System for plasma state estimation based on multiple ultra-high frequency source reflections and method therefor

Abstract

The present invention relates to a plasma state estimation system and method based on multiple ultra-high frequency source reflected waves, and more specifically, to a technology for estimating the plasma state in real time using the phase and magnitude of reflected waves reflected from multiple ultra-high frequency sources. The present invention relates to a plasma state estimation system based on multiple ultra-high frequency source reflected waves, comprising: a reflected signal measuring unit (100) that measures the phase and magnitude of a reflected signal reflected through an antenna connected to each of a plurality of ultra-high frequency generators; a transfer function generating unit (200) that generates a transfer function for estimating the plasma state based on the phase and magnitude of the reflected signal received from the reflected signal measuring unit; and a plasma state estimation unit (300) that estimates the plasma generation state inside a chamber by combining the transfer function and real-time reflected signal information. This eliminates the need to insert a separate sensor inside the chamber to estimate the plasma state in real time and has the effect of quickly identifying the plasma state based on the phase and magnitude information of the reflected waves.

Description

Technology Field

[0001] The present invention relates to a plasma state estimation system and method based on multiple ultra-high frequency source reflected waves, and more specifically, to a technology for estimating the plasma state in real time using the phase and magnitude of reflected waves reflected from multiple ultra-high frequency sources. Background Technology

[0002] In plasma equipment utilized in various industrial fields such as semiconductor manufacturing and environmental treatment, active technological development is underway to stably generate large-area, high-density plasma with excellent distribution uniformity.

[0003] In particular, to realize high-density, large-area plasma even under low-pressure conditions and to overcome the limitations of conventional magnetron methods, there is an increasing demand for switching to multi-RF generators that combine multiple RF generators to form high energy density.

[0004] To measure the plasma generation state inside a plasma chamber, a method of measuring the plasma state by inserting a separate sensor, such as a Langmuir probe, into the equipment is widely used.

[0005] In addition, electrical impedance changes depending on the plasma generation state, causing reflected waves to occur, and these reflected waves can return to the outside of the equipment. To minimize this, a technique is generally applied to adjust the reflected waves using an impedance tuner.

[0006] According to Public Patent No. 2024-0013053 (January 30, 2024), in estimating the state of plasma within a processing vessel, a plurality of sensors are provided to detect the state of plasma generated within the processing vessel, and a separate sensor is inserted inside the equipment to directly measure the plasma state.

[0007] However, the above-mentioned conventional method has the problem that it requires a separate sensor to be inserted inside the equipment to directly measure the plasma state, which creates structural complexity and a risk of contamination, and also limits real-time monitoring and control. Prior art literature

[0008] Published Patent No. 2024-0013053 (January 30, 2024) The problem to be solved

[0009] The present invention aims to solve the aforementioned problem and provides a system and method capable of precisely estimating the plasma generation state by utilizing phase and magnitude information of reflected waves reflected from multiple ultra-high frequency generators, without inserting a separate sensor inside the chamber to estimate the plasma state in real time. means of solving the problem

[0010] The present invention relates to a plasma state estimation system based on multiple ultra-high frequency source reflections, comprising: a reflection signal measuring unit (100) that measures the phase and magnitude of a reflection signal reflected through an antenna connected to each of a plurality of ultra-high frequency generators; a transfer function generating unit (200) that generates a transfer function for estimating a plasma state based on the phase and magnitude of a reflection signal received from the reflection signal measuring unit; and a plasma state estimation unit (300) that estimates a plasma generation state inside a chamber by combining the transfer function and real-time reflection signal information.

[0011] The reflected signal measuring unit synchronously measures the phase and magnitude of the reflected signal reflected in multiple frequency bands.

[0012] Additionally, the reflection signal measuring unit (100) includes a directional coupler connected to each of a plurality of ultra-high frequency generators, an IQ detector or phase / magnitude measuring unit for measuring the phase and magnitude of the reflection signal separated through the directional coupler, and an A / D converter and a signal processing unit for digitizing and processing the measured signal.

[0013] The reflection signal measuring unit (100) independently measures a reflection signal corresponding to each of a plurality of ultra-high frequency generators and outputs data in the form of a complex reflection coefficient (Γ = |Γ|∠θ) based on the phase (θ) and magnitude (|Γ|) of each reflection signal.

[0014] The transfer function generation unit (200) converts the phase and magnitude of the reflected signals received from each of the multiple ultra-high frequency generators into a complex reflection coefficient form, and generates a transfer function for estimating the plasma state by linearly combining the reflection coefficients with a weighting coefficient corresponding to the position information of the antenna.

[0015] The transfer function generation unit (200) includes a reflection signal input unit that receives reflection signal data of each ultra-high frequency generator received from the reflection signal measurement unit (100), a complex number conversion unit that converts the received reflection signal into a complex reflection coefficient using the phase and magnitude of the reflection signal, a weighted sum calculation unit that derives an intermediate calculation value of the transfer function by multiplying each reflection coefficient by a weighting coefficient set according to the antenna position information and combining them in a linear combination method, and a transfer function modeling unit that generates a transfer function capable of estimating the plasma state in the chamber based on the intermediate calculation value derived from the weighted sum calculation unit (230), wherein the transfer function modeling unit derives a transfer function corresponding to the plasma state in real time according to the characteristics of the reflection signal.

[0016] The transfer function generation unit (200) models the reflection characteristics according to the plasma generation state in the form of a function using the phase and magnitude of the reflection signal as variables, and the transfer function is dynamically generated through a regression model based on prior experimental data or a machine learning-based prediction model.

[0017] The above transfer function generating unit (200) receives complex reflection coefficients (Γ₁, Γ₂, ..., Γ) from each of a plurality of ultra-high frequency generating devices. n The transfer function H(f) is predicted or corrected using an artificial intelligence model that is pre-trained based on training data in which ) is used as an input variable and the corresponding plasma state (electron density, spatial distribution) is used as an output value.

