A method, system, and electronic device for coherent diffraction imaging aberration correction based on STEM modes.

Abstract

This application provides a coherent diffraction imaging aberration correction method based on STEM mode, comprising the following steps: acquiring raw data of sample coherent diffraction imaging and preprocessing it to obtain an initial diffraction pattern and Ronchigram pattern; extracting associated imaging features, including aberration type features, aberration intensity features, diffraction pattern quality features, and sample imaging features, to construct an aberration feature vector; obtaining aberration correction labels, each having a feature value of [aberration type feature + aberration intensity feature + diffraction pattern quality feature + sample imaging feature]; matching corresponding correction rules according to the aberration correction labels; and correcting the initial diffraction pattern according to the correction rules to obtain an image. Furthermore, this application also provides a system and electronic device for executing and storing the above aberration correction method. Compared with existing technologies, the aberration correction method, system, and electronic device of this application have the advantages of fast correction speed and high accuracy.

Description

Technical Field

[0001] This invention relates to the field of coherent diffraction imaging technology, and in particular to a method, system, and electronic device for correcting aberrations in coherent diffraction imaging based on STEM mode. Background Technology

[0002] Scanning transmission electron microscopy, often abbreviated as STEM, is a powerful microscopy technique that combines the core principles of scanning electron microscopy and transmission electron microscopy. In STEM mode, coherent diffraction imaging involves the following steps: focusing, where an electromagnetic lens system focuses the electron beam to form a probe; scanning, where the probe performs a grating scan on the sample; and image generation, where electrons that have penetrated the sample are collected, and a diffraction image is formed based on the different degrees of scattering of the electrons.

[0003] Coherent diffraction imaging in STEM mode records the two-dimensional convergent electron diffraction pattern at each scanning position, thereby obtaining atomic-scale characterization of the material's microstructure. Ideally, electron beams passing through the same point on the sample should converge into a single, sharp point at the corresponding point in the diffraction image. However, due to factors such as the curved shape and asymmetry of the lens, and external interference, some electron beams may fail to converge, dispersing into blurry spots within a small area, resulting in aberrations. Aberrations can lead to impaired image resolution, inaccurate phase, and signal loss; therefore, aberration correction is necessary to eliminate the discrepancy between the ideal and actual imaging.

[0004] Existing methods for coherent diffraction imaging aberration correction mainly fall into two categories: 1. Hardware correction, which involves introducing complex and precise aberration correctors into the system to physically cancel out imaging deviations. However, this method requires high stability of hardware and the imaging environment, often incurring significant costs and involving routine maintenance. 2. Software / algorithm correction, which uses existing algorithms to correct the imaging. It involves inversely calculating aberrations from the acquired data, then correcting the image based on the calculated aberrations, outputting a corrected image that approximates the ideal image. However, there are many types of aberrations, and different algorithms are required for different types. Currently, aberration type identification mainly relies on manual methods, leading to inefficiency and the risk of misjudgment resulting in incorrect algorithm usage. Summary of the Invention

[0005] To overcome the above shortcomings, this application provides a visual-tactile stress distribution sensing device, the technical solution of which is as follows:

[0006] A method for correcting aberrations in coherent diffraction imaging based on STEM mode includes the following steps: The raw data of coherent diffraction imaging of the sample were acquired and preprocessed to obtain the initial diffraction pattern and Ronchigram pattern. Based on the initial diffraction pattern and the Ronchigram pattern, associated imaging features are extracted. The associated imaging features include aberration type features, aberration intensity features, diffraction pattern quality features, and sample imaging features. Each associated imaging feature constitutes an aberration feature vector. Aberration correction labels are obtained based on each aberration feature vector. The aberration correction labels have feature values ​​of [aberration type feature + aberration intensity feature + diffraction pattern quality feature + sample imaging feature]. Match the corresponding correction rules according to the aberration correction labels; The initial diffraction pattern is corrected according to the correction rule to obtain the corrected image.

[0007] Compared with existing technologies, the STEM-based coherent diffraction imaging aberration correction method of this application can automatically match aberration correction rules according to multiple associated imaging features, reducing the inefficiency and large errors caused by manual correction. This STEM-based coherent diffraction imaging aberration correction method has high correction accuracy and avoids hardware modifications, resulting in low cost.

[0008] Furthermore, the method includes the following steps: correcting the dynamic error of the corrected imaging to obtain the dynamically error corrected imaging; the dynamic error includes probe drift, and / or sample thickness fluctuation, and / or parameter deviation.

