Radio frequency power adjusting system, adjusting method thereof and semiconductor process equipment

Abstract

The invention discloses a radio frequency power adjusting system, an adjusting method thereof and semiconductor process equipment. The adjusting system comprises an interference monitoring module which obtains a time domain signal of radio frequency power in a reaction cavity and obtains interference data based on frequency domain characteristics of the time domain signal; the adaptive control module takes the interference data as a state input variable, updates an action output variable by learning an optimal strategy, and determines an optimal action output variable based on an objective function of interference suppression and energy consumption, and the action output variable comprises a frequency regulation variable, an amplitude regulation variable and a pulse width regulation variable; and the feedback adjusting module is used for dynamically adjusting a plurality of process parameters in the reaction cavity based on the optimal action output variable so as to enable the process parameters to reach the target process parameter range. According to the invention, the radio frequency power density in the reaction cavity can be adaptively adjusted, and the optimal dynamic balance between interference suppression and energy consumption is realized, so that the radio frequency interference suppression effect, the process stability and the equipment efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of semiconductor manufacturing, and in particular to a radio frequency power regulation system, a radio frequency power regulator, a semiconductor process equipment, and a computer-readable storage medium. Background Art

[0002] In semiconductor manufacturing, the RF power module of the Plasma Enhanced Chemical Vapor Deposition (PECVD) equipment is the core component for plasma generation and maintenance. However, the RF power output process is often affected by various interference sources, including: high-frequency interference, environmental interference and process disturbances. High-frequency interference is that the RF power output will cause electromagnetic coupling and harmonic interference in the cavity, which will cause the plasma in the reaction chamber to be unstable. Environmental interference is that the electromagnetic noise generated by external equipment may be coupled into the PECVD equipment, thereby interfering with the output stability of the RF power. Process disturbances refer to the fluctuation of gas flow in the reaction chamber or the change of gas pressure in the cavity. The electron density, ion density and collision frequency of the plasma will change accordingly, causing the plasma impedance to change accordingly, which will affect the load characteristics of the RF power.

[0003] In the prior art, since the RF power module usually relies on fixed parameters to operate, it is unable to adapt to interference and process disturbances in real time, which easily leads to poor electromagnetic compatibility (EMC) of process equipment, thereby affecting equipment stability, and further, it is easy to cause fluctuations in the plasma state in the reaction chamber, thereby reducing the quality of thin film deposition. In addition, since the RF power module cannot respond to interference sources in a timely manner, the equipment maintenance cost will also increase.

[0004] In order to solve the above-mentioned problems existing in the prior art, the art is in urgent need of a RF power regulation technology that can adaptively adjust the RF power density in the reaction chamber to achieve an optimal dynamic balance between interference suppression and energy consumption, thereby improving the RF interference suppression effect of process equipment, ensuring the stability of the plasma process in the reaction chamber, and improving the process efficiency of the equipment and the quality of thin film deposition. Summary of the invention

[0005] A brief summary of one or more aspects is given below to provide a basic understanding of these aspects. This summary is not an exhaustive overview of all conceived aspects, and is neither intended to identify the key or decisive elements of all aspects nor to define the scope of any or all aspects. Its only purpose is to give some concepts of one or more aspects in a simplified form as a prelude to a more detailed description that will be given later.

[0006] In order to overcome the above-mentioned defects of the prior art, the present invention provides a radio frequency power regulation system, a radio frequency power regulation method, a semiconductor process equipment, and a computer-readable storage medium, which can adaptively adjust the radio frequency power density in the reaction chamber to achieve the optimal dynamic balance between interference suppression and energy consumption, thereby improving the radio frequency interference suppression effect of the process equipment, ensuring the stability of the plasma process in the reaction chamber, and improving the process efficiency and thin film deposition quality of the equipment.

[0007] Specifically, the above-mentioned RF power regulation system provided according to the first aspect of the present invention includes: an interference monitoring module, which obtains the time domain signal of the RF power in the reaction chamber, and obtains interference data based on the frequency domain characteristics of the time domain signal; an adaptive control module, which uses the interference data as a state input variable, updates the action output variable by learning the optimal strategy, and determines the optimal action output variable based on the objective function of interference suppression and energy consumption, wherein the action output variable includes the frequency adjustment amount, amplitude adjustment amount and pulse width adjustment amount of the RF power; and a feedback adjustment module, which dynamically adjusts the multiple process parameters in the reaction chamber based on the optimal action output variable so that the multiple process parameters reach the target process parameter range.

[0008] In addition, the above-mentioned RF power adjustment method provided according to the second aspect of the present invention includes the following steps: obtaining a time domain signal of the RF power in the reaction chamber, and obtaining interference data based on the frequency domain characteristics of the time domain signal; using the interference data as a state input variable, updating the action output variable by learning the optimal strategy; determining the optimal action output variable based on the objective function of interference suppression and energy consumption; and dynamically adjusting multiple process parameters in the reaction chamber based on the optimal action output variable so that the multiple process parameters reach the target process parameter range.

[0009] In addition, the semiconductor process equipment provided according to the third aspect of the present invention includes: a reaction chamber, in which wafers are placed for performing a thin film deposition process; an RF power module; for providing RF power; and the RF power adjustment system provided by the first aspect of the present invention, connected to the RF power module, for dynamically adjusting multiple process parameters in the reaction chamber according to the RF power corresponding to the optimal action output variable, so that the multiple process parameters reach the target process parameter range.

[0010] In addition, according to the fourth aspect of the present invention, there is also provided a computer-readable storage medium on which computer instructions are stored. When the computer instructions are executed by a processor, the above-mentioned radio frequency power adjustment method provided by the second aspect of the present invention is implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The above features and advantages of the present invention can be better understood after reading the detailed description of the embodiments of the present disclosure in conjunction with the following drawings. In the drawings, the components are not necessarily drawn to scale, and components with similar related properties or features may have the same or similar reference numerals.

