A method for ion implantation optimization of a semiconductor device

Abstract

The application provides an ion implantation optimization method for semiconductor devices, and relates to the technical field of ion implantation. The method comprises the following steps: receiving and configuring ion implantation control attributes based on ion beam implantation, plasma implantation components, etc. A fault tolerance threshold is set for the control attributes in combination with factors such as device model and wafer model. By analyzing wafer implantation logs, the deviation modulus of the control attributes is obtained, and a deviation prediction set is generated using a prediction network. According to the fault tolerance threshold, abnormal control attributes are calibrated, and the results are sent to the user end to optimize the equipment. Through the application, the problem that the uniformity and depth control of ion implantation are not accurate enough due to the lack of dynamic adjustment and optimization of process parameters in the prior art, which further affects the performance and manufacturing yield of semiconductor devices, can be solved. Real-time, dynamic and accurate control of ion implantation process parameters is realized, thereby improving the efficiency and product quality of wafer manufacturing.

Description

Technical Field

[0001] This application relates to the field of ion implantation technology, and more particularly to an ion implantation optimization method for semiconductor devices. Background Technology

[0002] Ion implantation is a crucial step in semiconductor manufacturing, using a high-energy ion beam to introduce dopant atoms onto the surface of a wafer to alter the material's electrical properties. This process demands extremely high levels of control over doping uniformity and depth, as this directly impacts the performance of the final semiconductor device. In traditional ion implantation equipment, adjusting control parameters typically relies on the operator's extensive experience and a series of physical experiments. These methods are not only time-consuming but may also fail to achieve optimal performance because human operators struggle to account for the complex interactions between all variables.

[0003] Currently, existing control methods for wafer ion implantation equipment mainly rely on operator experience and a set of fixed parameter settings, which are often determined based on limited experimental data or standard operating procedures. This method may not achieve optimal control results when dealing with different types of wafers, different ion beams, and different equipment models.

[0004] In summary, existing technologies lack dynamic adjustment and optimization of process parameters, resulting in insufficient precision in the uniformity and depth control of ion implantation, which further affects the performance and manufacturing yield of semiconductor devices. Summary of the Invention

[0005] The purpose of this application is to provide an ion implantation optimization method for semiconductor devices, which solves the problem that the lack of dynamic adjustment and optimization of process parameters in the prior art leads to insufficient precision in the uniformity and depth control of ion implantation, which further affects the performance and manufacturing yield of semiconductor devices.

[0006] In view of the above problems, this application provides an ion implantation optimization method for semiconductor devices.

[0007] This application provides an ion implantation optimization method for semiconductor devices, comprising: receiving an ion implantation control attribute set based on the ion beam implantation component, the plasma implantation component, the robotic arm, and the throttle valve; performing frequency allocation on the ion implantation control attribute set based on the ion beam type, wafer model, and wafer ion implantation equipment model to generate an ion implantation control attribute fault tolerance threshold; collecting wafer ion implantation logs according to the wafer ion implantation equipment tag number; performing deviation analysis on the wafer ion implantation logs to generate a control attribute deviation magnitude sequence set, wherein the control attribute deviation magnitude represents the mode of the deviation magnitude between the control parameter and the reference parameter; fitting the control attribute deviation magnitude sequence set through a deviation magnitude prediction network to generate a deviation magnitude prediction set; identifying abnormal attributes in the deviation magnitude prediction set based on the ion implantation control attribute fault tolerance threshold to generate an abnormal control attribute calibration result; and sending the abnormal control attribute calibration result to the user terminal for equipment optimization.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0009] The system receives a set of ion implantation control attributes based on the ion beam implantation component, the plasma implantation component, the robotic arm, and the throttle valve. Based on the ion beam type, wafer model, and wafer ion implantation equipment model, the set of ion implantation control attributes is frequency-matched to generate an ion implantation control attribute fault tolerance threshold. Wafer ion implantation logs are collected according to the wafer ion implantation equipment tag number. Deviation analysis is performed on the wafer ion implantation logs to generate a set of control attribute deviation magnitude sequences, where the control attribute deviation magnitude represents the mode of the deviation magnitude between the control parameter and the reference parameter. A deviation magnitude prediction network is used to fit the control attribute deviation magnitude sequence set to generate a deviation magnitude prediction set. Based on the ion implantation control attribute fault tolerance threshold, the deviation magnitude prediction set is anomaly attribute identified, generating anomaly control attribute calibration results. These abnormal control attribute calibration results are sent to the user terminal for equipment optimization. This effectively solves the problem that existing technologies lack dynamic adjustment and optimization of process parameters, resulting in insufficient precision in the uniformity and depth control of ion implantation, which further affects the performance and manufacturing yield of semiconductor devices. This enables real-time, dynamic, and precise control of ion implantation process parameters, thereby improving wafer manufacturing efficiency and product quality.

