A method for dynamically monitoring the thickness of an ITO coating

Abstract

The present application relates to the technical field of thin film deposition and vacuum coating monitoring control, in particular to a kind of ITO coating thickness dynamic monitoring method;Including: obtaining target material quality parameter, process gas parameter and substrate material quality parameter;Initial sheath equivalent circuit topology model containing impedance film thickness kinetics equation is constructed;RF sweep perturbation signal is injected to the substrate tray multiplexed as RF antenna;The phase difference of the RF voltage and current of the reflection signal is continuously collected, and the resonance peak frequency time sequence and its drift slope corresponding to the phase difference of zero are extracted;Substitute impedance film thickness kinetics equation to solve thin film geometric thickness and target material poisoning index;When thin film geometric thickness is less than target thickness, adjust oxygen flow, when greater than or equal to target thickness, fine-tune sputtering target power and oxygen flow, to complete coating control;Realize the synchronous perception of film thickness and target surface state, improve the thickness control precision and process endpoint consistency.

Description

Technical Field

[0001] This invention relates to the field of thin film deposition and vacuum coating monitoring and control technology, specifically to a method for dynamic monitoring of ITO coating thickness. Background Technology

[0002] With the increasing demands on the performance of transparent conductive films in products such as automotive displays and touch panels, indium tin oxide (ITO) coating has become an important preparation step in the field of vacuum coating. In order to ensure the consistency of sheet resistance, transmittance and thickness of conductive films, dynamic monitoring of film thickness changes and target conditions during the coating process has become particularly important.

[0003] In existing DC magnetron sputtering processes for indium tin oxide (ITO), film thickness control typically relies on preset deposition time, offline sampling, or external thickness probes. However, the target lifespan, oxygen flow rate, cavity condition, and substrate surface film formation process all affect the actual deposition rate. Especially when the target is poisoned by oxidation or the film transitions from island-like growth to continuous conductive film growth, time-based control alone cannot reflect changes in film thickness and electrical properties in a timely manner. External thickness measurement equipment may also have problems such as limited installation space, plasma interference, and increased complexity of the system hardware structure, making it difficult to meet the requirements of in-situ, continuous, and low-disturbance monitoring on the production line.

[0004] Therefore, how to utilize the existing substrate tray, RF signal link, and process execution components in the vacuum coating system to process the RF impedance changes during the coating process, simultaneously obtain information on film thickness, target state, and conductivity, and adjust oxygen flow and sputtering power accordingly, is crucial for improving the endpoint control accuracy and batch consistency of indium tin oxide coating. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a method for dynamically monitoring the thickness of ITO coatings. Specifically, the technical solution of this invention includes:

[0006] The method is applied to a vacuum coating system, the vacuum coating system including a substrate tray multiplexed as a radio frequency antenna, a radio frequency generator, an impedance matching network, a sputtering DC power supply, a mass flow meter, and a controller, the method being executed by the controller, the method comprising:

[0007] Obtain the preset target material parameters, process gas parameters, and substrate material parameters;

[0008] Based on the target material parameters, the process gas parameters, and the substrate material parameters, an initial sheath equivalent circuit topology model is constructed, which includes a preset impedance film thickness dynamics equation characterizing the relationship between RF impedance and thin film growth and target state.

[0009] Based on the radio frequency generator and the impedance matching network, a radio frequency sweep perturbation signal is injected into the base tray;

[0010] Within a preset time window, the phase difference between the radio frequency voltage and current of the radio frequency sweep perturbation signal reflected by the base tray is continuously acquired;

[0011] Based on the continuously acquired phase difference of the radio frequency voltage and current, the frequency corresponding to the zero phase difference is extracted, and this frequency is determined as the resonant peak frequency, and a time series of the resonant peak frequency is generated.

[0012] Extract the drift slope of the time series of the resonant peak frequency as a function of time;

[0013] Substituting the drift slope into the impedance film thickness dynamics equation, the current film geometric thickness and target poisoning index are calculated.

[0014] If the geometric thickness of the thin film is less than the preset target thickness, the oxygen flow rate of the mass flow meter is adjusted according to the poisoning index of the target material.

[0015] When the geometric thickness of the thin film is greater than or equal to the target thickness, the sputtering target power of the sputtering DC power supply and the oxygen flow rate of the mass flow meter are finely adjusted according to the derivative of the drift slope to complete the coating control.

[0016] Optionally, the step of constructing an initial sheath equivalent circuit topology model based on the target material parameters, the process gas parameters, and the substrate material parameters, including a preset impedance film thickness dynamics equation characterizing the relationship between RF impedance and thin film growth and target state, includes:

[0017] The initial plasma density is determined based on the target material parameters and the process gas parameters.

[0018] Based on the substrate material parameters, determine the initial substrate surface dielectric constant;

[0019] By combining the initial plasma density and the initial substrate surface dielectric constant, the initial sheath equivalent circuit topology model is generated;

[0020] The initial sheath equivalent circuit topology model includes thin film conductivity variables and sheath capacitance variables.

[0021] Optionally, injecting a radio frequency sweep perturbation signal into the substrate tray based on the radio frequency generator and the impedance matching network includes:

[0022] Based on the radio frequency generator, a wideband radio frequency signal within a preset radio frequency band range is generated;

[0023] The impedance matching network is used to perform impedance matching on the wideband radio frequency signal to generate the radio frequency sweep perturbation signal.

[0024] The radio frequency sweep perturbation signal is injected into the base tray at a preset power.

[0025] Optionally, extracting the drift slope of the time series of the resonant peak frequency over time includes:

[0026] Extract the current resonant peak frequency at the current moment and the initial resonant peak frequency at the initial moment from the time series of the resonant peak frequencies;

[0027] Calculate the frequency shift of the current resonant peak frequency compared to the initial resonant peak frequency;

[0028] If the frequency shift is greater than a preset positive frequency shift threshold, the film is determined to have entered the continuous conductive film growth stage.

[0029] If the frequency shift is less than or equal to the positive frequency shift threshold, the film is determined to be in the isolated island growth stage.

[0030] If the film is determined to be in the isolated island growth stage, a thickness monitoring waiting prompt is output, and the process returns to the step of continuously acquiring the RF voltage and current phase difference of the RF sweep frequency perturbation signal reflected by the substrate tray within a preset time window.

[0031] When it is determined that the thin film has entered the continuous conductive film growth stage, the drift slope is generated by linear fitting based on the time series of the resonant peak frequency.

[0032] Optionally, substituting the drift slope into the impedance film thickness dynamics equation to calculate the current film geometric thickness and target poisoning index includes:

[0033] The drift slope is input into the impedance film thickness dynamic equation to calculate the overall impedance change rate;

[0034] Based on the Fast Fourier Transform algorithm, the comprehensive impedance change rate is subjected to frequency domain separation processing to generate low-frequency thickness-related components and high-frequency oscillation components; a preset thickness conversion coefficient is obtained, which is a proportionality constant extracted by pre-collecting low-frequency thickness-related components under the same substrate and target material conditions from multiple historical calibration batches and performing linear regression analysis on the corresponding thin film geometric thickness data. The low-frequency thickness-related components are multiplied by the preset thickness conversion coefficient to calculate the thin film geometric thickness.

[0035] The amplitude characteristics of the high-frequency oscillation component, including its absolute peak value, root mean square value, or sliding peak-to-peak value, are extracted as the target poisoning index.

[0036] Optionally, adjusting the oxygen flow rate of the mass flow meter according to the target poisoning index when the geometric thickness of the thin film is less than the preset target thickness includes:

[0037] If the poisoning index of the target material is greater than the preset poisoning safety threshold, it is determined that the target surface has undergone excessive oxidation.