[0012] Figure 1 A structural block diagram of a radio frequency power regulation system provided according to some embodiments of the present invention is shown; and

[0013] Figure 2 A flow chart of a method for adjusting radio frequency power provided according to some embodiments of the present invention is shown;

[0014] Figure 3A A schematic diagram of fuzzy control rules for adjusting interference intensity and power amplitude in fuzzy logic control according to some embodiments of the present invention is shown; and

[0015] Figure 3B A schematic diagram of fuzzy control rules for interference intensity, frequency adjustment and amplitude adjustment in fuzzy logic control provided according to some embodiments of the present invention is shown.

[0016] Reference numerals:

[0017] 100 RF power regulation system;

[0018] 110 interference monitoring module;

[0019] 120 adaptive control module;

[0020] 130 feedback regulation module; and

[0021] Steps S210 to S240. DETAILED DESCRIPTION

[0022] The following specific embodiments illustrate the implementation of the present invention, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. Although the description of the present invention will be introduced in conjunction with the preferred embodiment, this does not mean that the features of this invention are limited to this implementation. On the contrary, the purpose of introducing the invention in conjunction with the implementation is to cover other options or modifications that may be extended based on the claims of the present invention. In order to provide a deep understanding of the present invention, the following description will include many specific details. The present invention can also be implemented without using these details. In addition, in order to avoid confusion or blurring the focus of the present invention, some specific details will be omitted in the description.

[0023] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0024] In addition, the terms "upper", "lower", "left", "right", "top", "bottom", "horizontal" and "vertical" used in the following description should be understood as the directions shown in the paragraph and the related drawings. Such relative terms are only used for the convenience of description and do not mean that the device described therein must be manufactured or operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0025] It is understood that although the terms "first", "second", "third", etc. may be used herein to describe various components, regions, layers and / or parts, these components, regions, layers and / or parts should not be limited by these terms, and these terms are only used to distinguish different components, regions, layers and / or parts. Therefore, the first component, region, layer and / or part discussed below may be referred to as a second component, region, layer and / or part without departing from some embodiments of the present invention.

[0026] As mentioned above, in the prior art, since the RF power module usually relies on fixed parameters to operate, it is unable to adapt to interference and process disturbances in real time, which easily leads to poor electromagnetic compatibility (EMC) of process equipment, thereby affecting equipment stability, and further, easily leads to fluctuations in the plasma state in the reaction chamber, thereby reducing the quality of thin film deposition. Moreover, since the RF power module cannot respond to interference sources in a timely manner, the equipment maintenance cost will also increase.

[0027] In order to solve the above-mentioned problems existing in the prior art, the present invention provides a radio frequency power regulation system, a radio frequency power regulation, a semiconductor process equipment, and a computer-readable storage medium, which can adaptively adjust the radio frequency power density in the reaction chamber to achieve the optimal dynamic balance between radio frequency interference suppression and energy consumption, thereby improving the radio frequency interference suppression effect of the process equipment, ensuring the stability of the plasma process in the reaction chamber, and improving the process efficiency of the equipment.

[0028] In some non-limiting embodiments, the RF power regulation system provided in the first aspect of the present invention can be configured in the semiconductor process equipment provided in the third aspect of the present invention, and used to implement the RF power regulation method provided in the second aspect of the present invention.

[0029] Specifically, in some non-limiting embodiments, the computer-readable storage medium provided in the fourth aspect of the present invention stores computer instructions thereon. When the computer instructions are executed by a processor, they can be used to implement the method for adjusting the half RF power provided in the second aspect of the present invention.

[0030] The following will describe the working principle of the above-mentioned RF power regulation system in conjunction with some semiconductor process equipment and embodiments of the RF power regulation method. Those skilled in the art will understand that these embodiments of semiconductor process equipment and the RF power regulation method are only some non-restrictive implementation methods provided by the present invention, which are intended to clearly demonstrate the main concept of the present invention and provide some specific solutions that are convenient for the public to implement, rather than to limit all working modes or all functions of the RF power regulation system. Similarly, the RF power regulation system is also only a non-restrictive implementation method provided by the present invention, and does not constitute a limitation on the various configuration objects in these semiconductor process equipment and the implementation subjects of each step in the RF power regulation method.

[0031] In some embodiments of the present invention, semiconductor process equipment may include: a reaction chamber, which holds a wafer inside and is used for a thin film deposition process; an RF power module; used to provide RF power; and an RF power adjustment system, connected to the RF power module, used to dynamically adjust multiple process parameters in the reaction chamber according to the RF power corresponding to the optimal action output variable, so that the multiple process parameters reach a target process parameter range.

[0032] Specifically, see Figure 1 , Figure 1 A schematic structural diagram of a semiconductor process equipment provided according to some embodiments of the present invention is shown.

[0033] like Figure 1 As shown, in some embodiments of the present invention, the RF power regulation system 100 may mainly include an interference monitoring module 110, an adaptive control module 120, and a feedback regulation module 130. The interference monitoring module 110 may obtain the time domain signal of the RF power in the reaction chamber, and obtain its frequency domain characteristics based on the time domain signal to obtain interference data. The adaptive control module 120 may use the interference data as the state input variable, update the action output variable by learning the optimal strategy, and determine the optimal action output variable based on the objective function of interference suppression and energy consumption, wherein the action output variable includes the frequency adjustment amount, amplitude adjustment amount, and pulse width adjustment amount of the RF power. The feedback regulation module 130 may dynamically adjust multiple process parameters in the reaction chamber based on the optimal action output variable so that the multiple process parameters reach the target process parameter range.