[0010] The above description is merely an overview of the technical solution of this application. To better understand the technical means of this application and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, specific embodiments of this application are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent through the following description. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0012] Figure 1 This is a schematic flowchart of an ion implantation optimization method for a semiconductor device according to this application;

[0013] Figure 2 This is a schematic diagram of the process for configuring the fault tolerance threshold of the ion implantation control attribute in an ion implantation optimization method for a semiconductor device according to this application. Detailed Implementation

[0014] This application provides an optimized ion implantation method for semiconductor devices, addressing the shortcomings of existing technologies that lack dynamic adjustment and optimization of process parameters. This results in insufficient precision in ion implantation uniformity and depth control, further impacting semiconductor device performance and manufacturing yield. The method achieves real-time, dynamic, and precise control of ion implantation process parameters, thereby improving wafer manufacturing efficiency and product quality.

[0015] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. It should be understood that this application is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. It should also be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all of them. Example 1

[0016] Please see the appendix Figure 1 This application provides an ion implantation optimization method for semiconductor devices, applied to a wafer ion implantation equipment. The wafer ion implantation equipment includes an ion beam implantation assembly, a plasma implantation assembly, a robotic arm, and a throttling valve, comprising:

[0017] S1: Receive a set of ion implantation control attributes based on the ion beam implantation assembly, the plasma implantation assembly, the robotic arm, and the throttle valve.

[0018] Specifically, data generated during the operation of the ion implantation equipment is collected. This includes, but is not limited to, parameters such as current, voltage, and beam intensity of the ion beam implantation components; parameters such as power, gas flow rate, and pressure of the plasma implantation components; data such as the movement speed and position of the robotic arm; and data such as the opening degree and flow rate of the throttle valve. The collected data is then integrated into a set of control attributes. This set is a multi-dimensional dataset containing all key parameters affecting the ion implantation process. Frequency values ​​are assigned to the set of control attributes based on the ion beam type, wafer model, and wafer ion implantation equipment model. For different ion beam types, wafer models, and equipment models, some control parameters may require monitoring and adjustment at different frequencies. For example, for high-energy ion beams, more frequent monitoring and adjustment of the beam intensity may be necessary. Based on the above values, a tolerance threshold for the ion implantation control attributes is generated. The tolerance threshold refers to the acceptable fluctuation range of the control parameters; exceeding this range may affect the quality of ion implantation. This threshold can be determined through historical data analysis and machine learning model prediction.

[0019] S2: Based on the ion beam type, wafer model, and wafer ion implantation equipment model, perform frequency allocation on the ion implantation control attribute set to generate ion implantation control attribute fault tolerance thresholds.

[0020] Specifically, this involves collecting various parameters involved in historical ion implantation processes, including parameters of the ion beam implantation components (such as beam intensity, energy, and focusing), parameters of the plasma implantation components (such as gas composition, pressure, and power), motion parameters of the robotic arm (such as speed, position, and acceleration), and adjustment parameters of the throttle valve (such as opening and flow rate). Control attributes are categorized and labeled according to ion beam type, wafer model, and wafer ion implantation equipment model. Control attribute data under different types and models are analyzed to identify key characteristics, such as which parameters change most significantly under different conditions. For each combination of ion beam type, wafer model, and equipment model, the monitoring and adjustment frequency of control attributes is determined. For example, certain wafer models with high precision requirements may require more frequent parameter adjustments. This frequency can be determined based on historical data analysis, equipment performance requirements, production experience, and other factors. Statistical methods and machine learning algorithms are used to analyze historical data and control attribute deviations to determine the tolerance threshold for each parameter. The tolerance threshold refers to the range within which a parameter can fluctuate without affecting product quality.

