Intelligent analysis system and method for crystal structure of electron microscopic images

Abstract

The present invention discloses an electron microscopic image crystal structure intelligent analysis system and method, which includes a data import module for importing electron microscopic images and crystal structure data; an image preprocessing module for preprocessing the imported electron microscopic images to obtain an experimental spectrum; a diffraction simulation module for simulating two-dimensional electron diffraction of a crystal based on the imported crystal structure data to obtain a diffraction simulation result, and processing the diffraction simulation result to obtain a theoretical spectrum; an intelligent matching module for matching the experimental spectrum with the theoretical spectrum and outputting the matching result; and a result output module for outputting the matching result and the diffraction simulation result to realize crystal structure visualization. Under the synergistic effect of each module, the electron microscopic image crystal structure can be automatically analyzed in the entire process, greatly improving the efficiency of crystal structure analysis; and under the action of the intelligent matching module, rapid matching can be achieved and the error rate can be reduced.

Description

Technical Field

[0001] The present invention relates to crystal structure analysis technology, and in particular to an intelligent analysis system and method for electron microscopic image crystal structures. Background Art

[0002] Traditional crystal structure analysis relies on manual annotation, simulation, and comparison. Complex systems (such as high-entropy alloys and perovskites) can take days to analyze, resulting in low efficiency. Complex systems have multiple structural possibilities, and manual analysis can easily overlook certain structures and orientations, leading to misjudgments.

[0003] Diffraction simulation relies on closed-source tools such as SingleCrystal, which cannot process data in batches.

[0004] Existing intelligent methods, such as MIT's AutoXRD, only support one-dimensional XRD (X-ray diffraction) spectrum analysis and are unable to process two-dimensional diffraction patterns or real-space TEM (transmission electron microscopy) imaging. They also require manual intervention to build a simulation database and rely on commercial software, making fully automated analysis impossible. Summary of the Invention

[0005] The purpose of the present invention is to overcome the above-mentioned deficiencies of the prior art and provide an electron microscopy image crystal structure intelligent analysis system and method thereof to improve analysis efficiency and accuracy.

[0006] To achieve the above object, the technical solution of the present invention is:

[0007] In a first aspect, the present invention provides an intelligent analysis system for electron microscopic image crystal structures, comprising:

[0008] Data import module, used to import electron microscopy images and crystal structure data;

[0009] An image preprocessing module is used to preprocess the imported electron microscopy images to obtain experimental maps;

[0010] A diffraction simulation module simulates two-dimensional electron diffraction of the crystal based on the imported crystal structure data to obtain a diffraction simulation result, and processes the diffraction simulation result to obtain a theoretical spectrum;

[0011] An intelligent matching module, used to match the experimental spectrum with the theoretical spectrum and output the matching result;

[0012] The result output module is used to display the output matching results and diffraction simulation results to achieve crystal structure visualization.

[0013] Optionally, the diffraction simulation module includes:

[0014] Crystal structure input interface, used to import crystal structure data;

[0015] Thickness Adaptive Optimizer, used to adjust crystal sample thickness when considering multi-beam interference effects;

[0016] The Bloch wave dynamics simulation submodule dynamically calculates the electron diffraction intensity distribution based on the imported crystal structure data based on the electron wave propagation equation in the crystal, takes into account the multi-beam interference effect, and obtains the diffraction simulation results based on the crystal sample thickness adjusted by the thickness adaptive optimizer; the diffraction simulation results are processed to obtain the theoretical spectrum.

[0017] Optionally, the Bloch wave dynamics simulation submodule includes:

[0018] The structure factor calculation unit calculates the structure factor of each crystal plane based on atomic scattering parameters and crystal structure data;

[0019] The excitation error modeling unit is used to combine the electron wavelength and the crystal sample thickness to derive the excitation error of each crystal plane;

[0020] The intensity simulation unit calculates the electron diffraction intensity distribution based on the excitation error of each crystal plane of the crystal to obtain a diffraction simulation result; and processes the diffraction simulation result to obtain a theoretical spectrum.

