Etching prediction method, device, equipment and medium

Abstract

The invention provides an etching prediction method and device, equipment and a medium. Etching process parameters and multiple pieces of first etching contour data obtained through etching under the etching process parameters can be obtained; inputting the etching process parameters and the plurality of pieces of first etching contour data into an etching contour prediction model, and outputting second etching contour data at a second moment; the second moment is after the first moment; the etching contour prediction model is determined based on a long short-term memory network and a cascade combination layer, is a cascade recurrent neural network, can reduce a large amount of calculation of modeling based on a complex physical mechanism in a traditional etching model, reduces repeated manual trial and error work in calibration of the traditional etching model, reduces modeling time, and improves simulation efficiency. And the prediction accuracy of the etching contour data is improved.

Description

Technical Field

[0001] The present application relates to the field of semiconductors, and particularly to an etching prediction method, device, equipment and medium. Background Art

[0002] Modeling and prediction of the etching process is a complex and difficult problem to solve. Traditional prediction methods for the etching process mainly use the original physical mechanisms and chemical reactions for modeling. Generally, the Monte Carlo model is used to simulate the particle trajectories and chemical reactions in the etching process. Although the model is already quite accurate, it still takes a lot of time to simulate a single etching process. For complex physical mechanisms, a large amount of calculation is required. In addition, model calibration also requires repeated manual trial and error, resulting in a high time cost.

[0003] As the process size decreases and the circuit scale increases, the complexity of the calculation multiplies, resulting in lower efficiency. Modeling of a single-time etching profile also cannot reflect the temporal variation of the entire etching process. Therefore, providing a suitable etching prediction method has become an urgent technical problem to be solved currently. Summary of the Invention

[0004] In view of this, the purpose of the present application is to provide an etching prediction method, device, equipment and medium, which reduces the modeling time, improves the simulation efficiency, and improves the prediction accuracy of the etching profile data. The specific solutions are as follows:

[0005] On the one hand, the present application provides an etching prediction method, including:

[0006] Obtain etching process parameters and a plurality of first etching profile data obtained by etching under the etching process parameters, each of the first etching profile data corresponding to each first moment;

[0007] Input the etching process parameters and the plurality of first etching profile data into an etching profile prediction model, and output second etching profile data at a second moment; the second moment is after the first moment; the etching profile prediction model is determined based on a long short-term memory network and a cascade combination layer.

[0008] Specifically, the training process of the etching profile prediction model includes:

[0009] Obtain a first etching profile data set; the first etching profile data set includes a plurality of historical etching profile data, and the plurality of historical etching profile data correspond one-to-one with a plurality of historical moments;

[0010] Based on the first etching profile data set, use the long short-term memory network for modeling to obtain an initial etching profile prediction model;

[0011] Based on the first etching profile dataset and the historical etching process parameters corresponding to the first etching profile dataset, the cascade combination layer is used to train the initial etching profile prediction model to obtain the trained etching profile prediction model.

[0012] Specifically, the method of using the cascade combination layer to train the initial etching profile prediction model based on the first etching profile dataset and the historical etching process parameters corresponding to the first etching profile dataset to obtain the trained etching profile prediction model includes:

[0013] Based on the first etching profile dataset and the historical etching process parameters corresponding to the first etching profile dataset, the cascade combination layer is used to train the initial etching profile prediction model to obtain the target etching profile prediction model;

[0014] Obtain the actual etching profile obtained by process means and the actual etching process parameters corresponding to the actual etching profile;

[0015] Use the actual etching profile and the actual etching process parameters to adjust the target etching profile prediction model to obtain the etching profile prediction model.

[0016] Specifically, the method of using the cascade combination layer to train the initial etching profile prediction model based on the first etching profile dataset and the historical etching process parameters corresponding to the first etching profile dataset to obtain the trained etching profile prediction model includes:

[0017] Perform splicing and fully connected operations on the historical etching profile data and the historical etching process parameters to obtain the trained etching profile prediction model.

[0018] Specifically, the method of obtaining the first etching profile dataset includes:

[0019] Generate the first etching profile dataset through simulation software.

[0020] Specifically, the method of generating the first etching profile dataset through simulation software includes:

[0021] Generate a second etching profile dataset through simulation software;

[0022] Perform normalization processing on the second etching profile dataset to obtain the first etching profile dataset.

[0023] Specifically, the cascade combination layer includes multiple splicing layers and multiple fully connected layers.

