Cleaning appratus for plasma chamber

Abstract

The present invention relates to a technology for efficiently removing by-products generated in a semiconductor process, and in particular, to a cleaning device for a plasma chamber that can shorten the decomposition time and improve the decomposition rate of by-products by estimating the amount of by-products through spectrum analysis technology and variably controlling the RF power accordingly. To this end, the cleaning device for a plasma chamber according to the present invention includes an OES (Optical Emission Spectroscopy) sensor that measures the optical spectrum emitted from the plasma inside the plasma chamber in real time; and a control unit that analyzes the optical spectrum data received from the OES sensor to calculate the concentration of by-products accumulated inside the plasma chamber, and calculates the RF power applied to the plasma chamber based on the concentration of by-products to control the operation of an RF generator.

Description

Technology Field

[0001] The present invention relates to a technology for efficiently removing by-products generated in a semiconductor process, and in particular, to a cleaning device for a plasma chamber that can shorten the decomposition time and improve the decomposition rate of by-products by estimating the amount of by-products through spectrum analysis technology and variably controlling the RF power accordingly. Background Technology

[0002] In semiconductor manufacturing, the Plasma Enhanced Chemical Vapor Deposition (PECVD) process is a core step for forming thin films on wafer surfaces. During this deposition process, byproducts continuously accumulate not only on the wafer surface but also on the inner walls of the plasma chamber. In particular, in the case of silicon dioxide (SiO2) deposition processes using silane (SiH4) and nitric oxide (N2O), a problem arises where a significant amount of silicon dioxide adheres to the inner walls of the chamber.

[0003] By-products accumulated on the inner walls of the chamber cause various problems. First, if by-products accumulate beyond a certain level, they delaminate, generating particles that contaminate the wafer surface and lead to a decrease in yield. Furthermore, changes in the chamber environment reduce process reproducibility and worsen process uniformity, while alterations in plasma characteristics make it difficult to achieve desired deposition results.

[0004] To solve these problems, it is essential to periodically perform a chamber cleaning process. Conventional cleaning methods mainly used a method of performing the cleaning process at fixed time intervals under fixed conditions. For example, this method involves injecting a fluorine-based cleaning gas, such as NF3 or CF4, into the chamber and generating plasma with a constant RF power to decompose and remove byproducts.

[0005] However, these conventional methods have the following limitations.

[0006] First, since cleaning is performed at fixed time intervals regardless of the actual amount of accumulated by-products, inefficiency occurs where cleaning is performed even when by-products have not accumulated sufficiently, or only after they have accumulated excessively.

[0007] Second, since fixed RF power is used regardless of the amount of byproduct, there is a problem where excessive energy is used when there is little byproduct and the cleaning time is prolonged when there is a lot of byproduct.

[0008] Third, the cleaning process may be incompletely finished or may take an unnecessarily long time because the progress of the cleaning process cannot be monitored in real time.

[0009] Therefore, there is a need to develop technology that can accurately determine the real-time accumulation amount of by-products and dynamically control RF power accordingly to maximize cleaning efficiency. Prior art literature

[0010] (Prior Art 1) Korean Registered Patent No. 0818561 (Prior Art 2) Korean Registered Patent No. 0885187 The problem to be solved

[0011] The problem that the present invention aims to solve is to provide a cleaning device for a plasma chamber that can maximize cleaning efficiency and shorten cleaning time by accurately measuring the amount of byproducts accumulated in the plasma chamber in real time and variably controlling the RF power according to the measured amount of byproducts.

[0012] Another problem that the present invention aims to solve is to provide a cleaning device for a plasma chamber that can monitor the degree of byproduct accumulation even during the deposition process and perform a cleaning process immediately when necessary.