[0038] In response to excessive oxidation of the target surface, a reduction command is generated and sent to the mass flow meter to reduce the oxygen flow rate;

[0039] If the poisoning index of the target material is less than or equal to the poisoning safety threshold, the target surface condition is determined to be normal, and the oxygen flow rate at the current moment is maintained.

[0040] Optionally, when the geometric thickness of the thin film is greater than or equal to the target thickness, fine-tuning the sputtering target power of the sputtering DC power supply and the oxygen flow rate of the mass flow meter according to the derivative of the drift slope includes:

[0041] Calculate the derivative of the drift slope over time to generate the impedance gradient derivative; obtain preset power adjustment ratio coefficients and flow adjustment ratio coefficients, which are pre-tuned based on the target thickness margin, the current poisoning index, and a preset historical batch thickness fluctuation variance threshold; multiply the impedance gradient derivative by the power adjustment ratio coefficients and flow adjustment ratio coefficients respectively to generate power attenuation curves and flow attenuation curves;

[0042] The sputtering target power is gradually reduced according to the power decay curve.

[0043] The oxygen flow rate is gradually reduced according to the flow rate decay curve until the sputtering target power drops to zero.

[0044] Optionally, the method further includes:

[0045] The amplitude of the radio frequency reflection coefficient of the radio frequency sweep perturbation signal reflected by the base tray is acquired;

[0046] The amplitude of the radio frequency reflection coefficient is input into a preset sheet resistance mapping model that characterizes the relationship between the reflection coefficient and the sheet resistance, and the in-situ sheet resistance at the current moment is calculated.

[0047] If the in-situ sheet resistance is higher than a preset poor conductivity threshold, an alarm message indicating abnormal conductivity is generated.

[0048] Optionally, the method further includes:

[0049] If the abnormal conductivity alarm message is generated, an interrupt command is sent to the sputtering DC power supply to terminate the coating process.

[0050] If the in-situ sheet resistance is lower than or equal to the poor conductivity threshold, the coating control continues.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] 1. The present invention provides a dynamic monitoring method for ITO coating thickness. By reusing the existing substrate tray in the vacuum coating system as a radio frequency antenna and injecting a radio frequency sweep perturbation signal into it, the phase difference of the radio frequency voltage and current reflected by the substrate tray is continuously collected to extract the time series of the resonant peak frequency and the drift slope. The time series and drift slope are then substituted into the impedance film thickness dynamics equation to calculate the geometric thickness of the film at the current moment. This method does not require the addition of an external thickness measurement probe that is susceptible to plasma interference and has limited installation space. It overcomes the shortcomings of the prior art that only relies on the preset deposition time. Without significantly changing the main sputtering process, it achieves high-precision, in-situ continuous monitoring of film thickness.

[0053] 2. This invention calculates the comprehensive impedance change rate by inputting the drift slope into the impedance film thickness dynamic equation, and performs frequency domain separation processing based on the Fast Fourier Transform algorithm. This successfully decouples the complex impedance signal into low-frequency thickness-related components and high-frequency oscillation components, thereby enabling the simultaneous and independent calculation of the film geometric thickness and the target poisoning index, which characterizes the target state. When the film has not yet reached the preset target thickness, the system can accurately determine whether excessive oxidation has occurred on the target surface, and promptly generate a reduction command based on the target poisoning index to adjust the oxygen flow rate of the mass flow meter, effectively suppressing the target poisoning phenomenon and maintaining the stability of the deposition rate and the continuity of the process.

[0054] 3. When the geometric thickness of the thin film is greater than or equal to the target thickness, this invention generates power attenuation curves and flow rate attenuation curves based on the derivative of the drift slope. It then fine-tunes and gradually reduces the sputtering target power and oxygen flow rate, achieving gradual convergence at the coating endpoint. This avoids the end-point runaway and uneven thickness caused by directly cutting off the power supply in traditional methods. Furthermore, this invention further collects the amplitude of the radio frequency reflection coefficient to calculate the in-situ sheet resistance at the current moment. When the in-situ sheet resistance exceeds a preset conductivity failure threshold, it actively generates an abnormal conductivity alarm message and sends an interrupt command to the sputtering DC power supply to terminate the coating process. This ensures that the film conductivity is also qualified while the film thickness meets the standard, significantly improving the consistency of production line batches and the pass rate of thin film preparation. Attached Figure Description

[0055] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0056] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0058] Example 1:

[0059] Please see Figure 1 A method for dynamically monitoring the thickness of ITO coating is disclosed. The method is applied to a vacuum coating system, which includes a substrate tray reused as a radio frequency antenna, a radio frequency generator, an impedance matching network, a sputtering DC power supply, a mass flow meter, and a controller. The method is executed by the controller and includes:

[0060] Obtain the preset target material parameters, process gas parameters, and substrate material parameters;

[0061] Based on the target material parameters, process gas parameters, and substrate material parameters, an initial sheath equivalent circuit topology model is constructed, which includes a preset impedance film thickness dynamics equation characterizing the relationship between RF impedance and thin film growth and target state.

[0062] Based on an RF generator and an impedance matching network, an RF sweep frequency perturbation signal is injected into the base tray;

[0063] Within a preset time window, the phase difference between the radio frequency voltage and current of the radio frequency sweep frequency perturbation signal reflected by the base tray is continuously acquired;

[0064] Based on the continuously acquired phase difference of radio frequency voltage and current, the frequency corresponding to the zero phase difference is extracted, and this frequency is determined as the resonant peak frequency, and a time series of the resonant peak frequency is generated.

[0065] Extract the drift slope of the time series of the resonant peak frequency as a function of time;

[0066] Substituting the drift slope into the impedance film thickness dynamics equation, the current film geometric thickness and target poisoning index are calculated.

[0067] When the geometric thickness of the thin film is less than the preset target thickness, the oxygen flow rate of the mass flow meter is adjusted according to the poisoning index of the target material.

[0068] When the geometric thickness of the thin film is greater than or equal to the target thickness, the sputtering target power of the sputtering DC power supply and the oxygen flow rate of the mass flow meter are finely adjusted according to the derivative of the drift slope to complete the coating control.

[0069] This embodiment provides a control mechanism for dynamic monitoring of ITO coating thickness. Specifically, this embodiment is applied to a DC magnetron sputtering production line for preparing conductive layers of automotive display glass. Within the same vacuum chamber, the workpiece to be coated consists of multiple soda-lime glass substrates. The substrate tray is a metal disk structure, which serves both as a substrate support and as an antenna structure for RF perturbation injection and reflection acquisition. The production line aims to form a 120-nanometer ITO conductive film in a single batch, while maintaining consistent thickness throughout the early and late stages of the target's lifespan. Here, the DC magnetron sputtering production line refers to production equipment where the main sputtering discharge is provided by a DC sputtering power supply. The RF generator is used only as a low-power sweep frequency perturbation signal source and is not used as the main sputtering power supply to ensure that the power supply division in this embodiment is consistent with the hardware configuration of the aforementioned vacuum coating system.

[0070] Specifically, before the coating process begins, the controller reads three types of input information from the preset process card. The first type is the target material parameters, such as the ratio of indium oxide to tin oxide in the indium tin oxide target, the initial thickness of the target, and the resistivity of the target. The second type is the process gas parameters, such as the basic flow rate of argon, the initial flow rate of oxygen, and the set value of the working pressure. The third type is the substrate material parameters, such as the dielectric constant of the glass, its thickness, its area, and whether it has a pre-cleaned surface activation layer. Based on this input information, the controller establishes an initial topological model of the substrate surface conductive channel—sheath capacitance—plasma coupling impedance, and writes a set of preset mapping relationships into this model to characterize the linkage between resonant frequency drift, film thickness growth, and target surface oxidation state. Preferably, the preset mapping relationships can be obtained from historical calibration batches and stored in the controller in the form of a lookup table, piecewise function, or parameterized model for recall at different target life stages.

