Semiconductor equipment electromagnetic compatibility modeling method and system combined with process parameters

Abstract

The invention relates to the technical field of semiconductor equipment, in particular to a semiconductor equipment electromagnetic compatibility modeling method and system combined with process parameters. The method comprises the following steps: acquiring key process parameters of semiconductor equipment, and preprocessing; a process neural network model is constructed, and the process neural network model is used for predicting and outputting electromagnetic interference characteristic data of the semiconductor equipment based on the key process parameters; a finite element simulation model is established, and the finite element simulation model takes electromagnetic interference characteristic data output by the process neural network model as a boundary condition and is used for simulating electromagnetic field distribution of the semiconductor equipment; and in combination with the process neural network model and the finite element simulation model, electromagnetic field dynamic distribution of the semiconductor equipment under different key process parameters is analyzed, and the electromagnetic interference influence is evaluated. According to the method, the dynamic correlation between the process parameters and the electromagnetic interference characteristics is established, and accurate prediction of the electromagnetic characteristics of the semiconductor equipment under different process conditions is realized.

Description

Technical Field

[0001] The present invention relates to the field of semiconductor equipment technology, and more specifically, to a semiconductor equipment electromagnetic compatibility modeling method and system combined with process parameters. Background Art

[0002] In the semiconductor manufacturing process, the electromagnetic compatibility (EMC) of PECVD (plasma enhanced chemical vapor deposition) thin film equipment is crucial to process stability and product quality. During the production process, PECVD thin film equipment needs to operate stably in a specific electromagnetic environment to ensure the uniformity and quality of thin film deposition. Therefore, it is necessary to model the EMC of PECVD thin film equipment to analyze the electromagnetic compatibility performance of PECVD thin film equipment.

[0003] Existing EMC modeling methods mostly focus on the physical characteristics of the equipment, such as the geometric structure, material properties or fixed operating parameters of the equipment, but fail to fully consider the dynamic impact of process parameters on the distribution of electromagnetic fields. These process parameters will continue to change during the production process, and their changes will directly affect the distribution and intensity of the electromagnetic field, which may potentially interfere with the electromagnetic compatibility performance of the equipment. Since the EMC model fails to effectively predict the potential interference of process changes on electromagnetic performance, it is impossible to accurately evaluate and control these interferences during the equipment design and optimization process, which affects the efficiency of equipment design and optimization.

[0004] However, existing modeling solutions are usually computationally intensive and cannot achieve real-time response to rapid process changes. In the semiconductor manufacturing process, process parameter adjustments often need to be completed in a very short time to adapt to production needs and improve efficiency. However, due to the high computational complexity, existing EMC modeling methods cannot complete the necessary calculations in a short time, and thus cannot provide real-time electromagnetic compatibility assessment.

[0005] Therefore, a semiconductor equipment electromagnetic compatibility modeling method is needed that can dynamically associate and respond in real time to the impact of process parameters on electromagnetic compatibility, so as to improve the stability and process adaptability of PECVD thin film equipment. Summary of the invention

[0006] The purpose of the present invention is to provide a method and system for electromagnetic compatibility modeling of semiconductor equipment combined with process parameters, so as to solve the problem that the electromagnetic compatibility modeling of semiconductor equipment in the prior art is difficult to dynamically associate process parameter changes and respond in real time to their impact on electromagnetic compatibility.

[0007] In order to achieve the above object, the present invention provides a method for electromagnetic compatibility modeling of semiconductor equipment combined with process parameters, comprising the following steps:

[0008] Collect key process parameters of semiconductor equipment and perform preprocessing;

[0009] Constructing a process neural network model, wherein the process neural network module is used to predict and output electromagnetic interference characteristic data of semiconductor equipment based on key process parameters;

[0010] Establishing a finite element simulation model, wherein the finite element simulation model uses the electromagnetic interference characteristic data output by the process neural network model as a boundary condition to simulate the electromagnetic field distribution of the semiconductor device;

[0011] Combining the process neural network model with the finite element simulation model, the dynamic distribution of the electromagnetic field of semiconductor equipment under different key process parameters is analyzed to evaluate the impact of electromagnetic interference.

[0012] In some embodiments, the semiconductor device is a thin film deposition device;

[0013] The key process parameters include film deposition rate, reaction gas flow rate and plasma power.

[0014] In some embodiments, the collected key process parameters are normalized.

[0015] In some embodiments, the process neural network model includes an input layer, a hidden layer and an output layer:

[0016] The input layer is used to receive key process parameters of semiconductor equipment;

[0017] The hidden layer performs nonlinear mapping through multiple neurons to capture the relationship between key process parameters and electromagnetic interference characteristics of the equipment;

[0018] The output layer is used to generate and output the electromagnetic interference characteristics of the semiconductor device under specific process parameter conditions.

[0019] In some embodiments, the hidden layer includes a first hidden sublayer and a second hidden sublayer:

[0020] The first hidden sublayer uses 128 neurons and uses the ReLU activation function;

[0021] The second hidden sublayer uses 64 neurons and the ReLU activation function.

[0022] In some embodiments, the training process of the process neural network model includes the following steps:

[0023] inputting key process parameters into the process neural network model;

[0024] The process neural network model outputs the electromagnetic interference characteristics of semiconductor devices through weighted calculation and activation function processing of each layer of neurons;

[0025] By comparing the predicted values ​​of the process neural network model with the true values, the loss function is calculated;

[0026] Use the Adam optimizer to adjust the network weights through the back-propagation algorithm to minimize the loss;

[0027] Continue to iterate the above process until the loss function converges or reaches the preset stopping condition.

[0028] In some embodiments, the loss function includes mean square error and weighted loss.

