CN119028882A

CN119028882A - Multi-frequency modulation plasma surface treatment method and system

Info

Publication number: CN119028882A
Application number: CN202411508972.1A
Authority: CN
Inventors: 杨月明; 朱成; 季云
Original assignee: Nanjing Jiayang Engineering Technology Co ltd
Current assignee: Nanjing Jiayang Engineering Technology Co ltd
Priority date: 2024-10-28
Filing date: 2024-10-28
Publication date: 2024-11-26

Abstract

The present invention discloses a multi-frequency modulation plasma surface treatment method and system, which relates to the field of semiconductor manufacturing technology; the system comprises an acquisition module, a processing module and a control module; the acquisition module is responsible for acquiring etching information and characteristic data of a target plasma, such as optical, electrical and gas composition data, and aligning these data to a unified time axis; the processing module analyzes these data through a preset machine learning model, generates process status evaluation information, and uses a genetic algorithm to generate a first process parameter of the target plasma based on these information; the control module multi-frequency modulates a plasma source according to these parameters to optimize the physical properties of the plasma; the present invention improves the semiconductor etching effect by optimizing the synchronization of multi-source data in the multi-frequency modulation plasma surface treatment process and the physical properties of the plasma.

Description

Multi-frequency modulation plasma surface treatment method and system

Technical Field

The invention relates to the technical field of semiconductor manufacturing, in particular to a multi-frequency modulation plasma surface treatment method and system.

Background

In the field of semiconductor fabrication, etching techniques are a critical step in the fabrication of integrated circuits, which involve transferring specific patterns to thin films or other underlying materials on silicon wafers. Plasma etching techniques are widely used because they provide highly anisotropic etching and are suitable for miniaturized processes. Plasma etching techniques typically rely on a single frequency rf source to ignite and sustain a plasma. While this approach works well in some applications, it faces some limitations in handling complex or large area wafers. In particular in high-end applications where high etch rates and fine pattern fidelity are required simultaneously, existing rf sources are typically fixed at a frequency that limits the ability to tailor the physical properties of the plasma, such as ion density and energy distribution, which are critical to the optimization of the etching process. Furthermore, the lack of power distribution and regulation of a single frequency rf source may result in wasted energy, particularly when the plasma source requires rapid adjustment of output according to different process steps.

For example, chinese patent publication No. CN117936425B discloses an etching apparatus and method for precisely etching and removing a thin film, comprising: the plasma etching cavity is used for carrying out film etching on the wafer to be removed with the film under the action of plasma; an observation window arranged at one side of the plasma etching cavity; the information acquisition module is arranged outside the observation window and acquires initial characteristic information of plasma in the plasma etching cavity through the observation window; the information processing module is used for processing the characteristic information of the plasmas to obtain final characteristic information after information processing; the information feedback module obtains adjustment parameters according to the final characteristic information after information processing, and feeds the adjustment parameters back to the plasma control module; the plasma control module adjusts a reaction parameter of the plasma etching chamber according to the adjustment parameter. The scheme is based on the information acquisition module and the information processing module, and the characteristic capture is performed on the plasma in the etching cavity accurately, so that the accuracy of etching and removing the film is improved.

The above method presents problems in the background art, and existing rf sources are typically fixed at a frequency, which limits the ability to tailor the physical properties of the plasma, such as ion density and energy distribution. Furthermore, the lack of power distribution and regulation of a single frequency rf source may result in wasted energy, particularly when the plasma source requires rapid adjustment of output according to different process steps.

Disclosure of Invention

Aiming at the defects of the prior art, the invention discloses a multi-frequency modulation plasma surface treatment method and a system, which are used for optimizing the physical properties of plasmas under multi-frequency modulation.

In a first aspect, embodiments of the present disclosure provide a multi-frequency modulated plasma surface treatment system comprising: the device comprises an acquisition module, a processing module and a control module;

the acquisition module is used for acquiring etching process information and acquiring characteristic data of target plasmas under multi-frequency modulation through the sensor, wherein the characteristic data comprise optical data, electrical data and gas composition data; aligning different types of characteristic data to a unified same time axis;

The processing module is used for inputting the characteristic data of the target plasma aligned to the same time axis into a preset machine learning model and generating process state evaluation information; optimizing by utilizing a genetic algorithm based on the process state evaluation information to generate a first process parameter of the target plasma;

The control module is used for performing multi-frequency modulation on a plasma source for generating the target plasma based on the first process parameter, and adjusting the physical properties of the target plasma.

As an alternative embodiment, the acquisition module includes: optical sensors, electrical sensors, and mass spectrometers;

The optical sensor is arranged at a window of the plasma etching cavity and is used for monitoring optical data of the target plasma in real time, and the optical sensor comprises: emission spectrum and light intensity;

The electrical sensor is directly connected to a plasma source for measuring electrical data of the plasma, comprising: voltage, current, and power;

the mass spectrometer is connected with the plasma etching cavity through a gas sampling system and is used for monitoring the composition and ion types of the reaction gas and analyzing the gas components in the plasma in real time to generate gas component data.

As an alternative embodiment, the acquisition module further comprises a data processing unit;

The data processing unit is used for preprocessing optical data, electrical data and gas composition data of the target plasma and aligning different types of characteristic data to the same time axis.

As an alternative embodiment, the aligning different types of feature data to the same time axis includes:

for the optical data, for each data acquisition time point, calculating optical data at the time point according to adjacent known optical data points by using a cubic spline interpolation method;

for the electrical data, for each data acquisition time point, the electrical data at that time point is calculated from adjacent known electrical data points using a linear interpolation method.

For the gas composition data, correction and interpolation are performed on the gas composition data using a kalman filter for each data acquisition time point, and the gas composition data at that time point is calculated.

As an alternative embodiment, the kalman filter includes:

An interactive noise covariance matrix, wherein the elements of the interactive noise covariance matrix are defined as influence values of different gas component changes on other gas component measurements in the processing process of the target plasma;

The Kalman filter is further configured with a state prediction mechanism and an error update mechanism;

The state prediction mechanism predicts the state of the current time step by applying a state transition matrix to the state estimation value of the previous time step;

The error updating mechanism updates the state estimation and the error covariance by observing the matrix, the actual measurement value and the actual measurement noise covariance, thereby adjusting the state estimation value of the gas component.