[0024] In another aspect, an embodiment of the present application further provides an etching prediction device, including:

[0025] An acquisition unit, configured to acquire etching process parameters and a plurality of first etching profile data obtained by etching under the etching process parameters, each of the first etching profile data corresponding to each first moment;

[0026] A prediction unit, configured to input the etching process parameters and the plurality of first etching profile data into an etching profile prediction model, and output second etching profile data at a second moment; the second moment is after the first moment; the etching profile prediction model is determined based on a long short-term memory network and a cascade combination layer.

[0027] In another aspect, an embodiment of the present application provides a computer device, which includes a processor and a memory:

[0028] The memory is used to store program codes and transmit the program codes to the processor;

[0029] The processor is configured to execute the method described in the above aspect according to the instructions in the program code.

[0030] In another aspect, an embodiment of the present application provides a computer-readable storage medium, which is used to store a computer program, and the computer program is used to execute the method described in the above aspect.

[0031] An embodiment of the present application provides an etching prediction method, device, equipment and medium, which can acquire etching process parameters and a plurality of first etching profile data obtained by etching under the etching process parameters, each of the first etching profile data corresponding to each first moment; the plurality of first etching profile data can reflect the profile change trend during the etching process, input the etching process parameters and the plurality of first etching profile data into an etching profile prediction model, and output second etching profile data at a second moment; the second moment is after the first moment; the etching profile prediction model is determined based on a long short-term memory network and a cascade combination layer, and is a cascade recurrent neural network, which can reduce a large amount of calculations based on complex physical mechanisms in traditional etching models, reduce the repetitive manual trial-and-error work in traditional etching model calibration, reduce the modeling time, improve the simulation efficiency, and in addition, improve the prediction accuracy of etching profile data. Description of the Drawings

[0032] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0033] Figure 1 The flowchart of an etching prediction method provided by an embodiment of the present application is shown;

[0034] Figure 2 The cross-sectional view of an etched trench provided by an embodiment of the present application is shown;

[0035] Figure 3 The structural diagram of an etching profile prediction model provided by an embodiment of the present application is shown;

[0036] Figure 4 The structural diagram of another etching profile prediction model provided by an embodiment of the present application is shown;

[0037] Figure 5 The structural block diagram of an etching prediction device provided by an embodiment of the present application;

[0038] Figure 6 The structural diagram of a computer device provided by an embodiment of the present application. Detailed implementation manners

[0039] To make the above objects, features, and advantages of the present application more obvious and understandable, the following will describe the detailed implementation manners of the present application in conjunction with the accompanying drawings.

[0040] Many specific details are set forth in the following description to facilitate a thorough understanding of the present application, but the present application may be implemented in other ways different from those described herein. Those skilled in the art can make similar generalizations without departing from the connotation of the present application, so the present application is not limited by the specific embodiments disclosed below.

[0041] Secondly, the present application will be described in detail with reference to the schematic diagrams. When describing the embodiments of the present application in detail, for the convenience of explanation, the cross-sectional views showing the device structure will be enlarged locally out of the general scale, and the schematic diagrams are only examples and should not limit the scope of protection of the present application herein. In addition, in actual production, three-dimensional spatial dimensions including length, width, and depth should be included.

[0042] For the convenience of understanding, the following will describe in detail an etching prediction method, device, equipment, and medium provided by an embodiment of the present application in conjunction with the accompanying drawings.

[0043] Referring to Figure 1 as shown, the flowchart of an etching prediction method provided by an embodiment of the present application is shown, and the method may include the following steps.

[0044] S101, obtain etching process parameters and a plurality of first etching profile data obtained by etching under the etching process parameters, and each first etching profile data corresponds to each first moment.

[0045] In the embodiments of the present application, the etching process parameters can be various environmental parameters during the etching process, parameters that can have a certain impact on the etching result, such as the power, voltage, pressure or temperature in the chamber applied during etching. The etching process parameters can be obtained. Generally, the etching process parameters do not change during the etching process. The etching process parameters can include at least one of the etching power, etching voltage, etching pressure or etching temperature. For example, they can include the etching power, etching voltage, etching pressure and etching temperature.