[0009] Furthermore, the correction rules include: Correction is performed based on the aberration type characteristics; Alternatively, when the aberration correction label has the characteristic value of [single spherical aberration + low or medium noise + thin sample], the cosine correction algorithm is used to calculate the corrected aberration coefficient C. new And based on the corrected aberration coefficient C new Correction is performed; the corrected aberration coefficient C new Calculate using the following formula: C new = C original × cosθ, where C new C represents the corrected aberration coefficient. original Here, θ represents the original aberration coefficients, and θ is the tilt angle of the electron beam. Alternatively, when the aberration correction label has the characteristic value of [single astigmatism + medium noise + thin sample], the offset compensation amount Δ(x,y) is calculated based on the gradient compensation algorithm, and the asymmetric offset in the diffraction pattern is corrected based on the offset compensation amount Δ(x,y). Alternatively, when the aberration correction label has the characteristic value of [multiple aberrations superimposed + medium or high noise], the noise is first eliminated, and then the joint correction parameter C is calculated according to the following formula. total Then, based on the combined correction parameter C total The correction is performed by first correcting spherical aberration, then astigmatism compensation, and finally coma correction: Combined correction parameter C total = α×C ball +β×C astigm + γ×C nomal C ball C is the coefficient of spherical aberration. astigm C is the astigmatism coefficient. nomal α represents the coma coefficient, and β and γ represent the spherical aberration weighting coefficient, aberration weighting coefficient, and coma weighting coefficient, respectively. Alternatively, when the aberration correction label has the characteristic value of [high noise + electron beam sensitive sample], a low-dose adaptive correction algorithm is used, combined with iterative optimization to perform correction.

[0010] Furthermore, when correcting the probe drift, the probe drift amount at each point is calculated based on the overlapping area of ​​the diffraction patterns of adjacent frames of the corrected imaging, and the coordinates of each point in the corrected diffraction pattern are calibrated according to the probe drift amount. When correcting sample thickness fluctuations, the thickness compensation value for each point is calculated according to the following formula: thickness compensation value = 0.015 × sample thickness + 0.05 pixels, and the coordinates of each point of the corrected diffraction pattern are compensated according to the sample thickness. When correcting parameter deviations, after obtaining the diffraction pattern of a standard sample with known lattice parameters and the theoretical correction value of a standard sample with known lattice parameters, aberration correction is performed on the standard sample to obtain the actual correction value of the standard sample with known lattice parameters. The theoretical correction value is compared with the actual correction value, and the spherical aberration weighting coefficient α, the aberration weighting coefficient β, and the coma weighting coefficient γ are dynamically adjusted according to the comparison result.

[0011] Furthermore, the aberration type features include single spherical aberration, single astigmatism, single coma, and multiple aberration superposition; the aberration intensity features include intensity, moderate, and weak; the diffraction pattern quality features include high noise, medium noise, and low noise; and the sample imaging features include thin samples, thick samples, and electron beam sensitive samples.

[0012] Furthermore, the preprocessing includes aligning diffraction spots, denoising, and normalizing grayscale on the diffraction pattern obtained from the original data.

[0013] Furthermore, the method also includes the following steps: storing the dynamically error-corrected image in the aberration feature library, determining the quality level of the dynamically error-corrected image, and establishing the correlation between the associated imaging features, the correction rules, and the quality of the dynamically error-corrected image.

[0014] Furthermore, this application also provides a STEM-based coherent diffraction imaging aberration correction system that implements the above-mentioned aberration correction method, the technical solution of which is as follows: A coherent diffraction imaging aberration correction system based on STEM mode includes: An imaging acquisition module is used to acquire the raw data of the sample coherent diffraction imaging and perform preprocessing. The aberration feature recognition module is used to extract the associated imaging features, construct each aberration feature vector, and obtain the aberration correction label; The intelligent correction decision module is used to match and execute the corresponding correction rules based on the aberration correction labels to obtain the corrected image.

[0015] Furthermore, it also includes: A dynamic error correction module is used to correct the dynamic error of the corrected image and obtain the image after dynamic error correction. The structured result storage and linkage analysis module is used to store the dynamically error-corrected image into the aberration feature library, determine the quality level of the dynamically error-corrected image, and establish the correlation between the associated imaging features, the correction rules, and the quality of the dynamically error-corrected image.

[0016] In addition, this application also provides an electronic device, the technical solution of which is as follows: An electronic device includes a chip, a processor, and a memory, the memory storing computer program code including computer instructions. When the chip executes the computer instructions, the electronic device performs the aforementioned STEM-based coherent diffraction imaging aberration correction method. Attached Figure Description

[0017] Figure 1 This is a schematic flowchart of the STEM-based coherent diffraction imaging aberration correction method of this application. Figure 2 This is a schematic diagram of the overall structure of the STEM-based coherent diffraction imaging aberration correction system of this application. Detailed Implementation

[0018] Based on the existing problems with aberration correction in STEM-based coherent diffraction imaging, the applicant analyzed and studied the characteristics and causes of aberrations in existing coherent diffraction imaging. The analysis revealed that, under STEM mode, aberrations in coherent diffraction imaging can be broadly categorized into spherical aberration, astigmatism, coma, and superposition of multiple aberrations, each with its own imaging characteristics. Therefore, the applicant studied and categorized the features of these aberrations, establishing an aberration feature database. A deep learning model can then be used to enable the correction system to automatically identify these aberration features. Furthermore, unlike existing algorithms that only select correction rules based on aberration type, aberration features also include aberration intensity, diffraction pattern quality, and sample imaging characteristics. By refining these parameters, the accuracy of aberration feature identification is improved, thereby enhancing the accuracy of selecting correction rules.

