after X-rays are scattered by the object, including: the scattering pattern of the object to be measured is formed after X-rays are scattered by the object at different azimuth angles and different incident angles, or the scattering pattern of the object to be measured is formed after X-rays are scattered by the object at the same azimuth angle and different incident angles. Specifically, the scattering pattern of the object to be measured is formed after X-rays irradiate and are scattered. The scattering pattern of the object to be measured can be formed by X-rays at different azimuth angles and different incident angles, or it can be formed by X-rays at the same azimuth angle and different incident angles.

[0062] The azimuth and incident angles used to irradiate the object under test by X-rays are determined by the structure of the object. If the object is a cylinder, and the critical dimension is its diameter, then since the diameter of the cylinder is the same at any cross-section, it is not necessary to obtain multiple azimuth and incident angles to obtain the diameter of the cylinder. If the object is an irregular object, and the critical dimension is its diameter, then to obtain the diameter of the irregular object at each cross-section, it is necessary to irradiate from different azimuths, and the incident angle at each azimuth must also be different. Only in this way can the diameter of the irregular object at each cross-section be obtained more accurately.

[0063] Different azimuth angles yield a set of key dimensions, which characterize the structure of an object under test. For example, if the object is placed in a three-dimensional space divided into x, y, and z directions, and the azimuth angle is divided into x and y directions; with the horizontal direction as the reference 0°, the range of the incident angle is generally within ±30°. Then, the key dimensions in the x and y directions can characterize the structure of the object. Multiple incident angles correspond to the x and y directions, and each incident angle generates a scattering pattern. For example... Figure 4 The image shows the scattering patterns of the object under different incident angles. Step 202: Extract the feature signals of the object under test from the scattering patterns.

[0064] Specifically, after obtaining the scattering map of the object to be measured, the feature signal of the object to be measured is extracted from the scattering map.

[0065] For example, extracting the feature signals of the object being measured from its scattering pattern. For instance, in... Figure 4 In the process, the central light spot is composed of multiple pixels. All pixels in the multiple light spots can be acquired as feature signals, or only the highest pixel in each light spot can be acquired from the multiple pixels and used to represent the light spot. In this case, the feature signal is composed of the highest pixel in each light spot.

[0066] Step 203: Based on the feature signals of the object to be measured, determine the key dimensions of the object to be measured through the target model; wherein, the target model is obtained by training an initial model through simulation samples, and then training the initial model again through measurement samples.

[0067] Specifically, after acquiring the feature signals of the object to be measured, the feature signals are input into the target model, and the target model outputs the key dimensions. Thus, the key dimensions of the object to be measured can be obtained based on the feature signals. The target model is obtained through two training processes: first, through training with simulation samples to obtain an initial model; then, through training with measurement samples, the initial model is trained again to obtain the target model.

[0068] The initial model can be a traditional machine learning model (support vector machine, random forest, decision tree, etc.), a neural network model (fully connected, convolutional network, recurrent neural network, etc.), or a deep neural network model (residual network, etc.).

[0069] In some embodiments, the target model is obtained by training an initial model using simulated samples, and then retraining the initial model using measured samples, such as... Figure 5 As shown, it includes the following steps:

[0070] Step 501: For any simulation sample, take the feature signal of the simulation sample as input and the key dimensions of the simulation sample as output to train the initial model.

[0071] Specifically, for any given simulation sample, the feature signal and key dimensions of the simulation sample are obtained. The feature signal of the simulation sample is used as input, and the key dimensions of the simulation sample are used as output to train the initial model.

[0072] The simulation samples are obtained through a simulation program. The key dimensions of the simulation samples are set to approximate the key dimensions of the object to be measured. The simulation program then generates a dataset of simulation samples near the object. The number of simulation samples needed to be determined based on the structure of the object. After obtaining the simulation samples, their feature signals are input into an initial model. The initial model is then iteratively trained using the key dimensions of the simulation samples. The measurement parameters of the simulation samples must be consistent with those of the measured samples.

