A method for analyzing concentration of acid copper plating solution additive based on double working electrode

Abstract

This application provides an analytical method for the concentration of additives in acid copper plating solutions based on a dual-working-electrode system. An electrochemical testing device with a nested ring electrode and a disk electrode is used to collect CV curves of the disk electrode and oxidation current curves of the ring electrode in a series of standard acid copper plating solutions. The CV curves and oxidation current curves are obtained by measuring the characteristic potential value E of the disk electrode at the same time point. d The characteristic current value I of the disk electrode d Characteristic oxidation current value I of the ring electrode r A one-to-one correspondence is established, with the ring electrode providing a fixed oxidation potential; a characteristic current value I is established. d Characteristic oxidation current value I r A machine learning model or standard working curve relating the three-dimensional mapping relationship between the additive concentration C and the target plating solution; the characteristic current value I of the plating solution to be tested. d Characteristic oxidation current value I r The additive concentration can be calculated by inputting the data into a machine learning model or a standard working curve. This dual-working-electrode method ensures that the analysis is unaffected by the electrode state, resulting in high detection accuracy.

Description

Technical Field

[0001] This invention relates to the field of plating solution additive analysis technology, and specifically to a method for analyzing the concentration of additives in acid copper plating solutions based on a dual-working-electrode system. Background Technology

[0002] Copper, due to its excellent conductivity, ductility, and high reliability, is widely used as an interconnect material in integrated circuits and printed circuit boards (PCBs). Currently, acidic copper plating is the most widely used electroplating copper process in the electronics manufacturing industry. In practical applications of acidic copper plating, suitable additives are generally added to enhance the performance of the copper layer. These additives can not only finely control the electrocrystallization process of copper, affecting its nucleation and growth kinetics, but also, under appropriate ratios, achieve efficient filling of complex via structures, thus meeting the needs of modern high-density interconnect technologies. However, if the concentration of the additives changes and becomes unbalanced during use, even slight deviations can lead to instability in the electrocrystallization process, directly affecting the uniformity of the plating layer, the filling effect of vias, and the reliability of the final product. Therefore, real-time monitoring and precise control of the additive concentration, ratio, and consumption status are crucial.

[0003] Existing technologies have proposed using cyclic voltammetric stripping (CVS) technology. Its core advantage lies in the process of "electrochemical enrichment-dissolution," which involves stripping away trace amounts of copper deposits. By utilizing the inhibitory or accelerating effects of additives on the copper deposition rate, the effective concentration of the additives can be indirectly and sensitively reflected, thus providing a basis for process control. However, CVS is inherently an indirect and relative analytical method. This method is extremely sensitive to the state of the electrode surface; even minute changes can directly cause fluctuations in the measurement signal, introducing significant data errors and severely affecting the repeatability and reliability of the results. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide an analytical method for the concentration of additives in acid copper plating solution based on dual working electrodes; aiming to solve the problem of poor detection reliability caused by the traditional CVS single working electrode method being easily affected by changes in electrode surface state.

[0005] The technical solution provided in this application is as follows:

[0006] An analytical method for additive concentration in acid copper plating solutions based on a ring-disk dual-working electrode system is proposed. This method utilizes an electrochemical testing apparatus with a nested ring electrode and a disk electrode as the dual working electrodes. It collects CV curves on the disk electrode and oxidation current curves on the ring electrode in a series of standard acid copper plating solutions. The CV curves are obtained by simultaneously measuring the characteristic potential E of the disk electrode. d The characteristic current value I of the disk electrode d Correspondingly, the oxidation current curve was established by measuring the characteristic potential value E of the disk electrode at the same time. dThe characteristic oxidation current value I of the ring electrode r Synchronous correspondence is established, with a fixed oxidation potential provided on the ring electrode; based on data from the CV curve and oxidation current curve, a characteristic current value I is established. d Characteristic oxidation current value I r A machine learning model or standard working curve relating the three-dimensional mapping relationship between the additive concentration C and the target plating solution; The characteristic current value I of the plating solution to be tested. d 'and characteristic oxidation current value I r 'Input the data into a machine learning model or a standard working curve to obtain the concentration of additives in the plating solution to be tested.'

[0007] Furthermore, based on the data from the CV curve and oxidation current curve, a dissolution current I was established. d and oxidation current I r Machine learning models mapping the relationship between additive concentration C and additive concentration C include:

[0008] The CV curves and oxidation current curves of multiple sets of standard acid copper plating solutions were preprocessed, and the characteristic potential value E related to the additive concentration C was extracted. d、 Characteristic current value I r and characteristic oxidation current value I r Multiple sets of standard acid copper plating solutions contain different combinations of additive concentrations.

[0009] The characteristic potential value E d、 Characteristic current value I d and characteristic oxidation current value I r The corresponding additive concentration C is input into a machine learning model for training, so as to establish a relationship between the additive concentration C and the characteristic current value I through machine learning model training. d and characteristic oxidation current value I r The three-dimensional mapping relationship.

[0010] Furthermore, based on the data from the CV curve and oxidation current curve, a dissolution current I was established. d and oxidation current I r Standard working curves showing the mapping relationship between additive concentration C and the additive concentration C include:

[0011] Data preprocessing was performed on the CV curves and oxidation current curves of multiple sets of standard acid copper plating solutions, based on the obtained characteristic potential value E. d、 Characteristic current value I d and characteristic oxidation current value I r Calculate the dissolution peak area Q of the CV curve related to additive concentration. d The oxidation current peak area Q of the oxidation current curve r Multiple sets of standard acid copper plating solutions contain different concentration ratios of additives.

