Quantitative analysis method and device for electroplating liquid additive

Abstract

The invention relates to the technical field of electroplate liquid analysis, in particular to a quantitative analysis method and device for an electroplate liquid additive. The method comprises the following steps: acquiring multi-dimensional electrochemical data of a plurality of groups of electroplating liquid samples; preprocessing the multi-dimensional electrochemical data and extracting multi-dimensional features related to the additive; inputting the multi-dimensional features and the corresponding additive concentration labels into a machine learning model for training, so as to establish a mapping relation between the additive concentration and the electrochemical signal through the machine learning model; and preprocessing electrochemical data of a to-be-detected electroplating solution, extracting multi-dimensional to-be-detected features, and inputting the to-be-detected features into the trained machine learning model to obtain the concentration of the additive in the to-be-detected electroplating solution. By means of the arrangement, the precision and robustness of quantitative analysis can be remarkably improved, and the method can be effectively suitable for electroplating liquid systems with different complexity degrees such as multiple components and high concentration difference.

Description

Technical Field

[0001] This invention relates to the field of electroplating solution chemical analysis technology, and in particular to a quantitative analysis method and apparatus for electroplating solution additives. Background Technology

[0002] In the field of high-end semiconductor device manufacturing, precision electroplating is a key process for preparing metal coatings with specific properties. The composition and concentration control of the electroplating solution play a decisive role in the macroscopic properties of the coating, such as quality and uniformity. Additives significantly affect the coating performance and morphology by regulating metal deposition kinetics. For example, inhibitors can form a current-suppressing film to ensure coating uniformity, brighteners can refine grains and improve brightness, and leveling agents can promote surface smoothing. Since both excessively high and low additive concentrations can lead to coating defects and affect product quality, their concentrations need to be monitored and quantitatively analyzed. However, there are synergistic or interfering effects between additives in complex electroplating solution systems. For example, in through-hole plating, leveling agents may weaken the anti-inhibition ability of brighteners, making precise concentration control even more challenging.

[0003] Currently, the main methods for quantitative analysis of additives in electroplating solutions include traditional instrumental analysis methods and electrochemical analysis methods. Traditional instrumental analysis methods, such as high-performance liquid chromatography (HPLC) and ion chromatography (IC), can measure the "absolute concentration" of additives, but they suffer from drawbacks such as offline sampling preventing real-time monitoring, expensive equipment and complex operation, low sensitivity to irreversibly adsorbed components, and inability to reflect the "effective concentration" of additives. Electrochemical quantitative methods, represented by cyclic voltammetry (CVS), are currently the mainstream testing methods, but they have limitations such as long testing cycles, high operational requirements, strict experimental conditions, limited measurement range, low testing efficiency, signal drift and poor repeatability, failure in high-concentration testing, and inability to accommodate multiple additives. Summary of the Invention

[0004] To address at least one deficiency in the existing methods for quantitative analysis of additives in electroplating solutions, this invention provides a method for quantitative analysis of additives in electroplating solutions, comprising the following steps: Multidimensional electrochemical data of multiple sets of electroplating solution samples were obtained, and the multiple sets of electroplating solution samples contained different concentration ratios of additives. The multidimensional electrochemical data is preprocessed and multidimensional features related to the additive are extracted; the multidimensional features and the corresponding additive concentration labels are input into a machine learning model for training, so as to establish a mapping relationship between additive concentration and electrochemical signal through machine learning model training; The electrochemical data of the electroplating solution to be tested are preprocessed and multi-dimensional features to be tested are extracted. The features to be tested are then input into the trained machine learning model to obtain the concentration of additives in the electroplating solution to be tested.

[0005] In some embodiments, multi-dimensional electrochemical data of multiple electroplating solution samples are collected and acquired using an electrochemical testing device; Collect multi-dimensional electrochemical data from multiple sets of electroplating solution samples, including the following steps: The types and concentration ranges of additives are determined, and the concentration range at least covers the dynamic range of additive concentration and edge concentration values ​​that can be achieved. Electroplating solution samples with uniformly distributed concentration ratios of multiple additives were generated using Latin hypercube sampling and orthogonal experimental design. The concentration ratio combinations met the following conditions: the concentration of each additive was randomly and uniformly distributed within a specified range; the total number of combinations should cover the synergistic / interference effects between additives; and the total volume of all combinations was consistent. Multiple electrochemical testing methods were applied to multiple sets of electroplating solution samples, and multi-dimensional electrochemical data of multiple sets of electroplating solution samples were collected.

[0006] In some embodiments, the multidimensional electrochemical data includes data acquired by at least one electrochemical testing method selected from cyclic voltammetry, electrochemical impedance spectroscopy, chronoamperometry, potential scan-step hybrid method, and multipotential step method.

[0007] In some embodiments, extracting multidimensional features related to the additive specifically includes the following steps: using at least one of principal component analysis, wavelet transform, or time-frequency analysis, extracting multidimensional features that are strongly correlated with the concentration of the additive from unnormalized multidimensional electrochemical data; the multidimensional features include at least one of cyclic voltammetry curve peak area, peak potential, electrochemical impedance spectroscopy phase angle change, and chronoamperometry current change rate.

[0008] In some embodiments, the machine learning model includes one or more of the following: linear regression model, ridge regression model, ridge regression cross-validation model, lasso regression model, lasso regression cross-validation model, LARS lasso regression model, LARS lasso regression cross-validation model, LARS lasso regression information criterion model, elastic network regression model, elastic network regression cross-validation model, minimum angle regression model, minimum angle regression cross-validation model, Bayesian ridge regression model, autocorrelation deterministic regression model, stochastic gradient descent regression model, passive attack regression model, RANSAC regression model, Theil-Sen regression model, Huber regression model, quantile regression model, Poisson regression model, gamma regression model, Tweedie regression model, orthogonal matching pursuit model, orthogonal matching pursuit cross-validation model, decision tree regression model, random forest regression model, gradient boosting regression model, AdaBoost regression model, K-nearest neighbor regression model, support vector regression model, and kernel ridge regression model.

[0009] In some embodiments, the method further includes automatically diluting high-concentration electroplating solution samples that exceed the preset detection range or preset sensitivity range of the electrochemical testing method using an electrochemical testing device; the dilution medium includes an original replenishment solution or electrolyte without additives, used to adjust the high-concentration electroplating solution sample to the linear response range of the electrochemical test.

