In-situ monitoring method and system for uniformity of growth of molecular beam epitaxial semiconductor thin films

Abstract

The present invention discloses a method and system for in-situ monitoring of the uniformity of semiconductor film growth by molecular beam epitaxy. The method includes obtaining spectral information of semiconductor films prepared under different molecular beam epitaxy growth states and characterization data reflecting the uniformity of semiconductor film growth, and constructing a database; obtaining a semiconductor film growth uniformity identification model based on database training; obtaining real-time spectral information of semiconductor films epitaxially grown on the sample surface, and combining the semiconductor film growth uniformity identification model to obtain the growth of the semiconductor film by molecular beam epitaxy. The present invention combines the in-situ monitored film spectral information with an artificial intelligence algorithm to realize non-contact, in-situ online monitoring of the doping concentration of semiconductor films. The present invention can collect optical property data of thin films in real time, quickly process a large amount of complex optical data, automatically extract features and perform efficient processing and analysis capabilities for pattern recognition, and realize efficient real-time monitoring of film uniformity.

Description

Technical Field

[0001] The invention relates to the technical field of semiconductor thin film measurement, and in particular to an in-situ monitoring method and system for the uniformity of growth of a molecular beam epitaxial semiconductor thin film. Background Art

[0002] Traditionally, molecular beam epitaxy (MBE) technology, as an important thin film growth technology, has been widely used in the field of semiconductor device preparation. MBE technology can achieve precise control of semiconductor material thin films by depositing atoms or molecules layer by layer under vacuum conditions, and is therefore widely used in the preparation of high-performance electronic devices, optoelectronic devices, and quantum devices.

[0003] As a key means of material regulation, thin film doping plays a vital role in the fields of semiconductor devices and optoelectronic devices. By introducing doping atoms into the thin film, the electrical, optical and chemical properties of the material can be effectively adjusted, thereby achieving precise control and optimization of the material properties. Doping can be used to change the electrical properties of the material, such as conductivity, carrier concentration, mobility, etc., and it can also adjust the optical absorption and emission characteristics of the material, as well as the chemical reaction activity. Therefore, thin film doping not only provides important technical support for the preparation of high-performance semiconductor devices and optoelectronic devices, but also provides a powerful means to achieve the functional design and performance optimization of new materials.

[0004] However, in practical applications, the uniformity of doping concentration of semiconductor thin films is one of the key factors affecting device performance and reliability. Currently, the widely used monitoring of thin film doping concentration uniformity mainly relies on indirect measurement methods, such as electron beam ion beam analysis (EBIC), atomic force microscopy (AFM), etc. These methods have the disadvantages of high time cost, complex operation, and inability to monitor in real time, which limits the real-time control and optimization of the thin film growth process. In order to solve the above problems, some real-time monitoring methods based on optical principles have emerged in recent years, such as reflected high-energy electron diffraction and Raman spectroscopy. However, these methods have challenges in practical applications, such as complex equipment, high cost, and cumbersome data processing, which limits their wide application in industrial production.

[0005] In recent years, artificial intelligence, as a research field that has attracted much attention, has shown great potential in processing large-scale data and providing accurate data predictions. In particular, in the analysis of optical spectral data, the use of machine learning algorithms provides new possibilities for exploring the relationship between spectral features and doping concentration. By training and modeling spectral data, machine learning algorithms can identify spectral features related to doping, thereby realizing the evaluation of doping concentration distribution and uniformity. Combining artificial intelligence algorithms with optical monitoring technology can more conveniently and comprehensively evaluate the doping concentration of thin films, thereby realizing accurate real-time in-situ monitoring of doping concentration. The application of this method helps to optimize the growth process, improve the performance and stability of thin film devices, and is of great significance for the evaluation of doping concentration uniformity during the growth of semiconductor films.

[0006] Therefore, in response to the need for monitoring the uniformity of doping concentration of semiconductor thin films, there is an urgent need for a method and system that can achieve in-situ, real-time, and efficient monitoring to improve the control accuracy and production efficiency of the semiconductor thin film growth process. Summary of the invention

[0007] The purpose of the present invention is to provide a method and system that can achieve in-situ, real-time and efficient monitoring to improve the control accuracy and production efficiency of the semiconductor thin film growth process and reduce the impact of thin film non-uniformity on the performance, stability and reliability of materials and devices.

