Exposure parameter determination method and device, photoetching method and device, storage medium and equipment

Abstract

The invention discloses an exposure parameter determination method, a photoetching method, a photoetching device, a storage medium and equipment. The exposure parameter determination method comprises the following steps: acquiring a focal length energy matrix of a wafer; exposing the wafer by using the obtained focal length energy matrix to obtain a CD measurement image corresponding to each exposure parameter combination in the focal length energy matrix; and inputting each exposure parameter combination and the CD measurement image corresponding to each exposure parameter combination into a first machine learning model, identifying the obtained CD measurement image by using the first machine learning model, and outputting an optimal exposure parameter combination. By adopting the scheme, the efficiency and accuracy of determining the exposure parameters can be improved.

Description

Technical Field

[0001] The present invention relates to the field of semiconductor technology, and particularly to a method for determining exposure parameters, a lithography method, an apparatus, a storage medium, and a device. Background Art

[0002] In the lithography process, it is often necessary to perform a Focus-Exposure Matrix to determine the optimal exposure parameters, such as the focal length and exposure energy of the exposure machine, so as to obtain the best image quality during lithography, improve the product yield, and shorten the R & D cycle.

[0003] However, the existing process of determining exposure parameters is inefficient and error-prone. Summary of the Invention

[0004] The problem to be solved by the present invention is: how to improve the efficiency and accuracy of determining exposure parameters.

[0005] To solve the above problem, an embodiment of the present invention provides a method for determining exposure parameters, the method comprising:

[0006] Obtaining a focus-exposure matrix of a wafer;

[0007] Exposing the wafer by using the obtained focus-exposure matrix to obtain CD measurement images corresponding to each exposure parameter combination in the focus-exposure matrix;

[0008] Inputting each exposure parameter combination and the CD measurement image corresponding to each exposure parameter combination into a first machine learning model, and using the first machine learning model to identify the obtained CD measurement images, and outputting an optimal exposure parameter combination.

[0009] In a possible embodiment, the using the first machine learning model to identify the obtained CD measurement images, and outputting an optimal exposure parameter combination, includes:

[0010] Based on the input exposure parameter combination, obtaining a complete process window corresponding to the CD measurement image;

[0011] Using the first machine learning model to identify the obtained CD measurement images, and obtaining a common process window from the complete process window, the common process window being an exposure parameter combination accepted by all CD measurement images on the wafer;

[0012] Based on the common process window, obtaining an optimal exposure parameter combination.

[0013] In a possible embodiment, the based on the common process window, obtaining an optimal exposure parameter combination, includes:

[0014] Obtain the exposure parameter combination corresponding to the CD measurement target value from the focal length energy matrix;

[0015] Determine whether the obtained exposure parameter combination is within the preset area of the common process window;

[0016] When the obtained exposure parameter combination is within the preset area of the common process window, use the obtained exposure parameter combination as the optimal exposure parameter combination; otherwise, the obtained exposure parameter combination is not the optimal exposure parameter combination.

[0017] In a possible embodiment, the preset area is the middle area of the common process window.

[0018] In a possible embodiment, the obtained exposure parameter combination includes: exposure energy and focal length.

[0019] In a possible embodiment, the first machine learning model is trained by the following method:

[0020] Collect different exposure parameter combinations and the CD measurement results corresponding to each exposure parameter combination;

[0021] Use the collected exposure parameter combinations and the CD measurement results corresponding to each exposure parameter combination to train an initial machine learning model to obtain the first machine learning model.

[0022] In a possible embodiment, using the collected exposure parameter combinations and the CD measurement results corresponding to each exposure parameter combination to train an initial machine learning model to obtain the first machine learning model includes:

[0023] Use the collected exposure parameter combinations and the CD measurement results corresponding to each collected exposure parameter combination to train multiple initial machine learning models implemented by different algorithms respectively to obtain multiple trained machine learning models;

[0024] Evaluate the multiple trained initial machine learning models, and obtain the machine learning model with the highest accuracy as the first machine learning model.

