A method for batch processing and imaging of electron energy loss spectra

Abstract

The present application relates to a kind of batch information processing and imaging method of electron energy loss spectrum, belong to the field of electron energy loss spectrum information processing.It is related to the feature extraction and analysis of electron energy loss spectrum, mainly for the characteristic information analysis of nanomaterial.By integrating power law background deduction algorithm, fourier ratio deconvolution multiple scattering algorithm, multi-order gaussian peak background fitting algorithm, peak integration algorithm and other batch processing algorithms, the problems of electron energy loss spectrum processing difficulty and low processing efficiency can be effectively solved, and support is provided for more extensive technical expansion.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of electron energy loss spectrum information processing, and particularly relates to a batch information processing and imaging method of electron energy loss spectrum. BACKGROUND

[0002] Electron energy loss spectrum (EELS) is an important means to characterize the electronic structure and valence bond information of materials, and is widely used in the fields of energy, catalysis, semiconductors, etc. Through spectrum analysis, the energy loss of electrons with known kinetic energy after interacting with atoms in the sample to be measured can be obtained, and the spatial environment information of the electrons in the sample to be measured can be obtained by using the distribution characteristics of the energy loss, and then the physical and chemical properties of the material can be analyzed.

[0003] Since the information extraction of electron energy loss spectrum involves complex mathematical operations and requires a professional theoretical knowledge base, the conventional spectrum processing and information extraction method mainly extracts single spectrum information or images a single two-dimensional array. Many documents also use this method for processing, such as the processing methods reported in documents J. Appl. Phys. 108, 063704 (2010) and Nat. Commun. 15, 378 (2024). However, as the number of materials increases and the analysis dimension of materials increases, researchers need to process a large number of electron energy loss spectra of the same type, which requires a batch processing technology. At the same time, analyzing a certain type of nanomaterials, a batch information extraction and imaging method for multiple two-dimensional electron energy loss spectra is needed. SUMMARY

[0004] The present application aims to solve the shortcomings in the batch processing of two-dimensional electron energy loss spectrum, fully considers the energy loss theory, electronic structure characteristics, spectral peak characteristic absorption, and other theoretical knowledge in the spectrum information, integrates the power law background subtraction algorithm, Fourier ratio deconvolution multiple scattering algorithm, multi-order Gaussian peak background fitting algorithm, peak integral algorithm, and other batch processing algorithms, and provides a batch information processing and imaging method of electron energy loss spectrum.

[0005] To achieve the above-mentioned purpose, the technical solution of the present application is: a batch information processing and imaging method of electron energy loss spectrum, including batch information processing and two-dimensional imaging of electron energy loss spectrum; wherein the batch information processing process integrates batch processing algorithms to realize batch information processing of electron energy loss spectrum.

[0006] In an embodiment of the present application, the electron energy loss spectrum is stored in a matrix structure before batch information processing, and the storage structure is [x, y, z, n], wherein x and y are the length x and width y of the electron energy loss spectrum, z is the energy collection length of the electron energy loss spectrum, and n is the number of batch processing; the same batch contains a zero-loss Low-loss spectrum [x, y, z, n] 0, an energy range m 0, and a high-energy loss High-loss spectrum [x, y, z, n] 1, an energy range m 1; in multi-batch processing, there are different batches k.

[0007] In an embodiment of the present application, the batch information processing process includes energy value correction, power law background subtraction, logarithmic ratio thickness calculation, Fourier ratio deconvolution to remove multiple scattering, multi-order Gaussian peak background fitting, and peak integration.

[0008] In an embodiment of the present application, the energy value correction is specifically implemented as follows:

[0009] The peak signal at each [x, y] 0 point in the Low-loss spectrum is extracted, and the difference Δx between the energy value at the strongest peak of each point and the 0 value is calculated. energy The energy value of each point in the high-loss spectrum [x, y] 1 is shifted by the corresponding Δx. energy .

[0010] In an embodiment of the present application, the power law background subtraction is specifically implemented as follows:

[0011] In the electron energy loss spectrum, the background presents an exponential decay characteristic. Based on this characteristic, a power law algorithm is used to subtract the background, and the calculation formula is:

[0012] I = a · x b energy (1)

[0013] Wherein, a and b are fitting parameters, x energy is the energy value of the fitting region; in order to avoid measurement error, the fitting signal is defined as the average value of a spectrum image; in the batch processing process, the energy range of 50eV-100eV before the peak is selected as the background fitting region, and the fitting signal is defined as the average value of all points in each spectrum image [x, y]; at the same time, considering the signal difference of each point, a multiplication parameter c i is used for optimization:

[0014] I = I acquire -c i · a · x b energy (2)

[0015] Wherein, I acquireI is the signal after background subtraction.