[0012] Establish the dissolution peak area Q d With oxidation current peak area Q r The ratio Q r / Q d Or Q d / Q r The working curve Y varies with the additive concentration C.

[0013] Optionally, a scanning potential E is applied to the disk electrode. d The value is 0.5~0.8V; a fixed oxidation potential E is applied to the ring electrode. r 0.1V≤E r <2.0V (vs. Hg / Hg2SO4).

[0014] Optionally, CV curves on the disk electrode and oxidation current curves on the ring electrode are collected in a series of standard acid copper plating solutions, specifically including:

[0015] The types and concentration ranges of additives in a series of standard acid copper plating solutions are determined, and the concentration range of additives should at least cover the dynamic change range and edge concentration values ​​of additive concentrations; each series of standard acid copper plating solutions includes a base solution with constant components and concentrations.

[0016] Set up a series of standard acid copper plating solutions with uniformly distributed concentration ratios of multiple additives.

[0017] Multiple electrochemical testing methods were applied to various series of standard acid copper plating solutions, and CV curves on the disk electrode and oxidation current curves on the ring electrode were collected for each series of standard acid copper plating solutions.

[0018] Optionally, various electrochemical testing methods are applied to each series of standard acid copper plating solutions, including electrode cleaning and activation of the electrochemical testing device during the electrochemical testing of the series of standard acid copper plating solutions.

[0019] Optionally, the base solution includes 20-80 g / L copper ions, 10-120 g / L sulfuric acid, and 10-100 g / L chloride ions; the additives include at least one of inhibitors, accelerators, and leveling agents.

[0020] Optionally, the electrochemical testing device is a three-electrode system, which also includes a reference electrode of any one of mercury / mercurous sulfate electrode, saturated calomel electrode, and silver / silver chloride electrode; a counter electrode of any one of inert metal counter electrodes such as platinum, gold, and graphite electrode; a disk electrode of any one of glassy carbon electrode, platinum electrode, and gold electrode; and a ring electrode of any one of glassy carbon electrode, platinum electrode, and gold electrode.

[0021] The method for analyzing the additive concentration in acid copper plating solution provided in this application employs a dual-working-electrode scheme combining a disk electrode and a ring electrode. A cyclic voltammetric voltage is applied to the disk electrode, while a fixed oxidation voltage is continuously supplied to the ring electrode. When metallic copper deposited on the disk electrode at a negative potential dissolves, monovalent copper ions are first generated, and then further oxidized to divalent copper ions. A portion of the monovalent copper ions generated during the dissolution process diffuses to the ring electrode and is collected, triggering a signal detector on the ring electrode. The current dissolution peak area on the disk electrode corresponds to the total amount of monovalent and divalent copper ions generated during the dissolution of deposited copper, while the oxidation peak current area collected by the ring electrode corresponds to the partial diffusion of monovalent copper ions generated during the dissolution stage of the disk electrode. When the distance and diameter between the disk electrode and the ring electrode are fixed, the collection coefficient k of this ion on the ring electrode is a constant value. Therefore, by obtaining the characteristic oxidation current value I on the ring electrode... r The characteristic current value I of the disk electrode d Characteristic current value I can be established. d Characteristic oxidation current value I r Three-dimensional mapping relationship with additive concentration C; due to monovalent copper ions CE + There is a highly correlated response relationship between the content of the additive and the concentration of the additive; therefore, the concentration of the additive in the copper plating system can be analyzed and detected through a three-dimensional mapping relationship. Compared with a single working electrode, the results of this application are not affected by changes in the electrode state, and the detection accuracy is higher. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of a ring-disk dual working electrode structure used in this application;

[0024] Figure 2(a) shows the CV curves on the disk electrode and the oxidation current curves on the ring electrode under different concentrations of inhibitors in the plating solution without accelerators and inhibitors.

[0025] Figure 2(b) shows the CV curves on the disk electrode and the oxidation current curves on the ring electrode in a plating solution saturated with both accelerator and inhibitor;

[0026] Figure 2(c) shows the CV curves on the disk electrode and the oxidation current curves on the ring electrode under different concentrations of accelerator in the plating solution with inhibitor saturation and accelerator fixation.

[0027] Figure 3(a) shows the relative error distribution of inhibitor concentrations predicted based on the machine learning model K1;

[0028] Figure 3(b) shows the relative error distribution of the leveling agent concentration predicted based on the machine learning model K2;

[0029] Figure 3(c) shows the relative error distribution of accelerator concentration predicted based on the machine learning model K3;

[0030] Figure 4(a) shows the standard working curve of the inhibitor established using the dual working electrode method;

[0031] Figure 4(b) shows the standard working curve of the leveling agent established using the dual working electrode method;

[0032] Figure 4(c) shows the standard working curve of the accelerator established using the dual working electrode method. Detailed Implementation

[0033] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0034] An analytical method for additive concentration in acid copper plating solutions based on a ring-disk dual-working electrode system is proposed. This method utilizes an electrochemical testing apparatus with a nested ring electrode and a disk electrode as the dual working electrodes. It collects CV curves on the disk electrode and oxidation current curves on the ring electrode in a series of standard acid copper plating solutions. The CV curves are obtained by simultaneously measuring the characteristic potential E of the disk electrode. d The characteristic oxidation current value I of the disk electrode d Correspondingly, the oxidation current curve was established by measuring the characteristic potential value E of the disk electrode at the same time. d The characteristic oxidation current value I of the ring electrode r Synchronous correspondence is established, and a fixed oxidation potential is provided on the ring electrode;

[0035] Based on the data from the CV curve and oxidation current curve, the characteristic current value I is established. d Characteristic oxidation current value I r A machine learning model or standard working curve relating the additive concentration C to a three-dimensional mapping.