[0046] The etching profile data can be the profile data of the cross-section of the etched trench. A polar coordinate system can be established for the trench. Refer to Figure 2 As shown, it is a schematic cross-sectional view of an etched trench provided by the embodiments of the present application, including the structure to be etched 101 and the mask layer 102 located above the structure to be etched. There is an etched trench in the structure to be etched 101. A polar coordinate system can be established with the center point on the surface of the etched trench as the origin O. At each etching moment, a set of etching profile data corresponds. The distance between any point P on the contour of the trench formed by etching and the origin O can be denoted as d, the angle between the line PO and the trench surface can be denoted as θ, and the point P is denoted as (d, θ).

[0047] Specifically, the entire etching process can be divided into multiple stages according to time. One stage corresponds to one moment. For example, it can be divided into 6 stages, corresponding to 0, 1, 2, 3, 4, 5 in the figure. For each stage, it can be represented by multiple polar coordinate points. For example, a point is taken every 10°. Then the etching profile data corresponding to each moment includes 19 polar coordinate points. In order to make the etching profile data more truly reflect the shape of the etched cross-section, a point can also be taken every 1°. The etching profile data of each stage is represented by 181 polar coordinate points. The distance d at each angle of the etching profile is evenly sampled. In this way, the etching profile at each moment can be discretized into a multi-dimensional vector, and the multiple profiles recorded during the entire etching process can be considered as a multivariate time series. For the convenience of expression, the etching profile data at the first moment can be denoted as the first etching profile data, and the first etching profile data can include multiple polar coordinate points.

[0048] Specifically, under the action of etching process parameters, multiple first etching profile data can be obtained. The multiple first etching profile data can reflect the profile change trend during the etching process. Each first etching profile data corresponds to a first moment. That is to say, multiple first etching profile data at multiple moments can be obtained as a set of etching profiles to predict the etching profile data at the next moment. For example, a set of etching profiles can include the etching profile data corresponding to each second (i.e., the first moment) within the time period from the 1st second to the 10th second, a total of 10 etching profile data. Subsequently, the etching profile data at the 11th second (i.e., the second moment) can be predicted based on these 10 etching profile data.

[0049] S102, input the etching process parameters and multiple first etching profile data into the etching profile prediction model, and output the second etching profile data at the second moment; the second moment is after the first moment; the etching profile prediction model is determined based on a long short-term memory network and a cascade combination layer.

[0050] In the embodiment of the present application, an etching profile prediction model can be determined based on a long short-term memory (LSTM) network and a cascade combination layer. The LSTM network is a variant model of a recurrent neural network (RNN). RNN is a type of recurrent neural network that takes sequence data as input, recurses in the evolution direction of the sequence, and all nodes (recurrent units) are connected in a chain. The LSTM network is commonly used in natural language processing tasks and can also be used in time series prediction. The LSTM introduces gate structures on the basis of the RNN, including an input gate, an output gate, and a forget gate, which alleviates the problems of gradient disappearance and gradient explosion existing in the RNN.

[0051] Specifically, the etching profile prediction model can be expressed as f(x p ,x e ) = y. f is the mapping relationship represented by the network, including two inputs x p and x e . x p is the historical etching profile data, x e is the etching process parameter, and y is the predicted etching profile data.

[0052] Taking the etching process parameters and multiple first etching profile data as the input of the etching profile prediction model, the second etching profile data at the second moment is output, where the second moment is a moment after the first moment. Among them, the etching environment at the second moment and the first moment has not changed, that is, the etching process parameters at the second moment are the same as those at the first moment. Therefore, based on the known etching profiles and etching process parameters of the trenches at multiple first moments, the etching profile of the trenches at the second moment can be predicted, that is, the second etching profile data can be obtained. Of course, the second moment can be one or multiple, that is, multiple second etching profile data can be predicted. For example, based on the etching profile data in the time period from the 1st second to the 10th second, the etching profile data from the 11th second to the 15th second can be predicted, and a set of etching profile data can be obtained to achieve accurate prediction of the etching profile at future moments and pre-understand the shape and size of the trenches that may be formed. In addition, by calculating the difference between the second etching profile data and the first etching profile data, the etching amount in the time period from the first moment to the second moment can be obtained. The etching profile prediction model is a cascaded recurrent neural network, which can reduce a large amount of calculations based on complex physical mechanisms in traditional etching models, reduce the repetitive manual trial-and-error work in the calibration of traditional etching models, reduce the modeling time, improve the simulation efficiency, and in addition, improve the prediction accuracy of the etching profile data.