[0019] In addition, the applicant also found in the analysis of existing coherent diffraction imaging that, under the influence of the working principle, coherent diffraction imaging in STEM mode also has some inherent and recurring external errors, such as probe drift, sample thickness fluctuations, and differences in electron beam tilt angle. If these external errors are identified and corrected, the imaging quality can be further improved.

[0020] In response to the above findings, the applicant proposes a method, system, and electronic device for aberration correction in coherent diffraction imaging based on STEM mode. This method establishes an aberration feature database and compares it with the raw data of coherent diffraction imaging to automatically identify aberration features. Then, aberrations are corrected based on the identification results. Please refer to [link to relevant documentation]. Figure 1 The STEM-based coherent diffraction imaging aberration correction method of this application includes the following steps: Step S10: Obtain the raw data of the sample coherent diffraction imaging and preprocess the raw data to obtain the initial diffraction pattern and Ronchigram pattern.

[0021] Specifically, after setting up the STEM system to scan the sample, the system acquires raw imaging data such as sample thickness information, probe position, and diffraction patterns and Ronchigrams corresponding to the probe position.

[0022] Preprocessing includes using existing technologies to perform diffraction spot alignment, denoising, and grayscale normalization on the diffraction pattern obtained from the original data to obtain the initial diffraction pattern, thereby eliminating background interference, ensuring the accuracy of subsequent aberration feature value extraction, and laying the foundation for subsequent aberration feature value recognition.

[0023] Step S20: Based on the initial diffraction pattern and Ronchigram pattern, extract the associated imaging features, including aberration type features, aberration intensity features, diffraction pattern quality features, and sample imaging features, to form each aberration feature vector.

[0024] Aberration feature database and deep learning model can be established based on existing technologies to automatically identify various associated imaging features.

[0025] Aberration type features include single spherical aberration, single astigmatism, single coma, and multiple aberration superposition, which are extracted based on the stripe features of the Ronchigram pattern.

[0026] When the Ronchigram pattern is disk-shaped with no internal fringes, the aberration type is identified as no aberration. When the Ronchigram pattern only shows symmetrical and concentric curved fringes, the aberration type is identified as a single spherical aberration. When the Ronchigram pattern only shows elliptical fringes stretched or compressed along the same axis, the aberration type is identified as a single astigmatism. When the Ronchigram pattern only shows tail-like fringes formed by unilateral offset and unilateral energy dispersion, the aberration type is identified as a single coma. When the Ronchigram pattern combines multiple single aberration type features, the aberration type is identified as multiple aberration superposition. For multiple aberration superposition, the superimposed aberration type can be identified by comparing it with each single aberration type feature one by one. For example, if it has both symmetrical and concentric curved fringes and elliptical fringes stretched or compressed along the same axis, it is a superposition of multiple aberrations, including spherical aberration and astigmatism.

[0027] Aberration intensity features include intensity, moderate, and weak. These are extracted based on the offset of the diffraction spots before and after alignment during preprocessing. If the offset is greater than 2, the aberration intensity feature is intensity; if the offset is between 0.5 and 2, the aberration intensity feature is moderate; and if the offset is less than 0.5, the aberration intensity feature is weak.

[0028] The quality characteristics of diffraction patterns are categorized into high noise, medium noise, and low noise. During preprocessing, denoising can be achieved using existing techniques such as Gaussian filtering. The quality characteristics of the diffraction patterns are then extracted based on the grayscale difference between the diffraction spots before and after denoising. If the grayscale difference is greater than 20, the diffraction pattern is classified as high noise; if the grayscale difference is between 8 and 20, it is classified as medium noise; and if the grayscale difference is less than 8, it is classified as low noise.

[0029] Sample imaging characteristics include thin samples, thick samples, and electron beam sensitive samples. The determination is based on the uniformity of the diffraction intensity distribution of the initial diffraction pattern. If the diffraction intensity distribution is uniform, the sample is identified as a thin sample; if the diffraction pattern shows localized attenuation of diffraction intensity, it is identified as a thick sample; if the diffraction pattern shows large fluctuations in diffraction intensity, it is identified as an electron beam sensitive sample. According to existing industry-standard methods for detecting electron beam sensitive samples, the intensity differences of multiple diffraction patterns acquired consecutively at the same scanning position are compared. Then, the relative rate of change of intensity is calculated using the following formula: If the relative rate of change of intensity exceeds 30%, and the relative rate of change of intensity of a pixel is greater than 20%, it is considered a large fluctuation in diffraction intensity: Relative rate of change of intensity = (Imax - Imin) / Iavg 100%, where Imax is the maximum intensity of the corresponding pixel in the diffraction pattern continuously acquired at the same location; Imin is the minimum intensity of the corresponding pixel in the diffraction pattern continuously acquired at the same location; and Iavg is the average intensity of the corresponding pixel in the diffraction pattern continuously acquired at the same location.

[0030] Step S30: Obtain aberration correction labels based on each aberration feature vector. The aberration correction labels have feature values ​​of [aberration type feature + aberration intensity feature + diffraction pattern quality feature + sample imaging feature].

[0031] Referring to existing technologies, the aberration feature vector, which includes aberration type features, aberration intensity features, diffraction pattern quality features, and sample imaging features, can be input into a random forest classifier to obtain aberration correction labels of [aberration type features + aberration intensity features + diffraction pattern quality features + sample imaging features].