[0018] The plasma state estimation unit (300) calculates the difference (ΔH) between the reference transfer function and the real-time transfer function, and quantitatively determines whether there is an abnormality in the plasma state by determining whether the magnitude or phase change amount of the ΔH exceeds a set threshold.

[0019] The plasma state estimation unit (300) analyzes the position-based change of the transfer function for each antenna connected to a plurality of ultra-high frequency generators and estimates the spatial distribution of the plasma generation state in a two-dimensional or three-dimensional form.

[0020] The above plasma state estimation unit (300) analyzes the sequence of changes in the transfer function over time to derive the direction of change in the plasma state, the amplitude, or the cumulative trend of the abnormal pattern.

[0021] The plasma state estimation unit (300) generates one or more control commands, including RF output adjustment, phase correction, frequency shift, gas flow rate control, or emergency stop command, based on the plasma state estimation result, and outputs them to an external controller.

[0022] A method for estimating a plasma state based on multiple ultra-high frequency source reflections comprises: (a) measuring the phase and magnitude of a reflected signal reflected through an antenna connected to each of a plurality of ultra-high frequency generators; (b) generating a transfer function that mathematically expresses the plasma generation state inside a plasma chamber based on the phase and magnitude of the measured reflected signal; and (c) quantitatively estimating the plasma state inside the chamber by combining the transfer function and real-time reflected signal information. Effects of the invention

[0023] According to the present invention, there is no need to insert a separate sensor inside the chamber to estimate the plasma state in real time, which simplifies the structure of the equipment and reduces the risk of contamination.

[0024] In addition, real-time control is possible as the plasma state can be rapidly identified based on the phase and magnitude information of the reflected waves.

[0025] This can improve the precision and stability of the plasma process and increase overall operational efficiency. Brief explanation of the drawing

[0026] FIG. 1 is a configuration diagram of a plasma state estimation system based on multiple ultra-high frequency source reflections according to one embodiment of the present invention. FIG. 2 shows the detailed configuration of the transfer function generation unit of a plasma state estimation system based on multiple ultra-high frequency source reflections according to one embodiment of the present invention. FIG. 3 is a flowchart illustrating the process of generating a transfer function of a plasma state estimation system based on multiple ultra-high frequency source reflections according to an embodiment of the present invention. Figure 4 shows the detailed configuration of the plasma state estimation unit of a plasma state estimation system based on multiple ultra-high frequency source reflections according to one embodiment of the present invention. FIG. 5 shows the operation flow of the plasma state estimation unit of a plasma state estimation system based on multiple ultra-high frequency source reflections according to one embodiment of the present invention. Figure 6 shows the overall flow of a plasma state estimation system based on multiple ultra-high frequency source reflections according to one embodiment of the present invention. Specific details for implementing the invention

[0027] The present invention relates to a technology that enables accurate determination of the plasma state without the insertion of a separate sensor by estimating the plasma generation state inside a plasma chamber in real time using the phase and magnitude of reflected waves received from each of multiple ultra-high frequency generators.

[0028] FIG. 1 is a configuration diagram of a plasma state estimation system based on multiple ultra-high frequency source reflections according to one embodiment of the present invention.

[0029] As illustrated in FIG. 1, a plasma state estimation system (10) based on multiple ultra-high frequency source reflections includes a reflection signal measurement unit (100), a transfer function generation unit (200), and a plasma state estimation unit (300).

[0030] The reflected signal measuring unit (100) measures the phase and magnitude of the reflected signal reflected through an antenna connected to each of the multiple ultra-high frequency generators.

[0031] This reflected signal measuring unit is configured to measure the magnitude and phase of the reflected signal measured from individual ultra-high frequency sources.

[0032] This reflection signal measuring unit (100) synchronously measures the phase and magnitude of the reflected waves in multiple frequency bands. Since the plasma changes its generated state depending on time and location, the reflected waves received from multiple ultra-high frequency sources may have different characteristics depending on the frequency band.

[0033] In particular, since the phase of reflected waves fluctuates sensitively depending on the electron density or generation distribution of the plasma, synchronously measuring them across multiple frequency bands improves temporal and spatial resolution, enabling more precise estimation of the plasma state.

[0034] Therefore, the present invention has the technical effect of being able to sensitively detect local changes in the plasma state or abnormal states by performing reflected wave measurements in multiple frequency bands in a real-time synchronous manner.

[0035] According to the present embodiment, the internal plasma state can be indirectly estimated through only electromagnetic waves reflected from outside the chamber without inserting a sensor such as a Langmuir probe into the plasma.

[0036] The plasma state estimation system of the present invention is applicable to various plasma generation devices and can be effectively applied, in particular, to microwave-based ECR (Electron Cyclotron Resonance) plasma systems.

[0037] For example, a microwave-based ECR plasma generation system includes a vacuum exhaust device, an Ar gas flow control unit, and a microwave power application device, and an operating pressure at which plasma is stably formed (e.g., 8×10⁻⁶). -4 ~ 3×10 -2 Various plasma distribution states are generated depending on Torr) and microwave power conditions (e.g., 0.7 to 2.3 kW).

[0038] At this time, by measuring the phase and magnitude of the microwave reflected signal received from outside the chamber and deriving a transfer function based on this, the generation status, spatial distribution, or abnormal state of plasma can be quantitatively evaluated in real time.

[0039] The system of the present invention has a reflection coefficient (S) under ECR conditions. 11It is possible to accurately estimate the plasma generation distribution based on changes and phase differences, and by modeling a transfer function in real time that reflects the plasma formation intensity or uniformity according to each operating condition (microwave power, gas flow rate, etc.), it provides high-precision plasma state monitoring and anomaly detection functions without the insertion of sensors.