[0021] S3: Collect wafer ion implantation logs based on the wafer ion implantation device tag number.

[0022] Specifically, each wafer ion implantation device should have a clearly defined tag number, which can be numbers, letters, or a combination of both. This tag number identifies the parameters to be collected, including all key variables affecting the ion implantation process, such as ion beam current, voltage, beam intensity, implantation time, implantation dose, plasma parameters, and robotic arm motion parameters. While the device is running, these defined parameters are recorded in real time and stored with timestamps. Timestamps ensure that each record can be traced back to a specific point in time. The collected data is then organized according to a specific format to generate a wafer ion implantation log. The log typically includes the device tag number, timestamp, process steps, and parameter values.

[0023] S4: Perform deviation analysis on the wafer ion implantation log to generate a set of control attribute deviation modulus sequences, wherein the control attribute deviation modulus represents the mode of the deviation modulus between the control parameter and the reference parameter.

[0024] Specifically, for each control parameter in the collected wafer ion implantation log, its deviation from the corresponding baseline parameter is calculated. The deviation calculation can be a simple difference or a more complex statistic, such as the standard deviation. The magnitude of the deviation for each control parameter is calculated to eliminate the influence of the sign and focus on the magnitude of the deviation. The magnitudes of the deviations of all control attributes are organized into a sequence set, with each sequence corresponding to the magnitude of the deviation of a control attribute at different time points. For each sequence of magnitudes of the deviations of a control attribute, its mode is calculated; this is the deviation magnitude that occurs most frequently. The mode can serve as a typical value for the deviation of that attribute.

[0025] S5: By using the deviation modulus prediction network, the set of control attribute deviation modulus sequences is traversed and fitted to generate a deviation modulus prediction set.

[0026] Specifically, prepare a dataset for training the prediction network. This dataset contains sequences of control attribute deviation moduli from historical wafer ion implantation logs. Choose a suitable prediction model. This could be a time series prediction model, such as an autoregressive integral moving average, or a more complex machine learning model, such as a long short-term memory network or a convolutional neural network. Train the prediction model using the prepared dataset. During training, the model will learn how to predict future deviation moduli based on historical deviation moduli. After training, use the model to iterate through the existing set of control attribute deviation moduli sequences, fitting each sequence. This means feeding the historical values ​​of each sequence into the model to predict future deviation moduli. The model will generate a series of predicted values ​​for each sequence, which constitute a set of predicted deviation moduli. This set contains the predicted deviation moduli for all control attributes over a future period.

[0027] S6: Based on the fault tolerance threshold of the ion implantation control attribute, identify abnormal attributes in the deviation modulus prediction set and generate abnormal control attribute calibration results.

[0028] Specifically, a tolerance threshold is set for each control attribute. This threshold defines the maximum allowable fluctuation of the parameter within the normal operating range. Each predicted value in the deviation modulus prediction set is compared with the corresponding tolerance threshold. If a predicted value exceeds the threshold range, the attribute is considered abnormal. For each predicted value exceeding the threshold, its corresponding control attribute is marked as abnormal. This can be achieved through marking, color coding, or any other form of visualization for quick operator identification. The results of marking all abnormal control attributes are compiled into a report or visualization interface; this is the abnormal control attribute labeling result. This result should clearly show which attributes are marked as abnormal and their predicted deviation modulus values.

[0029] S7: Send the abnormal control attribute calibration results to the user terminal for device optimization.

