Method, system, control device for controlling epitaxial growth of semiconductors and related products

Abstract

The application provides a control method and system of semiconductor epitaxial growth, a control device and related products, relates to the technical field of semiconductor detection, and acquires reflection spectrum data in real time and obtains an optical feature set by analysis, then inputs the optical feature set into an optical transmission model for iterative optimization to obtain a predicted growth rate; compares the predicted growth rate with a target growth rate and generates deviation information, and then outputs a process adjustment control instruction, so that the main control system performs closed-loop correction on the process parameters in the next process cycle. The optical feature set can dynamically represent the influence of the interface transition layer on the reflection spectrum, thereby improving the inversion accuracy and robustness of the real epitaxial structure state; based on the predicted growth rate, the process parameters are corrected in a closed loop, the growth rate drift can be timely inhibited, and the process disturbance can be adaptively compensated, the need for manual calibration and intervention is reduced, and the production stability, repeatability and batch consistency are improved.

Description

Technical Field

[0001] This application relates to the field of semiconductor testing technology, and in particular to a control method, system, control device, and related products for semiconductor epitaxial growth. Background Technology

[0002] In the epitaxial growth of compound semiconductor devices such as vertical-cavity surface-emitting lasers (VCSELs), the epitaxial layer thickness, composition, and interface quality directly determine the optical characteristics of the resonant cavity and key performance parameters such as device threshold, efficiency, and consistency. Therefore, in-situ monitoring of the growth state during metal-organic chemical vapor deposition (MOCVD) epitaxial growth, and subsequent closed-loop control, is a common technique for improving product yield and batch-to-batch consistency. Currently, the industry widely employs in-situ reflectometry to measure the change in the reflectance spectrum of the growth surface over time, and deduce the film thickness and growth rate based on the periodic oscillation characteristics of the reflectance spectrum, thereby achieving online monitoring of the epitaxial process.

[0003] Traditional in-situ spectral reflectance monitoring schemes typically rely on fixed optical constants and ideal layered structure models for parameter inversion. For example, they assume that the refractive index and extinction coefficient (n, k) of each material layer are constant and that the interfaces between layers are ideal abrupt transition interfaces. Then, they calculate growth parameters by fitting the oscillation period of the reflectance spectrum or selecting the phase change of characteristic wavelengths. These growth parameters usually include thickness and average growth rate.

[0004] However, in actual MOCVD epitaxial growth, due to memory effect and high-temperature diffusion, there are gradient regions at the interface, and the optical constants of the material will change with composition and temperature. Fixed model parameters cannot characterize these dynamic physical effects, resulting in systematic errors in the inverted growth parameters, which in turn affects the accuracy of process judgment and control. Summary of the Invention

[0005] The purpose of this application is to provide a control method, system, control device and related products for semiconductor epitaxial growth, so as to overcome the defects of traditional technology where fixed model parameters cannot characterize the dynamic physical effects in the epitaxial growth process, resulting in systematic errors in the inverted growth parameters, which in turn affect the process judgment and control accuracy.

[0006] In a first aspect, this application provides a method for controlling semiconductor epitaxial growth, including:

[0007] Acquire reflectance spectral data; wherein, the reflectance spectral data is generated based on the reflected light signal produced during semiconductor epitaxial growth;

[0008] The reflection spectrum data is analyzed to obtain an optical feature set; wherein, the optical feature set includes at least one of the following: the stop band center wavelength, the stop band width, the Fabry-Perot cavity mode center wavelength, the cavity mode depth, the cavity mode half-width at half maximum, and the absolute reflectivity value at a preset wavelength point of the distributed Bragg mirror.

[0009] The optical feature set is input into a preset optical transmission model, and the variable physical parameter set of the optical transmission model is iteratively optimized until the preset conditions are met to obtain the target parameter set.

[0010] A predicted growth rate is generated based on the target parameter set; the predicted growth rate and the target growth rate are compared to obtain deviation information.

[0011] Based on the deviation information, a process adjustment control command is generated so that the main control system executes the epitaxial growth process according to the adjusted process parameters in the next process cycle.

[0012] In one embodiment,

[0013] The process of analyzing the reflectance spectral data to obtain an optical feature set includes:

[0014] The reflectance spectral data are subjected to spectral feature identification to determine the stop band of the distributed Bragg mirror, and the center wavelength and / or stop band width of the stop band are extracted.

[0015] And / or, perform cavity mode feature extraction on the reflectance spectral data to determine the Fabry-Perot cavity mode, and extract the center wavelength, cavity mode depth and / or cavity mode half width at half maximum of the Fabry-Perot cavity mode;

[0016] And / or, select at least one preset wavelength point in the reflectance spectral data and extract the absolute reflectance value at the preset wavelength point.

[0017] In one embodiment, the variable set of physical parameters includes the thickness of the current layer, the thickness of the interface transition layer, the composition, and the optical constants of the material, wherein the optical constants include the refractive index and the extinction coefficient;

[0018] The process of inputting the optical feature set into a preset optical transmission model and iteratively optimizing the variable physical parameter set of the optical transmission model until a preset condition is met, thereby obtaining the target parameter set, includes:

[0019] Construct the objective function;

[0020] Based on the current set of variable physical parameters, a forward calculation is performed on the preset optical transmission model to obtain the predicted reflection spectrum data;

[0021] The predicted reflectance spectral data is analyzed to obtain a predicted feature set. The difference between the predicted feature set and the optical feature set is calculated according to the objective function. The current variable physical parameter set is iteratively optimized according to the difference until a preset condition is met. The variable physical parameter set at this time is then used as the target parameter set.

[0022] In one embodiment, the objective function is the squared L2 of the difference between the predicted feature set and the optical feature set;

[0023] The step of calculating the difference between the predicted feature set and the optical feature set according to the objective function, and iteratively optimizing the current variable physical parameter set based on the difference until a preset condition is met, includes:

[0024] Using the objective function, the predicted feature set and the optical feature set are differencing to obtain the feature residual; and the sensitivity matrix of the feature residual relative to the current variable physical parameter set is calculated.

[0025] Based on the feature residuals and the sensitivity matrix, an iterative optimization algorithm is used to solve for the parameter increments, and all parameter values ​​in the variable physical parameter set are updated synchronously based on the parameter increments until the objective function converges to less than a preset threshold.

[0026] In one embodiment, generating the predicted growth rate based on the target parameter set includes:

[0027] The iterative optimization process is performed at multiple consecutive sampling times to obtain the target parameter set corresponding to each sampling time, and the target layer thickness of the current layer is obtained from each target parameter set.

