Crystal growth uniformity dynamic monitoring method based on multi-mode photoelectron imaging

Abstract

The invention discloses a crystal growth uniformity dynamic monitoring method based on multi-mode photoelectron imaging, and relates to the technical field of crystal monitoring, and the method comprises the steps: integrating a plurality of photoelectron imaging methods to monitor a crystal growth process, and obtaining multi-mode imaging data; based on a deep learning model, multi-modal imaging data is subjected to data fusion analysis, a uniformity index is calculated, a thermodynamic diagram is generated, a non-uniform region is identified, growth parameters of the non-uniform region are adjusted, a non-uniform growth rate is calculated, and a crystal sample which is judged to be unqualified is transferred, so that a plurality of photoelectron imaging methods are combined; the dynamic correlation analysis of the crystal surface chemical components and the crystal structure overcomes the limitation that the single-mode in-situ characterization technology cannot comprehensively reflect the co-evolution of the crystal surface chemical components and the crystal structure, and solves the problem that the dynamic co-evolution data of the crystal growth cannot be comprehensively reflected. The technical effect of comprehensively and dynamically monitoring the crystal growth uniformity is achieved.

Description

Technical Field

[0001] The present application relates to the field of crystal monitoring technology, and in particular to a method for dynamic monitoring of crystal growth uniformity based on multimodal photoelectron imaging. Background Art

[0002] As materials science research deepens, there is an urgent need for deeper research into the mechanisms of action, structural changes, and factors influencing performance. Current crystal growth techniques often rely on post-growth offline testing or in-situ characterization techniques to characterize crystal quality uniformity.

[0003] However, offline detection cannot capture the dynamic changes in the crystal growth process in real time. Although existing in situ characterization technology can provide some real-time data, the limitations of a single mode make it impossible to fully reflect the coordinated evolution of the chemical composition and structure of the crystal surface. Summary of the Invention

[0004] The present application provides a method for dynamic monitoring of crystal growth uniformity based on multimodal photoelectron imaging, which at least solves the problem in related technologies that the limitations of a single mode make it impossible to fully reflect the coordinated evolution of the chemical composition and structure of the crystal surface.

[0005] The present application provides a method for dynamic monitoring of crystal growth uniformity based on multimodal photoelectron imaging, comprising:

[0006] Integrate multiple photoelectron imaging methods to monitor the crystal growth process and obtain multimodal imaging data;

[0007] Perform data fusion analysis on multimodal imaging data based on deep learning models and calculate the uniformity index;

[0008] Generate a heat map based on the uniformity index, identify non-uniform areas based on the heat map, and adjust the growth parameters of the non-uniform areas;

[0009] Based on the non-uniform area after adjusting the growth parameters, the growth non-uniformity rate is calculated, and whether the crystal is qualified is judged according to the growth non-uniformity rate, and the unqualified crystal samples are transferred.

[0010] The present application also provides a device for dynamic monitoring of crystal growth uniformity based on multimodal photoelectron imaging, comprising:

[0011] Multimodal imaging module, used to integrate multiple photoelectron imaging methods to monitor the crystal growth process and obtain multimodal imaging data;

[0012] Data fusion and analysis module, used to perform data fusion analysis on multimodal imaging data based on deep learning models and calculate uniformity index;

[0013] a parameter adjustment module configured to generate a heat map according to the uniformity index, identify a non-uniformity region according to the heat map, and adjust a growth parameter of the non-uniformity region;

[0014] a crystal transfer module configured to calculate a growth non-uniformity rate based on the non-uniformity region after the growth parameter is adjusted, determine whether the crystal is qualified according to the growth non-uniformity rate, and transfer a non-qualified crystal sample.

[0015] The application also provides an electronic device, comprising a memory configured to store a computer program, and a processor configured to execute the computer program to implement the steps of any of the above crystal growth uniformity dynamic monitoring methods based on multi-modal photoelectron imaging.

[0016] The application also provides a computer-readable storage medium having a computer program stored therein, wherein the computer program is executed by a processor to implement the steps of any of the above crystal growth uniformity dynamic monitoring methods based on multi-modal photoelectron imaging.

[0017] The application also provides a computer program product comprising a computer program, wherein the computer program is executed by a processor to implement the steps of any of the above crystal growth uniformity dynamic monitoring methods based on multi-modal photoelectron imaging.

[0018] By the application, multiple photoelectron imaging methods are combined to monitor the crystal growth process, to realize dynamic correlation analysis of the chemical composition on the crystal surface and the crystal structure, to overcome the limitation that a single modal in-situ characterization technology cannot comprehensively reflect the synergistic evolution of the chemical composition on the crystal surface and the crystal structure, to perform data fusion and analysis on multi-modal imaging data, to dynamically adjust the growth parameter of the non-uniformity region, to reduce the growth non-uniformity rate through dynamic regulation and control, to transfer non-qualified crystal samples, to reduce the invalid growth time, and to save raw materials and energy consumption. Therefore, the technical effect of comprehensively reflecting the dynamic synergistic data of crystal growth to realize comprehensive dynamic monitoring of crystal growth uniformity can be achieved. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the application, the drawings needed in the embodiments will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0020] Figure 1 A flowchart of a crystal growth uniformity dynamic monitoring method based on multi-modal photoelectron imaging provided by an embodiment of the application is shown in the figure.

[0021] Figure 2A flowchart of a multi-modal data fusion algorithm provided by the embodiment of the present application is shown in the figure;

[0022] Figure 3 An electron optical path diagram of deep ultraviolet laser emission electron microscopy provided by the embodiment of the present application is shown in the figure;

[0023] Figure 4 A dynamic feedback mechanism and transfer execution diagram provided by the embodiment of the present application is shown in the figure;

[0024] Figure 5 A structural diagram of a crystal growth uniformity dynamic monitoring device based on multi-modal photoelectron imaging provided by the embodiment of the present application is shown in the figure;

[0025] Figure 6 A structural diagram of an electronic device provided by the embodiment of the present application is shown in the figure.

