Arc detection method and apparatus for a radio frequency power supply load cavity

Abstract

The application discloses an arc detection method and device for a load cavity of a radio frequency power supply, in which a voltage signal and a current signal on the input side of the radio frequency power supply are acquired first, and multi-dimensional electrical characteristics including at least time domain characteristics, frequency domain characteristics and a power factor change rate are extracted from the voltage signal and the current signal, so that a fusion characteristic quantity is calculated by weighted summation of the multi-dimensional electrical characteristics, and whether an arc occurs is determined by comparing the fusion characteristic quantity with a preset threshold. Based on this, since the distortion of the electrical parameters caused by the change of the load impedance in the initial stage is directly acquired from the input side of the radio frequency power supply, the change of the characteristic quantity reflecting the arc occurrence can be detected in the arc formation stage of microsecond level, the timeliness of the arc detection is improved, meanwhile, the fusion calculation is combined with the multi-dimensional electrical characteristics, the specificity of the arc detection is improved, and then the accuracy of the arc detection is improved, so that the effect of reducing the damage risk of the semiconductor process equipment and the wafer and improving the yield is realized.

Description

Technical Field

[0001] This invention relates to the field of semiconductor manufacturing processes, and in particular to a method and apparatus for detecting electric arc in a radio frequency power supply load cavity. Background Technology

[0002] In semiconductor manufacturing processes, plasma processing equipment, such as etching machines and PECVD coating machines, is widely used in critical steps like wafer etching and thin film deposition. Radio frequency (RF) power supplies, as the energy source for generating plasma, transmit their output power to the process chamber via a matching network to maintain a stable plasma state. However, electric arcing is highly likely to occur within the process chamber. This arcing can lead to localized instantaneous high temperatures, material sputtering, particle contamination, and physical damage to the wafer surface, directly causing a decrease in chip yield and even scrapping entire batches of wafers, significantly increasing manufacturing costs.

[0003] One common method for arc detection in existing technologies is monitoring the reflected power at the output of an RF power supply. The basic principle is that when an arc occurs within the cavity, the load impedance changes abruptly, leading to impedance mismatch and a sharp increase in reflected power. The arc is identified by detecting whether the reflected power exceeds a threshold. However, a significant increase in reflected power usually occurs after the arc has fully formed and stabilized. From the microsecond-level arc breakdown to the stable detection of reflected power, multiple stages are involved, including arc development, transmission line reflection, coupler response, and jitter reduction by the control algorithm. During this delay, the arc may have already caused irreversible damage to the wafer. Therefore, this method cannot provide early warning, let alone detect the initial microsecond-level breakdown. Summary of the Invention

[0004] The purpose of this invention is to provide an arc detection method and apparatus for a radio frequency power supply load cavity. By directly obtaining the distortion of electrical parameters caused by the initial change in load impedance from the input side of the radio frequency power supply, the changes in characteristic quantities reflecting the occurrence of the arc can be detected at the microsecond level during the arc formation stage, thus improving the timeliness of arc detection. At the same time, by combining multi-dimensional electrical characteristics for fusion calculation, the specificity of arc detection is improved, thereby improving the accuracy of arc detection. This reduces the risk of damage to semiconductor process equipment and wafers and improves yield.

[0005] To address the aforementioned technical problems, this invention provides a method for detecting electric arc in a radio frequency power supply load cavity, comprising: Acquire the voltage and current signals from the RF power input side; Based on the voltage signal and the current signal, extract multidimensional electrical features including at least time-domain features, frequency-domain features, and power factor change rate; The fused feature quantity is obtained by weighted summation of the multidimensional electrical features; The fused feature quantity is compared with a preset threshold to determine whether an electric arc has occurred.

[0006] Preferably, the time-domain features include the rate of change of voltage and the rate of change of current; The process of extracting the time-domain features includes: Real-time monitoring of the voltage effective value drop depth in the voltage signal and the current peak value of the current signal; When the drop depth is greater than a preset drop depth threshold, the voltage change rate is calculated based on the voltage signal; When the peak current is greater than a preset peak current threshold, the rate of change of the current is calculated based on the current signal.

[0007] Preferably, the process of extracting the power factor change rate includes: Instantaneous active power and instantaneous reactive power are calculated based on the voltage signal and the current signal, and the power factor is calculated based on the instantaneous active power and the instantaneous reactive power. If the instantaneous active power decreases to less than a preset active power threshold, and the instantaneous reactive power increases to greater than a preset reactive power threshold, then the power factor change rate is calculated based on the power factor.

