RF sensor and method for performing plasma state monitoring
The RF sensor with AI algorithm accurately monitors plasma states by processing electromotive force data, addressing inaccuracies in existing methods and enhancing process control.
Patent Information
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2025-09-29
- Publication Date
- 2026-04-08
AI Technical Summary
Existing plasma state monitoring in semiconductor manufacturing processes is inaccurate, leading to frequent misjudgments of operational states and reduced productivity.
An RF sensor equipped with an artificial intelligence algorithm that processes induced electromotive force data through Fourier transforms and learning to predict plasma states, with verification and retraining capabilities to ensure accuracy.
Enhances the accuracy of plasma state monitoring, reducing misjudgments and improving productivity by providing real-time predictive data and process control.
Smart Images
Figure 2026060951000001_ABST
Abstract
Description
Technical Field
[0005] , ,
[0001] The present invention relates to an RF sensor for performing plasma state monitoring and a method thereof.
Background Art
[0002] Generally, in the manufacturing process of semiconductor devices, plasma equipment that utilizes plasma generated through high-frequency power to perform etching, deposition, etc. of semiconductor substrates is widely used. Various sensors are attached to such plasma equipment, and based on the sensing data obtained from the sensors, the presence or absence of operations and the operating states of the plasma equipment or accessories are confirmed.
[0003] Currently, the sensing data obtained by sensors attached to plasma equipment is analyzed to determine the start and end points of the plasma process, the presence or absence of functional abnormalities such as the presence or absence of plasma generation, the end point of wafer etching, and the presence or absence of removal of process by-products generated inside the plasma equipment. However, the accuracy of this is very low, so the process may stop during the plasma process, be judged as abnormal operation even during normal operation, or be judged as normal operation even during abnormal operation frequently occur, resulting in a problem of reduced productivity.
[0004] Therefore, there is a growing need for the development of technology that can more accurately confirm the plasma state or plasma process state and solve the problem of reduced productivity.
Summary of the Invention
Problems to be Solved by the Invention
[0005] To address these conventional problems, an embodiment of the present invention provides an RF sensor and method for performing plasma state monitoring, which can monitor the plasma state by learning an artificial intelligence algorithm mounted on an RF sensor using sensing data acquired by at least one RF sensor attached to a plasma equipment, and deriving data related to the plasma state based on the learned results. [Means for solving the problem]
[0006] An RF sensor for performing plasma state monitoring according to an embodiment of the present invention is characterized by including a collection unit that collects induced electromotive force induced during the plasma process as sensing data, and a processor that records the induced electromotive force as a function of time, performs a Fourier transform on the recorded induced electromotive force function to derive an n-th order harmonic amplitude value (where n is a natural number greater than or equal to 1), and applies the derived amplitude value and the plasma equipment setting value confirmed at the time the sensing data was generated to an artificial intelligence algorithm to derive predictive data that can monitor the plasma state and plasma process state.
[0007] Furthermore, the processor is characterized by performing verification against the predicted data.
[0008] Furthermore, the processor is characterized by confirming whether or not the plasma process is canceled or whether or not the artificial intelligence algorithm is being retrained, based on the verification results for the verification.
[0009] Furthermore, the system further includes a communication unit, the communication unit being characterized by transmitting the predicted data, which has been verified by the processor, to an electronic device.
[0010] Furthermore, the method for performing plasma state monitoring according to an embodiment of the present invention is characterized by including the steps of: an RF sensor collecting induced electromotive force induced during the plasma process as sensing data; the RF sensor recording the induced electromotive force as a function of time, performing a Fourier transform on the recorded induced electromotive force function to derive an n-th order harmonic amplitude value (where n is a natural number greater than or equal to 1); and the RF sensor applying the derived amplitude value and the plasma equipment setting value confirmed at the time the sensing data was generated to an artificial intelligence algorithm to derive predictive data that allows monitoring of the plasma state and plasma process state.
[0011] Furthermore, the system is characterized by further including a step in which the RF sensor performs verification against the predicted data.
[0012] Furthermore, the RF sensor is further characterized by including a step of confirming whether the plasma process is canceled or whether the artificial intelligence algorithm is being retrained based on the verified verification results.
[0013] Furthermore, the system is characterized by further including a step in which the RF sensor transmits a message requesting the termination of the plasma process to the electronic device if it is necessary to terminate the plasma process.