As an alternative embodiment, the processing module, before inputting the characteristic data of the target plasma aligned to the same time axis into a preset machine learning model, further includes:

carrying out standardization processing on historical characteristic data of target plasmas under multi-frequency modulation, and converting the historical characteristic data into dimensionless data;

Extracting spectral characteristics, voltage characteristics, current characteristics, power characteristics, the concentration of different gas components and the variation trend thereof from the dimensionless data to generate standardized characteristic data;

Constructing a data matrix based on the standardized feature data; wherein each row of the data matrix represents a data sample of characteristic data of a target plasma, and each column of the data matrix represents a feature;

calculating a covariance matrix of the data matrix, and carrying out eigenvalue decomposition on the covariance matrix to obtain eigenvalues and eigenvectors; wherein the eigenvalue represents the importance of the principal component, and the eigenvector represents the direction of the principal component;

Selecting a main component with a characteristic value accumulation contribution rate of more than 90% based on the magnitude of the characteristic value;

projecting the data matrix onto the principal component, generating a first reduced-dimension data set;

And constructing the preset machine learning model by using the first dimension reduction data set.

As an alternative embodiment, the processing module performs the following steps when optimizing using a genetic algorithm to generate the first process parameter of the target plasma based on the process state evaluation information:

generating an initial population, wherein each individual represents a set of possible power output and frequency combinations;

Designing an fitness function based on the process state evaluation information and the target plasma etch process information;

performing selection, crossing and mutation operations to explore a solution space, and dynamically adjusting an fitness function based on the process state evaluation information;

Wherein the mutation operation is non-uniform mutation, comprising: performing Gaussian variation on the power output and performing exchange variation on the frequency combination;

Repeating the steps, evaluating the fitness value of the newly generated individual, and selecting the individual with the fitness meeting the first target threshold to enter the next generation until the fitness value converges or reaches the preset iteration times;

and generating a first process parameter of the target plasma based on the current individual in response to the fitness value converging or reaching a preset number of iterations.

As an alternative embodiment, the control module includes: a power amplifier, a radio frequency source, and a power distribution network;

The power amplifier is used for amplifying the power of a radio frequency signal generated by the radio frequency source, and each target frequency corresponds to an independent power amplifier;

The radio frequency source generates a radio frequency signal with target frequency and power based on the first technological parameter, and inputs the radio frequency signal to the power amplifier; the radio frequency source comprises a plurality of frequency generation units, wherein each frequency generation unit is used for generating a target frequency;

the power distribution network is used for distributing the power of the amplified radio frequency signal to the plasma source.

As an alternative embodiment, the control module performs the following steps when performing multi-frequency modulation on the target plasma based on the first process parameter, and adjusting the physical attribute of the target plasma:

The radio frequency source sets the frequency of the radio frequency signal generated by each frequency generating unit based on the target frequency and sends the frequency to the power amplifier corresponding to each target frequency;

the power amplifier amplifies the output power of the radio frequency signals of each frequency based on the output power of the target frequency;

The power distribution network receives the radio frequency signals of all frequencies after amplifying the output power and distributes the radio frequency signals to different electrodes or areas of the plasma source, and the physical properties of the target plasma are adjusted.

In a second aspect, the present invention also provides a method for multi-frequency modulated plasma surface treatment, comprising:

Acquiring etching process information, and acquiring characteristic data of target plasmas under multi-frequency modulation by a sensor, wherein the characteristic data comprise optical data, electrical data and gas composition data; aligning different types of characteristic data to the same time axis;

inputting the characteristic data of the target plasma aligned to the same time axis to a preset machine learning model to generate process state evaluation information; optimizing by utilizing a genetic algorithm based on the process state evaluation information to generate a first process parameter of the target plasma;

and based on the first technological parameter, performing multi-frequency modulation on a plasma source of the target plasma, and adjusting the physical property of the target plasma.

Compared with the prior art, the invention has the beneficial effects that: the present invention makes full use of the characteristic data to evaluate the current semiconductor etching condition by collecting the optical data, the electrical data and the gas composition data of the plasmas under the multi-frequency modulation. This integration of multi-dimensional data provides a more comprehensive monitoring and control capability for the etching process.

And (3) carrying out real-time evaluation on the etching process by using the established machine learning model, and optimizing the output power and frequency combination by using an improved genetic algorithm according to the evaluation result. The method can dynamically adjust the multi-frequency modulation parameters according to the target etching effect, actually change the physical properties of the plasma, and optimize the etching quality and consistency.

Aiming at the data for constructing the machine learning model, the invention adopts various methods to carry out time synchronization, and ensures the consistency and the accuracy of the data. In addition, the method and the device improve the processing efficiency and the prediction precision of the machine learning model by reducing the dimension of the data by applying the principal component analysis method, thereby optimizing the control and the prediction effect of the whole etching process.

Drawings

FIG. 1 is a schematic diagram of a multi-frequency modulated plasma surface treatment system according to an embodiment of the present invention;

FIG. 2 is a flow chart of a method for multi-frequency modulated plasma surface treatment according to an embodiment of the present invention;

Fig. 3 is a schematic structural diagram of a computer device according to an embodiment of the present invention.

Detailed Description

For the purposes of making the objects, technical solutions and advantages of the embodiments of the present disclosure more apparent, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the drawings in the embodiments of the present disclosure, and it is apparent that the described embodiments are only some embodiments of the present disclosure, but not all embodiments. The components of the disclosed embodiments generally described and illustrated herein may be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present disclosure is not intended to limit the scope of the disclosure, as claimed, but is merely representative of selected embodiments of the disclosure. All other embodiments, which can be made by those skilled in the art based on the embodiments of this disclosure without making any inventive effort, are intended to be within the scope of this disclosure.

The following describes a method and a system for multi-frequency modulated plasma surface treatment according to embodiments of the present disclosure.

Referring to fig. 1, a schematic diagram of a multi-frequency modulation plasma surface treatment system according to an embodiment of the present invention includes: the acquisition module 110, the processing module 120 and the control module 130;

the acquisition module 110 is used for acquiring etching process information and acquiring characteristic data of target plasma under multi-frequency modulation, including optical data, electrical data and gas composition data, through a sensor; aligning different types of characteristic data to the same time axis;

A processing module 120, configured to input characteristic data of the target plasma aligned to the same time axis to a preset machine learning model, and generate process state evaluation information; optimizing by utilizing a genetic algorithm based on the process state evaluation information to generate a first process parameter of the target plasma;

and the control module 130 is used for performing multi-frequency modulation on the plasma source of the target plasma based on the first process parameter, and adjusting the physical property of the target plasma.

Wherein the etching process information includes: etching depth information, etching profile and sidewall quality, etching target material, etching uniformity information, and the like, which are used to describe the etching target for the semiconductor device.