[0008] Preferably, the frequency domain features include total harmonic distortion, odd harmonic distortion, and even harmonic distortion; The process of determining the frequency domain features includes: In the absence of electric arc, the current signal is subjected to a fast Fourier transform to obtain the harmonic current values ​​of multiple harmonic orders within a specific frequency band, and a dynamic background noise reference is established. The dynamic background noise reference includes the reference harmonic current values ​​corresponding to each of the multiple harmonic orders. During real-time monitoring, the current signal is subjected to a fast Fourier transform to obtain the real-time harmonic current values ​​of multiple harmonic orders within the specific frequency band. Select at least one key harmonic order subset from the multiple harmonic orders, and compare each real-time harmonic current value corresponding to the key harmonic order subset with the reference harmonic current value of the corresponding order in the dynamic background noise reference by a preset multiple. When the proportion of times the real-time harmonic current value in the subset of key harmonics exceeds a preset multiple of the corresponding reference harmonic current value is greater than a preset proportion threshold, the total harmonic distortion, odd harmonic distortion, and even harmonic distortion are calculated. Wherein, the odd harmonic distortion is the result of taking the square root of the sum of the squares of the effective values ​​of the currents of all odd harmonics, dividing it by the effective value of the fundamental current, and then multiplying it by 100%; the even harmonic distortion is the result of taking the square root of the sum of the squares of the effective values ​​of the currents of all even harmonics, dividing it by the effective value of the fundamental current, and then multiplying it by 100%.

[0009] Preferably, after comparing the fused feature quantity with a preset threshold to determine whether an electric arc has occurred, the method further includes: If an electric arc is determined to occur, the fusion feature quantity is compared with multiple preset electric arc level thresholds to determine the level range into which the fusion feature quantity falls, and the electric arc level corresponding to the level range is determined as the electric arc level of the electric arc that occurred.

[0010] Preferably, after determining the arc level corresponding to the level range as the arc level of the occurring arc, the method further includes: Based on the determined arc level, perform the preset processing operation corresponding to the arc level.

[0011] Preferably, the fused feature quantity is obtained by weighted summation of the multidimensional electrical features, including: The formula for calculating the fusion feature is: ; Wherein, SCQ is the fused characteristic quantity, du / dt is the voltage change rate in the time domain characteristic, di / dt is the current change rate in the time domain characteristic, THD is the total harmonic distortion in the frequency domain characteristic, OHC is the odd harmonic distortion in the frequency domain characteristic, EHC is the even harmonic distortion in the frequency domain characteristic, and d(PF) / dt is the power factor change rate. This is the weighting coefficient for the voltage change rate. This is the weighting coefficient for the rate of change of the current. The weighting coefficients for the odd harmonic distortion are... The weighting coefficient for the even-order harmonic distortion is... The weighting factor for the total harmonic distortion is... This is the weighting coefficient for the rate of change of the power factor.

[0012] Preferably, before obtaining the fused feature quantity by weighted summation of the multidimensional electrical features, the method further includes: The time-domain feature, the frequency-domain feature, and the power factor change rate are compared with their respective preset limits; and when at least one of the features exceeds the corresponding preset limit, the process proceeds to the step of obtaining a fused feature quantity by weighted summation of the multidimensional electrical features.

[0013] To address the aforementioned technical problems, this invention provides an arc detection device for an RF power supply load cavity, comprising: Memory, used to store computer programs; A processor is configured to implement the steps of the arc detection method for the RF power supply load cavity as described above when executing a computer program.

[0014] Preferably, it also includes a voltage sensor, a current sensor, and an analog-to-digital converter module; The input terminal of the voltage sensor is connected to the input terminal of the radio frequency power supply to obtain the voltage signal on the input side of the radio frequency power supply. The input terminal of the current sensor is connected to the input terminal of the radio frequency power supply to obtain the current signal on the input side of the radio frequency power supply. The input terminal of the analog-to-digital converter module is connected to the output terminals of the voltage sensor and the current sensor, and the output terminal of the analog-to-digital converter module is connected to the processor, for converting the voltage signal and the current signal into analog signals and then transmitting them to the processor.

[0015] This application provides a method and apparatus for arc detection in a radio frequency (RF) power supply load cavity. The method first acquires voltage and current signals from the RF power supply input side. Multidimensional electrical features, including at least time-domain features, frequency-domain features, and power factor change rate, are extracted from these signals. A fused feature quantity is obtained by weighted summation of these multidimensional electrical features. The fused feature quantity is then compared with a preset threshold to determine whether an arc has occurred. Based on this, since the distortion of electrical parameters caused by the initial change in load impedance is directly obtained from the RF power supply input side, changes in feature quantities reflecting arc occurrence can be detected at the microsecond level during the arc formation stage, improving the timeliness of arc detection. Furthermore, the fusion calculation combining multidimensional electrical features enhances the specificity of arc detection, thereby improving its accuracy. This reduces the risk of damage to semiconductor process equipment and wafers, and improves yield. Attached Figure Description

[0016] To more clearly illustrate the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A schematic flowchart of an arc detection method for an RF power supply load cavity provided in this application; Figure 2 A schematic diagram of the structure of an arc detection system for a radio frequency power supply load cavity provided in this application; Figure 3 A schematic diagram of the structure of an arc detection device for a radio frequency power supply load cavity provided in this application; Figure 4 A schematic diagram of the structure of the computer-readable storage medium provided in this application. Detailed Implementation

[0018] The core of this invention is to provide an arc detection method and device for a radio frequency power supply load cavity. By directly obtaining the distortion of electrical parameters caused by the initial change of load impedance from the input side of the radio frequency power supply, the change of characteristic quantities reflecting the occurrence of arc can be detected at the microsecond level during the arc formation stage, which improves the timeliness of arc detection. At the same time, by combining multi-dimensional electrical features for fusion calculation, the specificity of arc detection is improved, thereby improving the accuracy of arc detection. This reduces the risk of damage to semiconductor process equipment and wafers and improves yield.

