Self-alignment process abnormity prediction method for silicon carbide high-voltage MOSFET

Abstract

The invention relates to the technical field of semiconductors, and provides a self-alignment process anomaly prediction method for a silicon carbide high-voltage MOSFET. The method comprises the following steps: acquiring deposition parameters of a gate oxide layer, generating a simulated gate oxide layer, and performing primary anomaly prediction to obtain a primary anomaly probability; acquiring a gate metal deposition parameter, and performing secondary anomaly prediction in combination with the simulated gate oxide layer to obtain a secondary anomaly probability; acquiring source and drain doping parameters, configuring a doping proportion according to the secondary abnormal probability, and predicting a doping abnormal probability; and calculating a prediction abnormal probability according to the primary, secondary and doping abnormal probabilities, and taking the prediction abnormal probability as a self-alignment process prediction result. According to the method, the technical problem that the abnormity is difficult to predict and control due to the complexity of multiple process steps in the self-alignment process of the silicon carbide high-voltage MOSFET is solved, and the accuracy and the prediction capability of process abnormity identification are improved through a multi-step combined abnormity prediction method, so that the effects of optimizing the manufacturing quality and the device performance are achieved.

Description

Technical Field

[0001] The present application relates to the field of semiconductor technology, and in particular to a self-aligned process anomaly prediction method for silicon carbide high-voltage MOSFETs. Background Art

[0002] Silicon carbide (SiC), as a wide bandgap semiconductor material, is increasingly used in high-voltage, high-power and high-frequency devices. In particular, in high-voltage MOSFET devices, silicon carbide has become an important material for realizing high-efficiency and miniaturized power electronic systems due to its excellent high-temperature resistance, high breakdown field strength and low conduction loss. However, in the manufacturing process of SiC high-voltage MOSFET, due to its unique material properties and precision manufacturing requirements, there are many technical difficulties, especially in the formation of the gate oxide layer, the deposition of the gate metal, and the doping of the source and drain regions. Abnormalities are prone to occur in key process steps. These abnormalities may lead to a decline in device performance and even affect its reliability.

[0003] As a high-precision manufacturing technology, the self-alignment process can effectively reduce alignment errors and improve the integration and performance of devices by accurately positioning the source and drain using the formed gate structure. However, the self-alignment process involves multiple complex manufacturing steps, and the parameter selection and control of each step directly affect the quality of the final device. For example, the thickness uniformity of the gate oxide layer, the integrity of the gate metal deposition, and the precise positioning of the doped area are all critical to the performance of SiC high-voltage MOSFETs. Due to the interdependence between the steps, any abnormality in a single link may trigger a chain reaction, resulting in the failure of the entire batch of devices.

[0004] At present, the anomaly prediction and control methods for the SiC high-voltage MOSFET self-alignment process are still relatively limited. Traditional quality control mainly relies on post-process detection, which has the problem of delayed anomaly detection and inability to prevent in advance. In addition, due to the high manufacturing cost of SiC materials and high-voltage MOSFET devices, the scrap caused by anomalies not only causes economic losses, but also affects production efficiency. Therefore, how to predict possible anomalies in real time during the process and provide a quantitative abnormality probability reference has become a key technical issue in improving production quality and process stability. Summary of the invention

[0005] The present application provides a self-alignment process anomaly prediction method for silicon carbide high-voltage MOSFET, aiming to solve the technical problem that anomalies of silicon carbide high-voltage MOSFET are difficult to predict and control due to the complexity of multiple process steps in the self-alignment process.

[0006] In view of the above problems, the present application provides a self-aligned process anomaly prediction method for silicon carbide high-voltage MOSFET.

[0007] The present application provides a self-aligned process anomaly prediction method for silicon carbide high-voltage MOSFETs, the method comprising: in the process of manufacturing silicon carbide high-voltage MOSFETs using a self-aligned process, obtaining deposition parameters of a gate oxide layer, simulating the generation of the gate oxide layer, obtaining a simulated gate oxide layer, performing a primary anomaly prediction, and obtaining a primary anomaly probability; obtaining deposition parameters of a gate metal layer deposition, combining the simulated gate oxide layer, performing a secondary anomaly prediction, and obtaining a secondary anomaly probability; obtaining doping parameters for self-aligned doping of a source and a drain, configuring a doping anomaly prediction ratio according to the secondary anomaly probability, performing an integrated doping anomaly probability prediction according to the doping parameters, and obtaining a doping anomaly probability; and calculating a predicted anomaly probability according to the primary anomaly probability, the secondary anomaly probability, and the doping anomaly probability as a self-aligned process anomaly prediction result.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages: The above-mentioned self-aligned process anomaly prediction method for silicon carbide high-voltage MOSFET obtains the deposition parameters of the gate oxide layer, simulates and generates the gate oxide layer through these parameters, and uses the generated gate oxide layer to make an anomaly prediction of the process, thereby obtaining the probability that there may be problems with the gate oxide layer. This step can effectively identify the impact that may be caused by uneven oxide layer thickness or fluctuations in deposition conditions, laying a stable foundation for subsequent processes; subsequently, by obtaining the deposition parameters of the gate metal layer and combining it with the gate oxide layer generated in the previous step, a secondary anomaly prediction is made to accurately analyze the quality problems of the gate metal deposition, identify potential problems in the metal layer thickness and coverage integrity, and thus reduce the gate metal deposition caused by the gate metal deposition. The electrical performance fluctuations caused by defects can be improved to improve the consistency of the process; then, for the self-aligned doping link of the source and drain, the corresponding doping parameters are obtained, and the doping anomaly prediction ratio is dynamically configured according to the anomaly prediction results of the first two steps. These configuration ratios are further combined with the doping parameters to perform integrated doping anomaly probability prediction, which helps to evaluate in real time whether the doping area is accurately located; finally, the primary anomaly probability, secondary anomaly probability and doping anomaly probability are combined to calculate the overall process anomaly prediction result. This multi-step joint prediction method can not only deeply identify potential problems in each link, but also dynamically adjust in the entire process flow to improve process reliability and device consistency, thereby optimizing production quality and efficiency.

