A parameter prediction model construction method, prediction method and prediction system for CD-SEM measurement

Abstract

The application discloses a parameter prediction model construction method, a prediction method and a prediction system for CD-SEM measurement. The model construction method comprises the following steps: obtaining a data set, wherein the data set comprises at least one group of data groups with proximity effect characteristics, each group of data groups comprises standard data and an SE signal group, the standard data comprises aspect ratio data and / or height data, and the SE signal group in the data group with proximity effect characteristics is a secondary electron signal generated by electron beam scanning on a geometric structure with adjacent structures; taking the SE signal group in the same group of data groups in the data set as input, taking the standard data in the same group of data groups as output, training a pre-constructed neural network model, and obtaining a parameter prediction model for predicting the aspect ratio data and / or height data in CD-SEM measurement. The application does not need to modify the hardware structure, thereby avoiding the problems of complicating the hardware structure and affecting reliability.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of semiconductor metrology, and in particular to a parameter prediction model construction method, a prediction method and a prediction system for CD-SEM metrology. BACKGROUND

[0002] Critical dimension scanning electron microscope (CD-SEM) is a key technology for measuring pattern structures on wafers or masks in the semiconductor industry. With the development of the integrated circuit industry, the size of device structures becomes smaller and smaller, and device structures also evolve into three-dimensional. CD-SEM faces greater challenges, especially in three-dimensional (3D) metrology, in which the height and aspect ratio of device structures are important parameters for characterizing three dimensions. However, CD-SEM cannot directly measure these important parameters at present, which makes it very inconvenient to obtain these parameters.

[0003] For the measurement of height and aspect ratio, there are several methods in the prior art: 1. Using two or more images with different incident electron beam angles, but this method increases the complexity of the measurement equipment and requires more time to collect images; 2. Using an exponential function to relate the BSE (backscattered electron) signal intensity to the depth of the HAR (high aspect ratio) hole, but this method usually requires the CD SEM to work at a high acceleration voltage state, which can easily damage the sample, and in addition, the image quality through BSE detection is lower than that of secondary electron (SE) images. 3. By tilting the sample workpiece table, multiple-angle incident electron beam images are obtained, but the mechanical movement of the workpiece table brings instability and strict alignment problems. For the measurement of height and aspect ratio, Applied Materials has developed and implemented electron beam tilting on the product V4i+ CD-SEM for height measurement. After calibration, low-angle (5°) and high-angle (14°) beams can obtain two sets of images for measuring the sidewall edge width to calculate the height. However, the method used by Applied Materials V4i+ makes the CD SEM hardware complex and brings reliability problems. SUMMARY

[0004] The embodiments of the present application provide a parameter prediction model construction method, a prediction method and a prediction system for CD-SEM metrology to solve the problem that it is difficult to measure the aspect ratio and height parameters of device structures in the prior art.

[0005] According to a first aspect of the present application, a parameter prediction model construction method for CD-SEM metrology is provided, comprising:

[0006] obtain a data set, the data set comprising at least one group of data sets with proximity effect characteristics, each group of data sets comprising standard data and SE signal groups, the standard data comprising aspect ratio data and / or height data, and the SE signal groups in the data sets with proximity effect characteristics being secondary electron signal groups generated by electron beam scanning on geometric structures with adjacent structures;

[0007] train a pre-constructed neural network model by taking the SE signal groups in the same group of data sets in the data set as input and taking the standard data in the same group of data sets as output, to obtain a parameter prediction model for predicting aspect ratio data and / or height data in CD-SEM measurement.

[0008] The parameter prediction model construction method of the present application trains the constructed neural network model by taking the SE signal groups as input and taking the corresponding standard data, i.e., the corresponding aspect ratio data and / or height data, as output, so that the obtained parameter prediction model can predict the aspect ratio data and / or height data of the corresponding device structure based on the SE signal. Compared with the prior art method, the hardware structure does not need to be modified, so there is no problem of complicating the hardware structure, no problem of affecting reliability, and also no need to adjust the working parameters of the CD-SEM, which will not cause damage to the sample and will not affect the image quality. The data set used by the parameter prediction model of the present application in training is a data set with proximity effect characteristics, so that the parameters related to the proximity effect characteristics can be predicted. After precision detection test, when predicting the aspect ratio and height, the average relative error between the obtained aspect ratio prediction value and the standard value is 2.4%, the percentage of prediction cases with an error within 20% is 100.0%, the average relative error between the obtained height prediction value and the standard value is 1.9%, and the percentage of prediction cases with an error within 10% is 99.5%.

[0009] In some embodiments, the method for constructing the neural network model comprises:

[0010] constructing the network architecture of the neural network model based on the Identity module and the Dense module;

[0011] constructing the neural network model according to the constructed network architecture and the models library under the Keras library encapsulating TensorFlow, wherein, when constructing, the model is compiled using the compile function, the loss function is specified as mse, Adam is specified as the optimizer of the neural network, and the metrics function is specified as mse.

[0012] Thus, by such setting, the network architecture of the parameter prediction model of the application adopts a fully connected layer instead of a convolutional layer and a pooling layer, compared with the traditional residual network more suitable for image processing, the modified model architecture is more suitable for nonlinear regression, so that the prediction result accuracy of the neural network model after training is higher, and the training efficiency is also higher.

