A CVS analysis parameter optimization method based on stability evaluation of volt-ampere curve

Abstract

The application discloses a CVS analysis parameter optimization method based on stability evaluation of volt-ampere curve, comprising: a data acquisition step, obtaining process parameters and stable interval volt-ampere curves recorded in the CVS analysis process; a data preprocessing step, performing outlier processing, calculating the stability evaluation index of the volt-ampere curve, and selecting key process parameters as model input features; a model training and selection step, constructing a machine learning model library, and selecting the best model with the root mean square error as the index; an intelligent optimization parameter searching step, adopting a Bayesian optimization algorithm based on a multi-objective tree structure Parzen estimator, taking the relative standard deviation of the stripping electric quantity, the relative standard deviation of the peak current, the standard deviation of the peak potential and the absolute value of the slope of the stripping electric quantity as the optimization targets, and performing collaborative optimization through the Pareto non-dominated relationship to obtain the optimal parameter combination. The application realizes automatic optimization of the CVS analysis parameters, and improves the analysis repeatability and stability.

Description

Technical Field

[0001] This invention relates to the field of CVS analysis technology, and in particular to a method for optimizing CVS analysis parameters based on the stability evaluation of volt-ampere curves. Background Technology

[0002] Cyclic Voltammetry Stripping (CVS) analysis is widely used in the semiconductor industry for monitoring the concentration of additives (including inhibitors, brighteners, and leveling agents) in copper plating baths. CVS analysis records cyclic voltammetry curves, integrates the stripping peaks in the curves to obtain the stripping charge, and then calculates the additive concentration based on different analytical methods for different additives (such as CVS-DT cyclic voltammetry stripping-dilution titration, MLAT modified linear approximation method, and RC response curve method).

[0003] However, existing CVS analysis techniques have the following drawbacks: First, process parameters are mainly set based on expert experience. During CVS analysis, parameters such as temperature, scan rate, potential range, stirring speed, and number of cycles are usually configured based on experience, lacking a systematic and quantitative basis for setting.

[0004] Second, a quantitative evaluation system for the stability of voltammetric curves has not been established. Existing methods typically focus only on the final concentration analysis results, lacking a systematic evaluation mechanism for the characteristics of voltammetric curves, making it difficult to quantitatively assess the repeatability and stability of voltammetric curves.

[0005] Third, parameter tuning is inefficient and global optima are difficult to obtain. Traditional methods rely on manual trial and error adjustments, which can easily lead to local optima, failing to fully utilize equipment performance and resulting in limited repeatability and consistency of analysis results.

[0006] Fourth, the one-time optimization fails quickly. The "optimal parameters" found by traditional methods are only applicable to the initial state of a new machine. As the running time increases, non-stationary factors such as sensor drift, pipe scaling, and pump and valve wear will cause the analysis accuracy to drop rapidly. After the accuracy drops, manual repeated debugging and verification are required. Every 1 to 2 months, process engineers need to stop the machine to readjust the parameters, which takes 3 to 5 days each time, seriously affecting the production line utilization rate.

[0007] Furthermore, there are complex nonlinear relationships between process parameters such as temperature, scan rate, potential range, and stirring speed and the characteristics of the voltammetric curve. Traditional methods are unable to accurately identify key influencing factors and are also unable to achieve rapid optimization of parameter configuration. Summary of the Invention