[0014] Furthermore, the method is characterized by further including a step of performing retraining of the artificial intelligence algorithm if the RF sensor requires retraining of the artificial intelligence algorithm.
[0015] Furthermore, the RF sensor is characterized by further including a step in which it transmits the verified prediction data to an electronic device. [Effects of the Invention]
[0016] As described above, the RF sensor and method for performing plasma state monitoring according to the present invention have the effect of monitoring the plasma state by learning an artificial intelligence algorithm mounted on the RF sensor with sensing data acquired by at least one RF sensor attached to the plasma equipment, and deriving data related to the plasma state based on the learned results. [Brief explanation of the drawing]
[0017] [Figure 1] This is a drawing showing a system including an RF sensor for performing plasma state monitoring according to an embodiment of the present invention. [Figure 2] This is a flowchart illustrating a method for monitoring the plasma state using an RF sensor according to an embodiment of the present invention. [Modes for carrying out the invention]
[0018] The following is information on the national research and development projects that supported this invention. [Project-Specific Number] 1711203519 [Issue Number] CRC20014-000 [Department Name] Department of Science, Technology and ICT [Project Management (Specialized) Organization Name] National Science and Technology Research Association [Research Project Name] National Science and Technology Research Association Research Operating Expense Support (Major Project Expenses) - Future-Leading Integrated Research Project [Research Project Title] Development and Demonstration of Intelligent Technology for Semiconductor Plasma Process Equipment [Project Implementation Organization] Korea Fusion Energy Research Institute [Research period] November 1, 2020 to October 31, 2026
[0019] Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. The detailed description disclosed below together with the accompanying drawings is intended to illustrate exemplary embodiments of the present invention, and is not intended to show the only embodiments in which the present invention can be implemented. Parts not related to the description can be omitted in order to clearly explain the present invention in the drawings, and the same reference numerals can be used for the same or similar components throughout the specification.
[0020] FIG. 1 is a drawing showing a system including an RF sensor that performs plasma state monitoring according to an embodiment of the present invention.
[0021] Referring to FIG. 1, a system 10 according to the present invention can include a plurality of RF sensors 100 and an electronic device 200.
[0022] Each of the plurality of RF sensors 100 can include a communication unit 110, a collection unit 120, a processor 130, and a memory 140. In addition, in the embodiment of the present invention, one RF sensor will be described, but it is to clarify that this applies to all of the plurality of RF sensors 100. The RF sensor 100 is a sensor that can sense a signal having a radio frequency (RF), and can include a coil. At this time, the coil is a wire member that forms one or more closed surfaces of the conductor, and can mean a wire member in which an electromotive force is induced so that an electric current can flow through the conductor when a change occurs in the magnetic field passing through the closed surface formed by the conductor.
[0023] At least one RF sensor 100 is arranged outside an inductively coupled plasma (ICP) generator including an antenna, that is, plasma equipment. The RF sensor 100 is not limited as long as it is arranged around the antenna, particularly at a position where an inductive electromotive force can be induced above the antenna. For example, a spiral antenna can be formed on one plane above the plasma equipment, and the RF sensor 100 can be arranged perpendicular to the plane on which the spiral antenna is formed.
[0024] The communication unit 110 transmits result data derived by the processor 130 to the electronic device 200 in order to monitor the plasma state and plasma process state through communication with the electronic device 200. For this purpose, the communication unit 110 can communicate with the electronic device 200 using Wi-Fi (wireless fidelity), Bluetooth, BLE (Bluetooth low energy), etc.
[0025] The data collection unit 120 collects the induced electromotive force induced in the RF sensor 100 during the plasma process as sensing data. At this time, the data collection unit 120 can measure the induced electromotive force by first measuring physical quantities such as current, which are not the induced electromotive force, and then calculating the electromotive force. The plasma equipment can generate plasma inside the equipment through the current applied to the antenna and changes in the current, and changes in the surrounding magnetic field may occur from the antenna. Therefore, the data collection unit 120 can collect the induced electromotive force induced from the changes in the magnetic field as sensing data.
[0026] The processor 130 performs preprocessing of the sensing data. More specifically, the processor 130 can perform a Fourier transform on the induced electromotive force collected by the acquisition unit 120 from a function of time of the induced electromotive force. The Fourier transform can decompose the function of time into frequency components, which can be numbered in order from the most prominent frequency component to the first harmonic, then the second harmonic. In the context of this specification, harmonic refers to the result of the Fourier transform, where the nth harmonic refers to the nth harmonic component in order of prominence of the frequency components shown in the Fourier transform result, and the amplitude value of the nth harmonic refers to the amplitude value of the harmonic wave.