[0071] During the coating process, the sputtering DC power supply maintains the target discharge, while the RF generator does not participate in the main sputtering. Instead, it injects a low-power sweep frequency perturbation signal into the substrate tray through an impedance matching network. The power of this perturbation signal can be set from 1 watt to 3 watts to avoid causing significant disturbance to the main process.

[0072] To illustrate with an example set of data, suppose that five discrete frequency points were detected within a certain frequency sweep range at a certain moment, namely 10 MHz, 15 MHz, 20 MHz, 25 MHz and 30 MHz, and the corresponding voltage and current phase differences were +12 degrees, +4 degrees, 0 degrees, -5 degrees and -11 degrees, respectively. Then the controller would identify 20 MHz as the resonant peak frequency at that moment. If the resonant peak frequencies obtained in the next three consecutive time windows are 20 MHz, 20.6 MHz and 21.3 MHz, respectively, then a time series of resonant peak frequencies arranged in time can be formed.

[0073] Furthermore, the controller calculates the drift slope based on this time series. Using a simplified example, assuming the resonant frequency is 20 MHz in the first minute and 21 MHz in the third minute, the average drift slope over those two minutes is 0.5 MHz / minute. This slope is then input into a preset impedance film thickness dynamics equation, yielding two outputs: the current geometric thickness and the target poisoning index. If, at this moment, the calculated geometric thickness is 92 nm, the target poisoning index is 0.76, and the target thickness is 120 nm, the system determines that the target thickness has not yet been reached and should prioritize adjusting the oxygen flow rate based on the target poisoning index.

[0074] If the geometric thickness is calculated to be 121 nanometers, the system will no longer execute the control strategy of continuous thickness increase. Instead, it will read the derivative of the drift slope and fine-tune the target power and oxygen flow rate in conjunction with it, and enter the endpoint gradual convergence control stage, so that the deposition rate of the stage decreases gradually rather than being cut off instantaneously. The endpoint gradual convergence control stage here means that after the target thickness condition is met, the controller continuously updates the power and flow rate setpoints based on the latest drift slope derivatives of multiple time windows, so that the deposition process gradually transitions to the end state.

[0075] It should be noted that two types of anomalies may occur in actual production. First, if no frequency with zero phase difference is collected within a certain time window, the controller can interpolate using the two adjacent frequency points with the closest phase difference to zero to obtain an approximate resonant frequency. If adjacent frequency points are missing or the signal-to-noise ratio is lower than a preset threshold, the data in that window is discarded, and the reliable resonant peak frequency of the previous window is used and marked as low confidence. Second, if the calculated geometric thickness shows an abnormal decrease, for example, it was 88 nanometers in the previous moment but is currently calculated to be 81 nanometers, it indicates that plasma fluctuations or external electromagnetic interference have caused instantaneous distortion. In this case, the controller does not directly perform adjustment, but adds a confirmation window for resampling. The control command is updated only after consistent results are obtained in two consecutive windows.

[0076] In this automotive display glass production line, each batch consists of 24 wafers, and the target material is nearing the end of its lifespan. Traditional time-controlled processes, even with a 12-minute timer, often result in a thinner film due to changes in the target surface condition. In this embodiment, the system primarily records impedance changes during the isolated island growth process in the first two minutes of coating. Around the third minute, as a continuous conductive film forms on the substrate surface, a stable blue shift occurs in the resonance peak, and the system begins to invert the thickness based on the drift slope. When the thickness reaches 118 nanometers and the target poisoning index increases, the oxygen flow rate is first reduced from 10 standard cubic centimeters per minute to 8 standard cubic centimeters per minute.

[0077] When the thickness exceeds 120 nanometers, the target power is gradually reduced from 4 kW to 2 kW according to a predetermined curve, and the oxygen flow rate is simultaneously reduced from 8 standard cubic centimeters / minute to 5 standard cubic centimeters / minute to make the final stage of coating more stable. The purpose of this step is to transform the traditional thickness measurement method that relies on external probes into an in-situ monitoring method based on the RF impedance reflected by the substrate tray, so as to achieve synchronous perception of film thickness and target surface status, and improve the control accuracy of ITO coating thickness and the consistency of process endpoint through closed-loop linkage of gas flow rate and sputtering power.

[0078] Based on the target material parameters, process gas parameters, and substrate material parameters, an initial sheath equivalent circuit topology model is constructed, including a pre-defined impedance film thickness dynamics equation characterizing the relationship between RF impedance and thin film growth and target state. This model includes:

[0079] The initial plasma density is determined based on the target material parameters and process gas parameters;

[0080] Determine the initial substrate surface dielectric constant based on the substrate material parameters;

[0081] By combining the initial plasma density and the initial substrate surface dielectric constant, an initial sheath equivalent circuit topology model is generated;

[0082] The initial sheath equivalent circuit topology model includes thin film conductivity variables and sheath capacitance variables.

[0083] This embodiment provides a modeling mechanism for establishing an initial sheath equivalent circuit topology model. Specifically, in the aforementioned automotive display glass ITO coating scenario, resonant frequency monitoring alone is insufficient to indicate whether the change comes from film thickness growth or discharge condition drift. To avoid the inability to distinguish different physical factors causing impedance changes in subsequent inversion, this embodiment first initializes the plasma state and substrate electrical boundary at the beginning of coating.

[0084] Specifically, the controller estimates the initial plasma density based on the target material parameters and process gas parameters. A pre-calibrated lookup table or empirical model can be used here. For example, suppose the lookup results for the resistivity of a certain batch of target materials, the remaining thickness of the target surface, and the argon / oxygen flow ratio are in three levels: low density, medium density, and high density. If the current input parameter falls into the medium density level, the system assigns a nominal value to the initial plasma density. For example, the medium density level can be mapped to coefficient 2, the low density level to coefficient 1, and the high density level to coefficient 3.

[0085] The controller determines the initial dielectric constant of the substrate surface based on the glass substrate material, substrate thickness, and surface cleaning state. If the substrate is ordinary soda-lime glass, a first dielectric constant range can be assigned; if it is high borosilicate glass, a second dielectric constant range can be assigned. The system does not require an infinitely precise absolute value during initial modeling, but rather aims to form a starting point that can be dynamically updated with the coating process.

[0086] In terms of topology, this embodiment simplifies the substrate tray, initial ITO nucleation layer, sheath region, and plasma region into a series-parallel combined network. The thin film conductivity variable characterizes the in-plane conductive channels that gradually increase with deposition, while the sheath capacitance variable characterizes the energy storage capacity of the electric field between the substrate and the plasma. The specific logical derivation is as follows: if a continuous film has not yet formed initially, the thin film conductivity variable takes a preset initial value, and the sheath capacitance variable dominates. As deposition progresses, the thin film conductivity variable continuously increases, and the resonant position of the entire equivalent network shifts towards higher frequencies. The controller only needs to update these two core variables to continuously track the complex discharge system with a low computational burden.

[0087] In the anomaly handling mechanism, if the material information of a certain batch of substrates is missing, the controller can first use the default dielectric constant of the previous batch of similar products to establish a temporary model, and mark the batch as a batch to be corrected; after obtaining the first stage sampling data, the initial parameters are recalibrated using the measured resonance characteristics; if the target material is close to the end of its life, causing the deviation between the pre-calibration table and the current state to exceed the preset calibration threshold, the system can introduce a short-time idle frequency sweep, and read a set of baseline resonance frequencies before the deposition begins to correct the initial plasma density estimate;

[0088] In the aforementioned automotive display glass production line, a certain batch used an ITO target that had been 80% consumed. The process was set to 60 standard cubic centimeters per minute for argon and 8 standard cubic centimeters per minute for oxygen. Based on this flow ratio and the remaining thickness of the target material, the system determined that the initial plasma density was within the preset range of the first density to the second density. Then, based on the glass substrate thickness of 0.7 mm and the material type, it determined that the initial surface dielectric constant was within the preset range.