[0010] In some embodiments, the method further includes cleaning and online activation of the electrodes during electrochemical testing of multiple sets of electroplating solution samples using an electrochemical testing device.

[0011] This invention also provides a quantitative analysis device for additives in electroplating solutions, including an electrochemical testing device and a data processing unit; The electrochemical testing device is used to prepare multiple sets of electroplating solution samples, perform various electrochemical testing methods on the multiple sets of electroplating solution samples, and collect multi-dimensional electrochemical data of the multiple sets of electroplating solution samples. The data processing unit is used to receive and acquire multi-dimensional electrochemical data of multiple sets of electroplating solution samples, each set of electroplating solution samples containing different concentration ratios of electroplating solution additives; preprocess the multi-dimensional electrochemical data and extract multi-dimensional features related to the additives; input the multi-dimensional features and corresponding additive concentration labels into a machine learning model for training, so as to establish a mapping relationship between additive concentration and electrochemical signal through machine learning model training; preprocess the electrochemical data of the electroplating solution to be tested and extract multi-dimensional features to be tested, input the features to be tested into the trained machine learning model, so as to obtain the concentration of additives in the electroplating solution to be tested.

[0012] In some embodiments, the electrochemical testing device includes a solution preparation unit; the solution preparation unit is used to automatically dilute high-concentration electroplating solution samples that exceed the preset detection range or preset sensitivity range of the electrochemical testing method; the dilution medium includes an original replenishment solution or electrolyte without additives, used to adjust the high-concentration electroplating solution sample to the linear response range of the electrochemical test.

[0013] In some embodiments, the electrochemical testing apparatus includes an electrode cleaning and online activation unit, which is used to clean and activate the electrodes online during the electrochemical testing of multiple sets of electroplating solution samples.

[0014] Based on the above, compared with the prior art, the quantitative analysis method for electroplating solution additives provided in this embodiment of the invention effectively analyzes the concentration of additives by utilizing multi-dimensional electrochemical data and training machine learning models. This not only significantly improves the accuracy and robustness of quantitative analysis but also effectively handles the problem of multi-component signal coupling in different complex electroplating solution systems. It effectively overcomes the problems of high cost, long processing time, and low accuracy of existing analytical methods, as well as the issue that traditional instrumental analysis methods can only measure "absolute concentration" and cannot provide "effective concentration." Furthermore, it solves the problems of existing testing techniques for the quantitative analysis of electroplating solution additives in multi-component systems with high concentration differences, such as inaccurate low-concentration testing due to excessive concentration differences, narrow test concentration range, signal drift and poor repeatability, and difficulties in quantitative analysis of multi-component synergistic effects.

[0015] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention 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 some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Unless otherwise specified, the positional relationships shown in the drawings in the following description are based on the direction in which the components are drawn in the figure.

[0017] Figure 1 A flowchart illustrating the steps of a quantitative analysis method for additives in electroplating solutions according to an embodiment of the present invention; Figure 2 These are exemplary results of the performance of the method provided in the embodiments of the present invention on different machine learning models on a test set; Figure 3 This is a schematic diagram of the structure of a quantitative analysis device for electroplating solution additives provided in an embodiment of the present invention; Figure 4 A flowchart illustrating the steps of a quantitative analysis method for electroplating solution additives provided in another embodiment of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments. The technical features designed in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0019] In the description of this invention, it should be noted that all terms used in this invention (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains, and should not be construed as limiting the invention; it should be further understood that the terms used in this invention should be understood to have the same meaning as those in the context of this specification and in the relevant field, and should not be understood in an idealized or overly formal sense, except as expressly defined in this invention.

[0020] Traditional instrumental analysis methods, such as high-performance liquid chromatography (HPLC) and ion chromatography (IC), can determine the "absolute concentration" of additives and distinguish specific components (such as SPS and MPS). However, these methods have the following shortcomings: (1) They require offline sampling and detection, which cannot meet the real-time monitoring needs of the dynamic changes in additive concentration during electroplating; (2) The instruments and equipment are expensive and the operation is complicated, relying on professional technicians; (3) They are only applicable to components that can be reversibly adsorbed on the chromatographic column, and have low detection sensitivity for irreversibly adsorbed components; (4) They ignore the influence of byproducts generated by chemical changes such as oxidation, reduction, cracking, and polymerization of additives on electroplating behavior, and can only mechanically reflect the "absolute concentration", and cannot provide the "effective concentration" that is directly related to the quality of electroplating.

[0021] In addition, the equipment currently used in large-scale commercial applications is based on cyclic voltammetric stripping (CVS). However, as the current mainstream electrochemical analysis technology, CVS equipment has the following significant shortcomings in practical applications: (1) Low testing efficiency: a single test takes several hours, which is difficult to cope with the frequently changing electroplating process; (2) Strong operational dependence: it requires highly skilled technicians to operate, and the experimental conditions (such as pH value, temperature, and electrode surface state) are strictly controlled, otherwise the repeatability of the experiment will easily decrease; (3) Limited concentration range: the measurement range of high-concentration additives (which are prone to entering the nonlinear response region due to concentration polarization or saturation effect) or low-concentration additives (weak signal) is limited; (4) Poor signal stability: organic additives and electrochemical byproducts in the electroplating solution are easily adsorbed and contaminated on the surface of the working electrode, resulting in the electrode active area being reduced. The reduction in electron transfer rate causes signal drift, decreased response sensitivity and poor measurement repeatability, making it difficult to distinguish small changes in concentration. Especially during long-term or continuous testing, the signal drift caused by changes in electrode surface state, solution environment and other factors is more obvious, which challenges the accuracy of quantitative analysis. (5) Insufficient multi-component analysis capability: Relying solely on a single parameter of the cyclic voltammetry curve (such as metal stripping charge), it is necessary to use dilution titration (DT), modified linear approximation (MLAT) and other techniques to group and quantify different additives, which is time-consuming and sample-intensive and cannot be analyzed quickly and synchronously. (6) Insufficient adaptability: Traditional single-parameter analysis methods are difficult to meet the high sensitivity, high accuracy and wide testing range requirements of multiple additives (which may act in different concentration ranges) at the same time.