In particular implementations, the acquisition module 110 acquires various parameter data in the plasma etch in real time and sends it to the processing module 120 and the control module 130 for use. The acquisition module 110 is equipped with a variety of high-precision sensors to cover the monitoring needs of critical parameters.

The characteristic data are used for reflecting the characteristics of the plasmas, the characteristics are physical properties of the plasmas, and the influence of different physical properties of the plasmas on etching effects can be further obtained by analyzing the physical properties.

As an alternative embodiment, the acquisition module 110 includes: optical sensors, electrical sensors, and mass spectrometers;

wherein the optical sensor is used for monitoring optical data of the plasma, including emission spectrum, light intensity and the like, and the parameters can reflect ion density and energy distribution of the plasma. The optical sensor is arranged at a window of the plasma etching cavity and can capture the state change of the plasma in real time.

The electrical sensor is used for measuring electrical characteristics such as voltage, current, power and the like of the plasma. These sensors are directly connected to the plasma source, i.e. the generator and the power supply system, and can provide accurate electrical data. These data facilitate analysis of the energy input and reactivity of the plasma.

Mass spectrometers for monitoring the composition and ionic species of the reactant gases. The mass spectrometer can analyze the gas composition in the plasma in real time to provide detailed chemical reaction information. The mass spectrometer is connected with the etching cavity through a gas sampling system, and can continuously collect and analyze the reaction gas sample.

Wherein the plasma etch chamber is a closed environment specifically designed for performing plasma processing, such as etching or deposition. It is the primary site of plasma generation and action. The interior of the etching chamber is typically equipped with an optical window, allowing an optical sensor to be mounted thereto, monitoring the state changes of the plasma, such as emission spectrum and light intensity, in real time.

A plasma source refers to a device that generates a plasma and typically includes a Radio Frequency (RF) source and other auxiliary devices, such as a gas flow control system. The plasma source is responsible for exciting an incoming gas (e.g., argon, hydrogen fluoride, etc.) to generate a plasma with energetic particles that are used to etch materials on the silicon wafer. The plasma source is directly connected to electrical sensors, and the voltage, current and power data measured by these sensors provides accurate information about the plasma energy input and reactivity.

The gas sampling system is a device connected to the mass spectrometer for extracting a gas sample from the etching chamber and delivering it to the mass spectrometer for analysis. The system is capable of monitoring and analyzing gas composition changes in the plasma, including the composition of the reactant gases and the ion species.

In a multi-frequency modulated plasma surface treatment system, these components interact to collectively optimize the etching process of the semiconductor. For example, a plasma generated by a plasma source reacts with silicon material within an etch chamber while monitoring the progress of the chemical reaction in real time by a gas sampling system and mass spectrometer to adjust the frequency and power of the radio frequency source. The data provided by the optical and electrical sensors is then used to further precisely control and monitor the physical and chemical state of the plasma, ensuring the efficiency and repeatability of the etching process.

In particular implementations, the plasma source generates a target plasma whose operating parameters, such as radio frequency power and frequency, directly affect the physical and chemical characteristics of the plasma. The target plasma interacts with the semiconductor material within the etch chamber to effect etching or deposition. The gas sampling system is used together with the mass spectrometer to analyze the gas components in the target plasma and provide data support for adjusting the parameters of the plasma source.

As an alternative embodiment, the acquisition module 110 further comprises a data processing unit.

In a specific implementation, the data acquisition unit is equipped with a data processing chip for rapidly processing large amounts of real-time data and transmitting the processed data to the processing module 120.

In addition, the data acquisition unit integrates the data synchronization and time stamping functions, so that all sensor data are ensured to be acquired and recorded at the same time point. Therefore, analysis errors caused by asynchronous data acquisition time can be avoided, and a data basis is provided for subsequent data processing and control decisions.

In a specific implementation, the data acquisition unit receives data packets of the optical sensor, the electrical sensor, and the mass spectrometer via a high-speed data bus or a network interface. Each data packet contains a measurement value and timestamp information.

The received data is first denoised and filtered to eliminate sensor noise and environmental interference. This step ensures the sharpness and accuracy of the data.

To ensure that different types of characteristic data can be analyzed at the same point in time, the data acquisition unit needs to time synchronize and align the data.

The data acquisition unit aligns all data according to the time stamp. By mapping data of different sources onto a unified time axis, the synchronism of the data is ensured. For example, if the time stamps of the optical data and the electrical data are slightly different, the system will interpolate or weight average to ensure that the various data at the same time can be matched.

As an alternative embodiment, aligning different types of characteristic data to the same time axis includes:

For the optical data, for each data acquisition time point, the optical data at that time point is calculated from adjacent known optical data points using a cubic spline interpolation method.

For electrical data, for each data acquisition time point, the electrical data at that time point is calculated from adjacent known electrical data points using a linear interpolation method.

For the gas composition data, for each data acquisition time point, the gas composition data at that time point is calculated by correcting and interpolating the gas composition data using a kalman filter.

In implementations, the acquisition time stamps for each sensor may not be the same, requiring alignment of the data to a uniform time axis.

There are various reasons for the different acquisition time stamps of the sensors, for example, each sensor, such as an optical sensor, an electrical sensor, and a mass spectrometer, has its specific response time and data processing speed. The electrical sensor may provide data in near real time, whereas the mass spectrometer may output data after collecting and analyzing the gas sample. These differences result in inconsistent times for data acquisition and processing to complete. For example, improper maintenance of the device or environmental disturbances may result in some sensors responding more slowly, thereby affecting the time of acquisition of the data. As another example, different types of sensors may be provided with different sampling rates, i.e. the number of measurements per unit time. Some sensors may require a higher frequency of data updates, such as electrical sensors, while other sensors may update slower, such as mass spectrometers. This difference in sampling rate directly affects the point in time of the data recording.

In practice, there are three different types of data sources: optical data, electrical data, and gas composition data.

Wherein the optical data typically has a high temporal and spatial resolution, but may be affected by ambient noise. To ensure accuracy of the data, a cubic spline interpolation method is used for alignment. The method comprises the following specific steps: for each time point, an optical data value at that time point is calculated from adjacent known optical data points using a cubic spline interpolation method. The interpolation method generates a high-precision interpolation result by considering smooth transition among data points, and reduces noise influence.

The electrical data generally varies smoothly and has high reliability and accuracy. For alignment, a linear interpolation method is used. The method comprises the following specific steps: for each time point, an electrical data value at that time point is calculated from adjacent known electrical data points. The linear interpolation method provides a rapid and reliable interpolation result through simple linear calculation, and is suitable for the gentle change characteristic of the electrical data.