[0029] In some embodiments, the establishing of the finite element simulation model further comprises:

[0030] Based on the simulation analysis objectives, a simulation equation group is established;

[0031] Create a 3D geometric model based on the device structure and define electromagnetic properties for different areas;

[0032] Set boundary conditions and perform meshing;

[0033] Set solution parameters and simulate the electromagnetic field distribution of semiconductor devices.

[0034] In some embodiments, defining electromagnetic properties for different regions further comprises:

[0035] For the plasma region, the Drude model is used to simulate the plasma in the semiconductor device cavity, and the dielectric constant is calculated based on the set plasma frequency damping coefficient.

[0036] In some embodiments, the setting of boundary conditions and meshing further includes:

[0037] The cavity is divided into finite grid units, and the numerical solution of electromagnetic field distribution is performed in each unit. The grid unit types include tetrahedral units and hexahedral units.

[0038] In some embodiments, the step of setting solution parameters and simulating and calculating the electromagnetic field distribution of semiconductor devices further includes:

[0039] COMSOL Multiphysics simulation software is used to solve the electromagnetic field distribution in the semiconductor device cavity and output the interference field strength.

[0040] In some embodiments, the step of combining the process neural network model with the finite element simulation model to analyze the dynamic distribution of the electromagnetic field of the semiconductor device under different key process parameters and evaluate the impact of electromagnetic interference further includes:

[0041] Use sensitivity analysis methods to evaluate the impact of different key process parameters on electromagnetic compatibility performance;

[0042] Based on the sensitivity analysis results, determine the key process parameters that have the greatest impact on electromagnetic compatibility performance;

[0043] Optimize the operating range of key process parameters to reduce electromagnetic interference.

[0044] In some embodiments, the sensitivity analysis method includes the Sobol method;

[0045] Sensitivity analysis was performed using the Sobol method, which included the following steps:

[0046] Determine the model to be analyzed and clarify the input parameters involved;

[0047] Generate a series of sample data for the input parameters;

[0048] Use sample data to simulate the model and output the corresponding simulation results;

[0049] According to the output simulation results, the sensitivity index of each input parameter is calculated;

[0050] The calculated sensitivity indices were analyzed to assess the degree of influence of each input parameter on the model output.

[0051] In some embodiments, the sensitivity index includes a first-order sensitivity index, a total effect sensitivity index, and an interaction effect parameter:

[0052] The first-order sensitivity index is used to measure the independent contribution of a single input parameter;

[0053] The total effect sensitivity index is used to measure the total impact of all input parameters;

[0054] The interaction effect parameter is used to measure the interaction effect between two input parameters.

[0055] In order to achieve the above object, the present invention provides a semiconductor device electromagnetic compatibility modeling system combined with process parameters, including a memory and a processor:

[0056] The memory is used to store instructions executable by the processor;

[0057] The processor is used to execute the instructions to implement the above method.

[0058] In order to achieve the above object, the present invention provides a computer storage medium on which computer instructions are stored, wherein when the computer instructions are executed by a processor, the above method is executed.

[0059] The electromagnetic compatibility modeling method and system for semiconductor equipment combined with process parameters proposed in the present invention establish a dynamic relationship between process parameters and electromagnetic interference characteristics through the linkage of a process neural network model and a finite element simulation model, thereby achieving accurate prediction of the electromagnetic characteristics of semiconductor equipment under different process conditions, and ultimately achieving precise electromagnetic interference suppression and optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] The above and other features, properties and advantages of the present invention will become more apparent through the following description in conjunction with the accompanying drawings and embodiments, in which the same reference numerals always represent the same features, wherein:

[0061] Figure 1 A step diagram of a method for electromagnetic compatibility modeling of semiconductor equipment combined with process parameters according to an embodiment of the present invention is disclosed;

[0062] Figure 2 A method step diagram of a process for training a process neural network model according to an embodiment of the present invention is disclosed;

[0063] Figure 3 A method step diagram of a process for establishing a three-dimensional finite element simulation model according to an embodiment of the present invention is disclosed;

[0064] Figure 4 A step diagram of a sensitivity analysis method according to an embodiment of the present invention is disclosed;

[0065] Figure 5 A step diagram of a Sobol analysis method according to an embodiment of the present invention is disclosed. DETAILED DESCRIPTION

[0066] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the invention and are not used to limit the invention.

[0067] Figure 1 A step diagram of a method for modeling electromagnetic compatibility of semiconductor equipment combined with process parameters according to an embodiment of the present invention is disclosed, such as Figure 1 As shown, the present invention proposes a semiconductor device electromagnetic compatibility modeling method combined with process parameters, comprising the following steps:

[0068] Step S1, collecting key process parameters of semiconductor equipment and performing preprocessing;

[0069] Step S2, constructing a process neural network model, wherein the process neural network module is used to predict and output electromagnetic interference characteristic data of semiconductor equipment based on key process parameters;

[0070] Step S3, establishing a finite element simulation model, wherein the finite element simulation model uses the electromagnetic interference characteristic data output by the process neural network model as a boundary condition to simulate the electromagnetic field distribution of the semiconductor device;

[0071] Step S4, combining the process neural network model with the finite element simulation model, analyzing the dynamic distribution of the electromagnetic field of the semiconductor device under different key process parameters, and evaluating the impact of electromagnetic interference.

[0072] The electromagnetic compatibility modeling method of semiconductor equipment combined with process parameters proposed in the present invention provides a systematic solution for EMC optimization of semiconductor equipment by combining process parameters with electromagnetic field simulation and using the linkage of process neural network model and finite element simulation model, thereby improving the electromagnetic compatibility of semiconductor equipment and ultimately achieving accurate electromagnetic interference suppression and optimization.