[0025] An embodiment of the present invention further provides a lithography method, and the method includes:

[0026] Obtain an optimal exposure parameter combination, and the optimal exposure parameter combination is determined by using any one of the exposure parameter determination methods described above;

[0027] Based on the optimal exposure parameter combination, expose the photoresist layer on the substrate to form a corresponding lithography pattern on the substrate;

[0028] Etch the positions on the substrate that are not covered by the lithography pattern, and transfer the lithography pattern onto the substrate.

[0029] An embodiment of the present invention further provides an exposure parameter determination device, which includes:

[0030] An acquisition unit, adapted to acquire the focus energy matrix of a wafer;

[0031] A first exposure unit, adapted to expose the wafer by using the acquired focus energy matrix to obtain CD measurement images corresponding to each exposure parameter combination in the focus energy matrix;

[0032] A determination unit, adapted to input each exposure parameter combination and the CD measurement image corresponding to each exposure parameter combination into a first machine learning model, use the first machine learning model to identify the obtained CD measurement images, and output the optimal exposure parameter combination.

[0033] An embodiment of the present invention further provides a lithography device, which includes:

[0034] A parameter determination unit, adapted to acquire the optimal exposure parameter combination, where the optimal exposure parameter combination is determined by using the above-mentioned exposure parameter determination device;

[0035] A second exposure unit, adapted to expose the photoresist layer on the substrate based on the optimal exposure parameter combination to form a corresponding lithography pattern on the substrate;

[0036] An etching unit, adapted to etch the positions on the substrate that are not covered by the lithography pattern, and transfer the lithography pattern onto the substrate.

[0037] An embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored, and the computer program is executed by a processor to implement the steps of any one of the above methods.

[0038] An embodiment of the present invention further provides an electronic device, including a memory and a processor, where a computer program capable of running on the processor is stored on the memory, and when the processor runs the computer program, it executes the steps of any one of the above methods.

[0039] Compared with the prior art, the technical solution of the embodiment of the present invention has the following advantages:

[0040] After obtaining the CD measurement results corresponding to each exposure parameter combination in the focal length energy matrix by applying the solution of the present invention, the first machine learning model is used to identify the obtained CD measurement results to determine the optimal exposure parameter combination, without the need for manual identification of the CD measurement results, thereby improving the efficiency of determining the exposure parameters. Moreover, by improving the performance of the first machine learning model, the accuracy of determining the exposure parameters can be significantly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 is a flowchart of a method for determining exposure parameters in an embodiment of the present invention;

[0042] Figure 2 is a flowchart of a method for training a first machine learning model in an embodiment of the present invention;

[0043] Figure 3 is a flowchart of a lithography method in an embodiment of the present invention;

[0044] Figures 4 to 6 is a schematic cross-sectional structure diagram of a semiconductor structure during lithography in an embodiment of the present invention;

[0045] Figure 7 is a schematic structural diagram of a device for determining exposure parameters in an embodiment of the present invention;

[0046] Figure 8 is a schematic structural diagram of a lithography device in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] In the semiconductor field, the focal length energy matrix is a tool used in lithography technology, mainly for testing the lithography process window and determining the optimal exposure parameters. Specifically, the focal length energy matrix is a set of data obtained by changing the focal length value in one direction with a fixed step length and changing the exposure energy in the other direction with a fixed step length during exposure. On the same wafer, the exposure parameters of each exposure area (shot) may be different. One part is the exposure energy step of each column of exposure areas, and the other part is the focal length step of each row of exposure areas, finally forming the focal length energy matrix of the focal length and exposure energy.

[0048] After the wafer with the focal length energy matrix is exposed, it is necessary to measure the critical dimension (CD) values (i.e., CD values) at the critical positions of the obtained critical dimension (CD) images (hereinafter referred to as CD measurement images), export the CD measurement data and CD measurement images, and then manually determine the optimal exposure parameters according to the CD measurement data and CD images.

[0049] However, the method of determining the optimal exposure parameters based on manual CD measurement data and CD images is inefficient and prone to overlooking the details of the CD measurement images, resulting in non-optimal exposure parameters.