The gas composition data reflects the chemical reaction process in the plasma in a multi-frequency modulated plasma etch process. However, because of the possible time delay and noise of the data, correction and interpolation of the data is required to ensure that it is synchronized with other types of data. The kalman filter is an effective way to achieve this goal.

As an alternative embodiment, to cope with complex gas dynamics and to enhance the performance of the kalman filter, an interaction noise covariance is introduced, and this parameter is designed to better handle interactions between gas components, especially in a multi-component gas environment where the variation of different gas components may depend on each other or interfere with each other.

The purpose of the introduction of the cross noise covariance is to take into account the interactions between the gas components during the measurement update of the kalman filter. In the process ofWhen gas is used, the change of single gas component can affect the measurement accuracy of other components due to the complexity of chemical reaction and physical interaction. For example, the number of the cells to be processed,The decomposition products of (2) may affectAnd vice versa. Conventional kalman filters may not accurately estimate gas concentration without taking such interactions into account, thereby affecting the accuracy of control of the overall etching process.

Wherein, (Carbon tetrafluoride) is a widely used gas for etching silicon and its oxides, particularly in etching processes for semiconductor device surfaces; Is oxygen.

In a specific implementation, the cross noise covariance S is a matrix whose structure is designed to describe the statistical dependency between the measurements of different gas components. For example, if primary tracking in the systemAndS will be a 2x2 matrix in which each elementIndicating how a change in component i affects the measurement error of component j. The specific values of such matrices may be estimated based on experimental data, chemical kinetics models, or by data-driven methods.

In an implementation, the system first initializes basic parameters including a state vector, an error covariance matrix, a state transition matrix, and an observation matrix. The state vector includes the estimated concentrations of all critical gas components. Alternatively, the cross noise covariance matrix described above may also be defined.

At each data acquisition cycle, the system first performs a prediction step, using a state transition matrix to estimate the new state of each gas constituent. Then, in an update step, the system collects actual gas concentration data and adjusts and optimizes the state estimation using the observation matrix and the cross noise covariance matrix. This process involves calculating the kalman gain and applying it to integrate the new measurement data, updating the state vector and the error covariance matrix.

Finally, the system aligns the corrected gas composition data with the data collected from other sensors (e.g., optical and electrical sensors) to ensure that all data is based on a uniform time axis. This is done by data interpolation or synchronization algorithms to ensure that all data used in the process and control is accurate and consistent.

In this way, the processing module 120 is able to accurately synchronize and align the optical data, electrical data, and gas composition data, and to fuse the data from different sources by weighted average to ensure overall consistency and accuracy of the data, providing reliable data support for feedback control and process optimization. In this way, the processing module 120 can effectively synchronize and align different types of data, and adopts a proper method according to respective characteristics, so that the accuracy and reliability of overall data fusion are improved, and accurate data support is provided for subsequent process control and optimization.

In semiconductor manufacturing, for example, two gas components are respectivelyAnd. These two gases interact during plasma etching and may affect each other's measurement accuracy.

The parameters were set as follows: state vectorComprisesAndExpressed as the concentration ofState transition matrixFor example, the gas concentration is changed slowly, provided thatIs a unit matrix; observation matrixAlso, an identity matrix, which represents a direct measurement of the concentration of each gas; process noise covariance matrixIn consideration of the stability of the system, is set as; Measuring noise covariance matrixBased on the sensor performance, set as; Interactive noise covariance matrixTaking into accountMay affect the decomposition of (C)Measurement is set as；

The data processing flow is as follows:

Initializing: (initial) AndConcentration estimation),(Initial error covariance);

Real-time data acquisition based on acquisition module 110, at time point k, measured AndThe concentration of (2) is 102ppm and 48ppm, respectively, i.e. Wherein ppm (part per million) denotes the parts per million,Is a new measurement obtained from the mass spectrometer.

Performing a Kalman filter;

the state prediction formula:

；

wherein; Representing an estimate of the system state at the instant of time step k, given all previously observed information up to time k-1 (i.e., the observed data not including time k). Representing the state estimate of the last time step k-1.

Based on the historical gas constituent concentration (e.gAnd) To predict the concentration at the current time point. Here a state transition matrixIt is described how the gas concentration shifts from one point in time to the next, often assuming that these shifts are linear or nearly linear over a short period of time, reflecting the natural course of gas reaction and physical diffusion in the plasma.

Error covariance prediction formula:

；

the above equation updates the uncertainty regarding the gas concentration estimate. Wherein, Is the error covariance of the previous step and represents the uncertainty of the estimated state in the last time step measurement.

Updating:

；

The above method is used for calculating Kalman gain, and determining new measurement data (actually measured AndConcentration) on the state estimate. Wherein, Is a prediction error covariance matrix. It represents the uncertainty of the system state estimate given the information at time k, given all previous times k-1; Is an observation matrix. Mapping the state space to a measurement space for converting the state variable into an observable output; is a measurement matrix For proper dimensional matching in matrix multiplication and remapping information of the observation space back to the state space; Is a measurement noise covariance matrix that describes uncertainty or errors in the measurement process, including sensor errors, etc.; Is the cross noise covariance matrix.

For the kalman gain, this matrix is the core of the filtering process, which determines how much "weight" should be given to the new measurement data at the time of the state estimate update. This gain helps balance the weights of the predictions and measurements, optimizing the accuracy of the state estimate.

；

The above equation is used to update the estimate of the gas concentration. The difference term in the formulaRepresenting the error between the prediction and the actual observation, the kalman gain is used to adjust the contribution of this error to the final state estimate.

；

The above equation is used to update the uncertainty of the state estimate. Introducing the identity matrix I ensures that the updated covariance maintains the correct dimensions and properties.

It will be appreciated that in the example of calculation of the Wen Kaer man filter above, each parameter is a matrix or vector, and k|k-1 represents the prediction of the state of time k given all the information of the deadline k-1, which is part of the prediction step of the kalman filter, for representing the state prediction prior to the current observation, unless otherwise specified; t represents a transpose of the matrix or vector; i represents an identity matrix, wherein elements on the diagonal line of the identity matrix are 1, and other positions are 0; -1 represents the inverse of the matrix, i.e. the inverse of one matrix is multiplied by the original matrix to obtain the identity matrix.

Using updatedTo correct and align the gas composition data while ensuring synchronization (optical, electrical) with other sensor data.