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please refer to Figure 1 , Figure 1 A flowchart illustrating an arc detection method for an RF power supply load cavity provided in this application, the method comprising: S11: Acquire the voltage and current signals from the RF power input side; In semiconductor manufacturing, radio frequency (RF) power supplies are the core equipment for generating plasma. Electric arcing within the process chamber is a major factor leading to decreased wafer yield and equipment damage. Existing arc detection technologies largely rely on monitoring changes in the chamber's optical signal or abrupt changes in reflected power at the output.

[0021] However, traditional reflected power monitoring methods can only detect arcs in the stage where a stable arc has already formed, resulting in a long detection delay of over 100ms, making it impossible to capture the microsecond-level initial stage of an arc. In contrast, arc monitoring schemes using optical sensors suffer from high false alarm rates due to contamination of the cavity window and interference from plasma luminescence. Furthermore, single-parameter threshold methods for a single electrical variable, such as relying solely on current mutations or total harmonic distortion for exceeding limits, lack specificity under fluctuating load conditions and struggle to distinguish between micro-arcs and normal impedance modulation. Therefore, there is an urgent need for a detection method that can respond quickly and accurately identify arcs.

[0022] In this embodiment, instantaneous voltage and current signals from the RF power input side or DC bus side can be simultaneously acquired using high-precision voltage and current sensors, such as high-voltage differential probes, Rogowski coils, or Hall effect sensors. Unlike traditional methods that detect reflected power at the RF power output side, this application moves the detection point forward to the RF power input. Based on this, since the transient change in load impedance during an arc is immediately reverse-propagated to the power input side, it causes early distortion of characteristic quantities, including voltage and current signals, such as voltage dips and current spikes. This electrical distortion occurs synchronously with arc breakdown, avoiding the delay caused by the tuning of the matching network on the RF power output side. This allows for the capture of initial arc characteristics at the microsecond or even nanosecond level, enabling immediate wafer protection.

[0023] S12: Based on voltage and current signals, extract multidimensional electrical features including at least time-domain features, frequency-domain features, and power factor change rate; After obtaining the voltage and current signals, feature extraction is performed on the signals from multiple dimensions, namely, extracting multi-dimensional electrical features including at least time-domain features, frequency-domain features, and power factor change rate. Time-domain features reflect the instantaneous changes in the waveforms of the voltage and current signals over time, such as voltage amplitude fluctuations and current pulse patterns. Frequency-domain features are obtained through spectrum analysis, such as Fast Fourier Transform, to obtain the magnitude distribution of the current signal at different frequency components, reflecting harmonic and noise levels. The power factor change rate reflects the time-differential characteristic of the ratio of active power to apparent power, characterizing the dynamic fluctuations in energy conversion efficiency. Based on this, considering that single-dimensional features are easily affected by normal process modulation, such as impedance matching and power step interference, leading to false alarms or missed alarms, this embodiment integrates electrical features from the time, frequency, and power domains for arc detection, significantly improving the sensitivity and specificity of arc detection and effectively distinguishing whether an arc, such as a micro-arc, has occurred, or whether a normal impedance tuning process is underway.

[0024] S13: The fused feature quantity is obtained by weighted summation of multidimensional electrical features; The extracted time-domain features, frequency-domain features, and power factor change rate are linearly weighted and summed according to preset weighting coefficients to obtain a comprehensive scalar value, namely the fused feature quantity. The weighting coefficients can be determined after system initialization or process formulation. Their values ​​can be preset based on historical experience, experimental calibration, or typical arc waveform characteristics to ensure that the fused feature quantity remains stable at a low level under normal operating conditions, while increasing significantly when an arc occurs.

[0025] However, to further improve the adaptability of the detection method to different process conditions, different cavity states, and equipment aging after long-term operation, an online learning module can be embedded to monitor the accuracy and reliability of arc detection results in real time. For example, it can compare the degree of matching between actual arc events and alarm records, and adaptively fine-tune the weighting coefficients accordingly. For instance, when frequent false alarms are detected, the weighting coefficients of voltage or current change rates in the time domain characteristics can be appropriately increased; when the false alarm rate increases, the weighting coefficients of frequency domain characteristics or power factor change rates can be relatively increased. Based on this, the calculation of fused feature quantities can continuously maintain fault sensitivity and specificity without frequent manual intervention.

[0026] S14: Compare the fused feature quantity with a preset threshold to determine whether an electric arc has occurred.

[0027] The fused feature value is compared with a preset threshold in real time. When the fused feature value exceeds the preset threshold, it is determined that an electric arc has occurred; if it does not exceed the threshold, it is determined that there is no electric arc. The preset threshold can be dynamically set through experimental calibration, historical data statistics, or online learning, etc., and this application does not limit it in this way.