[0050] To address this problem, the present invention provides a method for determining exposure parameters for trench-type devices. Using this method, after obtaining the CD measurement results corresponding to each exposure parameter combination in the focal length-energy matrix, the first machine learning model is used to identify the obtained CD measurement results to determine the optimal exposure parameter combination. Compared with manual identification of CD measurement results, it can effectively improve the efficiency and accuracy of determining exposure parameters.

[0051] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following detailed description of specific embodiments of the present invention will be given with reference to the accompanying drawings.

[0052] Refer to Figure 1 , an embodiment of the present invention provides a method for determining exposure parameters, and the method may include the following steps:

[0053] Step 11, obtain the focal length-energy matrix of the wafer.

[0054] In a specific implementation, the focal length-energy matrix is a set of data obtained by changing the focal length value in one direction with a fixed step length and changing the exposure energy in the other direction with a fixed step length during exposure. On the same wafer, the exposure parameters of each exposure area may be different. One part is the exposure energy step of each column of exposure areas, and the other part is the focal length step of each row of exposure areas, finally forming the focal length-energy matrix of focal length and energy.

[0055] Step 12, use the obtained focal length-energy matrix to expose the wafer to obtain CD measurement images corresponding to each exposure parameter combination in the focal length-energy matrix.

[0056] In a specific implementation, there are multiple exposure areas on the same wafer. The exposure parameter combinations corresponding to different exposure areas in the focal length-energy matrix can be used to expose the exposure area to obtain the corresponding CD measurement images. For example, when there are 10 exposure areas on the same wafer, each exposure area has an exposure parameter combination. At this time, there are 10 exposure parameter combinations in the obtained focal length-energy matrix, and one exposure parameter combination corresponds to one exposure area on the wafer, and the exposure area is exposed using the exposure parameter combination corresponding to the exposure area. After each exposure area on the wafer is exposed, the corresponding CD measurement images can be obtained.

[0057] Step 13: Input each exposure parameter combination and the CD measurement image corresponding to each exposure parameter combination into the first machine learning model. Use the first machine learning model to identify the obtained CD measurement image and output the optimal exposure parameter combination.

[0058] In an embodiment of the present invention, using the first machine learning model to identify the obtained CD measurement image and output the optimal exposure parameter combination may include: obtaining the complete process window corresponding to the CD measurement image based on the input exposure parameter combination; using the first machine learning model to identify the obtained CD measurement image, and obtaining a common process window from the complete process window, where the common process window is the exposure parameter combination accepted by all CD measurement images on the wafer; and obtaining the optimal exposure parameter combination based on the common process window.

[0059] In a specific implementation, the machine learning model may also output the complete process window where the CD measurement image is located, that is, the exposure parameter interval corresponding to the CD measurement image, and determine the common process window from this complete process window.

[0060] Specifically, the first machine learning model can automatically identify the quality of the critical dimension image, that is, automatically identify whether there are process problems such as residual glue, white edges, under-exposure or over-exposure in the critical dimension image. For different critical dimension images, their critical positions and corresponding critical dimension values may be different. The first machine learning model can also determine the critical position of the identified critical dimension image and measure the dimension of the critical position of the critical dimension image. That is to say, the first machine learning model can identify and obtain the critical dimension value of the critical dimension image based on the critical dimension image.

[0061] Further, after identifying the CD measurement image, the machine learning model can establish a mapping relationship between the CD measurement image and the exposure parameters, form a complete process window, and be able to determine the optimal parameter interval where the exposure parameter combination that meets the production requirements is located from this complete process window as the common process window. The so-called common process window is also the exposure parameter interval that can make all CD measurement images on the wafer meet the production requirements. Subsequently, the optimal exposure parameter combination can be obtained based on the common process window, that is, select a set of exposure parameters from the common process window as the optimal exposure parameter combination.

[0062] In a specific implementation, based on the common process window, multiple methods can be used to obtain the optimal exposure parameter combination.

[0063] In one embodiment, the exposure parameter combination corresponding to the CD measurement target value can be obtained from the focal length energy matrix first, and then it is determined whether the obtained exposure parameter combination is within the preset area of the common process window. When the obtained exposure parameter combination is within the preset area of the common process window, the obtained exposure parameter combination is used as the optimal exposure parameter combination; otherwise, the obtained exposure parameter combination is not the optimal exposure parameter combination.

