Intelligent control method and system for semiconductor plating machine

Abstract

This application provides an intelligent control method and system for a semiconductor electroplating machine, relating to the semiconductor manufacturing field. By acquiring conductivity measurement data of the main electroplating tank and combining it with the activity information of key organic components obtained after applying electrochemical excitation to the electroplating solution sample, the system can accurately judge abnormal states. When the conductivity measurement data is consistently low, but the activity of key organic components is still within the normal range, or when the conductivity measurement data is consistently low and the activity of key organic components is also below the normal range, the system can specifically adjust the electroplating process parameters of the wafer edge region. This effectively solves the problems in the prior art where sensor data drift leads to system error compensation, accelerated consumption of electroplating solution additives, and consequently, a decrease in coating uniformity and product yield. It significantly improves coating uniformity and product yield, overcomes the shortcomings of untimely response in the control methods of the prior art, and achieves more stable and efficient semiconductor electroplating production.

Description

Technical Field

[0001] This application relates to the field of semiconductor manufacturing technology, and more specifically, to an intelligent control method and system for a semiconductor electroplating machine. Background Technology

[0002] In the precision world of semiconductor manufacturing, electroplating machines are indispensable equipment, responsible for depositing metal layers on wafers. To ensure both uniform and efficient plating, these machines require an intelligent control system. However, existing control methods often suffer from inconsistent production efficiency and fluctuating product quality due to slow response times or insufficient adaptability to complex situations.

[0003] In long-term continuous production operations, the conductivity sensor probe used to monitor the key chemical activity of the electroplating solution is immersed in a complex electroplating solution containing trace amounts of metal ions and various organic additives for extended periods. Over time, the components in this solution slowly form a thin layer on the probe surface.

[0004] When the intelligent control system receives persistently drifting, low conductivity data, it makes logical judgments based on its built-in coating thickness prediction method and current density setpoint. These prediction methods and setpoints are based on ideal or normal sensor data; therefore, the system may incorrectly assume that the actual activity or concentration of key components in the plating solution is below standard levels. To correct this deviation and maintain the intended coating uniformity target, the system activates its internal compensation mechanism. This typically involves adjusting the feedback regulation parameters of the underlying control loop, resulting in a slight increase in plating current density or an extension of plating time. This, in turn, leads to unpredictable, slight fluctuations in the coating uniformity of the finished wafer, accompanied by intermittent decreases in product yield. Summary of the Invention

[0005] This application discloses an intelligent control method and system for a semiconductor electroplating machine, which aims to solve the problems of insufficient response and adaptability of existing semiconductor electroplating machine control methods when facing physical deviations of sensors and changes in the actual chemical state of electroplating solution, resulting in unstable production efficiency, fluctuations in product quality, and consequently, a decrease in coating uniformity and a reduction in product yield.

[0006] The technical solution of this application is as follows: In a first aspect, this application discloses an intelligent control method for a semiconductor electroplating machine, the method comprising: Obtain conductivity measurement data of the main electroplating tank; A sample of the electroplating solution was extracted from the main electroplating tank; Electrochemical excitation was applied to the electroplating solution sample, and the electrochemical response characteristics of the sample were analyzed to obtain the activity information of key organic components in the electroplating solution. Based on the conductivity measurement data and the activity information, a state determination operation is performed to obtain the state determination result. The state determination operation includes: When the conductivity measurement data shows a continuous downward trend, and the activity information indicates that the activity of the key organic components in the electroplating solution is within the preset normal range, it is determined that there is a physical deviation in the conductivity measurement device of the main electroplating tank; when the conductivity measurement data shows a continuous downward trend, and the activity information indicates that the activity of the key organic components in the electroplating solution is lower than the preset normal range, it is determined that the actual chemical state of the electroplating solution has changed. Based on the state assessment result, when it is determined that the actual chemical state of the electroplating solution has changed, the electroplating process parameters of the wafer edge area are adjusted.

[0007] Through this technical solution, this application can effectively distinguish between the physical deviation of the conductivity measuring device and the change in the actual chemical state of the electroplating solution, avoid incorrect adjustment of process parameters due to misjudgment, and thus improve the control accuracy and product yield of the semiconductor electroplating process.

[0008] Secondly, this application also discloses an intelligent control system for a semiconductor electroplating machine, the system comprising: The conductivity acquisition module is used to acquire conductivity measurement data of the main electroplating tank; the sample extraction module is used to extract electroplating solution samples from the main electroplating tank. The electrochemical analysis module is used to apply electrochemical excitation to the electroplating solution sample and analyze its electrochemical response characteristics to obtain the activity information of key organic components in the electroplating solution. The status judgment module is used to perform a status judgment operation based on the conductivity measurement data and the activity information to obtain a status judgment result. The status judgment operation includes: when the conductivity measurement data shows a continuous low trend and the activity information indicates that the activity of the key organic components of the electroplating solution is within a preset normal range, it is determined that there is a physical deviation in the conductivity measurement device of the main electroplating tank. When the conductivity measurement data shows a continuous downward trend, and the activity information indicates that the activity of the key organic components in the electroplating solution is lower than the preset normal range, it is determined that the true chemical state of the electroplating solution has changed. The process parameter adjustment module is used to adjust the electroplating process parameters of the wafer edge area based on the state judgment result when it is determined that the actual chemical state of the electroplating solution has changed.

[0009] Beneficial effects This application discloses an intelligent control method for a semiconductor electroplating machine. By acquiring conductivity measurement data of the main electroplating tank and combining it with the activity information of key organic components obtained after applying electrochemical excitation to an electroplating solution sample, the method can accurately determine abnormal states during the electroplating process. Specifically, when the conductivity measurement data is consistently low, but the activity of key organic components remains within the normal range, this method can identify a physical deviation in the main conductivity measurement device, thus avoiding misjudgments caused by sensor drift in traditional methods. Conversely, when the conductivity measurement data is consistently low and the activity of key organic components is also below the normal range, this method can accurately determine that the true chemical state of the electroplating solution has changed. Based on this accurate state judgment result, this method can specifically adjust the electroplating process parameters in the wafer edge region, effectively solving the problems in the prior art caused by sensor data drift leading to system error compensation, accelerated consumption of electroplating solution additives, and consequently, decreased coating uniformity and reduced product yield. By distinguishing between the physical deviation of the sensor and the actual chemical state changes of the electroplating solution, this method avoids unnecessary adjustments to process parameters, thereby maintaining the chemical balance of the electroplating solution, significantly improving the uniformity of the coating and the product yield. It overcomes the shortcomings of existing control methods, such as slow response and insufficient adaptability, and achieves more stable and efficient semiconductor electroplating production. Attached Figure Description

[0010] Figure 1 This is a schematic diagram of an intelligent control method for a semiconductor electroplating machine provided in this application.

[0011] Figure 2 This is a schematic diagram of an intelligent control system for a semiconductor electroplating machine provided in this application. Detailed Implementation

[0012] The technical solutions 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.

[0013] Reference Figure 1 The diagram illustrates an embodiment of an intelligent control method for a semiconductor electroplating machine according to the present invention, which may specifically include the following steps: S101, by acquiring conductivity measurement data of the main electroplating tank and extracting electroplating solution samples from the main electroplating tank; S102, apply electrochemical excitation to the electroplating solution sample and analyze the electrochemical response characteristics of the electroplating solution sample to obtain the activity information of key organic components in the electroplating solution; S103, based on conductivity measurement data and activity information, perform a state judgment operation to obtain a state judgment result. The state judgment operation includes: when the conductivity measurement data shows a continuous downward trend and the activity information indicates that the activity of the key organic components of the electroplating solution is within a preset normal range, determine that there is a physical deviation in the conductivity measurement device of the main electroplating tank; when the conductivity measurement data shows a continuous downward trend and the activity information indicates that the activity of the key organic components of the electroplating solution is lower than the preset normal range, determine that the true chemical state of the electroplating solution has changed. S104. Based on the state judgment result, when it is determined that the actual chemical state of the electroplating solution has changed, adjust the electroplating process parameters of the wafer edge area.

[0014] This application can effectively distinguish between the physical deviation of the conductivity measuring device and the change in the actual chemical state of the electroplating solution, thereby avoiding incorrect process adjustments and improving the accuracy of control and the stability of production.