[0013] In some embodiments, the network architecture of the constructed neural network model is set to consist of two basic modules connected in sequence, a batch normalization layer, a fully connected layer and a linear activation function, wherein each basic module is formed by sequentially connecting two Identity modules after a Dense module;

[0014] The Identity module comprises three basic units connected in sequence, wherein each basic unit consists of a fully connected layer, a batch normalization layer and a ReLU activation function connected in sequence, and a Dropout layer is inserted after the second basic unit;

[0015] The Dense module comprises three basic units connected in sequence, wherein a Dropout layer is inserted after the second basic unit, and a residual path unit is connected after the first basic unit and the second basic unit, respectively, the residual path unit consists of a fully connected layer and a batch normalization layer stacked in sequence, and the output of the residual path unit and the result between the batch normalization layer and the ReLU activation function in the third basic unit in the Dense module are configured to perform tensor summation operation.

[0016] Thus, by such setting, the network architecture can be further adapted to nonlinear regression, improving the accuracy of the prediction result and the training efficiency.

[0017] In some embodiments, the data set is set to obtain standard data according to the parameters of the constructed geometry structure by performing SEM simulation imaging on the simulated constructed geometry structure with adjacent structure, and obtain SE signal according to the parameters of the simulation imaging.

[0018] Thus, by such setting, the actual product can be replaced by the simulated constructed structure for testing, thereby greatly reducing the training cost, and the parameters of the simulated constructed structure are highly controllable, and various device structures with different parameter combinations can be provided as data sets to provide different types of data samples for model training.

[0019] In some embodiments, the data set includes a training set and a test set, and the training set and the test set are obtained by random number division on a plurality of data groups in the data set;

[0020] When the constructed neural network model is trained, the data groups in the training set are used to train the constructed neural network model to obtain model parameters, and the data groups in the test set are used to optimize the obtained model parameters.

[0021] Therefore, by such setting, the data set can be divided, the model can be trained by using the training set, the model parameters can be optimized by using the test set, the accuracy of the parameter prediction model obtained by training is ensured, and the data set can be shuffled so that the data groups in the data set are no longer consecutive, thereby reducing the consideration of the continuity of the data by the parameter prediction model during training, so that the finally obtained parameter prediction model can be more suitable for more different scenarios, and the prediction accuracy of the parameter prediction model is improved.

[0022] In some embodiments, the pre-constructed neural network model is trained by using the SE signal groups in the same group of data groups in the data set as input and the standard data in the same group of data groups as output to obtain a parameter prediction model, including:

[0023] According to the preset number of training times, the data groups in the training set are selected, and the SE signal groups in the same group of data groups are used as input and the standard data in the same group of data groups are used as output to train the pre-constructed neural network model to obtain first model parameters;

[0024] The neural network model is updated according to the first model parameters to obtain a first neural network model;

[0025] The SE signal groups in the data groups in the test set are input into the first neural network model, and the prediction result of the first neural network model for the input SE signal groups is obtained;

[0026] The error between the prediction result and the standard data in the corresponding data group is evaluated by using a loss function value, and the first neural network model is optimized according to the optimizer of the neural network and the error to obtain a parameter prediction model.

[0027] Therefore, by such setting, the setting in the neural network training process can be improved, the MSE specified by the loss function is used as an index to monitor the training effect in real time during the training process, and the neural network model is iteratively optimized to improve the accuracy of the finally obtained parameter prediction model.

[0028] In some embodiments, before the constructed neural network model is trained, the data set is preprocessed, including normalizing the data set.

[0029] Therefore, by setting in this way, the bias of the simulator or the electron microscope itself can be avoided. In addition, data normalization can solve the problem of inconsistent feature scales, making the influence of different features on the model more balanced. Furthermore, normalization can also prevent the problem of gradient vanishing or explosion, thereby accelerating the convergence speed of the model.

[0030] In some embodiments, further comprising determining the prediction accuracy of the parameter prediction model according to the data set and the parameter prediction model, comprising:

[0031] inputting the SE signal group in the selected data group in the data set into the parameter prediction model, and obtaining the prediction data output by the parameter prediction model;

[0032] determining the prediction accuracy of the parameter prediction model according to the prediction data and the standard data in the data group where the input SE signal group is located through the metrics function.

[0033] Therefore, by setting in this way, the prediction accuracy of the trained parameter prediction model can be further determined as a reference for accuracy.

[0034] According to a second aspect of the present application, a training device for a parameter prediction model for CD-SEM measurement is provided, comprising:

[0035] a data acquisition module configured to acquire a data set, the data set comprising at least one group of data groups with proximity effect characteristics, each group of data groups comprising standard data and an SE signal group, the standard data comprising aspect ratio data and / or height data, and the SE signal group in the data group with proximity effect characteristics being a secondary electron signal generated by electron beam scanning on a geometric structure with adjacent structures;

[0036] a training module configured to input the SE signal group in the same group of data groups in the data set as input, and input the standard data in the same group of data groups as output, to train a pre-constructed neural network model, and obtain a parameter prediction model for predicting aspect ratio data and / or height data in CD-SEM measurement.

[0037] According to a third aspect of the present application, a parameter prediction method for CD-SEM measurement is provided, which uses the parameter prediction model obtained by the parameter prediction model construction method for CD-SEM measurement according to the first aspect of the present application and the SE signal of the device structure to predict the aspect ratio and / or height of the device structure.