[0036] The characteristic current value I of the plating solution to be tested d 'and characteristic oxidation current value I r 'Input the data into a machine learning model or a standard working curve to obtain the concentration of additives in the plating solution to be tested.'

[0037] In some embodiments, a suitable regression algorithm is selected to construct the machine learning model. The algorithm for the machine learning model includes, but is not limited to, one or more of gradient-based optimization methods and second-derivative-based optimization methods. Gradient-based optimization methods include gradient descent and Adam algorithms; or, second-derivative-based optimization methods include Newton's method and L-BFGS algorithms.

[0038] In some embodiments, such as Figure 1 As shown, Figure 1 This is a schematic diagram of a ring-disk dual working electrode D-WE provided in an embodiment of this application. In the ring-disk dual working electrode D-WE, the ring electrode 01 and the disk electrode 02 are nested together, with the disk electrode 02 in the middle and the ring electrode 01 on the periphery. Preferably, the disk electrode 02 and the ring electrode 01 are arranged concentrically. The disk electrode 02 can be made of platinum, gold, glassy carbon, or other inert metals, and its diameter can be 2-5 mm. The inner diameter of the ring electrode 01 is 2.5-6 mm, and its outer diameter is 4.5-8 mm.

[0039] In some embodiments, the electrochemical testing device is a three-electrode system, which further includes a reference electrode of any one of a mercury / mercurous sulfate electrode, a saturated calomel electrode, and a silver / silver chloride electrode; a counter electrode of any one of an inert metal counter electrode such as a platinum, gold, or graphite electrode; a disk electrode of any one of a glassy carbon electrode, a platinum electrode, and a gold electrode; and a ring electrode of any one of a glassy carbon electrode, a platinum electrode, and a gold electrode.

[0040] In some embodiments, the scanning potential E applied to the disk electrode is... d The oxidation potential is 0.5~0.8V; a fixed oxidation potential Er is applied to the ring electrode, 0.1V≤E r <2.0V vs. Hg / Hg2SO4).

[0041] In some embodiments, the base plating solution includes 2-80 g / L copper ions, 10-120 g / L sulfuric acid, and 10-100 g / L chloride ions; the additives include at least one of inhibitors, accelerators, and leveling agents.

[0042] In some embodiments, the dissolution current I is established based on data from CV curves and oxidation current curves. d and oxidation current I r Machine learning models mapping the relationship between additive concentration C and additive concentration C include:

[0043] The CV curves and oxidation current curves of multiple sets of standard acid copper plating solutions were preprocessed, and characteristic current values ​​I related to additive concentration were extracted. d and characteristic oxidation current value I r Multiple sets of standard acid copper plating solutions contain different combinations of additive concentrations.

[0044] The characteristic current value I d and characteristic oxidation current value I r The corresponding additive concentration C is input into a machine learning model for training, so as to establish a relationship between the additive concentration C and the characteristic current value I through machine learning model training. d and characteristic oxidation current value I r The three-dimensional mapping relationship.

[0045] Optionally, data preprocessing includes converting all CV curves and oxidation current curves into "characteristic potential value - characteristic current value" data for processing, and establishing a one-to-one correspondence with the additive concentration; in order to facilitate the subsequent machine learning modeling process.

[0046] It should be noted that, in this embodiment, the concentration information of different additives and the corresponding CV curves and oxidation current curves can also be used to construct and train various machine learning models to obtain machine learning models for different additives. Furthermore, since the series of standard acid copper plating solutions already includes different concentration ratios of additives, the electrochemical test only needs to be performed once, without the need for repeated steps.

[0047] In some embodiments, the dissolution current I is established based on data from CV curves and oxidation current curves. d and oxidation current I r Standard working curves showing the mapping relationship between additive concentration C and the additive concentration C include:

[0048] The CV curves and oxidation current curves of multiple sets of standard acid copper plating solutions were preprocessed. Based on the obtained characteristic potential value Ed, characteristic current value Id, and characteristic oxidation current value Ir, the dissolution peak current area Q of the CV curve related to the additive concentration was calculated. d The oxidation peak current area Q of the oxidation current curve r Multiple sets of standard acid copper plating solutions contain different concentration ratios of additives.

[0049] Establish the dissolution peak current area Q d With oxidation peak current area Q r The ratio Q r / Q d Or Q d / Q r The working curve Y varies with the additive concentration C.

[0050] In some embodiments, CV curves on the disk electrode and oxidation current curves on the ring electrode are collected in a series of standard acid copper plating solutions, specifically including:

[0051] The types and concentration ranges of additives in a series of standard acid copper plating solutions are determined, and the concentration range of additives should at least cover the dynamic change range and edge concentration values ​​of additive concentrations; each series of standard acid copper plating solutions includes a base solution with constant components and concentrations.

[0052] Set up a series of standard acid copper plating solutions with uniformly distributed concentration ratios of multiple additives.

[0053] Multiple electrochemical testing methods were applied to various series of standard acid copper plating solutions, and CV curves on the disk electrode and oxidation current curves on the ring electrode were collected for each series of standard acid copper plating solutions.