[0013] Another problem that the present invention aims to solve is to provide a cleaning device for a plasma chamber that improves the precision and reproducibility of the cleaning process by monitoring the decomposition process of byproducts in real time using optical emission spectroscopy (OES) and calculating the optimal RF power through a machine learning algorithm. means of solving the problem

[0014] To this end, the cleaning device of a plasma chamber according to the present invention comprises: an RF generator that supplies high-frequency power; a plasma chamber that generates plasma by receiving high-frequency power from the RF generator; an upper electrode installed in the plasma chamber to which high-frequency power is applied; a lower heated plate disposed opposite to the upper electrode and on which a wafer is placed; a process gas injection pipe that injects process gas for a deposition process into the plasma chamber; a cleaning gas injection pipe that injects cleaning gas for a cleaning process into the plasma chamber; an OES (Optical Emission Spectroscopy) sensor that measures the optical spectrum emitted from the plasma inside the plasma chamber in real time; and a control unit that analyzes the optical spectrum data received from the OES sensor to calculate the concentration of byproducts, calculates RF power based on the concentration of byproducts, and feeds back to the RF generator.

[0015] In addition, the cleaning device of a plasma chamber according to the present invention includes an OES (Optical Emission Spectroscopy) sensor that measures the optical spectrum emitted from the plasma inside the plasma chamber in real time; and a control unit that analyzes the optical spectrum data received from the OES sensor to calculate the concentration of byproducts accumulated inside the plasma chamber and calculates the RF power applied to the plasma chamber based on the concentration of byproducts to control the operation of an RF generator.

[0016] In addition, the cleaning method of a plasma chamber according to the present invention is a method performed in a cleaning device of a plasma chamber, comprising the steps of: blocking the injection of a process gas and injecting a cleaning gas during or after the deposition process in the plasma chamber; applying a reference RF power to generate plasma with the cleaning gas to decompose by-products on the inner wall of the chamber; measuring the optical spectrum emitted from the plasma in real time through an OES sensor; and analyzing the optical spectrum to calculate the concentration of by-products and calculating the RF power based on the concentration of by-products to adjust the RF power of an RF generator. Effects of the invention

[0017] The cleaning device for a plasma chamber according to the present invention provides the following effects.

[0018] First, since the amount of accumulated by-products can be accurately measured in real time using an OES sensor, an efficient cleaning process based on the actual state of the by-products is possible.

[0019] Second, by dynamically adjusting RF power according to the amount of by-products, it is possible to quickly remove them with high power when there are many by-products and save energy with low power when there are few by-products, thereby optimizing cleaning time and energy consumption.

[0020] Third, since the degree of byproduct accumulation can be monitored even during the deposition process and immediate cleaning can be performed when necessary, the generation of particles caused by excessive accumulation of byproducts can be prevented in advance.

[0021] Fourth, through machine learning-based analysis algorithms, the relationship between byproduct concentration and optimal RF power can be continuously learned and improved, thereby enhancing the precision and reproducibility of the cleaning process.

[0022] Fifth, since the end point of the cleaning process can be determined based on the actual removal status of by-products, damage to chamber parts caused by excessive cleaning can be prevented and process defects caused by incomplete cleaning can be prevented.

[0023] Sixth, overall process efficiency is improved, leading to increased chamber utilization, reduced maintenance costs, and improved product yield and quality. Brief explanation of the drawing

[0024] FIG. 1 is a schematic diagram showing the configuration of a cleaning device for a plasma chamber according to the present invention. FIG. 2 is a flowchart showing the cleaning control process of a plasma chamber according to the present invention. Specific details for implementing the invention

[0025] Hereinafter, embodiments according to the present invention will be described in detail with reference to the attached drawings. The structure of the present invention and the resulting effects will be clearly understood through the following detailed description.

[0026] Prior to the detailed description of the present invention, it should be noted that identical components are indicated by the same reference numerals whenever possible, even if they are shown in different drawings, and that specific descriptions of known components are omitted if it is determined that such descriptions could obscure the essence of the present invention.

[0027] Throughout the specification, when a part is described as "including" a certain component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components.

[0029] Figure 1 shows a schematic configuration of a cleaning device for a plasma chamber according to the present invention.