It will be appreciated that, because the specific parameters of the various processes, systems, and apparatus are not the same, the specific parameter settings of the kalman filter described above should be combined with the actual requirements, for example, the state transition matrix may be determined according to the physical and chemical characteristics of the gas components. The state transition matrix may be set as an identity matrix or a linear equation coefficient matrix if the behavior of the gas composition over the time interval is approximately linear. If the gas concentration variation is known it can be described by a first order linear dynamic model and these coefficients can be fitted by experimental data. For more complex dynamics, such as non-linear behavior affected by temperature and pressure changes, advanced models (such as physics-based models or data-driven machine learning models) may be used to estimate these coefficients.

The observation matrix describes how observations are derived from state variables. In general, the observation matrix may be an identity matrix if each state variable is measured directly. If the observations are a function or combination of state variables, such a relationship needs to be reflected in the observation matrix. This may be determined by a physical understanding of the system or by experimental data.

The process noise covariance matrix reflects the uncertainty in the model predictions and external disturbances. This typically needs to be determined from historical data. In some scenarios, the process noise covariance matrix may be dynamically adjusted by comparing model prediction and actual measurement data.

The measurement noise covariance matrix reflects the statistical properties of the measurement errors and can be determined based on the specifications of the measurement device or by statistical analysis of the actual measurement errors. For a multi-sensor system, if the accuracy of the sensors is known, the measurement noise covariance matrix can be set directly to the variance of these accuracies.

The cross noise covariance matrix describes the interactions between different gas components. This can be determined by chemical kinetic modeling or control experiments by measuring the response of one gas component as the other component changes. In a data driven approach, the cross noise covariance matrix can be estimated by statistically analyzing the cross influence between multiple variables.

The initial value of the error covariance matrix is typically large, indicating uncertainty in the initial state estimate. The subsequent values are dynamically adjusted by an iterative update process of the kalman filter. The updating of this matrix reflects the accuracy of the system state estimation, progressively refined by the predictive and update formulas described above.

Thus, by integrating the cross noise covariance matrix into the Kalman filter update step, the accuracy of the gas component concentration estimation can be significantly improved. In particular, this allows the filter to take into account the effects caused by other constituent variations when updating the state estimate, thereby more truly reflecting the actual dynamics of the gas constituent. This approach can improve the accuracy of control of the etching process because accurate estimates of the gas composition concentration directly affect the chemical and physical behavior of the etchant.

For the processing module 120:

In particular implementations, the processing module 120 receives different types of characteristic data aligned to the same time axis and inputs the characteristic data into a machine learning model trained based on historical data to generate process state assessment information.

As an alternative embodiment, a Principal Component Analysis (PCA) is used to process and analyze a variety of characteristic data, simplify the data structure, identify key factors affecting process performance, and understand the relationships between the variables. And (3) carrying out standardization treatment on the data from different sources, and converting the data into dimensionless data. This can be achieved by dividing the mean value by the standard deviation. The normalization process can eliminate the difference between different dimension data, so that the dimension data are compared on the same scale. Spectral features, such as feature peaks, spectral intensities, line widths, etc., are extracted, which reflect information about ion density, energy distribution, and electron temperature of the plasma. Voltage, current and power characteristics, such as voltage fluctuation amplitude, power consumption rate, etc., are extracted, which reflect the energy input and reactivity. The concentration of different gas components and the variation trend thereof, such as the instantaneous concentration, the average concentration variation rate and the like of specific gases, are extracted, and the characteristics reflect the progress of chemical reaction.

The normalized feature data is constructed into a data matrix, where each row represents one sample (e.g., data of one etching process) and each column represents one feature (e.g., spectral intensity, power consumption rate, etc.). Covariance matrices of the data matrices are calculated to learn correlations between different features. And carrying out eigenvalue decomposition on the covariance matrix to obtain eigenvalues and eigenvectors. The eigenvalues represent the importance of the principal component and the eigenvectors represent the direction of the principal component. The first few principal components are selected according to the magnitude of the eigenvalues, which components are able to interpret the majority of the variance of the data. And selecting a main component with the characteristic value accumulation contribution rate reaching more than 90%. And projecting the original data onto the principal component to generate a dimensionality reduced data set. These datasets are easier to analyze and understand in the new space.

In a specific implementation, the data sent by the acquisition module 110 is first normalized. This step is necessary because the dimensions and magnitude of the data sources may vary, and direct use of such data may result in bias or inaccuracy in the machine learning model. Normalization typically involves de-averaging and variance scaling, i.e., subtracting the average value from each feature and dividing by the standard deviation, such that the resulting data has an average value of 0 and a standard deviation of 1. The data thus processed is referred to as dimensionless data, facilitating subsequent analysis and model training.

Next, key features are extracted from the normalized dimensionless data, including spectral features, voltage features, current features, power features, and concentrations of gas components and their trends. These features are important indicators describing the plasma state and they are used to capture critical physical and chemical information during processing. Feature extraction involves not only the direct use of raw data, but may also include some derivative calculations such as peak detection, average, coefficient of variation, etc.

The extracted features are organized into a data matrix, where each row represents a data sample (i.e., an observation of a point in time) and each column represents a feature. This form of data organization facilitates statistical analysis and training of machine learning models.

A covariance matrix of the data matrix is calculated, which is used to analyze the linear relationship between the features. Then, eigenvalue decomposition is performed, from which eigenvalues and eigenvectors are obtained. The eigenvalues represent the variance of each principal component, i.e., the importance of that component in the data; the feature vector then indicates the direction of each principal component.

Based on the magnitude of the characteristic value, the principal component with the accumulated contribution rate reaching more than 90% is selected, which is a dimension reduction technology, and can reduce the complexity of a model and simultaneously retain the most important information. These principal components constitute a new, dimensionality-reduced dataset.

The reduced dimension dataset is used to train a preset machine learning model. This model will be used to predict and evaluate the state during plasma processing in real time, such as predicting future quality trends or potential process anomalies.

As an alternative implementation, the machine learning model is trained by using the dimensionality reduced data to realize classification and prediction of the process state.

In a specific implementation, based on the reduced-dimension data set, the processing module 120 divides the data into a training set and a testing set, so that training and verification of the model are performed on different data, and overfitting is avoided. An appropriate machine learning model, such as a support vector, random forest, or neural network, is selected and trained using the training set to learn patterns and rules in the data. The performance of the model was evaluated using the test set.

In actual operation, the real-time collected reduced-dimension data is input into a trained machine learning model, and the model can classify and predict the technological state. For example, the model may predict that the current process state is "normal," "ion density is too high," or "power consumption is abnormal," etc. Based on the predicted results of the model, the processing module 120 generates process state estimation information.

In particular implementations, based on the process state evaluation information, the processing module 120 further generates a particular control strategy, i.e., a first process parameter, to achieve process optimization and stability.