[0054] In some embodiments, various electrochemical testing methods are performed on each series of standard acid copper plating solutions, including electrode cleaning and activation of the electrochemical testing device during the electrochemical testing of the series of standard acid copper plating solutions.

[0055] The series of standard acid copper plating solutions includes a base solution and a series of additives to be tested at different concentrations. The additives to be tested include any one of inhibitors, accelerators, and leveling agents.

[0056] A dual-working-electrode system was used to collect and calculate the ratio Q of the dissolution peak current area for any unknown plating solution. r ' / Q d 'or Q d ' / Q r ', the ratio of dissolution peak area Q r ' / Q d 'or Q d ' / Q r Substituting the values ​​into the standard working curve Y, we can obtain the concentration of the additive to be tested in the unknown plating solution.

[0057] Electrochemical analysis of acid copper plating solution is based on Cu 2+ As a two-electron reduction, but in the presence of additives, divalent copper ions Cu 2+ The transformation process with metallic copper atoms also involves monovalent copper ions (Cu) in the transition state. + Therefore, only divalent copper ions (Cu) are utilized. 2+ It is inaccurate to directly determine the amount of copper plating and the concentration of additives based on the stripping charge.

[0058] This application employs a dual-working-electrode scheme combining a disk electrode and a ring electrode. A cyclic voltammetric voltage is applied to the disk electrode, while a fixed oxidation voltage is continuously supplied to the ring electrode. When metallic copper deposited on the disk electrode at a negative potential dissolves, monovalent and divalent copper ions are generated. Some of the monovalent copper ions generated during the dissolution process diffuse to the ring electrode and are collected, triggering a signal detector on the ring electrode. The current dissolution peak area on the disk electrode corresponds to the total amount of monovalent and divalent copper ions generated during the dissolution of deposited copper, while the redox peak current area collected by the ring electrode corresponds to the partial diffusion of monovalent copper ions generated during the dissolution stage of the disk electrode. When the plating solution system is stable, its diffusion coefficient of monovalent copper ions is stable, and the collection coefficient k of this ion on the ring electrode is also a constant value. Therefore, by obtaining the characteristic oxidation current value I of the redox peak on the ring electrode... r The characteristic current value I of the dissolution peak of the disk electrode d Characteristic current value I can be established. d Characteristic oxidation current value I r Three-dimensional mapping relationship with additive concentration C; due to monovalent copper ions CE + There is a highly correlated response relationship between the content of the additive and the concentration of the additive; therefore, the concentration of the additive in the copper plating system can be analyzed and detected through a three-dimensional mapping relationship. Compared with a single working electrode, the results of this application are not affected by changes in the electrode state, and the detection accuracy is higher.

[0059] The preparation method and excellent effects of this application are illustrated in detail below through two specific embodiments.

[0060] Example 1

[0061] This embodiment uses a machine learning model;

[0062] 1. The electrochemical working device includes: dual working electrodes D-WE including: WE1, a 4mm platinum disk electrode; WE2, a platinum ring electrode with an inner diameter of 6.25mm and an outer diameter of 7.92mm, rotating at 3000rpm; reference electrode RE: Hg / Hg2SO4 (external salt bridge is saturated potassium nitrate); counter electrode: platinum rod.

[0063] 2. Electrode experimental conditions: On the disk electrode: CVS scan at a scan rate of 0.1V / s; on the ring electrode: a constant +0.8V is applied for chronocurrent scanning.

[0064] 3. Scheme for establishing a machine learning sample database:

[0065] With CV scanning applied to the disk electrode and a fixed oxidation potential applied to the ring electrode, CVS curves on the disk electrode and oxidation current curves on the ring electrode were acquired for each standard acid copper plating solution sample. Characteristic group data M under each curve were recorded. The characteristic potentials of each potential characteristic group M were scanned from -0.7V to 0.8V, corresponding to the current characteristic value group I on the disk electrode. 1M1 and the current characteristic value group I on the ring electrode 2M2 The data includes the CVS curve and oxidation current curve of each individual sample, along with their corresponding potential characteristic groups and current characteristic values. The CVS curves and oxidation current curves of all individual samples under each test scheme are compiled into a sample set database file D.

[0066] It should be noted that before performing electrochemical data acquisition on each single sample, the working electrode is pre-activated using a base plating solution. After each sample test, electrode cleaning is performed: the single-sample plating solution in the electrochemical measurement cell is drained, and the electrochemical measurement cell and each detection electrode are rinsed with deionized water in preparation for the next set of tests.

[0067] 4. Machine learning and model training scheme design:

[0068] The establishment and training of its machine learning model can refer to the technical solution disclosed in the invention CN202510715830.0 previously submitted by the inventor of this application. For example:

[0069] In the machine learning and data training unit, after data processing and model building on the sample database file D, a packaged model is generated; the steps include:

[0070] (1) Data preprocessing: The potential and current characteristic values ​​of the CVS curve and oxidation current curve of each single sample are normalized.

[0071] (2) Sample labeling: Label the concentration values ​​of each additive in each single sample.

[0072] (3) Data set partitioning: All sample data in sample database file D are divided into training set and validation set in a ratio of 8:2, with no duplicate sample data between the two datasets.