[0021] Optionally, the image preprocessing module is used to preprocess the imported electron microscopy image to obtain an experimental spectrum, including: performing fast Fourier transform on the imported electron microscopy image, fitting the amorphous background, dynamically subtracting amorphous scattering noise, and unifying the image resolution to obtain the experimental spectrum.

[0022] Optionally, the thickness adaptive optimizer adopts multi-threaded parallel computing and supports FFT / diffraction mode switching.

[0023] Optionally, the intelligent matching module includes:

[0024] Polar coordinate transformation submodule, used to transform experimental and theoretical maps from Cartesian coordinate system to polar coordinate system;

[0025] The cyclic cross-correlation submodule is used to calculate the fast Fourier cross-correlation between the experimental and theoretical images in polar coordinates to determine the optimal rotation angle;

[0026] The scoring optimization submodule is used to optimize the scale factor and crystal sample thickness parameters by minimizing the intersection-over-union loss function, and to screen the optimal crystal orientation and structure in combination with the optimal rotation angle.

[0027] In a second aspect, the present invention provides an intelligent analysis method for crystal structure of electron microscopic images, the characteristics of which include:

[0028] Import electron microscopy images and crystal structure data;

[0029] Preprocess the imported electron microscopy images to obtain experimental maps;

[0030] Simulating two-dimensional electron diffraction of the crystal based on the imported crystal structure data to obtain a diffraction simulation result, and processing the diffraction simulation result to obtain a theoretical spectrum;

[0031] The experimental spectrum and the theoretical spectrum are matched, and the matching result is output.

[0032] Optionally, simulating two-dimensional electron diffraction of a crystal based on the imported crystal structure data to obtain a diffraction simulation result, and processing the diffraction simulation result to obtain a theoretical spectrum includes:

[0033] Import crystal structure data;

[0034] When considering the multi-beam interference effect, adjust the crystal sample thickness;

[0035] The imported crystal structure data is based on the propagation equation of electron waves in the crystal, the multi-beam interference effect is taken into account, and the electron diffraction intensity distribution is dynamically calculated based on the adjusted crystal sample thickness to obtain the diffraction simulation results. The diffraction simulation results are processed to obtain the theoretical spectrum.

[0036] Optionally, the imported crystal structure data is based on the propagation equation of electron waves in the crystal, multi-beam interference effect is taken into account, and the electron diffraction intensity distribution is dynamically calculated based on the adjusted crystal sample thickness to obtain a diffraction simulation result, and the diffraction simulation result is processed to obtain a theoretical spectrum, including:

[0037] Calculate the structure factors of each crystal plane based on atomic scattering parameters and crystal structure data;

[0038] Combining the electron wavelength and crystal sample thickness, the excitation error of each crystal plane is derived;

[0039] The electron diffraction intensity distribution is calculated based on the excitation error of each crystal plane to obtain a diffraction simulation result, and the diffraction simulation result is processed to obtain a theoretical spectrum.

[0040] Optionally, matching the experimental spectrum with the theoretical spectrum includes:

[0041] Convert the experimental and theoretical maps from Cartesian coordinates to polar coordinates;

[0042] The fast Fourier cross-correlation between the experimental and theoretical images was calculated in polar coordinates to determine the optimal rotation angle;

[0043] By minimizing the intersection-over-union loss function, optimizing the scale factor and crystal sample thickness parameters, and combining the optimal rotation angle, the optimal crystal orientation and structure are screened.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] The electron microscopic image crystal structure intelligent analysis system provided in this embodiment can automatically analyze the electron microscopic image crystal structure throughout the entire process under the synergistic effect of various modules, greatly improving the efficiency of crystal structure analysis; based on the diffraction simulation module, diffraction simulation can be realized without relying on closed-source tools such as SingleCrystal; at the same time, under the action of the intelligent matching module, fast matching can be achieved and the error rate can be reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 A schematic diagram of the composition of the electron microscopy image crystal structure intelligent analysis system provided in an embodiment of the present application;

[0047] Figure 2 Schematic diagram of the composition of the diffraction simulation module;

[0048] Figure 3 Schematic diagram of the composition of the Bloch wave dynamics simulation submodule;

[0049] Figure 4 This is a schematic diagram of the composition of the intelligent matching module;

[0050] Figure 5 A flow chart of the method for intelligent analysis of crystal structure of electron microscopic images provided in an embodiment of the present application;

[0051] Figure 6 is a flowchart of step S3;

[0052] Figure 7 This is a flowchart of sub-step S33;

[0053] Figure 8 This is a flowchart of step S4. DETAILED DESCRIPTION

[0054] The technical solution of the present invention is further described below with reference to the accompanying drawings and embodiments.