[0030] As illustrated in FIG. 1, the cleaning device of the plasma chamber according to the present invention is implemented in a plasma chamber (100) where a plasma process is performed.

[0031] The plasma chamber (100) provides a sealed space in which a deposition process is performed on a semiconductor wafer (1). The chamber (100) is generally made of a metal material such as aluminum or stainless steel, and the interior can be maintained in a vacuum state.

[0032] The RF generator (10) generates high-frequency power required for plasma generation. The high-frequency power generated by the RF generator (10) is applied to the upper electrode (110) through a matching network (not shown). The RF generator (10) generally generates high-frequency power with a frequency of 13.56 MHz, and other frequencies (e.g., 27.12 MHz, 60 MHz) may be used as needed.

[0033] The upper electrode (110) is also called a showerhead and receives high-frequency power from an RF generator (10). The upper electrode (110) has a plurality of gas injection holes formed therein and serves to uniformly inject process gas or cleaning gas into the plasma chamber (100). The upper electrode (110) is generally made of aluminum and may have an anodized surface.

[0034] The upper heated plate (120) is positioned opposite the upper electrode (110) and serves as a support on which the wafer (1) is placed. The lower heated plate (120) has a built-in heater to maintain the wafer (1) at a constant temperature and is generally grounded. The gap between the upper electrode (110) and the lower heated plate (120) is generally maintained at a distance of several centimeters, and a plasma region (2) is formed in this space.

[0035] The process gas injection pipe (130) supplies process gas for a deposition process into the plasma chamber (100). A first valve (132) is installed in the process gas injection pipe (130) so as to control the flow rate of the process gas or cut off the supply.

[0036] In the case of the wafer (1) deposition process, silane (SiH4) and nitrogen oxide (N2O) may be used as process gases. Silane and nitrogen oxide can each be injected through separate supply lines, and their flow rates are precisely controlled by a mass flow controller (MFC).

[0037] The cleaning gas injection pipe (140) supplies cleaning gas for a cleaning process into the plasma chamber (100). A second valve (142) is installed in the cleaning gas injection pipe (140) so that the flow rate of the cleaning gas can be controlled or the supply can be cut off.

[0038] Fluorine-based gases such as NF3 (nitrogen trifluoride), CF4 (carbon tetrafluoride), SF6 (sulfur hexafluoride), and C2F6 (hexafluoroethane) can be used as cleaning gases. These fluorine-based gases can effectively decompose byproducts such as silicon dioxide by generating active fluorine radicals in a plasma state.

[0039] The exhaust pipe (150) serves to discharge gas from inside the plasma chamber (100) to the outside. A vacuum pump is connected to the exhaust pipe (150) to maintain the inside of the chamber at a constant pressure and continuously discharge byproduct gas generated during the process.

[0040] The OES sensor (20) is installed on the side or top of the plasma chamber (100) to measure light emitted from the plasma inside the chamber in real time.

[0041] The OES sensor (20) generally observes plasma through an optical window and transmits an optical signal to a spectrometer through an optical fiber. The spectrometer separates the received light by wavelength and measures the light intensity at each wavelength. The OES sensor (20) specifically measures the light intensity at the Si I (288 nm) and O I (777 nm) wavelengths to monitor the degree of decomposition of silicon dioxide byproducts.

[0042] The control unit (30) receives optical spectrum data from the OES sensor (20) and analyzes it to calculate the concentration of the byproduct. The control unit (30) may be implemented as a computer system including a microprocessor, memory, input / output interface, etc.

[0043] The control unit (30) is equipped with a machine learning algorithm and can learn and predict the relationship between optical spectrum data and byproduct concentration.

[0044] In addition, the control unit (30) calculates the optimal RF power based on the byproduct concentration and feeds this back to the RF generator (10) to dynamically adjust the RF power.

[0045] Looking at the plasma process performed in the plasma chamber (100), first, high-frequency power generated from the RF generator (10) is applied to the upper electrode (shower head) (110).

