Process parameter optimization methods, systems, in-situ online monitoring devices, computer equipment, and storage media

Abstract

This application provides a method, system, in-situ online detection device, computer equipment, and storage medium for optimizing process parameters. The method involves acquiring process parameters and measurement data within a preset window; extracting characteristic parameters representing the physical state of wafer surface growth based on the measurement data; inputting the characteristic parameters and process parameters into a pre-trained prediction model to obtain prediction results representing the growth state trend and photoelectric property change trend; optimizing the process parameters based on the prediction results, and using the optimized process parameters for the epitaxial growth process of the wafer in the next preset window. By employing this approach, the application extracts characteristic parameters representing the physical state of growth from the original measurement data. Based on these characteristic parameters, the application automatically analyzes the high-dimensional relationships between multiple parameters and finds the globally optimal process window without relying on prior models and parameters. The generated prediction results are consistent with the actual growth mechanism, effectively improving the accuracy of the process and product yield.

Description

Technical Field

[0001] This application relates to the field of semiconductor testing technology, and in particular to a method, system, in-situ online testing device, computer equipment, and storage medium for optimizing process parameters. Background Technology

[0002] Metal-organic chemical vapor deposition (MOCVD) is a core epitaxial technology for fabricating compound semiconductor optoelectronic and microelectronic devices. It achieves atomically precise single-crystal thin film growth by precisely controlling the chemical reaction between an organometallic source and a hydride on a heated substrate. This technology is widely used in the fabrication of high-end devices such as vertical-cavity surface-emitting lasers (VCSELs), gallium nitride-based high electron mobility transistors (HEMTs), and light-emitting diodes (LEDs). These devices typically possess complex structures with nanometer-thickness quantum wells, superlattices, and distributed Bragg mirrors (DBRs), thus placing extremely high demands on the precision of growth rate (GR), composition, thickness, and interface quality control during the growth process.

[0003] Traditional monitoring methods largely rely on prior physical models such as the Transfer Matrix Model (TMM). After collecting reflection spectrum data, it is fitted with TMM simulation results based on preset thickness and optical constants. The current growth rate and thickness are then calculated using inversion algorithms such as least squares. These models are highly dependent on precise prior optical constants. For applications with unknown or variable optical constants, such as new material systems, gradient layers, or complex interfaces, prediction biases can occur. These biases are further introduced into process optimization as state variables, leading to poor optimization of process parameters. Summary of the Invention

[0004] The purpose of this application is to provide a method, system, in-situ online detection device, computer equipment, and storage medium for optimizing process parameters, so as to overcome the problem that the current growth rate and thickness are calculated based on preset thickness and optical constants in traditional technologies. However, in application scenarios such as new material systems, gradient layers, or complex interfaces where the optical constants are unknown or change with the process, prediction deviations occur, resulting in poor optimization of process parameters.

[0005] In a first aspect, this application proposes a process parameter optimization method, the method comprising: acquiring process parameters of the equipment and measurement data of the process chamber within a preset window; wherein the measurement data is obtained by observing the wafer surface of the epitaxial growth process within the process chamber through an observation window; extracting feature parameters characterizing the physical state of the wafer surface growth based on the measurement data; inputting the feature parameters and the process parameters into a pre-trained prediction model to obtain a prediction result; the prediction result is used to characterize the growth state trend and the photoelectric property change trend; and optimizing the values ​​of the current process parameters within a pre-defined specification range based on the prediction result, the optimized process parameters being used for the epitaxial growth process of the wafer in the next preset window.

[0006] In one embodiment, the measurement data includes a reflectance spectral signal; the feature parameters include an instantaneous growth rate; the extraction of feature parameters characterizing the physical state of wafer surface growth based on the measurement data includes: performing narrowband weighted integration on the reflectance spectral signal at a preset monitoring wavelength to obtain a reflectance time series; processing the reflectance time series using a real-time moving window algorithm to obtain the current oscillation period; for each oscillation period, calculating the product of the thickness increment within that oscillation period and the number of periods per unit time to obtain the instantaneous growth rate; wherein the thickness increment is related to the refractive index of the material and the monitoring wavelength, and the refractive index and the monitoring wavelength are obtained based on the reflectance spectral signal.

[0007] In one embodiment, the feature parameters further include stop band feature data; the step of extracting feature parameters characterizing the physical state of wafer surface growth based on the measurement data further includes: for the periodic structure of epitaxial growth, based on the reflection spectrum signal, performing Lorentz or Gaussian line fitting on each frame of reflection spectrum obtained after calibration and normalization within a preset stop band wavelength range, dynamically tracking and outputting stop band feature data; wherein, the stop band feature data includes the center wavelength and / or full width at half maximum (FWHM).

[0008] In one embodiment, the measurement data further includes wafer temperature data and wafer curvature data; the feature parameters further include hybrid feature parameters; the step of extracting feature parameters characterizing the physical state of wafer surface growth based on the measurement data further includes: using the reflection spectrum signal as observation data, combining the constraints of the wafer temperature data on the optical constant temperature coefficient, and the constraints of the wafer curvature data on the stress state, performing inversion fitting using a transfer matrix model to obtain hybrid feature parameters; the hybrid feature parameters include equivalent optical constants and equivalent thickness; wherein, the wafer temperature data originates from infrared radiation on the wafer surface, and the wafer curvature data originates from the reflection image of the laser signal on the wafer surface.