[0055] Example:

[0056] See Figure 1 As shown, the electron microscopic image crystal structure intelligent analysis system 100 provided in this embodiment mainly includes:

[0057] The data import module 110 is used to import electron microscopy images and crystal structure data.

[0058] The image preprocessing module 120 is used to preprocess the imported electron microscopy image to obtain an experimental atlas.

[0059] The diffraction simulation module 130 simulates two-dimensional electron diffraction of the crystal based on the imported crystal structure data, generating diffraction simulation results. These diffraction simulation results are then transformed into a theoretical spectrum after polar coordinate transformation. This allows the diffraction simulation module to be used without relying on closed-source tools such as SingleCrystal.

[0060] The intelligent matching module 140 is used to match the experimental spectrum with the theoretical spectrum and output the matching result. In this way, under the action of the intelligent matching module, fast matching can be achieved and the error rate is reduced.

[0061] The result output module 150 is used to display the output matching results and diffraction simulation results, and integrates the PyMatgen library to realize crystal structure visualization.

[0062] It can be seen that the electron microscopy image crystal structure intelligent analysis system provided in this embodiment can automatically analyze the electron microscopy image crystal structure in the entire process under the synergistic effect of various modules, greatly improving the efficiency of crystal structure analysis; based on the diffraction simulation module, diffraction simulation can be realized without relying on closed-source tools such as SingleCrystal; at the same time, under the action of the intelligent matching module, fast matching can be achieved and the error rate can be reduced.

[0063] In a specific embodiment, if Figure 2 As shown, the diffraction simulation module 130 includes:

[0064] Crystal structure input interface 1301, used to import crystal structure data;

[0065] A thickness adaptive optimizer 1302 is used to adjust the crystal sample thickness when considering the multi-beam interference effect;

[0066] The Bloch wave dynamics simulation submodule 1303 dynamically calculates the electron diffraction intensity distribution based on the imported crystal structure data, based on the electron wave propagation equation in the crystal, taking into account the multi-beam interference effect, and based on the crystal sample thickness adjusted by the thickness adaptive optimizer. This results in a diffraction simulation result, which is then processed to produce a theoretical spectrum. This allows for accurate theoretical spectrum generation.

[0067] More specifically, if Figure 3 As shown, the Bloch wave dynamics simulation submodule 1303 includes:

[0068] The structure factor calculation unit 13031 calculates the structure factor of each crystal plane based on the atomic scattering parameters and crystal structure data ( hkl) structure factor ;

[0069] Excitation error modeling unit 13032, used to combine electron wavelength and crystal sample thickness t , derive the excitation error of each crystal plane ;The thickness of the crystal sample t is the thickness of the crystal sample adjusted based on the thickness adaptive optimizer;

[0070] Intensity simulation unit 13033, by formula , to calculate the electron diffraction intensity distribution and obtain the diffraction simulation results; where ξ is the extinction distance, which is determined by the electron wavelength and crystal structure parameters; then the theoretical spectrum is obtained by the following processing method: according to the calculated electron diffraction intensity distribution, the low-intensity peak is first removed to simulate the detection threshold in the experiment, and then the diffraction point broadening is simulated by Gaussian convolution. Finally, an annular mask is applied to limit the reciprocal space range and the intensity is normalized to obtain the theoretical spectrum.

[0071] In one embodiment, the crystal structure input interface supports importing CIF files or Material Project databases to enhance the versatility of the crystal structure input interface; that is, the crystal structure data can be imported from the MaterialProject database, and user-defined databases (CIF file collections) are also accepted;

[0072] The crystal structure input interface is used to extract lattice parameters, atomic coordinates, and symmetry information.