[0073] These steps will be described in detail below. It should be understood that within the scope of the present invention, the above-mentioned technical features of the present invention and the technical features specifically described below (such as embodiments) can be combined and related to each other to form a preferred technical solution.

[0074] Step S1, collecting key process parameters of semiconductor equipment and performing preprocessing.

[0075] In this embodiment, the semiconductor device is a thin film deposition device.

[0076] The key process parameters of semiconductor equipment are collected through a sensor network, which includes flow meters, power meters, and temperature sensors.

[0077] The key process parameters include film deposition rate, reaction gas flow rate and plasma power:

[0078] The film deposition rate reflects the stability of film growth, and the film thickness is measured in real time by an interferometer, and the unit is nm / min;

[0079] The reaction gas flow rate directly affects the concentration and uniformity of the plasma and is collected by a mass flow meter (MFC) in units of sccm;

[0080] The plasma power has a significant impact on the spectrum characteristics of electromagnetic interference (EMI), and a high-frequency power meter is used to monitor the radio frequency power output in W.

[0081] In this embodiment, the step of performing preprocessing further includes:

[0082] The collected key process parameters are normalized.

[0083] More specifically, the min-max normalization method is used to normalize the key process parameters to eliminate the dimensional differences of different parameters.

[0084] The formula for min-max normalization is as follows:

[0085] X norm =(XX min ) / (X max -X min );

[0086] Where, X norm is the normalized key process parameter data, X is the original key process parameter data, and X min is the minimum value of the key process parameter, X max is the maximum value of the key process parameter.

[0087] In this embodiment, the step of performing preprocessing further includes:

[0088] Use a low-pass filter to remove high-frequency noise.

[0089] Step S2, constructing a process neural network model, wherein the process neural network module is used to predict and output electromagnetic interference characteristic data of semiconductor equipment based on key process parameters;

[0090] The Process Neural Network (PNN) model is a special neural network model, which is mainly used for modeling, prediction and optimization of process. It can learn the relationship between process parameters and electromagnetic interference characteristics based on historical data, so as to predict electromagnetic interference under different process conditions.

[0091] In the electromagnetic compatibility modeling of PECVD thin film equipment, the PNN model optimizes the equipment performance by simulating and predicting the relationship between process parameters and electromagnetic interference (EMI) characteristics.

[0092] In this embodiment, the constructed process neural network (PNN) model is a three-layer neural network structure, including an input layer, a hidden layer and an output layer:

[0093] The input layer is used to receive the key process parameters of semiconductor equipment. The characteristic number is 3, corresponding to three key process parameters, namely, film deposition rate, reaction gas flow rate and plasma power;

[0094] The hidden layer uses multiple neurons to perform nonlinear mapping to capture the complex relationship between process parameters and equipment electromagnetic interference characteristics;

[0095] In this embodiment, the hidden layer includes a first hidden sublayer and a second hidden sublayer:

[0096] The first hidden sublayer uses 128 neurons, which is responsible for extracting preliminary features from the input layer and processing them;

[0097] The second hidden sublayer uses 64 neurons to further process the output of the first layer and extract more advanced features;

[0098] Each layer uses the ReLU activation function to ensure the learning of nonlinear relationships and improve the expressiveness of the network.

[0099] Neurons are the basic units of neural networks, receiving input data (from the input layer or the previous layer) and obtaining results through weighted sum operations. The results are then further processed by activation functions to generate outputs.

[0100] In this embodiment, the activation function is ReLU. ReLU (Rectified Linear Unit) is a common activation function that performs nonlinear transformation on the input of neurons. ReLU helps to speed up the training process and effectively avoid the gradient vanishing problem (especially in deep networks).

[0101] The output layer maps the features of the hidden layer to the corresponding semiconductor devices and generates the electromagnetic interference characteristics of the semiconductor devices under specific process parameter conditions.

[0102] The electromagnetic interference characteristics include spectrum distribution and power density:

[0103] The frequency spectrum distribution reflects the intensity distribution of electromagnetic interference generated by the device at different frequencies;

[0104] The power density represents the energy distribution density of electromagnetic interference in space.

[0105] The loss function is used to evaluate the difference between the model's predicted value and the true value. The function is minimized during the optimization process to improve the model's prediction accuracy. The loss function includes mean square error and weighted loss.

[0106] Mean Squared Error (MSE) is a commonly used regression loss function that measures the difference between the model's predicted value and the true value. It is defined as the average of the squared prediction errors.

[0107] Weighted loss is an improved mean square error (MSE) method that increases the focus on key frequency bands or specific features by assigning weights to different prediction targets.

[0108] Compared with the mean square error, the weighted loss sets different weights to impose higher penalties on errors in important frequency bands during the optimization process, allowing the model to pay more attention to the prediction accuracy of key frequency bands or specific areas. The weighted loss function is particularly suitable for dealing with unequal importance distribution of target characteristics, such as focusing on a specific spectrum range in electromagnetic interference characteristic modeling.

[0109] In this embodiment, the mean square error (MSE) is used as the loss function; in addition to the mean square error (MSE), the loss function of the process neural network model can also select the mean absolute error (MAE, Mean Absolute Error), the mean square logarithmic error (MSLE, Mean Squared Logarithmic Error), the log-cosine similarity loss (Log-Cosine Loss), etc.

[0110] The process neural network model is trained using the Adam (Adaptive Moment Estimation) optimizer, and the weights of the neural network are updated through back propagation to minimize the loss function, thereby improving the prediction accuracy of the network.

[0111] The Adam optimizer is a commonly used adaptive learning rate optimization algorithm that combines the momentum method and adaptive learning rate to increase training speed and provide better performance in deep learning.