[0009] In one embodiment, the prediction model includes a timing prediction model and a photoelectric characteristic prediction model; the feature parameters include instantaneous growth rate, process parameter statistical features, final wafer curvature value, mixed feature parameters, and stop band feature data; the measurement data includes wafer temperature data; the step of inputting the feature parameters and the process parameters into the pre-trained prediction model to obtain the prediction result includes: inputting the instantaneous growth rate, the wafer temperature data, and the process parameters into the encoder of the timing prediction model to obtain a context vector representing the historical sequence; inputting the context vector into the decoder of the timing prediction model to obtain a growth rate prediction value; performing integration processing on the growth rate prediction value to obtain a thickness prediction value; inputting the thickness prediction value, the process parameter statistical features, the final wafer curvature value, the mixed feature parameters, and the stop band feature data into the photoelectric characteristic prediction model to obtain a predicted value of the photoelectric characteristic parameters; wherein, the photoelectric characteristic parameters include the emission wavelength of the light-emitting device and / or the sheet resistance of the power semiconductor device.

[0010] In one embodiment, optimizing the values ​​of the current process parameters within a pre-defined specification range based on the prediction results includes: comparing the prediction results with a preset target specification to obtain deviation information; wherein the prediction results include predicted thickness values ​​and predicted values ​​of photoelectric characteristic parameters; constructing a state vector based on the deviation information and process parameters; processing the state vector using a reinforcement learning algorithm to obtain process parameter adjustment amounts; and constraining the process parameter adjustment amounts within a preset safe range of process parameters to obtain optimized process parameters.

[0011] In one embodiment, the method further includes: generating a display chart based on the prediction result; wherein the display chart includes at least one of a real-time growth rate curve, a stop band center wavelength evolution trend chart, and a predicted photoelectric property change trend distribution chart.

[0012] Secondly, this application proposes a process parameter optimization system, the system comprising:

[0013] The acquisition module is used to acquire the process parameters of the machine and the measurement data of the process chamber within a preset window; wherein, the measurement data is obtained by observing the wafer surface of the epitaxial growth process inside and outside the process chamber through the observation window.

[0014] The feature extraction module is used to extract feature parameters characterizing the physical state of wafer surface growth based on the measurement data;

[0015] The prediction module is used to input the feature parameters and the process parameters into a pre-trained prediction model to obtain prediction results; the prediction results are used to characterize the growth state trend and the photoelectric property change trend.

[0016] The optimization module is used to optimize the values ​​of the current process parameters within a pre-defined specification range based on the prediction results. The optimized process parameters are used for the epitaxial growth process of the wafer in the next preset window.

[0017] Thirdly, this application also provides an in-situ online detection device, comprising:

[0018] The data acquisition module is used to detect the wafer surface of the epitaxial growth process outside the process cavity through the observation window within a preset window and obtain measurement data.

[0019] The process parameter optimization system described in the second aspect is used to optimize the values ​​of process parameters based on the measurement data and the current process parameters, and to use the optimized process parameters in the epitaxial growth process of the wafer in the next preset window; wherein the current process parameters are sent to the in-situ online detection device by the manufacturing execution system or the machine.

[0020] Fourthly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method steps of the first aspect.

[0021] Fifthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the method steps of the first aspect.

[0022] The above-mentioned process parameter optimization methods, systems, in-situ online detection devices, computer equipment, and storage media have at least the following advantages:

[0023] This application acquires the process parameters of the equipment and measurement data within a preset window; based on the measurement data, it extracts feature parameters characterizing the physical state of wafer surface growth; the feature parameters and process parameters are input into a pre-trained prediction model to obtain prediction results, which are used to characterize the growth state trend and photoelectric property change trend; based on the prediction results, the values ​​of the current process parameters are optimized within a pre-defined specification range, and the optimized process parameters are used for the epitaxial growth process of the wafer in the next preset window. Using the above scheme, this application extracts feature parameters characterizing the physical state of growth from the original measurement data. Based on these feature parameters, this application does not rely on prior models and parameters, automatically analyzes the high-dimensional relationships between multiple parameters, finds the globally optimal process window, and generates prediction results consistent with the actual growth mechanism, effectively improving the accuracy of the process and product yield. Attached Figure Description

[0024] Figure 1This is a structural block diagram of an in-situ online detection device in one embodiment;

[0025] Figure 2 This is a flowchart illustrating a process parameter optimization method in one embodiment;

[0026] Figure 3 This is a schematic diagram of the prediction model training process in one embodiment;

[0027] Figure 4 This is a flowchart illustrating the steps for extracting feature parameters in one embodiment;

[0028] Figure 5 This is a flowchart illustrating the operation of a process parameter optimization method in one embodiment;

[0029] Figure 6 This is a block diagram of a process parameter optimization system in one embodiment;

[0030] Figure 7 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0031] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, unless otherwise specified, the following embodiments and features in the embodiments can be combined with each other.

[0032] Some exemplary embodiments of this application have been described for illustrative purposes. It should be understood that this application may be implemented in other ways not specifically shown in the accompanying drawings.

[0033] Please see Figure 1 Optionally, in one embodiment, this application provides an in-situ online detection device, including: a data acquisition module and a process parameter optimization system.

[0034] The data acquisition module, located above the reaction chamber, is used to observe the wafer surface of the epitaxial growth process inside and outside the process chamber through a viewing window within a preset window, and obtain measurement data.

[0035] Specifically, the machine is equipped with a reaction chamber, and a tray is placed inside the reaction chamber. The wafers to be processed are placed on the tray.

[0036] The preset window is determined based on the processing steps, which are predetermined and stored in the Manufacturing Execution System (MES). For example, if a processing step includes 10 steps, 10 preset windows are set, each corresponding to one processing step, and the duration of the preset window is the same as the duration of the corresponding processing step. Using this approach, the data acquisition module can repeatedly inspect the wafer surface within the preset window to obtain a series of measurement data values. Specifically, in this embodiment, the measurement data includes at least one of the following: reflectance spectral signal, wafer temperature data, and wafer curvature data.

[0037] It should be understood that the MES system pre-stores at least one process recipe, which defines the complete sequence of operations required to complete a processing step, including multiple processing steps arranged in sequence and a set of process parameters corresponding to each step. When a production task is triggered, the MES system calls the corresponding process recipe according to the scheduling strategy and sends the step instructions and process parameter set contained in the process recipe as control instructions to the corresponding machine. The machine receives and parses the above control instructions and automatically configures its execution mechanism based on the control instructions to complete the wafer processing process according to the preset steps and process parameters.