[0008] In order to achieve the above object, the present invention provides an in-situ monitoring method for the uniformity of growth of a molecular beam epitaxial semiconductor film, comprising:

[0009] Obtain spectral information of semiconductor films prepared under different molecular beam epitaxial growth conditions and characterization data reflecting the uniformity of semiconductor film growth, and build a database;

[0010] A semiconductor thin film growth uniformity recognition model is obtained by training the database;

[0011] The real-time spectrum information of the semiconductor film epitaxially grown on the sample surface is obtained, and the growth status of the molecular beam epitaxial semiconductor film is obtained by combining the semiconductor film growth uniformity identification model.

[0012] Furthermore, the characterization data of the semiconductor film growth uniformity includes the thickness, doping concentration and morphology of the semiconductor film.

[0013] Furthermore, the semiconductor thin film growth uniformity recognition model obtained by training the database includes:

[0014] An optical simulation model is established to perform simulation calculations, and the simulation results are extracted and compared with the characterization data of the uniformity of semiconductor thin film growth to establish related data, which is then cleaned, trained, and optimized using an artificial intelligence algorithm.

[0015] Furthermore, the cleaning includes missing value processing, noise processing and logic error cleaning as well as standardization preprocessing and dividing the data set and the test set;

[0016] The training is based on a GRU neural network;

[0017] The optimization is performed by adding the Adam optimization module.

[0018] Furthermore, the characterization data reflecting the uniformity of semiconductor thin film growth is obtained based on at least one of XRD, scanning electron microscopy, electron beam ion beam analysis, and atomic force microscopy.

[0019] Furthermore, after obtaining the growth status of the molecular beam epitaxial semiconductor film, the growth process is also controlled so that the film growth is close to the expected value.

[0020] Furthermore, the real-time spectrum information of the semiconductor thin film epitaxially grown on the sample surface is obtained according to a wide spectrum monitoring device arranged at the bottom of the molecular beam epitaxy equipment;

[0021] The molecular beam epitaxy device comprises:

[0022] A vacuum reaction chamber, provided with an installation port and having a bottom part or entirely made of a light-transmitting material, for providing a reaction environment, wherein the installation port is used to install a furnace source;

[0023] A sample stage, arranged in the vacuum reaction chamber;

[0024] A first driving assembly, disposed at the top of the outside of the vacuum reaction chamber and connected to the sample stage through a connecting rod, for driving the sample stage and adjusting the position of the sample stage;

[0025] The wide spectrum monitoring device comprises a second driving component and a monitoring component;

[0026] During operation, the monitoring component driven by the second driving component and the sample on the sample stage driven by the first driving component keep moving synchronously to obtain the information of thin film growth.

[0027] Further, the wide spectrum monitoring device includes a second driving component, a monitoring component and a processing and display component;

[0028] The monitoring assembly includes a fixing component, a polarizing light path, a light source, a polarizing light path, and a detector;

[0029] The fixing part includes a plurality of movable epitaxial ends above, the polarizing optical path and the light source and the analyzing optical path and the detector are installed on different epitaxial ends, the light beam generated by the light source generates a light beam with a specific polarization direction through the polarizing optical path to irradiate the sample surface, the analyzing optical path is used to analyze the polarization spectrum information of the reflected light generated by the sample, and the detector is used to receive, measure and output the polarization spectrum information of the reflected light;

[0030] The second driving assembly comprises a translation component, a rotation component and a lifting component respectively connected to the fixed component, and a workbench is further arranged between the rotation component and the translation component;

[0031] Wherein, the sample stage and the fixing component have the same axis;

[0032] The processing and display component is used to receive and process the signal output by the detector, and perform data analysis and result display.