[0028] In summary, by directly obtaining the distortion of electrical parameters caused by the initial change in load impedance from the RF power input side, characteristic quantities reflecting the occurrence of the arc can be detected at the initial stage of arc formation in microseconds, improving the timeliness of arc detection. At the same time, by combining multi-dimensional electrical characteristics for fusion calculation, the specificity of arc detection is improved, thereby improving the accuracy of arc detection. This reduces the risk of damage to semiconductor process equipment and wafers and improves yield.

[0029] Based on the above embodiments: As a preferred embodiment, the time-domain features include the rate of change of voltage and the rate of change of current; The process of extracting time-domain features includes: Real-time monitoring of the voltage effective value drop depth in the voltage signal and the current peak value in the current signal; When the drop depth exceeds a preset drop depth threshold, the voltage change rate is calculated based on the voltage signal. When the peak current exceeds the preset peak current threshold, the rate of change of current is calculated based on the current signal.

[0030] In this embodiment, the time-domain features may include, but are not limited to, the rate of change of voltage and the rate of change of current.

[0031] Specifically, firstly, the voltage RMS value drop depth and current peak value of the current signal are monitored in real time. The voltage RMS value drop depth refers to the decrease in the voltage RMS value relative to its normal operating reference within a continuous time window; while the current peak value refers to the instantaneous maximum value of the current detected in each RF cycle or sliding window. This monitoring process uses a high-speed sliding window algorithm to detect amplitude changes in the voltage and current signals with microsecond-level time resolution.

[0032] Based on this, when the detected drop depth exceeds a preset drop depth threshold, the calculation of the voltage change rate is triggered. Specifically, the voltage change rate can be calculated by recording the time interval from the start of the voltage drop until it reaches the maximum drop depth and then begins to rise again, and combining this with the maximum drop depth. Correspondingly, when the peak current exceeds a preset peak current threshold, the calculation of the current change rate is triggered. The current change rate is obtained by detecting the ratio of the rise amplitude of the current spike to its rise time.

[0033] As a preferred embodiment, the process of extracting the rate of change of power factor includes: Instantaneous active power and instantaneous reactive power are calculated based on voltage and current signals, and the power factor is calculated based on instantaneous active power and instantaneous reactive power. If the instantaneous active power decreases to less than the preset active power threshold and the instantaneous reactive power increases to more than the preset reactive power threshold, then the power factor change rate is calculated based on the power factor.

[0034] In this embodiment, instantaneous active power and instantaneous reactive power are first calculated based on the acquired voltage and current signals. Instantaneous active power is obtained by multiplying the instantaneous voltage and current values ​​at the same moment and averaging the results over one radio frequency cycle or a sliding window to determine the actual energy consumption rate of the load. Instantaneous reactive power reflects the energy exchanged between the power source and the load, and can be obtained by shifting the current signal by 90° and multiplying it by the voltage, and then averaging the results, or by summing the reactive components of the fundamental frequency and its harmonics.

[0035] Secondly, the power factor is calculated based on instantaneous active power and instantaneous reactive power. In AC circuits, the power factor PF is the ratio of active power to apparent power, i.e. PF represents active power and Q represents reactive power. PF ranges from 0 to 1, reflecting the degree to which electrical energy is effectively utilized.

[0036] Then, the instantaneous active power change trend is monitored in real time. When the active power decreases to below a preset active power threshold, and the instantaneous reactive power increases to above a preset reactive power threshold, the calculation of the power factor change rate is triggered. The preset active power threshold and preset reactive power threshold can be, but are not limited to, preset based on the statistical distribution of the power factor under normal process conditions. Based on this, since the load impedance changes abruptly when an electric arc occurs, and the active power decreases rapidly due to the bypass effect of the arc channel, while the reactive power surges instantaneously due to energy rebound caused by impedance mismatch, the simultaneous fulfillment of both conditions can effectively distinguish between power factor fluctuations caused by electric arcs and slow changes caused by normal power steps or matching network tuning.

[0037] Finally, at the moment when both of the above conditions are satisfied, the power factor is differentiated over time, i.e., calculated. d(PF) / dt is the rate of change of power factor. for Power factor at time t, for The power factor at time n, where n is a positive integer. This is the time difference used to calculate the rate of change of the power factor. The rate of change of the power factor directly reflects the degree of change in the power factor per unit time.

[0038] As a preferred embodiment, the frequency domain characteristics include total harmonic distortion, odd harmonic distortion, and even harmonic distortion; The process of determining frequency domain characteristics includes: In the absence of electric arc, a fast Fourier transform is performed on the current signal to obtain the harmonic current values ​​of multiple harmonic orders within a specific frequency band, and a dynamic background noise reference is established. The dynamic background noise reference includes the reference harmonic current values ​​corresponding to each of the multiple harmonic orders. During real-time monitoring, a fast Fourier transform is performed on the current signal to obtain the real-time harmonic current values ​​of multiple harmonic orders within a specific frequency band. Select at least one key harmonic order subset from multiple harmonic orders, and compare each real-time harmonic current value corresponding to the key harmonic order subset in the real-time harmonic current value with the reference harmonic current value of the corresponding order in the dynamic background noise reference by their respective preset multiples. When the proportion of times the real-time harmonic current value in the key harmonic order set exceeds a preset multiple of the corresponding reference harmonic current value is greater than a preset proportion threshold, the total harmonic distortion, odd harmonic distortion, and even harmonic distortion are calculated. Among them, odd harmonic distortion is the result of taking the square root of the sum of the squares of the effective values ​​of all odd harmonic currents, dividing it by the effective value of the fundamental current, and then multiplying it by 100%; even harmonic distortion is the result of taking the square root of the sum of the squares of the effective values ​​of all even harmonic currents, dividing it by the effective value of the fundamental current, and then multiplying it by 100%.