[0009] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0011] Figure 1 The figure is a flow chart of a method for predicting a self-aligned process abnormality for silicon carbide high-voltage MOSFET in one embodiment.

[0012] Figure 2 The present invention is a schematic diagram of a process of obtaining a simulated gate oxide layer in a self-aligned process anomaly prediction method for silicon carbide high-voltage MOSFET in one embodiment. DETAILED DESCRIPTION

[0013] The embodiment of the present application solves the technical problem that the abnormalities of silicon carbide high-voltage MOSFET are difficult to predict and control due to the complexity of multiple process steps in the self-alignment process by providing a self-alignment process abnormality prediction method for silicon carbide high-voltage MOSFET.

[0014] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0015] It should be noted that the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or inherent to these processes, methods, products or devices.

[0016] Examples, such as Figure 1 As shown, the present application provides a self-aligned process anomaly prediction method for silicon carbide high-voltage MOSFET, the method comprising: In the process of manufacturing silicon carbide high-voltage MOSFET by using a self-aligned process, deposition parameters of a gate oxide layer are obtained, simulation generation of the gate oxide layer is performed, a simulated gate oxide layer is obtained, an abnormality prediction is performed, and an abnormality probability is obtained.

[0017] In an embodiment of the present application, in the self-aligned process manufacturing of silicon carbide high-voltage MOSFET, first, by obtaining the deposition parameters of the gate oxide layer, a generative adversarial network is used to simulate the thickness distribution of the oxide layer to form a simulated gate oxide layer containing thickness information of multiple gate regions. This simulated gate oxide layer not only restores the thickness distribution characteristics after actual deposition, but can also be used in subsequent steps to detect and evaluate the uniformity of the oxide layer thickness. The uniformity is detected because if the oxide layer thickness is unevenly distributed, it may have a serious impact on subsequent process links. When the oxide layer thickness is thin in some areas, the subsequent metal layer deposition may not be able to completely cover it, and even the doped regions of the source and drain will invade the inside of the gate, resulting in electrical isolation failure. This situation will directly affect the performance of the MOSFET, including increased leakage current and deterioration of switching characteristics. To this end, through the simulation and thickness distribution analysis of the generative adversarial network, an abnormal prediction can be performed at an early stage, and the possible oxide layer unevenness problem can be accurately identified by combining a variation classifier to generate an abnormal probability. The completion of this step not only helps to intervene in potential manufacturing defects in advance, but also provides a reliable data reference for subsequent gate metal layer deposition and doping processes, thereby ensuring the stability and consistency of device performance from the source.

[0018] Further, if Figure 2 As shown, the present application provides a method for obtaining deposition parameters of a gate oxide layer, performing simulation generation of a gate oxide layer, and obtaining a simulated gate oxide layer, including: The deposition parameters of the gate oxide layer are obtained as the gate oxide layer deposition parameters; based on the generative adversarial network, an oxide layer simulation channel is trained; the gate oxide layer deposition parameters are input into the oxide layer simulation channel, and a simulated gate oxide layer is obtained by simulation generation.

[0019] Preferably, relevant parameters for depositing the gate oxide layer are collected from the actual production process as gate oxide deposition parameters, which include but are not limited to oxidation temperature, time, atmosphere composition (such as oxygen concentration), substrate surface cleanliness and other influencing factors of the deposition environment. The accurate acquisition of these parameters is the basis for subsequent simulation generation; then, the collected gate oxide deposition parameters are input into a pre-constructed oxide simulation channel for simulated generation of the gate oxide layer. This oxide simulation channel is constructed based on a generative adversarial network, which includes a generator and an adversary. The generator generates simulated data close to the actual thickness distribution by inputting the gate oxide deposition parameters, and the discriminator determines the difference between the thickness distribution output by the generator and the actual thickness distribution, thereby forming supervisory feedback for the generator; by inputting the gate oxide deposition parameters obtained in the actual process into the trained oxide simulation channel, the generator outputs a simulated gate oxide thickness distribution according to the input parameters, covering the entire gate area, and this distribution provides thickness information of the oxide layer in each area, providing accurate basic data for subsequent process optimization and abnormality prediction, while reducing the time cost and material waste of process experiments.