[0040] Since the structure uses multiple ultra-high frequency sources, the phase and magnitude of the reflected waves generated by each source are inevitably different. The reflected signal measurement unit independently tracks and measures each of these reflected waves.

[0041] The reflected signal measuring unit can measure not only a single frequency but also multiple frequency bands or sweep frequencies if necessary, and can precisely analyze spatial patterns by maintaining time synchronization between multiple channels.

[0042] The reflection signal measuring unit (100) in this embodiment is described in detail as follows. This reflection signal measuring unit (100) includes a directional coupler, an IQ detector or a phase / magnitude measuring unit, an A / D converter, and a signal processing unit (such as a DSP or FPGA).

[0043] A directional coupler is connected to each of the multiple ultra-high frequency generators, through which the power transmitted to the plasma chamber and the reflected signal returning from the chamber are separated.

[0044] At this point, the separated reflected signal is transmitted to a signal analysis device, such as an IQ detector or a phase / magnitude meter, which converts the phase difference and power level of the reflected signal into an electrical signal and outputs it as digital data.

[0045] The output signal is digitized through an A / D converter, and then analyzed in real time by a signal processing unit (such as a DSP or FPGA) to precisely calculate the phase and magnitude of the reflected signal.

[0046] Through this process, the characteristics of the reflected signal for each ultra-high frequency generator are measured independently and subsequently used as reference data for estimating the plasma state.

[0047] That is, the reflected signal measuring unit measures the magnitude (|Γ|, intensity of reflected power) and phase (θ, phase difference with respect to the original transmitted wave) of the reflected signal reflected back from each ultra-high frequency source in real time, and outputs data in the form of a complex reflection coefficient Γ = |Γ|∠θ. The value measured in this way is transmitted to the transfer function generating unit (200) and used as an input variable for generating a transfer function for plasma state estimation.

[0048] The transfer function generation unit (200) generates a transfer function for estimating the plasma state based on the phase and magnitude of the reflected signal received from the reflected signal measurement unit.

[0049] This transfer function generation unit is configured to derive a transfer function for estimating the plasma generation state inside the chamber by combining the magnitude and phase of the measured reflected signal.

[0050] For reference, the impedance characteristics of plasma vary depending on the electron density, and accordingly, the reflection coefficient (S 11 The phase and power loss vary depending on the frequency. The transfer function is a quantitative model of the relationship between the input (RF power) and the output (reflected wave).

[0051] For reference, here S 11 is one of the Scattering Parameters, representing the ratio of power reflected out to the same port among the RF signals applied to Port 1, and is a reflection coefficient that quantitatively represents the impedance characteristics that vary depending on the plasma state.

[0052] Reflection coefficient S 11 is a concept widely used in high-frequency (RF) systems or microwave circuits, where S 11is one of the Scattering Parameters, representing the ratio of power reflected out of Port 1 to the power input to Port 1.

[0053] Expressed as a formula, S 11 =V reflected at port 1 / V incident at port 1 .

[0054] In other words, it is a complex ratio indicating how much of the high-frequency power is reflected and returns to Port 1 when high-frequency power is applied to Port 1.

[0055] |S 11 | = 0 indicates that all is absorbed without any reflection, representing impedance matching (ideal). |S 11 | = 1 indicates that all power is reflected and is a perfect mismatch (open or short).

[0056] When the electron density changes within the plasma chamber, the internal impedance characteristics change, and consequently, the amount of high-frequency power reflected changes, so S 11 The value changes. |S 11 | is used as a real value between 0 and 1, in dB units (20log|S 11 It is also expressed as |). Therefore, S 11 It serves as an indirect indicator of the plasma state (density, uniformity, generation status, etc.).

[0057] The transfer function generation unit (200) derives a transfer function related to the plasma generation state inside the plasma chamber based on the phase and magnitude information of the reflected signal received from the reflected signal measurement unit (100).

[0058] This transfer function is the reflection coefficient (S) depending on the frequency (f). 11 It is modeled using multiple measurements such as the phase change amount (Δφ) and the magnitude of reflected power (|Γ|) as variables, and has unique functional response characteristics depending on the electron density distribution or generation location of the plasma.

[0059] For example, it can be assumed that the reflected signals measured for three ultra-high frequency sources (RF1 to RF3) in the chamber are as follows.

[0060] RF1: Phase φ₁ = 175°, Magnitude |Γ₁| = 0.2

[0061] RF2: Phase φ₂ = 180°, Magnitude |Γ₂| = 0.25

[0062] RF3: Phase φ₃ = 210°, Magnitude |Γ₃| = 0.7

[0063] Based on these measurements, a reflection characteristic matrix (R) can be constructed for each source, and the transfer function H(f) can be constructed in the following form:

[0064]

[0065] Here, W₁ to W₃ are weighting factors that account for antenna location, plasma influence area, etc., and this transfer function is used to mathematically estimate the plasma generation distribution.

[0066] The transfer function generation unit (200) may be configured with a regression model based on prior experimental data or a data-based machine learning algorithm (SVR, GPR, etc.) and may be a structure that receives real-time reflected wave data and is dynamically updated every cycle.

[0067] That is, the transfer function generating unit (200) is configured to mathematically combine the phase and magnitude of the reflected signals received from a plurality of ultra-high frequency generators with weighting coefficients corresponding to the position information of each antenna to derive a transfer function that models the reflection characteristics according to the plasma state in the form of a function.

[0068] Mathematically combining the phase and magnitude of reflected signals received from multiple ultra-high frequency generators with weighting factors corresponding to the position information of each antenna involves expressing each reflected signal as a complex reflection coefficient (Γ = |Γ|∠θ) and then using weighting factors (W₁, W₂, ..., W₀) set according to the spatial position occupied by the antenna or the correlation of the region affected by plasma. n It means performing a linear combination by multiplying ).