[0073] For example, if the critical dimension of the object to be measured is 200 nanometers, the critical dimension of the simulation sample is set to ±20%, that is, between 160 nanometers and 240 nanometers, to obtain the simulation sample. This acquisition method involves generating a simulation sample near the object to be measured. The characteristic signal of the simulation sample is used as input, and the critical dimension of the simulation sample is used as output to train the initial model. When acquiring the simulation sample, if the distance between the object to be measured and the X-ray source is 10 meters, then the simulation sample should also be set to be 10 meters away from the X-ray source. Since the simulated scattering pattern is calculated based on the hardware configuration of the measurement system (such as detector distance, light source size, detector size, etc.), the acquisition parameters of the simulation sample must be consistent with those of the measured sample.

[0074] In some embodiments, the number of simulated samples is much larger than the number of measured samples.

[0075] Specifically, when acquiring simulation samples, the number of simulation samples needs to be much larger than the number of measurement samples. For example, if the number of simulation samples is X and the number of measurement samples is Y, then XY > N, where N is 10. 6 .

[0076] Step 502: For any measurement sample, input the feature signal of the measurement sample into the initial model to obtain the predicted key size of the measurement sample; based on the predicted key size, determine the predicted feature signal corresponding to the predicted key size; adjust the initial model by the loss value between the feature signal of the measurement sample and the predicted feature signal until the training termination condition is met.

[0077] Specifically, after obtaining the initial model through training, it is retrained to obtain the target model. Measurement samples are acquired; these samples are feature signals of the object under test measured by hardware devices. The feature signals of the measurement samples are input into the initial model to obtain the predicted key dimensions of the measurement samples. Based on the predicted key dimensions of the measurement samples, the predicted feature signals of the measurement samples are determined. The loss value between the predicted feature signals and the feature signals of the measurement samples is determined. The initial model is adjusted according to the magnitude of the loss value. If the loss value is less than a preset threshold, the initial model is saved as the final target model; if the loss value is greater than or equal to the preset threshold, the initial model is retrained until it is less than the preset threshold. The measurement samples can be obtained through hardware devices such as detectors to acquire the feature signals of the object under test.

[0078] In some embodiments, determining the predicted feature signal corresponding to the predicted key size based on the predicted key size includes:

[0079] The predicted key dimensions are input into the prediction model to obtain the predicted feature signals corresponding to the predicted key dimensions. The prediction model is trained by taking the key dimensions of the simulation sample as input and the feature signals of the simulation sample as output.

[0080] Specifically, by inputting the predicted key dimensions into the prediction model, the predicted feature signals of the key dimensions can be obtained. When training the prediction model, the key dimensions of the simulated sample are used as input, and the feature signals of the simulated sample are used as output to train the prediction model. The prediction model can also accelerate the extraction of feature signals. In step 201, when acquiring the feature signals of the object to be measured, the prediction model can also be used to extract the feature signals of the object to be measured from the scattering map of the object.

[0081] In some embodiments, the characteristic signal of the simulated sample is obtained based on the distribution characteristics of the scattered signal in the scattering pattern of the simulated sample and the correlation between the critical size of the simulated sample and the scattering pattern of the simulated sample.

[0082] Specifically, when analyzing and calculating the characteristic signals of the simulated sample, the results are obtained based on the distribution characteristics of the scattered signals in the scattering pattern of the simulated sample, the critical dimensions of the simulated sample, and the correlation of the scattering pattern of the simulated sample. For example... Figure 6 As shown, the top 2000 most obvious feature measurement signals (black dots) found after feature extraction and the fitted simulation signals (gray dots) are shown. Figure 7 The key dimensions of a semiconductor device measured using this method are shown in the x and y directions (left image for x direction, right image for y direction).

[0083] In some embodiments, the feature signal of the measured sample is obtained by inputting the scattering map of the measured sample into the feature extraction model; the feature extraction model is obtained by training the model with the scattering map of the simulated sample as input and the feature signal of the simulated sample as output.