[0073] In a specific embodiment, the thickness adaptive optimizer adopts multi-threaded parallel computing and supports FFT (Fast Fourier Transform) / diffraction mode switching to improve its versatility.

[0074] In a specific embodiment, the image preprocessing module 120 is used to preprocess the imported electron microscopy image to obtain an experimental atlas, including:

[0075] After fast Fourier transform of the imported electron microscopy image, the amorphous background was fitted using the Pseudo-Voigt method, the amorphous scattering noise was dynamically subtracted, and the image resolution was unified by combining the Lanczos interpolation algorithm to obtain the experimental map.

[0076] In this way, the experimental spectrum can be obtained accurately and efficiently through the above operations.

[0077] like Figure 4 As shown, the intelligent matching module 140 includes:

[0078] Polar coordinate transformation submodule 1401 is used to transform the experimental spectrum and theoretical spectrum from the Cartesian coordinate system to the polar coordinate system;

[0079] The cyclic cross-correlation submodule 1402 is used to calculate the fast Fourier cross-correlation between the experimental image and the theoretical image in polar coordinates to determine the optimal rotation angle ψ for subsequent scoring and determination of crystal orientation.

[0080] The scoring optimization submodule 1403 is used to optimize the rescale factor (the rescale factor is the ratio of the maximum intensity of the experimental spectrum to the theoretical spectrum) and the crystal sample thickness parameter by minimizing the intersection-over-union (IoU) loss function, and to screen the optimal crystal orientation and structure in combination with the optimal rotation angle ψ determined by the cyclic cross-correlation submodule 1402.

[0081] In this way, through the synergistic effect of the above sub-modules, automatic matching of experimental maps and theoretical maps can be achieved, thereby solving the problems of low efficiency and insufficient accuracy caused by the existing manual comparison and matching.

[0082] Specifically, the scale factor was optimized by searching for a local optimum in the interval [0.1, 5] using the Brent algorithm. The sample crystal thickness parameters and crystal orientation were obtained by minimizing the intersection-over-union loss function using the Direct Search algorithm.

[0083] The sample crystal structure is obtained by calculating and screening a series of crystal structures in the crystal structure library, including:

[0084] Step 1: Establish a three-dimensional coordinate system

[0085] The crystal is oriented using three rotation angles: θ (tilt), φ (pitch), and ψ (rotation). θ and φ control the angle at which the electron beam penetrates the crystal, while ψ is the rotation angle.

[0086] Step 2: Adjust the angle by trial and error

[0087] Automatically try different angle combinations: First, fix the thickness and use a global optimization algorithm (such as Direct Search) to quickly scan the possible range of θ and φ; after each angle adjustment, the corresponding theoretical map is immediately generated.

[0088] Step 3: Fingerprint comparison

[0089] The experimental and theoretical spectra are converted into polar coordinates, and the matching degree is calculated by fast Fourier transform to find the best matching rotation angle, that is, the optimal rotation angle ψ.

[0090] Step 4: Fine-tune and confirm

[0091] After determining the optimal rotation angle ψ, fine adjustments are made and the optimal crystal sample thickness is automatically calculated. Finally, the intersection-over-union (IoU) score is used to select the angle combination with the highest matching degree, thereby screening the optimal crystal orientation and structure.

[0092] Accordingly, if Figure 5 As shown, this embodiment also provides a method for intelligent analysis of crystal structure of sub-microscopic images, which mainly includes the following steps:

[0093] S1, importing electron microscopy images and crystal structure data;

[0094] S2, preprocessing the imported electron microscopy images to obtain experimental maps;

[0095] S3, simulating two-dimensional electron diffraction of the crystal based on the imported crystal structure data to obtain a diffraction simulation result, and processing the diffraction simulation result to obtain a theoretical spectrum;

[0096] S4, matching the experimental spectrum with the theoretical spectrum, and outputting the matching result.