[0069] Meanwhile, the transfer function generation unit (200) may additionally include an artificial intelligence-based transfer function estimation algorithm that predicts the plasma state using the phase and magnitude of the reflected signal as input variables, in addition to the linear combination method described above.

[0070] For example, the transfer function modeling section (240) has a plurality of reflection coefficients (Γ₁, Γ₂, ..., Γ n It may include a configuration that takes ) as input and dynamically generates a transfer function H(f) using machine learning algorithms such as Support Vector Regression, Neural Network, and Gaussian Process Regression (GPR) learned based on prior experimental data or simulation data.

[0071] This AI-based transfer function estimation method has the advantage of further improving the accuracy and environmental adaptability of plasma state estimation, as it can precisely model the nonlinear and complex interactions between the plasma state and the reflected signal.

[0072] Furthermore, if the system is implemented to continuously learn or correct through data input, real-time dynamic estimation is possible that reflects even minute variations in the plasma process over time.

[0073] The transfer function H in this embodiment is H = W₁·Γ₁ + W₂·Γ₂ + ... + W n ·Γ nIt is defined in the form of, where each W n is a weighting value that reflects the effect on plasma density changes or reflection characteristics at the corresponding antenna location, and Γ n is a complex value representing the phase and magnitude of the reflected signal measured at that location.

[0074] The transfer function is constructed, for example, in the form of a linear combination or matrix operation based on the complex reflection coefficients (values ​​including magnitude and phase) of individual reflected signals, and is used as a reference function to quantitatively estimate states such as the generation distribution, density, or electron concentration of the plasma within the chamber.

[0075] This transfer function can be extended into a time-varying function form capable of reflecting the dynamic trend of the plasma state by reflecting the amount of change in the reflected signal over time.

[0076] The details of the previously explained content regarding the detailed configuration of the transfer function generation section are as follows.

[0077] FIG. 2 shows the detailed configuration of the transfer function generation unit of a plasma state estimation system based on multiple ultra-high frequency source reflections according to one embodiment of the present invention.

[0078] As shown in FIG. 2, the transfer function generation unit (200) is composed of a reflection signal input unit (210), a complex number conversion unit (220), a weighted sum calculation unit (230), and a transfer function modeling unit (240).

[0079] The reflection signal input unit (210) receives reflection signal data of each ultra-high frequency generator received from the reflection signal measurement unit (100).

[0080] The complex number conversion unit (220) converts the received reflected signal into the form of a complex reflection coefficient (Γ = |Γ|∠θ) using the phase and magnitude of the received reflected signal.

[0081] The complex number conversion unit (220) converts the phase (θ) and magnitude (|Γ|) of the reflected signal into the form of a complex reflection coefficient Γ = |Γ|∠θ, which is a mathematical expression integrating phase and magnitude information, and is an essential preprocessing step for quantitatively modeling the plasma state and utilizing it for linear operations or transfer function calculations.

[0082] The weighted sum calculation unit (230) calculates weighted coefficients (W₁, W₂, ..., W₁) set according to the antenna position information for each reflection coefficient. n After multiplying by ), the intermediate calculated value of the transfer function is derived by combining it in a linear combination manner.

[0083] For reference, the weighted sum calculation unit (230) calculates the complex reflection coefficients (e.g., Γ₁, Γ₂, ..., Γ) measured from each ultra-high frequency generator. n For ), pre-set weighting factors (W₁, W₂, ..., W₀) considering the installation location of each antenna or the plasma influence area, etc. n After multiplying by ), they are combined using the linear combination method (Σ W n ·Γ n It is a configuration that derives an intermediate calculated value of the transfer function by adding )

[0084] Here, the "linear combination method" refers to a mathematical operation in which each complex reflection coefficient is multiplied by its corresponding weight and then simply summed up; for example, H′(f) = W₁·Γ₁ + W₂·Γ₂ + ... + W n ·Γ n It is expressed in the form of.

[0085] The intermediate calculated value (H′(f)) at this time is a preprocessing result for the final transfer function modeling, and is a composite response value that integrates and reflects the spatial distribution and intensity differences of the reflection coefficients based on weights. This intermediate calculated value is subsequently converted into the final transfer function H(f) through normalization, reflection of time changes, and frequency characteristic correction in the transfer function modeling unit (240), and as a result, can be used as base data for quantitatively estimating the plasma state.

[0086] The transfer function modeling unit (240) forms a transfer function H(f) capable of estimating the plasma state within the chamber based on intermediate calculation values ​​derived from the weighted sum calculation unit (230), and the function is used for real-time state prediction, comparison with reference models, generation of control commands, etc.

[0087] This transfer function modeling unit (240) mathematically models a transfer function H(f) that can quantitatively express the plasma state based on the intermediate calculation value derived from the weighted sum calculation unit (230), and the said transfer function can be used as a key reference value for real-time state prediction, anomaly detection through comparison with a reference transfer function, and generation of plasma process control commands.

[0088] For reference, the transfer function modeling unit (240) forms a final transfer function H(f) that can quantitatively express the plasma state based on the intermediate calculated value H′(f) derived from the weighted sum calculation unit (230).

[0089] At this time, H(f) is modeled as an intrinsic response function that can mathematically express the plasma generation state, density, and distribution by integrating the reflection coefficient values ​​(e.g., |Γ|, θ) for each frequency and additionally reflecting the change pattern over time or the error value relative to the reference state.

[0090] The transfer function H(f) can be expressed as, for example, as follows.

[0091] H(f) = Σ W n · Γ n (f) + B(f). Here, W n is the weighting coefficient, Γ n (f) is the complex reflection coefficient, and B(f) is the correction bias or normalization term.

[0092] Or in machine learning-based methods Here, M is a learning-based model such as a regression model (SVR), a neural network (NN), or GPR.

[0093] In the quantitative estimation of plasma state, H(f) is used as a reference function to mathematically estimate the plasma electron density, generation uniformity, and distribution of abnormal regions, and can determine whether there is a state anomaly by comparing it with real-time input values.