[0038] According to a fourth aspect of the present application, a parameter prediction system for CD-SEM measurement is provided, comprising:

[0039] A prediction module is configured to predict the aspect ratio and / or height of the device structure according to the SE signal of the device structure and the parameter prediction model obtained by the parameter prediction model construction method for CD-SEM measurement according to any one of claims 1 to 8. BRIEF DESCRIPTION OF DRAWINGS

[0040] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0041] Figure 1 The flow chart of the parameter prediction model construction method for CD-SEM measurement according to an embodiment of the present application;

[0042] Figure 2 The overall structure schematic diagram of the simulated and constructed geometry structure when the data set is obtained in the parameter prediction model construction method for CD-SEM measurement according to an embodiment of the present application;

[0043] Figure 3 The cross-sectional structure schematic diagram of the simulated and constructed geometry structure when the data set is obtained in the parameter prediction model construction method for CD-SEM measurement according to an embodiment of the present application;

[0044] Figure 4 The SE signal comparison diagram of structures with different structure bottom widths;

[0045] Figure 5 The SE signal comparison diagram of structures with different side wall angles;

[0046] Figure 6 The SE signal comparison diagram of structures with different heights;

[0047] Figure 7 The flow chart of step S11 in the parameter prediction model construction method for CD-SEM measurement according to an embodiment of the present application;

[0048] Figure 8 The network architecture schematic diagram of the neural network model in the parameter prediction model construction method for CD-SEM measurement according to an embodiment of the present application;

[0049] Figure 9 The flow chart of step S12 in the parameter prediction model construction method for CD-SEM measurement according to an embodiment of the present application;

[0050] Figure 10A flow chart of a parameter prediction model construction method for CD-SEM measurement according to another embodiment of the present application;

[0051] Figure 11 A scatter plot of the aspect ratio in the standard data and the aspect ratio in the prediction data according to the embodiment of the present application; Figure 10 A flow chart of step S13 in the parameter prediction model construction method for CD-SEM measurement according to the embodiment of the present application;

[0052] Figure 12 A scatter plot of the height in the standard data and the height in the prediction data according to the embodiment of the present application;

[0053] Figure 13 A relative error distribution plot between the height in the standard data and the height in the prediction data according to the embodiment of the present application;

[0054] Figure 14 A scatter plot of the height in the standard data and the height in the prediction data according to the embodiment of the present application;

[0055] Figure 15 A relative error distribution plot between the height in the standard data and the height in the prediction data according to the embodiment of the present application;

[0056] Figure 16 A principle block diagram of a training device of a parameter prediction model for CD-SEM measurement according to an embodiment of the present application;

[0057] Figure 17 A structural schematic diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0058] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0059] It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0060] Finally, it is to be understood that the terms such as first and second, and the like, merely denote different instances of an entity or an operation, and do not necessarily require or imply any actual relationship or order between such entities or operations. Moreover, the terms "comprises", "comprising", "includes", "including" and the like, specify the presence of stated elements, but do not preclude the presence or addition of other elements, or preclude adaption of other stated elements. An element procured "including" another element, does not, without more restrictions, foreclose the existence of additional identical elements in the process, method, article, or apparatus that comprises the utilized element.

[0061] The application will be further described in detail below with reference to the accompanying drawings.

[0062] Figure 1 The overall step flow of the parameter prediction model construction method for CD-SEM measurement according to an embodiment of the application is schematically shown, referring to FIG. 1, the parameter prediction model construction method for CD-SEM measurement according to the application comprises the following steps: Figure 1

[0063] Step S11: obtaining a data set, the data set comprising at least one group of data groups with proximity effect characteristics, each group of data groups comprising standard data and a SE signal group, the standard data comprising aspect ratio data and / or height data, the SE signal group in the data group with proximity effect characteristics being a secondary electron signal generated by electron beam scanning on a geometric structure with adjacent structures;

[0064] Step S12: taking the SE signal group in the same group of data groups in the data set as input, taking the standard data in the same group of data groups as output, training a pre-constructed neural network model to obtain a parameter prediction model for predicting aspect ratio data and / or height data in CD-SEM measurement.

[0065] Step S11 is a step of obtaining a data set, specifically, the data set comprises a plurality of groups of data, each group of data representing a structure parameter of a device structure and a SE signal obtained by performing SEM simulation imaging, that is, each group of data comprises standard data and a corresponding SE signal group, the standard data comprising aspect ratio data and / or height data. Among them, at least one group of data is a data group with proximity effect characteristics, the SE signal group in the data group with proximity effect characteristics being a secondary electron signal generated by electron beam scanning on a geometric structure with adjacent structures. Specifically, a two-dimensional array can be used to represent the data set. For example, each row represents a group of data, the first two columns of each row are standard data, and the third column is the SE signal group, thereby realizing one-to-one correspondence between the SE signal and its corresponding aspect ratio data and height data.​

[0066] Each group of data in the data set can be measured and tested by actual samples. In some possible embodiments, the data set with proximity effect characteristics in the embodiments of the present application can be obtained by SEM simulation imaging of the simulation-constructed geometric structure with adjacent structures. Referring to Figure 2 and Figure 3 , schematic overall structure diagrams and cross-sectional structure diagrams of the simulation-constructed geometric structure with adjacent structures are shown. Figure 2 and Figure 3 Generally, it can be considered that the arc of the round corner is tangent to the adjacent edge, so the position of the arc of the round corner can be obtained according to swa and r. In CD-SEM metrology, the influence of proximity effect cannot be ignored, so the proximity effect performance of different structure distances, side wall angles and heights is analyzed. The reference structure parameters used are: bcd = 20, swa = 86°, d = 20, h = 100. When the values of d, swa or h are changed, the other parameters remain unchanged. All the above scanning signal profiles are from SE signals. Exemplarily, the data set derived in the embodiments of the present application contains 5906 groups of data, that is, 5906 rows of data, wherein the first column is the aspect ratio data, the second column is the height data, and each subsequent column is the SE signal, and the total number of columns is 245.