[0022] In particular, for the quantitative analysis of additives in multi-component systems and electroplating solutions with strong concentration differences, existing technologies still face the following serious limitations: (1) Large concentration differences lead to inaccurate low-concentration tests: In multi-component systems, the concentrations of each additive differ significantly. The signal of low-concentration additives may be masked by high-concentration components, resulting in inaccurate test results. For example, the concentration of some additives may be several orders of magnitude lower than that of other components. This makes traditional instrumental analysis methods (such as high-performance liquid chromatography) insufficiently sensitive when detecting low-concentration components, making it difficult to achieve accurate measurements.

[0023] (2) Excessive components lead to significant synergistic effects and make component identification difficult: There are many types of additives in the electroplating solution, and there may be complex synergistic or interfering effects between the components. For example, in hole-filling electroplating, leveling agents may weaken the anti-inhibition ability of brighteners. This interaction makes quantitative analysis of single components difficult. In addition, chemical reactions between additives may lead to the formation of new substances, further increasing the complexity of component identification.

[0024] (3) Narrow test concentration range: Electrochemical methods such as CVS can usually obtain accurate results within a specific concentration range. For electroplating solutions with strong concentration differences, the test results may be inaccurate or invalid when the additive concentration exceeds this range. For example, in the presence of high concentrations of additives, the response of the CVS method may reach saturation and cannot accurately reflect the actual concentration of the additives; while in the presence of low concentrations of additives, the signal may be too weak to be accurately detected.

[0025] (4) Challenges of real-time monitoring: Traditional instrumental analysis methods usually require offline sampling and cannot achieve real-time monitoring. Although electrochemical analysis methods can monitor in real time, the interpretation and analysis of real-time data still face challenges due to the complexity of the electroplating solution system. For example, in electroplating solutions with large concentration differences, the real-time changes of low-concentration additives may be masked by the fluctuations of high-concentration components, resulting in inaccurate monitoring results.

[0026] (5) Difficulty in quantitative analysis of synergistic effects among multi-component systems: In multi-component systems, the synergistic effects between additives can significantly affect coating performance, but quantitative analysis of such effects is very complex. For example, some additives may exhibit synergistic effects within a specific concentration range, while exhibiting antagonistic effects within other concentration ranges. This nonlinear relationship makes it difficult to establish accurate quantitative models, further increasing the complexity of the analysis.

[0027] In summary, traditional techniques either fail to achieve real-time monitoring due to offline detection and high costs, or suffer from limited accuracy in analyzing complex systems due to single parameters and poor stability. Therefore, there is an urgent need for a more efficient, accurate, and adaptable quantitative analysis technique for additives. This invention provides a quantitative analysis method for electroplating solution additives, aiming to achieve a wider frequency range measurement, higher measurement accuracy and flexibility, and easier customization and expansion by users according to specific applications. In particular, it effectively addresses the quantitative analysis of additives in multi-component systems and electroplating solutions with significant concentration differences. This solves problems such as inaccurate low-concentration testing, narrow test concentration range, signal drift and poor repeatability, and difficulties in quantitative analysis of multi-component synergistic effects caused by excessive concentration differences in existing multi-component systems and / or additives with high concentration differences.

[0028] The present invention provides a quantitative analysis method and apparatus for electroplating solution additives, which are described in detail below with reference to specific embodiments.

[0029] Example 1 Please see Figure 1 The quantitative analysis method for electroplating solution additives provided in one embodiment of the present invention includes at least the following steps: Step S1: Obtain multi-dimensional electrochemical data of multiple sets of electroplating solution samples, wherein the multiple sets of electroplating solution samples contain different concentration ratios of additives; wherein, the multi-dimensional electrochemical data of multiple sets of electroplating solution samples can be collected and obtained by an electrochemical testing device.

[0030] This embodiment preferably employs a high-throughput automated electrochemical testing device. Multidimensional electrochemical data includes, but is not limited to, data acquired using at least one of the following electrochemical testing methods: cyclic voltammetry (CV), electrochemical impedance spectroscopy (EIS), chronoamperometry (CA), sweep-step functions (SSF), and multi-potential steps (MPS). In other words, the electrochemical testing device is used to perform the aforementioned chemical testing methods to obtain data, thereby constructing a comprehensive dataset containing information from multiple sets of electroplating solution samples in the time domain, frequency domain, and interface domain.

[0031] Step S2: Preprocess the multidimensional electrochemical data and extract multidimensional features related to the additive.

[0032] In practice, preprocessing may include, but is not limited to, noise filtering, smoothing, and baseline correction. If necessary, techniques such as linear interpolation can be used to convert all electrochemical data into potential-current data for processing, facilitating subsequent machine learning modeling. Preferably, extracting multidimensional features related to the additive specifically includes: using at least one of principal component analysis (PCA), wavelet transform, or time-frequency analysis to extract multidimensional features strongly correlated with the concentration of the additive from the unnormalized multidimensional electrochemical data; the multidimensional features include at least one of cyclic voltammetry peak area, peak potential, electrochemical impedance spectroscopy phase angle change, and chronoamperometry current change rate. Of course, multidimensional features may also include other numerical features at specific potentials, which should be reasonably selected according to actual needs; this embodiment does not limit this.

[0033] Step S3: Input the multi-dimensional features and corresponding additive concentration labels into the machine learning model for training, so as to establish the mapping relationship between additive concentration and electrochemical signal through machine learning model training.

[0034] In practical implementation, appropriate regression algorithms can be selected to construct the machine learning model. These models include, but are not limited to, one or more of the following: linear regression, ridge regression, ridge regression cross-validation model, lasso regression, lasso regression cross-validation model, LARS lasso regression, LARS lasso regression cross-validation model, LARS lasso regression information criterion model, elastic network regression, elastic network regression cross-validation model, minimum angle regression, minimum angle regression cross-validation model, Bayesian ridge regression, autocorrelation deterministic regression, stochastic gradient descent regression, passive attack regression, RANSAC regression, Theil-Sen regression, Huber regression, quantile regression, Poisson regression, gamma regression, Tweedie regression, orthogonal matching pursuit, orthogonal matching pursuit cross-validation model, decision tree regression, random forest regression, gradient boosting regression, AdaBoost regression, K-nearest neighbor regression, support vector regression, and kernel ridge regression. Of course, based on the concept of this invention, those skilled in the art can also employ other suitable regression algorithms to construct the model.