[0028] For the current sampling time, obtain the target layer thickness corresponding to the previous sampling time adjacent to it, and determine the time interval between adjacent sampling times;

[0029] The difference in the target layer thickness is calculated and divided by the time interval to obtain the predicted growth rate at the current sampling time.

[0030] In one embodiment, generating process adjustment control instructions based on the deviation information includes:

[0031] Based on the deviation information, a preset control algorithm is used to calculate the control quantity, wherein the preset control algorithm includes a proportional-integral-derivative control algorithm and / or a model predictive control algorithm.

[0032] Based on the control quantity, a process adjustment control command is generated for adjusting the epitaxial growth process parameters, including the flow rate parameter of the metal-organic precursor and / or the growth temperature parameter.

[0033] Secondly, this application provides a control system for semiconductor epitaxial growth, disposed within a control device, wherein the control device is partially disposed within the reaction chamber of a semiconductor device; the system includes:

[0034] An acquisition module is used to acquire reflectance spectral data; wherein the reflectance spectral data is generated based on the reflected light signal produced during the semiconductor epitaxial growth process;

[0035] The analysis module is used to analyze the reflection spectrum data to obtain an optical feature set; wherein, the optical feature set is associated with the layer thickness, aluminum composition and interface change state of the interface transition layer located between adjacent material layers;

[0036] The prediction module is used to input the optical feature set into a preset optical transmission model, iteratively optimize the variable physical parameter set of the optical transmission model until the preset conditions are met, and obtain the target parameter set.

[0037] The processing module is used to generate a predicted growth rate based on the target parameter set; compare the predicted growth rate with the target growth rate to obtain deviation information; and generate a process adjustment control command based on the deviation information so that the main control system executes the epitaxial growth process according to the adjusted process parameters in the next process cycle.

[0038] Thirdly, this application also provides a control device, partially disposed within the reaction chamber of a semiconductor device; the control device includes:

[0039] Target light source, used to output target optical signals during semiconductor epitaxial growth;

[0040] The sensing module is partially disposed within the reaction chamber of the semiconductor device, and is used to guide the target optical signal to the wafer surface within the reaction chamber; and continuously acquire the reflected optical signal generated on the wafer surface based on the target optical signal at a preset frequency, and generate reflection spectral data based on the reflected optical signal;

[0041] The semiconductor epitaxial growth control system described in the second aspect is used to calculate the deviation information between the predicted growth rate and the target growth rate based on the reflection spectrum data, and to generate process adjustment control instructions based on the deviation information, so that the main control system executes the epitaxial growth process according to the adjusted process parameters in the next process cycle.

[0042] In one embodiment, the sensing module includes an optical fiber probe; the device also includes a cleaning module, which includes a cooling assembly and a purging assembly.

[0043] The cooling component is used to actively cool the fiber optic probe, and the active cooling process includes water cooling and / or air cooling.

[0044] The purging assembly includes a purging gas path for outputting purging gas to the lens end of the fiber optic probe to reduce the deposition of contaminants on the lens during the acquisition process.

[0045] In one embodiment, the number of fiber optic probes is multiple.

[0046] In one embodiment, the device further includes an interface module comprising a plurality of industry-standard communication interfaces, each of which is used to communicate with the main control system of an epitaxial growth device having different communication protocols.

[0047] 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.

[0048] 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.

[0049] The aforementioned methods, systems, control devices, and related products for controlling semiconductor epitaxial growth have at least the following advantages:

[0050] This application acquires reflectance spectral data in real time during the epitaxial growth process and analyzes it to obtain an optical feature set related to the interface transition layer. This optical feature set is then input into a preset optical transmission model, and the variable physical parameter set of the optical transmission model is iteratively optimized to obtain a predicted growth rate. The predicted growth rate is compared with the target growth rate to generate deviation information, and then a process adjustment control command is output, enabling the main control system to perform closed-loop correction of process parameters in the next process cycle. Using this approach, the optical feature set of this application can dynamically characterize the influence of the layer thickness, composition, and interface change state of the interface transition layer on the reflectance spectrum without relying on fixed optical constants and the assumption of an ideal abrupt interface change, thereby improving the inversion accuracy and robustness of the actual epitaxial structure state. Furthermore, closed-loop correction of process parameters based on the predicted growth rate obtained through iterative optimization can achieve timely suppression of growth rate drift and adaptive compensation for process disturbances, reducing the need for manual calibration and intervention, and improving the stability, repeatability, and batch-to-batch consistency of the epitaxial growth process. Attached Figure Description

[0051] Figure 1 This is a structural block diagram of the control device in one embodiment;

[0052] Figure 2This is a flowchart illustrating a method for controlling semiconductor epitaxial growth in one embodiment;

[0053] Figure 3 This is a flowchart illustrating the steps for obtaining an optical feature set in one embodiment;

[0054] Figure 4 This is a flowchart illustrating the steps for obtaining the target parameter set in one embodiment;

[0055] Figure 5 This is a block diagram of the control system for semiconductor epitaxial growth in one embodiment;

[0056] Figure 6 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0057] 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.

[0058] 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.

[0059] Please see Figure 1 In one exemplary embodiment, this application provides a control device partially disposed within the reaction chamber of a semiconductor device. The reaction chamber includes a substrate for placing a wafer, made of high-purity graphite, supported by a rotating shaft driven by a motor, allowing continuous rotation during semiconductor epitaxial growth. Under reaction conditions where a vapor precursor is introduced into the reaction chamber and suitable temperature and pressure are provided, semiconductor material is directionally deposited layer by layer on the wafer surface using epitaxial growth methods such as atomic layer epitaxy (ALE) or vapor phase epitaxy (VPE) to form an epitaxial thin film with controllable thickness.

[0060] Specifically, the control device includes a target light source, a sensing module, and a control system for semiconductor epitaxial growth.

[0061] A target light source is used to output a target optical signal during the semiconductor epitaxial growth process. For example, in this embodiment, the target light source is a broadband halogen tungsten lamp with an output wavelength range of 600-1100 nm, and is equipped with constant current drive and temperature control to output a stable target optical signal.

[0062] The sensing module is used to guide the target light signal to the wafer surface in the reaction cavity; and continuously acquire the reflected light signal generated by the wafer surface based on the target light signal according to a preset frequency, and generate reflection spectrum data based on the reflected light signal; specifically, the sensing module includes an optical probe and a detector.

[0063] An optical probe, partially disposed within the reaction chamber of a semiconductor device, is used to receive and transmit target optical signals. Exemplarily, the optical probe is made of high-purity quartz or sapphire.

[0064] A detector, positioned outside the reaction chamber, is used to continuously collect reflected light signals at a preset frequency and then convert the reflected light signals into reflected light spectral data. For example, in this embodiment, the preset frequency is not less than 1 Hz.