[0026] Reference signs

[0027] 1, ultraviolet light or deep ultraviolet laser; 2, sample; 3, objective lens; 4, transfer lens; 5, electron entrance slit; 6, micro-area diffraction diaphragm; 7, selected area diaphragm; 8, 90° deflection prism chamber; 9, projection lens; 10, detector. DETAILED DESCRIPTION

[0028] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the protection scope of the present application.

[0029] It should be noted that, in the description of the present application, the terms “include”, “contain” or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or device. The terms “first”, “second” and the like in the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence.

[0030] In order for those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0031] At present, the characterization methods commonly used for crystal growth include X-ray diffraction (XRD), Raman spectroscopy, infrared spectra analysis, scanning electron microscopy (SEM), atomic force microscopy (AFM), scanning tunneling microscopy (STM), transmission electron microscopy (TEM), and square resistance mapping.

[0032] Among them, X-ray diffraction technology usually tests the same position of the same material during the entire testing process, so it can only observe changes in the crystal structure of the material in a smaller area.

[0033] Compared to X-ray diffraction technology, Raman spectroscopy can be used to study amorphous or poorly crystalline materials, and can also be used to study reactions at interfaces (i.e., the boundary area between different materials).

[0034] Compared with other methods, the biggest advantage of infrared spectroscopy is that it can measure samples in any state, including liquids, solutions, viscous samples, powders, films, fibers, and gases, under the condition of selecting appropriate sample preparation methods.

[0035] Scanning electron microscopy can detect changes in nanomaterials in real time, while atomic force microscopy and scanning tunneling microscopy have the advantages of contact measurement and atomic-level resolution.

[0036] Transmission electron microscopy not only has the advantages of high spatial resolution and high energy resolution of ex situ transmission electron microscopy, but also introduces external excitations such as force, heat, electricity, magnetism, and chemical reactions inside the electron microscope. It can directly observe the microstructural evolution and characterization process of the sample under external excitation at the atomic level. By real-time observation and recording of the dynamic response process of the sample inside the electron microscope to different external excitation signals, relevant sample information can be obtained.

[0037] Sheet resistance mapping is used to determine the distribution of sheet resistance across the surface of a semiconductor material. This test method typically uses an electron microscope and a computer program to acquire and analyze the data. By selecting multiple points on the material surface and measuring the sheet resistance at each point, a sheet resistance mapping image can be generated to show the sheet resistance distribution across different areas of the material surface. Sheet resistance mapping can be used to study the effects of surface treatment, thin film deposition, doping, and other processes on semiconductor materials, as well as to evaluate the performance of semiconductor devices. It can also be used to characterize the uniformity of material surface quality.

[0038] However, traditional non-in situ research methods can no longer meet current needs. In situ characterization technology, which has been developed in recent years, can use different instruments and equipment to perform "online" analysis on specific reaction processes. As a technology that uses specific devices to continuously and synchronously analyze substances, it ultimately obtains a series of results with time or other related conditions as parameters, realizing the analysis of changes in the reaction process of substances. It has the characteristics of being dynamic, real-time, and intuitive. In addition, using in situ characterization methods, the process of chemical reactions, material structure, and morphology can be observed in real time, and information such as reaction intermediates can be obtained, which helps to analyze the reaction mechanism and promote the further development of chemistry and materials science.

[0039] In current crystal growth technologies (such as MOCVD, MBE, CVD, etc.), the characterization of crystal quality uniformity mostly relies on the offline detection method mentioned above after the growth is completed. However, offline detection cannot capture the dynamic changes in the growth process in real time, resulting in delayed adjustment and optimization of process parameters, affecting the consistency of crystal quality. Although existing in-situ characterization technologies (such as in-situ XRD and in-situ Raman) can provide some real-time data, the limitations of a single mode make it impossible to fully reflect the coordinated evolution of the chemical composition and structure of the crystal surface. In addition, the existing technology lacks a dynamic feedback mechanism and an automated quality control system based on real-time data, making it difficult to achieve real-time optimization of growth parameters and rapid elimination of unqualified samples.

[0040] In response to the above problems, the present invention provides a method for dynamic monitoring of crystal growth uniformity based on multimodal photoelectron imaging, which includes: integrating multiple photoelectron imaging methods to monitor the crystal growth process to obtain multimodal imaging data; performing data fusion analysis on the multimodal imaging data based on a deep learning model to calculate a uniformity index; generating a heat map based on the uniformity index, identifying non-uniform areas based on the heat map, and adjusting the growth parameters of the non-uniform areas; calculating the growth non-uniformity rate based on the non-uniformity area after adjusting the growth parameters, judging whether the crystal is qualified based on the growth non-uniformity rate, and transferring unqualified crystal samples. The method provided by the above scheme monitors the crystal growth process based on the integration of multiple electronic imaging methods, which solves the limitation of the single-modal in-situ characterization technology in the related art that cannot fully reflect the co-evolution of the chemical composition and crystal structure of the crystal surface. After obtaining the multimodal imaging data, the multimodal imaging data is fused and analyzed, the growth parameters of the non-uniform area are dynamically adjusted, the growth non-uniformity rate is dynamically controlled to reduce, and unqualified crystal samples are transferred, thereby reducing the ineffective growth time, saving raw materials and energy consumption, and achieving the technical effect of comprehensive dynamic monitoring of crystal growth uniformity.

[0041] In combination with the specific application environment architecture or specific hardware architecture on which the execution of the crystal growth uniformity dynamic monitoring method based on multimodal photoelectron imaging depends, the specific application environment architecture or specific hardware architecture is described herein.

[0042] The embodiments of the present application provide a method for dynamic monitoring of crystal growth uniformity based on multimodal photoelectron imaging. Figure 1 FIG. 1 is a flow chart of a method for dynamic monitoring of crystal growth uniformity based on multimodal photoelectron imaging according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:

[0043] Step S101 : Integrate multiple photoelectron imaging methods to monitor the crystal growth process and obtain multimodal imaging data.