[0064] In one embodiment, the preset area is the middle area of the common process window.

[0065] Specifically, the CD measurement target value may include the target image quality and the target CD value. The exposure parameter combination that satisfies the CD measurement target value in the focal length energy matrix is used as the candidate exposure parameter combination. If the candidate exposure parameter combination is located in the middle area of the common process window, the candidate exposure parameter combination can be used as the exposure parameter combination; otherwise, the candidate exposure parameter combination can be re-determined.

[0066] In one embodiment, the exposure parameter combination in the very middle of the common process window can also be directly used as the candidate exposure parameter combination.

[0067] Through the first machine learning model, the optimal exposure parameter combination can be quickly obtained, thereby improving the efficiency and accuracy of determining the optimal exposure parameter combination.

[0068] Referring to Figure 2 , the embodiment of the present invention also provides a training method for a first machine learning model. The method may include the following steps:

[0069] Step 21, collect different exposure parameter combinations and the CD measurement results corresponding to each exposure parameter combination.

[0070] In a specific implementation, the CD measurement results may include: the quality of the critical dimension measurement image, and the critical dimension value of the critical dimension measurement image. Among them, the quality of the critical dimension measurement image includes: whether there are process problems such as residual glue, white edges, under-exposure or over-exposure in the critical dimension measurement image. The critical dimension value of the critical dimension measurement image is the dimension value of the critical dimension image at the critical position.

[0071] During the data collection process, the CD measurement results such as the line width and pattern quality of the photoresist pattern under the exposure energy and focal length settings can be collected. Each group of CD measurement results corresponds to a specific combination of exposure energy (E) and focal length (F). The CD measurement result and the corresponding exposure energy and focal length combination are used as a training sample.

[0072] Step 22: Use the collected exposure parameter combinations and the CD measurement results corresponding to each exposure parameter combination to train the initial machine learning model to obtain the first machine learning model.

[0073] In some embodiments, before training the initial machine learning model, it is also possible to label and classify the quality of the critical dimension measurement images in the collected CD measurement results. For example, it can be determined which critical dimension measurement images in the critical dimension measurement images have residual glue, which have white edges, and which have insufficient exposure, etc. This can facilitate the initial machine learning model to accurately identify images.

[0074] In some embodiments, after collecting the data, it is also possible to preprocess the collected data to achieve data augmentation. Specifically, the preprocessing can include data cleaning (such as removing outliers and filling missing values), standardization (to ensure that each feature is on the same scale), normalization (adjusting the data distribution to a specific interval), etc. High-quality data is the basis for the success of machine learning. Preprocessing can improve the efficiency and effectiveness of model training and avoid biases or overfitting caused by data quality problems. Use the preprocessed data to train the initial machine learning model.

[0075] In some embodiments, the original exposure parameter combination data collected may also include other factors that may affect the lithography result, such as photoresist characteristics, mask design, environmental parameters, etc. Subsequently, feature data that has a greater impact on the CD measurement result can be extracted from the original exposure parameter combination data, such as exposure energy, focal length, photoresist type, wafer material, etc. These features will be used as the input of the initial machine learning model. The initial machine learning model can determine the corresponding critical dimension measurement image appearance and critical dimension position based on the photoresist type and wafer material. The initial machine learning model can establish a mapping relationship between the exposure parameters and the CD measurement results based on the exposure energy and focal length.

[0076] In specific implementation, before training the initial machine learning model, it is also possible to create new derivative feature data based on the extracted exposure parameter feature data. For example, interaction terms (E×F) or feature data based on physical models can be created to better represent complex interactions, thereby reducing the computational burden, improving the interpretability of the first machine learning model, and at the same time increasing the sensitivity of the first machine learning model to the key feature data.

[0077] In specific implementation, a suitable machine learning model is selected according to the characteristics of the problem and the data, such as machine learning models like linear regression, support vector machine, random forest, neural network, etc. Different models have different advantages. To improve the accuracy of the first machine learning model, multiple initial machine learning models can be established using different algorithms respectively, and then the multiple initial machine learning models implemented by different algorithms are respectively trained using the collected training samples to obtain multiple trained machine learning models. Subsequently, the multiple trained initial machine learning models can be evaluated to obtain the machine learning model with the highest accuracy as the first machine learning model.