[0065] It should be understood that the aforementioned sensing module also includes a beam splitter for separating the received reflected light signal into multiple different specific wavelength bands. For example, the beam splitter may be a beam splitter grating or a filter wheel. In this case, the detector will synchronously or at high speed switch to measure the reflected light signal in each wavelength band.

[0066] A control system for semiconductor epitaxial growth is used to acquire reflectance spectral data, which is generated based on reflected light signals produced during the semiconductor epitaxial growth process. The reflectance spectral data is analyzed to obtain an optical feature set, which is associated with the layer thickness, composition, and interface change state of the interface transition layer between adjacent material layers. The optical feature set is input into a preset optical transmission model, and the variable physical parameter set of the optical transmission model is iteratively optimized until preset conditions are met to obtain a predicted growth rate. The predicted growth rate is compared with the target growth rate to obtain deviation information. Based on the deviation information, a process adjustment control command is generated so that the main control system executes the epitaxial growth process according to the adjusted process parameters in the next process cycle.

[0067] Optionally, the control device further includes a cleaning module, which includes a cooling component and a purging component.

[0068] Cooling components are used to actively cool the fiber optic probe, wherein the active cooling process includes water cooling and / or air cooling.

[0069] The purging assembly includes a purging gas path for outputting purging gas to the lens end of the fiber optic probe to reduce contaminant deposition on the lens during acquisition and ensure long-term stable operation in the harsh reaction chamber environment.

[0070] Optionally, multiple fiber optic probes can be used to achieve synchronous monitoring of different locations on the wafer and to monitor uniformity. With multiple fiber optic probes, a switching switch can be configured to sequentially send the acquired signals from each probe into the detector according to channel polling.

[0071] Optionally, the control device further includes an interface module, which comprises multiple industry-standard communication interfaces. Each industry-standard communication interface is used to communicate with the main control system of epitaxial growth equipment with different communication protocols. By adopting the above solution, the control device of this application can be compatible with various models of epitaxial growth equipment, avoiding the customization and modification costs caused by protocol inconsistencies. For example, the interface module of this application supports multiple communication protocols to ensure low-latency, high-reliability transmission of control commands; wherein the communication protocols include real-time Ethernet protocols such as EtherCAT and TCP / IP.

[0072] Optionally, the interface module also includes safety interlock logic to prevent damage to expensive main equipment due to misoperation. Specifically, before sending any process control command, the interface module performs steps such as status verification, command whitelist detection, step consistency detection, and handshake confirmation to determine the interlock.

[0073] The aforementioned control device outputs a broadband target light signal from a target light source. The sensing module guides this signal to the wafer surface within the reaction chamber and continuously acquires reflected light signals at a preset frequency, generating reflectance spectral data. This enables in-situ, real-time spectral monitoring during semiconductor epitaxial growth. Furthermore, the semiconductor epitaxial growth control system acquires reflectance spectral data in real time during the epitaxial growth process and analyzes it to obtain an optical feature set related to the interface transition layer. This optical feature set is then input into a preset optical transmission model, and the variable physical parameter set of the optical transmission model is iteratively optimized to obtain a predicted growth rate. The predicted growth rate is compared with the target growth rate to generate deviation information, which in turn outputs process adjustment control commands, allowing the main control system to perform closed-loop correction of process parameters in the next process cycle. Using this scheme, the optical feature set of this application can dynamically characterize the influence of the layer thickness, composition, and interface change state of the interface transition layer on the reflectance spectrum without relying on fixed optical constants and the assumption of an ideal abrupt interface change, thereby improving the accuracy and robustness of inversion of the actual epitaxial structure state. Furthermore, by performing closed-loop correction of process parameters based on the predicted growth rate obtained through iterative optimization, it is possible to promptly suppress growth rate drift and adaptively compensate for process disturbances, thereby reducing the need for manual calibration and intervention and improving the stability, repeatability, and batch-to-batch consistency of the epitaxial growth process.

[0074] Please see Figure 2In one exemplary embodiment, this application provides a method for controlling semiconductor epitaxial growth, including:

[0075] Step 202: Obtain reflectance spectral data; wherein, the reflectance spectral data is generated based on the reflected light signal generated during the semiconductor epitaxial growth process.

[0076] Specifically, reflectance spectral data refers to the time-series reflectance spectral data composed of multiple frames of reflectance spectra collected by a detector at a preset frequency after the broadband target light signal output by the target light source is incident on the wafer surface in the reaction cavity during the semiconductor epitaxial growth process.

[0077] Step 204: Analyze the reflection spectrum data to obtain an optical feature set; wherein, the optical feature set is associated with the layer thickness, composition and interface change state of the interface transition layer located between adjacent material layers.

[0078] Specifically, epitaxial growth on the wafer surface within the reaction cavity forms a multilayer thin-film structure, primarily comprising a distributed Bragg reflector (DBR) periodic layer, a cavity layer, and a transition layer. When broadband light is incident, reflection and transmission occur at the interfaces between layers, and phase accumulation occurs during propagation within the layers. Multiple reflected beams coherently superimpose at the exit end, ultimately forming a wavelength-dependent reflection spectrum. Therefore, the reflection intensity at each wavelength point in the reflection spectrum contains relevant information about the interface transition layer.

[0079] Based on the optical feature set obtained from the above reflection spectrum analysis, this set of multiple characteristic quantities characterizes the optical response of the epitaxial structure and can reflect the changes in the optical interference and optical resonance properties of the epitaxial layer and its interface transition layer during the growth process. For example, the DBR stop band features include the stop band center wavelength and stop band width, which are mainly sensitive to changes in layer thickness and composition-related average refractive index, and are the basis for inverting the growth rate and average composition; the Fabry-Perot (FP) cavity mode features include the center wavelength, cavity mode depth, and cavity mode half-width at half-maximum, which are extremely sensitive to the optical thickness of the cavity, and their depth and shape strongly depend on the interface sharpness of the gradient layer and the optical constants (n, k) of the material, serving as the key basis for inverting the interface gradient thickness and real-time optical constants; the reflectivity at a specific wavelength point provides an absolute intensity calibration reference and contains information on material absorption (k value).