[0038] Furthermore, the aforementioned data acquisition module includes: a spectrometer, a pyrometer, and a curvature sensor.

[0039] A spectrometer is used to detect the reflection spectral signal on the wafer surface. The oscillation frequency of this reflection spectral signal is directly related to the growth rate of epitaxial growth. Specifically, in this embodiment, the wavelength range is 280-1600 nm.

[0040] A pyrometer is used to detect the temperature data generated by infrared radiation on the surface of a wafer. This temperature data is a key parameter that affects the rate of chemical reactions and the composition of components.

[0041] Curvature sensors are used to detect reflected images on the wafer surface to generate curvature data, which is used to monitor quality issues caused by mismatch stress.

[0042] The process parameter optimization system is connected to the output of the data acquisition module and receives the measurement data acquired by the data acquisition module. Simultaneously, the process parameter optimization system also obtains the process parameters corresponding to the current processing technology from the machine tool side or the EMS system. Specifically, in this embodiment, the process parameters include at least one of the following: flow rate of each source bottle (e.g., TMGa, TMIn, NH3), reaction chamber pressure, wafer carrier disk rotation speed, and temperature setpoint.

[0043] Furthermore, the process parameter optimization system optimizes the values ​​of process parameters based on measurement data and current process parameters, and applies the optimized process parameters to the epitaxial growth process of the wafer in the next preset window. Specifically, based on the aforementioned measurement data, the process parameter optimization system extracts characteristic parameters representing the physical state of wafer surface growth; inputs the characteristic parameters and process parameters into a pre-trained prediction model to obtain prediction results, which characterize the growth state trend and photoelectric property change trend; based on the prediction results, the system optimizes the values ​​of the current process parameters within a pre-defined specification range, and the optimized process parameters are applied to the epitaxial growth process of the wafer in the next preset window. Using this approach, the process parameter optimization system predicts the growth state trend and photoelectric property change trend based on measurement data within the current preset window, and optimizes the process parameters according to the prediction results. In the next preset window, the optimized process parameters are applied to the next step of the epitaxial growth process, thereby achieving a progressive positive optimization effect within a process batch, improving process accuracy and increasing product yield.

[0044] Please see Figure 2 Optionally, in one embodiment, this application provides a method for optimizing process parameters, specifically including the following steps:

[0045] Step 202: Obtain the process parameters of the machine and the measurement data of the process chamber within the preset window; wherein, the measurement data is obtained by observing the wafer surface of the epitaxial growth process inside and outside the process chamber through the observation window.

[0046] Step 204: Based on the measurement data, extract characteristic parameters that characterize the physical state of wafer surface growth.

[0047] Specifically, the physical state of wafer surface growth refers to the comprehensive characterization of the instantaneous and cumulative behavior of the wafer surface during epitaxy, determined by multiple influencing factors. It represents the coupling vector between multi-source measurements and process parameters on the same time axis. These influencing factors include material deposition rate, cumulative layer thickness, composition, refractive index, stress, curvature, and optical interference response.

[0048] By quantifying the information related to the aforementioned influencing factors in the measurement data into physically interpretable feature vectors, the prediction model can utilize the joint constraints between short-term changes and long-term accumulations to automatically learn the high-dimensional coupling relationships between various feature parameters in both time and feature dimensions, thereby reducing the rigid dependence on prior constants. Furthermore, the prediction model can achieve multi-objective consistency in the structural dimension, outputting trend predictions consistent with the actual mechanism, thereby accurately identifying the global process window and significantly improving the stability and achievement rate of multi-objective process parameter optimization. Specifically, in this embodiment, the feature parameters include at least one of instantaneous growth rate, stop zone feature data, mixed feature parameters, process parameter statistical features, and the final value of wafer curvature.

[0049] Step 206: Input the feature parameters and process parameters into the pre-trained prediction model to obtain the prediction results; the prediction results are used to characterize the growth state trend and the photoelectric property change trend.

[0050] Specifically, the prediction model is trained based on historical process data, which includes historical measurement data and historical process parameters.

[0051] Please see Figure 3 For example, the steps of pre-training a prediction model include: acquiring historical process data; cleaning and labeling the data, then extracting features to obtain a training set; constructing a model architecture and training the model based on this training set to make the model's output value approach the target value; evaluating the model's output value, and if the evaluation meets the requirements, deploying the trained model. In practical applications, real-time measured data and current process parameters are input into the trained prediction model to obtain prediction results. Through processing, feature extraction, and model training of a large amount of historical data, a high-performance model suitable for real-time prediction and control is finally obtained.

[0052] Step 208: Based on the prediction results, optimize the values ​​of the current process parameters within the pre-defined specification range. The optimized process parameters are used for the epitaxial growth process of the wafer in the next preset window.

[0053] Specifically, the pre-defined specification range refers to the predetermined process thresholds, which typically include process output specification boundaries and machine safety boundaries. The process output specification boundaries are used to constrain the quality of the process, such as the allowable deviation of the target emission wavelength, thickness tolerance, upper limit of uniformity, upper limit of defect density, and inter-wafer / intra-wafer consistency requirements. The machine safety boundaries are used to constrain the parameter adjustments of the machine, such as the upper and lower limits of temperature and the upper and lower limits of the rise and fall rate, the upper and lower limits of the flow rate of each source and the V / III ratio, the controllable cavity pressure zone, the upper limit of the carrier gas, and the limits of the single-step amplitude and total change of parameters.

[0054] Based on the pre-defined specification range, constraints are constructed. Within these constraints, the values ​​of process parameters are gradually adjusted to reduce the deviation of the predicted results from the target specifications and bring them closer to meeting the target specifications, thereby completing the optimization of process parameters.