[0033] The present invention also provides an in-situ monitoring system for the uniformity of growth of a molecular beam epitaxial semiconductor thin film, comprising:

[0034] A database module is used to obtain spectral information of semiconductor thin films prepared under different molecular beam epitaxial growth conditions and characterization data reflecting the uniformity of semiconductor thin film growth, and to construct a database;

[0035] A model building module, used to obtain a semiconductor thin film growth uniformity recognition model based on the database training;

[0036] The identification module is used to obtain real-time spectral information of the semiconductor film epitaxially grown on the sample surface, and obtain the growth status of the molecular beam epitaxial semiconductor film in combination with the semiconductor film growth uniformity identification model.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] The present invention combines the thin film spectrum information monitored in situ with the artificial intelligence algorithm to realize non-contact, in-situ online monitoring of the doping concentration of the semiconductor film, and obtains the growth of the molecular beam epitaxial semiconductor film without affecting the growth of the film. At the same time, the present invention can collect the optical property data of the film in real time, quickly process a large amount of complex optical data, automatically extract features and perform pattern recognition efficient processing and analysis capabilities, and realize efficient real-time monitoring of the uniformity of the film. The present invention is also applicable to various material systems and growth conditions, without excessive human intervention and adjustment, and has strong versatility and applicability. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0040] Figure 1 A flow chart showing the in-situ monitoring method for the uniformity of semiconductor thin film growth by molecular beam epitaxy according to Embodiment 1 of the present invention is shown;

[0041] Figure 2 It shows a flow chart of training a semiconductor thin film growth uniformity recognition model according to Embodiment 1 of the present invention;

[0042] Figure 3 The structure diagram of the molecular beam epitaxy equipment and the wide spectrum monitoring device used in Example 1 of the present invention is shown;

[0043] Figure 4 A detailed schematic diagram of the wide spectrum monitoring device used in Example 1 of the present invention is shown;

[0044] Figure 5 A schematic diagram showing the principle of an artificial intelligence algorithm according to Embodiment 2 of the present invention is shown;

[0045] Figure 6 A schematic diagram of an in-situ monitoring system for the uniformity of growth of a molecular beam epitaxial semiconductor thin film according to Embodiment 3 of the present invention is shown;

[0046] Description of reference numerals:

[0047] 1. Vacuum reaction chamber; 2. Sample stage; 3. First driving assembly; 31. Rotation control component; 32. Lifting control component; 4. Broad spectrum monitoring device; 41. Housing; 42. Monitoring assembly; 421. Fixing component; 422. Polarizing light path; 423. Light source; 424. Polarizing light path; 425. Detector; 43. Second driving assembly; 431. Translation component; 432. Rotation component; 433. Lifting component; 434. Workbench; 44. Processing and display component; 441. Computer; 442. Display screen; 5. First furnace source; 6. Second furnace source; 7. Sample transfer port; 8. Second gate valve; 9. Cold pump; 10. First gate valve; 11. High energy electron gun; 12. Fluorescent screen; 13. Sample; 14. Extension part. DETAILED DESCRIPTION

[0048] The following will be combined with the specific embodiments of the present invention and the drawings of the specification to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] Example 1

[0050] A method for in-situ monitoring of the uniformity of semiconductor thin film growth by molecular beam epitaxy, such as Figure 1 As shown, the following steps are included:

[0051] S101. Acquire spectral information of semiconductor thin films prepared under different molecular beam epitaxial growth conditions and characterization data reflecting the growth uniformity of semiconductor thin films, and construct a database.

[0052] First, according to the preliminary experiment, the semiconductor film obtained under different molecular beam epitaxial growth states can be obtained, and then the obtained semiconductor film is characterized to obtain the spectral information of the wide spectrum measurement of the semiconductor film, as well as the characterization data reflecting the uniformity of the growth of the semiconductor film, including thickness and doping concentration. These characterizations are at least one of XRD, scanning electron microscopy, electron beam ion beam analysis, and atomic force microscopy. Different molecular beam epitaxial growth states can be semiconductor films at different growth times, doping concentrations of semiconductor films at different times, etc., and corresponding preliminary experiments can be constructed according to the needs of monitoring to construct a database. The subsequent steps of this embodiment are mainly based on doping concentration.

[0053] S102, obtaining a semiconductor thin film growth uniformity recognition model according to the database training.

[0054] like Figure 2 As shown, an optical simulation model is established to perform simulation calculations, and the simulation results are extracted and standardized with the characterization data of the uniformity of semiconductor thin film growth to establish related data for comparison, and then the related data are cleaned, trained, and optimized in combination with an artificial intelligence algorithm.