[0035] Based on the constructed machine learning model, the multi-dimensional features and corresponding additive concentration labels are input into the machine learning model for training to obtain a quantitative relationship model of "additive concentration-electrochemical signal". Specifically, machine learning libraries such as Scikit-learn can be used for training, and the model performance can be optimized through parameter tuning and other steps to finally obtain a model suitable for quantitative concentration analysis.

[0036] Preferably, this embodiment can also select the optimal hyperparameters through cross-validation (such as K-fold cross-validation) and evaluate the model performance based on metrics such as R² and root mean square error (RMSE) to ensure its robustness and generalization ability in complex systems. This step, through algorithm optimization and validation, achieves a high-precision, highly adaptable machine learning model. Figure 2 Exemplary results show the performance of the method provided in embodiments of the present invention on different machine learning models on a test set. Figure 2 The various subgraphs illustrate the performance of different machine learning models (such as LinearRegression, LassoCV, LassoLarsCV, BayesianRidge, ARDRegression, TheilSenRegressor, etc.) on the test set (measured by metrics such as RMSE and R²). Based on the distribution of data points in each subgraph (corresponding to different evaluation metrics), it can be inferred that each model exhibits good stability (i.e., the data points are relatively concentrated and the trend is stable, indicating relatively stable model performance). Therefore, the machine learning model trained using the embodiments of this invention demonstrates good performance.

[0037] It should be noted that, in this embodiment, the concentration information of different additives and the corresponding electrochemical data can also be used to construct and train various machine learning models, so as to enable machine learning models for different additives. Furthermore, since the electroplating solution sample already contains a combination of different additive concentration ratios, the electrochemical test only needs to be performed once, without the need for repeated steps.

[0038] Step S4: Preprocess the electrochemical data of the electroplating solution to be tested and extract multi-dimensional features to be tested. Input the features to be tested into the trained machine learning model to obtain the concentration of additives in the electroplating solution to be tested.

[0039] Specifically, when quantitatively analyzing additives in an electroplating solution, an electrochemical testing device and appropriate electrochemical testing methods can be used to collect electrochemical data from the solution. The collected electrochemical data is then preprocessed to extract multi-dimensional features. Finally, this data is input into a trained machine learning model to obtain the concentration information of the additives in the electroplating solution under the current real-time collected electrochemical data. By repeatedly inputting the electrochemical data into the optimized model corresponding to different additives, the predicted concentration values ​​of the additives to be quantitatively analyzed can be obtained.

[0040] The above steps can effectively obtain the concentration of additives in electroplating solutions, which can not only significantly improve the accuracy and robustness of quantitative analysis, but also effectively handle the problem of multi-component signal coupling under different complex electroplating solution systems. It can effectively overcome the problems of high cost, long time consumption and low accuracy of existing analytical methods, as well as solve the problem that traditional instrumental analysis methods can only measure "absolute concentration" and cannot provide "effective concentration".

[0041] In step S2, to avoid overwhelming weak signal features and to retain absolute signal strength information, normalization processing cannot be used. This invention first uses unnormalized multidimensional electrochemical data to extract multidimensional features with clear physicochemical significance or empirically proven validity related to additive concentrations. These features are then processed to more fully retain rich information related to the concentrations of each additive, providing high-quality input for constructing a more accurate "concentration-signal" mapping relationship in subsequent machine learning models. This strategy is particularly crucial for ensuring the sensitivity and accuracy of electroplating systems with multiple coexisting components, large differences in component concentrations, and complex interactions.

[0042] Furthermore, in step S1, multi-dimensional electrochemical data of multiple sets of electroplating solution samples are collected, including the following steps: Step S11: Determine the type and concentration range of the additive, wherein the concentration range at least covers the dynamic change range and edge concentration values ​​that the additive concentration can reach.

[0043] Specifically, based on the actual application requirements of the target electroplating solution system, the types and concentration ranges of the additives to be tested are determined. This range needs to cover the dynamic variation range of additive concentrations that can be achieved in actual production processes, and include edge concentration values ​​(such as the range of each additive concentration from 0 to 20% higher than the actual electroplating solution additive concentration) to construct a comprehensive calibration dataset. This step provides the basic framework for subsequent testing and model training, ensuring that the selected parameters can reflect the synergistic / interference effects of additives in complex electroplating solution systems.

[0044] Step S12: Using Latin hypercube sampling and orthogonal experimental design, electroplating solution samples with uniformly distributed concentration ratios of multiple additives are generated.

[0045] Specifically, the concentration generation strategy combining Latin hypercube sampling (LHS) and orthogonal experimental design (DOE) can construct a representative concentration sample set covering a high-dimensional concentration space, thereby improving the modeling efficiency of complex systems and expanding the coverage of concentration responses. The concentration combinations satisfy the following conditions: the concentration of each additive is randomly and uniformly distributed within a specified range to ensure the generality of model training; the total number of combinations should cover the synergistic / interference effects between additives to avoid the influence of reactions between additives; and the total volume of all combinations is consistent to eliminate the influence of volume deviations on electrochemical testing. Through these steps, a sample set covering a complex concentration space can be efficiently generated, providing highly representative data for model training.

[0046] Step S13: Perform various electrochemical testing methods on multiple groups of electroplating solution samples and collect multi-dimensional electrochemical data of multiple groups of electroplating solution samples.

[0047] In practice, an electrochemical testing device is used to perform various electrochemical testing methods to collect multi-dimensional electrochemical data from a series of electroplating solution samples containing specified additives with different concentration ratios. These electrochemical testing methods include, but are not limited to, cyclic voltammetry (CV), electrochemical impedance spectroscopy (EIS), chronoamperometry (CA), potential scan-step hybrid method (SSF), and multipotential step method (MPS). The specific required equipment and operating methods are existing conventional designs and will not be elaborated upon in this embodiment.

[0048] In one embodiment, the quantitative analysis method for the electroplating solution additives further includes automatically diluting the high-concentration electroplating solution sample that exceeds the preset detection range or preset sensitivity range of the electrochemical testing method using an electrochemical testing device; the dilution medium includes the original replenishment solution or electrolyte without additives, used to adjust the high-concentration electroplating solution sample to the linear response range of the electrochemical test.