The energy distribution of ions in the plasma is adjusted according to specific surface treatment requirements. For example, during deep etching, ions of higher energy are required to penetrate the material surface. The processing module 120 instructs to increase the high frequency power output to increase the average energy of the plasma based on the model prediction.

In the surface cleaning process, ions with lower energy can meet the requirements. The processing module 120 directs the adjustment of the low frequency power to reduce the average energy of ions in the plasma and avoid excessive damage to the material surface.

In practice, the processing module 120 dynamically adjusts the output and frequency combination of the high-frequency power and the low-frequency power by monitoring the ion energy distribution in real time and combining the model prediction result, so as to realize the optimal energy distribution control.

In a specific implementation, the control strategy is optimized using a genetic algorithm to generate a first process parameter of the target plasma. The frequency combination of the multi-frequency power output is adjusted in real time according to different surface treatment requirements in the etching process information acquired by the acquisition module 110. The genetic algorithm calculates the optimal frequency combination according to the real-time data and the process requirement.

As an alternative embodiment, the processing module 120 performs the following steps when optimizing using a genetic algorithm to generate the first process parameter of the target plasma based on the process state evaluation information:

In a specific implementation, a genetic algorithm is used to adjust the frequency combination of the multi-frequency power output in real time according to different surface treatment requirements.

During the multi-frequency modulated plasma surface treatment, possible power allocation and frequency combinations are covered by population initialization, thus ensuring a broad search of the algorithm in the solution space.

Each individual represents a scheme of power allocation and frequency combining. Considering the importance of high frequency power, low frequency power and frequency combinations in this scenario, an individual may be represented in the form:

；

Wherein, : The high-frequency power is supplied to the power supply,: The power of the low frequency is such that,: And (5) frequency combination.

High frequency powerAnd low frequency powerThe value range of (2) is required to be set according to the actual process requirement. For example, the high frequency power may range from 20-100 watts and the low frequency power may range from 10-50 watts.

Frequency combinationThe value of (2) is based on the process requirements. For example, common frequencies may include 2.45MHz, 13.56MHz, 27.12MHz, etc. Two or more frequencies may be selected for combining.

The size of the population determines how many individuals in each generation are involved in the evolution. The population size is set to 50 to 100 individuals to ensure diversity and computational efficiency.

Within a set range, individuals of the initial population are randomly generated. The high frequency power, the low frequency power and the frequency combination of each individual are randomly selected within a corresponding range.

In the setting of the fitness function, the main objectives of the optimization may include improving etching efficiency, optimizing material and energy utilization, and improving product quality. These goals reflect key performance metrics that are sought after in operation of the plasma processing system, such as etch rate, energy consumption, material usage efficiency, and surface quality of the final product.

To materialize these objectives, the fitness function may include a plurality of quantization indices, such as: etching efficiency: evaluating by material removal rate per unit time; material utilization rate: the effective use of materials is measured, and the waste is reduced; energy efficiency: calculated by the energy consumption of the system, higher efficiency means lower energy consumption; the product quality is as follows: by detecting the surface quality of the etched product.

These indices may be combined into a comprehensive fitness function, such as:

；

wherein E, U, P, G respectively represent etching efficiency, material utilization, energy efficiency and product quality ,,,Is a corresponding weight, and is adjusted according to the priorities of different targets.

In the implementation process, the fitness function can be dynamically adjusted according to real-time process feedback or external environment change. For example, if a significant change in energy costs is detected, the weight of the energy efficiency may be adjusted appropriately to accommodate the new cost structure.

By repeatedly performing the steps of selecting, crossing, and mutating, and evaluating the performance of each new born individual according to the fitness function, the system can gradually find the optimal solution. The whole process is continuously carried out until the preset iteration times or the convergence of the adaptability is reached, and finally, the parameter configuration capable of achieving the optimal technological effect is determined.

As an alternative embodiment, non-uniform variation, a variation method to better adapt to different types of data characteristics, may be used in the genetic algorithm optimization process. For the optimization problem in the multi-frequency modulation plasma surface treatment process, different mutation treatments can be carried out on the high-frequency power, the low-frequency power and the frequency combination.

The core idea of non-uniform variation is to use different variation strategies for different parameters. For example, the power value is subjected to gaussian variation (variation in a small range around the current value), and the frequency combination is subjected to crossover variation (crossover of two frequencies in the frequency combination).

In a specific implementation, the variation of the high frequency power and the low frequency power is varied by a gaussian distribution. The formula:

；

Wherein, For a new individual value to be obtained,For the current individual value(s),Representing a mean of 0 and a variance ofIs a gaussian noise of (c). The high frequency power and the low frequency power of each individual are subjected to gaussian variation, respectively. Selecting an appropriate varianceTo control the variation amplitude.

By way of example, a certain individual has a high frequency power of 60 watts and a low frequency power of 30 watts. After the gaussian variation, the high frequency power may become:

；

The low frequency power may become:

；

Wherein P _high-new、P_low-new represents the high-frequency power and the low-frequency power after Gaussian mutation, respectively.

In a specific implementation, the variation of the frequency combination is achieved by exchanging two frequencies in the frequency list. This approach ensures that the characteristics of the frequency combination remain unchanged, but the order changes. The two frequencies in the frequency combination are randomly selected and their positions are swapped. Ensuring that the mutated frequency combination is within the allowable frequency range.

Illustratively, assume that the frequency combination of a certain individual is [13.56MHz,27.12MHz ]. After the crossover mutation, the frequency combination may become:

；

Wherein F _comb-new represents the frequency combination after the crossover mutation.

Therefore, by the non-uniform variation method, the genetic algorithm can be more effectively adapted to different data characteristics in the surface treatment of the multi-frequency modulation plasma, and the optimization effect and efficiency are improved.

For the control module 130:

In a specific implementation, the control module 130 controls the power output at different frequencies based on the first process parameter output by the processing module 120 to optimize the ion energy distribution and density of the plasma, thereby achieving the best etching effect.

As an alternative embodiment, the control module 130 includes: a power amplifier, a radio frequency source, and a power distribution network; the power amplifier is used for increasing the power of a radio frequency signal generated by the radio frequency source, and each target frequency corresponds to an independent power amplifier; the radio frequency source is used for generating a radio frequency signal of target frequency and power based on the first process parameter and inputting the radio frequency signal to the power amplifier; the radio frequency source comprises a plurality of frequency generation units, wherein each frequency generation unit is used for generating a target frequency; the power distribution network is used for distributing the amplified radio frequency signals to the plasma source.