[0112] Figure 2 A method step diagram of a process for training a process neural network model according to an embodiment of the present invention is disclosed, such as Figure 2 As shown, the training process of the process neural network model in step S2 includes the following steps:

[0113] Step S21, inputting key process parameters into the process neural network model;

[0114] Step S22, the process neural network model outputs the electromagnetic interference characteristics of the semiconductor device through weighted calculation and activation function processing of each layer of neurons;

[0115] Step S23, by comparing the predicted value of the process neural network model with the true value, a loss function is calculated. In this embodiment, the loss function is a mean square error;

[0116] Step S24, using the Adam optimizer to adjust the network weights through the back propagation algorithm to minimize the loss;

[0117] Step S25, continue to iterate the above process until the loss function converges or reaches a preset stop condition.

[0118] Through this training method, the process neural network model can gradually learn the complex relationship between the input process parameters and the electromagnetic interference characteristics, and optimize its prediction ability, ultimately providing a reliable model for the electromagnetic compatibility analysis and optimization of PECVD equipment.

[0119] Step S3, establishing a finite element simulation (FEM) model, wherein the finite element simulation model uses the electromagnetic interference characteristic data output by the process neural network model as a boundary condition to simulate the electromagnetic field distribution of the semiconductor device;

[0120] In this embodiment, a three-dimensional finite element simulation model is established based on the actual structure of the semiconductor device;

[0121] Figure 3 A method step diagram of a process for establishing a three-dimensional finite element simulation model according to an embodiment of the present invention is disclosed, such as Figure 3 As shown, the process of establishing the three-dimensional finite element simulation model includes the following steps:

[0122] Step S31, establishing a simulation equation group based on the simulation analysis target;

[0123] Step S32, creating a three-dimensional geometric model according to the device structure and defining electromagnetic properties for different areas;

[0124] Step S33, setting boundary conditions and performing grid division;

[0125] Step S34, setting solution parameters, and simulating and calculating the electromagnetic field distribution of the semiconductor device.

[0126] More specifically, in step S31, the electromagnetic field distribution in the semiconductor device cavity is described by Maxwell's equations, which mainly include Faraday's law of electromagnetic induction and Ampere-Maxwell equations:

[0127] Faraday's law of electromagnetic induction, which describes how a time-varying magnetic field produces an electric field;

[0128] The Ampere-Maxwell equations describe how electric current and electric fields produce magnetic fields.

[0129] These equations are solved using the finite element method inside the semiconductor device cavity to determine the spatial distribution of the electromagnetic fields.

[0130] More specifically, the step S32 further comprises:

[0131] According to the actual structure of PECVD equipment, a three-dimensional geometric model is established;

[0132] Select appropriate materials for the different areas and define their electromagnetic properties.

[0133] Among them, the three-dimensional geometric model includes areas such as the device cavity, metal electrodes, and plasma area.

[0134] In geometric modeling, we try to maintain the accuracy of the actual structure while simplifying unimportant parts to reduce computational complexity.

[0135] Select appropriate materials for each area (such as cavity, metal electrode, plasma area, etc.) and define their electromagnetic properties (such as dielectric constant, magnetic permeability, and conductivity, etc.)

[0136] Among them, the metal electrode is assumed to be a perfect conductor (PEC), that is, its conductivity is very large and the penetration of the electric field can be ignored.

[0137] For the plasma region, the Drude model is used to simulate the plasma in the semiconductor device cavity, and the dielectric constant is calculated based on the set plasma frequency damping coefficient;

[0138] For non-magnetic materials, the magnetic permeability is usually taken as the vacuum permeability;

[0139] For a metal cavity, for example, the conductivity of copper is 5.96×10 7 S / m.

[0140] More specifically, the Drude model is a classic plasma physics model used to describe the response of free electrons to electromagnetic waves, and its dielectric constant is defined by the following formula:

[0141]

[0142] Among them, ω p is the plasma frequency; γ is the damping coefficient, which indicates the collision frequency of electrons; ∈0 is the vacuum dielectric constant; ω is the angular frequency of the electromagnetic wave;

[0143] ω p is the plasma frequency, which is related to the electron density n and the electron mass m;

[0144] More specifically, the plasma frequency ω p Defined as:

[0145]

[0146] Among them, n is the electron density; e is the electron charge; m is the electron mass.

[0147] The Drude model can accurately describe the dynamic response of plasma free electrons to electromagnetic waves in PECVD equipment, focusing on the collision dissipation effect (γ) and collective oscillation (ω p), which is an important feature of PECVD plasma. The Drude model provides an efficient electromagnetic field solution method through a simplified dielectric function form.

[0148] More specifically, the step S33 further comprises:

[0149] Inputting the electromagnetic interference characteristics output by the PNN model in step S2 into the finite element simulation model in step S3 as boundary conditions;

[0150] The cavity is divided into finite grid units so that the electromagnetic field distribution can be numerically solved in each unit. The grid unit types include tetrahedral units, hexahedral units, etc.

[0151] More specifically, the step S34 further comprises:

[0152] Choose a numerical solver suitable for electromagnetic field problems;

[0153] Set solution parameters;

[0154] Calculate the electric field intensity distribution, power density distribution and interference field intensity distribution in the cavity.

[0155] The solver includes a direct solver and an iterative solver:

[0156] The direct solver is suitable for small and medium-sized models, such as the PARDISO solver;

[0157] The iterative solver uses GMRES (generalized minimum residual method) or multigrid method to accelerate convergence for large-scale models.

[0158] The solution parameters include the solution frequency range, frequency step size, and iterative convergence conditions, etc.