[0044] In an embodiment of the present invention, a plurality of photoelectron imaging methods are combined to monitor the growth process of the crystal material in the crystal sample growth chamber, obtain photoelectron imaging data of the crystal surface, and obtain multimodal imaging data.

[0045] Step S102: performing data fusion analysis on the multimodal imaging data based on the deep learning model to calculate the uniformity index.

[0046] In an embodiment of the present invention, a multi-source data fusion algorithm is adopted to fuse and analyze the collected multimodal imaging data based on a deep learning model, and a uniformity index is calculated according to the data fusion and analysis results. The uniformity index is used to characterize the surface uniformity during the crystal growth process.

[0047] Step S103 : generating a heat map according to the uniformity index, identifying non-uniform regions according to the heat map, and adjusting growth parameters of the non-uniform regions.

[0048] Optionally, a heat map can be generated based on the uniformity index using an FPGA (Field Programmable Gate Array) accelerator. The FPGA accelerator processes data in milliseconds to generate dynamic heat maps. Furthermore, crystal growth trend prediction data can be generated to provide a basis for forward-looking analysis of crystal growth status.

[0049] Specifically, different colors in the thermal map represent uniform areas and non-uniform areas, respectively. Non-uniform areas are identified based on the colors in the thermal map, and a dynamic feedback control mechanism is used to adjust the growth parameters of the non-uniform areas (such as local temperature gradient optimization).

[0050] Specifically, the control system of the growth equipment (such as MOCVD) integrates a parameter control interface to adjust parameters such as temperature, gas pressure, and precursor flow rate. To improve the overall uniformity of the crystal, local growth conditions are independently controlled to address the differences in uniformity between different areas of the crystal surface (such as the edge and the center).

[0051] It can be understood that a heat map is generated based on the uniformity index to present the uniformity distribution of the crystal growth surface in a visual manner in real time, providing a data basis for dynamic feedback control.

[0052] Step S104 , calculating the growth non-uniformity rate based on the non-uniform area after adjusting the growth parameters, judging whether the crystal is qualified according to the growth non-uniformity rate, and transferring unqualified crystal samples.

[0053] In an embodiment of the present invention, after adjusting the growth parameters in the non-uniformity region, a growth non-uniformity rate (GNR) is defined based on the uniformity index. An automated elimination mechanism is used to determine whether a crystal is qualified based on the GNR. If a crystal is determined to be unqualified, the unqualified crystal sample is transferred to a waste bin.

[0054] The crystal growth uniformity dynamic monitoring method based on multi-modal photoelectron imaging provided in the embodiment of the application combines multiple photoelectron imaging methods to monitor the crystal growth process, realizes dynamic correlation analysis of the chemical composition and crystal structure of the crystal surface, overcomes the limitation that a single modal in-situ characterization technology cannot comprehensively reflect the synergistic evolution of the chemical composition and crystal structure of the crystal surface, performs data fusion and analysis on the multi-modal imaging data, dynamically adjusts the growth parameters of the non-uniform area, reduces the growth non-uniformity rate through dynamic regulation and control, transfers the unqualified crystal sample, reduces the invalid growth time, saves raw materials and energy consumption, and therefore the technical effect of comprehensively reflecting the dynamic synergistic data of the crystal growth and realizing comprehensive dynamic monitoring of the crystal growth uniformity can be achieved.

[0055] The embodiment of the application provides a crystal growth uniformity dynamic monitoring method based on multi-modal photoelectron imaging.

[0056] In step S201, multiple photoelectron imaging methods are integrated to monitor the crystal growth process, and multi-modal imaging data is obtained.

[0057] Specifically, the above step S201 includes the following steps.

[0058] In step S2011, an X-ray diffraction method is used to monitor the crystal structure evolution in real time, and XRD structure data is collected.

[0059] In step S2012, a Raman spectrum method is used to extract the crystal characteristic peak intensity.

[0060] In step S2013, a Raman spectrum method and an infrared spectrum method are combined to obtain the chemical composition and molecular bonding state of the crystal surface.

[0061] In step S2014, a deep ultraviolet laser is emitted to the crystal surface by using a deep ultraviolet laser photoemission electron microscope, and a photoemission electron map of the crystal surface is obtained.

[0062] In step S2015, a high-speed scanning electron microscope is used to scan the crystal surface topography.

[0063] In the embodiment of the application, for the crystal material grown in the MOCVD sample growth chamber, an in-situ XRD probe and a Raman spectrometer, a deep ultraviolet laser photoemission electron microscope, and a high-speed scanning electron microscope are integrated in the MOCVD reaction cavity. The X-ray diffraction method, the Raman spectrum method, the infrared spectrum method, the deep ultraviolet laser photoemission electron microscope, and the high-speed scanning electron microscope are combined to synchronously collect data.

[0064] X-ray diffraction method is used to detect the evolution of crystal structure in real time. Specifically, the X-ray diffraction technique is a characterization means for qualitatively analyzing the crystal type, crystal parameters, crystal defects of a material and quantitatively analyzing the relative content of different structure phases by using the position and intensity of diffraction peaks through the diffraction phenomenon of X-rays in the sample.

[0065] Raman spectroscopy method is used to extract the intensity of crystal characteristic peaks. Specifically, the Raman spectroscopy method relies on the scattering phenomenon of the incident light by the molecules of the sample to obtain relevant information of the sample. When the incident light interacts with the sample, a small part of the photons is inelastically scattered, i.e. Raman scattering, different from the frequency of the incident light, which provides information about the internal and intermolecular vibrations of the molecules, and further studies the composition or structural characteristics of the sample.

[0066] Raman spectroscopy method and infrared spectroscopy method are combined to synchronously obtain the chemical composition and molecular bonding state of the crystal surface. Specifically, the infrared spectroscopy technique is an effective means for studying and characterizing the molecular structure, and is used to determine the change law of a sample or a reaction system with time, temperature, pressure and environmental changes.