[0055] In the next preset window, the optimized process parameters will be applied to the wafer epitaxial growth process.

[0056] The aforementioned process parameter optimization method acquires the process parameters of the equipment and measurement data within a preset window; based on the measurement data, it extracts feature parameters characterizing the physical state of wafer surface growth; the feature parameters and process parameters are input into a pre-trained prediction model to obtain prediction results, which are used to characterize the growth state trend and photoelectric property change trend; based on the prediction results, the values ​​of the current process parameters are optimized within a pre-defined specification range, and the optimized process parameters are used for the epitaxial growth process of the wafer in the next preset window. Using this scheme, this application extracts feature parameters characterizing the physical state of growth from the original measurement data. Based on these feature parameters, this application does not rely on prior models and parameters, automatically analyzes the high-dimensional relationships between multiple parameters, finds the globally optimal process window, and generates prediction results consistent with the actual growth mechanism, effectively improving the accuracy of the process and product yield.

[0057] Please see Figure 4 Optionally, when the measurement data includes a reflectance spectral signal and the characteristic parameters include the instantaneous growth rate, characteristic parameters characterizing the physical state of wafer surface growth are extracted based on the measurement data, including:

[0058] Step 402: Perform narrowband weighted integration on the reflection spectral signal at a preset monitoring wavelength to obtain the reflection time series.

[0059] Step 404: The reflection time series is processed using a real-time moving window algorithm to obtain the current oscillation period.

[0060] Step 406: For each oscillation cycle, calculate the product of the thickness increment within that oscillation cycle and the number of cycles within the sampling time to obtain the instantaneous growth rate; wherein, the thickness increment is related to the refractive index of the material and the monitoring wavelength, and the refractive index and monitoring wavelength are obtained based on the reflection spectrum signal.

[0061] Specifically, the preset monitoring wavelength is one or more monitoring wavelengths pre-selected according to the test requirements, used to generate a time-varying reflection intensity trajectory. Generally, a band sensitive to thickness changes and with a high signal-to-noise ratio should be selected as the monitoring wavelength. Typically, a complete spectral scan of the spectrometer constitutes one frame, which includes multiple wavelength samples, and the monitoring wavelength is selected from these multiple wavelengths.

[0062] For example, for each monitoring wavelength in the k-th frame, a narrow bandwidth wavelength window is selected as the monitoring channel, centered on that wavelength. The intensity of each wavelength point within this window is summed according to a preset weight to obtain the channel value corresponding to that monitoring wavelength. This channel value represents the intensity of the monitoring channel. The channel values ​​are then concatenated in chronological order to form the reflection time series of the k-th frame. The above steps are repeated to obtain the reflection time series of each frame.

[0063] Furthermore, the monitored light waves illuminate the growing thin film, and the reflected light from the surface and interface coherently superimposes, forming interference fringes on the detector. The growth rate and thickness can be estimated based on the period of the interference fringes. Since the interference fringes are essentially oscillations over time, the reflection spectrum signal is converted into a reflection time series characterizing the channel values ​​as a function of time, allowing for peak detection and thus the oscillation period. Further, at each sampling arrival, the latest channel value is written into a sliding window of a preset duration. The sequence within the sliding window is denoised and baseline corrected. Subsequently, dynamic peak detection or the Hilbert instantaneous phase method is used to calculate the interval between adjacent fringes and estimate the current oscillation period accordingly. The thickness increment corresponding to one oscillation period is related to the material's refractive index n and the monitoring wavelength λ, and its expression is λ / 2n. Here, the monitoring wavelength is obtained by normalizing and scaling the reflection spectrum signal, and the refractive index is obtained by looking up a table based on the correspondence between the monitoring wavelength and temperature.

[0064] The above scheme uses narrowband weighted integration to compress each frame of broadband into the reflection intensity of the monitoring channel, and combines it with a sliding window for time-domain operations, avoiding frame-by-frame full-spectrum fitting and significantly reducing the computational load. At the same time, the sliding window is not sensitive to short-term disturbances, resulting in a more stable instantaneous growth rate. The rate is directly converted using the time beat of the interference fringes, which reduces the dependence on accurate optical constants and complex interface modeling compared to the full-parameter inversion of TMM in traditional techniques. It is suitable for application scenarios such as new material systems with unknown optical constants or those that vary with the process, gradient layers, or complex interfaces.

[0065] Optionally, if the feature parameters also include stop band feature data, extracting feature parameters characterizing the physical state of wafer surface growth based on measurement data further includes:

[0066] For periodic structures grown epitaxially, based on the reflection spectrum signal, for each frame of reflection spectrum obtained after calibration and normalization, Lorentz or Gaussian line shape fitting is performed within a preset stop band wavelength range, and stop band feature data is dynamically tracked and output; wherein, the stop band feature data includes the center wavelength and / or half width at half maximum (WHM).

[0067] Specifically, periodic structures in epitaxial growth refer to the alternating deposition of two or more thin layers with different refractive indices during the epitaxial growth process. These mainly include distributed Bragg reflectors (DBRs), multi-periodic superlattices, and periodic stress / refractive index modulation layers. A stop band exists in the reflection spectrum of these periodic structures, which can be characterized by the center wavelength and / or full width at half maximum (FWHM).

[0068] The reflection spectrum represents the curve corresponding to reflectance and wavelength. First, wavelength calibration is performed on the original reflection spectrum signal, followed by reference normalization to obtain the reflection spectrum on the physical wavelength axis.

[0069] The preset stop band wavelength range refers to the expected wavelength range in which the stop band will appear, given the known materials and design. For each frame of the reflection spectrum, the waveform of the stop band is described within this wavelength range using a Lorentz function or a Gaussian function, for example, by calculating the stop band characteristic data using the minimum error method. The drift of the center wavelength directly reflects the total thickness error of the grown stack and is a key basis for feedforward compensation.