[0089] An initial model is established, consisting of the tray induction branch, the thin film conductivity branch, and the sheath capacitance branch. The thin film conductivity variable is set to an initial value close to zero, and the sheath capacitance variable is set to an initial value determined by the substrate area and the initial sheath thickness. All subsequent resonant frequency changes are interpreted and tracked within this model framework. The purpose of this step is to provide initial conditions with clear physical boundaries for subsequent impedance inversion, thereby distinguishing between film thickness growth and target surface state changes, and improving the interpretability and stability of the entire dynamic monitoring process.

[0090] Based on an RF generator and impedance matching network, an RF sweep perturbation signal is injected into the substrate tray, including:

[0091] Based on the radio frequency generator, a wideband radio frequency signal within a preset radio frequency band is generated;

[0092] Impedance matching of wideband radio frequency signals is achieved by using an impedance matching network to generate radio frequency sweep perturbation signals.

[0093] The radio frequency sweep perturbation signal is injected into the base tray at a preset power.

[0094] This embodiment provides a radio frequency sweep perturbation signal injection mechanism. Specifically, after the aforementioned model is established, if the radio frequency source is directly electrically connected to the substrate tray, it is easy to generate reflections exceeding the preset tolerance, resonant point distortion, or even perturbation signals that cause additional disturbances to the main process due to the impedance fluctuations in the cavity during discharge. Therefore, this embodiment introduces an injection method that combines wideband signal generation with impedance matching, so that the perturbation measurement can cover the effective resonant range without significantly changing the deposition process.

[0095] Specifically, the RF generator generates a wideband signal covering a preset frequency range. This frequency range can be predetermined based on the device size and substrate tray structure, for example, from 1 MHz to 50 MHz. The controller can use point-by-point frequency sweeping or segmented frequency sweeping. To simplify the example, if a device is found to have its main resonant region concentrated between 15 MHz and 28 MHz after pre-calibration, the system can prioritize denser sampling in this frequency band, for example, sampling once every 0.5 MHz, while sampling once every 2 MHz in other frequency bands. This ensures both the accuracy of resonant peak location and the control of sampling time. The generated wideband signal is then injected into the substrate tray after passing through an impedance matching network. The function of the impedance matching network is to match the impedance at the output of the RF source with the impedance of the tray-plasma composite load to a preset tolerance range, thereby improving the available injected energy and the stability of reflection measurement.

[0096] To illustrate with a specific example, if the reflection coefficient amplitude is close to 1 when mismatched at a certain moment, it means that more than a preset proportion of radio frequency energy is reflected, and the resonance characteristics will be difficult to distinguish. At this time, the controller adjusts the variable capacitor or inductor in the matching network, the reflection coefficient amplitude decreases, and the phase zero crossing characteristics in the resonance region become clearer.

[0097] Finally, the perturbation signal is injected into the substrate tray at a preset power, which is controlled at a low level to prevent additional heating or secondary discharge. In the anomaly handling mechanism, if an abnormal rise in the tray surface potential is detected during the injection process, it indicates that the current perturbation power is higher than the preset matching power limit relative to the process state. The system should automatically reduce the output power of the RF generator and re-execute a low-power frequency sweep. If the impedance matching network cannot achieve stable matching in a certain frequency band, the system can narrow the current scanning range, retain only the frequency bands that have been verified to be stable for monitoring, and record the status of this batch of equipment for subsequent maintenance to check whether the matching components are aging.

[0098] In this automotive display glass production line, after the equipment starts coating, the RF generator first outputs a wideband signal from 1 MHz to 50 MHz; since the previous calibration showed that a stable resonance peak is more likely to appear in the 18 MHz to 24 MHz band, the controller sweeps the frequency at finer intervals in this band; the impedance matching network automatically adjusts the variable capacitor according to the real-time reflection situation to stabilize the perturbation signal injected into the substrate tray at 1 watt.

[0099] In actual operation, even if an ITO deposition layer is attached to the edge of the tray, the system can still read a clear phase change curve through the matched reflection characteristics. The purpose of this step is to enable the measurement signal to be stably coupled to the virtual resonant structure formed by the substrate tray and the plasma, so as to achieve continuous tracking of the resonant peak position and reduce the interference of the measurement process on the original coating process.

[0100] Extract the drift slope of the time series of the resonant peak frequency as a function of time, including:

[0101] Extract the current resonant peak frequency at the current moment and the initial resonant peak frequency at the initial moment from the time series of resonant peak frequencies;

[0102] Calculate the frequency shift of the current resonant peak frequency compared to the initial resonant peak frequency;

[0103] If the frequency shift is greater than the preset positive frequency shift threshold, the film is determined to have entered the continuous conductive film growth stage.

[0104] If the frequency shift is less than or equal to the positive frequency shift threshold, the film is determined to be in the isolated island growth stage.

[0105] If the film is determined to be in the isolated island growth stage, a thickness monitoring waiting prompt is output, and the process returns to the step of continuously acquiring the RF voltage and current phase difference of the RF sweep frequency perturbation signal reflected by the substrate tray within a preset time window.

[0106] Once the film is determined to have entered the continuous conductive film growth stage, the drift slope is generated by linear fitting based on the time series of the resonant peak frequency.

[0107] This embodiment provides a mechanism for extracting the resonant peak frequency drift slope. Specifically, in the early stage of actual ITO deposition, the film often does not immediately form a continuous conductive layer, but grows first in a discrete island-like nucleation manner. If the resonant frequency change is directly used for film thickness inversion at this stage, the jump caused by the discontinuous nucleation may be misjudged as effective thickness increase.

[0108] Therefore, this embodiment first distinguishes between the isolated island growth stage and the continuous conductive film growth stage, and only when the latter is entered does the drift slope generate; specifically, the controller reads the initial resonant peak frequency and the current resonant peak frequency from the time series and calculates the frequency shift.

[0109] To illustrate with an example set of data, assuming the initial frequency is 18.0 MHz, the frequency is 18.2 MHz at the first minute, 18.4 MHz at the second minute, and 19.3 MHz at the third minute; if the forward frequency shift threshold is set to 0.8 MHz, the frequency shifts in the first two minutes are 0.2 MHz and 0.4 MHz respectively, neither exceeding the threshold, therefore it is determined that it is still in the isolated island growth stage; by the third minute, the frequency shift is 1.3 MHz, exceeding the threshold, and the system determines that it has entered the continuous conductive film growth stage; when in the isolated island growth stage, the controller does not output the actual thickness value, but instead outputs a thickness monitoring waiting prompt and continues sampling;

[0110] This setting prevents premature and unstable adjustments to oxygen flow rate and target power during the nucleation transition. After entering the continuous conductive film growth stage, the system takes the resonant peak frequencies within the most recent time windows and performs linear fitting to obtain the drift slope. For example, if the frequency increases from 19.3 MHz and 19.8 MHz to 20.4 MHz between the 3rd and 5th minutes, the system fits an upward straight line to these three points, and its slope can be used as the effective drift velocity for the current stage.

[0111] Furthermore, the number of recent time windows can be preset to 3 to 10; when the sampling noise exceeds the preset noise threshold, the controller preferably increases the number of fitting windows to improve slope stability; when it is necessary to shorten the response period, the controller preferably decreases the number of fitting windows to shorten the stage recognition delay.

[0112] It should be noted that if the frequency shift fluctuates repeatedly around the threshold, for example, if the frequency shifts corresponding to three consecutive windows are 0.78 MHz, 0.82 MHz and 0.79 MHz, the system will not immediately switch stages. Instead, it will require two or three consecutive windows to be stably higher than the threshold before determining that it has entered the continuous conductive film growth stage.