[0080] For example, variations in the thickness and material composition of the interface transition layer, as well as interface changes such as the degree of interface gradation, diffusion, or roughness, can cause changes in the refractive index and extinction coefficient, further leading to shifts or deformations in the position and width of the stop band, the position and shape of the cavity mode, and the reflectance amplitude at characteristic wavelengths in the reflection spectrum. Therefore, the aforementioned optical feature set can serve as a comprehensive characterization input of the interface transition layer thickness, composition, and interface change state, used for subsequent parameter inversion and growth rate prediction of the optical transmission model. The optical feature set includes at least one of the following: the center wavelength of the stop band of the distributed Bragg reflector, the stop band width, the center wavelength of the Fabry-Perot cavity mode, the cavity mode depth, the cavity mode half-width at half-maximum, and the absolute reflectance value at a preset wavelength point. Step 206: Input the optical feature set into a preset optical transmission model, iteratively optimize the variable physical parameter set of the optical transmission model until the preset conditions are met, and obtain the target parameter set.

[0081] Specifically, an optical transmission model refers to a physical model used to describe how a broadband target optical signal is incident on an epitaxial multilayer structure of a semiconductor and forms a reflected spectrum. For example, the optical transmission model in this embodiment can employ the transfer matrix method (TMM) or rigorous coupled-wave analysis (RCWA).

[0082] A variable physical parameter set refers to a set of physical quantities that dynamically change with the process during the inversion of the optical transmission model. The target parameter set is the variable physical parameter set that satisfies the preset conditions. For example, in this embodiment, the variable physical parameter set includes the thickness of the current layer, the thickness of the interface transition layer, the composition, and the optical constants of the material, and the optical constants include the refractive index n and the extinction coefficient k.

[0083] Under the constraints of a given set of variable physical parameters, the optical transmission model performs forward calculations based on this set of variable physical parameters and outputs the corresponding predicted reflection spectrum.

[0084] Step 208: Generate a predicted growth rate based on the target parameter set; compare the predicted growth rate with the target growth rate to obtain deviation information; generate process adjustment control instructions based on the deviation information so that the main control system executes the epitaxial growth process according to the adjusted process parameters in the next process cycle.

[0085] Specifically, the growth rate is calculated based on the current layer thickness within the target parameter set, characterizing the increase in epitaxial thickness grown by the semiconductor per unit time. Typically, during semiconductor epitaxial growth, key process parameters such as effective precursor supply and surface reaction kinetics are difficult to measure directly, but are reflected in the growth rate; furthermore, rate deviations can be directly mapped to the input variables of the process parameters. Based on this, this application uses the growth rate as a real-time indicator for closed-loop feedback, enabling rapid and effective correction and improving calibration accuracy.

[0086] Generally speaking, the target growth rate is determined by both the process formulation target and the process capability of the semiconductor equipment.

[0087] The main control system refers to the main control platform of the epitaxial growth equipment, which is used to execute the process formula and control the execution units such as mass flow controllers, heaters, and pressure control valves to complete the epitaxial growth process.

[0088] The aforementioned control method for semiconductor epitaxial growth acquires reflectance spectral data in real time during the epitaxial growth process and analyzes it to obtain an optical feature set related to the interface transition layer. This optical feature set is then input into a preset optical transmission model, and the variable physical parameter set of the optical transmission model is iteratively optimized to obtain a predicted growth rate. The predicted growth rate is compared with the target growth rate to generate deviation information, and then a process adjustment control command is output, enabling the main control system to perform closed-loop correction of process parameters in the next process cycle. Using this scheme, the optical feature set of this application can dynamically characterize the influence of the layer thickness, composition, and interface change state of the interface transition layer on the reflectance spectrum without relying on fixed optical constants and the assumption of an ideal abrupt interface change, thereby improving the inversion accuracy and robustness of the actual epitaxial structure state. Furthermore, closed-loop correction of process parameters based on the predicted growth rate obtained through iterative optimization can achieve timely suppression of growth rate drift and adaptive compensation for process disturbances, reducing the need for manual calibration and intervention, and improving the stability, repeatability, and batch consistency of the epitaxial growth process.

[0089] Please see Figure 3 Optionally, when the optical feature set includes at least one of the following: the center wavelength of the stop band of the distributed Bragg reflector, the stop band width, the center wavelength of the Fabry-Perot cavity mode, the cavity mode depth, the cavity mode half-width at half-maximum, and the absolute reflectance value at a preset wavelength point, the reflection spectrum data is analyzed to obtain the optical feature set, including:

[0090] Step 302: Perform spectral feature identification on the reflectance spectral data, determine the stopping band of the distributed Bragg mirror, and extract the center wavelength and / or stopping band width of the stopping band.

[0091] Step 304, and / or, extract cavity mode features from the reflectance spectral data, determine the Fabry-Perot cavity mode, and extract the center wavelength, cavity mode depth, and / or cavity mode half-width.

[0092] Step 306, and / or, select at least one preset wavelength point in the reflectance spectral data and extract the absolute reflectance value at the preset wavelength point.

[0093] Optionally, spectral feature identification is performed on the reflectance spectral data to determine the stopping band of the distributed Bragg mirror, and the center wavelength and / or stopping band width are extracted, including:

[0094] The reflectance spectral data is preprocessed, including smoothing filtering to suppress random noise and baseline correction to reduce the impact of light source spectral shape and link attenuation on the overall spectral envelope. Optionally, the smoothing filtering includes moving average filtering, Savitzky-Golay filtering, or median filtering.

[0095] Within a preset wavelength search range, calculate the maximum reflectance Rmax and average reflectance Rref of the preprocessed reflectance spectral data, where the average reflectance is the average value at both ends of the wavelength search range. Construct a stop band determination threshold using the following formula: Rth = Rref + α × (Rmax - Rref), where α is a preset scaling factor, typically set to 0.5-0.8. Bands with reflectance greater than the stop band determination threshold and continuously covered by a wavelength span greater than a preset minimum span are identified as candidate stop bands.

[0096] For each candidate stop zone band, the two wavelengths corresponding to the intersection of the reflectivity curve and the stop zone determination threshold are determined and designated as the short-wave boundary wavelength λL1 and the long-wave boundary wavelength λH1 of the stop zone, respectively. When multiple candidate stop zone bands exist, the band with the highest average reflectivity or the largest band span can be selected as the stop zone. The center wavelength of the stop zone is calculated using the expression: λC = (λL1 + λH1) / 2; the width of the stop zone is calculated using the expression: Δλ = λH1 - λL1.

[0097] Optionally, cavity mode feature extraction is performed on the reflectance spectral data to determine the Fabry-Perot cavity mode, and the center wavelength, cavity mode depth, and / or cavity mode half-width at half-maximum (WHM) of the Fabry-Perot cavity mode are extracted, including:

[0098] After identifying the DBR stop band, a preset sub-interval within the stop band is set as the cavity mode search interval. Within this cavity mode search interval, extreme value detection is performed on the preprocessed reflectance spectral data to locate local peaks or valleys as cavity mode candidate points. For example, this embodiment uses a method of zero-crossing of the first derivative and sign discrimination of the second derivative to identify local extreme points. It should be noted that the above identification of the DBR stop band and preprocessing of the reflectance spectral data both adopt the method steps provided in the above embodiments, which will not be repeated here for the sake of brevity.