[0070] By adopting the above scheme, limiting the fitting interval and using linear fitting to estimate frame by frame, stop band feature data can be obtained stably and with low time delay without relying on accurate prior optical constants, which facilitates thickness error assessment and feedforward compensation.

[0071] Optionally, when the measurement data also includes wafer temperature data and wafer curvature data, and the characteristic parameters also include mixed characteristic parameters, based on the measurement data, the extraction of characteristic parameters characterizing the physical state of wafer surface growth further includes:

[0072] Using the reflection spectrum signal as observation data, combined with the constraints of wafer temperature data on the optical constant temperature coefficient and wafer curvature data on the stress state, the transfer matrix model is used for inversion fitting to obtain mixed characteristic parameters. The mixed characteristic parameters include equivalent optical constant and equivalent thickness. Among them, the wafer temperature data comes from the infrared radiation on the wafer surface, and the wafer curvature data comes from the reflection image of the laser signal on the wafer surface.

[0073] Specifically, using reflectance spectral signals as observation data means using the reflectance spectrum generated based on the reflectance spectral signals as the reference curve for fitting the transfer matrix model (TMM). Known process parameters or layer structure information are used as TMM inputs, while wafer temperature and wafer curvature data are used to constrain the model's adjustable parameters. The equivalent ray constant and equivalent thickness when the simulated spectrum output by the TMM best fits the reflectance spectrum are used as hybrid feature parameters. These hybrid feature parameters serve as supplementary features, providing physically meaningful inputs to the machine learning model and compensating for the shortcomings of purely data-driven methods in terms of physical interpretability.

[0074] Optionally, when the prediction model includes a time-series prediction model and a photoelectric property prediction model, and the feature parameters include instantaneous growth rate, statistical characteristics of process parameters, final value of wafer curvature, mixed feature parameters, and stop band feature data, and the measurement data includes wafer temperature data, the feature parameters and process parameters are input into the pre-trained prediction model to obtain the prediction results, including:

[0075] Instantaneous growth rate, wafer temperature data, and process parameters are input into the encoder of the timing prediction model to obtain a context vector representing the historical sequence.

[0076] The context vector is input into the decoder of the time series prediction model to obtain the growth rate prediction value;

[0077] The predicted growth rate is integrated to obtain the predicted thickness.

[0078] The predicted thickness, statistical characteristics of process parameters, final wafer curvature, mixed characteristic parameters, and stop band characteristic data are input into the optoelectronic characteristic prediction model to obtain the predicted values ​​of optoelectronic characteristic parameters; among which, optoelectronic characteristic parameters include the emission wavelength of the light-emitting device and / or the sheet resistance of the power semiconductor device.

[0079] Specifically, the aforementioned time-series prediction model is a Seq2Seq LSTM model employing an encoder-decoder architecture. The encoder receives the instantaneous growth rate sequence, process parameter sequence, and temperature sequence from a past period, converting these sequences into a context vector representing the historical sequence. The decoder not only receives the encoder's context vector, but its input at each step also includes the predicted instantaneous growth rate value from the previous step, forming a recursive prediction loop. This enables accurate prediction of the growth rate and cumulative thickness for a predetermined number of future steps, making it particularly suitable for high-precision control of ultrathin layers such as VCSEL quantum wells.

[0080] Furthermore, the optoelectronic property prediction model is constructed based on a Gradient Boosting Decision Tree (GBDT). The layer thickness sequences predicted by the bottom-level timing prediction model, the statistical characteristics of process parameters, the final wafer curvature value, mixed feature parameters, and stopband feature data are fed into the top-level Gradient Boosting Decision Tree model to predict the final device's optoelectronic property parameters, ultimately realizing the causal logic from microscopic growth to macroscopic performance. The statistical characteristics of process parameters are obtained by statistically analyzing data such as average temperature and flow rate fluctuations throughout the entire growth process.

[0081] By adopting the above scheme and simultaneously introducing in-situ measurement data and machine process parameters, the temporal evolution information is preserved while incorporating full-process statistics and final value constraints, resulting in predictions that better align with the actual growth mechanism. Furthermore, the hybrid feature parameters and stop zone feature data can provide physical anchors, reducing reliance on prior precise optical constants in new materials, gradient layers, and complex interface scenarios, thus improving resistance to operating condition drift and noise. Moreover, a multi-scale model combining a time-series prediction model and a photoelectric property prediction model is employed. Leveraging the powerful nonlinear fitting capabilities of machine learning models, the process rules can be directly learned from the data without the need for complex prior physical models, achieving high-precision predictions of growth rate, thickness, and photoelectric properties, thereby reducing model dependence.

[0082] Optionally, based on the prediction results, the values ​​of the current process parameters are optimized within the pre-defined specification range, including:

[0083] The prediction results are compared with the preset target specifications to obtain deviation information. The prediction results include the predicted thickness and the predicted photoelectric characteristic parameters. A state vector is constructed based on the deviation information and process parameters. The state vector is processed by a reinforcement learning algorithm to obtain the process parameter adjustment amount. The process parameter adjustment amount is constrained within the preset safe process parameter range to obtain the optimized process parameters.

[0084] Specifically, reinforcement learning algorithms (Proximal Policy Optimization, PPO) are used to train an agent to learn the optimal policy through interaction with the environment. In this embodiment, its state consists of the deviation information between the current prediction result and the target rule, and the current process parameters; its action is the adjustment amount of the process parameters, such as TMI flow rate, temperature setpoint, etc.; its reward function is the pre-designed process constraints and objectives. For example, the reward function = -(wavelength prediction deviation)² - α * (uniformity deviation)² - β * (temperature change rate)² - γ * (flow rate exceeding safety range penalty). Under the premise of satisfying various process safety constraints, the parameters are adjusted collaboratively to achieve multi-objective optimization. Furthermore, this application also uploads the optimized process parameters to the MES system so that the MES system updates the corresponding process formula and reissues it to the machine. The machine then executes the epitaxial growth process according to the updated process parameters in the next preset window.