[0053] In the embodiment of the present application, before using the etching profile prediction model for profile prediction, an initial etching profile prediction model can be constructed, and the initial etching profile prediction model is trained using a training set. After the training is completed, the etching profile prediction model is obtained. Next, the model establishment and training process will be described, including steps S201 - S203.

[0054] S201, obtain the first etching profile data set.

[0055] At each historical moment, there is corresponding historical etching profile data. Multiple historical etching profile data at multiple historical moments can be obtained, that is, multiple historical etching profile data correspond to multiple historical moments one by one. The multiple historical etching profile data can be used as the first etching profile data set.

[0056] In a possible implementation manner, the first etching profile data set can be obtained from the etching profiles in the actual etching process, which can improve the accuracy of the first etching profile data set.

[0057] In another possible implementation, a simulation model can be established using simulation software, and historical etching profile data can be obtained through simulation calculations as the first etching profile data set. In this way, a large amount of historical etching profile data can be provided, with a large data volume, enabling a large number of model trainings, improving the accuracy of the etching profile prediction model. Moreover, the actual etching profile is often obtained by cutting and scanning, which is highly destructive to the etching structure and the amount of data obtained is relatively small. Obtaining the first etching profile data set through simulation can reduce the destructiveness to the etching structure.

[0058] Reference Figure 3 As shown, the figure is a schematic structural diagram of an etching profile prediction model provided by an embodiment of the present application, including a preprocessing module, an LSTM module, a cascaded combination layer, and an output module. The preprocessing module can be used for normalization processing, the LSTM module is used to model the etching profile, the cascaded combination layer can integrate etching process parameters into the network structure in a cascaded manner, and the model is trained with etching profile data of different environmental parameters. The output module is used to output the prediction result.

[0059] In actual application, the etching profile data set can be normalized to simplify the calculation. That is, the first etching profile data set can be generated through simulation software. Specifically, the etching profile data set generated by the simulation software is denoted as the second etching profile data set, and the second etching profile data set is normalized to obtain the first etching profile data set.

[0060] Specifically, data normalization can adopt the mean standard deviation method, and the calculation method can be expressed as:

[0061]

[0062] where x i is the historical etching profile data at the i-th historical moment before normalization, x i ′ is the historical etching profile data at the i-th historical moment after normalization processing, μ is the mean of multiple historical etching profile data, and s is the standard deviation of multiple historical etching profile data.

[0063] Specifically, the first etching profile data set obtained after normalization processing can be denoted as X = {x 0 ′, x 1 ′,..., x i ′,... x t ′}, and x t ′ represents the historical etching profile data at the t-th historical moment after normalization processing.

[0064] S202, based on the first etching profile data set, use a long short-term memory network for modeling to obtain an initial etching profile prediction model.

[0065] The long short-term memory network can be used to model the first etching profile dataset to obtain an initial etching profile prediction model. The first etching profile dataset can be expressed as M×N, where M represents that the historical profile data at each historical moment has M points, and N represents a total of N groups of data, that is, N historical moments.

[0066] Specifically, refer to Figure 4 As shown, it is a schematic structural diagram of another etching profile prediction model provided by an embodiment of the present application. The input data is the first etching profile dataset, which can be 181×48 groups of data. 181 means that the etching profile data at each historical moment is represented by 181 points, and 48 groups means that the first etching profile dataset contains a total of 48 groups of data, that is, there are 48 historical moments. LSTM can be used to model these 48 groups of data to obtain 128×48 groups of data after modeling, that is, the output data width is 128 for 48 groups of data.

[0067] Specifically, after the first etching profile dataset passes through two layers of LSTM networks connected, the corresponding hidden layer state matrix H is obtained, and the calculation method is H = F(F(x)+x).

[0068] Specifically, assume that the hidden layer state matrix is H = {h 1 ,h 2 ,…,h t}, where h i represents the hidden layer state at the i historical moment, and the calculation method of h i is mainly calculated by the following equations of the LSTM unit:

[0069] i t =σ(W i ·[x t ,h t-1 +b i )

[0070] f t =σ(W f ·[x t ,h t-1 +b f )

[0071]

[0072]

[0073] o t =σ(W o ·[x t ,h t-1 +b o )

[0074] ht = o t *tanh(C t )

[0075] Wherein, i t , f t , o t respectively represent the outputs of the input gate, forget gate, and output gate at the t-th historical moment, x t represents the input at the t-th historical moment, h t-1 , h t respectively represent the hidden layer states at the (t - 1)-th historical moment and the t-th historical moment, C t represents the state of the LSTM neuron cell at the t-th historical moment, W and b are both trainable parameters, tanh is the hyperbolic tangent function, and σ is the sigmoid function.