[0015] To facilitate a clearer understanding of the technical solutions in this application, some key terms and implementation environments are first explained. The main electroplating tank is the core container in a semiconductor electroplating machine used for the electroplating reaction, filled with electroplating solution. Conductivity measurement data reflects the overall conductivity of the electroplating solution and is typically related to the concentration and activity of ions in the solution. An electroplating solution sample is a small amount of solution extracted from the main electroplating tank for further analysis. Electrochemical excitation refers to applying a specific potential or current signal to the electroplating solution sample to induce its electrochemical reaction. Electrochemical response characteristics are the changes in current, potential, or impedance of the electroplating solution sample under electrochemical excitation; these characteristics reflect the activity of key organic components in the electroplating solution. Key organic components typically refer to organic additives in the electroplating solution that significantly affect the coating quality and uniformity, such as brighteners and leveling agents. Activity information is obtained based on the analysis of electrochemical response characteristics and is used to assess the effective function of key organic components in the electroplating solution. The preset normal range refers to the acceptable range of activity of key organic components set according to process requirements and experience. Physical deviations refer to measurement errors caused by contamination, aging, or other factors within the conductivity measuring device itself. Actual chemical state changes refer to actual alterations in the concentration and ratio of key components in the electroplating solution. The wafer edge region is one of the most sensitive areas for electroplating reactions on the wafer, and adjustments to its process parameters significantly affect the uniformity of the plating layer.

[0016] The core of the intelligent control method for semiconductor electroplating machines in this application lies in achieving accurate identification of the state of the electroplating solution and fine adjustment of process parameters through multi-dimensional data fusion and intelligent judgment.

[0017] Specifically, this method first requires acquiring conductivity measurement data of the main electroplating tank. This can be achieved by directly immersing a conductivity measuring device in the main electroplating tank to monitor the conductivity of the electroplating solution in real time. For example, a traditional two-electrode or four-electrode conductivity sensor can be used to obtain conductivity data by measuring the resistance or conductivity of the electroplating solution under an AC electric field of a specific frequency. This data is continuously acquired and transmitted to the control system for further processing.

[0018] Simultaneously, it is necessary to extract electroplating solution samples from the main electroplating tank. This can be achieved through an automated sampling system that can extract small amounts of electroplating solution from the main electroplating tank at preset time intervals or based on specific triggering conditions, and transport them to an auxiliary electrochemical measurement device for analysis. For example, a peristaltic pump or syringe pump can be used to extract the electroplating solution sample from the main electroplating tank into a microcell via pipeline.

[0019] Subsequently, an electrochemical excitation is applied to the electroplating solution sample, and the electrochemical response characteristics of the sample are analyzed to obtain the activity information of key organic components in the electroplating solution. This can be accomplished using auxiliary electrochemical measurement devices. For example, techniques such as cyclic voltammetry (CV) or electrochemical impedance spectroscopy (EIS) can be employed. In cyclic voltammetry, a cyclically changing potential is applied to the working electrode, and the corresponding current response curve is recorded. The activity information of key organic components is extracted from characteristic parameters such as peak current and peak potential of the curve. In electrochemical impedance spectroscopy, small AC signals of different frequencies are applied to the working electrode, and its impedance response is measured. The activity of key organic components is analyzed from the shape and characteristic frequencies of the impedance spectrum.

[0020] Based on conductivity measurement data and activity information, a state judgment operation is performed to obtain the state judgment result. This state judgment operation is one of the key steps in this application, as it can distinguish between physical deviations of the conductivity measuring device and changes in the actual chemical state of the electroplating solution. Specifically, when the conductivity measurement data shows a continuous downward trend, and the activity information indicates that the activity of key organic components in the electroplating solution is within the preset normal range, the system will determine that there is a physical deviation in the conductivity measuring device of the main electroplating tank. For example, if the conductivity reading is below the normal range for several hours or days, but electrochemical analysis reveals that the activity of key organic components such as brighteners and leveling agents in the electroplating solution is still within the normal process window, then it can be determined that there is a problem with the conductivity measuring device itself, such as contamination or aging of the probe surface. Conversely, when the conductivity measurement data shows a continuous downward trend, and the activity information indicates that the activity of key organic components in the electroplating solution is below the preset normal range, the system will determine that the actual chemical state of the electroplating solution has changed. For example, if the conductivity reading remains low and electrochemical analysis also shows a significant decrease in the activity of key organic components, it indicates that the actual composition or performance of the electroplating solution has changed and needs to be adjusted accordingly.

[0021] Finally, based on the status assessment results, when it is determined that the actual chemical state of the electroplating solution has changed, the electroplating process parameters in the wafer edge region are adjusted. For example, if the assessment result indicates that the actual chemical state of the electroplating solution has changed, and the activity information indicates that the activity of key organic components is lower than the preset normal range, then the control system will adjust process parameters such as the electroplating current density, electroplating time, or additive replenishment rate in the wafer edge region according to the preset control strategy to compensate for the impact of decreased activity and ensure coating uniformity and quality. If the assessment result indicates that there is a physical deviation in the conductivity measuring device, the system will not adjust the process parameters, but will trigger an alarm or prompt maintenance personnel to check and calibrate the conductivity measuring device.

[0022] The intelligent control method for semiconductor electroplating machines in this application effectively solves the problem of misjudgment that may be caused by a single conductivity data in traditional electroplating machine control methods by introducing the activity information of key organic components in the electroplating solution as an auxiliary judgment basis.

[0023] Specifically, in traditional control systems, when conductivity measurement data shows a persistently low trend, the system often interprets this directly as a decrease in the activity of the electroplating solution and immediately activates a compensation mechanism, such as increasing the electroplating current density or extending the electroplating time. However, as described in the background section, this low trend may not be caused by a change in the actual chemical state of the electroplating solution, but rather by physical deviations caused by deposits on the probe surface of the conductivity measuring device. In this case, the traditional control system will make incorrect compensations, which will instead accelerate the consumption of key organic additives in the electroplating solution, leading to an imbalance in the electroplating solution composition and ultimately affecting the uniformity of the coating and the product yield.

[0024] This application introduces a step of applying electrochemical excitation to an electroplating solution sample and analyzing its electrochemical response characteristics to obtain activity information of key organic components in the electroplating solution, providing another independent and crucial dimension for state assessment. Electrochemical response characteristics can directly reflect the actual activity of key organic components (such as brighteners and leveling agents) in the electroplating solution, and this activity information is less affected by physical deviations in the conductivity measurement device.

[0025] When conductivity measurement data shows a persistently low trend, the method in this application will make a comprehensive judgment by combining the activity information of key organic components: If the activity information indicates that the activity of the key organic components in the electroplating solution is still within the preset normal range, it means that the chemical state of the electroplating solution itself is normal. In this case, the low conductivity trend is likely caused by a physical deviation of the conductivity measuring device. Instead of blindly adjusting the electroplating process parameters, the system can trigger an alarm to prompt the operator to check or calibrate the conductivity measuring device, thereby avoiding incorrect compensation and process imbalance caused by misjudgment.

[0026] If the activity information indicates that the activity of the key organic components in the electroplating solution is lower than the preset normal range, it means that the true chemical state of the electroplating solution has indeed changed. In this case, the lower conductivity trend truly reflects the decrease in the activity of the electroplating solution. The system will adjust the electroplating process parameters in the wafer edge area based on the judgment result, such as increasing the supply of key organic components or adjusting the current density, to restore the normal state of the electroplating solution and ensure the quality of the coating.

[0027] Therefore, the method of this application can accurately distinguish between physical deviations in the conductivity measuring device and changes in the actual chemical state of the electroplating solution, avoiding misjudgments and incorrect compensations caused by insufficient information in traditional methods. This dual judgment mechanism significantly improves the accuracy and robustness of intelligent control of semiconductor electroplating machines, effectively ensuring the uniformity of wafer plating and product yield, thereby improving the overall efficiency and quality of semiconductor manufacturing.

[0028] This application proposes a more comprehensive status assessment mechanism, which improves the accuracy and reliability of status assessment by introducing an assessment of the health status of the measuring device itself.

[0029] In some embodiments of this application, the steps of performing a state determination operation based on conductivity measurement data and activity information to obtain a state determination result include: Obtain health status information of the main conductivity measuring device; Acquire activation status information of the working electrode of the auxiliary electrochemical measurement device; the activation status information includes the passivation index; Receive conductivity measurement data from the main electroplating tank; Receive activity information of key organic components in the electroplating solution; Based on the health status information of the main conductivity measuring device, the activation status information of the working electrode of the auxiliary electrochemical measuring device, the conductivity measurement data, and the activation information, it is determined whether there is a physical deviation in the conductivity measuring device of the main electroplating tank, whether there is a deviation in the measurement results of the auxiliary electrochemical measuring device by obtaining the activation status information of the working electrode, and whether the true chemical state of the electroplating solution has changed, thus obtaining the state judgment result.