[0085] It should be noted that the above optimization action is only triggered when the deviation information exceeds the preset threshold. If the deviation information is within the allowable range, there is no need to calculate the process parameter adjustment amount, and the current measurement data is reacquired.

[0086] By adopting the above scheme, since the prediction results are the growth results of multiple future steps, this application can adjust process parameters before or after deviations occur, overcoming the lag of traditional feedback control, and is particularly suitable for the growth of ultrathin layers and complex structures. Furthermore, the above-mentioned time-series prediction model focuses on the time-series evolution of process dynamics, while the photoelectric property prediction model focuses on the mapping from structure to performance. Both are adaptively updated by continuously learning data from different batches to adapt to changes in the state of the reaction chamber and provide stable, low-deviation target evaluation. Based on the above target evaluation, the reinforcement learning algorithm searches for multi-parameter strategies within safety constraints, handles the complex coupling relationship between multiple input parameters and multiple output indicators, achieves global optimization under constraints, automatically finds the optimal process window, and significantly improves product yield and batch repeatability.

[0087] Optionally, the above-mentioned process parameter optimization method further includes:

[0088] A display chart is generated based on the prediction results; wherein the display chart includes at least one of the following: real-time growth rate curve, stop band center wavelength evolution trend chart, and predicted photoelectric property change trend distribution chart.

[0089] Specifically, this application also provides users with curves and trend charts of key physical quantities through a human-machine interface, which makes it easier for users to quickly identify anomalies, trigger corrections or interlocks in advance, and shorten processing delays.

[0090] Optionally, the above-mentioned process parameter optimization method further includes:

[0091] The system stores all data involved in each epitaxial growth process, including measurement data, process parameters and their adjustments, and characteristic parameters. Furthermore, the data also includes process formulations and cavity cleaning data, used to analyze the performance drift of the model under different cavity conditions. This data is used to predict model optimization.

[0092] Please see Figure 5 The following is combined Figure 5 The workflow of the process parameter optimization method in this application is described as follows:

[0093] After the epitaxial growth process begins, the process parameters of the equipment and the measurement data within the preset window are acquired. The process parameters include at least one of the following: flow rate of each source bottle, pressure of the reaction chamber, rotation speed of the wafer carrier disk, and temperature setpoint. The measurement data includes at least one of the following: reflection spectrum signal, wafer temperature data, and wafer curvature data.

[0094] The characteristic parameters of the measurement data are extracted, and the extracted characteristic parameters and process parameters are input into the prediction model to obtain the prediction results. The characteristic parameters include at least one of the following: instantaneous growth rate, stop zone characteristic data, mixed characteristic parameters, process parameter statistical characteristics, and final wafer curvature value. The prediction results include the predicted thickness value and the predicted value of the photoelectric characteristic parameters, and the photoelectric characteristic parameters include the emission wavelength of the light-emitting device and / or the sheet resistance of the power semiconductor device.

[0095] The deviation between the predicted result and the target value is compared. If the deviation exceeds a preset threshold, a reinforcement learning algorithm is used to calculate the adjustment amount of the process parameters and generate an optimization command. In response to this optimization command, the process parameters are adjusted, and the optimized process parameters are uploaded to the MES system. The MES system then sends the new process parameters to the machine tool, which executes the epitaxial growth process based on the new process parameters in the next preset window, achieving adaptive closed-loop control. If the deviation does not exceed the preset threshold, measurement data is reacquired, and the epitaxial growth process continues to be monitored.

[0096] Repeat the above steps until a complete process batch is completed.

[0097] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0098] Based on the same inventive concept, this application also provides a process parameter optimization system. This device is applicable to the above-described process parameter optimization method. The solution provided by this device is similar to the solution described in the above-described method. Therefore, the specific limitations of one or more device embodiments provided below can be found in the limitations of the method above, and will not be repeated here.

[0099] Please see Figure 6 In one embodiment, this application provides a process parameter optimization system, including: an acquisition module, a feature extraction module, a prediction module, and an optimization module.

[0100] The acquisition module is used to acquire the process parameters of the machine and the measurement data of the process chamber within a preset window; wherein, the measurement data is obtained by observing the wafer surface of the epitaxial growth process inside and outside the process chamber through the observation window.

[0101] The feature extraction module is used to extract feature parameters that characterize the physical state of wafer surface growth based on measurement data.

[0102] The prediction module is used to input feature parameters and process parameters into a pre-trained prediction model to obtain prediction results; the prediction results are used to characterize the growth state trend and the photoelectric property change trend.

[0103] The optimization module is used to optimize the values ​​of the current process parameters within the pre-defined specification range based on the prediction results. The optimized process parameters are used for the epitaxial growth process of the wafer in the next preset window.

[0104] Optionally, the above-mentioned process parameter optimization system also includes a training module.

[0105] The training module is used to train the prediction model, which includes a time-series prediction model and a photoelectric property prediction model.

[0106] Optionally, when the measurement data includes a reflectance spectral signal and the feature parameters include the instantaneous growth rate, the feature extraction module extracts feature parameters characterizing the physical state of wafer surface growth based on the measurement data, including: performing narrowband weighted integration on the reflectance spectral signal at a preset monitoring wavelength to obtain a reflectance time series; processing the reflectance time series using a real-time moving window algorithm to obtain the current oscillation period; and for each oscillation period, calculating the product of the thickness increment within that oscillation period and the number of periods within the sampling time to obtain the instantaneous growth rate; wherein the thickness increment is related to the refractive index of the material and the monitoring wavelength, and the refractive index and monitoring wavelength are obtained based on the reflectance spectral signal.