[0049] In practice, an electrochemical testing device is used to precisely and automatically dilute the collected electroplating solution sample according to a preset program or real-time feedback. The dilution factor can be optimized based on the target analyte, base solution type, or approximate concentration range (for example, for the analysis of high-concentration inhibitors like PEG, a dilution of 10-50 times can be set; for trace accelerators like SPS, a lower dilution factor or no dilution may be used). The dilution medium is typically the original replenishment solution (VMS) without additives or a specific supporting electrolyte. Through precise dilution, the response signal of high-concentration additives can be moved out of the saturation region or the strong nonlinear region, allowing it to fall within the good linear response range of the detector. At the same time, the signal intensity of the main component is reduced, thereby reducing the masking effect on the trace component signal, allowing the previously masked low-concentration additive signal to be revealed and accurately measured. Of course, other suitable methods can also be used to achieve dilution, which is not limited in this embodiment.

[0050] In another embodiment, the quantitative analysis method for the electroplating solution additives further includes cleaning and online activation of the electrodes during the electrochemical testing of multiple sets of electroplating solution samples using an electrochemical testing device.

[0051] In practice, a standardized multi-step electrode cleaning and activation procedure can be executed via an electrochemical testing device at the intervals between each electroplating sample test or within a set test cycle. This procedure is programmable and typically includes the following steps: a chemical cleaning step, where specific solvents such as dilute acids, organic solvents, or specialized cleaning solutions flow through the electrochemical test cell to act on the electrode surface to remove organic adsorbates; an electrochemical cleaning / regeneration step, where a specific potential program is applied via an electrochemical workstation (e.g., cyclic voltammetry scanning in the -0.8V to +1.2V vs Ag / AgCl potential range, or maintaining a specific duration at a strong oxidation / reduction potential) to remove stubborn adsorbates and metal deposits through redox reactions; a thorough rinsing step, where the electrode and cell are thoroughly rinsed with deionized water or a base electrolyte; and an online activation / stabilization step, where a brief electrochemical treatment, such as a specific potential pretreatment of the electrode in the base electrolyte, is performed before the start of a new test to ensure the electrode surface reaches a stable and highly active initial state. This automated cleaning and online activation process significantly ensures the consistency and high activity of the working electrode surface, effectively eliminating the "memory effect" and accumulated contamination of the previous electroplating sample. Furthermore, experimental data shows that when the same standard electroplating solution sample is repeatedly tested for 24 consecutive hours, the relative standard deviation of the additive concentration can be reduced from 15% before cleaning to less than 5%, which greatly improves the long-term stability and batch-to-batch repeatability of the measurement signal and lays the foundation for high-resolution monitoring of concentration fluctuations of ±5%.

[0052] Based on the above, in order to solve the technical challenge that traditional single-parameter electrochemical analysis methods cannot simultaneously meet the requirements of high sensitivity, high accuracy, and wide testing range for multiple additives, the embodiments of the present invention achieve an optimized balance of these three aspects through the following synergistic technical means: (1) Multidimensional electrochemical data fusion; This invention breaks through the limitations of traditional single-parameter analysis by using a variety of electrochemical testing techniques (such as cyclic voltammetry (CV), electrochemical impedance spectroscopy (EIS), chronoamperometry (CA), potential scanning-step hybrid method (SSF), multipotential step method (MPS), etc.) to collect multidimensional electrochemical data, making the response characteristics of different techniques complementary. For example, cyclic voltammetry (CV) can provide the macroscopic influence of additives on redox reaction kinetics (such as peak current and peak potential changes), and is suitable for preliminary characterization over a wide concentration range; electrochemical impedance spectroscopy (EIS) can accurately detect electrode / solution interface characteristics (such as double layer capacitance, charge transfer resistance, and adsorption behavior), and is highly sensitive to changes in the interface behavior of low-concentration surface-active additives (such as leveling agents and some inhibitors); chronoamperometry (CA) or chronopotentialography (CP) can be used to study the kinetics of additive consumption, diffusion, and adsorption / desorption processes. Therefore, by integrating the aforementioned multidimensional electrochemical data as input to the machine learning model, the machine learning model can automatically learn and extract the most relevant feature combinations from these complementary multidimensional data, dynamically assign differentiated weights to different technical features for different additives or concentration ranges, and significantly improve the overall analytical performance.

[0053] (2) Automated Precise Sample Dilution: Addressing the issue that existing high-concentration samples cannot be effectively tested, this invention automatically and precisely dilutes high-concentration samples that exceed the optimal detection linear range or sensitivity region of a certain electrochemical testing method, adjusting their concentration to within the optimal operating range of the subsequent electrochemical testing method. This operation not only expands the upper limit of accurate measurement for high-concentration additives (avoiding nonlinear responses caused by concentration polarization or saturation effects), but also indirectly improves the relative sensitivity of low-concentration components in the same sample by reducing matrix interference, achieving simultaneous and accurate analysis of high and low concentration components.

[0054] (3) LHS-optimized sampling constructs a wide-range calibration sample set; through the Latin hypercube sampling (LHS) strategy, a uniform and representative concentration spatial distribution is achieved with fewer experimental points within a wide range covering additive concentrations from extremely low to excessive concentrations (e.g., 0.1 to 2 times the actual production dynamic range). This strategy ensures that the machine learning model has sufficient learning basis within the target test breadth, avoiding the problem of a sharp drop in predictive performance in the edge regions of the concentration range (e.g., extremely low or extremely high concentrations), and significantly improving the model's adaptability to the entire concentration range.

[0055] (4) Selection and Integration of Machine Learning Models: This embodiment of the invention provides diverse machine learning models, allowing for the selection or integration of models with different characteristics based on specific application scenarios. For example, independent optimization models can be trained for different additives, or complex models capable of handling multiple outputs can be constructed. Furthermore, for additives exhibiting different response characteristics in different concentration ranges, segmented modeling or ensemble learning strategies (such as gradient boosting trees) can be employed to leverage the model's inherent ability to handle complex nonlinear and large dynamic range data. Therefore, compared to traditional single models based on physicochemical formulas, machine learning models (especially ensemble models) are better at capturing nonlinear responses over a wide dynamic range and accurately distinguishing the contributions of different additives in complex signals.