[0039] In this embodiment, the frequency domain characteristics may include, but are not limited to, total harmonic distortion, odd harmonic distortion, and even harmonic distortion.

[0040] First, when the process chamber is in a normal, arc-free operating state, such as in the initial stage of stable plasma ignition, a Fast Fourier Transform (FFT) is performed on the acquired current signal to obtain its spectral distribution. Based on this spectrum, for a pre-selected specific frequency band, such as a high-frequency noise band that may contain arc characteristics, this band can be pre-calibrated experimentally based on the fundamental frequency of the RF power supply, the switching frequency, and the chamber structure to obtain the harmonic current values ​​for multiple harmonic orders within the specific frequency band, such as the harmonic current values ​​for the 2nd to 39th harmonics. These harmonic current values ​​are used as the reference harmonic current values ​​for that harmonic in the arc-free state, and the set of all reference harmonic current values ​​constitutes the dynamic background noise reference.

[0041] During real-time monitoring, a sliding window Fast Fourier Transform is performed on the continuously acquired current signals to continuously calculate the real-time harmonic current values ​​within the same specific frequency band. The real-time harmonic current values ​​reflect the real-time harmonic current values ​​of the same harmonic order within that specific frequency band at the current moment. To reduce computational load and improve detection specificity, a subset of key harmonic orders can be selected from the multiple harmonic orders, such as odd harmonics sensitive to electric arcs, such as the 3rd, 5th, 7th, and 9th orders. For each harmonic order in the subset of key harmonic orders, its real-time harmonic current value is compared with the reference harmonic current value of the corresponding order in the dynamic background noise reference by a preset multiple. The preset multiple can be set independently according to different orders; for example, the multiple is smaller for lower-order harmonics and larger for higher-order harmonics.

[0042] Based on this, the number of times the real-time harmonic current value in the critical harmonic number subset exceeds the corresponding reference harmonic current value by a preset multiple is counted, and the proportion of this exceeding number to the total number of critical harmonic number subsets is calculated. When this proportion is greater than a preset proportion threshold, such as 60%, a significant anomaly in the frequency domain is determined, which then triggers the calculation of total harmonic distortion, odd harmonic distortion, and even harmonic distortion.

[0043] In one specific embodiment, harmonics from the 2nd to the 39th order are selected as the monitoring range, with the key harmonic order subset consisting of the 3rd, 5th, 7th, and 9th orders (4 in total). For the 3rd harmonic, the preset multiple is 2.5; for the 5th harmonic, the preset multiple is 2.0; for the 7th harmonic, the preset multiple is 1.8; and for the 9th harmonic, the preset multiple is 1.5. That is, different orders correspond to different preset multiples. The preset ratio threshold is 60%, meaning that if at least 3 of the 4 key harmonics have real-time values ​​exceeding their respective reference values ​​by a multiple, subsequent calculations are triggered.

[0044] Specifically, total harmonic distortion (THD) is the percentage of the square root of the sum of the squares of the effective values ​​of all harmonic components to the effective value of the fundamental frequency, reflecting the severity of the overall waveform distortion. Odd-order harmonic distortion is the square root of the sum of the squares of the effective values ​​of all odd-order harmonic currents, divided by the effective value of the fundamental frequency current, and then multiplied by 100%. Even-order harmonic distortion is the square root of the sum of the squares of the effective values ​​of all even-order harmonic currents, divided by the effective value of the fundamental frequency current, and then multiplied by 100%.

[0045] Specifically, THD represents total harmonic distortion, and h represents the harmonic order, which can be 2, 3, 4, ..., H. H can be set according to actual needs. This is the effective value of the fundamental current. This represents the effective value of the current for the h-th harmonic.

[0046] h takes values ​​of 1, 2, 3, ..., H, and EHC represents even-order harmonic distortion.

[0047] h takes values ​​of 1, 2, 3, ..., H, and OHC represents odd harmonic distortion.

[0048] As a preferred embodiment, after comparing the fused feature quantity with a preset threshold to determine whether an electric arc has occurred, the method further includes: If an electric arc is determined to have occurred, the fused feature quantity is compared with multiple preset electric arc level thresholds to determine the level range into which the fused feature quantity falls, and the electric arc level corresponding to the level range is determined as the electric arc level of the electric arc that occurred.