[0034] Furthermore, please combine Figure 2The flowchart of a radio frequency power regulation method provided by some embodiments of the present invention is shown to help understand the various modules in the radio frequency power regulation system 100.

[0035] Combination Figure 1 and Figure 2 As shown, in some embodiments of the present invention, the method for adjusting RF power may include the following steps. First, step S210 may be executed: obtaining a time domain signal of the RF power in the reaction chamber, and obtaining interference data based on the frequency domain characteristics of the time domain signal.

[0036] In some embodiments, the interference monitoring module 110 can be mainly used to collect radio frequency power signals in real time, generate spectrum data, and analyze its output characteristics, so as to monitor the internal and external interference of the reaction chamber.

[0037] Specifically, the interference monitoring module 110 may include a high-frequency probe and a spectrum analyzer. The high-frequency probe can be used to collect the time domain signal of the RF power in the reaction chamber, wherein the time domain signal P(t) may include the main frequency and its harmonic frequencies and signals within other bandwidths. Optionally, the main frequency of the RF power may be 13.56 MHz, the second harmonic frequency may be 27.12 MHz, the third harmonic frequency may be 40.68 MHz, etc. Afterwards, the collected real-time time domain signal may be subjected to data analysis to obtain the input model of the time domain signal, wherein the input model of the time domain signal may be as follows:

[0038]

[0039] Among them, P(t) represents the time domain signal, A(t), ω, Indicates the amplitude, frequency and phase of the main frequency, A n (t) and Indicates the amplitude and phase of each harmonic frequency.

[0040] Then, the interference characteristics can be extracted by a spectrum analyzer and spectrum data can be generated. Specifically, in some preferred embodiments, the collected time domain signal data can be pre-processed by filtering, removing DC bias, selecting a suitable window function (such as Hamming window, Blackman window, etc.), etc., to reduce the influence of edge effects or high-frequency noise on the FFT results.

[0041] After that, the preprocessed time domain signal data can be subjected to Fast Fourier Transform (FFT) to obtain the amplitude and phase information of each frequency component. Through FFT decomposition, the output model of the frequency domain signal of the RF power can be obtained, where the output model of the frequency domain signal is as follows:

[0042]

[0043] Wherein, S(f) represents the spectrum of the frequency domain signal, and 2πf represents the spacing between two spectrum lines of the frequency domain signal. According to the output model of the frequency domain signal, the frequency domain characteristics can be extracted. For example, the power density of the main frequency and harmonics can be extracted to identify the harmonic power density and interference peak.

[0044] Optionally, the power density or energy distribution of the main frequency (13.56 MHz), the second harmonic frequency (27.12 MHz) and the third harmonic frequency (40.68 MHz) in the transformation result and the frequency bands nearby may be focused on.

[0045] Then, the frequency domain characteristics of the amplitude, phase, or power spectrum density at the main frequency point and each harmonic can be extracted, as well as the frequency domain characteristics of the frequency position and amplitude of other possible interference peaks. After that, a threshold judgment can be made. Optionally, a safety threshold or warning threshold can be set for each frequency band by comparing with a reference spectrum or baseline data under normal system conditions. If a frequency band or harmonic peak exceeds the set threshold, it can be marked as a potential interference area.

[0046] In some optional embodiments, after the above spectrum analysis (FFT) of the radio frequency signal, in order to effectively identify and classify various interferences, a series of steps can be performed to correspond the spectrum results with possible interference types. The interference monitoring module can also be configured to: identify the interference source based on the interference data, and classify the identified interference source; and comprehensively evaluate the classification results, and determine the adjustment method according to the interference source.

[0047] Specifically, the spectrum results over a period of time can be dynamically tracked, including cumulative or sliding window analysis, to identify whether the interference is periodic, instantaneous or random. Then, the interference pattern can be identified by combining time series features and frequency domain features through statistical methods or machine learning models (such as PCA, clustering, neural networks, etc.), so as to further determine the source and severity of the interference. The identified interference sources can also be classified, and then the classification results can be comprehensively evaluated. Specifically, the frequency position, amplitude, and duration of the interference event can be recorded. The interference type is then associated with the system operating status (process steps, power setting values, etc.). According to different interference types and severity, the corresponding control or alarm strategy is triggered: if it is an internal controllable interference (such as matching network mismatch), automatic tuning or power adjustment is performed; if it is external interference, it can be reported to the upper system, or anti-interference measures (shielding, filtering, isolation, etc.) can be performed.

[0048] Through the above process executed by the interference monitoring module 110, various types of interference can be identified and classified according to their characteristic frequency, amplitude, occurrence mode and other information after spectrum analysis. On this basis, the RF power adjustment system 100 can take corresponding compensation or suppression measures for different interference sources to ensure the stability of RF output and plasma process.

[0049] Continue as Figure 2 As shown, step S220 may be performed afterwards: using the interference data as the state input variable, and updating the action output variable by learning the optimal strategy.

[0050] Combination Figure 1 As shown, the adaptive control module 120 can be mainly used to analyze the interference characteristics of the RF power signal, optimize the adjustable parameters of the RF power, and then output the optimized RF signal to suppress interference in real time.

[0051] The adaptive control module 120 may include a data processing unit (CPU) and a control interface. The data processing unit may be used to run an adaptive algorithm, including a reinforcement learning algorithm and a fuzzy logic control algorithm. The control interface may be used to adjust the RF output (frequency, amplitude, pulse width). The control interface may be used to adjust the RF output, including frequency, amplitude, and pulse width.