[0078] In specific implementation, the goal of training the initial machine learning model is: to enable the initial machine learning model to learn how exposure energy and focal length affect the critical dimension and quality of the final critical dimension measurement image, thereby establishing a mapping relationship between exposure energy and focal length and the critical dimension and quality of the critical dimension measurement image. Among them, the critical dimension of the specific critical dimension measurement image mainly refers to the line width of the pattern. Therefore, the goal of training the initial machine learning model can be expressed as: f(E, F) → line width + pattern quality.

[0079] In specific implementation, the corresponding loss function can be designed in advance to optimize the initial machine learning model during training. When training the initial machine learning model, the cross-validation method can also be used to split the collected data to train and validate the initial machine learning model, thereby adjusting the parameters of the initial machine learning model, which can ensure that the finally obtained first machine learning model can achieve the best generalization ability, neither overfitting nor underfitting.

[0080] In specific implementation, after the training of the initial machine learning model is completed, suitable evaluation metrics (such as mean squared error, R² score) can be used to measure the performance of the trained initial machine learning model, and the hyperparameters of the initial machine learning model can be further adjusted through methods such as grid search, random search, or Bayesian optimization to find the optimal configuration, so that the accuracy and performance of each trained initial machine learning model are the best, thereby improving the accuracy and reliability of the prediction of the finally obtained first machine learning model.

[0081] In specific implementation, the data in the actual production process can also be collected in real time, and the first machine learning model can be updated according to the collected data, and the model can be iteratively optimized, which can ensure that the first machine learning model can remain effective with the change of the process. Thereby, the first machine learning model can adapt to the change of the production environment, continuously optimize the process, reduce the defective rate, and improve the production efficiency.

[0082] The finally obtained first machine learning model can not only significantly accelerate the exploration process of the energy focus matrix, but also maintain a high degree of process control accuracy in a complex and changeable production environment, thereby improving the overall efficiency and product quality of semiconductor manufacturing.

[0083] Referring to Figure 3 , an embodiment of the present invention further provides a lithography method, and the method may include the following steps:

[0084] Step 31, obtaining an optimal exposure parameter combination, where the optimal exposure parameter combination is determined by using any one of the above-described exposure parameter determination methods.

[0085] Specifically, referring to the above description of the exposure parameter determination method, an optimal exposure parameter combination can be obtained, and the optimal exposure parameter combination includes the optimal exposure energy and the focal length.

[0086] Step 32, based on the optimal exposure parameter combination, exposing the photoresist layer on the substrate to form a corresponding lithography pattern on the substrate.

[0087] In a specific implementation, referring to Figure 4 , a photoresist can be uniformly coated on the substrate 41 to form a photoresist layer 42 on the substrate 41. Align the mask 43 with the substrate 41, and then use a light source with a specific wavelength (such as ultraviolet light, deep ultraviolet light, or extreme ultraviolet light) through a lithography machine to expose the photoresist layer 42 on the substrate 41. During the exposure process, the lens group of the lithography machine will transfer the pattern of the mask 43 to the photoresist layer 42 to form a patterned photoresist layer. Referring to Figure 5 , a subsequent development operation can be performed to form a corresponding lithography pattern 44 on the substrate 41.

[0088] Step 33, etching the positions on the substrate that are not covered by the lithography pattern to transfer the lithography pattern to the substrate.

[0089] Specifically, referring to Figure 6 , the area on the substrate 41 that is not covered by the lithography pattern 44 can be etched, whereby the lithography pattern 44 can be transferred to the substrate 41.

[0090] It should be noted that Figures 4 to 6 is only an example of the lithography process. After determining the optimal exposure parameter combination by using the exposure parameter determination method in the embodiment of the present invention, lithography operations can also be performed on other semiconductor structures, and no further examples will be given here.

[0091] As can be seen from the above, the solution in the embodiment of the present invention significantly reduces the workload of lithography process personnel, improves efficiency, reduces costs, enhances chip reliability and performance, and improves the user experience through machine learning-assisted deployment and application of the energy focus matrix.