[0020] Furthermore, the present application provides a method for training an oxide layer simulation channel based on a generative adversarial network, including: Based on the gate oxide deposition data in historical time, a sample gate oxide deposition parameter set is collected, and the thickness distribution of the gate oxide obtained by deposition under different sample gate oxide deposition parameters is collected and marked as a sample gate oxide set; based on a generative adversarial network, an oxide simulation channel is constructed, wherein the oxide simulation channel includes a generator and an adversary; using the sample gate oxide deposition parameter set and the sample gate oxide set, the generator and the adversary are alternately supervised and trained until convergence, so as to obtain a trained oxide simulation channel.

[0021] Optionally, a set of gate oxide deposition parameters is extracted from production data in historical time as a sample gate oxide deposition parameter set, including oxidation temperature, time, atmosphere composition, etc. These parameters should cover different process conditions to ensure the diversity and comprehensiveness of the sample data. Under each set of deposition parameter conditions, the corresponding oxide thickness distribution data is collected, including the thickness information of each gate region, and these thickness data are annotated as a sample gate oxide set; then, based on a generative adversarial network (GAN), an oxide simulation channel is constructed, which includes a generator and a discriminator. The generator is designed to generate an oxide thickness distribution that is close to the real one based on the input deposition parameters, while the discriminator is responsible for determining whether the input data is a real thickness distribution or a generated thickness distribution. During the training process, the parameters of the generator and the discriminator are randomly initialized, and related hyperparameters are set, such as learning rate, batch size, etc. In each round of training, the sample gate oxide deposition parameter set is first input into the generator to generate an initial simulated thickness distribution, and then the generated simulated thickness distribution is input into the adversary together with the real sample thickness distribution, and the probability output of the discriminator for the authenticity of the data is calculated. The discriminator optimizes the parameters through the cross entropy loss function so that it can accurately distinguish between real data and generated data. At the same time, the generator parameters are updated by backpropagation so that the output of the generator gradually approaches the real data distribution. The training process is carried out through the generator. The generator and the discriminator are optimized iteratively, the generator constantly tries to deceive the discriminator, and the discriminator continuously improves its ability to distinguish between real and generated data. As the training progresses, the output quality of the generator gradually improves, and the difference between the final generated thickness distribution and the real thickness distribution gradually narrows. The convergence conditions of the training include that the distribution difference between the generator output data and the real data reaches a preset threshold, or the discrimination accuracy of the discriminator on the real data and the generated data is close to random guessing (about 50%). When the convergence conditions are met, the trained oxide simulation channel can be used to input any gate oxide deposition parameters to generate high-quality simulated gate oxide thickness distribution. This simulation channel provides a basis for subsequent process optimization and anomaly prediction, and realizes a high degree of digitization of the deposition process through deep learning technology.

[0022] Furthermore, the present application provides a method for performing an abnormal prediction and obtaining an abnormal probability, including: According to the simulated gate oxide thickness distribution, the gate oxide thickness variance is calculated; based on the gate oxide deposition data in the historical time, a sample gate oxide thickness variance set is collected, and according to the proportion of abnormalities in the silicon carbide high-voltage MOSFET under different sample gate oxide thickness variances, a sample primary abnormality probability set is identified; using the sample gate oxide thickness variance set and the sample primary abnormality probability set, a primary variation classifier is constructed; the gate oxide thickness variance is input into the primary variation classifier, and the primary abnormality probability is obtained by classification prediction.