[0052] Specifically, in some optional embodiments, the sum of the interference power density values ​​of the main frequency and each harmonic frequency may be obtained as the first reward factor. For example, the interference power density value of the main frequency may be P 13.56 (t), the interference power density value of the second harmonic can be P 27.12 (t), the interference power density value of the third harmonic can be P 40.68 (t), etc. Then, a reward function can be constructed based on at least the first reward factor. In order for the algorithm to obtain "minimized interference", that is, the lower the value of the interference power density, the better, the reward function can be defined as the negative value of the interference power density. The formula of the reward function can be as follows:

[0053]

[0054] Among them, ∑ i P i (t) represents the sum of the interference power density values. In this embodiment, the single-step reward calculation method is adopted, R t It only reflects the degree of interference suppression. Therefore, when the interference power is larger, ∑ i P i The higher (t) is, the overall reward R t The smaller the interference, the smaller the i P i The lower (t) is, the greater the overall reward Rt The bigger it is.

[0055] Optionally, in some other embodiments, based on the above-mentioned “interference suppression” embodiment, “output stability” may be additionally introduced as a second reward factor.

[0056] Specifically, the output fluctuation of RF parameters or the fluctuation of process indicators can be obtained, and statistical calculations (such as standard deviation, variance, absolute change, etc.) can be performed on them, and then converted into a directly usable stability score through appropriate functions (such as inverse, exponential decay, normalization) to obtain the value of output stability. The value of output stability is usually mapped to a range of 0 to 1 or a wider positive number, and the larger the value, the more stable the system. The output fluctuation of RF parameters can include changes in frequency, amplitude, and pulse width, while the fluctuation of process indicators can include changes in plasma density, power coupling efficiency, etc.

[0057] A reward function may be constructed based on the first reward factor and the second reward factor, wherein the formula of the reward function may be as follows:

[0058]

[0059] Among them, σ represents output stability, and α represents the weight of "output stability" in the reward function, that is, the contribution of output stability. In this embodiment, R t ' can reflect the degree of interference suppression and output stability. Therefore, the larger the σ value, the smaller the system fluctuation and the more stable it is.

[0060] It can be understood that when we only need to focus on the interference suppression effect and do not consider system stability or other goals, the reward function R can be preferred. t When both interference suppression effect and output stability need to be taken into account, the reward function R can be preferred. t 'How rewards are calculated.

[0061] Furthermore, in some optional embodiments, the energy consumption of the radio frequency power can also be considered in the reward calculation process. Specifically, the energy consumption of the radio frequency power can be obtained as a third reward factor. A reward function can be constructed based on the first reward factor, the second reward factor, and the third reward factor, wherein the formula of the reward function can be as follows:

[0062]

[0063] Where E(t) represents the energy consumption of the radio frequency power, and β represents the influence of E(t) on the overall reward. t" can simultaneously reflect the degree of interference suppression, output stability, and energy consumption of RF power. Therefore, the smaller the energy consumption of RF power, the greater the value of the reward function.

[0064] Those skilled in the art can understand that the scheme of the first reward factor, the second reward factor and the third reward factor included in the above reward function is only a non-limiting implementation mode provided by the present invention, which is intended to clearly demonstrate the main concept of the present invention and provide a specific scheme that is convenient for the public to implement, rather than to limit the protection scope of the present invention. Optionally, in other embodiments, those skilled in the art can also add or change the items of the reward factor based on the concept of the present invention to meet the actual process requirements.

[0065] According to the updated reward function, the optimal learning strategy (Q-learning / strategy gradient, etc.) can be updated, where the formula for learning the optimal strategy can be as follows:

[0066]

[0067] Among them, Q(S,A) represents the cumulative reward Q that can be obtained in the future after performing a certain action A in a specific state S, γ represents the discount factor, R represents the value of the reward function, and π*(S) represents the action selected in state S according to the learning optimal strategy π. Based on the updated learning optimal strategy, the updated action output variable can be obtained to determine the frequency adjustment amount Δf, amplitude adjustment amount ΔA and pulse width adjustment amount ΔPW for the RF power in the next time step. Optionally, the frequency adjustment amount Δf can jump within the range of ±0.1MHz (such as +0.05MHz or -0.06MHz, etc.), the amplitude adjustment amount ΔA can be adjusted within the range of ±10%, and the pulse width adjustment amount ΔPW can appropriately adjust the duty cycle or on-off duration according to the set range.

[0068] Since the output parameters of RF power, especially frequency, amplitude, and pulse modulation method, will directly affect the generation and maintenance of plasma, these three parameters also have a significant impact on the ability to suppress interference (such as harmonics, standing waves, etc.). In the process, it is very important to maintain plasma stability. If the interference is too large, it will cause plasma unevenness and even cause process failure. By optimizing these parameters, better suppression effects can be obtained in different interference scenarios. The optimized RF power parameters can maximize the process yield and reduce the failure rate under dynamic conditions or interference environments.

[0069] At the same time, the amplitude and pulse width of the power output will determine the energy consumption of the RF system. If we only pursue interference suppression and ignore energy consumption, it will cause unnecessary cost increases or cause other overheating and stability problems. Therefore, power efficiency can also be considered comprehensively during optimization.

[0070] Continue back Figure 2 ,like Figure 2 As shown, step S230 may also be performed: based on the objective function of interference suppression and energy consumption, to determine the optimal action output variable.

[0071] Specifically, in some embodiments, the strategy can be optimized by reinforcement learning (RL). First, an objective function based on interference suppression and energy consumption can be constructed, where the objective function can be as follows:

[0072]

[0073] Among them, R(t) represents the reward function at time t (it can also be replaced by other types of reward functions mentioned above), E(t) represents the energy consumption of the RF (that is, power consumption), and λ represents the power efficiency weight. The objective function J can be used to measure the comprehensive indicators of interference suppression and energy consumption in the entire time domain at a higher level, and its input R(t) and E(t) can be single-step or continuous time integration, and output a "cumulative reward" or "long-term performance" value to test the pros and cons of the results of the above-mentioned optimal learning strategy.