[0113] Conversely, if the continuous conductive film growth stage has been entered, but the frequency curve drops briefly due to process fluctuations, the system will prioritize retaining the current stage state and only reassess whether there is abnormal discharge or film rupture when significant mismatch is observed in multiple consecutive windows.

[0114] In the aforementioned batch of automotive display glass, for approximately the first 150 seconds after the coating process begins, the system only displays a thickness monitoring wait; during this time, the system does not stop operating but continues to record radio frequency phase data; around the 180th second, the frequency shift first stably exceeds the preset positive frequency shift threshold, indicating that the ITO island-like structures on the glass surface have formed a continuous conductive network; only then does the system begin to linearly fit the resonance peaks of the past few windows and output an effective thickness curve; thus outputting a smoother thickness growth trajectory that conforms to the actual deposition pattern;

[0115] The purpose of this step is to distinguish the different impedance behaviors during the nucleation period and the continuous thickening period of the thin film, so as to output the drift slope only during the stage where it can be stably inverted, and avoid the thickness measurement error caused by the initial nucleation.

[0116] Substituting the drift slope into the impedance film thickness dynamics equation, the current film geometric thickness and target poisoning index are calculated, including:

[0117] The drift slope is input into the impedance film thickness dynamics equation to calculate the overall impedance change rate;

[0118] Based on the Fast Fourier Transform algorithm, the comprehensive impedance change rate is subjected to frequency domain separation processing to generate low-frequency thickness-related components and high-frequency oscillation components. A preset thickness conversion coefficient is obtained. The preset thickness conversion coefficient is a proportionality constant extracted by pre-collecting low-frequency thickness-related components under the same substrate and target material conditions from multiple historical calibration batches and performing linear regression analysis on the corresponding thin film geometric thickness data. The low-frequency thickness-related components are multiplied by the preset thickness conversion coefficient to calculate the thin film geometric thickness.

[0119] Extract any one of the amplitude characteristics of the high-frequency oscillation component, such as the absolute peak value, root mean square value, or sliding peak-to-peak value, as the target poisoning index.

[0120] This embodiment provides a mechanism for simultaneously calculating geometric thickness and target poisoning index. Specifically, relying solely on the drift slope may still result in the superposition of smooth changes caused by film thickness and rapid disturbances caused by target oxidation. Without separation, the controller may misjudge the high-frequency disturbances caused by target poisoning as rapid changes in thickness. Therefore, this embodiment decomposes the comprehensive impedance change rate into low-frequency thickness-related components and high-frequency oscillation components, which correspond to film thickness calculation and target condition assessment, respectively.

[0121] Specifically, the controller first inputs the drift slope into the impedance film thickness dynamics equation to obtain a time-ordered sequence of comprehensive impedance change rates. The impedance film thickness dynamics equation is a transfer function characterizing the change of radio frequency impedance with plasma parameters and film thickness. Specifically, the equation establishes a differential relationship with the thin film conductivity and sheath capacitance variables in the initial sheath equivalent circuit topology model as independent variables. By extracting the input drift slope as the change derivative term of the frequency domain response, the total equivalent impedance at both ends of the equivalent circuit network is calculated as a percentage of change over time, thereby obtaining the corresponding comprehensive impedance change rate.

[0122] The specific mathematical expression of the impedance film thickness dynamic equation is as follows:

[0123] ;

[0124] in, For time, The total equivalent impedance is... For the thin film conductivity variable, For the sheath capacitance variable, The resonant peak frequency, The input drift slope, The preset frequency domain response coupling coefficient;

[0125] It should be noted that the input for frequency domain separation is not a single value at an isolated moment, but a short time sequence composed of the current window and several consecutive windows preceding it. In other words, each time the controller enters a new sampling window, it constructs an analysis vector from the comprehensive impedance change rates within the most recent N consecutive windows in chronological order, and then performs frequency domain separation on this analysis vector. Here, N represents the number of consecutive sampling windows used to construct the analysis vector, which is a natural number greater than 1, and can be preset or adaptively adjusted by the controller according to real-time requirements and frequency domain resolution requirements.

[0126] To illustrate with example values, assume the overall impedance change rate for eight consecutive windows is 2, 3, 3, 4, 8, 4, 3, 3. Most of the data shows a gradual upward trend, while the deviation of the fifth window (8) exceeds the preset normal fluctuation range, potentially containing high-frequency components caused by abrupt changes in the target surface state. The system performs frequency domain separation processing on this sequence. After separation, a low-frequency thickness-related component is obtained, such as 2.5, 2.8, 3.1, 3.4, 3.6, 3.5, 3.3, 3.2; simultaneously, a high-frequency oscillation component is obtained, such as -0.5, 0.2, -0.1, 0.6, 4.4, 0.5, -0.3, -0.2.

[0127] Furthermore, to avoid spectral leakage caused by window boundaries, the controller can first perform mean removal and weighting on the short time series, and then perform a fast Fourier transform; after completing the frequency domain separation, the low-frequency part and the high-frequency part are mapped back to the thickness information and target surface state information corresponding to the current time, respectively.

[0128] In other words, the geometric thickness and target poisoning index output at the current moment are essentially a structured interpretation of the recent continuous change process, rather than a direct copy of a certain instantaneous peak. This can improve the physical reliability of the solution results. When calculating the film thickness, the system multiplies the low-frequency thickness-related components by a preset thickness conversion factor to obtain the current geometric thickness.

[0129] For example, if the thickness conversion factor is 30 nanometers / unit, then the low-frequency component 3.6 corresponds to a thickness of approximately 108 nanometers. At the same time, the system extracts amplitude characteristics from the high-frequency oscillation components, such as the absolute peak value, root mean square value, or sliding peak value as the target poisoning index. If, in the aforementioned example, the high-frequency amplitude of the fifth window reaches 4.4, which is significantly higher than the normal range, the system determines that the oxidation degree of the target surface has increased at this time, and the risk of poisoning due to excessive oxygen should be taken seriously.

[0130] In the anomaly handling mechanism, if the sampling window is less than the first preset duration, resulting in insufficient frequency domain separation accuracy, the system can automatically extend the analysis window by one, expanding the original 8 points to 16 points before performing separation. If the window is greater than the second preset duration, which would reduce real-time performance, the controller will prioritize retaining the most recent data and use an overlapping window method to balance resolution and response speed. If the high-frequency oscillation component is close to zero overall, it indicates that the current process is relatively stable, and the target poisoning index can be recorded as a low-risk state without forcibly triggering any gas flow correction. If there is a single-point gap in the comprehensive impedance change rate sequence within a certain window due to noise or missing measurements, the controller can first use the mean of adjacent windows or linear interpolation to fill the gap before performing frequency domain separation, and mark the result as a reduced confidence output to avoid the direct introduction of pseudo-high-frequency components from missing measurement points.

[0131] During the aforementioned ITO coating process for automotive display glass, when the equipment was running in the middle to later stages, the system observed that the resonant frequency was still blue-shifted, but the frequency curve showed obvious small oscillations. After frequency domain separation, the thickness corresponding to the low-frequency component smoothly increased from 95 nanometers to 110 nanometers, while the high-frequency component suddenly amplified at a certain time. The system did not identify this as an abnormal increase in deposition rate, but rather as signs of excessive oxidation appearing on the target surface, thus shifting the monitoring focus to oxygen supply regulation. The purpose of this step is to distinguish between thickness information and target surface state information in the same impedance change, thereby achieving synchronous output of geometric thickness and target poisoning index, providing a more targeted basis for subsequent closed-loop regulation.

[0132] When the film's geometric thickness is less than the preset target thickness, the oxygen flow rate of the mass flow meter is adjusted according to the target poisoning index, including:

[0133] If the poisoning index of the target material exceeds the preset poisoning safety threshold, it is determined that the target surface has undergone excessive oxidation.