[0008] According to an embodiment of the present invention, a method for optimizing CVS analysis parameters based on the stability evaluation of volt-ampere curves is provided, comprising the following steps: Step S1, Data Acquisition: Obtain the process parameters recorded during the CVS analysis and their corresponding stable range volt-ampere curves. The stable range volt-ampere curves are determined based on the minimum number of cycles and the charge tolerance set during the CVS analysis. Step S2, Data Preprocessing: Outlier handling is performed on the collected data, and the volt-ampere curves in the stable interval within the same group are divided according to the volume of standard sample added during the analysis. The stability evaluation index of the volt-ampere curve for each subset is calculated, including the mean stripping charge, the standard deviation of the stripping charge, the slope of the stripping charge, the mean peak current, the standard deviation of the peak current, and the standard deviation of the peak potential. In addition, key process parameters affecting the stability of the volt-ampere curve in the stable interval are selected as model input features based on the analysis method of the additives. Step S3, Model Training and Selection: Construct a machine learning model library, train each model on the training set, and select the best-performing model as the optimal model using the root mean square error as the evaluation index. This model is used to establish the mapping relationship between process parameters and the stability evaluation index of the volt-ampere curve. Step S4, Intelligent Parameter Optimization: The Bayesian optimization algorithm based on the Parzen estimator with a multi-objective tree structure is used to optimize the process parameters. The four optimization objectives are the relative standard deviation of the stripped charge, the relative standard deviation of the peak current, the standard deviation of the peak potential, and the absolute value of the stripped charge slope. The Pareto non-dominated relationship is used for collaborative optimization to obtain the Pareto optimal solution set, and the recommended combination of process parameters is determined according to the preset decision rules.

[0009] Furthermore, outlier handling in step S1 includes: deleting samples with a missing process parameter rate exceeding a preset threshold; deleting data with obvious anomalies, where obvious anomalies refer to data that do not meet the basic process constraints of CVS analysis; and filling missing values ​​in the remaining data with the median.

[0010] Furthermore, in step S3, the machine learning model library includes four models: XGBoost, LightGBM, CatBoost, and RandomForest. Each model is trained on a training set with K-fold cross-validation, and the default value of K is 5.

[0011] Furthermore, step S4 specifically includes: dividing the sample set into a superior sample set and a non-superior sample set based on the Pareto non-dominated ranking results of historical samples; establishing the probability density functions of the process parameters corresponding to the two types of samples using Parzen kernel density estimation; performing probability sampling on the process parameter space based on the probability density functions to generate new candidate process parameters; inputting the candidate process parameters into the optimal model and calculating four optimization objectives based on the model prediction results; adding the candidate process parameters and their corresponding calculation results as new historical samples to the historical sample set, re-performing the Pareto non-dominated ranking and updating the probability density functions corresponding to the two types of samples; repeating the above process until the preset stopping condition is met.

[0012] Furthermore, the preset stopping condition is reaching a preset number of iterations.

[0013] Furthermore, the four optimization objectives are expressed by the following formula:

[0014] in, This indicates the average amount of electricity removed. This indicates the standard deviation of the stripped charge. This represents the average peak current. Indicates the standard deviation of peak current. Indicates the standard deviation of peak potential. This indicates the slope of the stripped charge.

[0015] Furthermore, the CVS analysis includes inhibitor analysis, brightener analysis, and leveling agent analysis; among which, the inhibitor analysis adopts the cyclic voltammetric stripping-dilution titration method; the brightener analysis adopts the modified linear approximation method; and the leveling agent analysis adopts the response curve method.

[0016] Furthermore, the model input features for inhibitor analysis include: temperature, analysis low potential, analysis high potential, scan rate, integration upper limit, minimum number of cycles, charge tolerance, stirring speed, support liquid volume, copper chloride concentration, standard inhibitor concentration, and standard added volume. The model input features for brightener analysis include: temperature, analysis low potential, analysis high potential, scan rate, integration upper limit, minimum number of cycles, charge tolerance, stirring speed, support liquid volume, copper chloride concentration, inhibitor concentration in support liquid, standard brightener concentration, standard added volume, and test sample added volume. The model input features for leveling agent analysis include: temperature, analysis low potential, analysis high potential, scan rate, integration upper limit, minimum number of cycles, charge tolerance, stirring speed, support liquid volume, copper chloride concentration, inhibitor concentration in support liquid, brightener concentration in support liquid, standard leveling agent concentration, and standard sample addition volume.

[0017] Furthermore, the preset decision-making rules are determined based on the priority of the optimization objectives.

[0018] Furthermore, in step S4, the search range of each optimization variable is determined based on the allowable setting range of the CVS analyzer or the statistical range of historical process data, preferably based on the minimum and maximum values ​​of historical data.