[0067] Figures 4 to 6 Schematic diagrams of SE signals corresponding to changes in different parameters are shown. Figure 4 The SE signal comparison diagrams of structures with different structure bottom widths are shown. Specifically, the values of d in these structures are set to 20, 30, 50, 100 and 200 nm, respectively. From Figure 4 , it can be seen that the SE signal generally shows a trend of weakening as the line structure distance decreases, because when the adjacent line distance is closer, the SE signal is more easily blocked, that is, the proximity effect is more obvious. Figure 5 The SE signal comparison diagrams of structures with different side wall angles are shown. Specifically, the values of swa in these structures are 86°, 87°, 88°, 89° and 90°, respectively. From Figure 5 , it can be seen that the SE signal peak value decreases as the value of swa increases, because the possibility of blocking the SE signal by the adjacent structure increases. The SE signal peak value moves to the right as swa increases, because the upper edge width increases. In addition, when swa is relatively small, a second peak value of the signal curve appears. When swa increases to 88°, the second peak value is almost invisible. Figure 6The SE signal comparison graphs of structures with different heights. The h values of these structures are 70, 80, 90 and 100 nm respectively. It can be seen from Figure 6 that the peak value of the SE signal decreases as the height decreases, and the peak position shifts to the right, which is due to the different upper groove widths caused by different heights when swa is fixed. In addition, it can be found that the SE signal from both sides, the path between the upper edge and the lower edge, increases as the height decreases, which is due to the decrease of the height makes the adjacent structure less blocking the SE signal. Further, in the present application, the data set with proximity effect characteristics is obtained for training, so that the parameters related to the proximity effect characteristics can be predicted.

[0068] In some possible embodiments, the data set can be split to form a training set and a test set, wherein the training set and the test set are obtained by random number division of a plurality of groups of data in the data set. Specifically, in the data set, the training set accounts for 90%, and the remaining 10% is the test. When training the constructed neural network model, the constructed neural network model is trained with the training set obtained by division, the test set obtained by division is used to optimize the model during training of the neural network model, and the prediction accuracy of the parameter prediction model is determined after the parameter prediction model is trained. It can be understood that in the present application, the data set is obtained by SEM simulation imaging of the simulated geometric structure with adjacent structures, and there is a certain continuity between each group of data in the data set. Therefore, when training, the neural network model may

[0069] Step S13 is a step of training the pre-constructed neural network model with the obtained data set to obtain a parameter prediction model. Specifically, the pre-constructed neural network is constructed based on the network architecture constructed by the Identity module and the Dense module. Figure 7 The steps of the construction method of the pre-constructed neural network model in the parameter prediction model construction method for CD-SEM metrology according to an embodiment of the present application are schematically shown in FIG. 13. As shown in FIG. 13, the steps can be implemented as follows: Figure 7

[0070] ​Step S21: constructing a network architecture of the neural network model based on the Identity module and the Dense module;

[0071] Step S22: constructing the neural network model according to the constructed network architecture and the models library under the Keras library encapsulating TensorFlow, wherein, when constructing, the model is compiled using the compile function, the loss function is specified as mse, Adam is specified as the optimizer of the neural network, and the metrics function is specified as mse.

[0072] In step S21, the network architecture of the neural network model is first constructed. In the present application, the network architecture of the constructed neural network model is the network architecture of the residual network. The difference between the present application and the residual network is that the convolutional layer and the pooling layer of the residual network are replaced by the fully connected layer. Figure 8 The network architecture of the parameter prediction model for CD-SEM measurement in an embodiment of the present application is schematically shown, referring to Figure 8 As shown, the architecture of the Identity module, the architecture of the Dense module and the overall network architecture are shown. The overall network architecture is set to be composed of two basic modules connected in sequence, one batch normalization layer, one fully connected layer and one linear activation function. Each basic module is formed by sequentially connecting two Identity modules after one Dense module. The Identity module is set to be composed of three basic units connected in sequence, each basic unit is composed of a fully connected layer, a batch normalization layer and a ReLU activation function connected in sequence, and a Dropout layer is inserted after the second basic unit. The Dense module is set to be composed of three basic units connected in sequence, a Dropout layer is inserted after the second basic unit, and a residual path unit is additionally connected after the first basic unit and the second basic unit, respectively, the residual path unit is composed of a fully connected layer and a batch normalization layer stacked in sequence, and the output of the residual path unit and the result between the batch normalization layer and the ReLU activation function in the third basic unit in the Dense module are configured to perform tensor summation operation. The network architecture of the present application is set in this way. Compared with the network architecture of the traditional neural network model, the traditional residual network is more suitable for image processing, and the network architecture of the present application is improved and optimized, and can be more suitable for nonlinear regression.