[0030] Specifically, obtaining the health status information of the main conductivity measuring device refers to assessing whether the device used to measure the conductivity of the main electroplating tank is in normal working condition through specific detection methods, such as whether there are physical problems like probe contamination, damage, or calibration deviations. The purpose is to ensure the accuracy and reliability of conductivity measurement data. This includes obtaining the activation status information of the working electrode of the auxiliary electrochemical measuring device. This activation status information includes a passivation index, which can be understood as an assessment of the surface activity of the working electrode used to analyze the electrochemical response characteristics of the electroplating solution sample. The passivation index is a quantitative indicator measuring the degree of decline in the surface activity of the working electrode. For example, when the surface of the working electrode is covered by impurities or reaction products in the electroplating solution, its electrochemical response will be inhibited, leading to inaccurate activity measurement. By obtaining the passivation index, it can be determined whether the measurement results of the auxiliary electrochemical measuring device are affected by the passivation of the working electrode. In practical applications, receiving conductivity measurement data from the main electroplating tank and receiving activity information of key organic components in the electroplating solution means that the system continuously obtains real-time conductivity data and activity data of key organic components from the corresponding sensors or analysis modules as the basic input for status judgment. Furthermore, based on the health status information of the main conductivity measuring device, the activation status information of the working electrode of the auxiliary electrochemical measuring device, the conductivity measurement data, and the activity information, the system determines whether there is a physical deviation in the conductivity measuring device of the main electroplating tank, whether there is a deviation in the measurement results of the auxiliary electrochemical measuring device obtained by acquiring the activation status information of the working electrode, and whether the true chemical state of the electroplating solution has changed, thus obtaining a state judgment result. This means comprehensively analyzing the above four types of information. For example, if the conductivity measurement data is consistently low, but the health status information of the main conductivity measuring device shows that it has a physical deviation, the system will prioritize judging it as a problem with the measuring device rather than a change in the state of the electroplating solution. Similarly, if the activity information indicates that the activity of key organic components is low, but the passivation index of the working electrode is high, the system will judge that the measurement results of the auxiliary electrochemical measuring device may be biased, thereby avoiding misjudging the true chemical state of the electroplating solution.

[0031] Through the above technical solution, this application can significantly improve the accuracy and reliability of state judgment in the intelligent control method of semiconductor electroplating machines. By assessing the health status of the measuring device itself, misjudgments caused by sensor failure or electrode passivation can be effectively avoided, thereby reducing unnecessary process parameter adjustments and lowering production costs. Furthermore, this solution enables the system to more accurately identify the true chemical state changes of the electroplating solution, ensuring timely and correct process intervention when necessary, thus guaranteeing wafer electroplating quality and production efficiency, and improving the overall intelligence level of the electroplating process.

[0032] In some preferred embodiments, a specific example is given below. Suppose that during semiconductor electroplating, the system continuously receives conductivity measurement data from the main plating bath, showing a persistently low trend. If the system were to judge solely based on the aforementioned conductivity data and activity information, it might directly determine that the true chemical state of the plating solution has changed and trigger adjustments to the plating process parameters in the wafer edge region. However, in the solution of this application, when performing the status judgment operation, the system first obtains the health status information of the main conductivity measuring device. If this information indicates that the probe of the main conductivity measuring device has severe contamination or damage, the system will also obtain the activation status information of the working electrode of the auxiliary electrochemical measuring device, such as its passivation index showing that the working electrode surface is severely passivated. In this case, even if the activity information also indicates that the activity of key organic components is below the preset normal range, the system will not immediately determine that the true chemical state of the plating solution has changed. Instead, based on this comprehensive information, the system will determine that the low conductivity data and low activity information are likely caused by physical deviations of the measuring device and passivation of the working electrode. Therefore, the system will prioritize triggering maintenance or calibration procedures for the measuring device, rather than immediately adjusting the plating process parameters. Only when the measuring devices are confirmed to be in a healthy and activated state, and the conductivity data and activity information still indicate abnormalities, will the system determine that the true chemical state of the electroplating solution has changed, and then adjust the electroplating process parameters of the wafer edge area accordingly.

[0033] This application further proposes steps for obtaining health status information of the main conductivity measuring device, including: When the main conductivity measuring device is in a non-electroplating period, an AC signal with a preset frequency and amplitude is applied to the main conductivity measuring device; While applying the AC signal, the local potential of the electroplating solution in the main electroplating tank near the probe area of ​​the main conductivity measuring device is monitored by the auxiliary electrode to obtain the potential change information of the local microenvironment of the electroplating solution. Based on the potential change information, the frequency or amplitude of the AC signal is adjusted to suppress the electrochemical reaction induced by the AC signal near the probe surface; The electrical response of the probe of the main conductivity measuring device at the interface with the electroplating solution was measured. The health status information of the main conductivity measuring device is obtained based on the electrical response.

[0034] Specifically, when the main conductivity measuring device is not in the electroplating period, an AC signal with a preset frequency and amplitude is applied to the main conductivity measuring device. The purpose is to diagnose the measuring device without interfering with the normal electroplating process. The non-electroplating period can be understood as a period of time when no actual electroplating operation is performed, such as when the equipment is shut down, under maintenance, or when the wafer is replaced.

[0035] Furthermore, while applying an alternating current signal, the local potential of the electroplating solution near the probe of the main conductivity measuring device in the main electroplating tank is monitored using an auxiliary electrode to obtain information on potential changes in the local microenvironment of the electroplating solution. This step aims to monitor the electrochemical environment around the probe in real time and identify side reactions that may be induced by the alternating current signal. The auxiliary electrode can be a small, highly sensitive reference electrode or working electrode used to accurately capture potential fluctuations near the probe.

[0036] Specifically, based on potential change information, the frequency or amplitude of the AC signal is adjusted to suppress electrochemical reactions induced by the AC signal near the probe surface. This adjustment mechanism is dynamic, using feedback control to ensure that the diagnostic signal does not cause additional contamination or damage to the probe surface, thereby guaranteeing the accuracy of subsequent electrical response measurements. For example, when a drastic potential change is detected, indicating a significant electrochemical reaction, the amplitude of the AC signal can be reduced or the frequency adjusted to a range less likely to induce a reaction.

[0037] In practical applications, measuring the electrical response of the probe of a main conductivity measuring device at the interface with the electroplating solution aims to directly assess the physical and chemical state of the interface. The electrical response can include parameters such as impedance, capacitance, or resistance, which reflect the probe's surface cleanliness, coating integrity, or the presence of a passivation layer, among other health conditions. Thus, information about the health status of the main conductivity measuring device is obtained from the electrical response. This information can be a comprehensive health index or a specific indication of whether the probe is contaminated, damaged, or has calibration errors.

[0038] The above technical solution significantly improves the accuracy and reliability of acquiring health status information from the main conductivity measuring device. Compared to simply acquiring health status information, this solution actively suppresses electrochemical reactions, avoiding potential damage or misjudgment to the probe caused by the diagnostic process itself, thus ensuring more accurate judgment of physical deviations in the conductivity measuring device. This is crucial for the effectiveness of the entire intelligent control method for semiconductor electroplating machines, preventing erroneous process adjustments due to sensor false alarms, and improving production stability and product yield.

[0039] In some preferred embodiments, a specific example is given below. Assume the probe of the main conductivity measuring device may have surface contamination or slight corrosion after prolonged use. During non-electroplating periods, the system applies an AC signal with an initial frequency of 1 kHz and an amplitude of 10 mV to the probe. Simultaneously, the auxiliary electrode detects periodic fluctuations in the potential near the probe with relatively large amplitude fluctuations, which may indicate a slight electrochemical reaction. Based on this potential change information, the system automatically adjusts the amplitude of the AC signal to 5 mV and attempts to adjust the frequency to 500 Hz to reduce the activity of the electrochemical reaction. After the signal stabilizes, the impedance spectrum of the probe-electroplating solution interface is measured. If the impedance spectrum shows a significant capacitive arc in the high-frequency region, and its characteristic parameters deviate significantly from the reference data for a clean probe, it can be determined that the probe of the main conductivity measuring device has surface contamination or passivation, thus obtaining information about its health status. For example, a significantly increased impedance value may indicate the formation of an insulating layer on the probe surface; an abnormal phase angle may indicate the adsorption of electrochemically active substances on the surface. This information will be used for subsequent status assessment to distinguish whether the low conductivity is due to probe malfunction or changes in the actual state of the electroplating solution.

[0040] In some embodiments described above in this application, information on the activation status of the working electrode of an auxiliary electrochemical measurement device is proposed. Specifically, the steps for obtaining information on the activation status of the working electrode of the auxiliary electrochemical measurement device may include the following:

[0041] Before applying a potential step signal to the working electrode, the auxiliary electrochemical measurement device is controlled to inject cleaning solution into the microcell to rinse the working electrode, thereby removing electrochemically active substances remaining on the electrode surface and in the electrolyte. After the rinsing operation is completed, the cleaning solution is drained, and a small amount of fresh electroplating solution sample is injected into the microcell by the auxiliary electrochemical measurement device. A standard potential step signal is applied to the working electrode, and the current decay curve generated on the working electrode is recorded; the current decay curve is the curve of the current generated on the working electrode decaying over time. The characteristic parameters of the current decay curve are extracted, and the difference between the currently measured decay curve characteristic parameters and the reference curve characteristic parameters in the clean and activated state of the working electrode is compared to obtain the activation status information of the working electrode of the auxiliary electrochemical measurement device.