[0032] When setting up a random forest classifier, a model can be built based on an aberration feature database using a random forest classification algorithm. That is, each decision tree of the random forest is constructed based on the aberration feature database to obtain the random forest classifier.

[0033] Step S40: Match the corresponding correction rule according to the aberration correction label.

[0034] Correction rules include existing methods of correction based on aberration type characteristics, or correction rules corresponding to aberration correction labels that are one of the following: When the aberration correction label has the characteristic value of [single spherical aberration + low or medium noise + thin sample], the correction rule that can be selected is: use the cosine correction algorithm to calculate the corrected aberration coefficient C. new And based on the corrected aberration coefficient C new Existing technologies such as analytical direct correction or iterative optimization correction are employed to compensate for distortions caused by electron beam tilt. The corrected aberration coefficient C... newCalculate using the following formula: C new = C original × cosθ, where C new C represents the corrected aberration coefficient. original The original aberration coefficients can be obtained by fitting the diffraction pattern using existing techniques, and θ is the tilt angle of the electron beam.

[0035] When the aberration correction label has the characteristic values ​​of [single astigmatism + medium noise + thin sample], the correction rule that can be selected is: calculate the offset compensation amount Δ(x,y) based on the gradient compensation algorithm in the existing technology, and correct the asymmetric offset in the diffraction pattern based on the offset compensation amount Δ(x,y). Specifically, the offset compensation amount Δ(x,y) is calculated according to the following formula: Offset compensation amount Δ(x,y) = k× I(x,y), where k is the compensation coefficient, obtained through standard sample calibration combined with dynamic adaptation. I(x,y) represents the gray-level gradient of the diffraction pattern, which can be directly calculated using existing image gradient algorithms after preprocessing. Common gradient algorithms in image processing, such as the Sobel operator, Prewitt operator, and Roberts operator, are employed to calculate the horizontal / vertical gray-level change rate pixel-by-pixel on the preprocessed diffraction pattern gray-level matrix, ultimately yielding a two-dimensional gray-level gradient field. I (x,y).

[0036] The above formula makes the offset compensation amount in the gradient compensation algorithm directly related to the gray-scale gradient of the diffraction pattern, thus achieving accurate quantitative correction of asymmetric offset.

[0037] When the aberration correction label has the characteristic value of [multiple aberrations superimposed + medium or high noise], the correction rule that can be selected is: first eliminate the interference of noise on aberration correction, and then perform correction. Specifically, noise is first removed by morphological filtering in the prior art, and then the joint correction parameter C is calculated according to the following formula. total Then, based on the combined correction parameter C total Perform spherical aberration correction first, then astigmatism compensation, and finally coma correction in the order of aberration type "spherical aberration → astigmatism → coma": Combined correction parameter C total = α×C ball + β×C astigm + γ×C nomal C ball C is the spherical aberration coefficient, characterizing the original distortion degree of a single spherical aberration; astigm C is the astigmatism coefficient, representing the degree of original distortion of a single astigmatism; nomalThe coma coefficient represents the original distortion degree of a single coma. α, β, and γ are the spherical aberration weighting coefficient, aberration weighting coefficient, and coma weighting coefficient, respectively, and their magnitudes depend on the aberration intensity characteristics (intense / moderate / weak); the higher the aberration intensity, the larger the corresponding weighting coefficient. Subsequently, based on the joint correction parameter C... total Analytical direct correction or iterative optimization correction, which are methods used in existing technologies, can be employed.

[0038] Because medium and high noise in diffraction patterns can mask the true characteristics of aberrations, directly correcting different types of aberrations can easily lead to parameter calculation errors and distorted correction results. Therefore, eliminating noise interference in aberration correction first can improve correction accuracy. Morphological filtering first removes noise, restoring the true shape of the diffraction pattern and Ronchigram fringes. This makes subsequent step-by-step corrections of spherical aberration, astigmatism, and coma more closely match the actual aberration characteristics, avoiding correction errors caused by noise.

[0039] By employing a step-by-step correction approach based on aberration type, the mutual interference of multiple superimposed aberrations is avoided, achieving precise compensation. When multiple aberrations are superimposed, the distortion characteristics of different aberration types can overlap and interfere with each other. If a single algorithm is used to correct all aberrations simultaneously, it is easy to encounter problems such as incomplete correction or amplification of the distortion of one aberration after the correction of another. The "spherical aberration priority" sequential step-by-step correction approach aligns with the physical characteristics of STEM imaging. This is because spherical aberration is the most fundamental and critical aberration affecting resolution in STEM mode electron coherence diffraction imaging. Its distortion directly leads to symmetrical distortion of the entire diffraction pattern. If secondary aberrations such as astigmatism and coma are corrected first, the fundamental distortion of spherical aberration renders the compensation parameters for secondary aberrations meaningless. Prioritizing the correction of spherical aberration restores the basic resolution of the image first, making subsequent compensation for astigmatism and coma more targeted, conforming to the physical laws of STEM imaging, and improving the overall correction effect.