[0016] In an embodiment of the present application, the logarithmic ratio thickness calculation is implemented as follows:

[0017] According to the integral intensity of the signal in the low loss spectrum, the thickness is calculated using the logarithmic ratio method, which is a multiple of the inelastic mean free path, and the calculation method is as follows:

[0018] t / μ = ln(I t / I0) (3)

[0019] Where I0 is the total sum of zero loss peak counts, I t is the total sum of the spectrum counts, i.e. the zero loss peak plus the extrapolated 50 eV loss part, and μ is the mean free path of inelastic scattering, t is the sample thickness; using the relative thickness, the intensity after thickness correction is calculated, and the calculation is as follows:

[0020] I high-loss = I·e^(-t / μ) (4)

[0021] Where I high-loss is the corrected high energy loss spectrum.

[0022] In an embodiment of the present application, the Fourier ratio deconvolution to remove multiple scattering is implemented as follows:

[0023] The energy loss function ELF of all points in each spectral image [x, y] is calculated by two-dimensional Fourier transform; the high energy loss spectrum is converted from the energy domain to the frequency domain by Fourier transform, in the frequency domain, the Fourier transform of the high energy loss spectrum is divided by the energy loss function for deconvolution, and the deconvoluted frequency domain data is inverse Fourier transformed to convert it back to the energy domain, eliminating the influence of multiple scattering effects on the high energy loss spectrum, and the calculation formula is as follows:

[0024] I corr = IFT((FT(I high-loss )) / FT(I low-loss )) (5)

[0025] Where FT represents Fourier transform, I low-loss is the zero loss spectrum, IFT represents inverse Fourier transform, and I corr is the corrected signal intensity.

[0026] In an embodiment of the present application, the multi-order Gaussian peak background fitting is implemented as follows:

[0027] The baseline of the element peak in the material is fitted according to the corrected signal, the fitting region is selected as the extended region of the energy value x1 before the peak and the energy value x2 after the peak, and the baseline fitting uses a third-order Gaussian model:

[0028] I g = a1 * e^((-((x energy -b1) / c1)^2))+a2 * e^((-((x energy -b2) / c2)^2))…+a3 * e^((-((x energy -b3) / c3)^2) (6)

[0029] Wherein, I g is the baseline value, a i ,b i ,c i are fitting parameters, i=1,2,3.

[0030] In an embodiment of the present application, the peak integral is specifically implemented as follows:

[0031] The peak integral of each point in each spectral image [x,y] is calculated, and the value is the material information characteristic value at the corresponding position; the characteristic values of each image are imaged to obtain a batch of two-dimensional images; the integral value is:

[0032] Intensity = ∑(I corr -I g ) (7).

[0033] The present application also provides an electron energy loss spectrum batch information processing and imaging system, comprising a memory, a processor and computer program instructions stored in the memory and capable of being executed by the processor, when the processor executes the computer program instructions, the method steps as described above can be realized.

[0034] Compared with the prior art, the present application has the following beneficial effects: the method fully considers the energy loss theory, electron structure characteristics, spectral peak characteristic absorption and other theoretical knowledge in the spectral information, integrates the power law background subtraction algorithm, Fourier ratio deconvolution multiple scattering algorithm, multi-order Gaussian peak background fitting algorithm, peak integral algorithm and other batch processing algorithms, the method can realize batch information processing and imaging of electron energy loss spectrum, effectively solve the problems of large processing difficulty and low processing efficiency of electron energy loss spectrum, and provide support for more extensive technical expansion. BRIEF DESCRIPTION OF DRAWINGS

[0035] Figure 1 It is the batch processing flow of the electron energy loss spectrum of the present application.

[0036] Figure 2 It is a baseline fitting schematic diagram.

[0037] Figure 3 It is a schematic diagram after spectral correction.

[0038] Figure 4 to batch extract information of copper in nanomaterials and image.

[0039] Figure 5 to batch extract information of selenium in nanomaterials and image. DETAILED DESCRIPTION

[0040] The technical solutions of the present application will be specifically described below with reference to the drawings.