[0055] Because concentration doping cannot be directly reflected in optical simulation software, only an equivalent model can be established for the construction of concentration doping identification model. By changing other parameters of the model, a simulation spectrum similar to the measured spectrum is obtained, and this simulation model is used as an equivalent model under the doping concentration; the measurement sample is a semiconductor thin film grown on the same substrate, and the growth time can form a continuous time series, and multiple measurement data samples of the same material with different growth times are equivalently replaced by data of the doping concentration of a sample changing with time; in the simulation process, an equivalent geometric model of proportional size is established according to the measurement sample through optical simulation software, the material properties of the semiconductor material are input, the light source information and measurement configuration are set, and finally the simulation analysis is performed to output the diffraction result; the spectrum output by the simulation is fitted with the measured spectrum to determine whether the optical model is accurate. If there is a deviation, the optical simulation model needs to be readjusted and the simulation calculation is re-performed until the model error value is less than 5%; if the spectrum output by the simulation fits the measured spectrum and the model error value is less than 5%, the doping concentration and spectrum parameters under the growth time are output, and the data are included in the simulation spectrum database and the doping concentration model database.

[0056] The data needs to be cleaned first, including missing value processing, noise processing and logic error cleaning, as well as standardized preprocessing and dividing the data set and test set. Then, deep learning is performed based on the GRU neural network to output the predicted doping concentration data. During the training process, an optimization module is added to optimize the neural network model. Finally, the real-time measured spectral data is input into the trained neural network model to output the parameters for predicting the change of the film doping concentration over time. By inputting the real-time data of multiple measurement points, the change of the film doping concentration over time at each point can be output.

[0057] Missing value processing includes using interpolation to handle partial data missing and using data from adjacent time periods to fill in all missing data. Noise processing includes using basic data statistical description methods and data visualization methods to identify noise data and then analyze it. Logical error cleaning includes filtering duplicate data, removing obviously unreasonable values, correcting contradictory data, and other outlier cleaning methods.

[0058] In the data standardization preprocessing, the collected data is normalized and the data is linearly transformed. The transformation formula is: norm =(xx mean ) / x std ;

[0059] Where: x norm is the data after standardization, x is the collected data, and x mean is the mean within a single sequence, x std is the standard deviation of all values ​​in a single series.

[0060] The specific way to divide the data set is to divide the data into a training set T tr and the test set T te , where the training set accounts for 80% and the test set accounts for 20%. i Each has its corresponding target label y i .

[0061] The prediction model uses the GRU neural network. As a variant of the LSTM neural network, GRU simplifies the network structure while retaining the excellent performance of LSTM, thereby greatly reducing the training time. The GRU neural network changes the design of the gate structure and only uses the update gate and the reset gate. The update gate determines how much previous information can continue to be passed on, and the reset gate determines how much information needs to be forgotten. This structure enables the GRU neural network to have excellent performance in model convergence, parameter updating and generalization capabilities. Since the information transmission between neurons in the GRU neural network is highly dependent, it is very suitable for processing upper interval time series problems, and the prediction of semiconductor film doping concentration happens to fall into this category.

[0062] Neural network training uses the mean square error (MSE) function to evaluate the accuracy of model prediction. The mean square error reflects the deviation between the predicted value and the true value. The calculation formula is:

[0063]

[0064] In the formula Represents the predicted value, y i represents the predicted value, and n represents the number of prediction periods;

[0065] The parameters adjusted during model training mainly include: number of hidden layers, number of nodes in the hidden layer, training rounds, batch size, and optimizer;

[0066] The optimizer uses the Adam optimizer. At this time, the prediction effect of the GRU neural network model is generally good, and the error function reaches the minimum value. The choice of the optimizer is determined according to the experimental results. GRU selects 100 experimental rounds and a batch size of 30. When the optimizer is Adam, the prediction effect of the network model is relatively accurate.

[0067] S103, obtaining real-time spectrum information of the semiconductor thin film epitaxially grown on the sample surface, and combining the semiconductor thin film growth uniformity identification model to obtain the growth status of the molecular beam epitaxial semiconductor thin film.