[0073] After the network architecture is constructed, step S22 is performed. Step S22 is a step of constructing a neural network model based on the constructed network architecture. Specifically, in the present embodiment, it can be implemented to construct a neural network model according to the constructed network architecture and the models library under the Keras library encapsulating TensorFlow. Keras is an open-source neural network library that can run with TensorFlow, CNTK, or Theano as a backend. Among them, when constructing the neural network model, the compile function is used to compile the model, the loss function is specified as mse, Adam is specified as the optimizer of the neural network to control the iteration direction of the neural network parameters, and the metrics function is specified as mse. The loss function, i.e. the loss function (Loss Function), is an important concept in machine learning and deep learning, which is used to measure the difference or error between the model prediction result and the true result. It is a numerical evaluation index that provides a measure of model performance by comparing the model output and the true label. The main function of the optimizer is to make the gradient faster and better in the process of gradient descent, so as to find the minimum value of the objective function as soon as possible, that is, to find the optimal "loss" (loss) by which way. The metrics function, i.e. the evaluation function, is used to evaluate the performance of the model during and after training. When the model is compiled (compile), the evaluation function is input as the parameter of metrics. The evaluation function is similar to the loss function, but the result of the evaluation function is not used in the training process. The accuracy evaluation function is used to evaluate the accuracy of the model prediction. Among the commonly used regression evaluation indicators, the relative error of the model prediction using MSE as the loss function is smaller, specifically the relative error using MSE is 4.8%, the relative error using MAE is 6.3%, and the relative error using MSLE is 56.6%. For the regression evaluation indicators selected by the metrics function, the same function as the loss function is generally selected to make it easier to observe the training effect.

[0074] In the implementation of dividing the data set into a training set and a test set when training the pre-constructed neural network model using the obtained data set, the training set can be used to train the neural network model. In some possible implementations, before training, the data set can also be preprocessed, and the preprocessing includes normalizing the data set. The data set is normalized to avoid the bias of the simulator or the electron microscope itself. In addition, data normalization can solve the problem of inconsistent scales of different features, so that the influence of different features on the model is more balanced. In addition, normalization can also prevent the problem of gradient vanishing or explosion, thereby accelerating the convergence speed of the model. When training the neural network model, the fit function can be used for training, epochs, batch_size, validation_split are specified, and the loss value and the validation loss value (metrics value) in the training process are recorded. The epoch refers to the process of passing the neural network once and returning once when a complete data set passes through the neural network, that is, an epoch is the process of training all training samples once. The batch_size is the number of samples input into the model at the same time in each iteration when training the model. The validation_split is a floating point number between 0 and 1, which is used to specify a certain proportion of data in the data set as a test set. For example, in this embodiment, the training set accounts for 90%, and the remaining 10% is the test set, so the validation_split is specified as 0.1. After obtaining the parameter prediction model, the parameter prediction model is saved. Figure 9 The step flow of step S12 of an embodiment of the application is schematically shown, with reference to Figure 9 which can be specifically implemented as comprising the following steps:

[0075] Step S31: According to the preset number of training times, select the data group in the training set, and train the pre-constructed neural network model by taking the SE signal group in the same data group as input and taking the standard data in the same data group as output, to obtain first model parameters;

[0076] Step S32: Update the neural network model according to the first model parameters to obtain a first neural network model;

[0077] Step S33: Input the SE signal group in the data group in the test set into the first neural network model, and obtain the prediction result of the first neural network model for the input SE signal group;

[0078] Step S34: Evaluate the error between the prediction result and the standard data in the corresponding data group by using the loss function value, and optimize the first neural network model according to the optimizer of the neural network and the error, to obtain a parameter prediction model.

[0079] In step S31, the preset sample number of the simultaneous input model can be determined by a specified batch_size value, and the preset training number can be determined by a specified epoch value. According to the specified batch_size value and the epoch value, the pre-constructed neural network model is trained by taking the SE signal group in the same group of data sets as input and taking the standard data in the same group of data sets as output, so that the first model parameter of the corresponding trained neural network model is obtained.

[0080] Step S32 is a step of updating the neural network model based on the first model parameter. After updating, the first neural network model based on the first model parameter is obtained.

[0081] In step S33, by inputting the input data of the data group in the test set, i.e., the SE signal group, into the first neural network model, the corresponding prediction result is obtained.

[0082] Finally, step S34 is performed to optimize the first neural network model based on the prediction result and the standard data. Specifically, the error of the first neural network model can be determined by using the loss function specified when the neural network model is constructed. After obtaining the error, the first neural network model is optimized according to the error and the optimizer Adam specified when the neural network model is constructed, so that the parameter prediction model is obtained.

[0083] Figure 10 The overall step flow of the parameter prediction model construction method for CD-SEM measurement of another embodiment of the present application is schematically shown, and Figure 10 As shown in FIG. 13, in some possible embodiments, the parameter prediction model construction method for CD-SEM measurement of the present application can further include the following steps:

[0084] Step S13: determining the prediction accuracy of the parameter prediction model according to the data set and the parameter prediction model.