[0030] Specifically, determine the communication protocol used to send the results. This could be email, SMS, enterprise instant messaging tools such as Slack and Microsoft Teams, or a dedicated user interface. Ensure the security of the results during transmission. For transmissions involving sensitive data, encrypted channels and secure authentication mechanisms should be used. Push the anomaly control attribute labeling results to the user using the selected communication protocol. The user receives and processes this information. If the anomaly labeling results indicate an emergency or a problem requiring immediate attention, issue an alert or notification to ensure the user can respond promptly.

[0031] Furthermore, step S1 of this application also includes:

[0032] The ion beam implantation component includes ion beam emission energy, ion beam current, mass spectrometer magnetic field strength, ion type, ion beam implantation energy, and ion beam implantation path timing information; the plasma implantation component includes plasma energy and plasma current; the robotic arm includes wafer movement path timing information; the throttle valve includes space pressure; the ion beam emission energy, ion beam current, mass spectrometer magnetic field strength, ion type, ion beam implantation energy, ion beam implantation path timing information, plasma energy, plasma current, wafer movement path timing information, and space pressure are added to the ion implantation control attribute set.

[0033] Specifically, identify the key parameters for each component. For the ion beam implantation component, this includes ion beam emission energy, ion beam current, mass spectrometer magnetic field strength, ion type, ion beam implantation energy, and ion beam implantation path timing information. For the plasma implantation component, key parameters include plasma energy and plasma current. The key parameter for the robotic arm is the wafer movement path timing information. The key parameter for the throttle valve is the space pressure. Acquire data for these parameters from data interfaces in the equipment control system, sensor readings, or other monitoring devices. Integrate the acquired data into a central database. Ensure data format consistency and timestamp accuracy. Add newly acquired parameters to the existing set of ion implantation control attributes. Update the data structure or database schema to accommodate additional information.

[0034] Furthermore, such as Figure 2 As shown, step S2 of this application further includes:

[0035] Based on the ion beam type, wafer model, and wafer ion implantation equipment model, qualified samples of the ion implantation control attribute set are assigned values ​​to obtain an ion implantation control attribute assigned dataset; frequency analysis is performed on the ion implantation control attribute assigned dataset to obtain a distribution frequency coefficient set; the ion implantation control attribute assigned dataset is sorted according to the distribution frequency coefficient set to obtain ion implantation control attribute assigned data sorting results; ion implantation control attribute benchmark data is received through the user terminal; based on the ion implantation control attribute benchmark data and the ion implantation control attribute assigned data sorting results, the ion implantation control attribute fault tolerance threshold is configured.

[0036] Specifically, qualified production samples are selected from historical data based on ion beam type, wafer model, and equipment model. These samples should be data points that performed well during the production process. For each selected sample, its corresponding ion implantation control attribute value is assigned as a qualified sample value and added to the ion implantation control attribute assignment dataset. Frequency analysis is performed on the ion implantation control attribute assignment dataset to calculate the distribution frequency coefficient of each attribute value. Based on the set of distribution frequency coefficients, the ion implantation control attribute assignment dataset is sorted. The dataset is divided into several parts, each containing attribute values ​​with similar distribution frequency coefficients. Benchmark data for ion implantation control attributes is received through the user terminal. Based on the received ion implantation control attribute benchmark data and the sorting results of the ion implantation control attribute assignment data, the fault tolerance threshold for the ion implantation control attributes is configured.

[0037] Furthermore, this application also includes:

[0038] Obtain first ion implantation control attribute assignment data from the ion implantation control attribute assignment dataset; extract second ion implantation control attribute assignment data from the ion implantation control attribute assignment dataset, wherein the second ion implantation control attribute assignment data is different from the first ion implantation control attribute assignment data; perform similarity analysis on the first ion implantation control attribute assignment data and the second ion implantation control attribute assignment data to generate a first similarity ratio; when the first similarity ratio is greater than or equal to the first similarity ratio threshold, increment the first distribution frequency coefficient of the first ion implantation control attribute assignment data by one, wherein the initial value of the first distribution frequency coefficient is equal to 0; until the ion implantation control attribute assignment dataset is traversed completely, add the first distribution frequency coefficient to the distribution frequency coefficient set.