[0040] Therefore, eliminating noise first and then performing linkage correction in a certain order can adapt to complex imaging scenarios with high noise and multiple aberrations, solve the drawbacks of single correction and one-size-fits-all processing, and achieve a triple improvement in correction accuracy, efficiency and robustness.

[0041] When the aberration correction label has the characteristic values ​​of [high noise + electron beam sensitive sample], the correction rule that can be selected is: to use a low-dose adaptive correction algorithm, combined with iterative optimization of existing technologies to perform correction. The iterative optimization formula is: I n+1 = I n + λ×(I target - I n ×H), where I n+1 I represents the intensity of the corrected diffraction pattern in the (n+1)th iteration. nLet I be the intensity of the current diffraction pattern in the nth iteration. target Let I be the intensity of the target ideal diffraction pattern, i.e., the theoretical intensity under aberration-free and sample-damage-free conditions; λ be the low-dose adaptation coefficient; and H be the transfer function of the STEM imaging system, used to describe the system response of electron beam imaging. When determining the value, the low-dose adaptation coefficient λ is determined by standard sample calibration and the characteristics of the electron beam-sensitive sample; the target ideal diffraction pattern intensity I... target The reference values ​​are fixed using existing technology and obtained based on the theoretical diffraction intensity of aberration-free standard samples; the transfer function H of the STEM imaging system is an intrinsic parameter of the STEM system, calculated through fitting using existing technology; the intensity I of the current diffraction pattern in the nth iteration is... n The intensity of the diffraction pattern is obtained by collecting the actual diffraction pattern intensity from the current iteration; the corrected diffraction pattern intensity I in the (n+1)th iteration is calculated according to the above formula. n+1 The intensity is used as the input for the next iteration, thus successively approximating the ideal target intensity. Reducing the number of iterations throughout the process avoids prolonged electron beam irradiation that could damage sensitive samples. After iteration terminates, I... n+1 This is the intensity of the final corrected diffraction pattern, which can be directly used to generate a clear image.

[0042] Compared to traditional layered imaging correction algorithms, which increase electron beam irradiation time and cumulative dose through multiple iterations, easily leading to sample structural distortion and inaccurate imaging results, the aforementioned correction rules for electron beam-sensitive samples reduce the number of iterations, decrease imaging damage from the electron beam, and ensure that the corrected imaging results accurately reflect the atomic-scale microstructure of the sample. Furthermore, the introduction of a low-dose adaptation coefficient λ suppresses noise interference, enhances the algorithm's adaptability to high-noise diffraction patterns, strengthens noise robustness, and adapts to high-noise imaging environments, ultimately resolving the contradiction between "fewer iterations" and "high-precision correction."

[0043] Furthermore, based on existing technologies for correction based on aberration type features, and the correction rules corresponding to the aforementioned aberration correction labels, the algorithm's strength, weight, and magnitude of correction can be determined based on aberration intensity features. This is a key detail for achieving accurate, adaptive, and overcorrection-free correction. For example, to determine the weight coefficients for the aforementioned multi-aberration superposition correction, based on existing multivariate nonlinear correlation models, aberration intensity features are quantified into numerical values ​​(e.g., weak = 0.1, moderate = 0.3, strong = 0.5), which are used as core independent variables in the calculation of correction coefficients and error correction coefficients. The correction magnitude R = R0 × [1 - ∑(Ki × Ci × (1 - Pi × Mi))] is directly calculated, where Ci is the quantified aberration intensity value. The greater the intensity, the greater the adjustment magnitude of the correction parameters. Therefore, when multiple aberrations are superimposed, the formula combines the correction parameter C... total = α×C ball+ β×C astigm + γ×C nomal In this model, the spherical aberration weighting coefficient α, the aberration weighting coefficient β, and the coma weighting coefficient γ can be completely determined by the aberration intensity characteristics. The higher the aberration intensity, the larger the corresponding weighting coefficient.

[0044] Step S50: Correct the initial diffraction pattern according to the correction rules to obtain the corrected image.

[0045] In addition, correction rules can be applied only to aberration-sensitive areas in the diffraction pattern, such as the non-edge noise areas of the diffraction pattern, as needed.

[0046] Step S60: Correct the dynamic error of the corrected image according to the physical imaging principle to obtain the image after dynamic error correction.

[0047] Dynamic error correction includes correction for probe drift, sample thickness fluctuation, and parameter deviation. These corrections can be performed independently, or one or both can be selected and performed, with the final results being summed.

[0048] When correcting probe drift, the probe drift amount (Δx, Δy) (in pixels) of each point is calculated based on the overlapping area of ​​the diffraction patterns of adjacent frames after correction. The coordinates of each point in the corrected diffraction pattern are then calibrated according to the probe drift amount. The corrected coordinates (x', y') = (x - Δx, y - Δy).

[0049] When correcting for sample thickness fluctuations, the thickness compensation value for each point is calculated using the following formula: Thickness compensation value = 0.015 × Sample thickness + 0.05 pixels, where the sample thickness is in nanometers. The coordinates of each point in the corrected diffraction pattern are then compensated based on the sample thickness. The corrected coordinates (x', y') = (x + thickness compensation value, y + thickness compensation value). The sample thickness can be obtained in step S10.