[0068] In this embodiment, a wide spectrum monitoring device 4 is arranged at the bottom of a conventional molecular beam epitaxy device to obtain real-time spectrum information of the semiconductor thin film epitaxially grown on the sample surface. Figure 3 and Figure 4As shown, the molecular beam epitaxy device includes a vacuum reaction chamber 1, a sample stage 2 located in the vacuum reaction chamber 1, and a first driving assembly 3. The side wall of the vacuum reaction chamber 1 is provided with a mounting port, a sample transfer cavity, a sample transfer port 7, a cold pump 9, and a reflection high-energy electron diffraction device, and a bottom portion of the vacuum reaction chamber 1 is extended inward to form an extension portion 14, and the extension portion 14 provides space for the installation of the wide spectrum monitoring device 4, and the extension height is 1 / 4 to 1 / 3 of the height of the vacuum reaction chamber 1, and the material of the extension portion 14 is a light-transmitting material. Among them, the installation port is used to install the furnace source, which includes the first furnace source 5 and the second furnace source 6. The installation port and the vacuum reaction chamber 1 are at a certain angle in the axial direction to ensure the uniform deposition of the growing material; the sample transfer port 7 is connected to the vacuum reaction chamber 1 through a flange and a sealing ring, and a second gate valve 8 is arranged at the connection. The sample 13 can be transferred from the sample transfer chamber to the sample stage 2 through a ring manipulator, wherein the sample stage 2 can also heat the sample 13; the cold pump 9 is connected to the vacuum reaction chamber 1 through a flange and a sealing ring to maintain the vacuum degree inside the vacuum reaction chamber 1; the first gate valve 10 is arranged at the connection between the vacuum reaction chamber 1 and the cold pump 9, and the first gate valve 10 is used to control the pressure atmosphere inside the vacuum reaction chamber 1 and protect the cold pump 9; the reflection type high energy electron diffraction device includes a high energy electron gun 11 and a fluorescent screen 12. The first driving component 3 is arranged at the top of the outside of the vacuum reaction chamber 1 and is connected to the sample stage 2 through a connecting rod, and is used to drive the sample stage 2 and adjust the position of the sample stage 2. The first driving assembly 3 includes a rotation control component 31 and a lifting control component 32, which are respectively used to control the rotation and lifting of the sample stage 2. The rotation control component 31 includes a stepper motor, a turntable, a transmission device and a controller; the lifting control system includes a stepper motor, a linear guide rail, a transmission device and a controller.

[0069] The wide spectrum monitoring device 4 is arranged at the bottom of the vacuum reaction chamber 1, and includes a second driving component 43, a monitoring component 42, a housing 41 and a processing display component 44. The second driving component 43 and the monitoring component 42 are arranged in the housing 41, the upper end of the outer shell 41 abuts against the extension portion 14 of the vacuum reaction chamber 1, and the lower end of the outer shell 41 is located below the bottom of the vacuum reaction chamber 1. Figure 4As shown, the monitoring assembly 42 includes a fixed component 421, a polarizing optical path 422, a light source 423, a polarizing optical path 424, and a detector 425. The upper portion of the fixed component 421 includes a plurality of movable epitaxial ends. In this embodiment, there are two epitaxial ends. The fixed component 421 presents a "Y"-shaped structure. The left epitaxial end is provided with a polarizing optical path 424 and a detector 425, and the right epitaxial end is provided with a polarizing optical path 422 and a light source 423. The light beam generated by the light source 423 generates a light beam with a specific polarization direction through the polarizing optical path 422 and irradiates the surface of the sample 13. The polarizing optical path 424 is used to analyze the polarization spectrum information of the reflected light generated by the sample 13, and the detector 425 is used to receive, measure and output the polarization spectrum information of the reflected light. A small electric motor can also be provided at the left / right epitaxial end to achieve precise control of the left / right epitaxial end and fine adjustment of the optical path angle. The second driving assembly 43 includes a translation component 431, a rotation component 432 and a lifting component 433 respectively connected to the fixed component 421. A workbench 434 is also arranged between the rotation component 432 and the translation component 431. The axis of the sample stage 2 and the fixed component 421 are the same. The processing and display component 44 includes a computer 441 and a display screen 442. The computer 441 is used to receive and process the signal output by the detector 425 and perform data analysis. The computer 441 also stores a semiconductor film growth uniformity recognition model, which can calculate the uniformity of the semiconductor film according to the spectral information of the test point. The display screen 442 is used to display the result.