[0039] Specifically, a first set of ion implantation control attribute assignment data is extracted from the ion implantation control attribute assignment dataset. Then, a second set of different ion implantation control attribute assignment data is extracted, and a similarity analysis is performed on the extracted first and second sets of ion implantation control attribute assignment data. This is achieved by calculating the correlation coefficient, cosine similarity, or other similarity measures between the two sets of data. The generated similarity ratio is called the first similarity ratio. If the first similarity ratio is greater than or equal to a preset first similarity ratio threshold, it indicates that the two sets of data are sufficiently similar, and therefore the distribution frequency coefficient of the first set of data is incremented by one. The initial value of the first distribution frequency coefficient is set to 0, and this coefficient will increase as similar data is discovered. The above process is repeated, traversing the entire ion implantation control attribute assignment dataset, performing similarity analysis and updating the frequency coefficient for each set of data. After the traversal is complete, all first distribution frequency coefficients are added to the distribution frequency coefficient set. This set will contain the distribution frequency information for each attribute value in the dataset.

[0040] Furthermore, this application also includes:

[0041] Extract the first assigned value data of the first control attribute from the first ion implantation control attribute assignment data, and extract the second assigned value data of the first control attribute from the second ion implantation control attribute assignment data; calculate the product of the first assigned value data and the second assigned value data, and then multiply by 2 to generate a first similarity evaluation factor; calculate the first square value of the first assigned value data and the second square value of the second assigned value data, and sum the first square value and the second square value to generate a second similarity evaluation factor; set the ratio of the first similarity evaluation factor to the second similarity evaluation factor as the first control attribute similarity ratio; when the first control attribute similarity ratio is greater than or equal to the second similarity ratio threshold, increment the number of similar control attributes by one, wherein the initial value of the number of similar control attributes is equal to 0; after traversing all attributes, obtain the proportion of the number of similar control attributes in the total number of attributes, and set it as the first similarity ratio.

[0042] Specifically, the first assigned value data for the first control attribute is extracted from the first ion implantation control attribute assignment data. The second assigned value data for the same first control attribute is extracted from the second ion implantation control attribute assignment data. The product of the first and second assigned values ​​is calculated and then multiplied by 2 to generate a first similarity evaluation factor. This factor reflects the degree of similarity of the control attribute values ​​in the two sets of data. The squares of the first and second assigned values ​​are calculated separately, and then the two squares are added together to generate a second similarity evaluation factor. This factor is used to standardize the similarity evaluation. The ratio of the first and second similarity evaluation factors is set as the first control attribute similarity ratio. This ratio is a dimensionless similarity measure that can be used to compare the similarity between different control attributes. If the first control attribute similarity ratio is greater than or equal to a preset second similarity ratio threshold, it indicates that the two data points are sufficiently similar in that control attribute, and therefore the number of similar control attributes is incremented by one. The initial value of the number of similar control attributes is set to 0, and this number will increase as the number of similar attributes increases. The above process is repeated, iterating through all control attributes, calculating the similarity ratio for each attribute, and updating the number of similar control attributes. After iterating through all attributes, the percentage of similar control attributes out of the total number of attributes is obtained; this percentage is the first similarity ratio. The first similarity ratio can be used to evaluate the degree of similarity between two sets of data across the entire attribute set.

[0043] Furthermore, this application also includes:

[0044] Based on the sorting results of the ion implantation control attribute assignment data, a first control attribute assignment data sorting result set is extracted; based on the ion implantation control attribute benchmark data, first control attribute benchmark data is extracted; based on the first control attribute assignment data sorting result set and the first control attribute benchmark data, deviation extreme values ​​are extracted to obtain positive deviation extreme values ​​and negative deviation extreme values; the minimum modulus of the positive deviation extreme value and the negative deviation extreme value is selected and set as the first control attribute fault tolerance threshold, and added to the ion implantation control attribute fault tolerance threshold.