[0067] Since molecules can selectively absorb infrared light of certain wavelengths, causing the transition of vibration energy levels and rotation energy levels in the molecules, the absorption of infrared light can be detected, and the infrared absorption spectrum of the substance can be obtained, which is also called molecular vibration spectrum or vibration-rotation spectrum.

[0068] Deep ultraviolet laser photoemission electron microscopy is used to emit deep ultraviolet laser to the surface of the crystal to obtain the photoemission electron pattern of the surface of the crystal. Deep ultraviolet laser photoemission electron microscopy (DUV-PEEM) emits deep ultraviolet laser to the surface of the crystal sample to obtain the in-situ deep ultraviolet laser photoelectron morphology of the crystal sample.

[0069] Specifically, by irradiating the crystal surface with ultraviolet / deep ultraviolet laser light, the emission of photoelectrons is stimulated on the crystal surface, and microscopic imaging is performed through an electron optical system. The ultraviolet / deep ultraviolet light emission electron imaging used in the embodiment of the present invention belongs to electron projection microscopy imaging. There is no scanning link during imaging, which is suitable for real-time dynamic imaging observation and analysis of the surface of crystal samples. When light is incident on the surface of the thermal diffusion sample, the photoemission electrons from the surface of the crystal sample are first collected by the objective lens, and after passing through different apertures, slits and a series of electromagnetic lens components, they are finally received and collected by the detector. By scanning the crystal surface with a deep ultraviolet laser light emission electron microscope, a light emission electron map of the surface of the crystal material can be obtained. The grayscale value of the light emission electron imaging is mainly determined by the difference in the local work function of the surface micro-region. The distribution of the grayscale value of the light emission electron imaging on the sample surface can realize the in-situ characterization of the uniformity of the growth surface of the crystal material.

[0070] The crystal surface topography is scanned using a high-speed scanning electron microscope. Specifically, a high-speed scanning electron microscope scans the sample surface at high speed using an electron beam, collecting secondary electrons or backscattered electron signals in real time to form a crystal surface topography with nanometer-level resolution.

[0071] The embodiment of the present application provides a dynamic monitoring method for crystal growth uniformity based on multimodal photoelectron imaging, which integrates multiple detection methods such as X-ray diffraction, Raman spectroscopy, infrared spectroscopy, deep ultraviolet laser light emission electron microscopy, and high-speed scanning electron microscopy to monitor the chemical composition and crystal structure of the crystal surface.

[0072] In some optional embodiments, after integrating multiple photoelectron imaging methods to monitor the crystal growth process and obtain multimodal imaging data, the method further includes:

[0073] Step S2016: performing a noise removal process on the XRD structure data by wavelet transform and performing a normalization process.

[0074] Step S2017: Using filtering and smoothing technology, the intensity of the crystal characteristic peak is converted into Raman chemical adsorption amount.

[0075] Step S2018: Divide the light-emission electron spectrum into micro-regions using an image segmentation algorithm, and calculate the work function of each region.

[0076] In the embodiment of the present invention, after the multimodal imaging data is acquired, data preprocessing is performed on the multimodal imaging data.

[0077] Specifically, if Figure 2As shown, after collecting the multimodal imaging data, the multimodal imaging data is preprocessed. There are noises in the XRD structural data caused by X-ray source fluctuations, detector thermal noise and crystal sample vibration. The lattice constant (such as the c-axis length of GaN) in the structural data collected in real time by XRD is de-noised by wavelet transformation. Specifically, the XRD structural data is first decomposed into low-frequency components and high-frequency components, and the high-frequency components are threshold-processed using a soft threshold function to suppress noise. The denoised high-frequency components and low-frequency components are merged to obtain the XRD structural data after denoising. After denoising, normalization is performed to the range of [0, 1].

[0078] The intensity of the crystal characteristic peak extracted by the Raman spectrometer (such as the E2 peak of GaN) is smoothed using Savitzky-Golay filtering to convert the crystal characteristic peak intensity into Raman chemical adsorption amount (such as NH3 coverage). Specifically, a sliding window scan is performed on the characteristic peak region of the Raman spectrum, and a second-order polynomial is used to fit the spectral curve within each window to replace the original data points and suppress fluorescence background and random noise. By finding the mapping relationship between the crystal characteristic peak intensity and the Raman chemical adsorption amount, the Raman chemical adsorption amount corresponding to the crystal characteristic peak intensity is determined.

[0079] The photoemission electron spectrum collected using a deep ultraviolet laser photoemission electron microscope is divided into microregions using an image segmentation algorithm, and the work function of each region is calculated. The image segmentation algorithm used is the U-Net algorithm, which is used for example only and not as a limitation. Specifically, the U-Net algorithm is used to divide the photoemission electron spectrum into microregions with different work functions, and the work function of each region is calculated.

[0080] The grown sample is imaged microscopically using a deep ultraviolet laser. The imaging process does not involve a scanning process and is more suitable for real-time dynamic imaging observation and analysis of the surface. The obtained photoemission electron spectrum of the crystal material surface can be used to determine the distribution of grayscale values ​​of the photoemission electron imaging of the sample surface. The distribution of grayscale values ​​reflects the differences in local work functions in the surface micro-regions. A crystal material with a uniform grayscale value has a higher uniformity of crystal quality. Deep ultraviolet photoelectron spectroscopy can be used to study the chemical adsorption, molecular structure, chemical reaction and other properties of the material surface. It can be used to monitor chemical changes on the material surface, such as oxidation, reduction, adsorption and other processes. It can also be used to study the physical properties of the material surface, such as surface state and interface state.

[0081] In addition, the advantages of using deep ultraviolet laser light source as laser source are: (1) The energy of deep ultraviolet photons is relatively high, reaching 7eV, which can produce photoelectron excitation for a wider range of materials, especially suitable for the characterization of wide bandgap and ultra-wide bandgap materials such as gallium nitride, silicon carbide, diamond, aluminum nitride, and gallium oxide; (2) The difference between the theoretical value of the spatial resolution of photoemission electron imaging and the energy distribution of photoemission electrons isΔ E is related, and the energy distribution difference of the light-emitted electrons excited by high photon energy is Δ E is larger, and using deep ultraviolet laser as the excitation light source can achieve better spatial resolution; (3) Deep ultraviolet laser is monochromatic and has high energy resolution, while the ultraviolet light produced by the mercury lamp is a mixed beam of multiple wavelengths. During imaging, the electrons excited by the single wavelength deep ultraviolet laser have fewer stray signals, which is beneficial to improving imaging quality and spatial resolution.