[0070] In this embodiment, the light-transmitting material is quartz; the light source 423 is a white light source, which has a uniform spectral distribution and stable light intensity output, is suitable for general thin film thickness uniformity monitoring, and has a long life and low power consumption, and is easy to install and maintain; the detector 425 is a spectrometer, which can finely analyze optical characteristic data and complete the data acquisition process in a shorter time, thereby improving the efficiency and performance of the monitoring system. The spectrometer is connected to the processor via a data line to extract the signal.

[0071] During measurement, it is necessary to ensure that the angular velocity of the light source 423 and the test point is the same to ensure that they move synchronously, so that the angle difference between the light source 423 and the test point is constant, and the distance between the light source and the test point is constant. In order to improve the accuracy of the test, the influence of nonlinear disturbance is considered while ensuring the distance between the light source 423 and the test point on the surface of the sample 13 is constant:

[0072]

[0073] The three-dimensional coordinate system is constructed with the initial positions of the sample stage 2 and the fixing component 421, D(t) is the distance between the light source 423 and the test point on the surface of the sample 13; n is the number of tests, n = 1, 2, 3, ...; r n is the distance from the light source 423 to the origin; rn is the distance from the test point to the origin; t is the time; θ n (t) is the angle of the light source 423; θ n is the angle of the test point; H is the vertical distance between the light source 423 and the sample stage 2; α is the amplitude of the sinusoidal disturbance; β is the frequency of the sinusoidal disturbance; γ is the initial amplitude of the exponential decay; δ is the speed of the exponential decay; the corresponding parameters at time t can be measured in real time.

[0074] According to the real-time data of multiple measuring points input into the computer, the change of film doping concentration at each point over time can be output, and the growth of molecular beam epitaxial semiconductor film can be obtained by comparing the change curves of each measuring point.

[0075] S104, determining whether the molecular beam epitaxy semiconductor thin film growth can reach the expected value, and regulating the growth process so that the thin film growth is close to the expected value.

[0076] The operator can judge whether the real-time doping concentration uniformity of the semiconductor film reaches the expected value based on the real-time thin film doping concentration uniformity. If the growth process parameter thin film doping concentration uniformity reaches the expected value, the growth is continued until the process is completed; if the growth process parameter thin film doping concentration uniformity does not reach the expected value, it is necessary to adjust the process parameters, such as the temperature of the source furnace or the temperature of the sample stage and the rotation rate, so that the film growth is close to the expected value.

[0077] Example 2

[0078] The embodiment is basically the same as the first embodiment, except that the semiconductor film growth uniformity of the present embodiment is for the thickness of the semiconductor film. Compared with the doping concentration of the first embodiment, the thickness can be directly simulated by optical simulation software, which is simpler. Figure 5 As shown, other steps will not be repeated.

[0079] Example 3

[0080] like Figure 6 As shown, a molecular beam epitaxial semiconductor thin film growth uniformity in-situ monitoring system comprises:

[0081] A database module is used to obtain spectral information of semiconductor thin films prepared under different molecular beam epitaxial growth conditions and characterization data reflecting the uniformity of semiconductor thin film growth, and to construct a database;

[0082] A model building module, used to obtain a semiconductor thin film growth uniformity recognition model based on the database training;

[0083] The identification module is used to obtain real-time spectral information of the semiconductor film epitaxially grown on the sample surface, and obtain the growth status of the molecular beam epitaxial semiconductor film in combination with the semiconductor film growth uniformity identification model.