[0045] Specifically, a set of sorted assignment data for the first control attribute is extracted from the sorting results of the ion implantation control attribute assignment data. This set contains the first control attribute assignment data selected during the sorting process. Based on the ion implantation control attribute baseline data, baseline data for the first control attribute is extracted. Deviation extreme values ​​are extracted from the sorting result set of the first control attribute assignment data and the baseline data. The deviation between each data point and the baseline data is calculated, and the maximum value (positive deviation extreme value) and minimum value (negative deviation extreme value) among these deviations are found. The minimum modulus value is selected from the positive and negative deviation extreme values. This modulus value represents the minimum possible value of the control attribute deviation. The selected minimum modulus value is set as the first control attribute tolerance threshold. This threshold defines the maximum acceptable deviation range for the control attribute. The first control attribute tolerance threshold is added to the ion implantation control attribute tolerance threshold set. This set contains the tolerance thresholds for all control attributes.

[0046] Furthermore, step S5 of this application also includes:

[0047] Based on the control attribute deviation magnitude sequence set, a first control attribute deviation magnitude fluctuation curve, a second control attribute deviation magnitude fluctuation curve, and so on up to the Nth control attribute deviation magnitude fluctuation curve are constructed. Using the deviation magnitude prediction network, the first control attribute deviation magnitude fluctuation curve, the second control attribute deviation magnitude fluctuation curve, and so on up to the Nth control attribute deviation magnitude fluctuation curve are traversed to predict curve trends, thereby obtaining the deviation magnitude prediction set. The construction steps of the deviation magnitude prediction network include: collecting a deviation magnitude curve recording dataset; segmenting the deviation magnitude curve recording dataset to generate an input training curve dataset and an output supervised magnitude dataset; and training a recurrent neural network based on the input training curve dataset and the output supervised magnitude dataset to generate the deviation magnitude prediction network.

[0048] Specifically, a fluctuation curve is constructed for each control attribute based on a set of control attribute deviation magnitude value sequences. This includes time series data, where each data point represents the deviation magnitude value at a specific time point. A deviation magnitude prediction network is used to traverse the fluctuation curve of each control attribute. The network learns the fluctuation pattern of each attribute and predicts future deviation magnitude value trends based on historical data. First, a dataset of deviation magnitude value curve records is collected, containing time series data of historical deviation magnitude values. These datasets are split to generate an input training curve dataset and an output supervised magnitude value dataset. The input dataset is used to train the network, and the output dataset is used to validate the network's predictive ability. A recurrent neural network, such as LSTM or GRU, is trained based on the split input training curve dataset and output supervised magnitude value dataset. Recurrent neural networks can process time series data and learn patterns and trends within it. Through training, the recurrent neural network learns how to predict the future deviation magnitude values ​​of control attributes. This trained network is the deviation magnitude prediction network.

[0049] Furthermore, this application also includes:

[0050] Based on the control attribute deviation modulus sequence set, extract the first control attribute deviation modulus set; when the endpoint modulus of the first control attribute deviation modulus set does not meet the ion implantation control attribute fault tolerance threshold, or / and the number of modulus values ​​in the first control attribute deviation modulus set that do not meet the ion implantation control attribute fault tolerance threshold is greater than a preset number, add the first control attribute to the abnormal control attribute calibration result; traverse the second control attribute up to the Nth control attribute, and output the abnormal control attribute calibration result.

[0051] Specifically, the deviation modulus set of the first control attribute is extracted from the control attribute deviation modulus sequence set. This set contains the deviation modulus values ​​of the first control attribute within a specific time period. The endpoint modulus value of the first control attribute deviation modulus set, i.e., the deviation modulus value at a specific point in time, is monitored. It is then checked whether this endpoint modulus value meets the ion implantation control attribute tolerance threshold. If the endpoint modulus value exceeds the tolerance threshold, or if the number of modulus values ​​in the set that do not meet the tolerance threshold exceeds a preset number, the first control attribute is considered abnormal. If the first control attribute deviation modulus set is abnormal, the first control attribute is added to the abnormal control attribute calibration results. This is achieved through marking, color coding, or other visualization methods. The above process is repeated, traversing the second control attribute up to the Nth control attribute. The same monitoring and abnormal attribute addition steps are performed for each control attribute. After traversing all control attributes, the abnormal control attribute calibration results are output. This may include generating a report or visualization interface listing all control attributes marked as abnormal.