[0084] Specifically, to obtain feature signals more quickly, a feature extraction model can be used. The scattering map of the simulated sample is used as input, and the feature signal of the simulated sample is used as output to train the feature extraction model. Feature extraction models can include: finding the maximum value, non-maximum suppression (NMS), Fourier transform, principal component analysis (PCA), independent component analysis (ICA), correlation analysis, autoencoder networks, etc.

[0085] In some embodiments, the feature extraction model is a model that can perform feature extraction on scattering maps under different azimuth angles and different incident angles, or there are multiple feature extraction models, each of which can perform feature extraction on scattering maps under a single azimuth angle and a single incident angle.

[0086] Specifically, the feature extraction model can be a model that performs feature extraction on scattering maps under different incident angles and azimuth angles, or it can be a functional model that performs feature extraction on scattering maps under one azimuth angle and one incident angle for different azimuth angles and incident angles.

[0087] For example, the feature extraction model takes scattering maps with different incident and azimuth angles as input and the feature signal as output to train the model. Alternatively, the feature extraction model can take the scattering map with azimuth angle x and incident angle 0° as input and the feature signal as output, resulting in Feature Extraction Model 1; the scattering map with azimuth angle x and incident angle 30° as input and the feature signal as output, resulting in Feature Extraction Model 2; the scattering map with azimuth angle y and incident angle 0° as input and the feature signal as output, resulting in Feature Extraction Model 3; and the scattering map with azimuth angle y and incident angle 30° as input and the feature signal as output, resulting in Feature Extraction Model 4. Thus, different feature extraction models are obtained for each azimuth and incident angle.

[0088] To better explain the embodiments of this application, the following flowchart, in conjunction with a specific implementation scenario, describes a method for measuring a key dimension provided by an embodiment of this application, including the following steps: Figure 8 As shown:

[0089] First, let's introduce the model training phase:

[0090] Step 801: Obtain simulation samples under different incident angles and azimuth angles through a simulation program. The number of simulation samples is much larger than the number of measurement samples. The incident angle, azimuth angle and other parameters of the simulation samples are consistent with the parameters of the measurement samples.

[0091] Step 802: Use the feature signals of the simulation samples as input and the key dimensions of the simulation samples as output to train the initial model.

[0092] Step 803: Input the feature signal of the measurement sample into the initial model to obtain the predicted key dimensions of the measurement sample.

[0093] Step 804: The predicted key dimensions of the measured sample are obtained through a simulation program to obtain the predicted feature signal of the measured sample.

[0094] Step 805: Determine whether the loss values ​​of the feature signals and predicted feature signals of the measured sample are less than a preset threshold. If they are less, proceed to step 806; otherwise, proceed to step 802.

[0095] Step 806: Use the initial model as the final target model.

[0096] The following describes the stages of model usage:

[0097] Step 807: Measure the object to be measured using a detector to obtain the scattering pattern of the object.

[0098] Step 808: Analyze and calculate the scattering signal of the scattering pattern of the object to be measured through a simulation program.

[0099] Step 809: Input the scattering signal into the target model to obtain the key dimensions.

[0100] Based on the same technical concept, embodiments of this application provide a measuring device for critical dimensions, such as... Figure 9 As shown, the device 900 includes:

[0101] The acquisition module 901 is used to acquire the scattering pattern of the object to be measured; the scattering pattern of the object to be measured is formed after X-rays are scattered by the object to be measured.

[0102] Extraction module 902 is used to extract the feature signal of the object to be measured from the scattering map of the object to be measured;

[0103] The determination module 903 is used to determine the key dimensions of the object to be measured based on the feature signals of the object to be measured through a target model; wherein the target model is obtained by training an initial model with simulation samples and then retraining the initial model with measurement samples.

[0104] Optionally, the scattering pattern of the object to be measured is formed after X-rays are scattered by the object at different azimuth angles and different incident angles, or the scattering pattern of the object to be measured is formed after X-rays are scattered by the object at the same azimuth angle and different incident angles.