[0099] Candidate points that meet the preset cavity mode determination conditions are selected as the center wavelength λcav of the FP cavity mode; wherein, the preset cavity mode determination conditions include the candidate point being located within the cavity mode search interval, the spectral width of the candidate feature falling within the preset range, and the reflectivity difference of the candidate point relative to the surrounding baseline meeting the preset amplitude threshold.

[0100] A preset wavelength window is selected on both sides of the center wavelength. The average reflectance within the window is calculated as the baseline reflectance Rbase. The reflectance at the center wavelength is taken as R(λcav), then the cavity mode depth D = Rbase - R(λcav). Based on the cavity mode depth and the baseline reflectance, a half-height level is constructed, which is expressed as: Rhalf = Rbase - D / 2. Wavelength positions satisfying reflectance equal to Rhalf are searched on both sides of the center wavelength, obtaining the left intersection wavelength λL2 and the right intersection wavelength λH2. The cavity mode half-height width is calculated, which is expressed as: W = λH2 - λL2.

[0101] Optionally, at least one preset wavelength point is selected from the reflectance spectral data, and the absolute reflectance value at the preset wavelength point is extracted, including:

[0102] Based on the design center wavelength of the epitaxial structure to be monitored, the location of the DBR stop band, and the optimal signal-to-noise ratio range of the spectrometer, one or more preset wavelength points λp are pre-selected. For example, each preset wavelength point can be set within the DBR stop band, near the edge of the stop band, and near the cavity mode to improve the sensitivity to changes in layer thickness, composition, or interface state.

[0103] The reflectance spectral data includes measured light intensity data from multiple wavelength channels. The measured light intensity at a preset wavelength point and the reference light intensity at the same wavelength point are normalized to obtain the absolute reflectance value R(λp,t) at the preset wavelength point. The reference spectrum is obtained under the conditions of a standard reflector or a reference sheet with known reflectance.

[0104] By employing the above scheme, through hierarchical spectral analysis of the reflectance spectral data and extraction of multiple optical feature sets with clear physical meanings, key information in the original reflectance spectrum can be structurally expressed without adding additional hardware sensors. Furthermore, this application can quickly and accurately lock and quantify optical feature sets from complex reflectance spectra, improving the real-time performance and repeatability of feature extraction. Specifically, the extracted center wavelength and / or stop band width of the DBR stop band can characterize the reflectance band drift and deformation caused by changes in DBR period thickness, refractive index contrast, and interface quality, facilitating the identification of process anomalies such as component drift and intensified interface changes. The extracted center wavelength, cavity mode depth, and / or cavity mode half-width at half-maximum (FWHM) of the FP cavity mode can characterize changes in cavity length and cavity loss, providing higher resolution for epitaxial layer thickness evolution, scattering enhancement, and interface state changes. The absolute reflectance value extracted at the preset wavelength point can form a rapid monitoring index for spectral amplitude, reflecting link attenuation, overall changes in reflection intensity, and changes in local optical response. The resulting multidimensional optical feature set can significantly improve the observability and robustness of subsequent optical transmission model inversion, reduce the systematic errors caused by relying solely on fixed model parameter fitting, and provide more stable and quantifiable inputs for growth rate prediction and closed-loop process adjustment, thereby improving the monitoring accuracy, process consistency and yield of the epitaxial growth process.

[0105] Please see Figure 4 Optionally, when the variable physical parameter set includes the thickness of the current layer, the thickness of the interface transition layer, the composition, and the optical constants of the material, and the optical constants include refractive index and extinction coefficient, the optical feature set is input into a preset optical transmission model, and the variable physical parameter set of the optical transmission model is iteratively optimized until the preset conditions are met, thus obtaining the target parameter set, including:

[0106] Step 402: Construct the objective function.

[0107] Step 404: Based on the current set of variable physical parameters, perform forward calculations on the preset optical transmission model to obtain the predicted reflection spectrum data.

[0108] Step 406: Analyze the predicted reflectance spectrum data to obtain the predicted feature set. Calculate the difference between the predicted feature set and the optical feature set according to the objective function. Iteratively optimize the current variable physical parameter set based on the difference until the preset conditions are met. Use the variable physical parameter set at this time as the target parameter set.

[0109] Specifically, the set of variable physical parameters to be inverted is θ = [d, x, σ, n, k], where d is the thickness of the current layer; σ is the thickness of the interface transition layer; x is the composition; n is the refractive index; and k is the extinction coefficient. The optical feature set extracted using the methods described in steps 302-306 above is denoted as F_meas.

[0110] Optionally, the objective function is the squared L2 norm of the difference between the predicted feature set and the optical feature set, expressed as: L(θ) = ||F_pred(θ) - F_meas||², where F_pred(θ) is the predicted feature set.

[0111] Optionally, the difference between the predicted feature set and the optical feature set is calculated based on the objective function, and the current set of variable physical parameters is iteratively optimized based on the difference until a preset condition is met, including:

[0112] The objective function is used to differiate the predicted feature set and the optical feature set to obtain the feature residual; the sensitivity matrix of the feature residual relative to the current variable physical parameter set is calculated; based on the feature residual and the sensitivity matrix, an iterative optimization algorithm is used to solve the parameter increment, and all parameter values ​​in the variable physical parameter set are updated synchronously based on the parameter increment until the objective function converges to less than the preset threshold.

[0113] Specifically, the purpose of iteration is to find a set of variable physical parameters θ such that the objective function L(θ) converges to satisfy a preset condition. In this embodiment, the preset condition is convergence to less than a preset threshold. For example, the iterative optimization algorithm in this embodiment can be a nonlinear least squares optimization algorithm (Levenberg–Marquardt, LM) or a Bayesian optimization algorithm based on a surrogate model. The following explanation uses the LM algorithm as an example for the iterative optimization algorithm.

[0114] In each iteration, based on the current set of variable physical parameters θ, a forward calculation is performed using the optical transmission model to obtain the predicted reflectance spectrum data. Using the same method as in steps 302-306, the predicted feature set F_pred(θ) is extracted from the predicted reflectance spectrum data. The difference between the predicted feature set F_pred(θ) and the optical feature set F_meas is calculated to obtain the feature residual δF, which is then used as the baseline residual vector.