[0082] Photoemission electron microscopy (PEEM) testing uses ultraviolet light to illuminate the sample, obtaining an image of UV-emitted photoelectrons from the sample surface. This image is primarily used to qualitatively analyze differences in the surface work function after activation. This method is primarily suitable for measuring surface electron energy levels, determining the energy difference between surface electrons and the vacuum energy level. The UV light source used is a 100W mercury arc lamp, which emits photons with a maximum energy of 4.9eV. Using this lamp as an excitation source can excite photoelectrons on the surface of most materials.

[0083] like Figure 3 As shown, Figure 3 This is a partial electron optical path diagram of a deep ultraviolet laser light emission electron microscope. The specific microscopic imaging process is as follows:

[0084] (1) Ultraviolet light or deep ultraviolet laser 1 is incident on the surface of sample 2 at a certain angle, and electrons are emitted from the surface of sample 2;

[0085] (2) There is an accelerating electric field between the objective lens 3 and the sample 2, and the light / thermal electrons are accelerated to 15 keV by the electric field;

[0086] (3) The electrons are collected by the objective lens 3 and form a first image on the transfer lens 4;

[0087] (4) After being transmitted by the transmission lens 4, the electron beam passes through the electron incident slit 5 and is deflected 90° by the 90° deflection prism chamber 8 (the 90° deflection prism chamber 8 realizes electron beam separation and electron beam deflection);

[0088] (5) It then enters the projection lens 9 and is finally imaged on the detector 10.

[0089] In the electron optical system, the projection lens 9 includes different lenses and performs different functions. These include acquiring a sample image from the prism array and performing preliminary image magnification; converting between real space and k-space; adjusting the image magnification; and obtaining upright or inverted images in different states.

[0090] To optimize microscopic imaging, the system features various apertures. The selection aperture 7 is required for photoelectron imaging. This aperture features apertures of varying sizes, each corresponding to a specific microregion on the sample surface. During imaging, the system adjusts the selection aperture 7 to select the appropriate microregion, limiting the entry of electrons outside the microregion into the projection lens system, effectively improving image quality.

[0091] The work function of a semiconductor refers to the difference between the energy E0 of a stationary electron in a vacuum at a thermodynamic temperature of zero degrees and the energy of the semiconductor's Fermi level. For metals, it represents the minimum energy required for an electron with an initial energy equal to the Fermi level to escape from the inside of the metal into a vacuum. The size of the work function indicates the strength of the electron's binding in the metal (semiconductor). The larger the work function, the less likely the electron is to leave the metal (semiconductor). The factors that affect the work function of a semiconductor are mainly the Fermi level, and the Fermi level is related to the semiconductor temperature, impurity concentration and impurity type. In addition, the size of the work function is also related to the surface condition of the semiconductor. The surface roughness, defects and adsorbed substances will affect the size of the work function. These influences also have an important impact on the performance of semiconductor devices. In the present invention, we characterize the uniformity of the growth of crystalline materials by obtaining the work function distribution on the surface of the crystalline material.

[0092] Deep ultraviolet laser is used to irradiate the sample surface to obtain the in-situ deep ultraviolet laser photoelectron morphology of the sample. This electron image further determines the position of the photoelectron emission area on the sample surface. The deep ultraviolet laser light source is a 177.3nm all-solid-state laser source, which mainly converts 355nm laser into 177.3nm deep ultraviolet laser through direct frequency doubling technology. When this new all-solid-state laser source is used as a microscope excitation source, the output power is usually adjusted to 0.5-0.55mW.

[0093] Step S202: Perform data fusion analysis on the multimodal imaging data based on the deep learning model and calculate the uniformity index.

[0094] In some optional implementations, the above step S202 includes:

[0095] In step S2021 , the XRD structure data, Raman chemical adsorption amount, and photoemission electron spectrum are mapped to the same spatial grid by affine transformation using the crystal surface topography as a reference coordinate system for spatial alignment.

[0096] Step S2022: Use historical growth data to train the lightweight convolutional neural network model.

[0097] In step S2023, the XRD structure data, Raman chemical adsorption amount, and work function are used as inputs, and the trained lightweight convolutional neural network model is used to output the uniformity index.

[0098] In the embodiment of the present invention, spatial alignment is performed first, and then multimodal fusion is performed.

[0099] Spatial alignment: Using the SEM morphology image as the reference coordinate system, the XRD structural data, Raman chemical adsorption amount, and PEEM photoemission electron spectrum were mapped to the same spatial grid (resolution of 1 μm × 1 μm) through affine transformation.

[0100] Specifically, a scanning electron microscope (SEM) generates a high-resolution topography image by scanning the sample surface with an electron beam. Its pixel coordinates directly correspond to the physical spatial position (XY axis) of the sample surface, which is selected as the reference coordinate system. Each pixel in the SEM topography image corresponds to the actual position on the sample surface. XRD structural data, Raman chemisorption data, and PEEM photoemission electron spectra are uniformly mapped to the reference coordinate system through a transformation matrix.

[0101] Multimodal fusion: 500 sets of historical growth data were collected to train a lightweight convolutional neural network model. Using a test set, the model achieved an accuracy rate exceeding 92%. This lightweight convolutional neural network replaces standard convolution with depthwise separable convolution, reducing the number of parameters and using fixed-point operations instead of floating-point operations, adapting to the millisecond-level real-time processing requirements of FPGA accelerators.

[0102] A lightweight convolutional neural network model (3 layers of convolution + 2 layers of full connection) is used, with the input being: XRD structural data (Sxrd), Raman chemical adsorption (Craman), work function (Dpeem), and the output being the uniformity index HI.