[0159] More specifically, in step S34, based on the established three-dimensional finite element simulation model, COMSOL Multiphysics simulation software is used to solve the electromagnetic field distribution in the cavity, and the interference field strength is output, and the unit of the interference field strength is dBμV / m.

[0160] COMSOL Multiphysics is a software widely used in multi-physics simulation. It has significant advantages in solving coupling problems such as electromagnetic fields, thermal fields, and fluid fields. It not only supports efficient electromagnetic field solutions, but also can easily couple multi-physics fields (such as thermal effects and plasma dynamics) to meet the complex needs of semiconductor thin film devices. For some specific needs, more professional electromagnetic simulation software such as ANSYS HFSS or CST Studio Suite can also be considered. The specific choice needs to be weighed according to the project characteristics and goals.

[0161] Step S3 of this embodiment utilizes the finite element method, the Drude model and the output of the PNN model to construct an efficient and accurate electromagnetic field simulation model, which can simulate the electromagnetic interference characteristics in the cavity of the semiconductor device in detail and provide a reliable theoretical basis for the optimal design of the semiconductor device.

[0162] Step S4, combining the process neural network model with the finite element simulation model, analyzing the dynamic distribution of the electromagnetic field of the semiconductor device under different key process parameters, and evaluating the impact of electromagnetic interference.

[0163] In this step S4, through dynamic association and optimization, the EMC performance of the semiconductor device can be further improved, and the stable operation of the device under different process conditions can be ensured.

[0164] Figure 4 A sensitivity analysis method step diagram according to an embodiment of the present invention is disclosed. Figure 4 As shown, the step S4 further comprises:

[0165] Step S401, using a sensitivity analysis method to evaluate the impact of different key process parameters on electromagnetic compatibility performance;

[0166] Step S402, determining the key process parameters that have the greatest impact on electromagnetic compatibility performance according to the sensitivity analysis results;

[0167] Step S403, optimizing the operating range of key process parameters to reduce electromagnetic interference.

[0168] Furthermore, the sensitivity analysis method includes the Sobol method, which calculates the sensitivity index of the input parameters by decomposing the total variance, and evaluates the contribution of each input parameter and its interaction to the output.

[0169] Figure 5 A step diagram of a Sobol analysis method according to an embodiment of the present invention is disclosed, such as Figure 5 As shown, the sensitivity analysis was performed using the Sobol method, which included the following steps:

[0170] Step S411, determining the model and input parameters;

[0171] Identify the model to be analyzed and clarify the input parameters involved.

[0172] Determine the model that needs to be subjected to sensitivity analysis. In this embodiment, the models that need to be subjected to sensitivity analysis are a process neural network model and a finite element simulation model;

[0173] The input parameters are key process parameters, such as film deposition rate, reaction gas flow rate, plasma power, etc.

[0174] Changes in these input parameters may affect the output results of the model

[0175] Step S412: Generate samples of input parameters;

[0176] Generate a series of sample data for the input parameters.

[0177] Typically, Sobol sequences or Latin Hypercube Sampling (LHS) are used to generate random samples of input parameters to ensure the uniformity and coverage of the samples.

[0178] Step S413: Simulate the model and output the calculation results;

[0179] Use the sample data to simulate the model and output the corresponding simulation results.

[0180] For each set of input parameter samples, run the model and calculate the corresponding output results. The output simulation results can be electromagnetic interference characteristic data related to electromagnetic compatibility (EMC), such as electromagnetic interference field strength, power density, etc., and electromagnetic field distributions.

[0181] Step S414: Calculate the sensitivity index;

[0182] Calculate the sensitivity index of each input parameter based on the output simulation results.

[0183] The sensitivity index of the input parameter includes the first-order sensitivity index, the total effect sensitivity index, and the interaction effect parameter:

[0184] The first-order sensitivity index is used to measure the independent contribution of a single input parameter;

[0185] The total effect sensitivity index is used to measure the total impact of all input parameters, including their interaction with other input parameters;

[0186] The interaction effect parameter refers to the interaction effect between two input parameters and measures the contribution of the input parameter combination.

[0187] The first-order sensitivity index S i , corresponding expression:

[0188]

[0189] where Var is the total variance; X i is a single input parameter; Y is the output result; E(Y|X i ) is the conditional expectation of the output result Y when X i is fixed.

[0190] The total effect sensitivity index S Ti , corresponding expression:

[0191]

[0192] Among them, E(Y|X - i ) is the conditional expectation of the output result Y when X - i is fixed.

[0193] Interaction effect parameter S ij , and the corresponding expression:

[0194]

[0195] Among them, X i and X i are two input parameters.

[0196] Step S415, analyze the sensitivity index.

[0197] Analyze the calculated sensitivity index to evaluate the influence degree of each input parameter on the model output.

[0198] Based on the results of the sensitivity analysis, identify the key process parameters, and optimize the operating range or conditions of the key process parameters to reduce electromagnetic interference or improve electromagnetic compatibility.

[0199] Specifically, for the key process parameters with larger first - order sensitivity indices, it indicates that these key process parameters have a significant impact on the output. When optimizing, these key process parameters should be given priority, and the setting ranges of these parameters should be concentrated on for optimization.

[0200] If the interaction effects of two or more key process parameters are large, this means that the relationships between these parameters need to be jointly optimized to reduce the impact of their interaction effects on electromagnetic compatibility.

[0201] Evaluate the overall impact of all key process parameters through the total - effect sensitivity index to understand which parameters and their interactions have a greater comprehensive impact on the output.

[0202] The Sobol method can effectively help evaluate the influence of different input parameters on the model output and provide a scientific basis for subsequent optimization.

[0203] Next, a specific embodiment will be used to illustrate the semiconductor device electromagnetic compatibility modeling method combining process parameters proposed by the present invention.