[0046] At this time, a strong alternating electric field is formed between the upper electrode (110) and the lower heated plate (120). Simultaneously, the process gas injected through the process gas injection pipe (130) is evenly sprayed into the plasma chamber (100) through the showerhead (110). In the case of a wafer (1) deposition process, silane (SiH4) and nitrogen oxide (N2O) may be used as the process gas.

[0047] A strong alternating electric field formed within the plasma region (2) causes electrons in the process gas to vibrate rapidly and accelerate, and the accelerated electrons collide with neutral gas atoms or molecules to cause ionization.

[0048] As accelerated electrons repeatedly collide with neutral gas atoms or molecules, an ionized gas state is formed in which electrons, positive ions, neutral particles, and excited atoms are mixed. This state becomes the plasma state. A deposition process on a wafer (1) can be performed using this plasma state within the plasma chamber (100).

[0049] The cleaning device of the plasma chamber according to the present invention can measure the amount of by-products generated during the deposition process in real time, remove a portion of the by-products, and remove the remaining by-products even after the deposition process is completed.

[0050] This explains the process of wafer deposition using silane (SiH4) and nitrogen oxide (N2O) as process gases.

[0051] When silane (SiH4) and nitric oxide (N2O) are sprayed into the plasma region (2) through the showerhead (110), they repeatedly collide with electrons accelerated by a strong alternating electric field, becoming a plasma state and generating radicals (Si*, O*).

[0052] Radicals are highly reactive and adsorb to and move across the surface of the wafer (1), causing a surface reaction that forms a thin film (SiO2) on the wafer surface as shown below, and byproducts (N2, H2) are discharged through a pump.

[0054] Si* + O* → SiO₂(solid)

[0055] SiH₄ + 2N2O → SiO₂ + 2N₂↑ + 2H₂↑

[0057] At this time, silicon dioxide (SiO2) is deposited on the surface of the wafer (1), and at the same time, silicon dioxide is attached to the inner wall of the chamber (100) as a byproduct.

[0058] The cleaning device of the plasma chamber according to the present invention can perform a cleaning process during the deposition process. That is, the first valve (132) can be closed to block the injection of the deposition process gas, and the second valve (142) can be opened to inject a cleaning gas (NF3, CF4, etc.).

[0059] A cleaning process using plasma is performed by applying a preset high-frequency power simultaneously with the injection of cleaning gas. That is, as shown below, solid SiO2 can be removed from the inner wall of the chamber as it is decomposed by the plasma and converted into gaseous SiF4.

[0061] NF₃ → NF* + 2F*

[0062] 3SiO₂ + 4NF₃ → 3SiF₄↑ + 2N₂↑ + 3O₂

[0064] The OES sensor (20) collects light signals emitted as byproducts exposed to plasma generated inside the plasma chamber (100) decompose in real time and analyzes the intensity of specific wavelength ranges.

[0065] In particular, when silicon dioxide decomposes, silicon atoms and oxygen atoms exist in an excited state, and as they fall to a ground state, they emit light of a specific wavelength. Si I exhibits strong luminescence at a wavelength of 288 nm, and O I at a wavelength of 777 nm. The OES sensor (20) can quantitatively determine the decomposition rate and remaining amount of byproducts by measuring the light intensity of these specific wavelengths.

[0066] The control unit (40) inputs spectrum data received from the OES sensor (20) into a machine learning-based analysis algorithm to calculate the degree of accumulation of byproducts, calculates the optimal RF power (P_opt) from this, and feeds it back to the RF generator (10).

[0067] Machine learning algorithms learn from past cleaning process data to identify correlations between specific spectrum patterns and byproduct concentrations. For example, algorithms such as Artificial Neural Networks (ANN), Support Vector Machines (SVM), and Random Forest can be used. The trained model predicts the current concentration of byproducts from real-time input spectrum data and calculates the optimized RF power.

[0068] Silicon dioxide (SiO2) deposited on the surface of the wafer (1) exists stably in a solid state, and even when plasma is applied, the entire surface is hardly decomposed.