[0027] In this way, the processor 130 can perform preprocessing by recording the induced electromotive force as a function of time, performing a Fourier transform on the recorded induced electromotive force function to separate each frequency component, and deriving the amplitude value of each frequency component or a specific nth-order (where n is a natural number greater than or equal to 1) harmonic. At this time, the value of n can be used to determine how close the main frequency component of the induced electromotive force function as a function of time is to the harmonic frequency.
[0028] The processor 130 applies the derived amplitude values and the set values used when acquiring sensing data to the artificial intelligence algorithm to perform learning of the artificial intelligence algorithm. At this time, the set values may include the plasma state (e.g., electron density and electron temperature) and the plasma process state (e.g., processes such as etching processes) when the sensing data was acquired.
[0029] The processor 130 can adjust the number of training iterations, the size of the sensing data load, the number of layers, etc., to improve training accuracy. It can also provide a function to repeatedly input settings within a certain range to find the settings that produce the best result. In this case, the settings within a certain range may refer to the values set through the GridSearchCV technique, which searches for training conditions that can most accurately predict the result value for training.
[0030] The processor 130 inputs test data into the artificial intelligence algorithm, checks the prediction accuracy and the time spent on prediction, and can use only the learning results that match the accuracy and response time set by the user of the RF sensor 100. At this time, the processor 130 can set one or more learning execution results that match the accuracy and response time set by the user. Through this, the predicted values are continuously logged as learning is repeated, and the artificial intelligence algorithm with the highest accuracy can be selected. The processor 130 stores the completed artificial intelligence algorithm in memory 140.
[0031] The processor 130 applies pre-processed amplitude values to the trained artificial intelligence algorithm to generate predictive data for the plasma state and plasma process state. At this time, the predictive data may include the amplitude values applied to the artificial intelligence algorithm, the prediction time, the artificial intelligence algorithm used for prediction, the prediction results (plasma state, plasma process state), and actual measured values.
[0032] The processor 130 compares the actual measurement results obtained during the inspection process performed after the plasma process is actually completed with the predicted data predicted by the processor 130, and performs verification of the predicted data. At this time, the processor 130 can compare the predicted data with the actual measurement results periodically or in real time.
[0033] If the error between the predicted data and the actual measurement result exceeds a critical number of times beyond the tolerance range included in the pre-set verification information, the processor 130 can confirm that a process abort is necessary and transmit a message to the electronic device 200 informing it that the plasma process must be aborted.
[0034] The processor 130 can determine that the artificial intelligence algorithm needs to be retrained if the error between the predicted data and the actual measurement result exceeds the permissible error range included in the pre-set verification information by a critical number of cycles. At this time, the criteria for determining whether or not to stop the plasma process and the criteria for determining whether or not to retrain the artificial intelligence algorithm may differ. If the processor 130 determines that the artificial intelligence algorithm needs to be retrained, it proceeds with retraining the artificial intelligence algorithm.
[0035] Furthermore, if the artificial intelligence algorithm does not require retraining, or if retraining is complete, the processor 130 applies the pre-processed amplitude values to the artificial intelligence algorithm trained by the RF sensor 100 and transmits the resulting predicted data for the plasma state and plasma process state to the electronic device 200. Through this, the electronic device 200 can monitor the plasma state and plasma process state predicted by the RF sensor 100.
[0036] Memory 140 stores the operating program for operating the RF sensor 100. More specifically, memory 140 can store artificial intelligence algorithms learned by the processor 130.
[0037] The electronic device 200 is a device capable of controlling the plasma equipment (not shown) through communication with the plasma equipment, and may be an electronic device such as a computer, laptop computer, or tablet PC. The electronic device 200 displays predictive data received from at least one RF sensor 100 so that the user can check the plasma state and plasma process state. The electronic device 200 also controls the operation of the plasma equipment based on the predictive data received from the RF sensor 100.
[0038] More specifically, the electronic device 200 can plan parameter values to be controlled during process progression in the plasma equipment based on predictive data received from at least one RF sensor 100. When the plasma state value confirmed based on the predictive data differs from a preset range by more than a critical value, the electronic device 200 can receive input from the user regarding the changed items of the plasma state, the amount of change, process parameters, number of process steps, and process execution time.