As an alternative embodiment, the control module 130 performs the following steps when performing multi-frequency modulation on the target plasma based on the first process parameter, and adjusting the physical properties of the target plasma: setting the frequency of the radio frequency signal generated by each frequency generating unit based on the target frequency included in the first power parameter by using the radio frequency source, and transmitting the radio frequency signal to a power amplifier corresponding to each target frequency; the power amplifier amplifies the output power of the radio frequency signals of each frequency based on the output power of the target frequency; the power distribution network receives the radio frequency signals of all frequencies after amplifying the output power and distributes the radio frequency signals to different electrodes or areas of the plasma source to adjust the physical properties of the target plasma.

Among other things, a power amplifier is a device for increasing the power of an electrical signal, particularly in radio frequency applications, so that the signal can effectively drive a load such as an antenna or a plasma source.

The rf source is a device that generates rf electromagnetic waves for generating signals of a specific frequency and power to control various process equipment, such as a plasma generator.

The power distribution network is a network system that receives an input power and distributes it to a plurality of outputs, ensuring that the components receive the proper energy.

A plasma source is a device for generating a plasma, i.e. a gas containing free electrons and ions, commonly used for material processing and surface treatment.

In a specific implementation, each frequency generation unit in the radio frequency source generates a desired specific frequency based on the first process parameter. These parameters include the frequency and power level required to meet specific physical processing requirements. The radio frequency signal generated by the radio frequency source is first input to the corresponding power amplifier. Each target frequency corresponds to a separate power amplifier to ensure that the signal reaches a sufficient power level before being transmitted to the plasma source. The amplified signals are sent to different electrodes or regions of the plasma source through a power distribution network. This step ensures that each portion of the plasma source receives the proper energy to uniformly and efficiently ignite a plasma.

Thus, by precisely controlling the power and frequency of each portion of the plasma source, the generation and maintenance of the plasma can be optimized, thereby improving the processing efficiency and uniformity of the results. The multi-frequency modulation allows the system to be quickly and flexibly adjusted for different materials and processing requirements, enhancing the application range and efficiency of the system. By optimizing the power output and frequency usage, the system can reduce unnecessary energy consumption and reduce operating costs.

In a multi-frequency modulated plasma surface treatment system, an efficient power distribution network is critical to ensure that the rf power is accurately and uniformly distributed to each region of the plasma source according to the process requirements.

As an alternative embodiment, the goal of the power distribution network is to achieve a uniform distribution of the radio frequency signal and to allow flexible adjustment of the power output according to process variations.

In implementations, a ring coupler or wilkinson power splitter may be used. These passive power splitters provide low insertion loss and good uniformity. The output of each splitter may be connected to an adjustable attenuator to achieve fine tuning of the output power.

The layout of the power distribution network should be designed according to the geometry of the plasma source. For example, if the plasma source has multiple independent reaction zones, a centralized distribution node may be designed to draw the main line from the rf source and then distribute the power to the various zones through splitters. Each branch should be designed with sufficient shielding and proper impedance matching to reduce reflection and attenuation of the radio frequency signal.

In order to achieve dynamic regulation, the adjustable attenuator of each branch is managed by a control system. The control system can automatically adjust the attenuator settings based on sensor data (e.g., current, voltage, and spectral data) returned from the plasma source in response to changes in the process.

It will be appreciated that the control system manages each of the adjustable attenuators to achieve automatic adjustment, which is not the subject of the present invention and will not be described in detail herein.

Illustratively, in a plasma cleaning scenario for semiconductor wafers, the wafer diameter is 300 millimeters. The designed power distribution network uses a central splitter to divide the main rf signal into four regions. Each region is precisely regulated and controlled through an independent adjustable attenuator so as to adapt to the cleaning requirements of different regions on the wafer. The uniformity and efficiency of the cleaning process are ensured by monitoring the cleaning effect of the surface of the wafer in real time and adjusting the power output of the corresponding area.

Based on the same inventive concept, the embodiment of the present invention further provides a multi-frequency modulation plasma surface treatment method, and since the principle of solving the problem in the method in the embodiment of the present disclosure is similar to that of the multi-frequency modulation plasma surface treatment system in the embodiment of the present disclosure, the implementation of the method may refer to the implementation of the system, and the repetition is omitted.

Referring to fig. 2, a flow of a multi-frequency modulation plasma surface treatment method according to an embodiment of the disclosure includes steps S101 to S103, which are specifically as follows:

S101: acquiring etching process information, and acquiring characteristic data of target plasmas under multi-frequency modulation by a sensor, wherein the characteristic data comprise optical data, electrical data and gas composition data; aligning different types of characteristic data to the same time axis;

S102: inputting the characteristic data of the target plasma aligned to the same time axis to a preset machine learning model to generate process state evaluation information; optimizing by utilizing a genetic algorithm based on the process state evaluation information to generate a first process parameter of the target plasma;

S103: and based on the first technological parameter, performing multi-frequency modulation on a plasma source of the target plasma, and adjusting the physical property of the target plasma.

The disclosed embodiment of the invention also provides a computer device, as shown in fig. 3, which is a schematic structural diagram of the computer device provided by the disclosed embodiment of the invention, comprising:

A processor 31 and a memory 32; the memory 32 stores machine readable instructions executable by the processor 31, the processor 31 being configured to execute the machine readable instructions stored in the memory 32, the machine readable instructions when executed by the processor 31, the processor 31 performing the steps of:

acquiring target plasma etching process information, and acquiring characteristic data of target plasma under multi-frequency modulation, including optical data, electrical data and gas composition data, through a sensor; aligning different types of characteristic data to the same time axis;

The memory 32 includes a memory 321 and an external memory 322; the memory 321 is also referred to as an internal memory, and is used for temporarily storing operation data in the processor 31 and data exchanged with an external memory 322 such as a hard disk, and the processor 31 exchanges data with the external memory 322 via the memory 321.

The specific implementation process of the above instruction may refer to the implementation content of a multi-frequency modulation plasma surface treatment system in the embodiments of the disclosure, which is not described herein.

It should be understood that, in various embodiments of the present invention, the sequence numbers of the foregoing processes do not mean the order of execution, and the order of execution of the processes should be determined by the functions and internal logic thereof, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

In the description of the present specification, the descriptions of the terms "one embodiment," "example," "specific example," and the like, mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiments or examples. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

The preferred embodiments of the invention disclosed above are intended only to assist in the explanation of the invention. The preferred embodiments are not intended to be exhaustive or to limit the invention to the precise form disclosed. Obviously, many modifications and variations are possible in light of the above teaching. The embodiments were chosen and described in order to best explain the principles of the invention and the practical application, to thereby enable others skilled in the art to best understand and utilize the invention. The invention is limited only by the claims and the full scope and equivalents thereof.