[0076] S203. Based on the first etching profile dataset and the corresponding historical etching process parameters of the first etching profile dataset, use the cascaded combination layer to train the initial etching profile prediction model to obtain a trained etching profile prediction model.

[0077] In the embodiments of the present application, the corresponding historical etching process parameters of the first etching profile dataset can be obtained. The historical etching process parameters can represent the etching environment at the historical moment, and the historical etching process parameters can be normalized. The cascaded combination layer can be used to train the initial etching profile prediction model, and when the training stop condition is met, a trained etching profile prediction model can be obtained.

[0078] Specifically, the historical etching profile data and the historical etching process parameters can be concatenated and fully connected for operation to obtain a trained etching profile prediction model. Among them, the cascaded combination layer can include multiple concatenation layers and multiple fully connected layers. The concatenation layer can concatenate the historical etching profile data and the historical etching process parameters, and the fully connected layer can perform a fully connected process on the data formed after concatenation. In Figure 4 , the output data of the LSTM network can be processed into a one-dimensional data with a width of 128N, that is, 6144 bits. Since the width limit of the output data is 1024, 1024 bits are taken from 128N data each time, and concatenated with the corresponding 4 historical etching process parameters (such as etching power, etching voltage, etching pressure, and etching temperature), and then fully connected for operation to obtain 512-bit data. Then, the 512-bit data is concatenated with the 4 historical etching process parameters again, and fully connected for operation to obtain 256-bit data. The 256-bit data is concatenated with the 4 historical etching process parameters again, and fully connected for operation to obtain 256-bit data.

[0079] For example, assume that the 1024-bit input data is denoted as X = {x 1 , x2 , …, x 1024}, the input data is concatenated with the historical etching process parameters, and the concatenated input data can be denoted as X′ = {x 1 , x 2 , …, x 1024 , Power, Voltage, Pressure, Temperature}. Then, a fully connected operation is performed between layers. A total of 3 fully connected operations are performed, and 256 points are output. Finally, 181 points are sorted out, corresponding to an etching profile.

[0080] In a possible implementation, to improve the accuracy of the etching profile prediction model, transfer learning can be performed on the model, and the model is adjusted using real etching data with a small data volume to obtain the final etching profile prediction model.

[0081] Specifically, based on the first etching profile dataset and the corresponding historical etching process parameters, the initial etching profile prediction model can be trained using a cascade combination layer, and the obtained model is denoted as the target etching profile prediction model. Then, etching can be performed using a process method. The cross-section of the structure to be etched can be observed through a scanning electron microscope to obtain the actual etching profile. Among them, the actual etching process parameters corresponding to the actual etching profile are the same as the historical etching process parameters.

[0082] The target etching profile prediction model is adjusted using the actual etching profile and the actual etching process parameters. When the expectation is met, the etching profile prediction model is obtained. In this way, using real process profile data to fine-tune the model under the same etching process parameters can improve the prediction accuracy of the model.

[0083] Specifically, the mean absolute error (MAE) can be used to evaluate the performance of the etching profile prediction model, and its definition is:

[0084]

[0085] where y i represents the historical etching profile data, represents the predicted etching profile data.

[0086] Specifically, the neural network model in the present invention is based on RNN. Replacing it with a deep neural network (DNN), a convolutional neural network (CNN), their combination, or changing the number of network layers is within the scope of the present invention.

[0087] The present invention can be accelerated using CUDA (Compute Unified Device Architecture) tools, which can further improve the efficiency of etching modeling.

[0088] The present invention uses the mean absolute error to evaluate the model, and any form of replacing the error function is within the scope of the present invention.

[0089] The present invention uses polar coordinates to represent the etching profile data, realizing a simplified representation of the etching data. Except for the two-dimensional profiles in the examples, the profile representations in other dimensions are within the scope of the present invention.

[0090] The present invention uses simulation data for model pre-training and real data for model transfer learning. The modeling optimization methods for processes such as etching are within the scope of the present invention.