[0204] In this embodiment, the semiconductor device is a PECVD device, mainly used for depositing silicon nitride thin films, with a frequency of 13.56 MHz.

[0205] In step S1, process parameters are collected to determine the key process parameters of the PECVD equipment to ensure the integrity and accuracy of the data.

[0206] Key process parameters are collected through a sensor network.

[0207] The key process parameters collected include:

[0208] film deposition rate, with values ​​ranging from 80–120 nm / min;

[0209] Reagent gas flow rate, with values ​​ranging from 40–60 sccm;

[0210] Plasma power, with values ​​ranging from 200–400W.

[0211] In addition, the data acquisition frequency is 10 times per second to ensure high dynamic response.

[0212] In step S1, a preprocessing step is performed, which includes normalization and filtering of the data.

[0213] In step S2, process neural network (PNN) modeling is performed, and the correlation between process parameters and electromagnetic interference characteristics is predicted by the process neural network model.

[0214] The process neural network model includes an input layer, a hidden layer and an output layer:

[0215] The input layer consists of three nodes, corresponding to the three key process parameters of film deposition rate, gas flow rate and plasma power;

[0216] The hidden layer includes the first hidden sublayer and the second hidden sublayer:

[0217] The first hidden sublayer consists of 128 nodes, and the activation function is ReLU.

[0218] The second hidden sublayer consists of 64 nodes, and the activation function is ReLU.

[0219] The output layer is used to predict the electromagnetic interference characteristics.

[0220] When training the input data of the process neural network model, 300 sets of historical data (equipment operation records) and 200 sets of simulation data (generated based on the FEM model) are used as the training data sources.

[0221] The training algorithm uses the Adam optimizer, and the learning rate is set to 0.001. The loss function is the mean square error (MSE), and the training accuracy reaches 98.5%, while the test set error is kept below 2%.

[0222] In this embodiment, the output of the process neural network model is a set of spectrum characteristics (power density distribution in the frequency range of 10 MHz-1 GHz).

[0223] In step S3, finite element modeling simulation is performed to simulate the electromagnetic field distribution inside the PECVD equipment, and dynamic analysis is performed in combination with the output of the PNN model.

[0224] The finite element simulation model includes the following:

[0225] A geometric model of the PECVD chamber to simulate the size and shape of the real device;

[0226] Material properties, involving the plasma model and the cavity metal.

[0227] In this embodiment, the plasma model is the Drude model, and the plasma frequency ω p

[0228] =10 12 HZ, damping coefficient γ = 10 9 HZ. The conductivity of the cavity metal is set to 5.96X10 7 S / m.

[0229] The boundary condition of the finite element simulation model is to use the frequency spectrum characteristics output by the PNN model as the input excitation source.

[0230] The solution method of the finite element simulation model is to use COMSOL Multiphysics software to perform frequency domain simulation and solve the electromagnetic field distribution.

[0231] Outputs from the finite element simulation model include:

[0232] Electric field strength distribution (unit: V / m).

[0233] Interference field strength distribution (unit: dBμV / m).

[0234] Analyze the output results of the finite element simulation model:

[0235] The maximum interference field strength is about 120dBμV / m, and its location is mainly concentrated near the electrodes in the cavity;

[0236] The electric field distribution cloud diagram shows the non-uniform characteristics of the plasma generation region.

[0237] Step S4, combining the joint analysis results of the process neural network model and the finite element simulation model to dynamically correlate and optimize the process parameters and equipment structure.

[0238] Furthermore, the sensitivity analysis method in step S4 uses the Sobol method to calculate the sensitivity index of each key process parameter to the electromagnetic compatibility performance:

[0239] Among them, the first-order sensitivity index of the film deposition rate is 0.35, which means that the independent contribution of the film deposition rate to the electric field intensity is 35%;

[0240] The first-order sensitivity index of gas flow is 0.45, which means that the independent contribution of gas flow to the electric field strength is 45%;

[0241] The first-order sensitivity index of plasma power is 0.20, which means that the independent contribution of plasma power is 20%;

[0242] The interaction effect parameter between film deposition rate and gas flow rate is 0.1, which means that there is a 10% interaction effect between film deposition rate and gas flow rate;

[0243] It can be seen that gas flow has the greatest impact on the electromagnetic field distribution.

[0244] Based on the results of sensitivity analysis, the process parameters were optimized.

[0245] The process parameters before optimization were set as: plasma power 300W, gas flow 40sccm, and film deposition rate 100nm / min. After optimization, the plasma power was adjusted to 250W, the gas flow was adjusted to 50sccm, and the film deposition rate was increased to 115nm / min. The electromagnetic interference field intensity after optimization was reduced by 15% compared with that before optimization, and the uniformity of film deposition was significantly improved to 95%.

[0246] Based on the above optimization strategy, the operating range of key process parameters is output and applied to the automatic control system of the equipment to achieve the optimization of real-time parameters.

[0247] Specifically, the operating interval of the film deposition rate was set to 110-120 nm / min;

[0248] The operating range of the gas flow rate is 45-55 sccm.

[0249] The operating range of plasma power is 240-260W.

[0250] The present invention proposes a semiconductor device electromagnetic compatibility modeling method combined with process parameters, which accurately predicts the electromagnetic interference characteristics of semiconductor devices under different process conditions by combining process neural network (PNN) and finite element simulation (FEM) methods. The PNN model can dynamically associate key process parameters with electromagnetic compatibility performance, revealing the complex nonlinear relationship between the two, and combined with sensitivity analysis methods (such as the Sobol method), quickly identify the key parameters that have the greatest impact on electromagnetic compatibility, providing a scientific basis for process optimization, thereby reducing the number of experimental iterations and improving optimization efficiency.