[0056] Through the synergistic effect of multi-dimensional data fusion, automated dilution, LHS-optimized wide-range sampling, and intelligent machine learning model selection and integration, this invention effectively improves the testing range of additive concentrations to a wider extent (e.g., 10-900 ppm for inhibitors like PEG and 0.5-30 ppm for accelerators like SPS), while ensuring high sensitivity to low-concentration components (e.g., a detection limit of 0.1 ppm for SPS) and high accuracy in predicting the concentrations of all additive components (e.g., within their respective calibration ranges, the root mean square error (RMSE) between the predicted and actual concentrations of the main additives is less than 3%–5% of their measurement range, and the R² value is greater than 0.95). This overall performance improvement is difficult to achieve with a single technical means, providing technical support for the real-time and accurate monitoring of complex electroplating solution systems.

[0057] Example 2 Please see Figure 3 The present invention also provides a quantitative analysis device for additives in electroplating solutions. Specifically, the quantitative analysis device includes an electrochemical testing device and a data processing unit.

[0058] The electrochemical testing device is used to configure multiple sets of electroplating solution samples and perform various electrochemical testing methods on the multiple sets of electroplating solution samples, and to collect multi-dimensional electrochemical data of the multiple sets of electroplating solution samples; the data processing unit is used to receive and acquire multi-dimensional electrochemical data of multiple sets of electroplating solution samples, each of which contains different concentration ratios of electroplating solution additives; the multi-dimensional electrochemical data is preprocessed and multi-dimensional features related to the additives are extracted; the multi-dimensional features and corresponding additive concentration labels are input into a machine learning model for training, so as to establish a mapping relationship between additive concentration and electrochemical signal through machine learning model training; the electrochemical data of the electroplating solution to be tested is preprocessed and multi-dimensional features to be tested are extracted, and the features to be tested are input into the trained machine learning model to obtain the concentration of additives in the electroplating solution to be tested.

[0059] The data processing unit is electrically connected to the electrochemical testing device to receive multidimensional electrochemical data transmitted by the device, process the data, and construct and run machine learning models. An optional LCD screen can be installed to facilitate the display of experimental progress. The specific processing methods and functions are detailed in the above-described embodiments and will not be elaborated further here.

[0060] The electrochemical testing device includes, but is not limited to, a control unit and a solution preparation unit, an electrochemical testing cell unit, an electrochemical workstation, an electrode unit, and an electroplating solution sample injection unit electrically connected to the control unit.

[0061] Specifically, the solution preparation unit is used to accurately and rapidly prepare electroplating solution samples, standard solutions, or support solutions for testing. The unit may include multiple (e.g., five or more) detachable syringe pumps or other precision fluid control devices to achieve automated, high-throughput solution preparation.

[0062] The electrochemical test cell unit is used to contain the electroplating solution to be tested and to perform electrochemical tests. The design of the test cell should be reasonably configured to enable solution injection, testing, rapid drainage, and cleaning functions, thereby supporting high-throughput operation.

[0063] An electrochemical workstation is used to apply the potential or current program required for various electrochemical tests and to measure the corresponding electrochemical response signal. At least one electrochemical workstation can be configured in the device, preferably multiple workstations to improve testing efficiency or to verify the reliability between different devices. The specific circuit design, structural design, and functional design of the electrochemical workstation can be reasonably configured with reference to existing designs.

[0064] The electrode unit includes a working electrode (e.g., a rotating disk electrode), a reference electrode, and a counter electrode, used to perform electrochemical reactions and signal measurements.

[0065] The electroplating solution sample introduction unit is used to introduce the actual electroplating solution into the electrolytic cell of the electrochemical testing device. This unit can be implemented by an injection pump connected to the solution preparation unit, or by additional liquid introduction modules (such as an additional peristaltic pump).

[0066] The control unit is used for centralized control of various units. Specifically, it can be programmed through the API interface provided by the device to control the liquid inlet volume and mixing ratio of the solution preparation unit, control the electrochemical workstation to execute preset test programs (such as CV, EIS, CA, etc.), control the rotation speed of the rotating disk electrode, and control the liquid drainage and cleaning processes. Furthermore, the control unit also supports the stable and accurate acquisition of various multidimensional electrochemical data and transmits the data to the data processing unit.

[0067] The solution preparation unit is also used to automatically dilute high-concentration electroplating solution samples that exceed the preset detection range or preset sensitivity range of the electrochemical testing method; the dilution medium includes original replenishment solution or electrolyte without additives, used to adjust the high-concentration electroplating solution sample to the linear response range of the electrochemical test.

[0068] The electrochemical testing apparatus further includes an electrode cleaning and online activation unit, which is used to clean and activate the electrodes online during the electrochemical testing of multiple sets of electroplating solution samples. Specific cleaning and activation steps are described in the foregoing method embodiments and will not be repeated here.

[0069] Example 3 Please see Figure 4 This invention provides a detailed explanation of the specific implementation steps using the monitoring of additive consumption during the electroplating process of a copper-based acidic electroplating solution system as an example. Specifically, the electroplating solution may include: a Virgin Makeup Solution (VMS) containing 120 g / L copper sulfate pentahydrate, 60 g / L sulfuric acid, and 209 mg / L hydrochloric acid with a mass fraction of 37%; and various additives, including polyethylene glycol (PEG, used as an inhibitor), sodium polydithiopropane sulfonate (SPS, used as an accelerator), and Janus Green B (JGB, used as a leveling agent). In the exemplary initial formulation, the concentration of PEG can be 600 ppm, the concentration of SPS can be 20 ppm, and the concentration of JGB can be 15 ppm. The following preferred workflow can be implemented: Step 1. Establish a machine learning model for quantitative analysis of additives in electroplating solutions. Step 1.1: Prepare various solutions as electroplating solution samples for the electroplating solution system. For example, the following solutions can be prepared: 500 mL of initial replenishment VMS; 250 mL of VMS solution containing 9000 ppm PEG; 250 mL of VMS solution containing 300 ppm SPS; and 250 mL of VMS solution containing 300 ppm JGB.

[0070] Step 1.2: Using Latin hypercube sampling (LHS) technology, 125 different volume combinations of electroplating solution samples were generated, consisting of the solutions from Step 1.1, with the total volume of each combination controlled to be 10 mL. By adjusting the volume ratios, the concentrations of PEG, SPS, and JGB were randomly distributed and uniformly filled within the concentration ranges of 0-900 ppm, 0-30 ppm, and 0-30 ppm, respectively. It should be noted that the upper limit of the test concentration range of the additives should be higher than the upper limit of the corresponding additive concentration range in the actual electroplating solution to be monitored (in this embodiment, the upper limits of the test concentrations of PEG, SPS, and JGB set for the 125 solutions are 900 ppm, 30 ppm, and 30 ppm, respectively, which are higher than their initial concentrations of 600 ppm, 20 ppm, and 15 ppm in the actual electroplating solution, respectively).