[0085] Step S13 is a step of detecting the prediction accuracy of the trained parameter prediction model. By determining the prediction accuracy, the prediction result can be compared. Figure 11 The step flow of step S13 of the parameter prediction model construction method for CD-SEM measurement of one embodiment of the present application is schematically shown, and Figure 11 As shown in FIG. 13, step S13 can be specifically implemented by including the following steps:

[0086] Step S41: taking the SE signal group in the selected data group in the data set as input, inputting the parameter prediction model, and obtaining the prediction data output by the parameter prediction model;

[0087] Step S42: determining the prediction accuracy of the parameter prediction model according to the prediction data and the standard data in the data set where the input SE signal set is located through the metrics function.

[0088] Step S41 is a step of simulating prediction by using the parameter prediction model. Specifically, in step S41, the test set can be used for simulating prediction, and the SE signal set in the test set is taken as the input of the parameter prediction model to determine the prediction data. The prediction data includes the predicted aspect ratio and / or the predicted height.

[0089] Step S42 is a step of calculating the mean square error of the prediction data and the corresponding standard data. Specifically, the calculation can be performed based on the metrics function specified when the neural network model is constructed, so that the error rate of the two can be calculated. Specifically, the change of loss with epoch in the training process can be derived, the corresponding scatter plot of the standard data and the prediction data can be derived, and the distribution diagram of the prediction data error can be derived to form visualized data for comparison and observation.

[0090] Exemplarily, the following example shows the whole process of the parameter prediction model construction method for CD-SEM measurement. In the process of constructing the network architecture and the neural network model, the codes of data processing, deep learning model architecture design, training, verification and testing are written in Python programming language, and Keras is realized by using the high-level API of TensorFlow. In the training process, MSE is used as the training and verification loss, and the formula of MSE is as follows:

[0091]

[0092] The data set used contains 5906 data sets, of which 90% of the data sets are used as the training set and 10% of the data sets are used as the test set by using random number division. In the data set, the range of bcd is from 20 nm to 36 nm, the range of swa is from 80° to 90°, the range of h is from 48 nm to 144 nm, the range of aspect ratio is from 4 to 20, r is fixed at 1 nm, and the electron beam scanning tilt angle is fixed at 2.5°.

[0093] Then the neural network model can be trained by using the training set to obtain the parameter prediction model. After obtaining the parameter prediction model, the parameter prediction model can be saved, and the prediction accuracy of the parameter prediction model can be detected by using the test set. Specifically, the detection result is as shown in Figures 12 to 15 Figure 12 is the scatter plot of the aspect ratio in the standard data and the aspect ratio in the prediction data, and the shadow part represents that the relative error is within 20%. Figure 13 ​A relative error distribution plot between the aspect ratio in the standard data and the aspect ratio in the predicted data. Figure 14 A scatter plot of the height in the standard data and the height in the predicted data, wherein the shaded part represents the relative error within 10%. Figure 15 A relative error distribution plot between the height in the standard data and the height in the predicted data.

[0094] Predicted result data:

[0095] 1. Aspect ratio: The final training and validation loss are 0.0052 and 0.00035, respectively, and the loss of the test set is 0.00036. The square root of the test loss (RMSE) is 0.0190. By testing the test set (10% of the data set), the average relative error between the predicted data and the standard data is: 2.4%, and the percentage of predicted cases with an error within 20% is 100.0%.

[0096] Height: The final training and validation loss are 0.0052 and 0.00067, respectively, and the loss of the test set is 0.00068. The square root of the test loss (RMSE) is 0.0262. By testing the test set (10% of the data set), the average relative error between the predicted data and the standard data is: 1.9%, and the percentage of predicted cases with an error within 10% is 99.5%.

[0097] Figure 16 The overall module composition of the training device of the parameter prediction model for CD-SEM measurement according to an embodiment of the present application is schematically shown, referring to Figure 16 The training device of the parameter prediction model for CD-SEM measurement according to the present application includes the following modules, as shown in the figure.

[0098] The data acquisition module 1 is used to acquire a data set, wherein the data set includes at least one group of data sets with proximity effect characteristics, each group of data sets includes standard data and a group of SE signals, the standard data includes aspect ratio data and / or height data, and the group of SE signals in the data set with proximity effect characteristics is a secondary electron signal generated by electron beam scanning on a geometric structure with adjacent structures.

[0099] The training module 2 is used to take the group of SE signals in the same group of data sets in the data set as input, take the standard data in the same group of data sets as output, train the pre-constructed neural network model, and obtain a parameter prediction model for predicting the aspect ratio data and / or height data in CD-SEM measurement.

[0100] It should be noted that the implementation process and implementation principle of the training device of the parameter prediction model for CD-SEM measurement of the embodiments of the present application can be referred to the corresponding description of the above method embodiments, for example, the corresponding description of the data set acquisition, model training and the like in the method embodiment part, and therefore will not be described here. Exemplarily, the training device of the parameter prediction model for CD-SEM measurement of the embodiments of the present application can be any intelligent device with a processor, including but not limited to a computer, a smart phone, a personal computer, a robot, a cloud server and the like.

[0101] The present application also provides a parameter prediction method for CD-SEM measurement, which uses the parameter prediction model obtained by the parameter prediction model construction method for CD-SEM measurement of any one of the above embodiments and the SE signal of the device structure to predict the aspect ratio and / or height of the device structure. Specifically, only the SE signal of the device structure needs to be input into the parameter prediction model to predict the aspect ratio and / or height of the device structure. Compared with the method of the prior art, the hardware structure does not need to be modified, so there is no problem of complicating the hardware structure, no problem of affecting reliability, and also no need to adjust the working parameters of the CD-SEM, which will not cause damage to the sample and will not affect the image quality of the detection.