[0091] An embodiment of the present application provides an etching prediction method, which can obtain etching process parameters and a plurality of first etching profile data obtained by etching under the etching process parameters. Each first etching profile data corresponds to each first moment; input the etching process parameters and the plurality of first etching profile data into an etching profile prediction model, and output second etching profile data at a second moment; the second moment is after the first moment; the etching profile prediction model is determined based on a long short-term memory network and a cascade combination layer, and is a cascade recurrent neural network, which can reduce a large amount of calculations based on complex physical mechanisms in traditional etching models, reduce the repetitive manual trial-and-error work in traditional etching model calibration, reduce the modeling time, improve the simulation efficiency, and in addition, improve the prediction accuracy of the etching profile data.

[0092] Based on the above etching prediction method, an embodiment of the present application further provides an etching prediction device. Refer to Figure 5 As shown, it is a structural block diagram of an etching prediction device provided by an embodiment of the present application. The device may include:

[0093] An acquisition unit 201, configured to acquire etching process parameters and a plurality of first etching profile data obtained by etching under the etching process parameters. Each first etching profile data corresponds to each first moment;

[0094] A prediction unit 202, configured to input the etching process parameters and the plurality of first etching profile data into an etching profile prediction model, and output second etching profile data at a second moment; the second moment is after the first moment; the etching profile prediction model is determined based on a long short-term memory network and a cascade combination layer.

[0095] Specifically, the training process of the etching profile prediction model includes:

[0096] Obtain a first etching profile data set; the first etching profile data set includes a plurality of historical etching profile data, and the plurality of historical etching profile data correspond to a plurality of historical moments one by one;

[0097] Based on the first etching profile data set, use the long short-term memory network for modeling to obtain an initial etching profile prediction model;

[0098] Based on the first etching profile data set and the corresponding historical etching process parameters of the first etching profile data set, use the cascade combination layer to train the initial etching profile prediction model to obtain the trained etching profile prediction model.

[0099] Specifically, the step of using the cascade combination layer to train the initial etching profile prediction model based on the first etching profile data set and the corresponding historical etching process parameters of the first etching profile data set to obtain the trained etching profile prediction model includes:

[0100] Based on the first etching profile data set and the corresponding historical etching process parameters of the first etching profile data set, use the cascade combination layer to train the initial etching profile prediction model to obtain a target etching profile prediction model;

[0101] Obtain the actual etching profile obtained by process means and the actual etching process parameters corresponding to the actual etching profile;

[0102] Use the actual etching profile and the actual etching process parameters to adjust the target etching profile prediction model to obtain the etching profile prediction model.

[0103] Specifically, the step of using the cascade combination layer to train the initial etching profile prediction model based on the first etching profile data set and the corresponding historical etching process parameters of the first etching profile data set to obtain the trained etching profile prediction model includes:

[0104] Perform splicing and fully connected operations on the historical etching profile data and the historical etching process parameters to obtain the trained etching profile prediction model.

[0105] Specifically, the step of obtaining the first etching profile data set includes:

[0106] Generate the first etching profile data set through simulation software.

[0107] Specifically, the step of generating the first etching profile data set through simulation software includes:

[0108] Generate a second etching profile data set through simulation software;

[0109] Normalize the second etching profile data set to obtain the first etching profile data set.

[0110] Specifically, the cascaded combination layer includes a plurality of splicing layers and a plurality of fully connected layers.

[0111] An embodiment of the present application provides an etching prediction device. The acquisition unit is used to acquire etching process parameters and a plurality of first etching profile data obtained by etching under the etching process parameters, and each of the first etching profile data corresponds to each first moment; the prediction unit is used to input the etching process parameters and the plurality of first etching profile data into an etching profile prediction model, and output second etching profile data at a second moment; the second moment is after the first moment; the etching profile prediction model is determined based on a long short-term memory network and a cascaded combination layer, and is a cascaded recurrent neural network, which can reduce a large amount of calculations based on complex physical mechanisms in traditional etching models, reduce repetitive manual trial-and-error work in traditional etching model calibration, reduce modeling time, improve simulation efficiency, and in addition, improve the prediction accuracy of etching profile data.

[0112] On the other hand, an embodiment of the present application provides a computer device. Refer to Figure 6 As shown, it is a structural diagram of a computer device provided by an embodiment of the present application. The computer device includes a processor 310 and a memory 320:

[0113] The memory 320 is used to store program codes and transmit the program codes to the processor 310;

[0114] The processor 310 is used to execute the method provided by the above embodiment according to the instructions in the program code.