[0094] In addition, when comparing with a reference model, the difference ΔH(f, t) = H(t, f) - H0(f) from the reference transfer function H0(f) is calculated to determine whether it is normal or abnormal, and error analysis can be performed based on the phase difference and magnitude difference.

[0095] In addition, when generating process control commands, if ΔH(f, t) exceeds a threshold value at a specific location or frequency band, process control commands can be generated for RF output correction, phase adjustment, frequency switching, gas flow rate adjustment, alarm, or emergency stop signal transmission.

[0096] The weighted sum calculation unit (230) basically calculates the complex reflection coefficients (Γ₁, Γ₂, ..., Γ) measured from each ultra-high frequency generator. n For ), weighting factors (W₁, W₂, ..., W) set considering the installation location of each antenna or the plasma influence region, etc. n After multiplying by ), use the linear combination method (ΣW n ·Γ n It is a configuration that derives the intermediate calculated value H′(f) of the transfer function by combining ).

[0097] Meanwhile, in another embodiment of the present invention, in addition to the linear combination-based method, a plurality of complex reflection coefficients (Γ₁, Γ₂, ..., Γ n An artificial intelligence-based function estimation algorithm can be applied to derive an intermediate or final expression of a transfer function using ) as input.

[0098] For example, when measuring the magnitude (|Γ|) and phase (θ) of reflected signals received through multiple antennas in plasma equipment, different reflection characteristics appear for each antenna.

[0099] The plasma state can be estimated by multiplying these signals by preset weighting factors (W₁, W₂, ...) and linearly adding them (linear combination).

[0100] However, since the plasma state can vary complexly depending on various process conditions such as gas type, flow rate, pressure inside the chamber, and temperature, there may be cases where it is difficult to ensure sufficient accuracy in all situations using only linear combination-based estimation.

[0101] Accordingly, in another embodiment of the present invention, along with this linear-based method, an artificial intelligence-based transfer function estimation method capable of more flexibly reflecting complex nonlinear relationships or multivariate interactions may be additionally included.

[0102] Accordingly, one embodiment of the present invention may include an artificial intelligence-based transfer function estimation method.

[0103] Meanwhile, the linear combination-based method has the advantage of a simple structure and fast computation speed, making it advantageous for real-time processing or environments with limited system resources. In particular, it can demonstrate effective performance under conditions where the relationship between the plasma state and the reflected signal is relatively stable and linear. Therefore, the present invention may also include a hybrid method that balances accuracy and speed by utilizing a linear combination-based model and an artificial intelligence-based model in parallel or selectively as needed.

[0104] Specifically, reflected signal data is set as the input variable (X), and the corresponding plasma state (e.g., electron density, generation location, etc.) is set as the output (Y) to pre-train an AI model (e.g., artificial neural network, support vector regression, Gaussian process regression, etc.).

[0105] The AI ​​model trained in this way is configured to analyze input reflection signal patterns and perform predictions such as, "When such a reflection signal is received, the plasma state has these characteristics."

[0106] During the actual process, new reflection signals are input in real time, and at this time, the AI ​​model uses the input reflection coefficients (Γ₁, Γ₂, ..., Γ n The transfer function H(f) is immediately estimated based on ).

[0107] The transfer function is a value that can mathematically express the current plasma state and can be utilized for various plasma diagnostic and control functions, such as electron density, spatial distribution, detection of anomalies, and generation of control signals.

[0108] Input (X): [|Γ₁|, θ₁, |Γ₂|, θ₂, ..., |Γ n |, θ n If ] is applied to an AI model (such as a pre-trained artificial neural network), the state can be estimated (electron density, spatial distribution, anomaly detection, etc.) using the output (H(f)).

[0109] For reference, the weighted sum calculation unit (230) or the transfer function modeling unit (240) can directly generate or correct intermediate calculated values ​​or the transfer function H(f) using a pre-trained regression model (Support Vector Regression), artificial neural network, Gaussian Process Regression, etc. based on the input reflection coefficient pattern.

[0110] This AI-based approach can more precisely reflect nonlinear relationships between multivariates or complex spatial distribution patterns that are difficult to express with linear models, thereby improving the accuracy and real-time adaptability of plasma state estimation.

[0111] That is, in another embodiment of the present invention, in addition to a linear calculation method that simply multiplies reflected signals by weights and adds them, the plasma state can be predicted using artificial intelligence (AI) technology.

[0112] For example, it is a method in which a computer receives magnitude and phase data of reflected signals from multiple antennas and predicts the current plasma state through an AI model (e.g., an artificial neural network) that has been pre-trained in it.

[0113] This is akin to how a computer learns from numerous past reflected signals and their corresponding plasma states to independently determine the state when a new signal arrives, just as a human infers a situation by observing specific patterns based on extensive experience. This approach can account for the complex and non-linear relationships between plasma states and reflected signals, allowing for higher precision and predictive performance.

[0114] FIG. 3 is a flowchart illustrating the process of generating a transfer function of a plasma state estimation system based on multiple ultra-high frequency source reflections according to an embodiment of the present invention.

[0115] As illustrated in FIG. 3, the transfer function generation unit (200) converts the input reflection signal (s31) into a complex number form (s32), multiplies the position-based weighting coefficient corresponding to each reflection coefficient (s33), and linearly combines them (s34) to generate a transfer function (s35) that mathematically models the plasma state.

[0116] For reference, |Γ| (Gamma Magnitude) represents the magnitude of the reflection coefficient (Γ). It indicates how much power is reflected from the plasma chamber, i.e., the intensity of the reflected power, and is generally a real value between 0 and 1. For example, |Γ| = 0.2 means that 20% of the input power was reflected.