[0041] Microscopic characterization of nanomaterials is an important means to study and optimize the functional properties of materials. Electron energy loss spectroscopy (EELS) in scanning transmission electron microscopy has a unique ability to characterize the electronic structure and valence bond information of nanomaterials, with an energy resolution of approximately 0.1 eV, which can accurately record the fine structure of the spectrum, and is an important frontier analysis method. However, with the demand for high-throughput and high-latitude characterization, the single-spectrum single-image analysis method has become an important constraint factor for the development of this technology. The information analysis of electron energy loss spectroscopy requires a large amount of time for professional personnel and a large amount of data processing, which is expensive in terms of labor cost. The development of batch processing technology can effectively solve this defect, improve the analysis efficiency of the spectrum, and further improve the research process of the material.

[0042] As shown in Figure 1 The present application provides a batch information processing and imaging method of electron energy loss spectroscopy, comprising batch information processing of electron energy loss spectroscopy and two-dimensional imaging; wherein the batch information processing process integrates a batch processing algorithm to realize batch information processing of electron energy loss spectroscopy. Specifically, the batch information processing process includes energy value correction, power law background subtraction, logarithmic ratio thickness calculation, Fourier ratio deconvolution to remove multiple scattering, multi-order Gaussian peak background fitting, and peak integration.

[0043] First, the electron energy loss spectrum is stored in a matrix structure before batch information processing, and the storage structure is [x, y, z, n], wherein x and y are the length x and width y of the electron energy loss spectrum, z is the energy collection length of the electron energy loss spectrum, and n is the number of batch processing; the same batch contains a zero-loss Low-loss spectrum [x, y, z, n]0, an energy range m0, and a high-energy loss High-loss spectrum [x, y, z, n]1, an energy range m1; in multi-batch processing, there are different batches k.

[0044] Second, the energy value correction batch processing. Extract the peak value signal at each [x, y]0 point in the Low-loss spectrum, calculate the difference Δx between the energy value at the strongest peak of each point and the 0 value energy , and shift the energy value of each point in the high-loss spectrum [x, y]1 by Δx energy .

[0045] Then, background subtraction batch processing. In the electron energy loss spectrum, the electron inelastic scattering, loss of part of the energy occurred randomly small deflection, additional multiple scattering, the influence of thickness effect, background exponential decay characteristics. Based on this feature, the power law algorithm for background subtraction, the calculation formula is:

[0046] I = a x b energy (1)

[0047] Where, a and b are fitting parameters, x energy is the energy value of the fitting region; to avoid measurement error, the fitting signal is defined as the average value of a spectrum image; in the batch processing process, the energy range of 50eV~100eV before the peak is selected as the background fitting region, and the fitting signal is defined as the average value of all points in each spectrum image [x, y]. At the same time, considering the signal difference of each point, the multiplier parameter c i is used for optimization:

[0048] I = I acquire -c i a x b energy (2)

[0049] Where, I acquire is the measured value, I is the signal after background subtraction.

[0050] In addition, logarithmic ratio thickness correction. According to the integral intensity of the signal in the low loss spectrum, the thickness is calculated using the logarithmic ratio method, which is the multiple of the inelastic mean free path. The calculation method is as follows:

[0051] t / μ = ln(I t / I0) (3)

[0052] Where, I0 is the total count of zero loss peak value, I t is the total spectrum count total, that is, the zero loss peak plus extrapolation 50eV loss part, μ is the average free path of inelastic scattering, t is the sample thickness; using the relative thickness, the intensity after thickness correction is calculated as follows:

[0053] I high-loss = I e^(-t / μ) (4)

[0054] Where, I high-loss is the corrected high energy loss spectrum.

[0055] Then, Fourier ratio deconvolution is used to remove multiple scattering. The energy loss function ELF of all points in each spectral image [x, y] is calculated by two-dimensional Fourier transform; the high energy loss spectrum is converted from energy domain to frequency domain by Fourier transform, in which the Fourier transform of the high energy loss spectrum is divided by the energy loss function for deconvolution, and the deconvoluted frequency domain data is converted back to the energy domain by inverse Fourier transform, so as to eliminate the influence of multiple scattering on the high energy loss spectrum, and the calculation formula is as follows:

[0056] I corr high-loss low-loss

[0057] Wherein, FT represents Fourier transform, I low-loss is zero loss spectrum, IFT represents inverse Fourier transform, and I corr is the corrected signal intensity.

[0058] Further, multi-order Gaussian peak background fitting. The baseline of the element peak in the material is fitted according to the corrected signal, the fitting region is selected as the extended region of the energy value x1 before the peak and the energy value x2 after the peak, and the baseline fitting adopts a three-order Gaussian model:

[0059] I g energy energy energy

[0060] Wherein, I g is the baseline value, a i ,b i ,c i are fitting parameters, and i=1, 2, 3.