[0073] (4) Model Building and Training: The sample database was converted into computer language using the Python programming language. Then, the training set was imported into the Python-based machine learning model to train the machine learning model, and the test set was used to test the machine learning model to observe the test accuracy. Specifically, the PyTorch module in Python was used to build the neural network model; the pandas module was used to read the training set data and preprocess it before feeding it into the built neural network structure, and the gradient descent algorithm (Adam) was used to train the model; and the matplotlib module was used to draw images to visualize the fitting of the model training set. The final encapsulated model had a pre-concentration accuracy of less than 5% on the validation set.

[0074] 5. Single-sample component experimental design within the sample database:

[0075] Single samples of a series of standard acid copper plating solutions: 1) All samples contain a base plating solution (VMS): 40 g / L copper ions + 60 g / L sulfuric acid + 50 ppm chloride ions. 2) For S, L, and A concentration measurements, different types and amounts of corresponding additives were gradually added to the base plating solution under the following conditions. Following the sample database experimental acquisition plan, each single sample was tested in the electrochemical working device, and the characteristic current value I of the CV curve for each standard acid copper plating solution single sample at each additive concentration was obtained. d Characteristic oxidation current value I on the oxidation current curve of the ring electrode r The data.

[0076] Table 1. Partial target variables and grouped potential characteristic method voltammetric test results for each individual sample in Sample Set 1.

[0077]

[0078] Table 2. Partial target variables and grouped potential characteristic method voltammetric test results for each individual sample in Sample Set 2.

[0079]

[0080] Table 3. Partial target variables and grouped potential characteristic method voltammetric test results for each individual sample in sample set 3.

[0081] (1) When measuring inhibitor S: in a base plating bath (VMS) without accelerator and leveling agent (40 g / L Cu 2+ +120g / L H2SO4+ 50ppm Cl -In the experiment, an experimental data acquisition scheme was established to vary the accelerator concentration C(S) between 0 and 6 ppm according to a gradient change with a step size of 0.1 ppm, as shown in Sample Set 1 in Table 1. By adjusting the concentration of the active target variable accelerator (S), 60 single samples of standard acid copper plating solutions were set. The CV curve and oxidation current curve corresponding to each single sample were collected sequentially according to step 3 above, as shown in Figure 2(a); then, a machine learning model K1 for measuring the concentration of inhibitor S in acid copper plating solution was established according to step 4 above; Figure 3(a) is the relative error distribution of the inhibitor concentration predicted by the machine learning model K1. When testing unknown plating solutions, a test sample T1 was taken, whose composition included: VMS (40 g / L Cu 2+ + 120g / L H2SO4 + 50ppmCl - First, dilute the unknown plating solution 100 times or more until the concentration of leveling agent and accelerator can be ignored. Then, measure the sample to be tested (actual concentrations of S: 150 ppm, 200 ppm, and 300 ppm) according to the test conditions described above. After performing a voltammetric scan, obtain the current characteristic value corresponding to the potential characteristic range, which is recorded as the sample data file TD1. The characteristic potential value E contained in the CVS curve and oxidation current curve of the sample data file TD1 is... d Characteristic current value I d 'and characteristic oxidation current value I r 'data.

[0082] The machine learning model K1 was called, and the sample data files TD1 of the three test samples T1 were input into the model K1. The predicted concentrations C(S') of additive S in the test samples were 145 ppm, 204 ppm and 294 ppm, respectively. The concentration prediction errors Δε(S) were -3.3%, 2.0% and -2.0%, respectively. The absolute values ​​of the errors were all less than 4%.

[0083] (2) When measuring the leveling agent (L): Under the saturated conditions of the base plating bath (VMS) with an inhibitor content C(S) of 300 pm and an accelerator content C(A) of 20 ppm, an experimental data acquisition scheme was established to change the leveling agent concentration C(L) between 0 and 15 ppm according to the gradient change in step size of 0.5 ppm; as shown in Table 2, sample set 2; by adjusting the target variable of the leveling agent concentration C(L), 31 standard acid copper plating baths were set as single samples. The CV curve and oxidation current curve corresponding to each single sample were collected in sequence according to the above step 3, as shown in Figure 2(b); then, the machine learning model K2 for measuring the leveling agent (L) concentration in the acid copper plating bath samples was established according to the above step 4; Figure 3(b) is the relative error distribution of the leveling agent concentration predicted by the machine learning model K2. When testing an unknown plating bath, a sample T2 was taken, the composition of which included: VMS (40 g / L Cu2+ + 120g / L H2SO4 + 50ppm Cl - First, an inhibitor (S) was added to the unknown plating solution to a saturation state of 300 ppm and an accelerator (A) to 20 ppm. Then, the test samples (actual concentrations of L were 5 ppm, 7 ppm, and 10 ppm, respectively) were measured according to the test conditions described above. After performing a voltammetric scan, the current characteristic values ​​at the corresponding potential characteristic values ​​were obtained and recorded as the test sample data file TD2. The test sample data file TD2 contains the characteristic potential value E contained in the CVS curve and oxidation current curve of the test sample T2. d Characteristic current value I d 'and characteristic oxidation current value I r 'data.

[0084] The machine learning model K2 was called, and the sample data file TD1 of the three sets of test samples T2 was input into the model K2. The predicted concentrations C(L') of additive L in the test samples were 4.94 ppm, 7.11 ppm and 9.89 ppm, respectively. The prediction errors Δε(L) were -1.2%, 1.6% and 1.1%, respectively. The absolute values ​​of the errors were all less than 2%.