[0050] The imaging physical errors of thick samples are mainly due to the multiple elastic and inelastic scattering caused by the electron beam passing through the sample. The scattering effect increases linearly with increasing sample thickness, leading to aberration deviations that also increase linearly with thickness. The thickness compensation value in the above formula increases linearly with increasing sample thickness, achieving dynamic and precise matching compensation for thickness-related aberration deviations. This ensures that the thickness compensation amount closely matches the actual error value as the sample thickness increases. Furthermore, analysis of extensive imaging experimental data revealed that even at the theoretical minimum sample thickness, minor errors still exist in the optical system and electron beam emission end of the device, such as slight lens offsets and initial electron beam divergence. These errors introduce an inherent deviation of approximately 0.05 pixels, independent of sample thickness. The above formula precisely offsets this inherent deviation, ensuring full-range compensation even at the critical transition point from thin to thick samples. Therefore, the above correction method primarily targets electron scattering interference from thick samples, providing precise and linear quantitative compensation for aberration deviations caused by electron scattering in thick samples.

[0051] When correcting parameter deviations, the diffraction pattern of a standard sample with known lattice parameters and the theoretical correction value of the standard sample with known lattice parameters are obtained. Aberration correction is performed on the standard sample according to steps S10 to S40, and the actual correction value of the standard sample with known lattice parameters is obtained. The theoretical correction value is compared with the actual correction value, and the spherical aberration weighting coefficient α, aberration weighting coefficient β, and coma weighting coefficient γ are dynamically adjusted based on the comparison result. During adjustment, initial weights are determined, and dynamic iterative correction is performed in the dynamic error correction stage. Based on the magnitude and direction of the deviation between the correction result and the theoretical value, the spherical aberration weighting coefficient α, aberration weighting coefficient β, and coma weighting coefficient γ are adjusted and optimized in real time, so that the actual correction result infinitely approaches the theoretical value. Standard samples can use WSe2 or gold nanoparticles, etc. The theoretical correction value is the ideal reference value of the standard sample, a known value obtained from the standard sample with known lattice parameters under ideal conditions of no aberrations, no systematic errors, and no algorithmic deviations.

[0052] Step S70: Store the dynamically corrected image in the aberration feature library, determine the image quality level after dynamic error correction, and establish the correlation between imaging features, correction rules and the image quality after dynamic error correction based on existing algorithms to achieve intelligent correction.

[0053] Establishing the correlation between imaging features, correction rules, and image quality after dynamic error correction includes: Step S71: Input associated imaging features into the aberration feature library, including aberration type features, aberration intensity features, diffraction pattern quality features, and sample imaging features; Step S72: Adjust the magnitude of the correction parameters in the correction algorithm of the corresponding correction rule, such as spherical aberration weight coefficient α, aberration weight coefficient β, coma weight coefficient γ, dynamic error correction amount, etc.

[0054] Step S73: Obtain the image quality under different correction parameters, and determine whether the correction parameters are suitable or whether the dynamic error has been completely corrected based on the image quality.

[0055] Imaging quality is primarily measured by imaging resolution, with residual aberrations, diffraction spot sharpness, lattice matching, and correction bias serving as secondary indicators. Based on the steps outlined above, the following is a specific implementation example: Step S10: Acquire raw imaging data such as probe position, corresponding diffraction pattern, Ronchigram pattern, and sample thickness information, and preprocess the diffraction pattern and Ronchigram pattern. Step S20: Based on the Ronchigram pattern morphology, the sample aberration type is identified as single spherical aberration; the grayscale difference of the diffraction spot before and after denoising is 12, therefore the diffraction pattern quality is determined to be medium noise; the diffraction spot offset is 1.2, therefore the aberration intensity is determined to be medium intensity; the diffraction intensity distribution of the diffraction pattern is uniform, therefore the sample imaging characteristics are determined to be a thin sample.

[0056] Step S30: Obtain aberration correction labels based on each aberration feature vector. The aberration correction labels have feature values ​​of [single spherical aberration + medium intensity + medium noise + thin sample]. Step S40: Based on the aberration correction label, select the cosine correction algorithm to counteract the distortion caused by the electron beam tilt. Since the electron beam tilt angle θ measured from the raw imaging data is 32°, and C is obtained by fitting the diffraction pattern... original =1.56, therefore C is calculated according to the formula. new =1.323 Step S50: Locate the aberration-sensitive region in the diffraction pattern, based on C new Aberrations are corrected to obtain a corrected diffraction pattern. In this embodiment, the resolution of the corrected diffraction pattern can reach 0.6 Å.

[0057] Step S60: Simultaneously perform corrections for probe drift, sample thickness fluctuations, and parameter deviations.

[0058] Specifically, the probe drift amounts are calculated as Δx = 0.02 pixels and Δy = 0.01 pixels based on the overlapping area of ​​the diffraction patterns after correction in adjacent frames, and the coordinate values ​​in the corrected diffraction patterns are corrected based on the probe drift amounts.

[0059] The sample thickness is 0.8, and the thickness compensation value is calculated to be 0.062 pixels according to the formula. The imaging of the corrected diffraction pattern is compensated based on the thickness compensation value.