[0023] Optionally, the thickness distribution in the gate oxide layer is simulated to calculate its thickness variance. The simulated thickness distribution data in the gate oxide layer is derived from the output of the oxide layer simulation channel, and includes thickness information of multiple gate regions. The calculation process uses a statistical method to extract the thickness values ​​of all gate regions and calculate the square of their standard deviations to obtain the thickness variance of the gate oxide layer. This variance value can quantify the uniformity of the oxide layer thickness, thereby reflecting the stability of the deposition process. Subsequently, based on the gate oxide deposition data collected over a historical period of time, the corresponding oxide layer thickness variance is recorded to construct a sample gate oxide thickness variance set. These sample data cover a variety of thickness distribution characteristics. At the same time, according to these thickness variance data, the proportion of abnormalities of the corresponding silicon carbide high-voltage MOSFET is counted to mark the sample primary abnormality probability set. This set is obtained by analyzing the abnormality occurrence rate under each group of variance values, which can provide a clear mapping relationship for subsequent abnormality prediction. After that, a mutation classifier is constructed using the sample gate oxide thickness variance set and the sample primary abnormality probability set. The classifier is constructed using a machine learning algorithm, such as a support vector machine (SVM), a random forest or a neural network. Taking the fully connected neural network (FCNN) in the neural network as an example, the fully connected neural network (FCNN) is used. The structure of a mutation classifier is constructed through the network, including an input layer, a hidden layer, an output layer, etc. The input layer is used to receive the gate oxide thickness variance data, the output layer generates a predicted primary anomaly probability value, and the hidden layer captures the complex relationship between thickness variance and anomaly probability through multi-layer nonlinear transformations. The weights and biases of the network are initialized with random numbers to give the network a certain degree of randomness and avoid the symmetry problem caused by the initial weights. The collected gate oxide thickness variance set is used as the input feature, and the sample primary anomaly probability set is used as the target output and input into the initialized primary mutation classifier; during forward propagation, the input data enters the neural network through the input layer The network is transformed layer by layer through weighted operations and nonlinear activation functions (such as ReLU) of the hidden layer to extract features. In the output layer, the result is converted into a probability distribution through the Softmax function, and the probability of an abnormality corresponding to the variance of the current gate oxide thickness is output. The binary cross entropy loss function is then used to calculate the loss between the predicted probability of an abnormality and the probability of an abnormality of the sample. The gradient of the loss function for the weight and bias of each layer is calculated layer by layer through the back propagation algorithm, and the network parameters are adjusted to reduce the loss value. The Adam optimizer is used to optimize the network, and the weight and bias of the network are adjusted using the adaptive learning rate to make the network converge to the global optimum faster.Repeat the above forward propagation, loss calculation, back propagation and parameter optimization process until the maximum number of iterations is reached or the loss value no longer decreases significantly. After the training is completed, the performance of the current single-variation classifier is tested using data not used for training, and the accuracy of the single-variation classifier in predicting the probability of a single abnormality is evaluated. If the performance of the classifier on the data not used for training reaches the preset accuracy target, the currently trained single-variation classifier is output. If the performance does not meet expectations, the hyperparameters such as learning rate, number of hidden layers and number of neurons are adjusted according to the feedback of the data not used for training, and the model is retrained until the requirements are met. Finally, the actual generated gate oxide thickness variance value is input into the trained single-variation classifier. The classifier predicts the probability of abnormality under the current thickness variance based on the mapping relationship learned from the training data. This single abnormality probability is used as an early indicator of the process and provides a reference for parameter adjustment in subsequent processes, which can effectively reduce device performance problems caused by uneven oxide thickness. ;

[0024] Deposition parameters of the gate metal layer deposition are obtained, and secondary anomaly prediction is performed in combination with the simulated gate oxide layer to obtain the secondary anomaly probability.

[0025] In one embodiment, when performing secondary anomaly prediction of silicon carbide high-voltage MOSFET, the deposition parameters of the gate metal layer are first obtained, including the thickness of the metal deposition material, deposition rate, temperature, vacuum environment, etc. These parameters directly affect the uniformity and coverage quality of the gate metal layer. At the same time, the previously generated simulated gate oxide layer is used as basic data and combined with the gate metal layer deposition parameters to be input into the secondary variation classifier for analyzing potential anomalies in the metal layer deposition process. In the prediction process, the simulated gate oxide layer provides basic information on the distribution of oxide layer thickness, while the gate metal layer deposition parameters reflect the deposition conditions of the metal layer in different areas. If the oxide layer thickness distribution is uneven, it may cause the metal layer to be deposited in thinner areas. At the same time, if there are deviations in the deposition rate and uniformity of the metal layer, the instability of the overall deposition may be further aggravated. These problems will directly affect the electrical performance of the MOSFET, for example, the leakage current will increase or the switching characteristics will deteriorate, resulting in abnormal performance. By combining the gate metal layer deposition parameters and the simulated gate oxide layer, the secondary variation classifier can evaluate the uniformity and completeness of the metal layer deposition, identify potential defects in the deposition process, and output the corresponding secondary anomaly probability. This probability reflects the risk of abnormal device performance under the current deposition conditions, provides a scientific basis for process optimization and parameter adjustment, and helps prevent the occurrence of MOSFET performance anomalies at an early stage.

[0026] Furthermore, the present application provides a method for obtaining deposition parameters of a gate metal layer deposition, combining the simulated gate oxide layer, performing secondary anomaly prediction, and obtaining a secondary anomaly probability, including: The deposition parameters of the gate metal layer deposition are obtained as the gate metal layer deposition parameters; based on the production data of silicon carbide high-voltage MOSFET in the historical time, a sample gate oxide thickness variance set and a sample gate metal layer deposition parameter set are collected, and according to the proportion of abnormalities of the silicon carbide high-voltage MOSFET under different sample gate oxide thickness variances and sample gate metal layer deposition parameters, a sample secondary anomaly probability set is obtained by identifying; a secondary variation classifier is constructed using the sample gate oxide thickness variance set, the sample gate metal layer deposition parameter set and the sample secondary anomaly probability set; the gate metal layer deposition parameter and the gate oxide thickness variance are input into the secondary variation classifier, and the secondary anomaly probability is obtained by classification prediction.