[0251] In addition, this method uses three-dimensional finite element simulation based on the actual structure of semiconductor equipment, comprehensively considers the material properties in the cavity and multi-field coupling effects (such as plasma dynamics and electromagnetic field distribution), and achieves accurate solution of electromagnetic field strength. The optimized process parameters effectively reduce the risk of electromagnetic interference and improve the electromagnetic compatibility of the equipment.

[0252] The electromagnetic compatibility modeling method of semiconductor equipment combined with process parameters proposed in the present invention is suitable for PECVD equipment. It has strong versatility and expansibility and can be extended to the electromagnetic compatibility design of other semiconductor equipment, providing the industry with efficient and low-cost modeling and optimization solutions.

[0253] Although the above methods are illustrated and described as a series of actions for simplicity of explanation, it should be understood and appreciated that these methods are not limited by the order of the actions, because according to one or more embodiments, some actions may occur in a different order and / or concurrently with other actions from those illustrated and described herein or not illustrated and described herein but understandable to those skilled in the art.

[0254] The present invention also provides a semiconductor device electromagnetic compatibility modeling system combined with process parameters. The semiconductor device electromagnetic compatibility modeling system combined with process parameters includes a memory and a processor, the memory can store a program, and the processor can execute the program. When the processor executes the program, the semiconductor device electromagnetic compatibility modeling method combined with process parameters provided by the present invention is implemented.

[0255] The present invention also provides a computer-readable storage medium, which is a non-volatile storage medium or a non-transient storage medium, on which computer instructions are stored. When the computer instructions are executed, the steps corresponding to any of the above methods are executed, which will not be repeated here.

[0256] When the implementation process file of the electromagnetic compatibility modeling method of semiconductor equipment combined with process parameters is a computer program, it can also be stored in a computer-readable storage medium as a product. For example, a computer-readable storage medium may include, but is not limited to, a magnetic storage device (e.g., a hard disk, a floppy disk, a magnetic strip), an optical disk (e.g., a compact disk (CD), a digital versatile disk (DVD)), a smart card, and a flash memory device (e.g., an electrically erasable programmable read-only memory (EPROM), a card, a stick, a key drive). In addition, the various storage media described herein can represent one or more devices and / or other machine-readable media for storing information. The term "machine-readable medium" may include, but is not limited to, wireless channels and various other media (and / or storage media) that can store, contain and / or carry code and / or instructions and / or data.

[0257] The present invention provides a semiconductor device electromagnetic compatibility modeling method and system combined with process parameters, which specifically has the following beneficial effects:

[0258] 1) Improve modeling accuracy and realize dynamic correlation between process parameters and electromagnetic characteristics. The present invention introduces process parameters (such as film deposition rate, gas flow rate, plasma power) into electromagnetic compatibility (EMC) modeling, establishes a dynamic correlation between process parameters and electromagnetic interference characteristics, and realizes accurate prediction of electromagnetic characteristics under different process conditions;

[0259] 2) Optimize modeling efficiency and process stability. The present invention combines process neural network (PNN) with finite element method (FEM), uses PNN for fast dynamic response, reduces the computational burden of complex electromagnetic field simulation, and adjusts the global electromagnetic field distribution through FEM, thereby optimizing modeling efficiency and process compatibility, significantly shortening modeling time, and improving the real-time performance of electromagnetic characteristic simulation of semiconductor equipment.

[0260] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0261] Those skilled in the art will appreciate that information, signals, and data may be represented using any of a variety of different technologies and techniques. For example, data, instructions, commands, information, signals, bits, symbols, and chips cited throughout the above description may be represented by voltage, current, electromagnetic waves, magnetic fields or magnetic particles, optical fields or optical particles, or any combination thereof.

[0262] Those skilled in the art will further appreciate that the various illustrative logic blocks, modules, circuits, and algorithm steps described in conjunction with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or a combination of the two. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps are generally described above in terms of their functionality. Whether such functionality is implemented as hardware or software depends on the specific application and the design constraints imposed on the overall system. The technician may implement the described functionality in different ways for each specific application, but such implementation decisions should not be interpreted as resulting in a departure from the scope of the present invention.

[0263] The various illustrative logic modules and circuits described in conjunction with the embodiments disclosed herein may be implemented or executed with a general purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general purpose processor may be a microprocessor, but in the alternative, the processor may be any conventional processor, controller, microcontroller, or state machine. The processor may also be implemented as a combination of computing devices, such as a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors in cooperation with a DSP core, or any other such configuration.

[0264] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. The software module may reside in a RAM memory, a flash memory, a ROM memory, an EPROM memory, an EEPROM memory, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. An exemplary storage medium is coupled to a processor so that the processor can read and write information from / to the storage medium. In an alternative, a storage medium may be integrated into a processor. The processor and the storage medium may reside in an ASIC. The ASIC may reside in a user terminal. In an alternative, the processor and the storage medium may reside in a user terminal as discrete components.

[0265] The above embodiments are provided for persons familiar with the art to implement or use the present invention. Persons familiar with the art can make various modifications or changes to the above embodiments without departing from the inventive concept of the present invention. Therefore, the protection scope of the present invention is not limited to the above embodiments, but should be the maximum scope of the innovative features mentioned in the claims.

Claims

1. A semiconductor device electromagnetic compatibility modeling method combined with process parameters, characterized in that: The following steps are involved: Collect key process parameters of semiconductor equipment and perform preprocessing; Constructing a process neural network model, wherein the process neural network module is used to predict and output electromagnetic interference characteristic data of semiconductor equipment based on key process parameters; Establishing a finite element simulation model, wherein the finite element simulation model uses the electromagnetic interference characteristic data output by the process neural network model as a boundary condition to simulate the electromagnetic field distribution of the semiconductor device; Combining the process neural network model with the finite element simulation model, the dynamic distribution of the electromagnetic field of semiconductor equipment under different key process parameters is analyzed to evaluate the impact of electromagnetic interference.