[0052] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0053] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application also intends to include such modifications and variations.

Claims

1. A method of ion implantation optimization for a semiconductor device, comprising: Applied to a wafer ion implantation device, the wafer ion implantation device includes an ion beam implantation assembly, a plasma implantation assembly, a machine arm and a throttle valve, comprising: Based on the ion beam implantation assembly, the plasma implantation assembly, the machine arm and the throttle valve, a set of ion implantation control attributes is received; Based on the ion beam type, the wafer model and the wafer ion implantation device model, the ion implantation control attribute set is frequency valued to generate an ion implantation control attribute fault tolerance threshold; According to the wafer ion implantation device model, the wafer ion implantation log is collected; The wafer ion implantation log is analyzed for deviation to generate a control attribute deviation modulus value sequence set, wherein the control attribute deviation modulus value represents the mode of the deviation modulus value of the control parameter and the reference parameter; Through the deviation modulus value prediction network, the control attribute deviation modulus value sequence set is fitted to generate a deviation modulus value prediction set; Based on the ion implantation control attribute fault tolerance threshold, the deviation modulus value prediction set is identified for abnormal attributes to generate an abnormal control attribute calibration result; The abnormal control attribute calibration result is sent to the user end for device optimization.

2. The method of claim 1, wherein the step of optimizing the ion implantation process for the semiconductor device is performed by a computer system. Based on the ion beam implantation assembly, the plasma implantation assembly, the machine arm and the throttle valve, a set of ion implantation control attributes is received, comprising: The ion beam implantation assembly includes ion beam emission energy, ion beam current, mass spectrometer magnetic field strength, ion type, ion beam implantation energy, ion beam implantation path timing information; The plasma implantation assembly includes plasma energy and plasma current; The machine arm includes wafer movement path timing information; The throttle valve includes space pressure; The ion beam emission energy, the ion beam current, the mass spectrometer magnetic field strength, the ion type, the ion beam implantation energy, the ion beam implantation path timing information, the plasma energy, the plasma current, the wafer movement path timing information and the space pressure are added to the ion implantation control attribute set.

3. The method of claim 1, wherein the step of optimizing the ion implantation process is performed by a computer program. Based on the ion beam type, the wafer model and the wafer ion implantation device model, the ion implantation control attribute set is frequency valued to generate an ion implantation control attribute fault tolerance threshold, comprising: Based on the ion beam type, the wafer model and the wafer ion implantation device model, the ion implantation control attribute set is assigned to qualified samples to obtain an ion implantation control attribute assignment data set; The ion implantation control attribute assignment data set is iterated for frequency analysis to obtain a distribution frequency coefficient set; According to the distribution frequency coefficient set, the ion implantation control attribute assignment data set is sorted to obtain an ion implantation control attribute assignment data sorting result; Through the user end, ion implantation control attribute reference data is received; According to the ion implantation control attribute reference data, based on the ion implantation control attribute assignment data sorting result, the ion implantation control attribute fault tolerance threshold is configured.

4. The method of claim 3, wherein the step of determining the ion implantation parameters is performed by a computer program. The ion implantation control attribute assignment data set is iterated for frequency analysis to obtain a distribution frequency coefficient set, comprising: obtaining first ion implantation control attribute assignment data of the ion implantation control attribute assignment data set; extracting second ion implantation control attribute assignment data from the ion implantation control attribute assignment data set, wherein the second ion implantation control attribute assignment data is different from the first ion implantation control attribute assignment data; performing similarity analysis on the first ion implantation control attribute assignment data and the second ion implantation control attribute assignment data to generate a first similarity ratio; when the first similarity ratio is greater than or equal to a first similarity ratio threshold, adding one to a first distribution frequency coefficient of the first ion implantation control attribute assignment data, wherein the initial value of the first distribution frequency coefficient is equal to 0; until the ion implantation control attribute assignment data set is traversed, adding the first distribution frequency coefficient to the distribution frequency coefficient set.