[0074] From the formula of the above objective function K, we can see that since R(t) represents the "interference suppression reward", the smaller the interference, the larger R(t), and λ·E(t) represents the "energy consumption penalty", therefore, it is necessary to increase E(t) and decrease λ·E(t), so that the larger the value of the objective function J, the better.

[0075] Substitute the updated action output variables {Δf, ΔA, ΔPW} into the objective function J. The objective function J can be regarded as a global or offline evaluation of the entire process, or it can be continuously estimated in online iterations to determine whether the "current strategy" {Δf, ΔA, ΔPW} maximizes the objective function J as much as possible. When the objective function J is maximized, the action output variable corresponding to the maximized objective function is taken as the optimal action output variable.

[0076] For example, the frequency adjustment is based on the "action" result obtained by parameter optimization (reinforcement learning). In the reinforcement learning algorithm, the system will select an "action A" (including the adjustment of frequency, amplitude and pulse width) according to the current "state S" (interference intensity, main frequency and harmonic power density). Furthermore, the "state S" can include interference intensity and / or main frequency and its harmonic power density according to different usage scenarios.

[0077] When the algorithm decides to adjust the "frequency", it will select a specific change value within the allowed range (±0.1MHz). For example, when the action output by the agent in reinforcement learning includes the specific value of "changing the main frequency from 13.56MHz to 13.56+ΔfMHz" (for example, +0.05MHz), this "Δf" is the optimized frequency adjustment amount, which is commonly known as "frequency jump".

[0078] For example, if the algorithm finds at a certain moment that the current interference situation can be reduced by increasing the main frequency by 0.05MHz to reduce harmonics or stagger with external interference, then the optimal action can include the "+0.05MHz" item. Interference peaks can be avoided by adjusting the frequency. When the system detects significant external or internal interference at a specific frequency point (such as around 13.56MHz), the main frequency can be slightly offset to avoid the strong interference peak. In addition, fine-tuning the frequency can also obtain better energy coupling efficiency and lower standing wave ratio, enhance matching, and thus reduce interference.

[0079] Similarly, the amplitude is also based on the results of "parameter optimization (reinforcement learning)". Similar to frequency adjustment, when the action output by the reinforcement learning strategy includes the item "ΔA", it means that a certain percentage increase or decrease is made based on the original amplitude. For example, the current amplitude is 100 (which can be understood as 100V or a certain power scalar), and the "ΔA = -5%" given by reinforcement learning means adjusting the amplitude from 100 to 95. By adjusting the amplitude, interference can be reduced, because some interference sources are prone to nonlinearity or harmonics when the amplitude is high. If the harmonic peak is detected to be too high, the problem can be alleviated by appropriately reducing the amplitude. In addition, by adjusting the amplitude, the plasma can also be maintained stable. Since the larger the amplitude is within a certain range, the higher the plasma density is, but it is also more likely to cause transition discharge or nonlinear effects. Adaptively adjusting the amplitude can find a balance between suppressing interference and maintaining process requirements. In addition, energy can be saved by adjusting the amplitude. Since the higher the amplitude, the greater the power consumption is usually, therefore, reinforcement learning will combine the power consumption weight λ to dynamically balance between energy consumption and interference suppression.

[0080] When the action output by the reinforcement learning strategy includes the item "ΔPW", it means changing the "on" and "off" duration (or duty cycle) within the pulse period. For example, the current pulse period is 10μs and the duty cycle is 50% (i.e., 5μs on and 5μs off). Reinforcement learning may give "ΔPW = +10%", which means increasing the duty cycle value by 60% (6μs on and 4μs off). Transient power can be regulated by pulse width modulation. Since the power during the pulse on period is higher, plasma can be generated or maintained quickly, while the system can be relieved from interference peaks or energy consumption during the off period. Therefore, by dynamically changing the pulse width, certain specific types of interference can be suppressed, such as uneven discharge or local overheating. Moreover, energy consumption can be flexibly controlled through pulse width modulation. Compared with continuous waves, pulse mode can significantly reduce average power consumption. Therefore, moderate adjustment of the duty cycle can reduce the interference power peak to a certain extent and save energy at the same time. In addition, pulse width modulation is also helpful in matching process requirements, because in some thin film deposition or etching processes, it is desired to maintain a higher plasma density, followed by an appropriate shutdown period for chemical reaction or gas replacement, and the adjustment of the pulse width can optimize the process window and suppress interference.

[0081] Therefore, the above-mentioned reward function, objective function, and optimal strategy update in this embodiment are closely related to different levels: first, the single-step reward is calculated based on the interference measurement, then the reward (and energy consumption, stability, etc.) is incorporated into the overall objective function, and finally the optimal strategy is found through reinforcement learning or adaptive algorithms. In short, the "parameter optimization" performed by the above-mentioned adaptive control module 120 allows the system to achieve the best interference suppression and energy consumption balance by adaptively adjusting the RF power output (frequency, amplitude, pulse width) while meeting the dual requirements of interference suppression and power efficiency, thereby ensuring the stability and economy of the PECVD plasma process.

[0082] Further, in some preferred embodiments, after the above step S230, a fuzzy logic control (FLC) algorithm can also be performed. Specifically, the interference intensity can be determined according to the interference peak. For example, the interference peak is measured, the main frequency power density is measured to be -45dBm, the first harmonic power density is measured to be -35dBm, and the second harmonic power density is measured to be -25dBm. After that, the interference intensity is judged, and an interference threshold (such as -30dBm) can be preset. The power density of the main frequency and the first harmonic does not exceed -30dBm, and it can be judged that the two belong to medium interference, and the power density of the second harmonic exceeds -30dBm, so it is judged as strong interference.