[0115] A preset perturbation is applied sequentially to each parameter in the current set of variable physical parameters, forming corresponding perturbation parameters. Only one parameter is changed during each perturbation, while the remaining parameters remain at their baseline values. For each perturbation parameter, the above forward calculation and feature extraction are repeated once to obtain a perturbation prediction feature set. The difference between this perturbation prediction feature set and the aforementioned baseline residual vector is calculated, and this difference is compared with the perturbation amount of the parameter to obtain the degree of influence of the parameter change on each feature residual. This degree of influence is used as a row or column of a sensitivity matrix, ultimately yielding the sensitivity matrix J for all parameters in the set of variable physical parameters.

[0116] Solving the iterative optimization algorithm equations We obtain the parameter increment Δθ, and adjust all parameters in the variable physical parameter set θ according to the parameter increment. Here, T is the transpose, μ is the damping coefficient, and I is the identity matrix.

[0117] Repeat the above steps until the absolute value of the difference between the objective function values ​​of two adjacent iterations is less than the preset threshold. If the preset condition is met, the value of the variable physical parameter set at this time is taken as the value of the objective parameter set.

[0118] Optionally, this application supports an online learning mode, which can fine-tune the initial parameters of the optical transmission model based on historical successful inversion data to achieve self-optimization.

[0119] By employing the above scheme, the difference between the measured optical feature set and the predicted feature set calculated based on the optical transmission model is constructed as the objective function, transforming the epitaxial structure parameter inversion process from empirical parameter tuning into a quantifiable and convergent optimization solution process. Furthermore, by iteratively calculating the feature residuals and their sensitivity relationship relative to the variable physical parameter set, and simultaneously updating multiple coupling parameters in the variable physical parameter set, adaptive correction can be achieved under conditions of process dynamic drift and interface non-ideal, reducing the systematic errors caused by the fixed parameter model and improving the accuracy and robustness of the inversion results. At the same time, the optimized objective parameter set has clear physical meaning and can be used to output a more reliable growth rate in real time, providing a stable input for subsequent closed-loop control, thereby improving the stability, repeatability, and batch consistency of the epitaxial growth process, and reducing the costs of manual calibration and downtime maintenance.

[0120] Optionally, generating a predicted growth rate based on the target parameter set includes:

[0121] The iterative optimization process is performed at multiple consecutive sampling times to obtain the target parameter set corresponding to each sampling time, and the target layer thickness of the current layer in each target parameter set is obtained; for the current sampling time, the target layer thickness corresponding to the previous sampling time adjacent to it is obtained, and the time interval between adjacent sampling times is determined; the difference of the target layer thickness is calculated and divided by the time interval to obtain the predicted growth rate at the current sampling time.

[0122] Specifically, the target parameter set obtained from the above inversion is {d_opt, x_opt, σ_opt, n_opt, k_opt}. The expression for the predicted growth rate is then calculated as: r(t) = (d_opt(t) - d_opt(t-Δt)) / Δt, where d_opt(t) and d_opt(t-Δt) are the target layer thicknesses of the current layer at the current sampling time and the previous sampling time, respectively; Δt is the time interval. It should be understood that this predicted growth rate is an instantaneous value.

[0123] By adopting the above scheme, since the layer thickness is obtained by inversion under the condition of considering the changes in composition, interface gradient and material optical constants, the predicted growth rate calculated based on the layer thickness has stronger robustness and consistency against process drift, interface non-ideality and changes in optical constants. It can realize real-time and continuous tracking of growth rate, and provide more reliable feedback basis for subsequent closed-loop adjustment of process parameters such as precursor flow rate, temperature or pressure, thereby improving the accuracy of epitaxial thickness control and batch repeatability.

[0124] Optionally, based on the deviation information, process adjustment control instructions are generated, including:

[0125] Based on the deviation information, the control quantity is calculated using a preset control algorithm, which includes a proportional-integral-derivative control algorithm and / or a model predictive control algorithm.

[0126] Based on the control quantity, process adjustment control instructions are generated to adjust the epitaxial growth process parameters, including the flow rate parameters of the metal-organic precursor and / or the growth temperature parameters.

[0127] Specifically, when the preset control algorithm includes a proportional-integral-derivative (PI-DE) control algorithm, the PI-DE control algorithm calculates the control quantity based on the current value, cumulative value, and trend of the deviation information. When the preset control algorithm includes a model predictive control (MMC) algorithm, the MMC control algorithm predicts the future growth rate based on the deviation information and process state variables, and solves for the control quantity under constraints. Here, constraints refer to the safety boundaries of the process and equipment, and the control quantity is the parameter increment and / or target setpoint used to adjust the process parameters.

[0128] After receiving the process adjustment control command, the main control system adjusts the process parameters according to the control quantity, and executes the epitaxial growth process according to the adjusted process parameters in the next process cycle. For example, the metal-organic precursor includes trimethylaluminum (TMAl) or trimethylgallium (TMGa).

[0129] By adopting the above scheme, the growth rate deviation is converted into an executable control quantity, realizing closed-loop feedback regulation of the epitaxial growth process. This enables timely correction of deviations under conditions such as source bottle efficiency decay, mass transfer condition fluctuations, and cavity state drift, suppressing the accumulation of rate deviations and improving layer thickness control accuracy and batch-to-batch consistency. Furthermore, the control commands can be directly applied to the metal-organic precursor flow parameters and / or growth temperature parameters, making the process adjustment path clear and the execution efficiency high, thereby improving the real-time controllability of epitaxial growth and reducing the probability of manual intervention and rework.

[0130] The aforementioned control method for semiconductor epitaxial growth acquires reflectance spectral data in real time during the epitaxial growth process and analyzes it to obtain an optical feature set related to the interface transition layer. This optical feature set is then input into a preset optical transmission model, and the variable physical parameter set of the optical transmission model is iteratively optimized to obtain a predicted growth rate. The predicted growth rate is compared with the target growth rate to generate deviation information, and then a process adjustment control command is output, enabling the main control system to perform closed-loop correction of process parameters in the next process cycle. Using this scheme, the optical feature set of this application can dynamically characterize the influence of the layer thickness, composition, and interface change state of the interface transition layer on the reflectance spectrum without relying on fixed optical constants and the assumption of an ideal abrupt interface change, thereby improving the inversion accuracy and robustness of the actual epitaxial structure state. Furthermore, closed-loop correction of process parameters based on the predicted growth rate obtained through iterative optimization can achieve timely suppression of growth rate drift and adaptive compensation for process disturbances, reducing the need for manual calibration and intervention, and improving the stability, repeatability, and batch consistency of the epitaxial growth process.

[0131] 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.