[0097] It can be seen that the intelligent analysis method of the crystal structure of electron microscopic images provided in this embodiment can automatically analyze the crystal structure of electron microscopic images throughout the entire process under the synergistic effect of the steps, greatly improving the efficiency of crystal structure analysis; based on step S3, diffraction simulation can be achieved without relying on closed-source tools such as SingleCrystal; at the same time, under the action of step S4, fast matching can be achieved and the error rate can be reduced.

[0098] In a specific embodiment, Figure 6 As shown, the above step S3 includes the following sub-steps:

[0099] S31, import crystal structure data;

[0100] S32, adjust the crystal sample thickness when considering the multi-beam interference effect;

[0101] S33, based on the propagation equation of electron waves in the crystal, taking into account the multi-beam interference effect, the electron diffraction intensity distribution is dynamically calculated based on the adjusted crystal sample thickness for the imported crystal structure data to obtain the diffraction simulation results, and the diffraction simulation results are processed to obtain a theoretical spectrum.

[0102] In this way, under the synergistic effect of the above sub-steps, the theoretical map can be accurately obtained.

[0103] More specifically, if Figure 7 As shown, the above sub-step S33 includes the following sub-sub-steps:

[0104] S331, calculate the crystal planes based on atomic scattering parameters and crystal structure data ( hkl ) structure factor ;

[0105] S332, Combining Electron Wavelength and Crystal Sample Thickness t , derive the excitation error of each crystal plane ;The thickness of the crystal sample t is the thickness of the crystal sample adjusted based on the thickness adaptive optimizer;

[0106] S333, by formula , to calculate the electron diffraction intensity distribution; obtain the diffraction simulation results, where ξ is the extinction distance, which is determined by the electron wavelength and crystal structure parameters; then obtain the theoretical spectrum through the following processing method: according to the calculated electron diffraction intensity distribution, first remove the low-intensity peak to simulate the detection threshold in the experiment, then simulate the diffraction point broadening through Gaussian convolution, and finally apply a ring mask to limit the reciprocal space range and normalize the intensity to obtain the theoretical spectrum.

[0107] In a specific embodiment, if Figure 8 As shown, the above step S4 includes the following sub-steps:

[0108] S41, transform the experimental map from Cartesian coordinate system to polar coordinate system;

[0109] S42, calculating the fast Fourier cross-correlation between the experimental and theoretical images in polar coordinates to determine the optimal rotation angle ψ for subsequent scoring and determination of crystal orientation;

[0110] S43, by minimizing the intersection-over-union (IoU) loss function, optimizing the rescale factor (the rescale factor is the ratio of the maximum intensity of the experimental map and the theoretical map) and the crystal sample thickness parameter, and combining the optimal rotation angle to screen the optimal crystal orientation and structure.

[0111] In this way, through the synergistic effect of the above sub-steps, automatic matching of experimental maps and theoretical maps can be achieved, thereby solving the problems of low efficiency and insufficient accuracy caused by the existing manual comparison and matching.

[0112] The above embodiments are intended only to illustrate the technical concepts and features of the present invention. Their purpose is to enable those skilled in the art to understand the contents of the present invention and implement them accordingly. They are not intended to limit the scope of protection of the present invention. Any equivalent changes or modifications made based on the essence of the present invention are intended to be covered by the scope of protection of the present invention.