[0049] In a preferred embodiment, after determining that an electric arc has occurred based on a comparison of the fused feature quantity with a preset threshold, the severity of the electric arc can be further graded to perform differentiated subsequent processing. Specifically, when it is determined that an electric arc has occurred, the calculated fused feature quantity is compared with multiple preset electric arc level thresholds. Each preset arc level threshold divides the numerical range of the fused feature quantity into several consecutive level intervals. For example, the first preset arc level threshold and the second preset arc level threshold constitute the first level interval, and the arc level within the first level interval is the first arc level. The second preset arc level threshold and the third preset arc level threshold constitute the second level interval, and the arc level within the second level interval is the second arc level. Subsequent preset arc level thresholds divide more level intervals in sequence, and each level interval corresponds to an arc level. It can be set that the fused feature quantity is less than the first preset arc threshold, the first preset arc threshold is less than the second preset arc threshold, and so on. When the fused feature quantity is less than the preset threshold, it is determined that no arc has occurred. When the fused feature quantity is greater than the first preset arc threshold but less than the second preset arc threshold, it is determined that an arc of the first arc level has occurred, and so on, to determine the arc level of the arc that has occurred.

[0050] Therefore, since there is a continuous change in the arc from a micro-arc to a hard arc, different arc levels cause varying degrees of damage to the wafer and equipment, and the corresponding countermeasures also differ. That is, after determining the arc level corresponding to this range as the arc level of the occurring arc, preset processing operations corresponding to the determined arc level can be executed.

[0051] For example, for low-energy micro-arcs, such as arc levels less than the Mth arc level, their development can be suppressed by fine-tuning power or impedance matching, avoiding frequent RF power shutdowns that could affect production efficiency. However, for high-energy hard arcs, such as arc levels greater than the Mth arc level, power output needs to be immediately locked and an alarm triggered to protect the wafer and equipment. M is a positive integer greater than 1.

[0052] It should be noted that the specific number of preset arc level thresholds and the arc type corresponding to each level range, such as micro-arc, medium-arc, and hard-arc, can be set according to actual process requirements and experimental calibration; this application does not impose any limitations on this. Furthermore, after determining the arc level, corresponding control strategies can be implemented based on different levels, such as adjusting power, adaptively adjusting impedance matching, and shutting off the RF power supply. Based on this, not only can arcs be detected quickly and accurately, but the arc level, i.e., the severity of arc generation, can also be determined, allowing for targeted treatment.

[0053] As a preferred embodiment, the fused feature quantity is obtained by weighted summation of multidimensional electrical features, including: The formula for calculating the fusion feature is: ; Wherein, SCQ is the fused characteristic quantity, du / dt is the voltage change rate in the time domain characteristic, u is voltage, t is time, di / dt is the current change rate in the time domain characteristic, i is current, THD is the total harmonic distortion in the frequency domain characteristic, OHC is the odd harmonic distortion in the frequency domain characteristic, EHC is the even harmonic distortion in the frequency domain characteristic, and d(PF) / dt is the power factor change rate. The weighting coefficient for the rate of change of voltage. The weighting coefficient for the rate of change of current. The weighting coefficients for odd harmonic distortion. The weighting coefficient for even-order harmonic distortion. This is the weighting factor for total harmonic distortion. This is the weighting coefficient for the rate of change of the power factor.

[0054] In this embodiment, the step of obtaining the fused feature quantity by weighted summation of multidimensional electrical features can be achieved by the above-mentioned linear weighted summation formula. This weighted summation formula merges six feature quantities with different dimensions and physical meanings into a unified scalar through weighting coefficients, so that the arc can be judged by a preset threshold, which simplifies the calculation process.

[0055] As a preferred embodiment, before obtaining the fused feature quantity by weighted summation of multidimensional electrical features, the method further includes: The time-domain features, frequency-domain features, and power factor change rate are compared with their respective preset limits; and when at least one of the features exceeds the corresponding preset limit, the process proceeds to the step of obtaining the fused feature quantity by weighted summation of the multidimensional electrical features.

[0056] In this embodiment, before calculating the fused feature quantity, the extracted time-domain features, frequency-domain features, and power factor change rate are compared with their respective preset limits, such as the upper limit of voltage change rate (du / dt)max, the upper limit of current change rate (di / dt)max, and the upper limit of power factor change rate (d(PF) / dt)max. These preset limits can be pre-set based on the statistical distribution of each feature quantity under normal process conditions, such as the mean plus a certain number of standard deviations, or fixed thresholds calibrated based on experience or experiments. The subsequent step of weighted summation to calculate the fused feature quantity is only performed if at least one feature exceeds its corresponding preset limit; if all features do not exceed their respective limits, the current electrical state is determined to be normal, and there is no need to calculate the fused feature quantity, and the monitoring of the data at the next moment continues directly.

[0057] Since the process chamber operates normally for the vast majority of the time, all electrical characteristics remain stable within normal fluctuation ranges. Performing weighted summation calculations indiscriminately each time would continuously consume processor resources. Introducing preset limit comparisons significantly reduces invalid fusion calculations, saving computing resources. Secondly, the complete fusion decision-making process is only initiated when at least one characteristic exceeds the limit, ensuring a rapid response to arcing while avoiding frequent false alarms triggered by minor fluctuations in normal operating conditions.