[0102] The present application also provides a parameter prediction system for CD-SEM measurement, which comprises a prediction module for predicting the aspect ratio and / or height of the device structure according to the SE signal of the device structure and the parameter prediction model obtained by the parameter prediction model construction method for CD-SEM measurement of any one of the above embodiments.

[0103] In some embodiments, the embodiments of the present application provide a non-volatile computer readable storage medium, in which one or more programs including execution instructions are stored, which can be read and executed by an electronic device (including but not limited to a computer, a server, or a network device, etc.) to execute the parameter prediction model construction method for CD-SEM measurement of any one of the above embodiments of the present application.

[0104] In some embodiments, the embodiments of the present application also provide a computer program product, which comprises a computer program stored on a non-volatile computer readable storage medium, and the computer program comprises program instructions, which, when executed by a computer, cause the computer to execute the parameter prediction model construction method for CD-SEM measurement of any one of the above embodiments.

[0105] In some embodiments, the electronic device also provides a kind of electronic device, it includes: at least one processor, and the memory of communication connection with the at least one processor, wherein the memory has the instruction that can be executed by the at least one processor, the instruction is executed by the at least one processor, to enable the at least one processor to execute the parameter prediction model construction method for CD-SEM measurement of any one embodiment described above.

[0106] In some embodiments, the electronic device also provides a kind of electronic device, it includes: at least one processor, and the memory of communication connection with the at least one processor, wherein the memory has the instruction that can be executed by the at least one processor, the instruction is executed by the at least one processor, to enable the at least one processor to execute the parameter prediction model construction method for CD-SEM measurement of any one embodiment described above.

[0107] Figure 17 It is another embodiment of the application to provide the hardware structure schematic diagram of the electronic device for executing the parameter prediction model construction method for CD-SEM measurement, as shown in Figure Figure 17 The device includes:

[0108] One or more processors 310 and memory 320, Figure 17 In one processor 310 as an example.

[0109] The device for executing the parameter prediction model construction method for CD-SEM measurement can also include: input device 330 and output device 340.

[0110] Processor 310, memory 320, input device 330 and output device 340 can be connected by bus or other ways, Figure 17 In an example, it is connected by bus.

[0111] Memory 320 as a kind of non-volatile computer readable storage medium, it can be used to store non-volatile software programs, non-volatile computer executable programs and modules, such as the program instruction / module corresponding to the parameter prediction model construction method for CD-SEM measurement in the embodiments of the application.Processor 310 by running the non-volatile software program, instruction and module stored in memory 320, thereby executing the various functional applications and data processing of server, i.e. the parameter prediction model construction method for CD-SEM measurement of the method embodiment described above is realized.

[0112] The memory 320 can include a program storage area and a data storage area. The program storage area can store an operating system, application programs required by at least one function, etc. The data storage area can store data created according to the use of the parameter prediction model construction method for CD-SEM metrology, etc. In addition, the memory 320 can include a high-speed random access memory, and can also include a nonvolatile memory such as at least one magnetic disk storage device, flash memory device, or other nonvolatile solid-state memory device. In some embodiments, the memory 320 can optionally include a memory disposed remotely from the processor 310, which can be connected to the electronic device through a network. Examples of the network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0113] The input device 330 can receive input digital or character information, and generate signals related to user settings and function control of the image processing device. The output device 340 can include a display device such as a display screen.

[0114] The one or more modules are stored in the memory 320 and, when executed by the one or more processors 310, perform the parameter prediction model construction method for CD-SEM metrology in any of the method embodiments described above.

[0115] The above products can perform the methods provided in the embodiments of the present application, and have the corresponding function modules and beneficial effects of performing the methods. Technical details not described in detail in the embodiments can be referred to the methods provided in the embodiments of the present application.

[0116] The electronic device of the embodiments of the present application exists in various forms, including but not limited to:

[0117] (1) Mobile communication device: This type of device is characterized by having mobile communication function, and providing voice and data communication as the main target. This type of terminal includes: smart phone (such as iPhone), multimedia phone, functional phone, and low-end phone, etc.

[0118] (2) Ultra-mobile personal computer device: This type of device belongs to the category of personal computers, has computing and processing functions, and generally also has the characteristics of mobile Internet. This type of terminal includes: PDA, MID and UMPC devices, etc., such as iPad.

[0119] (3) Portable entertainment device: This type of device can display and play multimedia content. This type of device includes: audio and video player (such as iPod), handheld game console, electronic book, and smart toy and portable car navigation device.

[0120] (4) Server: a device providing computing services, the server is composed of a processor, a hard disk, a memory, a system bus, etc., the server is similar to a general computer architecture, but since it needs to provide high-reliable services, it has higher requirements in processing capability, stability, reliability, security, scalability, manageability, etc.

[0121] (5) Other electronic devices with data interaction function.

[0122] The device embodiments described above are only schematic, wherein the units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, i.e., can be located in one place, or can be distributed on multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the present embodiment.