[0132] Based on the same inventive concept, this application also provides a control system for semiconductor epitaxial growth. This system is applicable to the above-described control method for semiconductor epitaxial growth. The solution provided by this system is similar to the solution described in the above-described method. Therefore, the specific limitations in one or more system embodiments provided below can be found in the limitations of the method described above, and will not be repeated here.

[0133] Please see Figure 5 In one embodiment, a control system for semiconductor epitaxial growth is provided. The control system is disposed within a control device, which is partially disposed within the reaction chamber of a semiconductor device. The control system for semiconductor epitaxial growth includes an acquisition module, an analysis module, a prediction module, and a processing module.

[0134] The acquisition module is used to acquire reflectance spectral data; wherein, the reflectance spectral data is generated based on the reflected light signal generated during the semiconductor epitaxial growth process.

[0135] The analysis module is used to analyze the reflection spectrum data to obtain an optical feature set; the optical feature set is associated with the layer thickness, aluminum composition and interface change state of the interface transition layer located between adjacent material layers.

[0136] The prediction module is used to input the optical feature set into the preset optical transmission model, iteratively optimize the variable physical parameter set of the optical transmission model until the preset conditions are met, and obtain the target parameter set.

[0137] The processing module is used to generate a predicted growth rate based on the target parameter set; compare the predicted growth rate with the target growth rate to obtain deviation information; and generate process adjustment control instructions based on the deviation information so that the main control system can execute the epitaxial growth process according to the adjusted process parameters in the next process cycle.

[0138] Optionally, when the optical feature set includes at least one of the center wavelength of the stop band of the distributed Bragg reflector, the stop band width, the center wavelength of the Fabry-Perot cavity mode, the cavity mode depth, the cavity mode half-width at half-maximum (FWHM), and the absolute reflectance value at a preset wavelength point, the analysis module analyzes the reflection spectral data to obtain the optical feature set, including: performing spectral feature identification on the reflection spectral data to determine the stop band of the distributed Bragg reflector and extracting the center wavelength and / or stop band width of the stop band; and / or, performing cavity mode feature extraction on the reflection spectral data to determine the Fabry-Perot cavity mode and extracting the center wavelength, cavity mode depth, and / or cavity mode FWHM of the Fabry-Perot cavity mode; and / or, selecting at least one preset wavelength point in the reflection spectral data and extracting the absolute reflectance value at the preset wavelength point.

[0139] Optionally, when the variable physical parameter set includes the thickness of the current layer, the thickness of the interface transition layer, the composition, and the optical constants of the material, and the optical constants include refractive index and extinction coefficient, the prediction module inputs the optical feature set into a preset optical transmission model, iteratively optimizes the variable physical parameter set of the optical transmission model until the preset conditions are met, and obtains the target parameter set, including: constructing an objective function; performing forward calculation on the preset optical transmission model based on the current variable physical parameter set to obtain predicted reflectance spectrum data; parsing the predicted reflectance spectrum data to obtain a predicted feature set; calculating the difference between the predicted feature set and the optical feature set based on the objective function; iteratively optimizing the current variable physical parameter set based on the difference until the preset conditions are met, and using the variable physical parameter set at this time as the target parameter set.

[0140] Optionally, the prediction module calculates the difference between the predicted feature set and the optical feature set based on the objective function, and iteratively optimizes the current variable physical parameter set based on the difference until a preset condition is met. This includes: using the objective function to perform a difference between the predicted feature set and the optical feature set to obtain the feature residual; calculating the sensitivity matrix of the feature residual relative to the current variable physical parameter set; and using an iterative optimization algorithm to solve for the parameter increment based on the feature residual and the sensitivity matrix, and synchronously updating all parameter values ​​in the variable physical parameter set based on the parameter increment until the objective function converges to less than a preset threshold.

[0141] Optionally, the processing module generates a predicted growth rate based on the target parameter set, including: performing an iterative optimization process at multiple consecutive sampling times to obtain the target parameter set corresponding to each sampling time, and obtaining the target layer thickness of the current layer in each target parameter set; for the current sampling time, obtaining the target layer thickness corresponding to the previous sampling time adjacent to it, and determining the time interval between adjacent sampling times; calculating the difference in target layer thickness and dividing it by the time interval to obtain the predicted growth rate at the current sampling time.

[0142] Optionally, the processing module generates process adjustment control instructions based on the deviation information, including: calculating the control quantity based on the deviation information using a preset control algorithm, wherein the preset control algorithm includes a proportional-integral-derivative control algorithm and / or a model predictive control algorithm; and generating process adjustment control instructions for adjusting epitaxial growth process parameters based on the control quantity, wherein the process parameters include the flow rate parameters of the metal-organic precursor and / or the growth temperature parameters.

[0143] The aforementioned control system for semiconductor epitaxial growth acquires reflectance spectral data in real time during the epitaxial growth process and analyzes it to obtain an optical feature set related to the interface transition layer. This optical feature set is then input into a preset optical transmission model, and the variable physical parameter set of the optical transmission model is iteratively optimized to obtain a predicted growth rate. The predicted growth rate is compared with the target growth rate to generate deviation information, and then a process adjustment control command is output, enabling the main control system to perform closed-loop correction of process parameters in the next process cycle. Using this approach, the optical feature set of this application can dynamically characterize the influence of the layer thickness, composition, and interface change state of the interface transition layer on the reflectance spectrum without relying on fixed optical constants and the assumption of an ideal abrupt interface change, thereby improving the inversion accuracy and robustness of the actual epitaxial structure state. Furthermore, closed-loop correction of process parameters based on the predicted growth rate obtained through iterative optimization can achieve timely suppression of growth rate drift and adaptive compensation for process disturbances, reducing the need for manual calibration and intervention, and improving the stability, repeatability, and batch consistency of the epitaxial growth process.

[0144] Each module in the aforementioned semiconductor epitaxial growth control system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the operations corresponding to each module.

[0145] In one feasible embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 6 As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface 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 interface. 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 interface is 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 control method for semiconductor epitaxial growth. 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.

[0146] Those skilled in the art will understand that Figure 6 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.

[0147] 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 control method for semiconductor epitaxial growth.

[0148] 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 control method for semiconductor epitaxial growth described above.

[0149] 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 control method for semiconductor epitaxial growth described above.

[0150] 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.

[0151] 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.