[0069] On the other hand, silicon dioxide accumulated on the walls of the chamber (100) can be decomposed relatively more easily because it has a larger exposed surface area than the wafer or is exposed to plasma for a longer period of time. Therefore, the decomposition signal of the byproduct accumulated on the chamber walls may appear larger than the decomposition signal of the byproduct occurring on the wafer surface.

[0070] That is, the silicon dioxide deposited on the wafer surface remains mostly stable, and the Si / O signal detected by the OES sensor (20) mainly reflects fine byproducts decomposed on the chamber surface.

[0071] In this way, by-products adsorbed on the chamber surface during the deposition process are more easily decomposed by plasma and released as Si and O atoms, so the presence and concentration of by-products can be monitored by detecting the decomposed active species in real time using an OES sensor (10).

[0072] At the beginning of the cleaning process, the amount of byproducts is large, so the intensity of the Si I and OI signals is high. As cleaning progresses, the byproducts are gradually removed, and the intensity of the signals decreases. The control unit (30) monitors these signal changes in real time to determine the progress of the cleaning, and can determine that the cleaning is complete when the signal intensity drops below a reference value.

[0074] FIG. 2 is a flowchart schematically illustrating the cleaning control process of a plasma chamber according to the present invention.

[0075] The control unit (30) can quantitatively determine the degree of generation and accumulation of by-products during the process by monitoring the light emission spectrum of the plasma in real time through the OES sensor (20).

[0076] The control unit (30) optimizes plasma energy by automatically calculating and adjusting RF power based on sensing data from the OES sensor (20), thereby maximizing the efficiency of removing by-products from the chamber during the cleaning process. This enables dynamic control based on the plasma state and by-product concentration, providing the effect of stably maintaining process quality.

[0077] With reference to FIG. 2, the RF power control process during cleaning of a plasma chamber according to the present invention will be explained in detail.

[0078] Step 1: Perform deposition process (S100)

[0079] In the deposition process, plasma is formed to deposit silicon dioxide (SiO2) on the wafer, and at the same time, silicon dioxide byproducts are generated on the inner wall of the plasma chamber (100). The deposition process generally proceeds for tens of seconds to several minutes, during which time byproducts with a thickness of tens to hundreds of nanometers can accumulate on the inner wall of the chamber.

[0081] Step 2: Initiation of cleaning process and optical spectrum measurement (S110)

[0082] Subsequently, during the deposition process or after the deposition process is completed, the OES sensor (20) measures the light emission spectrum emitted from the plasma in real time and transmits it to the control unit (30). The light spectrum is generally measured in the wavelength range of 200 to 900 nm, and the intensity at the Si I (288 nm) and O I (777 nm) wavelengths is particularly important.

[0083] The OES sensor (20) can perform tens to hundreds of measurements per second, allowing for precise real-time tracking of changes in byproduct concentration.

[0085] Step 3: Optical spectrum analysis and signal intensity calculation (S120)

[0086] The control unit (30) analyzes the light emission spectrum to calculate the intensity of the Si I (288 nm) signal (ISi) and the intensity of the OI (777 nm) signal (IO). The signal intensity is generally expressed in an arbitrary unit (au), and only the pure emission intensity is extracted with background noise removed.

[0087] The control unit (30) can perform smoothing processing on the measured signal to remove noise and, if necessary, calculate the ratio between signals (ISi / IO) to determine changes in the chemical composition of the byproduct.

[0089] Step 4: Calculation of byproduct concentration (S130)

[0090] The control unit (30) calculates the byproduct concentration using the signal strength and calculates the relative value (Q) by comparing it with a reference value.

[0091] The byproduct concentration can be calculated as follows:

[0092] C = k1 × ISi + k2 × IO

[0093] Here, C is the byproduct concentration, and k1 and k2 are calibration factors. The calibration factors are determined through preliminary experiments and represent the correlation between the actual byproduct thickness or mass and the optical signal intensity.