[0134] In response to excessive oxidation of the target surface, a reduction command is generated and sent to the mass flow meter to reduce the oxygen flow rate;

[0135] If the poisoning index of the target material is less than or equal to the poisoning safety threshold, the target surface condition is determined to be normal, and the current oxygen flow rate is maintained.

[0136] This embodiment provides a mechanism for adjusting oxygen flow rate based on the target poisoning index before the target thickness is reached. Specifically, when the coating has not reached its endpoint, the primary task of the system is not to immediately reduce power, but to ensure that the target surface is in an oxidation equilibrium zone suitable for stable sputtering. If a fixed oxygen flow rate is maintained continuously during this stage, there is a high probability that excessive oxidation of the target surface will occur as the target material is consumed and the cavity state changes, leading to a sharp drop in deposition rate and drift in the electrical properties of the film. Therefore, this embodiment uses the target poisoning index as a direct basis for adjusting the gas flow rate during the process.

[0137] Specifically, the controller first compares the target poisoning index with the poisoning safety threshold. For example, the poisoning safety threshold is set to 0.70. If the calculated poisoning index is 0.76 and the geometric thickness is only 92 nanometers, which is less than the target's 120 nanometers, the system determines that the target surface has undergone excessive oxidation and generates a reduction command to send to the mass flow meter, reducing the oxygen flow rate from 10 standard cubic centimeters per minute to 9 standard cubic centimeters per minute. If the poisoning index is still higher than the threshold in the next window, for example, 0.74, it continues to be slightly reduced to 8.5 standard cubic centimeters per minute.

[0138] Conversely, if the current poisoning index is 0.58, the target surface is considered to be in normal condition, and the oxygen flow rate is kept constant to allow the system to continue to steadily increase in thickness. The control method here can be stepped or proportional. The stepped method is suitable for scenarios where the response time of the device actuator is greater than the preset response threshold. The proportional method is more suitable for production lines that require a preset high level of precision in flow rate fine-tuning. Regardless of which method is used, the core is to prioritize suppressing poisoning by increasing the oxygen flow rate before the film thickness reaches the target, rather than prematurely performing endpoint treatment on the target power.

[0139] In the anomaly handling mechanism, if the poisoning index occasionally exceeds the threshold once, but then falls back in the next window, the system can set a continuous threshold confirmation mechanism to avoid over-adjusting short-term noise; if the oxygen flow rate has dropped to the lower limit allowed by the process, but the poisoning index continues to rise, it indicates that the problem may not only come from the oxygen supply, but may also be related to the state of the target material or contamination of the cavity. At this time, the system can maintain the current lower limit flow rate and issue a process warning, and the host computer will record this batch as an abnormal batch.

[0140] At the 7th minute of the automotive display glass production line, the system reflected a geometric thickness of 101 nanometers, but the high-frequency oscillation amplitude increased significantly, and the poisoning index reached 0.79. If the original oxygen flow rate were maintained at this point, the deposition rate would further decrease in the following minutes, resulting in poor thickness consistency among the 24 glass pieces in the same batch. Therefore, the controller automatically sent a flow reduction command to the mass flow meter to reduce the oxygen flow rate in two steps. After the adjustment, the poisoning index dropped back to about 0.63 in three windows, and the thickness growth returned to a stable state. The purpose of this step is to suppress excessive oxidation of the target surface in time before the film layer reaches the target thickness, thereby achieving stable deposition rate and maintaining the continuity of the intermediate process.

[0141] When the film geometric thickness is greater than or equal to the target thickness, the sputtering target power of the sputtering DC power supply and the oxygen flow rate of the mass flow meter are finely adjusted based on the derivative of the drift slope, including:

[0142] Calculate the derivative of the drift slope over time to generate the impedance gradient derivative; obtain the preset power adjustment ratio coefficient and flow adjustment ratio coefficient, which are preset based on the target thickness margin, the current poisoning index and the preset historical batch thickness fluctuation variance threshold; multiply the impedance gradient derivative with the power adjustment ratio coefficient and flow adjustment ratio coefficient respectively to generate the power decay curve and flow decay curve.

[0143] The sputtering target power is gradually reduced according to the power decay curve.

[0144] The oxygen flow rate is gradually reduced according to the flow rate decay curve until the sputtering target power drops to zero.

[0145] This embodiment provides a mechanism for progressive convergence control after the target thickness is reached. Specifically, if the deposition is terminated by immediately cutting off the power supply after the film thickness reaches the target value, the end film thickness often cannot be accurately controlled due to discharge afterglow, particle retention and cavity response hysteresis. This is especially true in the preparation of large-area ITO, where differences between the edge and center regions are more likely to occur.

[0146] Therefore, this embodiment does not directly shut down the machine at the end stage. Instead, it generates an impedance gradient derivative based on the derivative of the drift slope and synchronously reduces the power and oxygen flow rate accordingly. Specifically, when the geometric thickness is greater than or equal to the target thickness, the controller continuously calculates the rate of change of the drift slope over time. To illustrate with a set of exemplary data, if the drift slopes at three consecutive moments are 0.50, 0.35, and 0.20 MHz / min respectively, it can be determined that the growth trend is slowing down and the derivative is negative. The system multiplies the derivative by the preset power adjustment ratio coefficient and the flow rate adjustment ratio coefficient respectively to generate two decay curves.

[0147] For example, if the power adjustment ratio coefficient is higher than the first preset value, the target power decreases at a greater rate; if the power adjustment ratio coefficient is lower than or equal to the first preset value, the target power decreases at a preset basic attenuation rate. If the flow rate adjustment ratio coefficient is lower than the second preset value, the oxygen flow rate decreases at a lower rate than the preset attenuation rate; if the flow rate adjustment ratio coefficient is higher than or equal to the second preset value, the oxygen flow rate decreases at a preset basic attenuation rate, thereby maintaining the stability of the final thin film stoichiometry.

[0148] Furthermore, the power adjustment ratio and flow rate adjustment ratio are set based on the target thickness margin, the current poisoning index, and the pre-set historical batch thickness fluctuation variance threshold, and remain unchanged within a batch, or are dynamically switched in the controller according to a preset segmentation rule; the controller adjusts the sputtering DC power supply output according to the power attenuation curve; for example, if the original power is 4 kW, it does not immediately drop to zero after the target thickness is reached, but gradually decreases in the order of 4 kW, 3.2 kW, 2.4 kW, 1.5 kW, 0.8 kW, and 0.

[0149] Meanwhile, the oxygen flow rate decreased synchronously according to curves of 8, 7.2, 6.5, 5.8, 5.2, and 5 standard cubic centimeters per minute; the combination of the two can avoid plasma abrupt changes caused by rapid power reduction alone, and also avoid changes in the composition at the end of the membrane layer due to excessively rapid flow reduction alone.

[0150] Additionally, if the impedance gradient derivative suddenly turns positive during the asymptotic convergence control phase at the endpoint, it indicates that the system may be subject to external disturbances or abnormal discharges. In this case, the controller can freeze the current decay step size and maintain the power and flow rate of the previous moment for 1 to 2 preset time windows. After the derivative returns to normal, the decay will continue. If the power is close to zero but the resonant frequency is still changing significantly, it indicates that there may be residual deposition effects in the cavity. The system can extend the monitoring period before termination until the signal stabilizes before completely ending the batch.

[0151] In this batch of automotive display glass, when the system determined that the thickness reached 120 nanometers, it did not immediately shut down the target power supply. Instead, it generated a 6-step decay curve based on the impedance gradient derivatives fitted from the most recent 4 windows. As the power and oxygen flow decreased simultaneously, the film thickness smoothly increased from 120 nanometers to 122 nanometers and then stabilized, without the end over-coating commonly seen during traditional emergency stops. As a result, the sheet resistance dispersion of the 24 glass pieces in the batch was also reduced.

[0152] The purpose of this step is to guide the synergistic decay of power and oxygen flow through the impedance gradient derivative, thereby achieving smooth convergence at the coating endpoint and improving the thickness consistency and film quality stability of each stage.