[0027] Optionally, deposition parameters of the gate metal layer are obtained as gate metal layer deposition parameters, including key process data such as thickness, deposition rate, temperature, and vacuum environment. These parameters directly affect the deposition uniformity and coverage performance of the metal layer. At the same time, based on the production data of silicon carbide high-voltage MOSFET in historical time, sample data related to these parameters are collected, including a sample gate oxide thickness variance set and a sample gate metal layer deposition parameter set; subsequently, according to the gate oxide thickness variance and gate metal layer deposition parameters of different samples, the proportion of abnormalities in silicon carbide high-voltage MOSFET is statistically analyzed to obtain a sample secondary abnormality probability set by annotation. This set is used to construct a secondary variation classifier. The training process of the classifier is similar to the training process of the aforementioned primary variation classifier, including steps such as model initialization, forward propagation, loss calculation, back propagation, and parameter optimization. After the training of the primary variation classifier is completed, the actually obtained gate metal layer deposition parameters and gate oxide thickness variance are input into the trained secondary variation classifier. Based on these two sets of input data and the mapping relationship learned during historical training, the classifier predicts and outputs the secondary anomaly probability. This probability is used to quantify the possible abnormal risk to MOSFET performance caused by the combined effect of metal layer deposition and oxide layer deposition, providing an accurate reference for subsequent process adjustments and quality optimization.

[0028] The doping parameters of the source and drain for self-aligned doping are obtained, and according to the secondary abnormality probability, a doping abnormality prediction ratio is configured to obtain the doping abnormality probability, and an integrated doping abnormality probability prediction is performed according to the doping parameters to obtain the doping abnormality probability.

[0029] In one embodiment, when performing anomaly prediction of source and drain self-aligned doping, firstly, relevant parameters in the doping process are obtained, including doping concentration, energy, dosage, scanning mode, etc. These doping parameters directly affect the formation quality of the source and drain; then, according to the secondary anomaly probability calculated previously, the number of doping anomaly prediction branches used to predict doping anomalies is dynamically configured. During the configuration process, the number of branches is adjusted according to the size of the secondary anomaly probability. When the secondary anomaly probability is high, more doping anomaly prediction branches are allocated. By comprehensively analyzing the potential anomalies in the doping process, the accuracy of the prediction is improved. When the secondary anomaly probability is low, the number of branches is reduced. This dynamic branch configuration mechanism reduces computing costs and saves computing power. It not only ensures prediction accuracy, but also optimizes resource utilization efficiency under low-risk conditions. In the integrated prediction stage, each branch receives doping parameters as input separately and calculates the doping anomaly probability based on its own analysis algorithm. The prediction results of all branches are combined through weighted average or other integration methods to output the overall doping anomaly probability. In this way, not only can a variety of potential anomalies in the doping process be captured, but also the computing intensity can be dynamically adjusted according to the level of abnormal risk, thereby improving the computing efficiency while ensuring prediction accuracy, and providing accurate data support for subsequent process improvement and quality optimization.

[0030] Further, the present application provides a method for obtaining doping parameters for self-aligned doping of a source and a drain, configuring a doping anomaly prediction ratio according to the secondary anomaly probability, and performing an integrated doping anomaly probability prediction according to the doping parameters to obtain a doping anomaly probability, including: The doping parameters of the source and drain for self-aligned doping are obtained to obtain the doping parameters; the secondary anomaly probability is configured as a doping anomaly prediction ratio; based on ensemble learning, a doping anomaly prediction channel including a doping anomaly prediction branch of a branch number is trained; the doping anomaly prediction ratio is multiplied by the branch number and the result is rounded to obtain a doping analysis number; the doping parameters are respectively input into the randomly selected doping anomaly prediction branches of the doping analysis number to predict the predicted doping anomaly probability of the doping analysis number, and the mean is calculated to obtain the doping anomaly probability.

[0031] Preferably, the doping parameters of the source and drain for self-aligned doping are obtained, including key process parameters such as doping concentration, energy, and dosage, which directly affect the accuracy and uniformity of the doping area. These parameters are recorded and used as input features for subsequent analysis; then, according to the secondary anomaly probability calculated in the previous step, it is configured as a doping anomaly prediction ratio, and the ratio is used to dynamically adjust the number of doping anomaly prediction branches used in the integrated prediction process, so as to achieve reasonable allocation of resources under different anomaly risk levels. The doping anomaly prediction branches are built into the doping anomaly prediction channel. These doping anomaly prediction branches are trained with historical doping data in an integrated training manner and can independently predict the doping anomaly probability. During the training process, a sample doping parameter set and the corresponding anomaly probability are collected as input and target, and multiple prediction branches are constructed using random forests, gradient boosting trees, deep learning, neural networks, etc. to ensure that each can ultimately accurately predict the anomaly. constant probability; in the prediction stage, the predicted proportion of doping anomaly is multiplied by the total number of branches in the prediction channel, and the number of doping analyses actually involved in the analysis is rounded up. The dynamic adjustment of the number of branches ensures that more branches are allocated to improve the analysis accuracy under high anomaly probability, while reducing the branches to optimize the calculation efficiency under low anomaly probability; then, the actually obtained doping parameters are input into the prediction branches of the randomly selected number of doping analyses, and each branch outputs a predicted doping anomaly probability separately. These prediction results are combined by mean calculation to obtain the final doping anomaly probability. This integrated prediction process can make full use of the flexibility and reliability of multi-branch channels, effectively manage computing resources while improving the accuracy of anomaly detection, and the final doping anomaly probability provides key data support for quality evaluation and process optimization of the self-aligned doping process, which is helpful to further improve the production quality and device performance of silicon carbide high-voltage MOSFETs.