[0042] The working electrode is the core component of the auxiliary electrochemical measurement device, and its surface condition directly affects the accuracy of the electrochemical measurement. Before electrochemical measurement, the surface of the working electrode may be adsorbed with organic or inorganic impurities from the electroplating solution, or residual electrochemically active substances from previous measurements. These substances can interfere with subsequent electrochemical responses. Therefore, the working electrode needs to be rinsed by injecting a cleaning solution. The cleaning solution can be selected according to the composition of the electroplating solution and the types of substances that may remain. For example, it can be deionized water, dilute acid, dilute alkali, or organic solvents to effectively dissolve or remove residues on the electrode surface and in the internal flow channels of the microcell. After ensuring the working electrode surface is clean, the cleaning solution needs to be completely drained to avoid interference with subsequent electrochemical measurements. Subsequently, a small amount of fresh electroplating solution sample is injected. This sample should be representative to simulate the actual electroplating environment and provide a stable electrolyte environment for the subsequent application of the potential step signal. The standard potential step signal refers to rapidly changing the potential of the working electrode from one stable value to another preset value within a short period of time. Under this potential step, a transient electrochemical reaction occurs on the surface of the working electrode, generating a current response that decays over time. The shape and decay rate of this current decay curve are closely related to the surface state of the working electrode, the electrode material, and the properties of the electrolyte. By recording this current decay curve, the dynamic response characteristics of the working electrode under specific electrochemical conditions can be obtained. The characteristic parameters may include, but are not limited to, peak current, decay time constant, steady-state current value, and charge. These parameters can quantify the characteristics of the current decay curve. The characteristic parameters of the reference curve under the clean and activated state of the working electrode are pre-measured and stored when the working electrode is in an ideal clean and activated state. By comparing the currently measured characteristic parameters with the reference parameters, the degree of activation of the working electrode can be evaluated. For example, if the decay rate of the current curve is significantly slower than that of the reference curve, or the peak current is significantly reduced, it may indicate that passivation or contamination exists on the surface of the working electrode, thereby affecting its electrochemical activity. This difference analysis can directly reflect the activation status of the working electrode, such as whether activation treatment or replacement is required.

[0043] This application's scheme ensures the optimal condition of the working electrode of the auxiliary electrochemical measurement device before electrochemical measurements through a series of meticulous operations. First, rinsing the working electrode with a cleaning solution effectively removes any residual substances that may exist on the electrode surface and within the microcell, thus preventing interference from these residues with the subsequent electrochemical response and ensuring the purity of the measurement environment. Second, after rinsing, the cleaning solution is drained and a fresh electroplating solution sample is injected, providing the working electrode with a standard and stable electrolyte environment, allowing subsequent electrochemical measurements to be performed under conditions close to actual operating conditions. Subsequently, by applying a standard potential step signal and recording the current decay curve, the dynamic response of the working electrode under transient electrochemical excitation can be captured, containing rich information about the working electrode surface state. Finally, by extracting the characteristic parameters of the current decay curve and comparing them with a reference curve, the activation status of the working electrode can be quantified, thereby accurately determining whether it is in its optimal working state. This series of steps works together to provide a reliable measurement basis for subsequent electroplating solution state determination based on activity information.

[0044] The above technical solution enables accurate assessment of the activation status of the working electrode in the auxiliary electrochemical measurement device. Thorough cleaning and rinsing before measurement effectively eliminates interference from electrode surface contamination and residues, significantly improving the accuracy and reliability of electrochemical measurement data. Furthermore, by applying a standard potential step signal and analyzing the current decay curve, the true electrochemical activity of the working electrode can be objectively and quantitatively reflected, avoiding measurement deviations caused by poor electrode condition. This ensures that the acquired activity information of key organic components in the electroplating solution is accurate and effective, thereby improving the accuracy of the state judgment in the entire intelligent control method of the semiconductor electroplating machine and providing a more reliable basis for subsequent process parameter adjustments.

[0045] In some embodiments described above, it is proposed to rinse the working electrode with a cleaning solution before applying a potential step signal to the working electrode, in order to remove residual electrochemically active substances on the electrode surface and in the electrolyte. However, in practice, if the cleaning process is not thorough enough, or if the cleaning solution is not properly selected, residues may remain on the electrode surface, affecting the accuracy of subsequent electrochemical measurements and leading to deviations in the judgment of the activity information of key organic components in the electroplating solution. Therefore, this application further proposes a scheme to optimize the working electrode cleaning process, aiming to ensure the clean and activated state of the working electrode through a more refined cleaning strategy, thereby improving measurement accuracy.

[0046] The steps described above, including controlling the auxiliary electrochemical measurement device to inject cleaning solution into the microcell before applying a potential step signal to the working electrode to rinse the working electrode and remove residual electrochemically active substances from the electrode surface and electrolyte, include: Based on the type of key organic components in the electroplating solution and the electrochemical characteristics of their decomposition products, determine the cleaning solution components with targeted dissolving or reacting capabilities. After the cleaning solution is injected, the working electrode is rinsed by applying a low-frequency alternating electric field or ultrasonic vibration to enhance the cleaning effect of the cleaning solution on the surface of the working electrode and the internal flow channels of the micro-pool. After rinsing is completed, the thoroughness of the cleaning process is determined by monitoring changes in the conductivity of the cleaning fluid or the spectral absorption characteristics of specific residues. If the cleaning process fails to meet the preset thoroughness standard, adjust the injection volume of cleaning fluid, rinsing time, and / or the intensity of low-frequency AC electric field or ultrasonic vibration, and repeat the rinsing process to remove electrochemically active substances remaining on the electrode surface and in the electrolyte.

[0047] Specifically, determining the cleaning solution components with targeted dissolving or reactivity capabilities refers to selecting or formulating a cleaning solution that can effectively dissolve, complex, or deactivate the key organic components (e.g., accelerators, inhibitors, leveling agents, etc.) that may be present in the electroplating solution and their decomposition products (e.g., oxidation products, polymerization products, etc.) that may be generated during the electroplating process, based on their chemical properties and electrochemical behavior. For example, organic solvents can be used for certain organic residues; acidic or alkaline solutions can be used for metal ion or inorganic salt residues. The aim is to ensure that the cleaning solution can efficiently remove the target residues.

[0048] The process of rinsing the working electrode by applying a low-frequency alternating electric field or ultrasonic vibration can be understood as using external physical aids to enhance the cleaning effect. The low-frequency alternating electric field can induce microscopic fluid disturbances and electrochemical reactions on the electrode surface, which helps to remove adhering substances; ultrasonic vibration generates microjets through cavitation effects, mechanically removing stubborn residues from the electrode surface and the internal channels of the micro-pool. The aim is to overcome the limitations of simple rinsing and improve cleaning efficiency and thoroughness.

[0049] In practical applications, assessing the thoroughness of the cleaning process by monitoring changes in the conductivity of the cleaning solution or the spectral absorption characteristics of specific residues involves real-time or near-real-time analysis of the cleaning solution within the micro-tank after cleaning is completed. For example, if residual electrochemically active substances in the cleaning solution are removed, its conductivity may stabilize or reach a preset low value; simultaneously, specific organic residues exhibit absorption peaks at specific wavelengths, and spectral analysis can detect whether their concentration has decreased to an acceptable level. The aim is to provide objective criteria for evaluating the cleaning effect, avoiding indiscriminate or insufficient cleaning.

[0050] Furthermore, when the cleaning process fails to meet the preset thoroughness standard, adjusting the injection volume of the cleaning solution, the rinsing time, and / or the intensity of the low-frequency AC electric field or ultrasonic vibration, and repeating the rinsing process, can be understood as an adaptive optimization cleaning strategy. This means that the cleaning process is not static but dynamically adjusted based on real-time monitoring results until the preset cleanliness standard is achieved. The aim is to ensure that the working electrode is always in an optimal activation state, providing a reliable basis for subsequent electrochemical measurements.