[0071] Step 1.3: Utilize the electrochemical testing device to perform automated preparation and electrochemical testing of the 125 different volume combinations of electroplating solution samples described in Step 1.2. Perform electrochemical testing on each prepared electroplating solution sample. In this embodiment, each electroplating solution sample undergoes two cyclic voltammetry (CV) tests. Specific electrochemical testing conditions can be set as follows: for the first CV test, the scan rate is 0.25 V / s; for the second CV test, the scan rate is 0.5 V / s. The remaining electrochemical testing conditions are kept consistent in both CV tests: the rotation speed of the rotating disk electrode (RDE) is 1500 RPM; the CV scan range is 1.55 V to -0.26 V vs Ag / AgCl (saturated potassium chloride solution); the number of scan cycles is 1.

[0072] Step 1.4: Perform necessary preprocessing on the acquired multidimensional electrochemical data. This preprocessing may include, but is not limited to, noise filtering, smoothing, interpolation, and baseline correction, but does not include normalization. This preprocessing is used to preserve absolute signal intensity information such as the absolute value of the peak current in the CV curve, the absolute value of the impedance modulus at a specific frequency in the EIS spectrum, the absolute value of the current reduction, and the true proportional relationship between features in the complete voltage-current data sequence of the CV curve. This ensures that all electrochemical data are fully aligned in the specified dimensions (potential, time, etc.).

[0073] Step 1.5: Using feature extraction techniques such as principal component analysis (PCA), extract multidimensional features related to additive concentration from the multidimensional electrochemical data preprocessed in Step 1.4.

[0074] Step 1.6: Using the electrochemical data containing multi-dimensional features obtained in Step 1.5 and the corresponding concentration labels of the additive PEG as input, a suitable regression model is selected using machine learning tools such as Scikit-learn. The model performance is then optimized through parameter tuning and other steps to obtain a machine learning model suitable for PEG concentration quantitative analysis. Similarly, this step is repeated using the electrochemical data and the corresponding concentration labels of the additives SPS or JGB as input to obtain machine learning models suitable for SPS and JGB concentration quantitative analysis, respectively.

[0075] Step 2. Quantitative analysis of the change in additive concentration in the electroplating solution during the electroplating process. Step 2.1: In the actual production process, electroplating is performed using an electroplating solution. During the electroplating process, the electroplating solution sample introduction module integrated into the electrochemical testing device is used to automatically sample the electroplating solution in real time or at set intervals.

[0076] Step 2.2: Electrochemical data are collected from the electroplating solution to be tested in Step 2.1 using the electrochemical testing method described in Step 1.3. Subsequently, the obtained electrochemical data are processed using the data preprocessing method described in Step 1.4 and the feature extraction method described in Step 1.5 to extract features related to the additive concentration.

[0077] Step 2.3: Input the processed and extracted features obtained in Step 2.2 into the machine learning model trained in Step 1, and obtain the "equivalent concentration" of the three additives in the electroplating solution to be tested according to the mapping relationship established by the model.

[0078] Step 2.4: Output and manage the analysis results (e.g., the "equivalent concentration" of the additive, the trend analysis of concentration changes over time, etc.). This output and management may include, but is not limited to, displaying the results on a human-machine interface, storing the data in a database or file, or transmitting the results to a higher-level electroplating solution monitoring system or production management system.

[0079] This invention combines multidimensional electrochemical data acquisition with machine learning algorithms to construct a highly efficient, accurate, adaptable, and low-cost quantitative analysis method and monitoring device for electroplating solution additives, providing strong support for solving key technical problems in the electroplating industry.

[0080] In summary, compared with the prior art, the quantitative analysis method and apparatus for electroplating solution additives provided by the present invention have the following advantages: I. By combining multi-dimensional electrochemical data and machine learning algorithms, the accuracy and robustness of quantitative analysis of additives in electroplating solutions are significantly improved, effectively addressing the problem of multi-component signal coupling in complex systems. It can dynamically focus on electrochemical characteristics strongly correlated with additive concentration, thereby achieving accurate analysis of multi-component synergistic / interference effects in complex systems such as through-hole plating. Furthermore, it effectively solves the problems of existing testing techniques for quantitative analysis of additives in multi-component systems with high concentration differences in electroplating solutions, such as inaccurate low-concentration testing due to large concentration differences, narrow testing concentration range, signal drift and poor repeatability, and difficulties in quantitative analysis of multi-component synergistic effects.

[0081] II. A multidimensional concentration ratio generation strategy combining Latin hypercube sampling (LHS) and orthogonal experimental design (DOE) was adopted to achieve "equivalent black box" modeling and "wide interval-high precision" synergistic optimization. This strategy, by constructing a highly representative sample set covering the additive concentration space, can equivalently replace the cumbersome process of traditional component-by-component titration analysis, directly outputting the comprehensive influence of each additive on electroplating behavior (i.e., "equivalent concentration") without distinguishing the absolute content of specific chemical components, effectively and significantly improving testing efficiency and modeling capabilities, and adapting to various process systems.

[0082] Third, the embodiments of this invention effectively solve the problem of weak and difficult-to-distinguish response signals in the low concentration range of additives through an adaptive dilution and electrode surface state reset mechanism. Automated processes such as acid washing, water washing, and electrode drying ensure consistent electrode state, and combined with dynamic adjustment of CV test parameters, the resolution is improved by 3 to 5 times, and the concentration detection error is as low as ±5%.

[0083] Fourth, in the data processing stage, unnormalized multidimensional electrochemical data is used to extract multidimensional features related to additive concentration, thus preserving trace information from the test data. Comparative experimental results show that the unnormalized data model significantly outperforms the traditional normalization process in terms of feature extraction accuracy and concentration prediction accuracy, especially in cooperative interference systems. In other words, it effectively avoids the loss of weak effective signals caused by normalization operations, ensuring that the model extracts true response features and effectively improving the accuracy of the method.