[0032] Furthermore, the present application provides a doping anomaly prediction channel based on ensemble learning to train doping anomaly prediction branches including the number of branches, including: According to the silicon carbide high-voltage MOSFET production data in historical time, a sample doping parameter set is collected, and the proportion of abnormalities in the silicon carbide high-voltage MOSFET under different sample doping parameters is obtained, and the sample doping anomaly probability set is identified; in the sample doping parameter set and the sample doping anomaly probability set, there are multiple doping supervision training data of the number of randomly selected branches with replacement; the multiple doping supervision training data are used respectively, based on ensemble learning, to train the doping anomaly prediction branches of the number of branches, and the doping anomaly prediction channels are obtained by integration.

[0033] Optionally, based on the production data in the historical time, a doping parameter set related to the silicon carbide high-voltage MOSFET is collected. These parameters include variables such as doping concentration, energy, and dosage, covering different combinations of doping processes in the historical production process. At the same time, the proportion of MOSFET abnormalities under each set of doping parameter conditions is counted, for example, the frequency of occurrence of phenomena such as electrical performance degradation and leakage current increase, and these proportions are marked as sample doping abnormality probability sets; in order to construct a multi-branch supervised training data set, random sampling is performed from the sample doping parameter set and the sample doping abnormality probability set, and a sampling method with replacement is used to extract multiple data according to the required number of branches. Each data set contains doping parameters and corresponding abnormality probability labels. In this way, multiple sets of independent supervised training data are generated to ensure that the sample characteristics of each data set have a certain degree of randomness while retaining the consistency of the overall data distribution; subsequently, each supervised training data set is used to train multiple doping abnormality prediction branches based on an ensemble learning method. These branches can be based on commonly used ensemble learning models, such as Random Forest, Gradient Boosting Decision Trees, GBDT) or deep neural network (DNN), each branch is trained independently, and by learning the mapping relationship between doping parameters and anomaly probabilities, a doping anomaly prediction branch that can predict doping anomaly probabilities is generated. Exemplarily, when a deep neural network is used, the training process can be completed through forward propagation, loss calculation, back propagation, and parameter optimization in the same way as the aforementioned construction of a variant classifier. When a random forest is used, a supervised training data set is input into the random forest model. The random forest generates multiple sub-data sets from the supervised training data set through a random sampling method with replacement. These sub-data sets are used to train different decision trees in the random forest. For the training of each decision tree, a subset is randomly selected from all input features as candidate split features, and at each node Make splitting decisions. The splitting criteria can be based on information gain, Gini coefficient or mean square error (MSE) to maximize data purity or minimize prediction error. Recursive splitting continues until the stopping condition is met. For example, when the maximum depth is reached, the number of node samples is insufficient, or the error reduction caused by splitting is not significant, the splitting stops and leaf nodes are generated. Each leaf node records the abnormal probability of the node as the prediction result. After all decision trees in the random forest are trained independently, the model is integrated. For new input doping parameters, the random forest makes a separate prediction through each decision tree to obtain multiple abnormal probability results. By averaging the prediction results of all decision trees, the final doping abnormal probability prediction value is output. The random forest uses the integration of multiple decision trees to improve the robustness and anti-overfitting ability of the model.Finally, all trained doping anomaly prediction branches are integrated in parallel to construct a complete doping anomaly prediction channel, which can efficiently process a variety of doping parameter combinations, provide reliable anomaly probability prediction, and provide strong support for subsequent process optimization and quality control. ;

[0034] According to the primary abnormality probability, the secondary abnormality probability and the doping abnormality probability, a predicted abnormality probability is calculated to obtain a self-alignment process abnormality prediction result.