[0117] θ (Phase Angle) is the phase component of the reflection coefficient. It indicates the phase difference between the reflected wave and the original transmitted wave, and is usually expressed in degrees or radians. For example, θ = 180° means that the reflected wave returns with the opposite phase to the transmitted wave.

[0118] The reason these two are used together is that reflected signals possess complex characteristics that include not only intensity but also phase. The phase and magnitude of the reflected signal refer to the complex reflection coefficient (Γ = |Γ|∠θ) mentioned above.

[0119] The generated transfer function can be used as a reference value to estimate plasma generation density, electron concentration, and spatial distribution, and can also be utilized for anomaly detection or real-time process control by analyzing the difference from the reference transfer function.

[0120] For example, the reference transfer function H0(f) is a transfer function measured in advance under a state where the plasma is normally and uniformly formed, and the real-time measured transfer function H(t, f) is a function generated based on the reflected signal during the current process.

[0121] If the magnitude of the reflection coefficient |Γ| in a specific frequency band at H(t, f) increases significantly above a reference value or the phase θ changes rapidly, this may be an abnormal sign indicating a potential decrease or extinction of plasma density at that location. By detecting this transfer function difference ΔH(f, t), process control commands can be applied in real time to increase the RF output in that region or to re-normalize the plasma through phase correction.

[0122] The plasma state estimation unit (300) estimates the plasma generation state inside the chamber by combining the transfer function and real-time reflection signal information.

[0123] This plasma state estimation unit estimates the plasma generation state through reflected signal information during normal operation.

[0124] The plasma state estimation unit (300) is configured to quantitatively estimate the plasma generation state inside the chamber based on the transfer function generated by the transfer function generation unit (200) and the reflected signal data received in real time.

[0125] The plasma state estimation unit (300) according to the present embodiment is described in detail as follows.

[0126] The plasma state estimation unit (300) is configured to perform quantitative estimation by comparing a reference transfer function generated from the transfer function generation unit (200) with a real-time transfer function calculated from the phase and magnitude of a reflected signal input in real time, and numerically evaluating the plasma state based on the difference value or residual between them.

[0127] For example, the complex reflection coefficient Γ measured at each antenna n For (t), the difference between the reference transfer function H0(f) and the real-time transfer function H(t, f) is calculated, and if the magnitude or phase change amount of this difference value ΔH(f, t) = H(t, f) - H0(f) exceeds a set threshold, the plasma state abnormality or abnormal distribution at the corresponding time or location can be quantitatively determined.

[0128] In addition, quantitative state estimation results can be derived by analyzing the absolute value or sum of squares (Σ|ΔH|²) of ΔH(f, t) and comparing the plasma electron density, local generation intensity, generation radius, etc. with a numerical model.

[0129] The detailed configuration of the plasma state estimation unit (300) is as follows.

[0130] Figure 4 shows the detailed configuration of the plasma state estimation unit of a plasma state estimation system based on multiple ultra-high frequency source reflections according to one embodiment of the present invention.

[0131] As illustrated in FIG. 4, the plasma state estimation unit (300) is configured to include a transfer function comparison unit (310), a state determination unit (320), a spatial distribution estimation unit (330), a dynamic trend analysis unit (340), and a control command generation unit (350).

[0132] The transfer function comparison unit (310) calculates ΔH(f, t) by comparing the reference transfer function H0(f) input from the transfer function generation unit (200) with the transfer function H(t, f) calculated based on the real-time reflection signal.

[0133] The state determination unit (320) determines whether the magnitude or phase change amount of ΔH(f, t) exceeds a preset threshold value, thereby quantitatively determining an abnormal state or abnormal region of the plasma.

[0134] The spatial distribution estimation unit (330) estimates the plasma generation location and spatial distribution state in a two-dimensional or three-dimensional form based on the transfer function change characteristics of each antenna connected to a plurality of ultra-high frequency generators.

[0135] Here, the spatial distribution estimation unit (330) is configured to quantitatively estimate the spatial non-uniformity or distribution pattern of plasma generation within the chamber by analyzing the change in the transfer function corresponding to the change in the phase and magnitude of the reflected signal received from each antenna connected to a plurality of ultra-high frequency generators according to location.

[0136] Transfer function H corresponding to each antenna n Based on the spatiotemporal differences of (f), the plasma's high-density generation region, generation radius, and local extinction point can be mapped and visualized or quantified in the form of a 2D or 3D distribution. Through this, the plasma formation state in the entire chamber area can be estimated in terms of area or volume, and local abnormal regions can also be precisely identified.

[0137] Optionally, a dynamic trend analysis unit (340) that analyzes the change pattern of the transfer function over time or a control command generation unit (350) that outputs a process control command may be additionally included.

[0138] The dynamic trend analysis unit (340) is configured to estimate the direction of change of plasma state, the rate of change, and the cumulative trend of abnormal patterns by analyzing the sequence of changes in the transfer function over time.

[0139] By analyzing the continuous change of the transfer function H(t, f) corresponding to each time interval Δt, signs of dynamic anomalies such as gradual degradation of the plasma state, density instability, and repetitive oscillation can be detected in real time. This analysis can be utilized as a function for predicting changes in the plasma state during the process or as a pre-processing alarm.

[0140] The control command generation unit (350) is configured to generate a control command to be provided to an ultra-high frequency generator or an external plasma process controller based on the plasma state abnormality determination result derived from the state determination unit (320) and the spatial distribution estimation unit (330).

[0141] Here, the control command is a correction signal for stabilizing or homogenizing the plasma state and may include the following.

[0142] It may include output power adjustment to correct plasma density by adjusting the RF output power of a specific antenna up or down, phase correction to control input to an RF phase controller to reduce phase difference, frequency shifting to fine-tune the operating frequency of a source when inefficiency in a specific frequency band is detected, chamber internal gas flow control to generate a flow control signal when plasma density imbalance is associated with process gas distribution, and emergency stop to generate a system interrupt or shot cutoff signal when a failure or transient response is detected.