2. The method for electromagnetic compatibility modeling of semiconductor equipment combined with process parameters according to claim 1, characterized in that: The semiconductor device is a thin film deposition device; The key process parameters include film deposition rate, reaction gas flow rate and plasma power.

3. The method for electromagnetic compatibility modeling of semiconductor equipment combined with process parameters according to claim 1, characterized in that: The step of pre-processing further comprises: The collected key process parameters are normalized.

4. The method for electromagnetic compatibility modeling of semiconductor equipment combined with process parameters according to claim 1, characterized in that: The process neural network model includes an input layer, a hidden layer and an output layer: The input layer is used to receive key process parameters of semiconductor equipment; The hidden layer performs nonlinear mapping through multiple neurons to capture the relationship between key process parameters and electromagnetic interference characteristics of the equipment; The output layer is used to generate and output the electromagnetic interference characteristics of the semiconductor device under specific process parameter conditions.

5. The method for electromagnetic compatibility modeling of semiconductor equipment combined with process parameters according to claim 4, characterized in that: The hidden layer includes a first hidden sublayer and a second hidden sublayer: The first hidden sublayer uses 128 neurons and uses the ReLU activation function; The second hidden sublayer uses 64 neurons and the ReLU activation function.

6. The method for electromagnetic compatibility modeling of semiconductor equipment combined with process parameters according to claim 1, characterized in that: The training process of the process neural network model includes the following steps: inputting key process parameters into the process neural network model; The process neural network model outputs the electromagnetic interference characteristics of semiconductor devices through weighted calculation and activation function processing of each layer of neurons; By comparing the predicted values ​​of the process neural network model with the true values, the loss function is calculated; Use the Adam optimizer to adjust the network weights through the back-propagation algorithm to minimize the loss; Continue to iterate the above process until the loss function converges or reaches the preset stopping condition.

7. The method for electromagnetic compatibility modeling of semiconductor equipment combined with process parameters according to claim 6, characterized in that: The loss function includes mean square error and weighted loss.

8. The method for electromagnetic compatibility modeling of semiconductor equipment combined with process parameters according to claim 1, characterized in that: The establishing of the finite element simulation model further comprises: Based on the simulation analysis objectives, a simulation equation group is established; Create a 3D geometric model based on the device structure and define electromagnetic properties for different areas; Set boundary conditions and perform meshing; Set solution parameters and simulate the electromagnetic field distribution of semiconductor devices.

9. The method for electromagnetic compatibility modeling of semiconductor equipment combined with process parameters according to claim 8, characterized in that: Defining electromagnetic properties for different regions further includes: For the plasma region, the Drude model is used to simulate the plasma in the semiconductor device cavity, and the dielectric constant is calculated based on the set plasma frequency damping coefficient.

10. The method for electromagnetic compatibility modeling of semiconductor equipment combined with process parameters according to claim 8, characterized in that: The step of setting boundary conditions and performing grid division further includes: The cavity is divided into finite grid units, and the numerical solution of electromagnetic field distribution is performed in each unit. The grid unit types include tetrahedral units and hexahedral units.

11. The method for electromagnetic compatibility modeling of semiconductor equipment combined with process parameters according to claim 8, characterized in that: The step of setting solution parameters and simulating and calculating the electromagnetic field distribution of semiconductor devices further includes: COMSOL Multiphysics simulation software is used to solve the electromagnetic field distribution in the semiconductor device cavity and output the interference field strength.

12. The method for electromagnetic compatibility modeling of semiconductor equipment combined with process parameters according to claim 1, characterized in that: The step of combining the process neural network model with the finite element simulation model to analyze the dynamic distribution of the electromagnetic field of the semiconductor device under different key process parameters and evaluate the impact of electromagnetic interference further includes: Use sensitivity analysis methods to evaluate the impact of different key process parameters on electromagnetic compatibility performance; Based on the sensitivity analysis results, determine the key process parameters that have the greatest impact on electromagnetic compatibility performance; Optimize the operating range of key process parameters to reduce electromagnetic interference.

13. The method for electromagnetic compatibility modeling of semiconductor equipment combined with process parameters according to claim 12, characterized in that: The sensitivity analysis method includes the Sobol method; Sensitivity analysis was performed using the Sobol method, which included the following steps: Determine the model to be analyzed and clarify the input parameters involved; Generate a series of sample data for the input parameters; Use sample data to simulate the model and output the corresponding simulation results; According to the output simulation results, the sensitivity index of each input parameter is calculated; The calculated sensitivity indices were analyzed to assess the degree of influence of each input parameter on the model output.

14. The method for electromagnetic compatibility modeling of semiconductor equipment combined with process parameters according to claim 13, characterized in that: The sensitivity index includes the first-order sensitivity index, the total effect sensitivity index and the interaction effect parameter: The first-order sensitivity index is used to measure the independent contribution of a single input parameter; The total effect sensitivity index is used to measure the total impact of all input parameters; The interaction effect parameter is used to measure the interaction effect between two input parameters.

15. A semiconductor device electromagnetic compatibility modeling system combined with process parameters, characterized in that: Including memory and processor: The memory is used to store instructions executable by the processor; The processor is used to execute the instructions to implement the method according to any one of claims 1-14.

16. A computer storage medium having computer instructions stored thereon, wherein when the computer instructions are executed by a processor, the method according to any one of claims 1 to 14 is executed.

Images