[0103] Specifically, the uniformity index HI is calculated according to the following formula:

[0104] HI=0.4×Sxrd+0.3×Craman+0.3×(1-Dpeem)

[0105] Step S203 : generating a heat map according to the uniformity index, identifying non-uniform regions according to the heat map, and adjusting growth parameters of the non-uniform regions.

[0106] In some optional implementations, generating a heat map according to the uniformity index in step S203, and identifying the non-uniform area according to the heat map includes:

[0107] Step S2031 : Map the uniformity index into a pseudo-color image according to the grid to generate a heat map.

[0108] Step S2032 : defining an area in the heat map whose uniformity index is less than a preset uniformity index threshold as a non-uniform area.

[0109] In the embodiment of the present invention, the uniformity index HI is mapped into a pseudo-color image according to the grid to obtain a heat map, in which red represents low uniformity and green represents high uniformity.

[0110] Specifically, the uniformity index (HI) output by the lightweight convolutional neural network model is divided into a 1μm×1μm spatial grid. Each grid stores an HI value, corresponding to the growth uniformity of a 1μm×1μm region on the sample surface. The HI values ​​are linearly mapped to the interval [0, 255], adapted to the pixel value range of the pseudo-color image. A "hot-to-cold" color gradient is used to map the HI values ​​based on the mapping relationship between the HI values ​​and the color mapping table.

[0111] Optionally, to avoid the mesh boundary pitch effect, bilinear interpolation is used for the uniformity index HI values ​​of adjacent meshes to smooth the color transition and improve the visualization accuracy.

[0112] The area in the thermal map where the uniformity index is less than a preset uniformity index threshold (in this embodiment, 0.6 is the preset uniformity index threshold, which is an empirical threshold corresponding to a lattice deviation greater than 1.5%) is defined as a non-uniform area.

[0113] The embodiment of the present application provides a dynamic monitoring method for crystal growth uniformity based on multimodal photoelectron imaging, which converts an abstract uniformity index into a visual heat map, intuitively presents the uniformity distribution of the crystal surface with a color gradient, quickly identifies non-uniform areas, and realizes an intuitive characterization of the uniformity of crystal growth.

[0114] Step S204 , calculating the growth non-uniformity rate based on the non-uniform area after adjusting the growth parameters, judging whether the crystal is qualified according to the growth non-uniformity rate, and transferring unqualified crystal samples.

[0115] In some optional embodiments, judging whether the crystal is qualified according to the growth non-uniformity rate and transferring the unqualified crystal sample in the above step S204 includes:

[0116] Step S2041 : sampling the non-uniform region after adjusting the growth parameters three times in a row.

[0117] Step S2042: If the growth non-uniformity rate in three consecutive samplings is greater than the preset growth non-uniformity rate threshold and the standard deviation is less than the preset standard deviation threshold, the crystal is judged as an unqualified crystal sample. The growth non-uniformity rate is the ratio of the area of ​​the non-uniform region to the total area in the thermal map.

[0118] Step S2043 , generating a transfer instruction for unqualified crystal samples, transferring unqualified crystal samples to a waste bin, and re-adjusting the growth parameters of the non-uniform area.

[0119] In the embodiment of the present invention, after adjusting the growth parameters of the non-uniform region, three consecutive samplings (at intervals of 30 seconds) are performed.

[0120] The growth non-uniformity rate GNR is calculated according to the following formula:

[0121]

[0122] It should be noted that the heat map and growth non-uniformity rate (GNR) value are refreshed every 10 seconds.

[0123] After the data is fused to obtain the growth non-uniformity rate GNR value, if the growth non-uniformity rate GNR in three consecutive samplings is greater than the preset growth non-uniformity rate threshold (20% is the preset growth non-uniformity rate threshold in this embodiment), and to eliminate accidental errors, the standard deviation is less than the preset standard deviation threshold (5% is the preset standard deviation threshold in this embodiment), then the crystal is judged to be an unqualified crystal sample.

[0124] like Figure 4 As shown in the figure, if the growth non-uniformity ratio (GNR) exceeds 20% in three consecutive samplings, a robotic arm is used to pick up unqualified crystal samples. The FPGA control board sends a CAN bus command to the robotic arm controller, and the robotic arm is raised along the Z axis from the standby area on the side of the reaction chamber to the same height as the sample stage. Then, it moves horizontally to the sample stage and uses vacuum adsorption to pick up unqualified crystal samples, moving them to the waste bin (taking 8 seconds), and an alarm is triggered. At the same time, a pressure sensor is used to ensure that the adsorption force is greater than 50N.

[0125] The embodiment of the present application provides a dynamic monitoring method for crystal growth uniformity based on multimodal photoelectron imaging. The method determines whether a crystal sample is qualified through the growth unevenness rate and standard deviation obtained by continuous sampling, eliminates possible accidental errors in single sampling, reduces the misjudgment rate, transfers unqualified crystal samples, reduces invalid growth time, reduces the amount of data, and reduces data monitoring costs.

[0126] In some optional embodiments, before transferring the unqualified crystal sample to the waste bin in step S204, the method includes:

[0127] Step S2043a, scanning obstacles in the transfer path of unqualified crystal samples by laser radar, performing collision detection, and generating a transfer path for unqualified crystal samples.

[0128] To prevent samples from falling, this embodiment uses a laser radar to scan obstacles along the transfer path for unqualified crystal samples, detect collisions, and generate a transfer path for the unqualified crystal samples. The unqualified crystal samples are then transferred according to the generated transfer path. Furthermore, the emergency braking response time is ensured to be less than 0.1 seconds.

[0129] To obtain more accurate information about the sample transfer path, an infrared Time of Flight (ToF) camera and force sensor were added to form a multi-sensor array. The LiDAR provides point cloud data of the sample transfer path, the infrared ToF camera identifies transparent and translucent objects and captures images of the sample transfer path, and the force sensor detects minor collisions. Kalman filtering is used to fuse multi-source data to detect obstacles such as metal brackets, improving obstacle detection capabilities and reducing the sample collision rate during the transfer of unqualified samples.