[0058] It should be noted that the arc detection scheme in this application cycles according to a time period, that is, it detects whether an arc occurs in real time, so as to provide timely protection when an arc occurs.

[0059] Please refer to Figure 2 , Figure 2 This is a schematic diagram of the structure of an arc detection system for a radio frequency power supply load cavity provided in this application. The device includes: Acquisition unit 21 is used to acquire the voltage signal and current signal on the RF power input side; Extraction unit 22 is used to extract multi-dimensional electrical features, including at least time-domain features, frequency-domain features, and power factor change rate, based on voltage and current signals. The calculation unit 23 is used to obtain the fused feature quantity by weighted summation of multidimensional electrical features; The determining unit 24 is used to compare the fused feature quantity with a preset threshold to determine whether an electric arc has occurred.

[0060] For a description of the arc detection system for the RF power supply load cavity provided by the present invention, please refer to the above method embodiments; the present invention will not be described again here.

[0061] Please refer to Figure 3 , Figure 3This is a schematic diagram of the structure of an arc detection device for a radio frequency power supply load cavity provided in this application. The device includes: Memory 31 is used to store computer programs; The processor 32 is configured to implement the steps of the arc detection method for the radio frequency power supply load cavity as described above when executing a computer program.

[0062] For a description of the arc detection device for the RF power supply load cavity provided by the present invention, please refer to the above method embodiments; the present invention will not be described again here.

[0063] In a preferred embodiment, the device further includes a voltage sensor, a current sensor, and an analog-to-digital converter module; The input terminal of the voltage sensor is connected to the input terminal of the radio frequency power supply to obtain the voltage signal on the input side of the radio frequency power supply. The input terminal of the current sensor is connected to the input terminal of the radio frequency power supply to acquire the current signal on the input side of the radio frequency power supply. The input terminal of the analog-to-digital converter module is connected to the output terminals of the voltage sensor and the current sensor, and the output terminal of the analog-to-digital converter module is connected to the processor 32. It is used to convert the voltage signal and the current signal into analog signals and then transmit them to the processor 32.

[0064] The arc detection device in this embodiment may include, but is not limited to, a voltage sensor, a current sensor, and an analog-to-digital converter (ADC). The input terminal of the voltage sensor is directly connected to the input side of the radio frequency (RF) power supply, such as a three-phase AC input or a DC bus, for real-time acquisition of the voltage signal from the RF power supply input side. The voltage sensor can be implemented using, but is not limited to, a high-voltage differential probe or a high-impedance voltage divider, with a bandwidth sufficient to capture high-frequency voltage distortion caused by the arc, such as at least 50MHz. The input terminal of the current sensor is also connected to the input side of the RF power supply for synchronous acquisition of the current signal from the input side. The current sensor can be implemented using, but is not limited to, a Rogowski coil or a Hall effect sensor. Rogowski coils have no magnetic saturation and a wide frequency response range, enabling the detection of rapidly changing current spikes. The input terminals of the ADC are connected to the output terminals of the voltage and current sensors, respectively, for converting the analog voltage and current signals into digital signals. The ADC includes two synchronous sampling channels with a resolution of at least 16 bits and a sampling rate of at least 100MS / s to ensure the reconstruction of microsecond-level transient signals. The output of the analog-to-digital converter module is connected to the processor 32, which may include, for example, a field-programmable gate array and a digital signal processor 32. The processor 32 receives digital voltage and current signals and performs the aforementioned time-frequency domain feature extraction, weighted fusion calculation and threshold comparison steps.

[0065] The field-programmable gate array in processor 32 can realize FFT and feature extraction, while the digital signal processor 32 performs the calculation of fused feature quantities and the comparison of preset thresholds, ensuring the real-time performance of detection.

[0066] Please refer to Figure 4 , Figure 4 This is a schematic diagram of the structure of the computer-readable storage medium provided in this application. The computer-readable storage medium 41 stores a computer program 42. When the computer program 42 is executed by the processor 32, it implements the steps of the arc detection method for the radio frequency power load cavity as described above.

[0067] For a description of the computer-readable storage medium provided by the present invention, please refer to the above method embodiments; the present invention will not be described again here.

[0068] It should also be noted that, in this specification, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0069] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for detecting electric arc in a radio frequency power supply load cavity, characterized in that, include: Acquire the voltage and current signals from the RF power input side; Based on the voltage signal and the current signal, extract multidimensional electrical features including at least time-domain features, frequency-domain features, and power factor change rate; The fused feature quantity is obtained by weighted summation of the multidimensional electrical features; The fused feature quantity is compared with a preset threshold to determine whether an electric arc has occurred.

2. The arc detection method for the RF power supply load cavity as described in claim 1, characterized in that, The time-domain features include the rate of change of voltage and the rate of change of current; The process of extracting the time-domain features includes: Real-time monitoring of the voltage effective value drop depth in the voltage signal and the current peak value of the current signal; When the drop depth is greater than a preset drop depth threshold, the voltage change rate is calculated based on the voltage signal; When the peak current is greater than a preset peak current threshold, the rate of change of the current is calculated based on the current signal.