[0019] According to the present invention, the CVS analysis parameter optimization method based on the stability evaluation of the current-voltage curve is to construct a predictive model between key process parameters and the stability index of the current-voltage curve, establish a stability evaluation system for the current-voltage curve, and combine it with an intelligent optimization algorithm to realize automatic parameter optimization, thereby forming a data-driven parameter optimization mechanism, thereby improving the repeatability and accuracy of CVS analysis, improving coating quality and process yield, and reducing production costs.

[0020] It should be understood that both the foregoing general description and the following detailed description are exemplary and intended to provide further illustration of the claimed technology. Attached Figure Description

[0021] Figure 1 This is a flowchart of a CVS analysis parameter optimization method based on stability evaluation of volt-ampere curves according to an embodiment of the present invention. Detailed Implementation

[0022] The preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, further illustrating the present invention.

[0023] First, combine Figure 1 This invention describes a method for optimizing CVS analysis parameters based on the stability evaluation of volt-ampere curves, which is applicable to the field of CVS analysis technology and has a wide range of applications.

[0024] like Figure 1 As shown, the CVS analysis parameter optimization method based on current-voltage curve stability evaluation according to an embodiment of the present invention includes the following steps: Step S1, Data Acquisition: Obtain the process parameters recorded during the CVS analysis and their corresponding stable range volt-ampere curves. The stable range volt-ampere curves are determined based on the minimum number of cycles and the charge tolerance set during the CVS analysis. Step S2, Data Preprocessing: Outlier handling is performed on the collected data, and the volt-ampere curves in the stable interval within the same group are divided according to the volume of standard sample added during the analysis. The stability evaluation index of the volt-ampere curve for each subset is calculated, including the mean stripping charge, the standard deviation of the stripping charge, the slope of the stripping charge, the mean peak current, the standard deviation of the peak current, and the standard deviation of the peak potential. In addition, key process parameters affecting the stability of the volt-ampere curve in the stable interval are selected as model input features based on the analysis method of the additives. Step S3, Model Training and Selection: Construct a machine learning model library, train each model on the training set, and select the best-performing model as the optimal model using the root mean square error as the evaluation index. This model is used to establish the mapping relationship between process parameters and the stability evaluation index of the volt-ampere curve. Step S4, Intelligent Parameter Optimization: The Bayesian optimization algorithm based on the Parzen estimator with a multi-objective tree structure is used to optimize the process parameters. The four optimization objectives are the relative standard deviation of the stripped charge, the relative standard deviation of the peak current, the standard deviation of the peak potential, and the absolute value of the stripped charge slope. The Pareto non-dominated relationship is used for collaborative optimization to obtain the Pareto optimal solution set, and the recommended combination of process parameters is determined according to the preset decision rules.

[0025] Furthermore, such as Figure 1 As shown, in this embodiment, outlier handling in step S1 includes: deleting samples with a missing process parameter rate exceeding a preset threshold; deleting data with obvious anomalies, where obvious anomalies refer to data that do not meet the basic process constraints of CVS analysis; and filling missing values ​​in the remaining data with the median.

[0026] Furthermore, such as Figure 1 As shown in this embodiment, in step S3, the machine learning model library includes four models: XGBoost, LightGBM, CatBoost, and Random Forest. Each model is trained on a training set with K-fold cross-validation, and the default value of K is 5.

[0027] Furthermore, such as Figure 1 As shown, in this embodiment, step S4 specifically includes: dividing the sample set into a superior sample set and a non-superior sample set based on the Pareto non-dominated ranking results of historical samples; establishing probability density functions for the process parameters corresponding to the two types of samples using Parzen kernel density estimation; performing probability sampling on the process parameter space based on the probability density functions to generate new candidate process parameters; inputting the candidate process parameters into the optimal model and calculating four optimization objectives based on the model prediction results; adding the candidate process parameters and their corresponding calculation results as new historical samples to the historical sample set, re-performing the Pareto non-dominated ranking and updating the probability density functions corresponding to the two types of samples; repeating the above process until a preset stopping condition is met.

[0028] Furthermore, such as Figure 1 As shown, in this embodiment, the preset stopping condition is reaching a preset number of iterations.