[0039] When the plasma process state confirmed based on the prediction data differs from a preset range of process states by more than a critical value, the electronic device 200 can receive input from the user regarding the process impact related to the change item that caused the change in the plasma state.
[0040] The electronic device 200 can generate control combinations based on control factors that can be adjusted using parameter control combination conditions, which have information on items that can be adjusted in at least one of the following processes: a process in which the confirmed predicted plasma state value deviates by a critical value from a preset range of plasma state values, and a process in which the confirmed process state deviates by a critical value from a preset range of process states.
[0041] The process control combination can learn by applying the sensing data acquired during the executed process, along with the predicted plasma state values and plasma process states corresponding to the sensing data, to a reinforcement learning algorithm, which is an artificial intelligence algorithm.
[0042] Once the electronic device 200 has completed predictions for process control combinations using a reinforcement learning algorithm, it performs a process impact evaluation for the predicted control combinations. The process impact evaluation determines whether the control combinations fall within the range defined by the pre-set parameter control constraints. If they do, the device can save the parameter control recommendation combination and parameter adjustment amount control information to be used as data for more stable control of the plasma equipment.
[0043] The electronic device 200 can control parameter values for the plasma equipment based on a set process control combination.
[0044] Figure 2 is a flowchart illustrating a method for monitoring the plasma state using an RF sensor according to an embodiment of the present invention.
[0045] In step 201, the processor 130 performs step 203 when it receives a start signal to monitor the plasma state and plasma process state of the plasma equipment, and waits for a start signal to be received if it does not. At this time, the start signal may be an activation signal of the RF sensor 100 received from the electronic device 200 via the communication unit 110.
[0046] In step 203, the collection unit 120 collects sensing data. At this time, the sensing data may represent the induced electromotive force induced in the RF sensor 100 during the plasma process. The collection unit 120 can also measure the induced electromotive force by measuring physical quantities such as current and then calculating the electromotive force.
[0047] In step 205, the processor 130 performs preprocessing of the sensing data. More specifically, the processor 130 records the induced electromotive force collected by the acquisition unit 120 as a function of time, performs a Fourier transform on the recorded induced electromotive force function to identify component frequencies, and derives the amplitude values of the identified component frequencies. In this way, the processor 130 can perform preprocessing by deriving the amplitude values of the nth order (where n is a natural number greater than or equal to 1) harmonic.
[0048] In step 207, the processor 130 applies the derived amplitude values and the setpoints used when acquiring sensing data to the artificial intelligence algorithm to perform learning of the artificial intelligence algorithm. At this time, the setpoints may include the plasma state (e.g., electron density and electron temperature) and the plasma process state (e.g., processes such as etching processes) when the sensing data was acquired.
[0049] The processor 130 can adjust the number of training iterations, the size of the sensing data load, the number of layers, etc., to improve training accuracy. It can also provide a function to repeatedly input settings within a certain range to find the settings that produce the best result. In this case, the settings within a certain range may refer to the values set through the GridSearchCV technique, which searches for training conditions that can most accurately predict the result value for training.
[0050] The processor 130 inputs test data into the artificial intelligence algorithm, checks the prediction accuracy and the time spent on prediction, and can use only the learning results that match the accuracy and response time set by the user of the RF sensor 100. At this time, the processor 130 can set one or more learning execution results that match the accuracy and response time set by the user. Through this, the predicted values are continuously logged as learning is repeated, and the artificial intelligence algorithm with the highest accuracy can be selected. The processor 130 saves the completed artificial intelligence algorithm in memory 140.
[0051] In step 209, the processor 130 applies pre-processed amplitude values to the learned artificial intelligence algorithm to generate predictive data for the plasma state and plasma process state. At this time, the predictive data may include the amplitude values applied to the artificial intelligence algorithm, the prediction time, the artificial intelligence algorithm used for prediction, the prediction results (plasma state, plasma process state), and actual measured values.
[0052] In step 211, the processor 130 compares the actual measurement results obtained during the inspection process performed after the plasma process is actually completed with the predicted data predicted by the processor 130, and performs verification of the predicted data. At this time, the processor 130 can compare the predicted data with the actual measurement results periodically or in real time.