[0094] The relative value (Q) of the byproduct concentration is calculated as follows:

[0095] Q = C / Cref

[0096] Here, Cref is the reference byproduct concentration. The reference byproduct concentration can be set as the average byproduct concentration under normal process conditions. If the Q value is greater than 1, it indicates a state where a large amount of byproduct has accumulated, and if it is less than 1, it indicates a state where a small amount of byproduct has accumulated.

[0098] Step 5: Calculate optimal RF power (S140)

[0099] Based on this, the control unit (30) can calculate the optimal RF power (Popt) according to the following mathematical formula 1.

[0100]

[0101] Here, P base is the reference power, Q is the relative value of the byproduct concentration, and α is the adjustment coefficient.

[0102] Reference power (P base The power is generally set in the range of 800 to 1500 W and can be adjusted according to the size of the chamber and the type of cleaning gas. The adjustment factor (α) indicates the sensitivity of RF power to changes in byproduct concentration and is generally set in the range of 0.3 to 0.8. The larger the value of α, the greater the range of change in RF power due to changes in byproduct concentration.

[0103] For example, P base For the case where = 1000 W, α = 0.5, Q = 1.5:

[0104] P opt = 1000 × (1 + 0.5 × 1.5) = 1750 W

[0105] In this way, when a large amount of byproducts accumulate (Q > 1), the RF power is increased to increase the decomposition rate. Conversely, when a small amount of byproducts accumulate (Q < 1), the RF power is reduced to save energy and prevent damage to chamber components.

[0107] Step 6: RF Power Feedback and Adjustment (S150)

[0108] The control unit (30) feeds back the calculated optimal RF power (Popt) to the RF generator (10) in real time. The RF generator (10) adjusts the output power according to the power setting value received from the control unit (30). The power adjustment is generally completed within 1 second and can be ramped to prevent plasma instability caused by rapid power changes.

[0110] Step 7: Determine cleaning completion (S160)

[0111] The control unit (30) continuously monitors the OES signal during the cleaning process and determines that the cleaning is complete when the byproduct concentration decreases to below a set target value. Generally, the cleaning can be determined to be complete when the Q value drops to 0.1 to 0.2 or lower.

[0112] After cleaning is complete, the control unit (30) stops the RF power supply and blocks the injection of cleaning gas. The inside of the chamber is prepared for the next deposition process by exhausting residual gas through a vacuum pump.

[0114] In another embodiment of the present invention, the degree of accumulation of by-products can be monitored in real time through the OES sensor (20) even before the cleaning process begins while the deposition process is in progress. Since plasma is generated even during the deposition process, fine decomposition signals can be generated from the by-products accumulated on the inner wall of the chamber.

[0115] The control unit (30) calculates the cumulative value of the Si I and OI signals measured during the deposition process and predicts the accumulation rate of byproducts. If it is determined that the accumulation rate of byproducts exceeds a threshold value, the control unit (30) can immediately suspend the deposition process and switch to a cleaning process. When the process is switched to a cleaning process, the process (S110 to S150) illustrated in FIG. 2 can then proceed sequentially.

[0116] This real-time monitoring and adaptive cleaning method prevents the excessive accumulation of by-products in advance, thereby minimizing the risk of particle generation and improving process stability.

[0117] The cleaning gas used in the present invention is most commonly NF3, but various fluorine-based gases such as CF4, SF6, and C2F6 may be used depending on process conditions and the type of byproduct.

[0118] NF3 is the most widely used cleaning gas due to its high decomposition efficiency and low global warming potential (GWP). CF4 has lower decomposition efficiency than NF3 but can form a more stable plasma. SF6 has strong etching capabilities and is effective for removing thick byproducts.

[0119] The control unit (30) can select the optimal cleaning gas according to the type and amount of byproduct and apply an RF power control algorithm suitable for each gas.

[0120] The plasma chamber cleaning device of the present invention can be applied not only to silicon dioxide deposition processes but also to the removal of byproducts generated in various thin film deposition processes, such as silicon nitride (SiN), aluminum oxide (Al2O3), and titanium oxide (TiO2).