Claims

1. A multi-frequency modulation plasma surface treatment system, characterized in that it comprises: a collection module, a processing module and a control module;

The acquisition module is used to obtain etching process information and obtain characteristic data of the target plasma under multi-frequency modulation through sensors, including optical data, electrical data and gas composition data; and align different types of characteristic data to the same time axis;

The processing module is used to input the characteristic data of the target plasma aligned to the same time axis into a preset machine learning model to generate process state evaluation information; based on the process state evaluation information, optimize using a genetic algorithm to generate a first process parameter of the target plasma;

The control module is used to perform multi-frequency modulation on a plasma source that generates the target plasma based on the first process parameter to adjust the physical properties of the target plasma.

2. The multi-frequency modulation plasma surface treatment system according to claim 1, characterized in that the acquisition module comprises: an optical sensor, an electrical sensor and a mass spectrometer;

The optical sensor is installed at the window of the plasma etching chamber and is used to monitor the optical data of the target plasma in real time, wherein the optical data includes: emission spectrum and light intensity;

The electrical sensor is directly connected to the plasma source and is used to measure the electrical data of the plasma, wherein the electrical data includes: voltage, current and power;

The mass spectrometer is connected to the plasma etching chamber through a gas sampling system, and is used to monitor the composition and ion types of the reaction gas, and to analyze the gas composition in the plasma in real time to generate gas composition data.

3. The multi-frequency modulation plasma surface treatment system according to claim 2, characterized in that the acquisition module also includes a data processing unit;

The data processing unit is used to pre-process the optical data, electrical data and gas composition data of the target plasma, and align different types of characteristic data to the same time axis.

4. The multi-frequency modulation plasma surface treatment system according to claim 3, wherein aligning different types of characteristic data to the same time axis comprises:

For the optical data, for each data collection time point, a cubic spline interpolation method is used to calculate the optical data at the time point according to adjacent known optical data points;

For the electrical data, for each data collection time point, a linear interpolation method is used to calculate the electrical data at that time point based on adjacent known electrical data points;

For the gas composition data, at each data collection time point, a Kalman filter is used to correct and interpolate the gas composition data to calculate the gas composition data at that time point.

5. The multi-frequency modulation plasma surface treatment system according to claim 4, characterized in that the Kalman filter comprises:

an interaction noise covariance matrix, wherein the elements of the interaction noise covariance matrix are defined as the impact of changes in different gas components on the measurement of other gas components during the processing of the target plasma;

The state prediction mechanism predicts the state of the current time step by applying the state transfer matrix to the state estimate of the previous time step;

The error updating mechanism updates the state estimation and error covariance through the observation matrix, the actual measurement value and the actual measurement noise covariance, and adjusts the state estimation value of the gas composition.

6. The multi-frequency modulation plasma surface treatment system according to claim 5, characterized in that, before the processing module is used to input the characteristic data of the target plasma aligned to the same time axis into a preset machine learning model, it also includes:

The historical characteristic data of the target plasma under multi-frequency modulation is standardized and converted into dimensionless data;

Extracting spectral features, voltage features, current features, power features, and concentrations of different gas components and their changing trends from the dimensionless data to generate standardized feature data;

Constructing a data matrix based on the standardized characteristic data; wherein each row of the data matrix represents a data sample of characteristic data of a target plasma, and each column of the data matrix represents a feature;

Calculate the covariance matrix of the data matrix, and perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues and eigenvectors; wherein the eigenvalues represent the importance of the principal component, and the eigenvectors represent the direction of the principal component;

Based on the size of the eigenvalue, a principal component whose cumulative contribution rate of the eigenvalue reaches more than 90% is selected;

Projecting the data matrix onto the principal components to generate a first dimensionally reduced data set;

The preset machine learning model is constructed using the first dimension reduction data set.

7. The multi-frequency modulation plasma surface treatment system according to claim 6, characterized in that the optimization based on the process state evaluation information and the use of a genetic algorithm to generate the first process parameter of the target plasma comprises the following steps:

Generate an initial population where each individual represents a set of power output and frequency combinations;

designing a fitness function based on the process state evaluation information and the target plasma etching process information;

Performing selection, crossover and mutation operations to explore the solution space, and dynamically adjusting the fitness function based on the process state evaluation information;

The mutation operation is a non-uniform mutation, including: performing Gaussian mutation on the power output and performing exchange mutation on the frequency combination;

Repeat the above steps to evaluate the fitness values of the newly generated individuals, and select individuals whose fitness meets the first target threshold to enter the next generation until the fitness value converges or reaches the preset number of iterations;

In response to the fitness value converging or reaching a preset number of iterations, a first process parameter of the target plasma is generated based on the current individual.

8. The multi-frequency modulation plasma surface treatment system according to claim 7, characterized in that the control module comprises: a power amplifier, a radio frequency source and a power distribution network;

The RF source generates a RF signal of target frequency and power based on the first process parameter, and inputs the signal into the power amplifier; wherein the RF source includes a plurality of frequency generating units, each of which is used to generate a target frequency;

The power amplifier amplifies the power of the radio frequency signal generated by the radio frequency source based on the target frequency, and each target frequency corresponds to an independent power amplifier;

The power distribution network is used to distribute the power of the amplified radio frequency signal to the plasma source.

9. The multi-frequency modulation plasma surface treatment system according to claim 8, characterized in that the multi-frequency modulation of the target plasma based on the first process parameter to adjust the physical properties of the target plasma comprises:

The RF source sets the frequency of the RF signal generated by each of the frequency generating units based on the target frequency, and sends the frequency to the power amplifier corresponding to each of the target frequencies;

The power amplifier amplifies the output power of the radio frequency signal at each frequency based on the output power of the target frequency;

The power distribution network receives radio frequency signals of various frequencies after amplifying the output power and distributes them to different electrodes or regions of the plasma source to adjust the physical properties of the target plasma.

10. A multi-frequency modulation plasma surface treatment method, which is implemented based on the multi-frequency modulation plasma surface treatment system according to any one of claims 1 to 9, characterized in that it comprises:

Obtain etching process information and obtain characteristic data of target plasma under multi-frequency modulation through sensors, including optical data, electrical data, and gas composition data; align different types of characteristic data to the same time axis;

Inputting the characteristic data of the target plasma aligned to the same time axis into a preset machine learning model to generate process state evaluation information; optimizing using a genetic algorithm based on the process state evaluation information to generate a first process parameter of the target plasma;

Based on the first process parameter, a plasma source of the target plasma is multi-frequency modulated to adjust the physical properties of the target plasma.