[0083] Then, through the fuzzy logic control algorithm, based on the interference intensity, the range of the frequency adjustment amount, the range of the amplitude adjustment amount and the range of the pulse width adjustment amount are determined; and based on the range of the frequency adjustment amount, the range of the amplitude adjustment amount and the range of the pulse width adjustment amount, the optimal action output variable can be fine-tuned on a local or short-time scale.

[0084] Specifically, in the fuzzy inference model of the fuzzy logic control algorithm, the so-called "input interference intensity" generally appears in the "fuzzification" step, and the corresponding membership function is used to convert the interference intensity of a "crisp value" into a "membership degree" on the fuzzy set. In other words, the interference intensity usually does not appear directly in a specific "formula", but is mapped to fuzzy sets such as "low, medium, and high" through the membership function, and then enters the fuzzy rule reasoning process. The fuzzy inference model can be μ 调整 =max[μ 低 , μ 中 , μ 高 ].

[0085] like Figure 3A As shown, in some embodiments, in the fuzzy control rules of interference intensity and power amplitude adjustment, the input variables are low, medium, and high interference intensity, and the output variables are small, medium, and large power amplitude adjustment. Figure 3B As shown, in some embodiments, in the fuzzy control rules of interference intensity, frequency adjustment and amplitude adjustment, the input variables are low, medium and high interference intensities, the frequency adjustment ranges in the output variables correspond to ±0.05, ±0.1 and ±0.15 respectively, and the amplitude adjustment ranges correspond to ±5, ±10 and ±15 respectively.

[0086] The fuzzy logic control algorithm and the reinforcement learning algorithm mentioned above belong to different time scales or control levels. Reinforcement learning / optimal strategies often perform global optimization over a larger range (frequency ±0.1MHz, amplitude ±10%) and a longer time period, while fuzzy control can respond quickly in a short time and local range, and make instant fine-tuning according to the interference intensity classification (low / medium / high) to improve the robustness and response speed of control. In addition, FLC improves redundancy and safety performance, and can provide expert experience rules. Even if the RL strategy has not converged or failed in extreme conditions, the fuzzy logic control algorithm can still provide basic protection to prevent large-scale mismatches. In addition, since the fuzzy logic control algorithm is easy to understand and modify, it is possible to use fuzzy control together with the above-mentioned new RL strategy to ensure the robustness and stability of the system.

[0087] Therefore, in the above embodiment, FLC does not conflict with the optimal strategy, but provides additional fine-tuning of RF parameters on a local or short-term scale, and can quickly and intuitively incorporate expert experience to ensure that the system still operates stably in the face of transient interference. The fuzzy reasoning model is used as a supplement to make fine adjustments based on the known global optimal setting value, compensate for transient interference or details, so as to make the system more robust, responsive and easy to maintain.

[0088] Please continue to return Figure 2 Finally, step S240 may be executed: based on the optimal action output variable, dynamically adjusting multiple process parameters in the reaction chamber so that the multiple process parameters reach the target process parameter range.

[0089] Specifically, it can be combined Figure 1 It is commonly understood that the feedback regulation module 130 can be mainly used to adjust the process parameters (such as gas flow and chamber pressure) in real time according to the optimized RF power signal to optimize the process parameters, thereby ensuring stable operation of the equipment.

[0090] The feedback regulation module 130 may include a sensor network and a closed-loop controller. The sensor network may be used to monitor process parameters such as gas flow, chamber pressure, and plasma density in the reaction chamber in real time. The closed-loop controller may be used to dynamically adjust the process parameters to match the RF power.

[0091] First, the process parameters in the reaction chamber can be collected in real time. The gas flow rate can be obtained through a mass flow controller (MFC), where the gas flow rate is usually in the range of 40 to 60 sccm. The chamber pressure can be obtained through a pressure sensor, where the chamber pressure is usually in the range of 1 to 5 Pa. The current density of the plasma can be measured by a spectrometer, where the target density of the plasma can be 10 12 cm -3 .

[0092] In response to the completion of the adjustment of the RF power based on the optimal action output variable, the current density of the plasma is continuously collected by the spectrometer to obtain the density deviation between the current density and the target density. Based on the density deviation, the gas flow rate and the cavity pressure can be closed-loop adjusted to make the current density reach the target density, wherein the closed-loop adjustment formula can be as follows:

[0093] Q=Q1+K p ·Δn e

[0094] Where Q1 represents the initial gas flow rate, K p Denotes the feedback gain coefficient, Δn erepresents the density deviation of the plasma, and Q is the dynamically adjusted gas flow rate. If the current density of the plasma is too high, the gas flow rate or chamber pressure can be reduced to reduce the discharge intensity, while if the current density of the plasma is too low, the flow rate or pressure can be increased to increase the collision and ionization opportunities.

[0095] In the above step S240, while adaptively controlling the output of higher RF energy, the gas flow rate and cavity pressure are matched with it through feedback adjustment, so that the reactants can be distributed more evenly in the reaction area, reducing the uneven film thickness caused by local excessive discharge or gas deficiency. In addition, with the help of closed-loop control, the plasma parameters will not deviate significantly from the target value due to external interference or process fluctuations, thereby significantly reducing the fluctuation amplitude of the plasma state, and the plasma density fluctuation can be effectively narrowed from ±15% to ±5%, improving the process stability.

[0096] In addition, the cavity pressure can be adjusted in real time by a vacuum pump to keep it within the target pressure range.