[0084] Fifth, the device and method are highly adaptable and can be quickly adapted to different electroplating solution systems (such as copper plating, nickel plating, gold-cobalt plating, gold plating, etc.). This not only effectively improves testing efficiency and reduces the complexity of manual operation, making additive monitoring more convenient and efficient, but also enables universal application across process scenarios through parameter adaptive optimization algorithms (such as genetic algorithm optimization model hyperparameters). The quantitative analysis device utilizes general-purpose equipment and a modular design, reducing system costs by more than 80%, which is conducive to the promotion and application of the technology and provides strong support for technological upgrading and innovative development in the electroplating industry.

[0085] Furthermore, those skilled in the art should understand that although many problems exist in the prior art, each embodiment or technical solution of the present invention can be improved in only one or a few aspects, without necessarily solving all the technical problems listed in the prior art or the background art simultaneously. Those skilled in the art should understand that any content not mentioned in a claim should not be construed as a limitation on that claim.

[0086] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for quantitative analysis of additives in electroplating solutions, characterized in that, Includes the following steps: Multidimensional electrochemical data of multiple sets of electroplating solution samples were obtained, and the multiple sets of electroplating solution samples contained different concentration ratios of additives. The multidimensional electrochemical data is preprocessed and multidimensional features related to the additive are extracted; the multidimensional features and the corresponding additive concentration labels are input into a machine learning model for training, so as to establish a mapping relationship between additive concentration and electrochemical signal through machine learning model training; The electrochemical data of the electroplating solution to be tested are preprocessed and multi-dimensional features to be tested are extracted. The features to be tested are then input into the trained machine learning model to obtain the concentration of additives in the electroplating solution to be tested.

2. The quantitative analysis method for electroplating solution additives according to claim 1, characterized in that: Multidimensional electrochemical data of multiple electroplating solution samples were collected and obtained using an electrochemical testing device. Collect multi-dimensional electrochemical data from multiple sets of electroplating solution samples, including the following steps: The types and concentration ranges of additives are determined, and the concentration range at least covers the dynamic range of additive concentration and edge concentration values ​​that can be achieved. Electroplating solution samples with uniformly distributed concentration ratios of multiple additives were generated using Latin hypercube sampling and orthogonal experimental design. The concentration ratio combinations met the following conditions: the concentration of each additive was randomly and uniformly distributed within a specified range; the total number of combinations should cover the synergistic / interference effects between additives; and the total volume of all combinations was consistent. Multiple electrochemical testing methods were applied to multiple sets of electroplating solution samples, and multi-dimensional electrochemical data of multiple sets of electroplating solution samples were collected.

3. The quantitative analysis method for electroplating solution additives according to claim 1, characterized in that: The multidimensional electrochemical data includes data acquired by at least one of the following electrochemical testing methods: cyclic voltammetry, electrochemical impedance spectroscopy, chronoamperometry, potential scan-step hybrid method, and multipotential step method.

4. The quantitative analysis method for electroplating solution additives according to claim 1, characterized in that, Extracting multidimensional features related to the additive specifically includes the following steps: using at least one of principal component analysis, wavelet transform, or time-frequency analysis, extracting multidimensional features that are strongly correlated with the concentration of the additive from unnormalized multidimensional electrochemical data; the multidimensional features include at least one of cyclic voltammetry curve peak area, peak potential, electrochemical impedance spectroscopy phase angle change, and chronoamperometry current change rate.

5. The quantitative analysis method for electroplating solution additives according to claim 1, characterized in that: The machine learning models include one or more of the following: linear regression model, ridge regression model, ridge regression cross-validation model, lasso regression model, lasso regression cross-validation model, LARS lasso regression model, LARS lasso regression cross-validation model, LARS lasso regression information criterion model, elastic network regression model, elastic network regression cross-validation model, minimum angle regression model, minimum angle regression cross-validation model, Bayesian ridge regression model, autocorrelation deterministic regression model, stochastic gradient descent regression model, passive attack regression model, RANSAC regression model, Theil-Sen regression model, Huber regression model, quantile regression model, Poisson regression model, gamma regression model, Tweedie regression model, orthogonal matching pursuit model, orthogonal matching pursuit cross-validation model, decision tree regression model, random forest regression model, gradient boosting regression model, AdaBoost regression model, K-nearest neighbor regression model, support vector regression model, and kernel ridge regression model.

6. The quantitative analysis method for electroplating solution additives according to claim 2, characterized in that: It also includes the automatic dilution of high-concentration electroplating solution samples that exceed the preset detection range or preset sensitivity range of the electrochemical testing method using an electrochemical testing device; The dilution medium includes the original replenishment solution or electrolyte without additives, used to adjust high-concentration electroplating solution samples to the linear response range of electrochemical tests.

7. The quantitative analysis method for electroplating solution additives according to claim 2, characterized in that: It also includes cleaning and online activation of electrodes during the electrochemical testing of multiple electroplating solution samples using an electrochemical testing device.

8. A quantitative analysis device for additives in electroplating solutions, characterized in that: Includes electrochemical testing equipment and data processing units; The electrochemical testing device is used to prepare multiple sets of electroplating solution samples, perform various electrochemical testing methods on the multiple sets of electroplating solution samples, and collect multi-dimensional electrochemical data of the multiple sets of electroplating solution samples. The data processing unit is used to receive and acquire multi-dimensional electrochemical data of multiple sets of electroplating solution samples, each set of electroplating solution samples containing different concentration ratios of electroplating solution additives; preprocess the multi-dimensional electrochemical data and extract multi-dimensional features related to the additives; input the multi-dimensional features and corresponding additive concentration labels into a machine learning model for training, so as to establish a mapping relationship between additive concentration and electrochemical signal through machine learning model training; preprocess the electrochemical data of the electroplating solution to be tested and extract multi-dimensional features to be tested, input the features to be tested into the trained machine learning model, so as to obtain the concentration of additives in the electroplating solution to be tested.

9. The quantitative analysis device for electroplating solution additives according to claim 8, characterized in that: The electrochemical testing device includes a solution preparation unit; the solution preparation unit is used to automatically dilute high-concentration electroplating solution samples that exceed the preset detection range or preset sensitivity range of the electrochemical testing method; The dilution medium includes the original replenishment solution or electrolyte without additives, used to adjust high-concentration electroplating solution samples to the linear response range of electrochemical tests.

10. The quantitative analysis device for electroplating solution additives according to claim 8, characterized in that: The electrochemical testing device includes an electrode cleaning and online activation unit, which is used to clean and activate the electrodes online during the electrochemical testing of multiple sets of electroplating solution samples.

Images