[0153] Example 2:

[0154] The method also includes:

[0155] The amplitude of the radio frequency reflection coefficient of the radio frequency sweep frequency perturbation signal reflected by the substrate tray is acquired;

[0156] The RF reflection coefficient amplitude is input into a preset sheet resistance mapping model that characterizes the relationship between the reflection coefficient and the sheet resistance, and the in-situ sheet resistance at the current moment is calculated.

[0157] If the in-situ sheet resistance is higher than the preset poor conductivity threshold, an alarm message indicating abnormal conductivity will be generated.

[0158] This embodiment provides a mechanism for adding in-situ sheet resistance monitoring on the basis of thickness monitoring; specifically, knowing only that the ITO film has reached the target thickness does not guarantee that its conductivity is qualified; under some working conditions, although the geometric thickness of the film layer meets the requirements, the actual sheet resistance is still too high due to oxygen partial pressure deviation or poor crystallization state at the end, forming a failure mode with thickness but poor conductivity.

[0159] Therefore, this embodiment further utilizes the RF reflection coefficient amplitude to establish a sheet resistance mapping, thereby enabling the synchronous acquisition of in-situ sheet resistance in the same measurement link. Specifically, while acquiring phase difference information, the controller also records the RF reflection coefficient amplitude within the corresponding frequency band. After pre-calibration, a mapping model between the reflection coefficient and the sheet resistance can be established. The pre-calibration here is preferably completed under the same tray structure, substrate size range, film material, and typical process pressure conditions as actual production, in order to reduce the mapping offset caused by differences in equipment geometry.

[0160] For a specific example, if the sheet resistance is approximately 20 ohms / square when the reflection coefficient amplitude is 0.30 at a certain frequency band, 40 ohms / square when the amplitude is 0.45, and 80 ohms / square when the amplitude is 0.60, then the system can use interpolation to convert the current measured amplitude into the in-situ sheet resistance; if the current measured amplitude is 0.52, then the mapped sheet resistance is approximately 55 ohms / square.

[0161] It should be noted that the conductivity anomaly alarm message in this embodiment is an alarm name executed on the production line. This alarm name can cover conductivity anomalies caused by factors such as insufficient crystallization, abnormal oxygen content, insufficient local densification, or poor conductivity network connectivity, and is not limited to being directly caused by a single crystallographic reason. The purpose of this setting is to ensure that the alarm logic is consistent with the production line quality judgment. That is, as long as the online inversion result shows that the film conductivity deviates significantly from the target range, the anomaly prompting process is initiated. When the in-situ sheet resistance is higher than the preset conductivity failure threshold, the system generates a conductivity anomaly alarm message.

[0162] This threshold can be preset according to product requirements; if the vehicle display glass requires the final sheet resistance to be no higher than 35 ohms / square, but the system still reflects 55 ohms / square when the film thickness is close to the target, it indicates that the current film layer may exhibit poor conductivity due to insufficient crystallization or abnormal oxygen content, and the operating system or host computer needs to be alerted immediately.

[0163] In the anomaly handling mechanism, if the reflection coefficient amplitude falls within the model interpolation range, such as below the minimum calibration value or above the maximum calibration value, the system can prioritize marking the sheet resistance result as exceeding the calibration range instead of directly outputting the precise value; at the same time, the alarm logic is retained, that is, when it shows a trend of change above the preset poor conductivity threshold, the anomaly can still be indicated; if the thickness result is normal but the sheet resistance result is abnormal at the same time, the system defaults to prioritizing the possibility of both existing, rather than simply judging one result as wrong, because this situation is one of the process defects that this embodiment needs to identify; if the difference in substrate size between different batches exceeds the preset tolerance, resulting in a system offset in the absolute sheet resistance corresponding to the same reflection coefficient amplitude, the controller can call the calibration sub-model corresponding to the product specifications of that batch, or perform normalization by the area correction factor before performing the mapping;

[0164] In the final stage of the aforementioned batch coating, the system detected that the geometric thickness of a certain time window had reached 119 nanometers, but the in-situ sheet resistance calculated from the corresponding RF reflection coefficient amplitude was still higher than the preset threshold; the host computer therefore received an alarm message indicating abnormal conductivity; based on this, the process engineer judged that although the batch was close to the target thickness, the quality of the conductive network might be insufficient, and the film preparation could not be judged to be up to standard simply because the thickness was qualified.

[0165] The purpose of this step is to obtain the film conductivity index simultaneously in addition to in-situ thickness monitoring, so as to realize online judgment of the dual-dimensional quality status of ITO film thickness and electrical performance.

[0166] The method also includes:

[0167] If an alarm message indicating abnormal conductivity is generated, an interrupt command is sent to the sputtering DC power supply to terminate the coating process.

[0168] If the in-situ sheet resistance is lower than or equal to the poor conductivity threshold, continue to perform coating control.

[0169] This embodiment provides a termination protection mechanism based on in-situ sheet resistance anomaly. Specifically, the aforementioned alarm mechanism can alert to conductivity risks, but if only an alert is issued without taking any action, deposition may continue on a continuous automated production line, resulting in the entire batch of glass not improving its electrical performance while its thickness increases, causing the overall electrical performance of the substrate to fail. Therefore, this embodiment further sets up an abnormal interruption logic so that the system can actively terminate the coating process when it detects obvious poor conductivity.

[0170] Specifically, after generating an alarm message indicating abnormal conductivity, the controller sends an interrupt command to the sputtering DC power supply to terminate the current coating process. To illustrate with a simplified example, if the poor conductivity threshold is set to 35 ohms / square, and the system calculates 52 ohms / square and 57 ohms / square in two consecutive windows respectively, while the film thickness has continued to increase from 118 nanometers to 123 nanometers, it indicates that further thickening has not improved conductivity and may instead form a high-resistivity thick film. At this point, the controller executes an interrupt, reducing the target power to zero and stopping subsequent thickening operations.

[0171] Conversely, if the in-situ sheet resistance is lower than or equal to the threshold, the system continues to execute the existing coating control process. For example, in the final progressive convergence control stage, when the thickness reaches the target and the sheet resistance has dropped to 30 ohms / square, the controller does not trigger an interrupt, but continues to finish according to the decay curve, so that the film meets the requirements in terms of both thickness and conductivity.

[0172] To further explain, in order to avoid false alarms caused by noise from a single measurement, the generation of the conductivity abnormality alarm message in this embodiment can be based on a preset alarm criterion. For example, it can require that the current window result and the previous window result are consistent in trend, or require that the conductivity failure condition be met continuously within a short confirmation period before the alarm is officially generated. Once the alarm message is officially generated, an interrupt command is sent to the sputtering DC power supply according to the process of this embodiment, instead of waiting additionally after the alarm is generated. This preserves the fault tolerance capability of the automated production line for instantaneous noise and maintains the consistency of the process between alarm generation and execution interruption.

[0173] In the anomaly handling mechanism, if the interruption occurs in the early stage of the batch, the system can mark the current batch as a failed batch and save the oxygen flow rate at that time, which is convenient for subsequent investigation of the root cause of poor crystallization. If the interruption occurs at the end of the progressive convergence control of the coating, in addition to saving the alarm record, the system can also record the trend of the change of the in-situ sheet resistance in the most recent windows before the interruption, so as to distinguish between two types of anomalies: continuous deterioration type of poor conductivity and critical edge type of poor conductivity.

[0174] In an abnormal batch of the automotive display glass production line, the system detected that after the film thickness reached 121 nanometers, the in-situ sheet resistance of a certain batch of 24 pieces remained above 50 ohms / square and continuously met the conductivity failure condition within a short confirmation period. The controller then officially generated a conductivity abnormality alarm message and immediately issued an interrupt command to the sputtering DC power supply to end the coating of this batch in advance. At the same time, the alarm message, the RF reflection coefficient amplitude curve during the abnormal period, and the oxygen flow rate change record were uploaded to the production line management system. In another normal batch, when the sheet resistance stabilized below 32 ohms / square, the system continued to execute the predetermined end control until the power dropped steadily to zero.