[0085] (3) When measuring accelerator A: Under the plating solution with inhibitor C(S) of 300 ppm and leveling agent C(L) of 5 ppm, a single-sample experimental scheme was established to vary the accelerator concentration C(A) between 0 and 10 ppm according to the gradient change in step size of 0.25 ppm. As shown in Table 3, sample set 3 was set up with 41 standard acid copper plating solutions as single samples by adjusting the active target variable of accelerator C(A). The CV curve and oxidation current curve corresponding to each single sample were collected in sequence according to step 3 above, as shown in Figure 2(c); a machine learning model K3 for measuring the concentration of accelerator A in acid copper plating solution was established according to step 4 above; Figure 3(c) is the relative error distribution of accelerator concentration predicted by machine learning model K3.

[0086] When testing an unknown plating solution, a sample T3 was taken, whose composition included: VMS (40 g / L Cu). 2+ + 120g / LH2SO4 + 50ppm Cl -First, the content of leveling agent (L) in the test sample T3 is measured. The test sample T3 is diluted with VMS until the leveling agent content C(L) is 5 ppm, and then 300 ppm of inhibitor (S) is added. The test sample T3 (the actual concentration of A C(A) is 3 ppm, 7 ppm, and 10 ppm respectively) is measured according to the test conditions of the above sample. After performing voltammetric scanning, the current characteristic value of the corresponding potential characteristic range is obtained and recorded as the test sample data file TD3. The test sample data file TD3 contains the characteristic potential value Ed', characteristic current value Id', and characteristic oxidation current value Ir' data contained in the CVS curve and oxidation current curve of the test sample T3.

[0087] The machine learning model K3 was called, and the sample data files TD3 of the three sets of test samples T3 were input into the model K3. The predicted concentrations C(A') of accelerator A in the test samples were 2.89 ppm, 7.05 ppm and 10.12 ppm, respectively. The concentration prediction errors Δε(A) were 3.7%, 0.7% and 1.2%, respectively. The absolute values ​​of the errors were all less than 4%.

[0088] The establishment and training of the machine learning model can refer to the technical solution disclosed in the invention CN202510715830.0 previously submitted by the inventors of this application. All the original data (CV curve and oxidation current curve on the ring electrode) of the disk electrode and ring electrode after the execution of the above technical solutions are input into the machine learning model for model fitting, establishing a machine learning model of the three-dimensional mapping relationship between the characteristic current value Id, the characteristic oxidation current value Ir, and the additive concentration C. 80% of the original data is used to learn and fit the data, and 20% is used to predict the accuracy of the data. After the trained model is input with the CV curve data on the disk electrode and the oxidation current curve data on the ring electrode in any unknown plating solution, it can directly output the predicted concentrations of S, L, and A, as shown in Figures 3(a), 3(b), and 3(c).

[0089] Based on the machine learning models established for S, L, and A as described above, the characteristic current value Id' and characteristic oxidation current value Ir' of the plating solution to be tested are input into the machine learning models to obtain the additive concentration in the plating solution. Specifically, the predicted concentrations of S, L, and A in three unknown plating solutions with different concentrations are tested in parallel. The comparison between the predicted concentrations S'\A'\L' obtained by machine learning and the actual concentrations S\A\L, as well as the relative error Δε, is shown in the table below. It can be seen that the machine learning scheme obtained in this embodiment controls the prediction error Δε to be less than 4%. It is evident that the machine learning model established based on the CV curves and oxidation current curves of the disk electrode and ring electrode, and the boundary conditions, can significantly characterize the global characterization ability of predicting the additive content in acid copper plating solutions.

[0090] category C(S')ppm C(S) ppm No (S) C(A') ppm C(A) ppm No (A) C(L') ppm C(L) ppm No (L) 1 145 150 -3.3 2.89 3 3.7 4.94 5 -1.2 2 204 200 2.0 7.05 7 0.7 7.11 7 1.6 3 294 300 -2.0 10.12 10 1.2 9.89 10 1.1

[0091] Example 2

[0092] This embodiment uses a standard working curve.

[0093] 1. The electrochemical working device includes: ring disk working electrode D-WE: 4mm platinum disk electrode WE1; platinum ring electrode WE2 with an inner diameter of 6.25 and an outer diameter of 7.92, rotating at 3000 rpm; reference electrode RE: Hg / Hg2SO4 (external salt bridge is saturated potassium nitrate); counter electrode: platinum rod.

[0094] 2. Electrode experimental conditions: On the disk electrode: CVS scanning was performed at a scan rate of 0.1V / s; on the ring electrode: a constant +0.8V was applied for scanning using the chronocurrent method.

[0095] 3. Experimental Design

[0096] S1. Preparation of base solution: The base plating solution includes 20~40g / L copper ions, 50~120g / L sulfuric acid, and 30~100g / L chloride ions;

[0097] S2. Establishment of the standard working curve Y: Different concentrations of the additive to be tested were added to the base plating solution in batches to prepare a series of standard acid copper plating solutions; the CV curve of the disk electrode and the oxidation current curve of the ring electrode in each standard acid copper plating solution sample were tested sequentially using an electrochemical working device, and the characteristic potential value E of the CV curve and oxidation current curve was obtained. d、 Characteristic current value I d and characteristic oxidation current value I r Calculate the peak area Q of the oxidation peak on the ring electrode. ri The peak area Q of the dissolution peak on the disk electrode di After all standard acid copper plating solution samples have been tested, the Q groups will be used to... ri / Q di Or Q di / Q ri Establish a standard working curve Y with the concentration of the additive to be tested, S / L / A;

[0098] S3. Detection of Unknown Plating Solution: Using the electrochemical working device and electrochemical testing conditions of step S2, collect data on the peak areas of the ring electrode and disk electrode in any unknown plating solution, and calculate the peak area ratio Q. ri ' / Q di 'or Q di ' / Q ri ', the peak area ratio Q ri ' / Q di 'or Q di ' / Q riSubstituting the values ​​into the standard working curve Y, we obtain the concentration of the additive to be tested in the unknown plating solution: S' / L' / A'.