5. The method of ion implantation optimization for a semiconductor device of claim 4, wherein, The similarity analysis on the first ion implantation control attribute assignment data and the second ion implantation control attribute assignment data to generate a first similarity ratio comprises: extracting first assignment data of a first control attribute from the first ion implantation control attribute assignment data, and extracting second assignment data of the first control attribute from the second ion implantation control attribute assignment data; calculating the product of the first assignment data and the second assignment data, and multiplying by 2 to generate a first similarity evaluation factor; calculating the first square value of the first assignment data and the second square value of the second assignment data, and adding the first square value and the second square value to generate a second similarity evaluation factor; setting the ratio of the first similarity evaluation factor to the second similarity evaluation factor as a first control attribute similarity ratio; when the first control attribute similarity ratio is greater than or equal to a second similarity ratio threshold, adding one to the number of similar control attributes, wherein the initial value of the number of similar control attributes is equal to 0; when all attributes are traversed, setting the proportion of the number of similar control attributes in the total number of attributes as the first similarity ratio.

6. The method of claim 3, wherein the step of determining the ion implantation parameters is performed by a computer program. According to the ion implantation control attribute reference data, based on the ion implantation control attribute assignment data sorting result, the ion implantation control attribute fault tolerance threshold is configured, comprising: According to the ion implantation control attribute assignment data sorting result, a first control attribute assignment data sorting result set is extracted; According to the ion implantation control attribute reference data, a first control attribute reference data is extracted; According to the first control attribute assignment data sorting result set and the first control attribute reference data, deviation extreme value extraction is performed to obtain positive deviation extreme value and negative deviation extreme value; selecting the minimum modulus value of the positive deviation extreme value and the negative deviation extreme value as the first control attribute fault tolerance threshold, and adding it to the ion implantation control attribute fault tolerance threshold.

7. The method of claim 1, wherein the step of optimizing the ion implantation process is performed by a computer system. Through the deviation modulus value prediction network, the control attribute deviation modulus value sequence set is traversed to generate a deviation modulus value prediction set, comprising: According to the control attribute deviation modulus value sequence set, a first control attribute deviation modulus value fluctuation curve, a second control attribute deviation modulus value fluctuation curve, and an Nth control attribute deviation modulus value fluctuation curve are constructed; According to the deviation modulus prediction network, curve trend prediction is performed on the first control attribute deviation modulus fluctuation curve, the second control attribute deviation modulus fluctuation curve, and the Nth control attribute deviation modulus fluctuation curve, and a deviation modulus prediction set is obtained; The construction of the deviation modulus prediction network comprises the following steps: Collecting deviation modulus curve record data sets; Segmenting the deviation modulus curve record data sets to generate input training curve data sets and output supervision modulus data sets; Training a recurrent neural network according to the input training curve data sets and the output supervision modulus data sets to generate the deviation modulus prediction network.

8. The method of ion implantation optimization for a semiconductor device of claim 7, wherein, According to the control attribute deviation modulus sequence set, a first control attribute deviation modulus fluctuation curve, a second control attribute deviation modulus fluctuation curve, and an Nth control attribute deviation modulus fluctuation curve are constructed, which comprises the following steps: According to the control attribute deviation modulus sequence set, a first control attribute deviation modulus set is extracted; When the terminal modulus value of the first control attribute deviation modulus set does not satisfy the ion implantation control attribute fault tolerance threshold, or / and the number of modulus values of the first control attribute deviation modulus set that do not satisfy the ion implantation control attribute fault tolerance threshold is greater than a preset number, the first control attribute is added to the abnormal control attribute calibration result; The second control attribute to the Nth control attribute are traversed, and the abnormal control attribute calibration result is output.

Images