[0107] Optionally, when the feature parameters also include stop band feature data, the feature extraction module extracts feature parameters characterizing the physical state of wafer surface growth based on the measurement data. This also includes: for the periodic structure of epitaxial growth, based on the reflection spectrum signal, performing Lorentz or Gaussian line fitting on each frame of reflection spectrum obtained after calibration and normalization within a preset stop band wavelength range, dynamically tracking and outputting stop band feature data; wherein, the stop band feature data includes the center wavelength and / or half-width at half-maximum.

[0108] Optionally, when the measurement data also includes wafer temperature data and wafer curvature data, and the feature parameters also include hybrid feature parameters, the feature extraction module extracts feature parameters characterizing the physical state of wafer surface growth based on the measurement data. This also includes: using the reflection spectrum signal as observation data, combining the constraints of the optical constant temperature coefficient by the wafer temperature data, and the constraints of the stress state by the wafer curvature data, performing inversion fitting using a transfer matrix model to obtain hybrid feature parameters; the hybrid feature parameters include equivalent optical constants and equivalent thickness; wherein, the wafer temperature data comes from the infrared radiation of the wafer surface, and the wafer curvature data comes from the reflection image of the laser signal on the wafer surface.

[0109] Optionally, when the prediction model includes a time-series prediction model and a photoelectric characteristic prediction model, and the feature parameters include instantaneous growth rate, process parameter statistical characteristics, final wafer curvature value, mixed feature parameters, and stop band feature data, and the measurement data includes wafer temperature data, the prediction model processes the feature parameters and process parameters to obtain the prediction result, including: inputting the instantaneous growth rate, wafer temperature data, and process parameters into the encoder of the time-series prediction model to obtain a context vector representing the historical sequence; inputting the context vector into the decoder of the time-series prediction model to obtain the predicted growth rate value; integrating the predicted growth rate value to obtain the predicted thickness value; and inputting the predicted thickness value, process parameter statistical characteristics, final wafer curvature value, mixed feature parameters, and stop band feature data into the photoelectric characteristic prediction model to obtain the predicted values ​​of the photoelectric characteristic parameters; wherein, the photoelectric characteristic parameters include the emission wavelength of the light-emitting device and / or the sheet resistance of the power semiconductor device.

[0110] Optionally, the optimization module optimizes the values ​​of the current process parameters based on the prediction results within the pre-defined specification range, including: comparing the prediction results with the preset target specifications to obtain deviation information; wherein, the prediction results include the predicted values ​​of thickness and photoelectric characteristic parameters; constructing a state vector based on the deviation information and process parameters; processing the state vector using a reinforcement learning algorithm to obtain the process parameter adjustment amount; and constraining the process parameter adjustment amount within the preset safe range of process parameters to obtain the optimized process parameters.

[0111] Optionally, the above-mentioned process parameter optimization system also includes a display module.

[0112] The display module is used to generate display charts based on the prediction results; wherein the display charts include at least one of the following: real-time growth rate curve, stop band center wavelength evolution trend chart, and predicted photoelectric property change trend distribution chart.

[0113] Optionally, the above-mentioned process parameter optimization system also includes a storage module.

[0114] The storage module is used to store all data involved in each epitaxial growth process, including measurement data, process parameters and their adjustments, characteristic parameters, process formulas, and cavity cleaning data.

[0115] The aforementioned process parameter optimization system acquires the process parameters of the equipment and measurement data within a preset window; based on the measurement data, it extracts feature parameters characterizing the physical state of wafer surface growth; the feature parameters and process parameters are input into a pre-trained prediction model to obtain prediction results, which are used to characterize the growth state trend and photoelectric property change trend; based on the prediction results, the values ​​of the current process parameters are optimized within a pre-defined specification range, and the optimized process parameters are used for the epitaxial growth process of the wafer in the next preset window. Using this scheme, this application extracts feature parameters characterizing the physical state of growth from the original measurement data. Based on these feature parameters, this application does not rely on prior models and parameters, automatically analyzes the high-dimensional relationships between multiple parameters, finds the globally optimal process window, and generates prediction results consistent with the actual growth mechanism, effectively improving the accuracy of the process and product yield.

[0116] Each module in the aforementioned process parameter optimization system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0117] In one feasible embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 7As shown, the computer device includes a processor, memory, input / output interfaces, a communication interface, a display unit, and an input device. The processor, memory, and input / output interfaces are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interfaces. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interfaces are used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements the aforementioned process parameter optimization method. The display unit is used to form a visually visible image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0118] Those skilled in the art will understand that Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0119] In one feasible embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method steps in the above-described process parameter optimization method.

[0120] In one feasible embodiment, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the method steps in the above-described process parameter optimization method.

[0121] In one feasible embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the method steps in the above-described process parameter optimization method.

[0122] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0123] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0124] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for optimizing process parameters, characterized in that, The method includes: The process parameters of the machine and the measurement data of the process chamber within a preset window are acquired; wherein, the measurement data is obtained by observing the wafer surface of the epitaxial growth process inside and outside the process chamber through the observation window. Based on the measurement data, characteristic parameters characterizing the physical state of wafer surface growth are extracted; The feature parameters and process parameters are input into a pre-trained prediction model to obtain prediction results; the prediction results are used to characterize the growth state trend and the photoelectric property change trend. Based on the prediction results, the values ​​of the current process parameters are optimized within the pre-defined specification range. The optimized process parameters are used for the epitaxial growth process of the wafer in the next preset window. The prediction model includes a time-series prediction model and a photoelectric property prediction model; the feature parameters include instantaneous growth rate, statistical characteristics of process parameters, final value of wafer curvature, mixed feature parameters, and stop band feature data; the mixed feature parameters are the equivalent ray constant and equivalent thickness when the simulated spectrum output by the transfer matrix model best matches the reflection spectrum; the measurement data includes wafer temperature data; The step of inputting the feature parameters and the process parameters into a pre-trained prediction model to obtain prediction results includes: The instantaneous growth rate, the wafer temperature data, and the process parameters are input into the encoder of the timing prediction model to obtain a context vector representing the historical sequence; the context vector is input into the decoder of the timing prediction model to obtain a predicted growth rate value; the predicted growth rate value is integrated to obtain a predicted thickness value; the predicted thickness value, the statistical characteristics of the process parameters, the final value of the wafer curvature, the mixed feature parameters, and the stop band feature data are input into the photoelectric characteristic prediction model to obtain predicted values ​​of photoelectric characteristic parameters; wherein, the photoelectric characteristic parameters include the emission wavelength of the light-emitting device and / or the sheet resistance of the power semiconductor device.