[0029] Furthermore, such as Figure 1 As shown, in this embodiment, the four optimization objectives are expressed by the following formulas:

[0030] in, This indicates the average amount of electricity removed. This indicates the standard deviation of the stripped charge. This represents the average peak current. Indicates the standard deviation of peak current. Indicates the standard deviation of peak potential. This indicates the slope of the stripped charge.

[0031] Furthermore, such as Figure 1 As shown, in this embodiment, CVS analysis includes inhibitor analysis, brightener analysis, and leveling agent analysis; wherein, inhibitor analysis adopts cyclic voltammetric stripping-dilution titration; brightener analysis adopts modified linear approximation method; and leveling agent analysis adopts response curve method.

[0032] Furthermore, such as Figure 1 As shown, in this embodiment, the model input features corresponding to inhibitor analysis include: temperature, analysis low potential, analysis high potential, scan rate, integration upper limit, minimum number of cycles, charge tolerance, stirring speed, support liquid volume, copper chloride concentration, standard sample inhibitor concentration, and standard sample addition volume. The model input features for brightener analysis include: temperature, analysis low potential, analysis high potential, scan rate, integration upper limit, minimum number of cycles, charge tolerance, stirring speed, support liquid volume, copper chloride concentration, inhibitor concentration in support liquid, standard brightener concentration, standard added volume, and test sample added volume. The model input features for leveling agent analysis include: temperature, analysis low potential, analysis high potential, scan rate, integration upper limit, minimum number of cycles, charge tolerance, stirring speed, support liquid volume, copper chloride concentration, inhibitor concentration in support liquid, brightener concentration in support liquid, standard leveling agent concentration, and standard sample addition volume.

[0033] Furthermore, such as Figure 1 As shown, in this embodiment, the preset decision rule is determined based on the priority of the optimization target.

[0034] Furthermore, such as Figure 1 As shown, in this embodiment, in step S4, the search range of each optimization variable is determined based on the allowable setting range of the CVS analyzer or the statistical range of historical process data, preferably based on the minimum and maximum values ​​of historical data.

[0035] Above, refer to Figure 1This paper describes a CVS analysis parameter optimization method based on current-voltage curve stability evaluation according to an embodiment of the present invention. By constructing a predictive model between key process parameters and current-voltage curve stability indicators, a current-voltage curve stability evaluation system is established. Combined with intelligent optimization algorithms, automatic parameter optimization is achieved, forming a data-driven parameter optimization mechanism. This improves the repeatability and accuracy of CVS analysis, enhances coating quality and process yield, and reduces production costs.

[0036] It should be noted that, in this specification, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes that element.

[0037] Although the present invention has been described in detail through the preferred embodiments above, it should be understood that the above description should not be considered as a limitation of the present invention. Various modifications and substitutions to the present invention will be apparent to those skilled in the art after reading the above description. Therefore, the scope of protection of the present invention should be defined by the appended claims.

Claims

1. A method for optimizing CVS analysis parameters based on the stability evaluation of volt-ampere curves, characterized in that, Includes the following steps: Step S1, Data Acquisition: Obtain the process parameters recorded during the CVS analysis and their corresponding stable range volt-ampere curves. The stable range volt-ampere curves are determined based on the minimum number of cycles and the charge tolerance set during the CVS analysis. Step S2, Data Preprocessing: Outlier handling is performed on the collected data, and the volt-ampere curves in the stable interval within the same group are divided according to the volume of standard sample added during the analysis. The stability evaluation index of the volt-ampere curve for each subset is calculated, including the mean stripping charge, the standard deviation of the stripping charge, the slope of the stripping charge, the mean peak current, the standard deviation of the peak current, and the standard deviation of the peak potential. In addition, key process parameters affecting the stability of the volt-ampere curve in the stable interval are selected as model input features based on the analysis method of the additives. Step S3, Model Training and Selection: Construct a machine learning model library, train each model on the training set, and select the best-performing model as the optimal model using the root mean square error as the evaluation index. This model is used to establish the mapping relationship between process parameters and the stability evaluation index of the current-voltage curve. Step S4, Intelligent Parameter Optimization: The Bayesian optimization algorithm based on the Parzen estimator with a multi-objective tree structure is used to optimize the process parameters. The four optimization objectives are the relative standard deviation of the stripped charge, the relative standard deviation of the peak current, the standard deviation of the peak potential, and the absolute value of the stripped charge slope. The Pareto non-dominated relationship is used for collaborative optimization to obtain the Pareto optimal solution set, and the recommended combination of process parameters is determined according to the preset decision rules.