[0105] Optionally, the determining module 903 is specifically used for:

[0106] For any simulation sample, the feature signal of the simulation sample is used as input and the key dimensions of the simulation sample are used as output to train an initial model.

[0107] For any measurement sample, the feature signal of the measurement sample is input into the initial model to obtain the predicted key size of the measurement sample; based on the predicted key size, the predicted feature signal corresponding to the predicted key size is determined; the initial model is adjusted by the loss value between the feature signal of the measurement sample and the predicted feature signal until the training termination condition is met.

[0108] Optionally, the determining module 903 is specifically used for:

[0109] The predicted key dimensions are input into the prediction model to obtain the predicted feature signals corresponding to the predicted key dimensions; the prediction model is trained by taking the key dimensions of the simulation samples as input and the feature signals of the simulation samples as output.

[0110] Optionally, the number of simulated samples is much larger than the number of measured samples.

[0111] Optionally, the feature signal of the simulated sample is obtained based on the distribution characteristics of the scattered signal in the scattering map of the simulated sample and the correlation between the key dimensions of the simulated sample and the scattering map of the simulated sample.

[0112] Optionally, the feature signal of the measured sample is obtained by inputting the scattering map of the measured sample into the feature extraction model; the feature extraction model is obtained by training the model with the scattering map of the simulated sample as input and the feature signal of the simulated sample as output.

[0113] Optionally, the feature extraction model is a model that can perform feature extraction on scattering maps under different azimuth angles and different incident angles, or there are multiple feature extraction models, each of which can perform feature extraction on scattering maps under a single azimuth angle and a single incident angle.

[0114] Based on the same technical concept, embodiments of this application provide a computer device, which may be a terminal or a server, such as... Figure 10 The embodiment includes at least one processor 1001 and a memory 1002 connected to the at least one processor. The specific connection medium between the processor 1001 and the memory 1002 is not limited in this embodiment. Figure 10 Taking the connection between processor 1001 and memory 1002 via a bus as an example, the bus can be divided into address bus, data bus, control bus, etc.

[0115] In this embodiment of the application, the memory 1002 stores instructions that can be executed by at least one processor 1001. By executing the instructions stored in the memory 1002, at least one processor 1001 can perform the steps included in the above-mentioned critical dimension measurement method.

[0116] The processor 1001 is the control center of the computer device, capable of connecting various parts of the computer device via various interfaces and lines. It performs interface aggregation by running or executing instructions stored in the memory 1002 and accessing data stored in the memory 1002. Optionally, the processor 1001 may include one or more processing units. The processor 1001 may integrate an application processor and a modem processor. The application processor primarily handles the operating system, user interface, and applications, while the modem processor primarily handles wireless communication. It is understood that the modem processor may not be integrated into the processor 1001. In some embodiments, the processor 1001 and the memory 1002 may be implemented on the same chip; in other embodiments, they may be implemented on separate chips.

[0117] The processor 1001 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.

[0118] Memory 1002, 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 1002 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 memory, magnetic disk, optical disk, etc. Memory 1002 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 1002 can also be a circuit or any other device capable of implementing storage functions for storing program instructions and / or data.

[0119] Based on the same inventive concept, embodiments of this application provide a computer-readable storage medium storing a computer program executable by a computer device, which, when run on the computer device, causes the computer device to perform the steps of the above-mentioned key dimension measurement method.

[0120] 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.

[0121] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to this application. It should 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 illustrations. Figure 1One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0122] 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.

[0123] 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 1 The steps of the function specified in one or more boxes.