[0143] In summary, the transfer function comparison unit (310) calculates the difference between the reference transfer function and the transfer function generated in real time, and calculates ΔH(f, t) based on the phase and magnitude difference of the complex reflection coefficient to quantitatively evaluate whether there is a change in the plasma state.

[0144] The state determination unit (320) is configured to determine whether the ΔH(f, t) derived from the transfer function comparison unit (310) exceeds a preset allowable error or threshold value, thereby determining whether there is an abnormality in the plasma state, an abnormal region, or a non-uniform distribution.

[0145] For reference, the tolerance or threshold is a numerical reference value used to determine whether the difference ΔH(f,t) between the reference transfer function H0(f) and the real-time transfer function H(t,f) is within the steady-state range. If this reference value is exceeded, the plasma state is determined to be abnormal or abnormal.

[0146] The threshold value is a numerical criterion for quantitatively determining the plasma state and can be set by one of the following or a combination thereof.

[0147] First, the reference state experimental result-based setting is a method of establishing an acceptable range by statistically analyzing the mean value and standard deviation of the transfer function after repeatedly performing the plasma process in a steady state.

[0148] Second, the boundary value learning-based setting is a method that uses machine learning or simulation models to learn the difference in transfer functions between normal and abnormal states, and automatically corrects or updates threshold values ​​in real time during the process.

[0149] Third, process specification-based fixed value setting is a method of setting pre-fixed numerical values ​​based on normal operating conditions defined in the design specifications of the reaction chamber or RF system.

[0150] For example, it can be determined that there is an issue if the phase difference exceeds ±10° or the reflection coefficient magnitude difference is 0.05 or greater, and these threshold values ​​can be preset based on experiments or equipment specifications, or automatically adjusted based on learning during the process.

[0151] The spatial distribution estimation unit (330) is configured to analyze the difference in relative intensity of the plasma generation location based on the change in transfer function of each antenna connected to a plurality of ultra-high frequency generators, and to estimate the plasma spatial distribution state in a two-dimensional or three-dimensional form.

[0152] The dynamic trend analysis unit (340) can determine the direction of change in plasma state and the cumulative pattern of abnormal signs in real time by analyzing the sequence of changes in the transfer function over time.

[0153] The control command generation unit (350) is configured to generate control commands for plasma process stabilization, such as RF output adjustment, phase correction, frequency shifting, gas flow rate control, or emergency stop, based on the above-mentioned state analysis results, and output them to an external controller or each RF generator.

[0154] FIG. 5 shows the operation flow of the plasma state estimation unit of a plasma state estimation system based on multiple ultra-high frequency source reflections according to one embodiment of the present invention.

[0155] As illustrated in FIG. 5, the plasma state estimation unit (300) calculates the difference between the reference transfer function and the transfer function generated in real time through the transfer function comparison unit (310), and calculates ΔH(f, t) based on the phase and magnitude difference of the complex reflection coefficients to quantitatively evaluate whether there is a change in the plasma state (s51).

[0156] Next, the plasma state estimation unit (300) determines whether the ΔH(f, t) derived from the transfer function comparison unit (310) through the state determination unit (320) exceeds a preset allowable error or threshold value, thereby determining whether there is an abnormality in the plasma state, an abnormal region, or a non-uniform distribution (s52).

[0157] Next, the plasma state estimation unit (300) analyzes the difference in relative intensity of the plasma generation location based on the change in transfer function for each antenna connected to a plurality of ultra-high frequency generators through the spatial distribution estimation unit (330), and estimates the plasma spatial distribution state in a two-dimensional or three-dimensional form (s53).

[0158] Next, the plasma state estimation unit (300) determines the direction of change of the plasma state and the cumulative pattern of abnormal signs in real time by analyzing the sequence of changes in the transfer function over time through the dynamic trend analysis unit (340) (s54).

[0159] And the plasma state estimation unit (300) generates control commands for plasma process stabilization, such as RF output adjustment, phase correction, frequency shifting, gas flow rate control, or emergency stop, based on the state analysis results through the control command generation unit (350), and outputs them to an external controller or each RF generator (s55).

[0160] The plasma state estimation unit (300) can determine the generation distribution, electron density, non-uniformity, or abnormal state of the current plasma by comparing and analyzing real-time phase and magnitude information of reflected signals received from multiple ultra-high frequency generators with the transfer function H(f) derived from said information.

[0161] In particular, by comparing the transfer function in the reference state with the transfer function in the current state, or by analyzing the trend of the transfer function change over time, it is possible to reflect even the dynamic changes in the plasma state in real time.

[0162] At this time, the plasma state estimation unit can identify local anomalies or abnormal regions by detecting specific reaction characteristics that fall outside the allowable error range compared with a set reference transfer function, and can provide control commands to the plasma process control device based on this information.

[0163] In addition, the plasma state estimation unit may also include a function to estimate the spatial distribution state of the plasma within the chamber in a two-dimensional or three-dimensional form by comprehensively analyzing the phase / magnitude change pattern of the reflected signal corresponding to each ultra-high frequency generator.

[0164] Accordingly, the plasma state estimation unit (300) of the present invention can realize high-precision, high-speed plasma state monitoring and anomaly detection functions without a separate intrusive sensor by integrating real-time diagnosis based on reflected waves and spatial distribution analysis.

[0165] Figure 6 shows the overall flow of a plasma state estimation system based on multiple ultra-high frequency source reflections according to one embodiment of the present invention.

[0166] As illustrated in FIG. 6, a plasma state estimation system based on multiple ultra-high frequency source reflections (hereinafter referred to as the system) measures the phase and magnitude of a reflected signal reflected through an antenna connected to each of a plurality of ultra-high frequency generators (s61).

[0167] Next, the system generates a transfer function to estimate the plasma state based on the phase and magnitude of the measured reflected signal (s62).