2. The CVS analysis parameter optimization method based on current-voltage curve stability evaluation as described in claim 1, characterized in that, The outlier handling in step S1 includes: deleting samples with a missing process parameter rate exceeding a preset threshold; deleting data with obvious anomalies, where obvious anomalies refer to data that do not meet the basic process constraints of CVS analysis; and filling missing values ​​in the remaining data with the median.

3. The CVS analysis parameter optimization method based on current-voltage curve stability evaluation as described in claim 1, characterized in that, In step S3, the machine learning model library includes four models: XGBoost, LightGBM, CatBoost, and RandomForest. Each model is trained on a training set with K-fold cross-validation, and the default value of K is 5.

4. The CVS analysis parameter optimization method based on current-voltage curve stability evaluation as described in claim 1, characterized in that, Step S4 specifically includes: dividing the sample set into a superior sample set and a non-superior sample set based on the Pareto non-dominated ranking results of historical samples; establishing probability density functions for the process parameters corresponding to the two types of samples using Parzen kernel density estimation; performing probability sampling on the process parameter space based on the probability density functions to generate new candidate process parameters; inputting the candidate process parameters into the optimal model and calculating the four optimization objectives based on the model prediction results; adding the candidate process parameters and their corresponding calculation results as new historical samples to the historical sample set, re-performing the Pareto non-dominated ranking and updating the probability density functions corresponding to the two types of samples; repeating the above process until a preset stopping condition is met.

5. The CVS analysis parameter optimization method based on current-voltage curve stability evaluation as described in claim 4, characterized in that, The preset stopping condition is reaching a preset number of iterations.

6. The CVS analysis parameter optimization method based on current-voltage curve stability evaluation as described in claim 1, characterized in that, The four optimization objectives are expressed by the following formula: ， in, This indicates the average amount of electricity removed. This indicates the standard deviation of the stripped charge. This represents the average peak current. Indicates the standard deviation of peak current. Indicates the standard deviation of peak potential. This indicates the slope of the stripped charge.

7. The CVS analysis parameter optimization method based on current-voltage curve stability evaluation as described in claim 1, characterized in that, The CVS analysis includes inhibitor analysis, brightener analysis, and leveling agent analysis; wherein, the inhibitor analysis adopts the cyclic voltammetric stripping-dilution titration method; the brightener analysis adopts the modified linear approximation method; and the leveling agent analysis adopts the response curve method.

8. The CVS analysis parameter optimization method based on current-voltage curve stability evaluation as described in claim 7, characterized in that, The model input features for inhibitor analysis include: temperature, analysis low potential, analysis high potential, scan rate, integration limit, minimum number of cycles, charge tolerance, stirring speed, support liquid volume, copper chloride concentration, standard inhibitor concentration, and standard added volume. The model input features for brightener analysis include: temperature, analysis low potential, analysis high potential, scan rate, integration upper limit, minimum number of cycles, charge tolerance, stirring speed, support liquid volume, copper chloride concentration, inhibitor concentration in support liquid, standard brightener concentration, standard added volume, and test sample added volume. The model input features for leveling agent analysis include: temperature, analysis low potential, analysis high potential, scan rate, integration upper limit, minimum number of cycles, charge tolerance, stirring speed, support liquid volume, copper chloride concentration, inhibitor concentration in support liquid, brightener concentration in support liquid, standard leveling agent concentration, and standard sample addition volume.

9. The CVS analysis parameter optimization method based on current-voltage curve stability evaluation as described in claim 1, characterized in that, The preset decision rules are determined based on the priority of the optimization objectives.

10. The CVS analysis parameter optimization method based on current-voltage curve stability evaluation as described in claim 1, characterized in that, In step S4, the search range of each optimization variable is determined based on the allowable setting range of the CVS analyzer or the statistical range of historical process data.

Images