[0123] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be realized by means of software plus a general hardware platform, or by hardware. Based on such understanding, the above technical solutions can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in each embodiment or some parts of the embodiments.

[0124] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements to some technical features; and these modifications or replacements do not make the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for generating a parameter prediction model for CD-SEM measurement, characterized in that, The application relates to a method for constructing a neural network model for predicting a parameter in CD-SEM measurement. The method comprises the following steps: acquiring a data set, wherein the data set comprises at least one group of data sets with proximity effect characteristics, each group of data sets comprises standard data and SE signal groups, the standard data comprises aspect ratio data and / or height data, and the SE signal groups in the data set with proximity effect characteristics are secondary electron signals generated by electron beam scanning on a geometric structure with adjacent structures, and the secondary electron signals contain changes of height parameters, side wall angle parameters and pitch parameters; 2. The method of claim 1, wherein, using the SE signal groups in the same group of data sets in the data set as input and using the standard data in the same group of data sets as output, training a pre-constructed neural network model to obtain a parameter prediction model for predicting aspect ratio data and / or height data in CD-SEM measurement. The method for constructing the neural network model comprises the following steps: constructing a network architecture of the neural network model based on an Identity module and a Dense module, wherein the Identity module is set to be composed of three basic units connected in sequence, each basic unit is sequentially connected by a full connection layer, a batch normalization layer and a ReLU activation function, and a Dropout layer is inserted after the second basic unit; the Dense module is set to be composed of three basic units connected in sequence, a Dropout layer is inserted after the second basic unit, and a residual path unit is additionally connected after the first basic unit and the second basic unit, respectively, the residual path unit is composed of a full connection layer and a batch normalization layer stacked in sequence, and the output of the residual path unit and the result between the batch normalization layer and the ReLU activation function in the third basic unit in the Dense module are configured to perform tensor sum operation; 3. The method of claim 2, wherein, the neural network model is constructed based on the constructed network architecture and a models library under a Keras library encapsulating TensorFlow, wherein, during construction, a compile function is used to compile the model, a loss function is specified as mse, Adam is specified as an optimizer of the neural network, and a metrics function is specified as mse.

4. The method of claim 1, wherein, The network architecture of the constructed neural network model is set to be composed of two basic modules connected in sequence, a batch normalization layer, a full connection layer and a linear activation function, wherein each basic module is formed by sequentially connecting two Identity modules after a Dense module.

5. The method of claim 1, wherein, The data set is set to be obtained by performing SEM simulation imaging on a simulated geometric structure with adjacent structures, obtaining standard data according to parameters of the constructed geometric structure, and obtaining SE signals according to parameters of the simulation imaging. The data set comprises a training set and a test set, and the training set and the test set are obtained by randomly dividing a plurality of data sets in the data set; during training of the constructed neural network model, the data sets in the training set are used to train the constructed neural network model to obtain model parameters, and the data sets in the test set are used to optimize the obtained model parameters.

6. The method of claim 5, wherein, The SE signal group in the same data group in the data set is input, and the standard data in the same data group is output, the pre-constructed neural network model is trained, and a parameter prediction model is obtained, including: According to the preset training times, the data group in the training set is selected, the SE signal group in the same data group is input, the standard data in the same data group is output, the pre-constructed neural network model is trained, and the first model parameter is obtained. The neural network model is updated according to the first model parameter to obtain a first neural network model; The SE signal group in the data group in the test set is input into the first neural network model, and the prediction result of the first neural network model to the input SE signal group is obtained; The error between the prediction result and the standard data in the corresponding data group is evaluated by using the loss function value, and the first neural network model is optimized according to the optimizer of the neural network and the error to obtain a parameter prediction model.

7. The method of claim 1, wherein, It also includes determining the prediction accuracy of the parameter prediction model according to the data set and the parameter prediction model, which includes: The SE signal group in the selected data group in the data set is input into the parameter prediction model as input, and the prediction data output by the parameter prediction model is obtained; The prediction accuracy of the parameter prediction model is determined by the metrics function according to the prediction data and the standard data in the data group where the input SE signal group is located.

8. A training device for a parameter prediction model used in CD-SEM measurement, characterized in that, It includes: The data acquisition module is used for acquiring a data set, the data set includes at least one data group with proximity effect characteristics, each data group includes standard data and an SE signal group, the standard data includes aspect ratio data and / or height data, the SE signal group in the data group with proximity effect characteristics is a secondary electron signal generated by electron beam scanning on a geometric structure with adjacent structure, and the secondary electron signal contains the change of height parameter, side wall angle parameter and spacing parameter; The training module is used for training a pre-constructed neural network model by inputting the SE signal group in the same data group in the data set and outputting the standard data in the same data group, to obtain a parameter prediction model for predicting aspect ratio data and / or height data in CD-SEM measurement.

9. A parameter prediction method for CD-SEM metrology, characterized in that, The parameter prediction model obtained by the parameter prediction model generation method for CD-SEM measurement according to any one of claims 1 to 7 and the SE signal of the device structure are used to predict the aspect ratio and / or height of the device structure.

10. A parameter prediction system for CD-SEM metrology, characterized in that, It includes: The prediction module is used for predicting the aspect ratio and / or height of the device structure according to the SE signal of the device structure and the parameter prediction model; The parameter prediction model is trained and generated by the method of any one of claims 1 to 7 or the device of claim 8.

Images