[0092] To enable those skilled in the art to better understand and implement the present invention, the apparatuses, test systems, electronic devices, and computer-readable storage media corresponding to the above methods are described in detail below.

[0093] Referring to Figure 7 , an exposure parameter determination apparatus 70 is further provided in an embodiment of the present invention. The exposure parameter determination apparatus 70 may include: an acquisition unit 71, a first exposure unit 72, and a determination unit 73. Among them:

[0094] The acquisition unit 71 is adapted to acquire a focal length energy matrix of a wafer;

[0095] The first exposure unit 72 is adapted to expose the wafer by using the acquired focal length energy matrix to obtain CD measurement images corresponding to each exposure parameter combination in the focal length energy matrix;

[0096] The determination unit 73 is adapted to input each exposure parameter combination and the CD measurement image corresponding to each exposure parameter combination into a first machine learning model, use the first machine learning model to identify the obtained CD measurement image, and output the optimal exposure parameter combination.

[0097] Regarding the acquisition unit 71, the first exposure unit 72, and the determination unit 73, specific implementation may refer to the description of the exposure parameter determination method above, and details are not described herein again.

[0098] Referring to Figure 8 , an optical lithography apparatus 80 is further provided in an embodiment of the present invention. The optical lithography apparatus 80 may include: a parameter determination unit 81, a second exposure unit 82, and an etching unit 83. Among them:

[0099] The parameter determination unit 81 is adapted to acquire the optimal exposure parameter combination, which is determined by using the exposure parameter determination apparatus described in claim 8;

[0100] The second exposure unit 82 is adapted to expose a photoresist layer on a substrate based on the optimal exposure parameter combination to form a corresponding lithography pattern on the substrate;

[0101] The etching unit 83 is adapted to etch positions on the substrate that are not covered by the lithography pattern, and transfer the lithography pattern to the substrate.

[0102] Regarding the parameter determination unit 81, the second exposure unit 82, and the etching unit 83, specific implementation may refer to the description of the optical lithography method above, and details are not described herein again.

[0103] An embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored, and the computer program is executed by a processor to implement the steps of any of the above methods.

[0104] In a specific implementation, the computer-readable storage medium may include: ROM, RAM, a magnetic disk, an optical disc, etc.

[0105] An embodiment of the present invention further provides an electronic device, the electronic device includes a memory and a processor, a computer program capable of running on the processor is stored on the memory, and when the processor runs the computer program, it executes the steps of any of the above methods.

[0106] Regarding each device and product described in the above embodiments, each module / unit included therein may be a software module / unit, a hardware module / unit, or may be partially a software module / unit and partially a hardware module / unit. For example, for each device and product applied to or integrated into a chip, each module / unit included therein may be implemented in a hardware manner such as a circuit, or at least some of the modules / units may be implemented in a software program manner, and the software program runs on a processor integrated inside the chip, and the remaining (if any) part of the modules / units may be implemented in a hardware manner such as a circuit; for each device and product applied to or integrated into a chip module, each module / unit included therein may be implemented in a hardware manner such as a circuit, and different modules / units may be located in the same component (such as a chip, a circuit module, etc.) or different components of the chip module, or at least some of the modules / units may be implemented in a software program manner, and the software program runs on a processor integrated inside the chip module, and the remaining (if any) part of the modules / units may be implemented in a hardware manner such as a circuit; for each device and product applied to or integrated into a terminal, each module / unit included therein may be implemented in a hardware manner such as a circuit, and different modules / units may be located in the same component (such as a chip, a circuit module, etc.) or different components inside the terminal, or at least some of the modules / units may be implemented in a software program manner, and the software program runs on a processor integrated inside the terminal, and the remaining (if any) part of the modules / units may be implemented in a hardware manner such as a circuit.

[0107] Although the present invention is disclosed as above, the present invention is not limited thereto. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention should be subject to the scope defined by the claims.