[0115] The computer device may include a terminal device or a server, and the foregoing device may be configured in the computer device.

[0116] On the other hand, an embodiment of the present application further provides a storage medium, which is used to store a computer program, and the computer program is used to execute the method provided by the above embodiment.

[0117] Those of ordinary skill in the art can understand that all or part of the steps of implementing the above method embodiments can be completed by program instructions in hardware. The foregoing program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps including the above method embodiments; and the foregoing storage medium can be at least one of the following media: read-only memory (English: Read-only Memory, abbreviation: ROM), RAM, magnetic disk, or optical disk and other media that can store program codes.

[0118] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the apparatus embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and reference can be made to the relevant parts of the method embodiments for the relevant content.

[0119] The above description is only a preferred embodiment of the present application. Although the present application has been disclosed above with preferred embodiments, it is not intended to limit the present application. Any person skilled in the art can make many possible changes and modifications to the technical solution of the present application, or modify it into an equivalent embodiment with equivalent changes, without departing from the scope of the technical solution of the present application. Therefore, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present application without departing from the content of the technical solution of the present application are still within the scope of the protection of the technical solution of the present application.

Claims

1. An etching prediction method, characterized in that, comprising: Obtaining etching process parameters and a plurality of first etching profile data etched under the etching process parameters, each of the first etching profile data corresponding to each first moment; Inputting the etching process parameters and the plurality of first etching profile data into an etching profile prediction model, and outputting second etching profile data at a second moment; The second moment is after the first moment; the etching profile prediction model is determined based on a long short-term memory network and a cascade combination layer.

2. The method according to claim 1, characterized in that, The training process of the etching profile prediction model includes: Obtaining a first etching profile data set; the first etching profile data set includes a plurality of historical etching profile data, and the plurality of historical etching profile data corresponds one-to-one to a plurality of historical moments; Based on the first etching profile data set, using the long short-term memory network for modeling to obtain an initial etching profile prediction model; Based on the first etching profile data set and the corresponding historical etching process parameters of the first etching profile data set, using the cascade combination layer to train the initial etching profile prediction model to obtain the trained etching profile prediction model.

3. The method according to claim 2, characterized in that, The step of training the initial etching profile prediction model with the cascade combination layer based on the first etching profile data set and the corresponding historical etching process parameters of the first etching profile data set to obtain the trained etching profile prediction model includes: Based on the first etching profile data set and the corresponding historical etching process parameters of the first etching profile data set, using the cascade combination layer to train the initial etching profile prediction model to obtain a target etching profile prediction model; Obtaining an actual etching profile obtained by process means and the actual etching process parameters corresponding to the actual etching profile; Using the actual etching profile and the actual etching process parameters to adjust the target etching profile prediction model to obtain the etching profile prediction model.

4. The method according to claim 2, characterized in that, The step of training the initial etching profile prediction model with the cascade combination layer based on the first etching profile data set and the corresponding historical etching process parameters of the first etching profile data set to obtain the trained etching profile prediction model includes: Performing splicing and fully connected operations on the historical etching profile data and the historical etching process parameters to obtain the trained etching profile prediction model.

5. The method according to claim 2, characterized in that, The step of obtaining the first etching profile data set includes: Generating the first etching profile data set through simulation software.

6. The method according to claim 5, characterized in that, The step of generating the first etching profile data set through simulation software includes: Generating a second etching profile data set through simulation software; Normalizing the second etching profile data set to obtain the first etching profile data set.

7. The method according to any one of claims 1-6, wherein, the cascaded combination layer includes a plurality of splicing layers and a plurality of fully connected layers.

8. An etching prediction device, wherein, it includes: an acquisition unit configured to acquire etching process parameters and a plurality of first etching profile data obtained by etching under the etching process parameters, each of the first etching profile data corresponding to each first moment; a prediction unit configured to input the etching process parameters and the plurality of first etching profile data into an etching profile prediction model, and output second etching profile data at a second moment; the second moment is after the first moment; the etching profile prediction model is determined based on a long short-term memory network and a cascaded combination layer.

9. A computer device, wherein, the computer device includes a processor and a memory: the memory is configured to store program code and transmit the program code to the processor; the processor is configured to execute the method according to any one of claims 1-7 according to the instructions in the program code.

10. A computer-readable storage medium, wherein, the computer-readable storage medium is configured to store a computer program, and the computer program is configured to execute the method according to any one of claims 1-7.

Images