[0168] Next, the system estimates the plasma generation state inside the chamber by combining the generated transfer function and real-time reflection signal information (s63).

[0169] According to the present invention, there is no need to insert a separate sensor inside the chamber to estimate the plasma state in real time, thereby simplifying the structure of the equipment and reducing the risk of contamination. In addition, real-time control is possible by rapidly identifying the plasma state based on the phase and magnitude information of the reflected waves. This improves the precision and stability of the plasma process and increases overall operational efficiency.

Claims

Claim 1 A plasma state estimation system based on multiple ultra-high frequency source reflections comprises: a reflection signal measuring unit (100) that measures the phase and magnitude of a reflection signal reflected through an antenna connected to each of a plurality of ultra-high frequency generators; a transfer function generating unit (200) that generates a transfer function for estimating the plasma state based on the phase and magnitude of the reflection signal received from the reflection signal measuring unit; and a plasma state estimation unit (300) that estimates the plasma generation state inside a chamber by combining the transfer function and real-time reflection signal information, wherein the plasma state estimation unit (300) analyzes the sequence of changes in the transfer function over time to derive the direction of change, amplitude, or cumulative trend of an abnormal pattern of the plasma state. Claim 2 A plasma state estimation system based on multiple ultra-high frequency source reflected waves, wherein, in claim 1, the reflected signal measuring unit synchronously measures the phase and magnitude of the reflected signal reflected in multiple frequency bands. Claim 3 A plasma state estimation system based on multiple ultra-high frequency source reflected waves, characterized in that, in claim 1, the reflected signal measuring unit (100) comprises a directional coupler connected to each of a plurality of ultra-high frequency generators, an IQ detector or phase / magnitude measuring unit for measuring the phase and magnitude of the reflected signal separated through the directional coupler, and an A / D converter and signal processing unit for digitizing and processing the measured signal. Claim 4 A plasma state estimation system based on multiple ultra-high frequency source reflected waves, characterized in that, in claim 1, the reflected signal measuring unit (100) independently measures a reflected signal corresponding to each of a plurality of ultra-high frequency generators and outputs data in the form of a complex reflection coefficient (Γ = |Γ|∠θ) based on the phase (θ) and magnitude (|Γ|) of each reflected signal. Claim 5 A plasma state estimation system based on multiple ultra-high frequency source reflected waves, wherein, in claim 1, the transfer function generation unit (200) converts the phase and magnitude of a reflected signal received from each of a plurality of ultra-high frequency generators into a complex reflection coefficient form, and generates a transfer function for estimating the plasma state by linearly combining the reflection coefficients with a weighting coefficient corresponding to the position information of the antenna. Claim 6 In claim 1, the transfer function generation unit (200) comprises: a reflection signal input unit that receives reflection signal data of each ultra-high frequency generator received from the reflection signal measurement unit (100); a complex number conversion unit that converts the received reflection signal into a complex reflection coefficient form using the phase and magnitude of the reflection signal; a weighted sum calculation unit (230) that derives an intermediate calculation value of the transfer function by multiplying each reflection coefficient by a weighting coefficient set according to the antenna position information and combining them in a linear combination method; and a transfer function modeling unit that generates a transfer function capable of estimating the plasma state within the chamber based on the intermediate calculation value derived from the weighted sum calculation unit (230), wherein the transfer function modeling unit derives a transfer function corresponding to the plasma state in real time according to the characteristics of the reflection signal. Claim 7 A plasma state estimation system based on multiple ultra-high frequency source reflected waves, characterized in that, in claim 1, the transfer function generation unit (200) models the reflection characteristics according to the plasma generation state in the form of a function using the phase and magnitude of the reflected signal as variables, and the transfer function is dynamically generated through a regression model based on prior experimental data or a machine learning-based prediction model. Claim 8 In claim 1, the transfer function generating unit (200) comprises complex reflection coefficients (Γ₁, Γ₂, ..., Γ) received from each of a plurality of ultra-high frequency generating devices. n A plasma state estimation system based on multiple ultra-high frequency source reflections, characterized by using a pre-trained artificial intelligence model based on training data that takes ) as an input variable and the corresponding plasma state (electron density, spatial distribution) as an output value to predict or correct the transfer function H(f). Claim 9 A plasma state estimation system based on multiple ultra-high frequency source reflections, wherein, in claim 1, the plasma state estimation unit (300) calculates the difference (ΔH) between a reference transfer function and a real-time transfer function, and determines whether the magnitude or phase change amount of the ΔH exceeds a set threshold value, thereby quantitatively determining whether there is an abnormality in the plasma state. Claim 10 A plasma state estimation system based on multiple ultra-high frequency source reflections, wherein, in claim 1, the plasma state estimation unit (300) analyzes the position-based change of the transfer function for each antenna connected to a plurality of ultra-high frequency generators and estimates the spatial distribution of the plasma generation state in a two-dimensional or three-dimensional form. Claim 11 delete Claim 12 A plasma state estimation system based on multiple ultra-high frequency source reflections, wherein, in claim 1, the plasma state estimation unit (300) generates one or more control commands, including RF output adjustment, phase correction, frequency shift, gas flow rate control, or emergency stop command, according to the plasma state estimation result and outputs them to an external controller. Claim 13 A method for estimating a plasma state based on multiple ultra-high frequency source reflections, comprising: (a) a step of measuring the phase and magnitude of a reflected signal reflected through an antenna connected to each of a plurality of ultra-high frequency generators; (b) a step of generating a transfer function that mathematically expresses the plasma generation state inside a plasma chamber based on the phase and magnitude of the measured reflected signal; and (c) a step of quantitatively estimating the plasma state inside the chamber by combining the transfer function and real-time reflected signal information, wherein step (c) is characterized by analyzing the sequence of changes in the transfer function over time to derive the direction of change, amplitude, or cumulative trend of an abnormal pattern of the plasma state.

Images