[0152] 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 of controlling the epitaxial growth of a semiconductor, characterized by, The method comprises: acquiring reflectance spectrum data; wherein the reflectance spectrum data is generated based on a reflected light signal generated in a semiconductor epitaxial growth process; analyzing the reflectance spectrum data to obtain an optical feature set; wherein the optical feature set comprises at least one of a stop band center wavelength, a stop band width, a Fabry-Perot cavity mode center wavelength, a cavity mode depth, a cavity mode full width at half maximum, and an absolute reflectance value at a preset wavelength point of a distributed Bragg reflector; inputting the optical feature set into a preset optical transfer model, iteratively optimizing a variable physical parameter set of the optical transfer model until a preset condition is met to obtain a target parameter set; the variable physical parameter set comprises a layer thickness of a current layer, a layer thickness, a component of an interface transition layer, and optical constants of a material, the optical constants comprising a refractive index and an extinction coefficient; generating a predicted growth rate based on the target parameter set; comparing the predicted growth rate with a target growth rate to obtain deviation information; generating a process adjustment control instruction according to the deviation information to enable a main control system to perform an epitaxial growth process according to an adjusted process parameter in a next process cycle.

2. The method of claim 1, wherein, The analyzing the reflectance spectrum data to obtain an optical feature set comprises: performing spectral feature recognition on the reflectance spectrum data to determine a stop band of a distributed Bragg reflector and extract a center wavelength and / or a stop band width of the stop band; and / or performing cavity mode feature extraction on the reflectance spectrum data to determine a Fabry-Perot cavity mode and extract a center wavelength, a cavity mode depth, and / or a cavity mode full width at half maximum of the Fabry-Perot cavity mode; and / or selecting at least one preset wavelength point in the reflectance spectrum data and extracting an absolute reflectance value at the preset wavelength point.

3. The method of claim 2, wherein: the inputting the optical feature set into a preset optical transfer model, iteratively optimizing a variable physical parameter set of the optical transfer model until a preset condition is met to obtain a target parameter set comprises: constructing a target function; performing forward calculation on the preset optical transfer model according to the current variable physical parameter set to obtain predicted reflectance spectrum data; analyzing the predicted reflectance spectrum data to obtain a predicted feature set, calculating a difference between the predicted feature set and the optical feature set according to the target function, and iteratively optimizing the current variable physical parameter set according to the difference until a preset condition is met, and taking the variable physical parameter set at this time as the target parameter set.

4. The method of claim 3, wherein, the target function is a square of a two-norm difference between the predicted feature set and the optical feature set; the calculating a difference between the predicted feature set and the optical feature set according to the target function, and iteratively optimizing the current variable physical parameter set according to the difference until a preset condition is met comprises: using the target function to perform difference between the predicted feature set and the optical feature set to obtain a feature residual, and calculating a sensitivity matrix of the feature residual with respect to the current variable physical parameter set; Based on the characteristic residual and the sensitivity matrix, a parameter increment is solved by using an iterative optimization algorithm, and all parameter values in the set of variable physical parameters are synchronously updated based on the parameter increment until the objective function converges to less than a preset threshold.

5. The method of claim 1, wherein, The generating a predicted growth rate based on the target parameter set comprises: The iterative optimization process is performed at each of a plurality of continuous sampling time points to obtain a target parameter set corresponding to each sampling time point, and a target layer thickness of a current layer in each target parameter set is obtained; For a current sampling time point, a target layer thickness corresponding to a previous sampling time point adjacent to the current sampling time point is obtained, and a time interval between the adjacent sampling time points is determined; A difference value of the target layer thickness is calculated and divided by the time interval to obtain a predicted growth rate of the current sampling time point.

6. The method of claim 1, wherein, The generating a process adjustment control instruction according to the deviation information comprises: Based on the deviation information, a control amount is calculated by using a preset control algorithm, wherein the preset control algorithm comprises a proportional-integral-derivative control algorithm and / or a model predictive control algorithm; A process adjustment control instruction for adjusting an epitaxial growth process parameter is generated according to the control amount, and the process parameter comprises a flow parameter of a metal organic precursor and / or a growth temperature parameter.

7. A control system for the epitaxial growth of semiconductors, characterised in that The system is arranged in a control device, and the control device is partially arranged in a semiconductor device reaction cavity; the system comprises: An acquisition module is configured to acquire reflectance spectrum data, wherein the reflectance spectrum data is generated based on a reflected light signal generated during an epitaxial growth process of a semiconductor; An analysis module is configured to analyze the reflectance spectrum data to obtain an optical feature set, wherein the optical feature set is associated with a layer thickness of an interface transition layer between adjacent material layers, an aluminum component, and an interface change state; A prediction module is configured to input the optical feature set into a preset optical transmission model, iteratively optimize a set of variable physical parameters of the optical transmission model until a preset condition is met, and obtain a target parameter set; the set of variable physical parameters comprises a layer thickness of a current layer, a layer thickness of an interface transition layer, a component, and optical constants of a material, and the optical constants comprise a refractive index and an extinction coefficient; A processing module is configured to generate a predicted growth rate based on the target parameter set, compare the predicted growth rate with a target growth rate to obtain deviation information, and generate a process adjustment control instruction according to the deviation information, so that a main control system performs an epitaxial growth process according to adjusted process parameters in a next process cycle.

8. A control device characterized by comprising: The control device is partially arranged in a semiconductor device reaction cavity, and the control device comprises: A target light source is configured to output a target light signal during an epitaxial growth process of a semiconductor; A sensing module is partially arranged in the semiconductor device reaction cavity and is configured to guide the target light signal to a wafer surface in the reaction cavity, and continuously collect a reflected light signal generated based on the target light signal on the wafer surface at a preset frequency, and generate reflectance spectrum data based on the reflected light signal; The control system for semiconductor epitaxial growth as claimed in claim 7, configured to calculate deviation information of the predicted growth rate and the target growth rate according to the reflection spectrum data, and generate a process adjustment control instruction according to the deviation information, so that the main control system executes epitaxial growth process according to the adjusted process parameters in the next process cycle.

9. The control device of claim 8, wherein The sensing module includes a fiber probe; the device further includes a cleaning module, the cleaning module including a cooling assembly and a purge assembly; The cooling assembly is configured to actively cool the fiber probe, and the active cooling process includes water cooling and / or air cooling; The purge assembly includes a purge gas path configured to output purge gas to the lens end of the fiber probe to reduce the deposition of contaminants on the lens during the acquisition process.

10. The control device of claim 9, wherein The number of the fiber probes is multiple.

11. The control device of claim 8, wherein The device further includes an interface module, the interface module including multiple industrial standard communication interfaces, each of the industrial standard communication interfaces being configured to communicate with the main control system of the epitaxial growth equipment having different communication protocols.

12. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that, The processor executes the computer program to implement the steps of the method of any one of claims 1-6.

13. A computer readable storage medium having stored thereon a computer program, characterized in that The computer program is executed by the processor to implement the steps of the method of any one of claims 1-6.

Images