[0121] Depending on the type of byproduct, the wavelength range monitored by the OES sensor (20) may vary, and the control unit (30) selects and applies an analysis algorithm and an RF power control algorithm suitable for the type of byproduct.

[0122] In addition, the present invention can be applied not only to PECVD equipment but also to various plasma process equipment such as etching equipment and sputtering equipment.

[0124] The above description is merely illustrative of the present invention, and various modifications may be made by those skilled in the art without departing from the technical spirit of the present invention.

[0125] Accordingly, the embodiments disclosed in the specification of the present invention are not intended to limit the present invention. The scope of the present invention should be interpreted by the claims below, and all technology within the equivalent scope should also be interpreted as being included within the scope of the present invention. Explanation of the symbols

[0126] 1: Wafer 2: Plasma region 10: RF Generator 20: OES Sensor 30: Control unit 100: Plasma chamber 110: Upper electrode (shower head) 120: Heated plate 130: Process gas injection pipe 132: First valve 140: Cleaning gas injection pipe 142: Second valve 150: Exhaust pipe

Claims

Claim 1 A cleaning device for a plasma chamber comprising: an RF generator that supplies high-frequency power; a plasma chamber that generates plasma by receiving high-frequency power from the RF generator; an upper electrode installed within the plasma chamber to which high-frequency power is applied; a lower heated plate disposed opposite to the upper electrode and on which a wafer is placed; a process gas injection pipe for injecting process gas for a deposition process into the plasma chamber; a cleaning gas injection pipe for injecting cleaning gas for a cleaning process into the plasma chamber; an OES (Optical Emission Spectroscopy) sensor that measures the optical spectrum emitted from the plasma inside the plasma chamber in real time; and a control unit that analyzes the optical spectrum data received from the OES sensor to extract the intensity of the Si I (288 nm) signal and the intensity of the OI (777 nm) signal from the optical spectrum, calculates the relative value (Q) of the byproduct concentration by comparing the signal intensity with a reference value, calculates the RF power based on the relative value of the byproduct concentration, and feeds back to the RF generator. Claim 2 In claim 1, the control unit, based on the relative value (Q), [describes] the RF power (P) according to the following mathematical formula opt A cleaning device for a plasma chamber characterized by calculating ).P opt = P base × (1 + α×Q) where, P base is the reference power, α is the adjustment coefficient, and Q is the relative value of the byproduct concentration. Claim 3 A cleaning device for a plasma chamber according to claim 1, wherein the control unit learns the correlation between the optical spectrum data and the byproduct concentration using a machine learning algorithm and predicts the byproduct concentration in real time through the learned model. Claim 4 A cleaning device for a plasma chamber comprising: an OES (Optical Emission Spectroscopy) sensor that measures the optical spectrum emitted from the plasma inside the plasma chamber in real time; and a control unit that analyzes the optical spectrum data received from the OES sensor to extract the intensity of the Si I (288 nm) signal and the intensity of the OI (777 nm) signal from the optical spectrum, calculates the relative value (Q) of the byproduct concentration by comparing the signal intensity with a reference value, and calculates the RF power applied to the plasma chamber based on the relative value of the byproduct concentration to control the operation of the RF generator. Claim 5 A method for cleaning a plasma chamber performed in a cleaning device for a plasma chamber, comprising: a step of blocking the injection of a process gas and injecting a cleaning gas during or after the completion of a deposition process within the plasma chamber; a step of generating plasma with the cleaning gas by applying reference RF power to decompose by-products on the inner wall of the chamber; a step of measuring the optical spectrum emitted from the plasma in real time through an OES sensor; and a step of extracting the intensity of a Si I (288 nm) signal and the intensity of an O I (777 nm) signal from the optical spectrum, calculating a relative value (Q) of by-product concentration by comparing the signal intensity with a reference value, and calculating RF power based on the relative value of the by-product concentration to adjust the RF power of an RF generator. Claim 6 delete

Images