2. The method according to claim 1, characterized in that, The measurement data includes reflectance spectral signals; the characteristic parameters include instantaneous growth rate; The extraction of characteristic parameters characterizing the physical state of wafer surface growth based on the measurement data includes: The reflection spectral signal is subjected to narrowband weighted integration at a preset monitoring wavelength to obtain a reflection time series. The reflection time series is processed using a real-time moving window algorithm to obtain the current oscillation period; For each oscillation cycle, the instantaneous growth rate is obtained by multiplying the thickness increment within that oscillation cycle by the number of cycles within the sampling time; wherein the thickness increment is related to the refractive index of the material and the monitoring wavelength, and the refractive index and the monitoring wavelength are obtained based on the reflection spectrum signal.

3. The method according to claim 2, characterized in that, The feature parameters also include stop band feature data; The step of extracting characteristic parameters representing the physical state of wafer surface growth based on the measurement data further includes: For the periodic structure of epitaxial growth, based on the reflection spectrum signal, for each frame of reflection spectrum obtained after calibration and normalization, Lorentz or Gaussian line shape fitting is performed within a preset stop band wavelength range, and stop band feature data is dynamically tracked and output; wherein, the stop band feature data includes the center wavelength and / or half width at half maximum (WHM).

4. The method according to claim 2, characterized in that, The measurement data also includes wafer temperature data and wafer curvature data; the characteristic parameters also include mixed characteristic parameters; The step of extracting characteristic parameters representing the physical state of wafer surface growth based on the measurement data further includes: Using the reflected spectrum signal as observation data, and combining the constraints of the wafer temperature data on the optical constant temperature coefficient and the constraints of the wafer curvature data on the stress state, a transfer matrix model is used for inversion fitting to obtain hybrid characteristic parameters; the hybrid characteristic parameters include equivalent optical constant and equivalent thickness; wherein, the wafer temperature data comes from the infrared radiation on the wafer surface, and the wafer curvature data comes from the reflection image of the laser signal on the wafer surface.

5. The method according to claim 1, characterized in that, The optimization of the current process parameters based on the prediction results, within the pre-determined specification range, includes: By comparing the predicted results with the preset target specifications, deviation information is obtained; wherein, the predicted results include the predicted thickness value and the predicted values ​​of photoelectric characteristic parameters; A state vector is constructed based on the aforementioned deviation information and process parameters; The state vector is processed using a reinforcement learning algorithm to obtain the process parameter adjustment amount; The process parameters are adjusted within the preset safe range to obtain optimized process parameters.

6. The method according to claim 1, characterized in that, The method further includes: A display chart is generated based on the prediction results; wherein the display chart includes at least one of the following: real-time growth rate curve, stop band center wavelength evolution trend chart, and predicted photoelectric property change trend distribution chart.

7. A process parameter optimization system, characterized in that, The system includes: The acquisition module is used to acquire the process parameters of the machine and the measurement data of the process chamber within a preset window; wherein, the measurement data is obtained by observing the wafer surface of the epitaxial growth process inside and outside the process chamber through the observation window. The feature extraction module is used to extract feature parameters characterizing the physical state of wafer surface growth based on the measurement data; The prediction module is used to input the feature parameters and process parameters into a pre-trained prediction model to obtain prediction results. The prediction results characterize the growth state trend and photoelectric property change trend. The prediction model includes a time-series prediction model and a photoelectric property prediction model. The feature parameters include instantaneous growth rate, statistical characteristics of process parameters, final wafer curvature value, mixed feature parameters, and stopband feature data. The mixed feature parameters are the equivalent ray constant and equivalent thickness when the simulated spectrum output by the transfer matrix model best matches the reflection spectrum. The measurement data includes wafer temperature data. The process of inputting the feature parameters and process parameters into the pre-trained prediction model to obtain prediction results includes... The process includes: inputting the instantaneous growth rate, the wafer temperature data, and the process parameters into the encoder of the timing prediction model to obtain a context vector representing the historical sequence; inputting the context vector into the decoder of the timing prediction model to obtain a predicted growth rate value; integrating the predicted growth rate value to obtain a predicted thickness value; and inputting the predicted thickness value, the statistical characteristics of the process parameters, the final value of the wafer curvature, the hybrid feature parameters, and the stop band feature data into the photoelectric characteristic prediction model to obtain predicted values ​​of photoelectric characteristic parameters; wherein, the photoelectric characteristic parameters include the emission wavelength of the light-emitting device and / or the sheet resistance of the power semiconductor device. The optimization module is used to optimize the values ​​of the current process parameters within a pre-defined specification range based on the prediction results. The optimized process parameters are used for the epitaxial growth process of the wafer in the next preset window.

8. An in-situ online detection device, characterized in that, include: The data acquisition module is used to detect the wafer surface of the epitaxial growth process outside the process cavity through the observation window within a preset window and obtain measurement data. The process parameter optimization system as described in claim 7 is used to optimize the values ​​of process parameters based on the measurement data and the current process parameters, and to use the optimized process parameters in the epitaxial growth process of the wafer in the next preset window; wherein the current process parameters are sent to the in-situ online detection device by the manufacturing execution system or the machine.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1-6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-6.

Images