3. The arc detection method for the RF power supply load cavity as described in claim 1, characterized in that, The process of extracting the power factor change rate includes: Instantaneous active power and instantaneous reactive power are calculated based on the voltage signal and the current signal, and the power factor is calculated based on the instantaneous active power and the instantaneous reactive power. If the instantaneous active power decreases to less than a preset active power threshold, and the instantaneous reactive power increases to greater than a preset reactive power threshold, then the power factor change rate is calculated based on the power factor.

4. The arc detection method for the RF power supply load cavity as described in claim 1, characterized in that, The frequency domain characteristics include total harmonic distortion, odd harmonic distortion, and even harmonic distortion; The process of determining the frequency domain features includes: In the absence of electric arc, the current signal is subjected to a fast Fourier transform to obtain the harmonic current values ​​of multiple harmonic orders within a specific frequency band, and a dynamic background noise reference is established. The dynamic background noise reference includes the reference harmonic current values ​​corresponding to each of the multiple harmonic orders. During real-time monitoring, the current signal is subjected to a fast Fourier transform to obtain the real-time harmonic current values ​​of multiple harmonic orders within the specific frequency band. Select at least one key harmonic order subset from the multiple harmonic orders, and compare each real-time harmonic current value corresponding to the key harmonic order subset with the reference harmonic current value of the corresponding order in the dynamic background noise reference by a preset multiple. When the proportion of times the real-time harmonic current value in the subset of key harmonics exceeds a preset multiple of the corresponding reference harmonic current value is greater than a preset proportion threshold, the total harmonic distortion, odd harmonic distortion, and even harmonic distortion are calculated. Wherein, the odd harmonic distortion is the result of taking the square root of the sum of the squares of the effective values ​​of the currents of all odd harmonics, dividing it by the effective value of the fundamental current, and then multiplying it by 100%; the even harmonic distortion is the result of taking the square root of the sum of the squares of the effective values ​​of the currents of all even harmonics, dividing it by the effective value of the fundamental current, and then multiplying it by 100%.

5. The arc detection method for the RF power supply load cavity as described in claim 1, characterized in that, After comparing the fused feature quantity with a preset threshold to determine whether an electric arc has occurred, the method further includes: If an electric arc is determined to occur, the fusion feature quantity is compared with multiple preset electric arc level thresholds to determine the level range into which the fusion feature quantity falls, and the electric arc level corresponding to the level range is determined as the electric arc level of the electric arc that occurred.

6. The arc detection method for the RF power supply load cavity as described in claim 5, characterized in that, After determining the arc level corresponding to the aforementioned level range as the arc level of the occurring arc, the method further includes: Based on the determined arc level, perform the preset processing operation corresponding to the arc level.

7. The arc detection method for the radio frequency power supply load cavity as described in any one of claims 1-6, characterized in that, The fused feature quantity is obtained by weighted summation of the multidimensional electrical features, including: The formula for calculating the fusion feature is: ； Wherein, SCQ is the fused characteristic quantity, du / dt is the voltage change rate in the time domain characteristic, di / dt is the current change rate in the time domain characteristic, THD is the total harmonic distortion in the frequency domain characteristic, OHC is the odd harmonic distortion in the frequency domain characteristic, EHC is the even harmonic distortion in the frequency domain characteristic, and d(PF) / dt is the power factor change rate. This is the weighting coefficient for the voltage change rate. This is the weighting coefficient for the rate of change of the current. The weighting coefficients for the odd harmonic distortion are... The weighting coefficient for the even-order harmonic distortion is... The weighting factor for the total harmonic distortion is... This is the weighting coefficient for the rate of change of the power factor.

8. The arc detection method for the radio frequency power supply load cavity as described in any one of claims 1-6, characterized in that, Before obtaining the fused feature quantity by weighted summation of the multidimensional electrical features, the method further includes: The time-domain feature, the frequency-domain feature, and the power factor change rate are compared with their respective preset limits; and when at least one of the features exceeds the corresponding preset limit, the process proceeds to the step of obtaining a fused feature quantity by weighted summation of the multidimensional electrical features.

9. An arc detection device for a radio frequency power supply load cavity, characterized in that, include: Memory, used to store computer programs; A processor, configured to, when executing a computer program, implement the steps of the arc detection method for a radio frequency power supply load cavity as described in any one of claims 1-8.

10. The arc detection device for the RF power supply load cavity as described in claim 9, characterized in that, It also includes a voltage sensor, a current sensor, and an analog-to-digital converter module; The input terminal of the voltage sensor is connected to the input terminal of the radio frequency power supply to obtain the voltage signal on the input side of the radio frequency power supply. The input terminal of the current sensor is connected to the input terminal of the radio frequency power supply to obtain the current signal on the input side of the radio frequency power supply. The input terminal of the analog-to-digital converter module is connected to the output terminals of the voltage sensor and the current sensor, and the output terminal of the analog-to-digital converter module is connected to the processor, for converting the voltage signal and the current signal into analog signals and then transmitting them to the processor.

Images