[0035] In one embodiment, during the self-alignment process abnormality prediction process, the final predicted abnormality probability is calculated by comprehensively considering the primary abnormality probability, the secondary abnormality probability and the doping abnormality probability. The primary abnormality probability mainly reflects the problems that may exist in the gate oxide deposition stage, such as uneven thickness distribution. The secondary abnormality probability combines the parameters of the gate oxide layer and the gate metal layer deposition to evaluate the abnormalities that may be caused during the metal layer deposition process. The doping abnormality probability is based on the parameters of the source and drain self-aligned doping to predict the accuracy and uniformity of the doping area. By comprehensively calculating these probability values, a predicted abnormality probability can be obtained as a self-alignment process abnormality prediction result. This predicted abnormality probability fully considers the impact of each key step in the process flow on the final device performance, reflects the independence of each link, and captures the linkage effect between them, and provides a quantitative assessment of the abnormality risk of the entire self-alignment process. It can be used as an important basis for production process monitoring and process optimization, and provides strong support for improving the manufacturing quality of silicon carbide high-voltage MOSFETs.

[0036] Further, the present application provides a method for calculating and obtaining a predicted abnormality probability according to the primary abnormality probability, the secondary abnormality probability and the doping abnormality probability as a self-alignment process abnormality prediction result, including: The average of the primary abnormality probability and the secondary abnormality probability is calculated to obtain the gate abnormality probability; the gate abnormality probability and the doping abnormality probability are added to obtain the predicted abnormality probability as the self-alignment process abnormality prediction result.

[0037] Preferably, the primary abnormality probability and the secondary abnormality probability are used as inputs, and the two are derived from the abnormality prediction results of the gate oxide layer deposition stage and the gate metal layer deposition stage respectively. The primary abnormality probability and the secondary abnormality probability are averaged to obtain the gate abnormality probability. The purpose of this step is to comprehensively evaluate the overall abnormality risk of the gate-related process and reflect the stability of the gate oxide layer and metal layer deposition links. Subsequently, the gate abnormality probability is added to the doping abnormality probability to calculate the final predicted abnormality probability. The doping abnormality probability is obtained by analyzing the parameters and abnormality risks of the source and drain doping links. By directly adding these two probability values, a comprehensive predicted abnormality probability is obtained. The predicted abnormality probability value comprehensively reflects the abnormality risk level of the three key stages (gate oxide layer deposition, gate metal layer deposition and doping) in the self-alignment process. The entire calculation process integrates the multi-stage prediction results into a unified quantitative indicator, which can provide a clear reference basis for the monitoring and abnormality management of the production process. If the predicted abnormality probability exceeds the preset threshold, it can be indicated that there is a high abnormality risk in the production process, and the process needs to be further optimized or the parameters adjusted to ensure that the final product quality meets the expected standards.

[0038] In summary, the embodiments of the present application have at least the following technical effects: The embodiment of the present application covers a complete process flow from gate oxide deposition, gate metal layer deposition to source and drain self-aligned doping. By obtaining the deposition parameters of the gate oxide layer, a generative adversarial network is used to simulate the distribution of oxide thickness, and a primary anomaly prediction is performed. The gate metal layer deposition parameters and the simulated gate oxide layer data are combined to perform a secondary anomaly prediction. The doping anomaly prediction ratio is configured according to the doping parameters and the secondary anomaly probability. The doping anomaly probability is predicted by an integrated learning model. After completing the anomaly probability prediction at each stage, the primary anomaly probability, the secondary anomaly probability and the doping anomaly probability are comprehensively calculated to generate the final self-aligned process anomaly prediction result. These technical effects jointly solve the technical problem that the anomalies of silicon carbide high-voltage MOSFETs in the self-aligned process are difficult to predict and control due to the complexity of multiple process steps, and realize the improvement of the accuracy and prediction ability of process anomaly identification through a multi-step joint anomaly prediction method, thereby optimizing the manufacturing quality and device performance.

[0039] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. The processes depicted in the accompanying drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0040] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0041] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A self-aligned process anomaly prediction method for silicon carbide high-voltage MOSFET, characterized in that: The method comprises: In the process of manufacturing silicon carbide high-voltage MOSFET by using a self-aligned process, the deposition parameters of the gate oxide layer are obtained, the gate oxide layer is simulated and generated, the simulated gate oxide layer is obtained, an abnormality prediction is performed, and an abnormality probability is obtained; Obtaining deposition parameters of the gate metal layer deposition, combining the simulated gate oxide layer, performing secondary anomaly prediction, and obtaining secondary anomaly probability; Acquire doping parameters for self-aligned doping of the source and drain, configure and obtain a doping anomaly prediction ratio according to the secondary anomaly probability, perform integrated doping anomaly probability prediction according to the doping parameters, and obtain a doping anomaly probability; According to the primary abnormality probability, the secondary abnormality probability and the doping abnormality probability, a predicted abnormality probability is calculated to obtain a self-alignment process abnormality prediction result.

2. The self-aligned process anomaly prediction method for silicon carbide high-voltage MOSFET according to claim 1, characterized in that: Obtaining deposition parameters of the gate oxide layer, performing simulation generation of the gate oxide layer, and obtaining a simulated gate oxide layer, including: Obtaining a deposition parameter of a gate oxide layer as a gate oxide layer deposition parameter; Based on the generative adversarial network, the oxide layer simulation channel is trained; The gate oxide layer deposition parameters are input into the oxide layer simulation channel, and a simulated gate oxide layer is generated by simulation.