[0051] The above technical solutions significantly improve the cleaning efficiency and thoroughness of the working electrode in auxiliary electrochemical measurement devices. The targeted selection of cleaning solutions and the introduction of physical auxiliary rinsing methods enable more effective removal of electrochemically active residues from the electrode surface and the internal flow channels of the microcell, avoiding measurement errors caused by incomplete cleaning. Furthermore, by real-time or near-real-time assessment of the cleaning process's thoroughness, combined with adaptive parameter adjustment and a repetitive cleaning mechanism, the working electrode is ensured to always be in an optimal clean and activated state. This greatly enhances the accuracy and reliability of measuring the activity information of key organic components in the electroplating solution, providing more precise data support for the intelligent control of semiconductor electroplating machines.

[0052] In some preferred embodiments, a specific example is given below. Assume the plating solution in the main electroplating tank contains copper ions, sulfuric acid, and accelerators (such as SPS) and inhibitors (such as PEG) as key organic components. When it is necessary to clean the working electrode of the auxiliary electrochemical measurement device: First, the composition of the cleaning solution is determined based on the chemical properties of SPS, PEG, and their decomposition products. For example, a mixed solution containing a weak acid and a surfactant can be used to dissolve organic residues and assist in stripping the adsorbed layer from the electrode surface.

[0053] Next, after injecting the cleaning solution, ultrasonic vibrations of 20-40 kHz are applied to the working electrode for 30 seconds using an ultrasonic generator placed outside or inside the microcell. The cavitation effect of the ultrasound effectively removes tiny particles and stubborn organic films from the electrode surface and the microcell channels.

[0054] Subsequently, after the ultrasonic rinsing operation is completed, the system automatically extracts the residual cleaning fluid in the micro-tank for analysis. For example, the absorption intensity of the cleaning fluid at 230 nm (characteristic absorption wavelength of SPS) and 280 nm (characteristic absorption wavelength of PEG) is monitored using a built-in spectrometer, while the conductivity of the cleaning fluid is measured.

[0055] If the spectral absorption intensity or conductivity is still higher than the preset threshold (e.g., absorption intensity is 5% higher than the background value, conductivity is 10% higher than that of pure cleaning fluid), then the cleaning is deemed not to have met the thoroughness standard. At this time, the system will automatically adjust the cleaning parameters, for example, increase the injection volume of cleaning fluid by 10%, extend the ultrasonic rinsing time to 45 seconds, and repeat the above rinsing and monitoring process until all indicators meet the preset standards, ensuring that there are no residues on the surface of the working electrode and that it is in a highly activated state.

[0056] This application further proposes a more accurate and comprehensive method for judging the thoroughness of the cleaning process. By integrating multi-wavelength spectral data and conductivity measurement data, more representative comprehensive residue characteristic information is obtained, thereby improving the accuracy of the judgment.

[0057] The steps for determining the thoroughness of the cleaning process mentioned above include: After the rinsing operation is completed, the residual cleaning solution in the micro tank is scanned by multi-wavelength spectrum to obtain absorption spectrum data at different wavelengths, thus obtaining multi-wavelength spectrum data. The conductivity of the residual cleaning fluid in the micro tank was measured to obtain conductivity measurement data; The multi-wavelength spectral data and the conductivity measurement data are fused together to obtain the comprehensive residue characteristics of the cleaning fluid; The thoroughness of the cleaning process is determined based on the comprehensive residue characteristics information.

[0058] Specifically, after rinsing, to comprehensively assess the types and concentrations of residual substances in the cleaning solution, a multi-wavelength spectral scan is first performed on the residual cleaning solution in the micro-tank. This scan aims to acquire absorption spectral data of the cleaning solution at different wavelengths. These data reflect the absorption characteristics of the residues to specific wavelengths of light, thus providing preliminary information about the chemical composition and concentration of the residues. Simultaneously, the conductivity of the residual cleaning solution in the micro-tank is measured to obtain conductivity data. Conductivity is an important indicator for measuring ion concentration in a solution, reflecting the total amount of soluble ion residues in the cleaning solution. Furthermore, to overcome the limitations of a single measurement method, the obtained multi-wavelength spectral data and conductivity measurement data are fused. This fusion process aims to comprehensively utilize the advantages of two different types of data, linking them through an algorithm to obtain comprehensive residue characteristic information of the cleaning solution. This comprehensive characteristic information more comprehensively and accurately characterizes the types, concentrations, and distribution of residues in the cleaning solution than a single data source. Finally, based on the obtained comprehensive residue characteristic information, the thoroughness of the cleaning process is assessed. For example, a threshold or model can be set so that when the comprehensive residual characteristics information is lower than a certain preset standard, the cleaning process is considered to be thorough.

[0059] Through the above technical solution, this application can significantly improve the accuracy and reliability of judging the thoroughness of the cleaning process. Compared with a single judgment method that relies solely on changes in conductivity or spectral absorption characteristics, the integration of multi-wavelength spectral data and conductivity measurement data enables a more comprehensive perception of the types and concentrations of residues in the cleaning solution. This not only allows for more accurate identification of trace residues, including organic or inorganic substances that are difficult to detect by a single method, but also effectively avoids misjudgments caused by the complexity of the electroplating solution composition or the diversity of decomposition products. Therefore, it ensures that the cleanliness of the working electrode surface and the internal flow channels of the micro-cell reaches its optimal state, providing a solid foundation for the accuracy of subsequent auxiliary electrochemical measurement devices, thereby improving the reliability and effectiveness of the entire intelligent control method for the semiconductor electroplating machine.

[0060] In some preferred embodiments, specifically, after rinsing the working electrode, a miniature spectrometer integrating a multi-wavelength light source and a spectral detector is first used to scan the residual cleaning solution in the miniature pool, acquiring absorption spectral data from the ultraviolet to near-infrared bands. Simultaneously, a high-precision conductivity sensor is used to measure the conductivity of the cleaning solution. Subsequently, these spectral and conductivity data are input into a pre-trained data fusion model. This model can be a machine learning-based algorithm, such as a support vector machine or neural network, which learns the joint spectral-conductivity features corresponding to different types and concentrations of residues during the training phase. The model outputs a comprehensive residue index, which quantifies the total amount and type of residues in the cleaning solution. For example, when the comprehensive residue index is higher than a preset threshold, the system determines that cleaning is incomplete and may trigger a re-rinsing instruction; conversely, if the index is lower than the threshold, cleaning is considered complete, and the next step can proceed. In this way, even trace amounts of residues that are difficult to detect by a single method can be effectively identified, thereby ensuring that the working electrode is in optimal activation condition.

[0061] This application further proposes the following steps for fusing multi-wavelength spectral data with conductivity measurement data to obtain comprehensive residual characteristic information of the cleaning fluid: Based on the electrochemical characteristics of key organic components and decomposition products in the electroplating solution, as well as the spectral absorption and conductivity response of key organic components and decomposition products at different temperatures and pH values, the multi-wavelength spectral data and conductivity measurement data are preliminarily calibrated and preprocessed to obtain calibrated multi-wavelength spectral data and conductivity measurement data. The correlation and mutual influence between the calibrated multi-wavelength spectral data and conductivity measurement data are analyzed, and the fusion weight or fusion algorithm parameters are dynamically adjusted to determine whether there is a nonlinear or time-varying relationship between the multi-wavelength spectral data and conductivity measurement data. When a nonlinear or time-varying relationship is detected between the multi-wavelength spectral data and the conductivity measurement data, a nonlinear mapping or time-series analysis mechanism is activated to extract a comprehensive feature vector that can reflect the types and concentrations of residues in the cleaning solution. Based on the comprehensive feature vector, the comprehensive residue characteristic information of the cleaning fluid is obtained.

[0062] Specifically, "preliminary calibration and preprocessing" refers to a series of correction and optimization operations performed on the raw multi-wavelength spectral data and conductivity measurement data before data fusion. This may include background noise removal, baseline correction, drift compensation, data normalization, and outlier detection. The aim is to eliminate errors and interference that may be introduced during the measurement process, ensuring the accuracy and consistency of the data. For example, the measured data can be corrected for temperature and pH based on the standard spectral absorption curves and conductivity response curves of known key organic components and their decomposition products at different temperatures and pH values ​​to eliminate the influence of environmental factors on the measurement results. Furthermore, electrochemical properties, such as the redox potential or charge transfer characteristics of specific residues, can be used to guide the calibration of spectral and conductivity data, ensuring that the data accurately reflects the chemical state of the residues.

[0063] "Correlation and mutual influence" refers to the correlation and interaction between multi-wavelength spectral data and conductivity measurement data in reflecting information about residues in the cleaning fluid. For example, some residues may significantly affect both spectral absorption and conductivity simultaneously, while others may only have a major impact on one of them. Specifically, the correlation between data can be quantified by calculating statistical indicators such as Pearson correlation coefficient and mutual information. Simultaneously, machine learning algorithms, such as support vector machines or decision trees, can be used to analyze the mutual influence patterns between data. When complex nonlinear relationships are detected between the data, such as a non-simple linear relationship between residue concentration and conductivity, or a time-varying relationship (e.g., residues may undergo chemical transformation during cleaning, causing their spectral and conductivity responses to change over time), the fusion strategy needs to be dynamically adjusted. For example, the weight of a certain type of data can be increased or decreased based on real-time analysis results, or a more complex fusion algorithm can be switched to better capture these complex relationships.