[0099] 5. Establishment of standard operating curves for inhibitors, leveling agents, and accelerators, and analysis of test results for unknown plating solutions:

[0100] (1) For inhibitor S: According to the gradient change with a step size of 1 ppm, the concentration of inhibitor S was tested and calculated. When the concentration of S changed between 0 and 6 ppm, the peak area Q of the oxidation peak on the ring electrode was measured. ri The peak area Q of the dissolution peak on the electrode di Using each group of Q ri / Q di A standard working curve was established to establish the relationship between (y) and the concentration S(x) of the inhibitor, y = 0.05x + 0.055, R0. 2 =0.99, R is the correlation between y and x in the standard working curve.

[0101] When testing an unknown plating solution, first dilute the unknown plating solution by 100 times or more until the concentration of leveling agent and accelerator can be ignored. Under the same electrochemical test conditions, test the peak area Qri'' of the oxidation peak on the ring electrode and the peak area Qdi'' of the dissolution peak on the disk electrode of the unknown plating solution, and obtain y''=Qri'' / Qdi''. Substitute y=0.05x+0.055, x''=(y''-0.055) / 0.05, which is to obtain the concentration of inhibitor S'' in the unknown plating solution, ConcentrationS''= x''.

[0102] (2) Regarding leveling agent L: Under the conditions of inhibitor S content of 300 pm and accelerator A content of 20 ppm, according to the gradient change of leveling agent L in step size of 1 ppm, the peak area Q of the oxidation peak on the ring electrode was tested and calculated when the concentration of leveling agent L changed between 0 and 10 ppm. ri The peak area Q of the dissolution peak on the electrode di Using each group of Q ri / Q di A standard working curve was established to establish the relationship between the concentration of leveling agent L (x) and the concentration of leveling agent L (y), y = 0.08x + 0.115, R. 2 =0.99, R is the correlation between y and x in the standard working curve.

[0103] When testing the unknown plating solution, an inhibitor was first added to the unknown plating solution to a saturation state of 300 ppm and an accelerator to 20 ppm. Under the same electrochemical testing conditions, the peak area Q of the oxidation peak on the ring electrode of the unknown plating solution was measured. ri The peak area Q of the dissolution peak on the electrode is... di '', so y''=Q ri '' / Q di Substituting y = -0.08x + 0.115, we get x'' = (0.115 - y'') / 0.08, which gives us the concentration of leveling agent L'' in the unknown plating solution: Concentration L'' = x''.

[0104] (3) For accelerator A: At a concentration of 300 ppm for inhibitor and 5 ppm for leveling agent, the peak area Q of the oxidation peak on the ring electrode was tested and calculated when the concentration of accelerator A varied between 0 and 10 ppm L, according to a gradient change of 1 ppm step size. ri The peak area Q of the dissolution peak on the electrode di Using each group of Q di / Q ri A standard working curve was established to establish the relationship between y (x) and the concentration of accelerator A, y = 0.122x + 6.41, R0. 2 =0.99, R is the correlation between y and x in the standard working curve.

[0105] When testing the unknown plating solution, the content of the leveling agent in the unknown plating solution was first measured. Then, the unknown plating solution was diluted with VMS until the leveling agent content was 5 ppm, and then 300 ppm of inhibitor was added. Under the same electrochemical testing conditions, the peak area Q of the oxidation peak on the ring electrode of the unknown plating solution was measured. ri The peak area Q of the dissolution peak on the electrode is... di '', so y''=Q ri '' / Q di Substituting y = 0.122x + 6.41, we get x'' = (y'' - 6.41) / 0.122, which gives us the concentration of the accelerator in the unknown plating solution: ConcentrationA'' = x''.

[0106] Based on the standard working curves established for S, L, and A as described above, the predicted concentrations of S'', L'', and A'' in three parallel tests of unknown plating solutions with different concentrations were obtained. The comparison between the predicted concentrations S'', A'', and L'' obtained from the standard working curves and the actual concentrations S, A, and L, as well as the relative errors Δε(S), Δε(L), and Δε(A), are shown in the table below. It can be seen that the standard working curve scheme obtained in this embodiment controls the prediction error Δε to be less than 5%. It is evident that the standard working curves established based on the CV curves and oxidation current curves of the disk electrode and ring electrode can accurately characterize and predict the additive content in the acid copper plating solution.