[0175] The purpose of this step is to terminate the process in a timely manner when an abnormal state is detected where the thickness meets the standard but the conductivity is out of control, so as to avoid the continuous deterioration of the target surface due to invalid deposition and to ensure that the automated production line has executable protective actions for abnormal batches.

[0176] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for dynamically monitoring the thickness of an ITO coating, characterized in that, The method is applied to a vacuum coating system, which includes a substrate tray multiplexed as a radio frequency antenna, a radio frequency generator, an impedance matching network, a sputtering DC power supply, a mass flow meter, and a controller. The method is executed by the controller and includes: Obtain the preset target material parameters, process gas parameters, and substrate material parameters; Based on the target material parameters, the process gas parameters, and the substrate material parameters, an initial sheath equivalent circuit topology model is constructed, which includes a preset impedance film thickness dynamics equation characterizing the relationship between RF impedance and thin film growth and target state. Based on the radio frequency generator and the impedance matching network, a radio frequency sweep perturbation signal is injected into the base tray; Within a preset time window, the phase difference between the radio frequency voltage and current of the radio frequency sweep perturbation signal reflected by the base tray is continuously acquired; Based on the continuously acquired phase difference of the radio frequency voltage and current, the frequency corresponding to the zero phase difference is extracted, and this frequency is determined as the resonant peak frequency, and a time series of the resonant peak frequency is generated. Extract the drift slope of the time series of the resonant peak frequency as a function of time; Substituting the drift slope into the impedance film thickness dynamics equation, the current film geometric thickness and target poisoning index are calculated. If the geometric thickness of the thin film is less than the preset target thickness, the oxygen flow rate of the mass flow meter is adjusted according to the poisoning index of the target material. When the geometric thickness of the thin film is greater than or equal to the target thickness, the sputtering target power of the sputtering DC power supply and the oxygen flow rate of the mass flow meter are finely adjusted according to the derivative of the drift slope to complete the coating control.

2. The method for dynamic monitoring of ITO coating thickness according to claim 1, characterized in that, The step of constructing an initial sheath equivalent circuit topology model based on the target material parameters, the process gas parameters, and the substrate material parameters, including a preset impedance film thickness dynamics equation characterizing the relationship between RF impedance and thin film growth and target state, includes: The initial plasma density is determined based on the target material parameters and the process gas parameters. Based on the substrate material parameters, determine the initial substrate surface dielectric constant; By combining the initial plasma density and the initial substrate surface dielectric constant, the initial sheath equivalent circuit topology model is generated; The initial sheath equivalent circuit topology model includes thin film conductivity variables and sheath capacitance variables.

3. The method for dynamic monitoring of ITO coating thickness according to claim 1, characterized in that, The step of injecting a radio frequency sweep perturbation signal into the substrate tray based on the radio frequency generator and the impedance matching network includes: Based on the radio frequency generator, a wideband radio frequency signal within a preset radio frequency band range is generated; The impedance matching network is used to perform impedance matching on the wideband radio frequency signal to generate the radio frequency sweep perturbation signal. The radio frequency sweep perturbation signal is injected into the base tray at a preset power.

4. The method for dynamic monitoring of ITO coating thickness according to claim 1, characterized in that, The extraction of the drift slope of the time series of the resonant peak frequency over time includes: Extract the current resonant peak frequency at the current moment and the initial resonant peak frequency at the initial moment from the time series of the resonant peak frequencies; Calculate the frequency shift of the current resonant peak frequency compared to the initial resonant peak frequency; If the frequency shift is greater than a preset positive frequency shift threshold, the film is determined to have entered the continuous conductive film growth stage. If the frequency shift is less than or equal to the positive frequency shift threshold, the film is determined to be in the isolated island growth stage. If the film is determined to be in the isolated island growth stage, a thickness monitoring waiting prompt is output, and the process returns to the step of continuously acquiring the RF voltage and current phase difference of the RF sweep frequency perturbation signal reflected by the substrate tray within a preset time window. When it is determined that the thin film has entered the continuous conductive film growth stage, the drift slope is generated by linear fitting based on the time series of the resonant peak frequency.

5. The method for dynamic monitoring of ITO coating thickness according to claim 1, characterized in that, The step of substituting the drift slope into the impedance film thickness dynamics equation to calculate the current film geometric thickness and target poisoning index includes: The drift slope is input into the impedance film thickness dynamic equation to calculate the overall impedance change rate; Based on the Fast Fourier Transform algorithm, the comprehensive impedance change rate is subjected to frequency domain separation processing to generate low-frequency thickness-related components and high-frequency oscillation components. A preset thickness conversion coefficient is obtained. The preset thickness conversion coefficient is a proportionality constant extracted by pre-collecting low-frequency thickness-related components under the same substrate and target material conditions from multiple historical calibration batches and performing linear regression analysis on the corresponding thin film geometric thickness data. The low-frequency thickness-related components are multiplied by the preset thickness conversion coefficient to calculate the thin film geometric thickness. The amplitude characteristics of the high-frequency oscillation component, including its absolute peak value, root mean square value, or sliding peak-to-peak value, are extracted as the target poisoning index.

6. The method for dynamic monitoring of ITO coating thickness according to claim 1, characterized in that, When the geometric thickness of the thin film is less than a preset target thickness, adjusting the oxygen flow rate of the mass flow meter according to the target poisoning index includes: If the poisoning index of the target material is greater than the preset poisoning safety threshold, it is determined that the target surface has undergone excessive oxidation. In response to excessive oxidation of the target surface, a reduction command is generated and sent to the mass flow meter to reduce the oxygen flow rate; If the poisoning index of the target material is less than or equal to the poisoning safety threshold, the target surface condition is determined to be normal, and the oxygen flow rate at the current moment is maintained.

7. The method for dynamic monitoring of ITO coating thickness according to claim 1, characterized in that, When the geometric thickness of the thin film is greater than or equal to the target thickness, fine-tuning the sputtering target power of the sputtering DC power supply and the oxygen flow rate of the mass flow meter based on the derivative of the drift slope includes: Calculate the derivative of the drift slope over time to generate the impedance gradient derivative; obtain preset power adjustment ratio coefficients and flow adjustment ratio coefficients, which are pre-tuned based on the target thickness margin, the current poisoning index, and a preset historical batch thickness fluctuation variance threshold; multiply the impedance gradient derivative by the power adjustment ratio coefficients and flow adjustment ratio coefficients respectively to generate power attenuation curves and flow attenuation curves; The sputtering target power is gradually reduced according to the power decay curve. The oxygen flow rate is gradually reduced according to the flow rate decay curve until the sputtering target power drops to zero.

8. The method for dynamic monitoring of ITO coating thickness according to claim 1, characterized in that, The method further includes: The amplitude of the radio frequency reflection coefficient of the radio frequency sweep perturbation signal reflected by the base tray is acquired; The amplitude of the radio frequency reflection coefficient is input into a preset sheet resistance mapping model that characterizes the relationship between the reflection coefficient and the sheet resistance, and the in-situ sheet resistance at the current moment is calculated. If the in-situ sheet resistance is higher than a preset poor conductivity threshold, an alarm message indicating abnormal conductivity is generated.

9. The method for dynamic monitoring of ITO coating thickness according to claim 8, characterized in that, The method further includes: If the abnormal conductivity alarm message is generated, an interrupt command is sent to the sputtering DC power supply to terminate the coating process. If the in-situ sheet resistance is lower than or equal to the poor conductivity threshold, the coating control continues.

Images