[0064] Furthermore, "nonlinear mapping" refers to transforming raw data into a new feature space through nonlinear functions to reveal hidden nonlinear patterns within the data. For example, this can be achieved using multinomial regression, kernel function methods (such as the radial basis function in support vector regression), or artificial neural networks (such as multilayer perceptrons or recurrent neural networks). The aim is to transform nonlinear relationships into linearly separable or more easily processed forms. "Time-series analysis mechanisms," on the other hand, focus on processing the characteristics of data changing over time. For example, methods such as Kalman filtering, hidden Markov models, or long short-term memory networks can be used to capture the dynamic patterns of residue concentration or composition changes over cleaning time. Through these mechanisms, a more representative and discriminative "comprehensive feature vector" can be extracted from calibrated multi-wavelength spectral data and conductivity measurement data. This vector can more comprehensively and accurately characterize the types and concentrations of residues in the cleaning fluid.

[0065] The "Comprehensive Residue Characteristic Information" is a comprehensive description of the state of residues in the cleaning solution, derived from the extracted comprehensive feature vectors and further analyzed and interpreted. This includes qualitative identification of residues, such as the presence of a specific organic component or its decomposition products; quantitative analysis, such as the concentration level of residues; and assessment of residue distribution or activity. Its purpose is to provide an accurate and reliable basis for determining the thoroughness of the cleaning process.

[0066] Through the above technical solution, this application can significantly improve the accuracy and robustness of residue detection in semiconductor electroplating machine cleaning solutions. Compared with methods that only perform simple data fusion, this solution effectively eliminates environmental interference and measurement errors through refined data calibration preprocessing. More importantly, by dynamically evaluating the complex correlation between multi-wavelength spectral data and conductivity measurement data, and intelligently adjusting the fusion strategy according to their nonlinear or time-varying characteristics, and even introducing advanced nonlinear mapping and time-series analysis techniques, the system can gain a deeper understanding of the chemical behavior and dynamic changes of residues. This not only ensures accurate extraction of comprehensive feature vectors reflecting the type and concentration of complex and variable residue components, but also greatly improves the reliability of judging the thoroughness of the cleaning process, thereby helping to optimize the electroplating process, reduce wafer defects, and improve product yield.

[0067] In some preferred embodiments, a specific example is given below. Suppose that during the cleaning process, trace amounts of organic additives and their decomposition products remain in the electroplating solution. These substances may exhibit nonlinear relationships in their conductivity and spectral absorption response at specific wavelengths at different concentrations, and their decomposition process is time-dependent.

[0068] First, after acquiring multi-wavelength spectral data and conductivity measurement data, the raw data is preliminarily calibrated and preprocessed according to the pre-established standard response curves of organic additives and their decomposition products at different temperatures and pH values. For example, temperature compensation and baseline correction are performed to eliminate the influence of environmental factors and instrument drift, so as to obtain calibrated data.

[0069] Next, the system analyzes this calibrated data. For example, it assesses the correlation between spectral and conductivity data by calculating nonlinear correlation coefficients, such as mutual information, or by training a small neural network. If a significant nonlinear relationship is found, such as conductivity changing little at low concentrations but drastically at high concentrations, or if time-varying characteristics are present, such as a certain decomposition product having a high concentration in the early stages of washing and gradually decreasing thereafter, the system will dynamically adjust the fusion strategy.

[0070] Specifically, when a nonlinear relationship is detected, a nonlinear mapping model based on multinomial regression or radial basis function kernels can be used to map the spectral and conductivity data to a high-dimensional feature space, in which the residue concentration and the feature vector may exhibit a better linear relationship. When a time-varying relationship is detected, Kalman filtering or long short-term memory networks can be used to analyze the data sequence to capture the dynamic trend of residue concentration changing over time.

[0071] These mechanisms allow for the extraction of a comprehensive feature vector. This vector not only incorporates independent information about the spectrum and conductivity but also integrates their nonlinear or time-varying correlations. For example, one dimension of the vector might represent the concentration of a specific organic additive, while another dimension might represent the cumulative amount of its decomposition products. Ultimately, based on this comprehensive feature vector, the system can accurately determine the type and concentration of residues in the cleaning solution, thereby concluding whether the cleaning process is thorough. For instance, if the comprehensive feature vector indicates that the concentration of a key residue is below a preset threshold, the cleaning is considered complete.

[0072] This application further proposes the following steps for determining whether there is a nonlinear or time-varying relationship between the multi-wavelength spectral data and conductivity measurement data after the above-mentioned analysis and calibration, dynamic adjustment of fusion weights or fusion algorithm parameters, and determination of the correlation and mutual influence between the multi-wavelength spectral data and conductivity measurement data: Under multiple preset conductivity thresholds, the calibrated multi-wavelength spectral data are segmented for analysis to identify the spectral characteristics of residues with different conductivity responses. Under multiple preset spectral absorption characteristic peaks or specific wavelength combinations, the calibrated conductivity measurement data are segmented for analysis to identify the conductivity response corresponding to residues with different spectral absorption characteristics. Based on the spectral characteristics corresponding to the residues with different conductivity responses and the conductivity responses corresponding to the residues with different spectral absorption characteristics, the correlation and mutual influence contribution of each residue to the calibrated multi-wavelength spectral data and conductivity measurement data are evaluated. Based on the contribution of the evaluation, the fusion weights or fusion algorithm parameters are dynamically adjusted to determine whether there is a nonlinear or time-varying relationship between the multi-wavelength spectral data and the conductivity measurement data.

[0073] Specifically, "segmented analysis of calibrated multi-wavelength spectral data under multiple preset conductivity thresholds" refers to dividing the conductivity range of the cleaning solution into several intervals, such as low conductivity, medium conductivity, and high conductivity intervals. Within each conductivity interval, the corresponding multi-wavelength spectral data is analyzed independently to identify the specific spectral characteristics exhibited by possible residues under that conductivity condition. The purpose is that different residues at different concentrations may have different effects on conductivity, and segmented analysis can capture these differences more precisely.

[0074] "Segmented analysis of calibrated conductivity measurement data under multiple preset spectral absorption characteristic peaks or specific wavelength combinations" refers to classifying spectral data based on typical spectral absorption peaks or representative wavelength combinations of known residues. For each spectral characteristic, the corresponding conductivity measurement data is analyzed to identify the specific conductivity response of the residue under that spectral characteristic. The purpose is that some residues may have strong absorption at specific wavelengths but have little impact on conductivity, and vice versa. This cross-segmented analysis allows for a more comprehensive understanding of the residue's characteristics.

[0075] "Assessing the correlation and mutual contribution of each residue to the calibrated multi-wavelength spectral data and conductivity measurement data" means, after completing the above segmented analysis, quantifying the influence of each potential residue (or residue category) on the correlation between the overall multi-wavelength spectral data and conductivity measurement data, using a known residue database or a pre-established model. For example, the contribution weight of each residue can be calculated using statistical methods (such as correlation coefficients, regression analysis) or machine learning models.

[0076] "Dynamically adjusting fusion weights or fusion algorithm parameters" refers to adjusting the algorithm parameters used to fuse multi-wavelength spectral data and conductivity measurement data in real time or periodically, based on the assessed contribution of each residue. For example, if a certain residue has a significant impact on conductivity but a relatively small impact on spectral absorption, the weight of the spectral data in the detection of that residue can be appropriately reduced during fusion, or the fusion algorithm can be adjusted to focus more on conductivity data.

[0077] Through the above technical solution, this application can significantly improve the accuracy and robustness of the fusion processing of residual characteristic information in cleaning solutions. Compared with methods that only perform overall correlation analysis, this application, by performing segmented analysis on multi-wavelength spectral data and conductivity measurement data and evaluating the contribution of each residue, can more precisely capture the characteristics of complex residues in the cleaning solution and their nonlinear or time-varying effects on the measurement results. Therefore, the adjustment of fusion weights or fusion algorithm parameters will be more accurate, making the judgment of the thoroughness of the cleaning process more reliable, effectively avoiding misjudgments or omissions caused by the complexity of residue characteristics, and further ensuring the quality and stability of the semiconductor electroplating process.

[0078] In some preferred embodiments, it is assumed that the cleaning solution may contain residual organic additive A and its decomposition product B. Organic additive A has a strong absorption peak at a specific wavelength, but has little effect on conductivity; decomposition product B has a significant effect on conductivity, but its spectral absorption characteristics are not obvious.