[0060] Since the theoretical correction value of the standard sample of the known lattice of WSe2 is a constant a = 3.28 Å, the spherical aberration weighting coefficient α, aberration weighting coefficient β, and coma weighting coefficient γ are dynamically adjusted according to the theoretical correction value to perform dynamic iterative correction of the diffraction pattern deviation after correction.

[0061] Step S70: Store the correction results and obtain the correlation.

[0062] To implement the above-described STEM-based coherent diffraction imaging aberration correction method, please refer to [link to relevant documentation]. Figure 2 This application also provides a coherent diffraction imaging aberration correction system based on STEM mode, comprising: The imaging acquisition module 100 is used to acquire the raw data of sample coherent diffraction imaging and perform preprocessing.

[0063] The aberration feature recognition module 200 is used to extract associated imaging features, including aberration type features, aberration intensity features, diffraction pattern quality features, and sample imaging features, to form aberration feature vectors and obtain aberration correction labels, thereby realizing automatic scene classification.

[0064] The intelligent correction decision module 300 is used to match and execute the corresponding correction rules according to the aberration correction label to obtain the corrected image, thereby completing the accurate correction of multiple types of aberrations.

[0065] The dynamic error correction module 400 is used to correct the dynamic error of the corrected image according to the physical imaging principle, obtain the image after dynamic error correction, and perform real-time compensation on the correction result to further reduce system error.

[0066] The structured result storage and linkage analysis module 500 is used to store the dynamically error-corrected images into the aberration feature library, determine the quality level of the dynamically error-corrected images, and establish the correlation between the imaging features, correction rules and the quality of the dynamically error-corrected images based on existing algorithms, so as to support multi-dimensional linkage analysis and subsequent optimization.

[0067] Furthermore, this application also provides an electronic device for storing the aforementioned STEM-based coherent diffraction imaging aberration correction method, which performs related calculations, data analysis, and comparisons. The electronic device includes, but is not limited to, a memory, a processor, and a network interface that can be interconnected via a system bus.

[0068] The memory includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. The memory can be an internal storage unit of the electronic device, such as the hard disk or RAM of the electronic device. The memory can also be an external storage device of the electronic device, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc. The memory may also include both internal storage units and external storage devices of the electronic device.

[0069] The processor can be a central processing unit (CPU), controller, microcontroller, microprocessor, or other data processing chip. This processor is typically used to control the overall operation of the electronic device, such as performing control and processing related to data interaction or communication with the electronic device. The processor is used to run program code stored in the memory or process data, such as running the chip detection method described above.

[0070] The network interface may include a wireless network interface or a wired network interface, which is typically used to establish communication connections between the electronic device and other electronic devices. For example, the network interface is used to connect the electronic device to an external data platform via a network, establishing a data transmission channel and communication connection between the electronic device and the external data platform. The network may be an intranet, the Internet, Global System for Mobile communication (GSM), Wideband Code Division Multiple Access (WCDMA), 4G network, 5G network, Bluetooth, Wi-Fi, or other wireless or wired networks.

[0071] Compared with existing technologies, the STEM-based coherent diffraction imaging aberration correction method, system, and electronic device of this application have the following advantages: 1. Improve aberration correction efficiency: Based on actual practical applications, the technical solution of this application can reduce the correction time of a single image from 30-60 minutes of manual adjustment to less than 5 seconds, and supports batch processing to meet the needs of large-scale representation.

[0072] 2. High adaptability: Even in complex scenarios such as multiple aberrations, high noise, and electron beam sensitive samples, the correction success rate has been increased from 60% of the existing technology to over 95%.

[0073] 3. High correction accuracy: The correction deviation in different scenarios is ≤1.5%, which is far lower than the 8% deviation of existing technologies, and the imaging resolution can reach 0.44Å.

[0074] 4. Improved ease of analysis: Structured storage improves the efficiency of aberration-correction effect correlation analysis by 60%, can quickly locate the optimal correction parameters, and supports continuous optimization of algorithms and equipment.

[0075] 5. Significant cost advantage: The pure algorithm solution does not require modification of STEM hardware and can basically be adapted to existing commercial 4D-STEM equipment, lowering the entry barrier for aberration correction and greatly reducing costs compared with hardware correction.

[0076] The embodiments described above are merely examples of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

Claims

1. A method for correcting aberrations in coherent diffraction imaging based on STEM mode, characterized in that, Includes the following steps: The raw data of coherent diffraction imaging of the sample is acquired, and the raw data is preprocessed to obtain the initial diffraction pattern and Ronchigram pattern. Based on the initial diffraction pattern and the Ronchigram pattern, associated imaging features are extracted. The associated imaging features include aberration type features, aberration intensity features, diffraction pattern quality features, and sample imaging features. Each associated imaging feature constitutes an aberration feature vector. Aberration correction labels are obtained based on each aberration feature vector. The aberration correction labels have feature values ​​of [aberration type feature + aberration intensity feature + diffraction pattern quality feature + sample imaging feature]. Match the corresponding correction rules according to the aberration correction labels; The initial diffraction pattern is corrected according to the correction rule to obtain the corrected image.