[0130] In some optional embodiments, such as Figure 4 As shown, after the unqualified crystal samples are transferred to the waste bin in step S204, the reaction chamber is cleaned. The specific cleaning steps include:

[0131] In step a1, HCl gas (200 sccm flow rate, 600° C. temperature, 5 minutes) is introduced.

[0132] Step a2: high-purity nitrogen purging (pressure 10 kPa, duration 3 minutes).

[0133] After cleaning, the MOCVD temperature control system was restored to the initial settings, that is, the temperature was 1050°C and the TMGa (TriMethylGallium) flow rate was 100 μmol / min.

[0134] After cleaning the reaction chamber, the parameters are reset and the growth is restarted.

[0135] It should be noted that the above-mentioned dynamic monitoring method of crystal growth uniformity based on multimodal photoelectron imaging is applicable to the growth monitoring of wide bandgap semiconductors (SiC, Ga, O3), two-dimensional materials (graphene) and superlattice structures.

[0136] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus the necessary general hardware platform, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method.

[0137] The embodiment of the present application also provides a device for dynamic monitoring of crystal growth uniformity based on multimodal photoelectron imaging, such as Figure 5 As shown, the device includes:

[0138] The multimodal imaging module 501 is used to integrate multiple photoelectron imaging methods to monitor the crystal growth process and obtain multimodal imaging data.

[0139] The data fusion and analysis module 502 is used to perform data fusion analysis on the multimodal imaging data based on the deep learning model and calculate the uniformity index.

[0140] The parameter adjustment module 503 is used to generate a heat map according to the uniformity index, identify non-uniform areas according to the heat map, and adjust the growth parameters of the non-uniform areas.

[0141] The crystal transfer module 504 is used to calculate the growth non-uniformity rate based on the non-uniform area after the growth parameters are adjusted, determine whether the crystal is qualified according to the growth non-uniformity rate, and transfer unqualified crystal samples.

[0142] In some optional embodiments, the multimodal imaging module 501 includes:

[0143] The acquisition unit is used to monitor the crystal structure evolution in real time using the X-ray diffraction method and collect XRD structure data.

[0144] The peak extraction unit is used to extract the intensity of the crystal characteristic peak using the Raman spectroscopy method.

[0145] The acquisition unit is used to obtain the chemical composition of the crystal surface and the molecular bonding state by combining Raman spectroscopy and infrared spectroscopy.

[0146] The laser emission unit is used to emit deep ultraviolet laser to the crystal surface using a deep ultraviolet laser light emission electron microscope to obtain a light emission electron spectrum of the crystal surface.

[0147] The scanning unit is used to scan the crystal surface topography using a high-speed scanning electron microscope.

[0148] In some optional embodiments, the device further comprises:

[0149] The noise processing module is used to remove noise from XRD structural data through wavelet transformation and perform normalization processing.

[0150] The conversion module is used to convert the crystal characteristic peak intensity into Raman chemical adsorption amount by using filtering and smoothing technology.

[0151] The work function calculation module is used to divide the light emission electron spectrum into micro-regions through an image segmentation algorithm and calculate the work function of each region.

[0152] In some optional implementations, the data fusion and analysis module 502 includes:

[0153] The mapping unit is used to use the crystal surface morphology map as the reference coordinate system and map the XRD structure data, Raman chemical adsorption amount, and light emission electron spectrum to the same spatial grid through affine transformation for spatial alignment.

[0154] The model training unit is used to train the lightweight convolutional neural network model using historical growth data.

[0155] The output unit is used to take XRD structural data, Raman chemical adsorption amount, and work function as input, and use the trained lightweight convolutional neural network model to output the uniformity index.

[0156] In some optional implementations, the parameter adjustment module 503 includes:

[0157] The heat map generation unit is used to map the uniformity index into a pseudo-color image according to the grid to generate a heat map.

[0158] The non-uniformity region definition unit is used to define a region in the thermal map whose uniformity index is less than a preset uniformity index threshold as a non-uniformity region.

[0159] In some optional embodiments, the crystal transfer module 504 includes:

[0160] The sampling unit is used to continuously sample the non-uniform area three times after the growth parameters are adjusted.

[0161] The judgment unit is used to judge the crystal as an unqualified crystal sample if the growth unevenness rate in three consecutive samplings is greater than a preset growth unevenness rate threshold and the standard deviation is less than the preset standard deviation threshold. The growth unevenness rate is the ratio of the area of ​​the non-uniform region in the thermal map to the total area.

[0162] The transfer execution unit is used to generate transfer instructions for unqualified crystal samples, transfer unqualified crystal samples to a waste bin, and readjust the growth parameters of the non-uniform area.

[0163] In some optional embodiments, the device further comprises:

[0164] The collision detection module is used to scan obstacles in the transfer path of unqualified crystal samples through laser radar, perform collision detection, and generate a transfer path for unqualified crystal samples.

[0165] For the description of the features in the embodiment corresponding to the device for dynamic monitoring of crystal growth uniformity based on multimodal photoelectron imaging, please refer to the relevant description of the embodiment corresponding to the method for dynamic monitoring of crystal growth uniformity based on multimodal photoelectron imaging, which will not be repeated here.

[0166] The embodiment of the present application also provides an electronic device, such as Figure 6 As shown, it includes a memory 610 and a processor 620, wherein the memory 610 stores a computer program, and the processor 620 is configured to run the computer program to execute the steps in any of the above-mentioned embodiments of the method for dynamic monitoring of crystal growth uniformity based on multimodal photoelectron imaging.

[0167] An embodiment of the present application further provides a computer-readable storage medium, in which a computer program is stored, wherein the computer program is configured to execute the steps of any one of the above-mentioned XX method embodiments when run.

[0168] In an exemplary embodiment, the computer-readable storage medium may include, but is not limited to, various media that can store computer programs, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk, or an optical disk.