[0079] First, the system performs segmented analysis on the calibrated multi-wavelength spectral data at several preset conductivity thresholds, such as 0.1 mS / cm, 0.5 mS / cm, and 1.0 mS / cm. When the conductivity is at a low threshold (e.g., 0.1 mS / cm), if a characteristic spectral peak of organic additive A is detected, it indicates that a small amount of additive A may be present.

[0080] Secondly, the system performs segmented analysis on the calibrated conductivity measurement data under multiple preset spectral absorption characteristic peaks (e.g., the characteristic peak wavelength of additive A) or specific wavelength combinations. When the characteristic spectral peak of additive A is detected, if the corresponding conductivity measurement data does not change significantly, it further confirms the characteristic that additive A has little effect on conductivity. Conversely, if the conductivity measurement data increases significantly, it may indicate the presence of decomposition product B.

[0081] Next, based on these segmented analysis results, the system evaluates the contribution of organic additive A and decomposition product B to the correlation between multi-wavelength spectral data and conductivity measurement data. For example, through the established database, the system knows that additive A mainly contributes to the spectral signal, while decomposition product B mainly contributes to the conductivity signal.

[0082] Finally, based on the contribution of this assessment, the system dynamically adjusts the fusion weights. For example, when judging the thoroughness of cleaning, if the primary focus is on the removal of decomposition product B, the system increases the weight of conductivity data in the fusion algorithm; if the primary focus is on the removal of additive A, it increases the weight of spectral data in the fusion algorithm. In this way, the system can more accurately determine the comprehensive residual characteristics of the cleaning solution, thereby more reliably judging the thoroughness of the cleaning process.

[0083] Specifically, the steps described above, which involve segmenting and analyzing calibrated multi-wavelength spectral data under multiple preset conductivity thresholds to identify the spectral characteristics corresponding to residues with different conductivity responses, may include the following operations: The step of segmenting and analyzing the calibrated multi-wavelength spectral data under multiple preset conductivity thresholds to identify the spectral characteristics corresponding to residues with different conductivity responses includes: Establish a database linking electrical conductivity with the spectral characteristics of various known residues; The currently measured multi-wavelength spectral data is matched with the spectral features in the associated database to obtain the matching result; The matching results are initially screened by combining the preset conductivity threshold to identify the spectral characteristics of residues with different conductivity responses.

[0084] Establishing a database linking conductivity with the spectral characteristics of various known residues involves pre-collecting and storing conductivity data and corresponding multi-wavelength spectral characteristics of various residues that may exist during the electroplating solution cleaning process (e.g., decomposition products of organic additives, metal ion complexes, etc.). This database can include data under different concentrations, temperatures, and pH values ​​to improve matching accuracy. For example, a set of characteristic parameters such as typical spectral absorption peaks, peak intensities, and full width at half maximum (FWHM) of specific residues within different conductivity ranges can be constructed through experimental measurements or theoretical calculations.

[0085] Matching the currently measured multi-wavelength spectral data with the spectral features in the associated database to obtain the matching result involves comparing the real-time or recently acquired multi-wavelength spectral data of the cleaning fluid with the spectral features of known residues stored in the associated database. The matching algorithm can employ various techniques, such as correlation coefficient-based matching, principal component analysis, or machine learning methods like support vector machines, to assess the similarity between the current spectral data and the spectral features of each known residue in the database, thereby initially identifying the types of residues that may be present in the cleaning fluid and their approximate concentrations.

[0086] Preliminary screening of matching results using the preset conductivity threshold to identify the spectral characteristics of residues with different conductivity responses refers to further verifying and refining the matching results based on the spectral matching, using the preset conductivity threshold. For example, if the spectral matching result indicates the presence of a certain residue, but its corresponding conductivity measurement data (obtained from the comprehensive residue characteristic information of the aforementioned cleaning solution) does not reach the preset conductivity threshold range for that residue in the database, then the matching result may be considered low confidence or excluded. Conversely, if both spectral matching and conductivity threshold support the presence of a certain residue, then the residue and its corresponding spectral characteristics can be identified more accurately. The preset conductivity threshold can be set according to the electrochemical properties of different residues and their influence on conductivity. For example, for ionic residues, the conductivity threshold may be higher; for non-ionic residues, the conductivity threshold may be lower or needs to be combined with other characteristics for judgment.

[0087] This application's solution establishes a database linking conductivity with the spectral characteristics of various known residues, providing a solid foundation for subsequent residue identification. Because this database contains rich and validated residue characteristic information, the currently measured multi-wavelength spectral data can be effectively matched with known residues, thus initially identifying possible residues in the cleaning solution. Based on this, by combining the matching results with a preset conductivity threshold for preliminary screening, false positives or low-confidence results that may exist in spectral matching can be effectively eliminated. Specifically, the introduction of the conductivity threshold provides an independent and complementary verification dimension for spectral analysis, ensuring that the identified residues not only have spectral similarity but also meet expectations in electrochemical properties (reflected by conductivity). This dual verification mechanism significantly improves the accuracy and reliability of residue identification, avoiding misjudgments that may arise from a single measurement method.

[0088] The above technical solution enables more accurate identification of the types and concentrations of residues in cleaning solutions. By introducing a correlation database of conductivity and spectral characteristics, and combining it with a preset conductivity threshold for screening, this solution effectively addresses the insufficient accuracy of traditional single-spectral analysis or conductivity measurement in complex residue systems. Specifically, this method can more accurately distinguish between residues with similar spectral characteristics but different conductivity responses, or residues with similar conductivity responses but different spectral characteristics. This provides more reliable data support for judging the thoroughness of the cleaning process, thereby optimizing the cleaning strategy and ensuring that the working electrode is in its optimal activation state.

[0089] In semiconductor manufacturing, intelligent control of electroplating machines is crucial for ensuring wafer plating quality. However, traditional control methods often struggle to accurately distinguish between physical deviations in conductivity measurement devices and changes in the actual chemical state of the plating solution. This leads to incorrect judgments and compensations by the control system, impacting product yield and production stability. For example, when an adhesion layer appears on the probe surface of the conductivity measurement device, the conductivity measurement data will remain consistently low. In this case, existing systems may misinterpret this as reduced plating solution activity, incorrectly adjusting process parameters and accelerating the consumption of key organic additives in the plating solution, ultimately resulting in decreased plating uniformity. Failure to address these issues will lead to unstable production efficiency and inconsistent product quality, ultimately affecting semiconductor product yield and manufacturing costs.

[0090] Secondly, referring to Figure 2 This application further proposes an intelligent control system for a semiconductor electroplating machine, the system comprising: The conductivity acquisition module 201 is used to acquire conductivity measurement data of the main electroplating tank; The sample extraction module 202 is used to extract electroplating solution samples from the main electroplating tank. The electrochemical analysis module 203 is used to apply electrochemical excitation to the electroplating solution sample and analyze its electrochemical response characteristics to obtain the activity information of key organic components in the electroplating solution. The state determination module 204 is used to perform a state determination operation based on the conductivity measurement data and the activity information to obtain a state determination result. The state determination operation includes: When the conductivity measurement data shows a continuous downward trend, and the activity information indicates that the activity of the key organic components in the electroplating solution is within the preset normal range, it is determined that there is a physical deviation in the conductivity measurement device of the main electroplating tank. When the conductivity measurement data shows a continuous downward trend, and the activity information indicates that the activity of the key organic components in the electroplating solution is lower than the preset normal range, it is determined that the true chemical state of the electroplating solution has changed. The process parameter adjustment module 205 is used to adjust the electroplating process parameters of the wafer edge region when it is determined that the actual chemical state of the electroplating solution has changed, based on the state judgment result.

[0091] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for intelligent control of a semiconductor electroplating machine, characterized in that, include: Obtain conductivity measurement data of the main electroplating tank; An electroplating solution sample was extracted from the main electroplating tank; Electrochemical excitation was applied to the electroplating solution sample, and the electrochemical response characteristics of the electroplating solution sample were analyzed to obtain the activity information of key organic components in the electroplating solution; Based on the conductivity measurement data and the activity information, a state determination operation is performed to obtain a state determination result. The state determination operation includes: When the conductivity measurement data shows a continuous downward trend, and the activity information indicates that the activity of the key organic components in the electroplating solution is within the preset normal range, it is determined that there is a physical deviation in the conductivity measurement device of the main electroplating tank; when the conductivity measurement data shows a continuous downward trend, and the activity information indicates that the activity of the key organic components in the electroplating solution is lower than the preset normal range, it is determined that the true chemical state of the electroplating solution has changed. Based on the state judgment result, when it is determined that the actual chemical state of the electroplating solution has changed, the electroplating process parameters of the wafer edge region are adjusted.