[0084] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for in-situ monitoring of the uniformity of semiconductor thin film growth by molecular beam epitaxy, characterized in that: include, Obtain spectral information of semiconductor films prepared under different molecular beam epitaxial growth conditions and characterization data reflecting the uniformity of semiconductor film growth, and build a database; A semiconductor thin film growth uniformity recognition model is obtained by training the database; Acquire real-time spectrum information of the semiconductor film epitaxially grown on the sample surface, and obtain the growth status of the molecular beam epitaxial semiconductor film by combining the semiconductor film growth uniformity identification model; The real-time spectrum information of the semiconductor thin film epitaxially grown on the sample surface is obtained according to a wide spectrum monitoring device arranged at the bottom of the molecular beam epitaxy equipment; The molecular beam epitaxy device comprises: A vacuum reaction chamber, provided with an installation port and having a bottom part or entirely made of a light-transmitting material, for providing a reaction environment, wherein the installation port is used to install a furnace source; A sample stage, arranged in the vacuum reaction chamber; A first driving assembly, disposed at the top of the outside of the vacuum reaction chamber and connected to the sample stage through a connecting rod, for driving the sample stage and adjusting the position of the sample stage; The wide spectrum monitoring device comprises a second driving component, a monitoring component and a processing display component arranged at the bottom end of the vacuum reaction chamber; When working, the monitoring component driven by the second driving component keeps synchronous movement with the sample on the sample stage driven by the first driving component to obtain the information of the thin film growth; The monitoring assembly includes a fixing component, a polarizing light path, a light source, a polarizing light path, and a detector; The fixing part includes a plurality of movable epitaxial ends above, the polarizing optical path and the light source and the analyzing optical path and the detector are installed on different epitaxial ends, the light beam generated by the light source generates a light beam with a specific polarization direction through the polarizing optical path to irradiate the sample surface, the analyzing optical path is used to analyze the polarization spectrum information of the reflected light generated by the sample, and the detector is used to receive, measure and output the polarization spectrum information of the reflected light; The second driving assembly comprises a translation component, a rotation component and a lifting component respectively connected to the fixed component, and a workbench is further arranged between the rotation component and the translation component; Wherein, the sample stage and the fixing component have the same axis; The processing and display component is used to receive and process the signal output by the detector, and perform data analysis and result display; During measurement, ensure that the angular velocity of the light source and the test point is the same to ensure that they move synchronously, so that the angle difference between the light source and the test point is constant, and the distance between the light source and the test point is constant. While ensuring that the distance between the light source and the test point on the sample surface is constant, consider the influence of nonlinear disturbance: ； ； Among them, a three-dimensional coordinate system is constructed with the initial positions of the sample stage and the fixed components. D ( t ) is the distance between the light source and the test point on the sample surface; n is the number of tests, n =1,2,3,…; r n is the distance from the light source to the origin; R n is the distance from the test point to the origin; t for the moment; θ n ( t ) is the angle of the light source; Θ n is the angle of the test point; H is the vertical distance between the light source and the sample stage; α is the amplitude of the sinusoidal disturbance; β is the frequency of the sinusoidal disturbance; γ The initial amplitude of the exponential decay; δ The speed of exponential decay; corresponding to t The time parameters can be measured in real time.

2. The method for in-situ monitoring of the uniformity of the growth of a molecular beam epitaxial semiconductor thin film according to claim 1, characterized in that: The characterization data of the semiconductor film growth uniformity include the thickness, doping concentration and morphology of the semiconductor film.

3. The method for in-situ monitoring of the uniformity of the growth of a molecular beam epitaxial semiconductor thin film according to claim 1, characterized in that: The semiconductor thin film growth uniformity recognition model obtained by training the database includes: An optical simulation model is established to perform simulation calculations, and the simulation results are extracted and compared with the characterization data of the uniformity of semiconductor thin film growth to establish related data, which is then cleaned, trained, and optimized using an artificial intelligence algorithm.

4. The method for in-situ monitoring of the uniformity of the growth of a molecular beam epitaxial semiconductor thin film according to claim 3, characterized in that: The cleaning includes missing value processing, noise processing and logic error cleaning as well as standardization preprocessing and dividing the data set and the test set; The training is based on a GRU neural network; The optimization is performed by adding the Adam optimization module.

5. The method for in-situ monitoring of the uniformity of the growth of a molecular beam epitaxial semiconductor thin film according to claim 1, characterized in that: The characterization data reflecting the uniformity of semiconductor thin film growth is obtained based on at least one of XRD, scanning electron microscopy, electron beam ion beam analysis, and atomic force microscopy.

6. The method for in-situ monitoring of the uniformity of semiconductor thin film growth by molecular beam epitaxy according to any one of claims 1 to 5, characterized in that: After obtaining the growth status of the molecular beam epitaxial semiconductor film, the growth process is also controlled to make the film growth close to the expected value.

Images