[0053] At step 213, the processor 130 executes step 215 if the verification results indicate that the plasma process needs to be stopped, and executes step 217 if the plasma process does not need to be stopped. More specifically, the processor 130 can confirm that the process needs to be stopped if the error between the predicted data and the actual measurement result exceeds the range of tolerance included in the pre-set verification information by a critical number of times.
[0054] At step 215, the processor 130 can generate a message indicating that the plasma process needs to be stopped and transmit it to the electronic device 200.
[0055] Conversely, at step 217, the processor 130 checks whether or not the artificial intelligence algorithm needs to be retrained. More specifically, the processor 130 can determine that the artificial intelligence algorithm needs to be retrained if the error between the predicted data and the actual measurement result exceeds the range of acceptable error included in the pre-set verification information by a critical number of steps. At this time, the criteria for determining whether or not to stop the plasma process and the criteria for determining whether or not the artificial intelligence algorithm needs to be retrained may differ.
[0056] If, after 217 steps of verification, it is determined that the artificial intelligence algorithm needs to be retrained, the processor 130 proceeds to step 219. If it is determined that retraining is not necessary, it proceeds to step 221. At step 219, the processor 130 performs the retraining of the artificial intelligence algorithm.
[0057] Conversely, if the artificial intelligence algorithm does not need to be retrained, or if retraining is complete, in step 221, the processor 130 applies the pre-processed amplitude values to the artificial intelligence algorithm that has been trained by the RF sensor 100 and transmits the predicted plasma state and plasma process state to the electronic device 200. Through this, the electronic device 200 can monitor the plasma state and plasma process state predicted by the RF sensor 100.
[0058] The embodiments of the present invention disclosed herein and in the drawings are merely examples provided to facilitate the technical content of the invention and to aid in its understanding, and are not intended to limit the scope of the invention. Therefore, the scope of the invention should be interpreted as including all modifications or variations derived from the technical idea of the invention, in addition to the embodiments disclosed herein.
Claims
1. A data collection unit that collects induced electromotive force generated during the plasma process using sensing data; and An RF sensor for performing plasma state monitoring, comprising: a processor that records the induced electromotive force as a function of time, performs a Fourier transform on the recorded induced electromotive force function to derive an n-th order harmonic amplitude value (where n is a natural number greater than or equal to 1), and applies the derived amplitude value and the plasma equipment setting value confirmed at the time of sensing data generation to an artificial intelligence algorithm to derive predictive data that can monitor the plasma state and plasma process state.
2. The aforementioned processor, An RF sensor for performing plasma state monitoring according to claim 1, characterized by performing verification of the aforementioned predicted data.
3. The aforementioned processor, An RF sensor for performing plasma state monitoring according to claim 2, characterized in that it confirms whether or not the plasma process is discontinued or whether or not the artificial intelligence algorithm is being retrained based on the verification results for the verification.
4. Communications Department; It further includes, The aforementioned communications unit is An RF sensor for performing plasma state monitoring according to claim 3, characterized in that it transmits the predicted data verified by the processor to an electronic device.
5. The RF sensor collects the induced electromotive force generated during the plasma process as sensing data; The RF sensor records the induced electromotive force as a function of time, and performs a Fourier transform on the recorded induced electromotive force function to derive the amplitude value of an n-th order harmonic (where n is a natural number greater than or equal to 1); and A method for performing plasma state monitoring, characterized by including the step of applying the derived amplitude value and the plasma equipment setting value confirmed at the time the sensing data was generated to an artificial intelligence algorithm to derive predictive data that can monitor the plasma state and plasma process state.
6. A method for performing plasma state monitoring according to claim 5, further comprising the step of the RF sensor performing verification against the predicted data.
7. A method for performing plasma state monitoring according to claim 6, further comprising the step of the RF sensor confirming whether the plasma process is discontinued or whether the artificial intelligence algorithm is being retrained based on the verified verification results.
8. A method for performing plasma state monitoring according to claim 7, further comprising the step of transmitting a message to an electronic device requesting the termination of the plasma process if the RF sensor indicates that it is necessary to terminate the plasma process.
9. A method for performing plasma state monitoring according to claim 7, further comprising the step of performing retraining of the artificial intelligence algorithm if the RF sensor requires retraining of the artificial intelligence algorithm;
10. A method for performing plasma state monitoring according to claim 7, further comprising the step of the RF sensor transmitting the verified prediction data to an electronic device.