Claims

1. An intelligent analysis system for electron microscopic image crystal structure, characterized in that: include: Data import module, used to import electron microscopy images and crystal structure data; An image preprocessing module is used to preprocess the imported electron microscopy images to obtain experimental maps; A diffraction simulation module simulates two-dimensional electron diffraction of the crystal based on the imported crystal structure data to obtain diffraction simulation results, and processes the diffraction simulation results to obtain a theoretical spectrum; An intelligent matching module, used to match the experimental spectrum with the theoretical spectrum and output the matching result; A result output module is used to display the output matching results and diffraction simulation results to achieve crystal structure visualization; The diffraction simulation module includes: Crystal structure input interface, used to import crystal structure data; Thickness Adaptive Optimizer, used to adjust crystal sample thickness when considering multi-beam interference effects; The Bloch wave dynamics simulation submodule dynamically calculates the electron diffraction intensity distribution based on the imported crystal structure data based on the electron wave propagation equation in the crystal, taking into account the multi-beam interference effect, and based on the crystal sample thickness adjusted by the thickness adaptive optimizer to obtain diffraction simulation results; the diffraction simulation results are processed to obtain a theoretical spectrum; The Bloch wave dynamics simulation submodule includes: The structure factor calculation unit calculates the structure factor of each crystal plane based on atomic scattering parameters and crystal structure data; The excitation error modeling unit is used to combine the electron wavelength and the crystal sample thickness to derive the excitation error of each crystal plane; an intensity simulation unit, which calculates the electron diffraction intensity distribution based on the excitation error of each crystal plane of the crystal to obtain a diffraction simulation result; and processes the diffraction simulation result to obtain a theoretical spectrum; The intelligent matching module includes: Polar coordinate transformation submodule, used to transform experimental and theoretical maps from Cartesian coordinate system to polar coordinate system; The cyclic cross-correlation submodule is used to calculate the fast Fourier cross-correlation between the experimental and theoretical images in polar coordinates to determine the optimal rotation angle; The scoring optimization submodule is used to optimize the scale factor and crystal sample thickness parameters by minimizing the intersection-over-union loss function, and to screen the optimal crystal orientation and structure in combination with the optimal rotation angle.

2. The electron microscopic image crystal structure intelligent analysis system according to claim 1, characterized in that: The thickness adaptive optimizer adopts multi-threaded parallel computing and supports FFT / diffraction mode switching.

3. The electron microscopic image crystal structure intelligent analysis system according to claim 1, characterized in that: The image preprocessing module is used to preprocess the imported electron microscopy image to obtain an experimental atlas, including: performing fast Fourier transform on the imported electron microscopy image, fitting the amorphous background, dynamically deducting amorphous scattering noise, and unifying the image resolution to obtain the experimental atlas.

4. A method for intelligent analysis of crystal structure of electron microscopic images, characterized in that: include: Import electron microscopy images and crystal structure data; Preprocess the imported electron microscopy images to obtain experimental maps; Simulating two-dimensional electron diffraction of the crystal based on the imported crystal structure data to obtain a diffraction simulation result, and processing the diffraction simulation result to obtain a theoretical spectrum; Matching the experimental spectrum with the theoretical spectrum and outputting a matching result; The method comprises simulating two-dimensional electron diffraction of the crystal based on the imported crystal structure data to obtain a diffraction simulation result, and processing the diffraction simulation result to obtain a theoretical spectrum, including: Import crystal structure data; When considering the multi-beam interference effect, adjust the crystal sample thickness; Based on the propagation equation of electron waves in the crystal, taking into account the multi-beam interference effect, the electron diffraction intensity distribution is dynamically calculated based on the adjusted crystal sample thickness for the imported crystal structure data to obtain a diffraction simulation result, and the diffraction simulation result is processed to obtain a theoretical spectrum; The imported crystal structure data is based on the propagation equation of electron waves in the crystal, taking into account the multi-beam interference effect, and dynamically calculating the electron diffraction intensity distribution based on the adjusted crystal sample thickness to obtain a diffraction simulation result, and the diffraction simulation result is processed to obtain a theoretical map, including: Calculate the structure factors of each crystal plane based on atomic scattering parameters and crystal structure data; Combining the electron wavelength and crystal sample thickness, the excitation error of each crystal plane is derived; Calculating the electron diffraction intensity distribution based on the excitation error of each crystal plane to obtain a diffraction simulation result, and processing the diffraction simulation result to obtain a theoretical spectrum; The matching of the experimental spectrum and the theoretical spectrum includes: Convert the experimental and theoretical maps from Cartesian coordinates to polar coordinates; The fast Fourier cross-correlation between the experimental and theoretical images was calculated in polar coordinates to determine the optimal rotation angle; By minimizing the intersection-over-union loss function, optimizing the scale factor and crystal sample thickness parameters, and combining the optimal rotation angle, the optimal crystal orientation and structure are screened.

Images