[0097] Next, a specific embodiment will be used to verify the effect of the adjustment method implemented by the RF power adjustment system provided by the present invention. The initial RF power of the experiment is set to 300W, the main frequency is 13.56MHz, the gas flow rate is 50sccm, and the chamber pressure is 2Pa. For the main frequency of 13.56MHz, before parameter optimization, its power density is 140dBμV / m. After parameter optimization, the power density of the main frequency can be reduced to 120dBμV / m, a reduction of 20dB. For the first harmonic of 27.12MHz, before parameter optimization, its power density is 85dBμV / m. After parameter optimization, the power density of the first harmonic can be reduced to 60dBμV / m, a reduction of 25dB. For the second harmonic of 40.68MHz, before parameter optimization, its power density is 60dBμV / m. After parameter optimization, the power density of the second harmonic can be reduced to 45dBμV / m, a reduction of 15dB. In other words, the power density of the main frequency and harmonic interference can be reduced by 20%-30%, thereby improving the interference suppression effect.

[0098] In addition, from the perspective of process performance, in one embodiment, the film deposition rate in the reaction chamber is 100nm / min before optimization, and can be increased to 120nm / min after optimization, and the deposition rate is increased by 20%. In addition, the coefficient of variation of film uniformity is 5% before optimization, and can be reduced to 2% after optimization, and the coefficient of variation is reduced by 60%. In other words, the rate of film deposition is significantly improved, and the uniformity is significantly improved, thereby improving equipment efficiency and film deposition quality.

[0099] Although the above methods are illustrated and described as a series of actions for simplicity of explanation, it should be understood and appreciated that these methods are not limited by the order of the actions, because according to one or more embodiments, some actions may occur in a different order and / or concurrently with other actions from those illustrated and described herein or not illustrated and described herein but understandable to those skilled in the art.

[0100] Those skilled in the art will further appreciate that the steps of the methods or algorithms described in conjunction with the embodiments disclosed herein may be directly embodied in hardware, in a software module executed by a processor, or in a combination of the two. The software module may reside in a RAM memory, a flash memory, a ROM memory, an EPROM memory, an EEPROM memory, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. An exemplary storage medium is coupled to a processor so that the processor can read and write information from / to the storage medium. In an alternative, a storage medium may be integrated into a processor. The processor and the storage medium may reside in an ASIC. The ASIC may reside in a user terminal. In an alternative, the processor and the storage medium may reside in a user terminal as discrete components.

[0101] In one or more exemplary embodiments, the functions described may be implemented in hardware, software, firmware, or any combination thereof. If implemented as a computer program product in software, each function may be stored on or transmitted by a computer-readable medium as one or more instructions or codes. Computer-readable media include both computer storage media and communication media, including any medium that facilitates the transfer of a computer program from one place to another. Storage media may be any available medium that can be accessed by a computer. As an example and not limitation, such a computer-readable medium may include RAM, ROM, EEPROM, CD-ROM or other optical disk storage, disk storage or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of an instruction or data structure and can be accessed by a computer. Any connection is also properly referred to as a computer-readable medium. For example, if the software is transmitted from a website, a server, or other remote source using a coaxial cable, a fiber optic cable, a twisted pair, a digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwaves, the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwaves are included in the definition of the medium. Disk and disc as used herein include compact disc (CD), laser disc, optical disc, digital versatile disc (DVD), floppy disk and Blu-ray disc, wherein disk often reproduces data magnetically, while disc reproduces data optically with lasers. Combinations of the above should also be included within the scope of computer-readable media.

[0102] In summary, the present invention provides a radio frequency power regulation system, a radio frequency power regulation method, a semiconductor process equipment, and a computer-readable storage medium, which can adaptively adjust the radio frequency power density in the reaction chamber to achieve the optimal dynamic balance between interference suppression and energy consumption, thereby improving the radio frequency interference suppression effect of the process equipment, ensuring the stability of the plasma process in the reaction chamber, and improving the process efficiency and thin film deposition quality of the equipment.

[0103] The previous description of the disclosure is provided to enable any person skilled in the art to make or use the disclosure. Various modifications to the disclosure will be apparent to those skilled in the art, and the general principles defined herein may be applied to other variations without departing from the spirit or scope of the disclosure. Thus, the disclosure is not intended to be limited to the examples and designs described herein, but should be granted the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A radio frequency power regulation system, characterized in that: include: An interference monitoring module, which obtains a time domain signal of the radio frequency power in the reaction chamber, and obtains interference data based on the frequency domain characteristics of the time domain signal; An adaptive control module, taking the interference data as a state input variable, updating the action output variable by learning the optimal strategy, and determining the optimal action output variable based on the objective function of interference suppression and energy consumption, wherein the action output variable includes the frequency adjustment amount, the amplitude adjustment amount and the pulse width adjustment amount of the radio frequency power; as well as The feedback adjustment module dynamically adjusts a plurality of process parameters in the reaction chamber based on the optimal action output variable so that the plurality of process parameters reach a target process parameter range.

2. The regulating system according to claim 1, characterized in that The interference monitoring module comprises: A high-frequency probe, used to collect a time-domain signal of the radio frequency power in the reaction chamber, wherein the time-domain signal includes a main frequency and its harmonic frequencies; and A spectrum analyzer is used to extract the frequency domain characteristics of the time domain signal and generate spectrum data to obtain the interference data, wherein the interference data includes harmonic power density and interference peak value.

3. The regulating system according to claim 2, characterized in that The input model of the time domain signal is as follows: Wherein, P(t) represents the time domain signal, A(t), ω, Indicates the amplitude, frequency and phase of the main frequency, A n (t) and Represents the amplitude and phase of each harmonic frequency, The step of extracting the frequency domain characteristics of the time domain signal and generating spectrum data to obtain the interference data includes: The output model of the frequency domain signal of the time domain signal is obtained by fast Fourier transform, wherein the output model of the frequency domain signal is as follows: Wherein, S(f) represents the frequency spectrum of the frequency domain signal, and 2πf represents the distance between two spectrum lines of the frequency domain signal; and The frequency domain characteristics are extracted according to the output model of the frequency domain signal for identifying the harmonic power density and the interference peak.