[0169] An embodiment of the present application further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements the steps in any of the above-mentioned embodiments of the method for dynamic monitoring of crystal growth uniformity based on multimodal photoelectron imaging.

[0170] An embodiment of the present application also provides another computer program product, including a non-volatile computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the steps in any of the above-mentioned embodiments of the method for dynamic monitoring of crystal growth uniformity based on multimodal photoelectron imaging.

[0171] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the above description has generally described the components and steps of each example according to their functions. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0172] The above is a detailed introduction to a method for dynamic monitoring of crystal growth uniformity based on multimodal photoelectron imaging provided by the present application. Specific examples are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea. It should be pointed out that for ordinary technicians in this technical field, without departing from the principles of the present application, several improvements and modifications can be made to the present application, and these improvements and modifications also fall within the scope of protection of the claims of the present application.

Claims

1. A method for dynamic monitoring of crystal growth uniformity based on multimodal photoelectron imaging, characterized in that: The method comprises: Integrate multiple photoelectron imaging methods to monitor the crystal growth process and obtain multimodal imaging data; Performing data fusion analysis on the multimodal imaging data based on a deep learning model to calculate a uniformity index; generating a heat map based on the uniformity index, identifying non-uniform regions based on the heat map, and adjusting growth parameters of the non-uniform regions; Based on the non-uniform area after the growth parameters are adjusted, the growth non-uniformity rate is calculated, and whether the crystal is qualified is judged according to the growth non-uniformity rate, and the unqualified crystal samples are transferred.

2. The method for dynamic monitoring of crystal growth uniformity based on multimodal photoelectron imaging according to claim 1, characterized in that: The method integrates multiple photoelectron imaging methods to monitor the crystal growth process and obtain multimodal imaging data, including: Use X-ray diffraction to monitor the evolution of crystal structure in real time and collect XRD structural data; The intensity of crystal characteristic peaks was extracted using Raman spectroscopy; Combine Raman spectroscopy and infrared spectroscopy to obtain the chemical composition of the crystal surface and the molecular bonding state; A deep ultraviolet laser light emission electron microscope is used to emit deep ultraviolet laser light onto the crystal surface to obtain a light emission electron spectrum of the crystal surface; The crystal surface morphology was scanned using a high-speed scanning electron microscope.

3. The method for dynamic monitoring of crystal growth uniformity based on multimodal photoelectron imaging according to claim 2, characterized in that: After integrating multiple photoelectron imaging methods to monitor the crystal growth process and obtain multimodal imaging data, the method further includes: The XRD structural data is subjected to a noise removal process and a normalization process by wavelet transformation; Using filtering and smoothing technology to convert the crystal characteristic peak intensity into Raman chemical adsorption amount; The light emission electron spectrum is divided into micro-regions by an image segmentation algorithm, and the work function of each region is calculated.

4. The method for dynamic monitoring of crystal growth uniformity based on multimodal photoelectron imaging according to claim 1, characterized in that: The performing data fusion analysis on the multimodal imaging data based on the deep learning model and calculating the uniformity index includes: Using the crystal surface topography as a reference coordinate system, the XRD structural data, Raman chemical adsorption amount, and photoemission electron spectrum are mapped to the same spatial grid through affine transformation for spatial alignment; Use historical growth data to train lightweight convolutional neural network models; Taking XRD structural data, Raman chemical adsorption amount and work function as input, the trained lightweight convolutional neural network model is used to output the uniformity index.

5. The method for dynamic monitoring of crystal growth uniformity based on multimodal photoelectron imaging according to claim 1, characterized in that: Generating a heat map according to the uniformity index and identifying a non-uniform area according to the heat map includes: Mapping the uniformity index into a pseudo-color image according to a grid to generate a heat map; The area in the heat map whose uniformity index is less than the preset uniformity index threshold is defined as a non-uniform area.

6. The method for dynamic monitoring of crystal growth uniformity based on multimodal photoelectron imaging according to claim 1, characterized in that: The step of judging whether the crystal is qualified according to the growth unevenness rate and transferring unqualified crystal samples comprises: The heterogeneous area after adjusting the growth parameters was sampled three times continuously; If the growth non-uniformity rate is greater than the preset growth non-uniformity rate threshold and the standard deviation is less than the preset standard deviation threshold in three consecutive samplings, the crystal is judged as an unqualified crystal sample. The growth non-uniformity rate is the ratio of the non-uniform area to the total area in the thermal map. Generate a transfer instruction for unqualified crystal samples, transfer the unqualified crystal samples to a waste bin, and readjust the growth parameters of the non-uniform area.

7. The method for dynamic monitoring of crystal growth uniformity based on multimodal photoelectron imaging according to claim 6, characterized in that: Before transferring the unqualified crystal sample to a waste bin, the method further comprises: Obstacles in the transfer path of unqualified crystal samples are scanned by laser radar, collision detection is performed, and a transfer path for unqualified crystal samples is generated.

8. A device for dynamic monitoring of crystal growth uniformity based on multimodal photoelectron imaging, characterized in that: include: Multimodal imaging module, used to integrate multiple photoelectron imaging methods to monitor the crystal growth process and obtain multimodal imaging data; A data fusion and analysis module, configured to perform data fusion analysis on the multimodal imaging data based on a deep learning model and calculate a uniformity index; a parameter adjustment module, configured to generate a heat map according to the uniformity index, identify non-uniform regions according to the heat map, and adjust growth parameters of the non-uniform regions; The crystal transfer module is used to calculate the growth non-uniformity rate based on the non-uniform area after adjusting the growth parameters, judge whether the crystal is qualified according to the growth non-uniformity rate, and transfer unqualified crystal samples.

9. An electronic device, characterized in that: include: memory for storing computer programs; A processor, configured to implement the steps of the method for dynamic monitoring of crystal growth uniformity based on multimodal photoelectron imaging as claimed in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, the steps of the method for dynamic monitoring of crystal growth uniformity based on multimodal photoelectron imaging according to any one of claims 1 to 7 are implemented.

Images