[0124] 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 method for measuring a critical dimension, characterized in that, include: Obtain the scattering map of the object to be measured; The scattering pattern of the object to be measured is formed by small-angle X-rays being scattered by the object at different azimuth angles and different incident angles, or the scattering pattern of the object to be measured is formed by small-angle X-rays being scattered by the object at the same azimuth angle and different incident angles. Extract the feature signal of the object to be measured from the scattering map of the object to be measured; Based on the feature signals of the object to be measured, the key dimensions of the object to be measured are determined by a target model; wherein, the target model is obtained by training an initial model with simulation samples, and then training the initial model again with measurement samples; The target model was trained in the following way: For any simulated sample, the feature signal of the simulated sample is used as input and the key dimensions of the simulated sample are used as output to train an initial model; the feature signal of the simulated sample is obtained based on the distribution characteristics of the scattered signal in the scattering map of the simulated sample and the correlation between the key dimensions of the simulated sample and the scattering map of the simulated sample. For any measurement sample, the feature signal of the measurement sample is input into the initial model to obtain the predicted key dimension of the measurement sample; based on the predicted key dimension, the predicted feature signal corresponding to the predicted key dimension is determined; the initial model is adjusted by the loss value between the feature signal of the measurement sample and the predicted feature signal until the training termination condition is met; the measurement parameters of the simulation sample and the measurement sample are kept consistent, and the structure of the measurement sample determines the number of simulation samples; the feature signal of the measurement sample is obtained by inputting the scattering map of the measurement sample into the feature extraction model; the feature extraction model is obtained by training the model with the scattering map and key dimension of the simulation sample as input and the feature signal of the simulation sample as output.

2. The method as described in claim 1, characterized in that, The step of determining the predicted feature signal corresponding to the predicted key size based on the predicted key size includes: The predicted key dimensions are input into the prediction model to obtain the predicted feature signals corresponding to the predicted key dimensions; the prediction model is trained by taking the key dimensions of the simulation samples as input and the feature signals of the simulation samples as output.

3. The method as described in claim 1, characterized in that, The number of simulated samples is greater than the number of measured samples.

4. The method as described in claim 1, characterized in that, The feature extraction model is a model that can perform feature extraction on scattering maps under different azimuth angles and different incident angles, or there are multiple feature extraction models, each of which can perform feature extraction on scattering maps under a single azimuth angle and a single incident angle.

5. A measuring device for a critical dimension, characterized in that, include: The acquisition module is used to acquire the scattering map of the object to be measured; The scattering pattern of the object to be measured is formed by small-angle X-rays being scattered by the object at different azimuth angles and different incident angles, or the scattering pattern of the object to be measured is formed by small-angle X-rays being scattered by the object at the same azimuth angle and different incident angles. An extraction module is used to extract the feature signals of the object to be measured from the scattering map of the object to be measured; The determination module is used to determine the key dimensions of the object to be measured based on the feature signals of the object to be measured and through a target model; wherein, the target model is obtained by training an initial model with simulation samples and then retraining the initial model with measurement samples; The target model was trained in the following way: For any simulated sample, the feature signal of the simulated sample is used as input and the key dimensions of the simulated sample are used as output to train an initial model; the feature signal of the simulated sample is obtained based on the distribution characteristics of the scattered signal in the scattering map of the simulated sample and the correlation between the key dimensions of the simulated sample and the scattering map of the simulated sample. For any measurement sample, the feature signal of the measurement sample is input into the initial model to obtain the predicted key dimension of the measurement sample; based on the predicted key dimension, the predicted feature signal corresponding to the predicted key dimension is determined; the initial model is adjusted by the loss value between the feature signal of the measurement sample and the predicted feature signal until the training termination condition is met; the measurement parameters of the simulation sample and the measurement sample are kept consistent, and the structure of the measurement sample determines the number of simulation samples; the feature signal of the measurement sample is obtained by inputting the scattering map of the measurement sample into the feature extraction model; the feature extraction model is obtained by training the model with the scattering map and key dimension of the simulation sample as input and the feature signal of the simulation sample as output.

6. A computer-readable storage medium, characterized in that, It stores a computer program executable by a computer device, which, when run on the computer device, causes the computer device to perform the steps of the method according to any one of claims 1 to 4.

Images