4. The regulating system according to claim 3, characterized in that The interference monitoring module is further configured to: identify interference sources based on the interference data, and classify the identified interference sources; and comprehensively evaluate the classification results, and determine an adjustment method based on the interference sources.

5. The regulating system according to claim 3, characterized in that: The step of using the interference data as the state input variable and obtaining the optimal action output variable by learning the optimal strategy comprises: Obtaining the sum of interference power density values ​​of the main frequency and each harmonic frequency as a first reward factor; Based at least on the first reward factor, a reward function is constructed, wherein the smaller the first reward factor is, the larger the value of the reward function is, wherein the formula of the reward function is as follows: Among them, ∑ i P i (t) represents the sum of the interference power density values.

6. The regulating system according to claim 5, characterized in that The step of determining a reward function based at least on the first reward factor further comprises: Obtaining output stability in the reaction chamber as a second reward factor, wherein the output stability includes output fluctuation of radio frequency parameters or fluctuation of process indicators; Based on the first reward factor and the second reward factor, a reward function is constructed, wherein the larger the second reward factor is, the larger the value of the reward function is, wherein the formula of the reward function is as follows: Wherein, σ represents the output stability, and α represents the contribution degree of the output stability.

7. The regulating system according to claim 6, characterized in that The step of determining a reward function based at least on the first reward factor further comprises: Obtaining energy consumption of the radio frequency power as a third reward factor; A reward function is constructed based on the first reward factor, the second reward factor, and the third reward factor, wherein the smaller the third reward factor is, the larger the value of the reward function is, wherein the formula of the reward function is as follows: Wherein, E(t) represents the energy consumption of the radio frequency power, and β represents the influence of E(t) on the overall reward.

8. The regulating system according to claim 5, 6 or 7, characterized in that: The step of using the interference data as the state input variable and updating the action output variable by learning the optimal strategy also includes: According to the reward function, the learning optimal strategy is updated, wherein the formula of the learning optimal strategy is as follows: Where Q(S,A) represents the cumulative reward Q that can be obtained in the future after performing an action A in a specific state S, γ represents the discount factor, R represents the value of the reward function, and π*(S) represents the action selected in state S according to the learned optimal strategy π; and Based on the updated learning optimal strategy, an updated action output variable is obtained to determine the frequency adjustment amount, amplitude adjustment amount and pulse width adjustment amount for the radio frequency power in the next time step.

9. The regulating system according to claim 8, characterized in that The step of determining the optimal action output variable based on the objective function of interference suppression and energy consumption includes: An objective function based on interference suppression and energy consumption is constructed, wherein the objective function is as follows: Wherein, R(t) represents the reward function at time t, E(t) represents the energy consumption of the radio frequency, and λ represents the power efficiency weight; Substituting the updated action output variable into the objective function to maximize the objective function; and In response to the objective function being maximized, the action output variable corresponding to the maximized objective function is used as the optimal action output variable.

10. The regulating system according to claim 2, characterized in that After the step of determining the optimal action output variable based on the objective function of interference suppression and energy consumption, the method further includes: Determining interference intensity according to the interference peak value; Determining the range of the frequency adjustment amount, the range of the amplitude adjustment amount, and the range of the pulse width adjustment amount based on the interference intensity by a fuzzy logic control algorithm; and Based on the range of the frequency adjustment amount, the range of the amplitude adjustment amount and the range of the pulse width adjustment amount, the optimal action output variable is fine-tuned on a local or short-time scale.

11. The adjustment system according to claim 1, characterized in that The feedback regulation module includes a sensor network, which is used to monitor the gas flow, cavity pressure, and plasma density in the reaction chamber in real time. The step of dynamically adjusting the multiple process parameters in the reaction chamber based on the optimal action output variable so that the multiple process parameters reach the target process parameter range includes: In response to completing the adjustment of the radio frequency power based on the optimal action output variable, collecting the current density of the plasma and obtaining a density deviation between the current density and a target density; Based on the density deviation, the gas flow rate and the cavity pressure are closed-loop regulated so that the current density reaches the target density, wherein the closed-loop regulation formula is as follows: Q=Q1+K p ·Δn e Where Q1 represents the initial gas flow rate, K p Denotes the feedback gain coefficient, Δn e represents the density deviation of the plasma.

12. A method for adjusting radio frequency power, characterized in that: The following steps are involved: Acquire a time domain signal of radio frequency power in the reaction chamber, and acquire interference data based on frequency domain characteristics of the time domain signal; Using the interference data as a state input variable, updating the action output variable by learning the optimal strategy; Determine the optimal action output variable based on the objective function of interference suppression and energy consumption; and Based on the optimal action output variable, a plurality of process parameters in the reaction chamber are dynamically adjusted so that the plurality of process parameters reach a target process parameter range.

13. A semiconductor process equipment, characterized in that: include: A reaction chamber, in which wafers are placed for thin film deposition process; RF power module; Used to provide radio frequency power; as well as The RF power regulation system according to any one of claims 1 to 11, connected to the RF power module, is used to dynamically adjust the multiple process parameters in the reaction chamber according to the RF power corresponding to the optimal action output variable, so that the multiple process parameters reach the target process parameter range.

14. A computer-readable storage medium having computer instructions stored thereon, characterized in that: When the computer instructions are executed by the processor, the method for adjusting the radio frequency power as claimed in claim 12 is implemented.

Images