Claims

1. A method for determining exposure parameters, characterized in that: include: Obtaining the focal energy matrix of the wafer; Exposing the wafer using the acquired focal length energy matrix to obtain a CD measurement image corresponding to each exposure parameter combination in the focal length energy matrix; Each exposure parameter combination and the CD measurement image corresponding to each exposure parameter combination are input into a first machine learning model, the first machine learning model is used to identify the obtained CD measurement image, and an optimal exposure parameter combination is output.

2. The exposure parameter determination method according to claim 1, characterized in that: The step of using the first machine learning model to identify the obtained CD measurement image and outputting an optimal exposure parameter combination includes: Based on the input exposure parameter combination, a complete process window corresponding to the CD measurement image is obtained; Using the first machine learning model to identify the obtained CD measurement image, a common process window is obtained from the complete process window, and the common process window is an exposure parameter combination accepted by all CD measurement images on the wafer; Based on the common process window, an optimal exposure parameter combination is obtained.

3. The exposure parameter determination method according to claim 2, characterized in that: The step of obtaining an optimal exposure parameter combination based on the common process window includes: Obtaining an exposure parameter combination corresponding to a CD measurement target value from the focal length energy matrix; Determining whether the acquired exposure parameter combination is within a preset area of ​​the common process window; When the acquired exposure parameter combination is within a preset area of ​​the common process window, the acquired exposure parameter combination is taken as the optimal exposure parameter combination; otherwise, the acquired exposure parameter combination is not the optimal exposure parameter combination.

4. The exposure parameter determination method according to claim 3, characterized in that: The preset area is a middle area of ​​the common process window.

5. The exposure parameter determination method according to claim 3, characterized in that: The acquired exposure parameter combination includes: exposure energy and focal length.

6. The exposure parameter determination method according to claim 1, characterized in that: The first machine learning model is trained using the following method: Collect different exposure parameter combinations and the CD measurement results corresponding to each exposure parameter combination; The initial machine learning model is trained using the collected exposure parameter combinations and the CD measurement results corresponding to each exposure parameter combination to obtain the first machine learning model.

7. The exposure parameter determination method according to claim 6, characterized in that: The initial machine learning model is trained using the collected exposure parameter combinations and the CD measurement results corresponding to each exposure parameter combination to obtain the first machine learning model, including: Using the collected exposure parameter combinations and the CD measurement results corresponding to each of the collected exposure parameter combinations, multiple initial machine learning models implemented by different algorithms are trained respectively to obtain multiple trained machine learning models; The multiple trained initial machine learning models are evaluated to obtain the machine learning model with the highest accuracy as the first machine learning model.

8. A photolithography method, characterized in that: include: Obtaining an optimal exposure parameter combination, wherein the optimal exposure parameter combination is determined by using the exposure parameter determination method according to any one of claims 1 to 7; Based on the optimal exposure parameter combination, exposing the photoresist layer on the substrate to form a corresponding photolithography pattern on the substrate; The positions on the substrate not covered by the photolithography pattern are etched to transfer the photolithography pattern to the substrate.

9. An exposure parameter determination device, characterized in that: include: An acquisition unit, adapted to acquire a focal energy matrix of a wafer; A first exposure unit, adapted to expose the wafer using the acquired focal length energy matrix to obtain a CD measurement image corresponding to each exposure parameter combination in the focal length energy matrix; The determination unit is adapted to input each exposure parameter combination and the CD measurement image corresponding to each exposure parameter combination into a first machine learning model, use the first machine learning model to identify the obtained CD measurement image, and output an optimal exposure parameter combination.

10. A photolithography apparatus, characterized in that: include: a parameter determination unit, adapted to obtain an optimal exposure parameter combination, wherein the optimal exposure parameter combination is determined by using the exposure parameter determination device according to claim 8; A second exposure unit, adapted to expose the photoresist layer on the substrate based on an optimal exposure parameter combination to form a corresponding photolithography pattern on the substrate; The etching unit is adapted to etch the position on the substrate not covered by the photolithography pattern, so as to transfer the photolithography pattern to the substrate.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that: The computer program is executed by a processor to implement the steps of the method according to any one of claims 1 to 8.

12. An electronic device comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, wherein: When the processor runs the computer program, the steps of the method according to any one of claims 1 to 8 are performed.

Images