[0107] category C(S'')ppm C(S) ppm No (S) C(A'') ppm C(A) ppm No (A) C(L'') ppm C(L) ppm No (L) 1 143 150 -4.7 3.2 3 6.7 2.88 3 -4 2 195 200 -2.5 7.12 7 1.7 5.11 5 2.2 3 312 300 -4.0 10.15 10 1.5 7.17 7 2.4

[0108] This application employs a dual-working-electrode scheme combining a disk electrode and a ring electrode, by obtaining the characteristic oxidation current value I of the oxidation current curve on the ring electrode. r The characteristic current value I of the CV curve on the disk electrode d Characteristic current value I can be established. d Characteristic oxidation current value I r Three-dimensional mapping relationship with additive concentration C; due to monovalent copper ions Cu + There is a highly correlated response relationship between the content of the additive and the concentration of the additive; therefore, the concentration of the additive in the copper plating system can be analyzed and detected through a three-dimensional mapping relationship. Furthermore, combining the analysis method of the ring disk dual working electrode with machine learning, and incorporating information from the ring electrode, enriches the data for machine learning and makes the response to additives more sensitive.

Claims

1. A method for analyzing the concentration of additives in acid copper plating solutions based on dual working electrodes, characterized in that, An electrochemical testing device using a ring electrode and a disk electrode as dual working electrodes was employed to collect CV curves on the disk electrode and oxidation current curves on the ring electrode in a series of standard acid copper plating solutions. The CV curves were obtained by simultaneously measuring the characteristic potential value E of the disk electrode. d The characteristic current value I of the disk electrode d The oxidation current curve is established by measuring the characteristic potential value E of the disk electrode at the same time. d The characteristic oxidation current value I of the ring electrode r Synchronous correspondence is established, with a fixed oxidation potential provided on the ring electrode; based on the data from the CV curve and the oxidation current curve, the characteristic current value I is established. d The characteristic oxidation current value I r A machine learning model or standard working curve relating the three-dimensional mapping relationship between the additive concentration C and the target plating solution; The characteristic current value I of the plating solution to be tested. d 'and characteristic oxidation current value I r 'Input the machine learning model or the standard working curve into the solution to be tested to obtain the concentration of additives in the solution.' 2. The analytical method according to claim 1, characterized in that, Based on the data from the CV curve and the oxidation current curve, the dissolution current I is established. d and the oxidation current I r Machine learning models mapping the relationship between additive concentration C and additive concentration C include: The CV curves and oxidation current curves of the obtained series of standard acid copper plating solutions were preprocessed to extract the characteristic potential value E related to the additive concentration C. d、 The characteristic current value I d and the characteristic oxidation current value I r The various sets of standard acid copper plating solutions each contain different combinations of additive concentrations. The characteristic potential value E d、 The characteristic current value I d and the characteristic oxidation current value I r The corresponding additive concentration C is input into a machine learning model for training. The machine learning model training establishes a relationship between the additive concentration C and the characteristic current value I. d and the characteristic oxidation current value I r The three-dimensional mapping relationship.

3. The analytical method according to claim 1, characterized in that, The characteristic current I is established based on the data from the CV curve and the oxidation current curve. d and the characteristic oxidation current value I r Standard working curves showing the mapping relationship between additive concentration C and the additive concentration C include: The CV curves and oxidation current curves of the multiple sets of the series of standard acid copper plating solutions were preprocessed, and the obtained characteristic potential value E was used as the basis for the data processing. d、 The characteristic current value I d and the characteristic oxidation current value I r Calculate the dissolution peak current area Q of the CV curve related to the additive concentration C. d and the oxidation peak current area Q of the oxidation current curve r The various sets of standard acid copper plating solutions each contain different concentration ratios of additives. Establish the dissolution peak current area Q d With the oxidation peak current area Q r The ratio Q r / Q d Or Q d / Q r The working curve Y varies with the additive concentration C.

4. The analytical method according to claim 1, characterized in that, The scanning potential E applied to the disk electrode d The oxidation potential E is 0.5~0.8V; a fixed oxidation potential E is applied to the ring electrode. r 0.1V≤E r <2.0V (vs. Hg / Hg2SO4).

5. The analytical method according to claim 1, characterized in that, The CV curves on the disk electrode and the oxidation current curves on the ring electrode in the series of standard acid copper plating solutions are specifically included as follows: The types and concentration ranges of additives in the series of standard acid copper plating solutions are determined, and the concentration range of the additives at least covers the dynamic change range and edge concentration values ​​of the additive concentration; each of the series of standard acid copper plating solutions includes a base solution with constant components and concentrations; The series of standard acid copper plating solutions are configured with a uniform distribution of multiple groups of additives in their concentration ratios. Multiple electrochemical testing methods were performed on each of the series of standard acid copper plating solutions, and CV curves on the disk electrode and oxidation current curves on the ring electrode were collected in each of the series of standard acid copper plating solutions.

6. The analytical method according to claim 5, characterized in that, The method of performing multiple electrochemical testing methods on each of the series of standard acid copper plating solutions also includes cleaning and activating the electrodes of the electrochemical testing device during the electrochemical testing of the series of standard acid copper plating solutions.

7. The analytical method according to claim 5, characterized in that, The base solution comprises 2-80 g / L copper ions, 10-120 g / L sulfuric acid, and 10-100 g / L chloride ions; the additives comprise at least one of inhibitors, accelerators, and leveling agents.

8. The analytical method according to claim 1, characterized in that, The electrochemical testing device is a three-electrode system, which also includes any one of the following: a reference electrode, a mercury / mercurous sulfate electrode, a saturated calomel electrode, and a silver / silver chloride electrode; a counter electrode, which is any one of an inert metal counter electrode, such as a platinum, gold, or graphite electrode; a disk electrode, which is any one of a glassy carbon electrode, a platinum electrode, and a gold electrode; and a ring electrode, which is any one of a glassy carbon electrode, a platinum electrode, and a gold electrode.

Images