2. The STEM-based coherent diffraction imaging aberration correction method according to claim 1, characterized in that, It also includes the following steps: The dynamic error of the corrected imaging is corrected to obtain the imaging after dynamic error correction; the dynamic error includes probe drift, and / or sample thickness fluctuation, and / or parameter deviation.

3. The STEM-based coherent diffraction imaging aberration correction method according to claim 2, characterized in that, The correction rules include: Correction is performed based on the aberration type characteristics; Alternatively, when the aberration correction label has the characteristic value of [single spherical aberration + low or medium noise + thin sample], the cosine correction algorithm is used to calculate the corrected aberration coefficient C. new And based on the corrected aberration coefficient C new Correction is performed; the corrected aberration coefficient C new Calculate using the following formula: C new = C original × cosθ, where C new C represents the corrected aberration coefficient. original Here, θ represents the original aberration coefficients, and θ is the tilt angle of the electron beam. Alternatively, when the aberration correction label has the characteristic value of [single astigmatism + medium noise + thin sample], the offset compensation amount Δ(x,y) is calculated based on the gradient compensation algorithm, and the asymmetric offset in the diffraction pattern is corrected based on the offset compensation amount Δ(x,y). Alternatively, when the aberration correction label has the characteristic value of [multiple aberrations superimposed + medium or high noise], the noise is first eliminated, and then the joint correction parameter C is calculated according to the following formula. total Then, based on the combined correction parameter C total The correction is performed by first correcting spherical aberration, then astigmatism compensation, and finally coma correction: Combined correction parameter C total = α×C ball + β×C astigm + γ×C nomal C ball C is the coefficient of spherical aberration. astigm C is the astigmatism coefficient. nomal α represents the coma coefficient, and β and γ represent the spherical aberration weighting coefficient, aberration weighting coefficient, and coma weighting coefficient, respectively. Alternatively, when the aberration correction label has the characteristic value of [high noise + electron beam sensitive sample], a low-dose adaptive correction algorithm is used, combined with iterative optimization to perform correction.

4. The STEM-based coherent diffraction imaging aberration correction method according to claim 3, characterized in that: When correcting the probe drift, the probe drift amount at each point is calculated based on the overlapping area of ​​the diffraction patterns of adjacent frames of the corrected imaging, and the coordinates of each point of the corrected diffraction pattern are calibrated according to the probe drift amount. When correcting for sample thickness fluctuations, the thickness compensation value at each point is calculated using the following formula: Thickness compensation value = 0.015 × sample thickness + 0.05 pixels, and compensate the coordinates of each point in the corrected diffraction pattern according to the sample thickness; When correcting parameter deviations, after obtaining the diffraction pattern of a standard sample with known lattice parameters and the theoretical correction value of a standard sample with known lattice parameters, aberration correction is performed on the standard sample to obtain the actual correction value of the standard sample with known lattice parameters. The theoretical correction value is compared with the actual correction value, and the spherical aberration weighting coefficient α, the aberration weighting coefficient β, and the coma weighting coefficient γ are dynamically adjusted according to the comparison result.

5. The STEM-based coherent diffraction imaging aberration correction method according to claim 1, characterized in that: The aberration type features include single spherical aberration, single astigmatism, single coma, and multiple aberration superposition; the aberration intensity features include intensity, moderate, and weak; the diffraction pattern quality features include high noise, medium noise, and low noise. The sample imaging features include thin samples, thick samples, and electron beam sensitive samples.

6. The STEM-based coherent diffraction imaging aberration correction method according to claim 1, characterized in that: The preprocessing includes aligning diffraction spots, denoising, and normalizing grayscale on the diffraction pattern obtained from the original data.

7. The STEM-based coherent diffraction imaging aberration correction method according to claim 2, characterized in that, It also includes the following steps: The dynamically error-corrected image is stored in the aberration feature library, the quality level of the dynamically error-corrected image is determined, and the correlation between the associated imaging features, the correction rules, and the quality of the dynamically error-corrected image is established.

8. A STEM-based coherent diffraction imaging aberration correction system, used to execute the STEM-based coherent diffraction imaging aberration correction method of claim 1, comprising: An imaging acquisition module is used to acquire the raw data of the sample coherent diffraction imaging and perform preprocessing. The aberration feature recognition module is used to extract the associated imaging features, construct each aberration feature vector, and obtain the aberration correction label; The intelligent correction decision module is used to match and execute the corresponding correction rules based on the aberration correction labels to obtain the corrected image.

9. A coherent diffraction imaging aberration correction system based on STEM mode, characterized in that, Also includes: A dynamic error correction module is used to correct the dynamic error of the corrected image and obtain the image after dynamic error correction. The structured result storage and linkage analysis module is used to store the dynamically error-corrected image into the aberration feature library, determine the quality level of the dynamically error-corrected image, and establish the correlation between the associated imaging features, the correction rules, and the quality of the dynamically error-corrected image.

10. An electronic device, characterized in that, include: A chip, a processor, and a memory, the memory being used to store computer program code, the computer program code including computer instructions, wherein, when the chip executes the computer instructions, the electronic device performs the STEM-based coherent diffraction imaging aberration correction method as described in any one of claims 1 to 7.

Images