2. The intelligent control method for a semiconductor electroplating machine according to claim 1, characterized in that, The step of performing a state determination operation based on the conductivity measurement data and the activity information to obtain the state determination result includes: Obtain health status information of the main conductivity measuring device; Acquire activation status information of the working electrode of the auxiliary electrochemical measurement device; the activation status information includes the passivation index; Receive conductivity measurement data from the main electroplating tank; Receive activity information of key organic components in the electroplating solution; Based on the health status information of the main conductivity measuring device, the activation status information of the working electrode of the auxiliary electrochemical measuring device, the conductivity measurement data, and the activation information, it is determined whether there is a physical deviation in the conductivity measuring device of the main electroplating tank, whether there is a deviation in the measurement results of the auxiliary electrochemical measuring device by obtaining the activation status information of the working electrode, and whether the true chemical state of the electroplating solution has changed, thus obtaining the state judgment result.

3. The intelligent control method for a semiconductor electroplating machine according to claim 2, characterized in that, The step of obtaining health status information of the main conductivity measuring device includes: When the main conductivity measuring device is in a non-electroplating period, an AC signal with a preset frequency and amplitude is applied to the main conductivity measuring device; While applying the AC signal, the local potential of the electroplating solution in the main electroplating tank near the probe area of ​​the main conductivity measuring device is monitored by the auxiliary electrode to obtain the potential change information of the local microenvironment of the electroplating solution. Based on the potential change information, the frequency or amplitude of the AC signal is adjusted to suppress the electrochemical reaction induced by the AC signal near the probe surface; The electrical response of the probe of the main conductivity measuring device at the interface with the electroplating solution was measured. The health status information of the main conductivity measuring device is obtained based on the electrical response.

4. The intelligent control method for a semiconductor electroplating machine according to claim 2, characterized in that, The step of obtaining the activation status information of the working electrode of the auxiliary electrochemical measurement device includes: Before applying a potential step signal to the working electrode, the auxiliary electrochemical measurement device is controlled to inject cleaning solution into the microcell to rinse the working electrode, thereby removing electrochemically active substances remaining on the electrode surface and in the electrolyte. After the rinsing operation is completed, the cleaning solution is drained, and a small amount of fresh electroplating solution sample is injected into the microcell by the auxiliary electrochemical measurement device. A standard potential step signal is applied to the working electrode, and the current decay curve generated on the working electrode is recorded; the current decay curve is the curve of the current generated on the working electrode decaying over time. The characteristic parameters of the current decay curve are extracted, and the difference between the currently measured decay curve characteristic parameters and the reference curve characteristic parameters in the clean and activated state of the working electrode is compared to obtain the activation status information of the working electrode of the auxiliary electrochemical measurement device.

5. The intelligent control method for a semiconductor electroplating machine according to claim 4, characterized in that, The step of controlling the auxiliary electrochemical measurement device to inject cleaning solution into the microcell before applying a potential step signal to the working electrode to rinse the working electrode and remove residual electrochemically active substances on the electrode surface and in the electrolyte includes: Based on the type of key organic components in the electroplating solution and the electrochemical characteristics of their decomposition products, determine the cleaning solution components with targeted dissolving or reacting capabilities. After the cleaning solution is injected, the working electrode is rinsed by applying a low-frequency alternating electric field or ultrasonic vibration to enhance the cleaning effect of the cleaning solution on the surface of the working electrode and the internal flow channels of the micro-pool. After rinsing is completed, the thoroughness of the cleaning process is determined by monitoring changes in the conductivity of the cleaning fluid or the spectral absorption characteristics of specific residues. If the cleaning process fails to meet the preset thoroughness standard, adjust the injection volume of cleaning fluid, rinsing time, and / or the intensity of low-frequency AC electric field or ultrasonic vibration, and repeat the rinsing process to remove electrochemically active substances remaining on the electrode surface and in the electrolyte.

6. The intelligent control method for a semiconductor electroplating machine according to claim 5, characterized in that, The step of determining the thoroughness of the cleaning process by monitoring changes in the conductivity of the cleaning fluid or the spectral absorption characteristics of specific residues after the rinsing operation is completed includes: After the rinsing operation is completed, the residual cleaning solution in the micro tank is scanned by multi-wavelength spectrum to obtain absorption spectrum data at different wavelengths, thus obtaining multi-wavelength spectrum data. The conductivity of the residual cleaning fluid in the micro tank was measured to obtain conductivity measurement data; The multi-wavelength spectral data and the conductivity measurement data are fused together to obtain the comprehensive residue characteristics of the cleaning fluid; The thoroughness of the cleaning process is determined based on the comprehensive residue characteristics information.

7. The intelligent control method for a semiconductor electroplating machine according to claim 6, characterized in that, The step of fusing the multi-wavelength spectral data with the conductivity measurement data to obtain the comprehensive residue characteristic information of the cleaning fluid includes: Based on the electrochemical characteristics of key organic components and decomposition products in the electroplating solution, as well as the spectral absorption and conductivity response of key organic components and decomposition products at different temperatures and pH values, the multi-wavelength spectral data and conductivity measurement data are preliminarily calibrated and preprocessed to obtain calibrated multi-wavelength spectral data and conductivity measurement data. The correlation and mutual influence between the calibrated multi-wavelength spectral data and conductivity measurement data are analyzed, and the fusion weight or fusion algorithm parameters are dynamically adjusted to determine whether there is a nonlinear or time-varying relationship between the multi-wavelength spectral data and conductivity measurement data. When a nonlinear or time-varying relationship is detected between the multi-wavelength spectral data and the conductivity measurement data, a nonlinear mapping or time-series analysis mechanism is activated to extract a comprehensive feature vector that can reflect the types and concentrations of residues in the cleaning solution. Based on the comprehensive feature vector, the comprehensive residue characteristic information of the cleaning fluid is obtained.

8. The intelligent control method for a semiconductor electroplating machine according to claim 7, characterized in that, The steps of analyzing the correlation and mutual influence between the calibrated multi-wavelength spectral data and conductivity measurement data, dynamically adjusting the fusion weights or fusion algorithm parameters, and determining whether there is a nonlinear or time-varying relationship between the multi-wavelength spectral data and conductivity measurement data include: Under multiple preset conductivity thresholds, the calibrated multi-wavelength spectral data are segmented for analysis to identify the spectral characteristics of residues with different conductivity responses. Under multiple preset spectral absorption characteristic peaks or specific wavelength combinations, the calibrated conductivity measurement data are segmented for analysis to identify the conductivity response corresponding to residues with different spectral absorption characteristics. Based on the spectral characteristics corresponding to the residues with different conductivity responses and the conductivity responses corresponding to the residues with different spectral absorption characteristics, the correlation and mutual influence contribution of each residue to the calibrated multi-wavelength spectral data and conductivity measurement data are evaluated. Based on the contribution of the evaluation, the fusion weights or fusion algorithm parameters are dynamically adjusted to determine whether there is a nonlinear or time-varying relationship between the multi-wavelength spectral data and the conductivity measurement data.

9. The intelligent control method for a semiconductor electroplating machine according to claim 8, characterized in that, The step of segmenting and analyzing the calibrated multi-wavelength spectral data under multiple preset conductivity thresholds to identify the spectral characteristics corresponding to residues with different conductivity responses includes: Establish a database linking electrical conductivity with the spectral characteristics of various known residues; The currently measured multi-wavelength spectral data is matched with the spectral features in the associated database to obtain the matching result; The matching results are initially screened by combining the preset conductivity threshold to identify the spectral characteristics of residues with different conductivity responses.

10. An intelligent control system for a semiconductor electroplating machine, characterized in that, The system includes: The conductivity acquisition module is used to acquire conductivity measurement data of the main electroplating tank; the sample extraction module is used to extract electroplating solution samples from the main electroplating tank. An electrochemical analysis module is used to apply electrochemical excitation to the electroplating solution sample and analyze its electrochemical response characteristics to obtain activity information of key organic components in the electroplating solution. The state determination module is used to perform a state determination operation based on the conductivity measurement data and the activity information to obtain a state determination result. The state determination operation includes: When the conductivity measurement data shows a continuous downward trend, and the activity information indicates that the activity of the key organic components in the electroplating solution is within the preset normal range, it is determined that there is a physical deviation in the conductivity measurement device of the main electroplating tank. When the conductivity measurement data shows a continuous downward trend, and the activity information indicates that the activity of the key organic components in the electroplating solution is lower than the preset normal range, it is determined that the true chemical state of the electroplating solution has changed. The process parameter adjustment module is used to adjust the electroplating process parameters of the wafer edge region when it is determined that the actual chemical state of the electroplating solution has changed, based on the state judgment result.

Images