3. The self-aligned process anomaly prediction method for silicon carbide high-voltage MOSFET according to claim 2, characterized in that: Based on the generative adversarial network, the oxide layer simulation channel is trained, including: Based on the gate oxide deposition data in the historical time, a sample gate oxide deposition parameter set is collected, and the thickness distribution of the gate oxide layer deposited under different sample gate oxide deposition parameters is collected and marked as a sample gate oxide set; Based on a generative adversarial network, an oxide layer simulation channel is constructed, wherein the oxide layer simulation channel includes a generator and an adversary; The sample gate oxide layer deposition parameter set and the sample gate oxide layer set are used to perform alternate supervised training on the generator and the adversary until convergence, thereby obtaining a trained oxide layer simulation channel.

4. The self-aligned process anomaly prediction method for silicon carbide high-voltage MOSFET according to claim 3, characterized in that: Perform an exception prediction and obtain an exception probability, including: According to the simulated gate oxide layer thickness distribution, a gate oxide layer thickness variance is calculated; Based on the gate oxide deposition data in the historical time, the sample gate oxide thickness variance set is collected, and according to the abnormal proportion of silicon carbide high-voltage MOSFET under different sample gate oxide thickness variances, the sample one-time abnormal probability set is identified; Using the sample gate oxide thickness variance set and the sample primary anomaly probability set, constructing a primary variation classifier; The gate oxide layer thickness variance is input into the primary variation classifier, and a primary abnormality probability is obtained through classification prediction.

5. The self-aligned process anomaly prediction method for silicon carbide high-voltage MOSFET according to claim 4, characterized in that: Obtaining deposition parameters of the gate metal layer deposition, combining the simulated gate oxide layer, performing secondary anomaly prediction, and obtaining secondary anomaly probability, including: Obtaining deposition parameters of the gate metal layer as gate metal layer deposition parameters; Based on the historical production data of silicon carbide high-voltage MOSFET, a set of sample gate oxide thickness variances and a set of sample gate metal layer deposition parameters are collected, and according to the proportion of abnormalities of silicon carbide high-voltage MOSFET under different sample gate oxide thickness variances and sample gate metal layer deposition parameters, a set of sample secondary abnormality probabilities is obtained by identification; A secondary variation classifier is constructed using the sample gate oxide layer thickness variance set, the sample gate metal layer deposition parameter set and the sample secondary anomaly probability set; The gate metal layer deposition parameters and the gate oxide layer thickness variance are input into the secondary variation classifier, and the secondary anomaly probability is obtained through classification prediction.

6. The self-aligned process anomaly prediction method for silicon carbide high-voltage MOSFET according to claim 1, characterized in that: Acquiring doping parameters for self-aligned doping of the source and drain, configuring a doping anomaly prediction ratio according to the secondary anomaly probability, and performing integrated doping anomaly probability prediction according to the doping parameters to obtain the doping anomaly probability, including: Obtaining doping parameters of the source and drain for self-aligned doping to obtain doping parameters; configuring the secondary anomaly probability as a doping anomaly prediction ratio; Based on ensemble learning, a doping anomaly prediction channel including doping anomaly prediction branches of a number of branches is trained; The doping anomaly prediction ratio is multiplied by the number of branches and the result is rounded to obtain the doping analysis number; The doping parameters are respectively input into the doping anomaly prediction branches of the randomly selected doping analysis quantity, the predicted doping anomaly probability of the doping analysis quantity is predicted, and the mean is calculated to obtain the doping anomaly probability.

7. The self-aligned process anomaly prediction method for silicon carbide high-voltage MOSFET according to claim 6, characterized in that: Based on ensemble learning, the doping anomaly prediction channel of the doping anomaly prediction branch including the number of branches is trained, including: According to the production data of silicon carbide high-voltage MOSFET in the historical period, a set of sample doping parameters is collected, and the proportion of abnormal silicon carbide high-voltage MOSFET under different sample doping parameters is obtained, and a set of sample doping abnormality probability is obtained by identification; There are multiple copies of doping supervision training data of randomly selected branches with replacement in the sample doping parameter set and the sample doping abnormal probability set; The plurality of doping supervision training data are respectively used to train a number of doping anomaly prediction branches based on ensemble learning, and the doping anomaly prediction channels are obtained through integration.

8. The self-aligned process anomaly prediction method for silicon carbide high-voltage MOSFET according to claim 1, characterized in that: According to the primary abnormality probability, the secondary abnormality probability and the doping abnormality probability, a predicted abnormality probability is calculated as a self-alignment process abnormality prediction result, including: Calculate the average of the primary abnormality probability and the secondary abnormality probability to obtain the gate abnormality probability; The gate abnormality probability and the doping abnormality probability are added to obtain a predicted abnormality probability as a self-alignment process abnormality prediction result.

Images