[0061] Figure 2 is a schematic diagram of baseline fitting. Figure 3 is a schematic diagram of spectrum correction.

[0062] Finally, peak integration and two-dimensional imaging. The peak integral of each point in each spectral image [x, y] is calculated, and the value is the material information characteristic value at the corresponding position. The characteristic values of each image are imaged to obtain a batch of two-dimensional images; the integral value is:

[0063] Intensity=∑(I corr -I g ) (7)。

[0064] ​​​​​​​The application further provides a batch information processing and imaging system of electron energy loss spectrum, comprising a memory, a processor and computer program instructions stored in the memory and capable of being executed by the processor, when the processor executes the computer program instructions, the method steps as described above can be realized.

[0065] Example 1

[0066] Batch extraction of copper information in nanomaterials and imaging.

[0067] (1) Use the electron energy loss tester of the transmission electron microscope to collect five electron energy loss spectra of copper materials and store them in a matrix, the matrix of the low-loss spectrum is A[42, 43, 2020, 5], the energy range is -150eV to 1365eV, and the matrix of the high-loss spectrum is B[42, 43, 2020, 5], the energy range is -650eV to 2165eV.

[0068] (2) Extract the peak signal at each point in the low-loss spectrum, calculate the energy value at the strongest peak of each point and the difference Δxenergy from 0, and store the difference value in matrix C[42, 43, 5], and shift the energy value of each point in the high-loss spectrum by the corresponding Δxenergy, to obtain matrix D[42, 43, 2020, 5].

[0069] (3) According to the integral intensity of the signal in the low loss spectrum, the relative thickness t / μ of each spectrum is calculated, and the relative thickness is stored in matrix E[42, 43, 5], and the high-loss spectrum is corrected by using the relative thickness, to obtain matrix E[42, 43, 2020, 5].

[0070] (4) Using the low loss spectrum matrix A and the high-loss spectrum matrix E, the multiple scattering is removed by Fourier ratio deconvolution to obtain matrix F[42, 43, 2020, 5].

[0071] (5) According to the signal matrix F[42, 43, 2020, 5] of the high-loss spectrum, the signal peak background is fitted in the extension region of the energy value 100eV before the copper peak (931eV) and the energy value 100eV after the peak (981eV).

[0072] (6) Accumulate the energy range 931-981eV, and subtract the energy value of the signal peak background to obtain the element signal intensity of each point, and image it, and the obtained image is as shown in Figure 4 .

[0073] Example 2

[0074] Batch extraction of selenium information in nanomaterials and imaging.

[0075] Steps (1) to (4) are the same as described in Example 1, except that 5 electron energy loss spectra of selenium materials are collected and stored in a matrix.

[0076] (5) According to the signal matrix G[42, 43, 2020, 5] of the high-loss spectrum of the selenium material, the signal peak background is fitted in the extension region with an energy value of 50 eV before the selenium peak (1436 eV) and an energy value of 50 eV after the peak (1486 eV).

[0077] (6) The energy value of the signal peak background is subtracted in the energy range of 1436-1486 eV to obtain the element signal intensity of each point, and an image is formed, as shown in Figure 5 .

[0078] The above is the preferred embodiment of the present application. Any changes made in accordance with the technical solutions of the present application, as long as the resulting functional effects do not exceed the scope of the technical solutions of the present application, are within the scope of protection of the present application.

Claims

1. A batch information processing and imaging method for electron energy loss spectra, characterized in that, This includes batch processing of electron energy loss spectra and two-dimensional imaging. The batch processing integrates batch processing algorithms to achieve batch processing of electron energy loss spectra. Before batch processing, the electron energy loss spectra are stored in a matrix structure [x,y,z,n], where x and y are the acquisition length x and width y of the electron energy loss spectrum, z is the energy acquisition length of the electron energy loss spectrum, and n is the number of batches. Each batch contains a low-loss spectrum [x,y,z,n]0 with an energy range of m0 and a high-loss spectrum [x,y,z,n]1 with an energy range of m1. Multiple batches have different batches k. The batch processing includes energy value correction, power-law background subtraction, logarithmic ratio thickness calculation, Fourier ratio deconvolution to remove multiple scattering, multi-order Gaussian peak background fitting, and peak integration. The energy value correction is specifically implemented as follows: Extract the peak signal at each [x,y]0 point in the Low-loss spectrum, and calculate the difference Δx between the energy value of the strongest peak and the 0 value at each point. energy The energy value at each point in the high-loss spectrum [x,y]1 is shifted by the corresponding Δx. energy ; The specific implementation of the power-law background subtraction is as follows: In the electron energy loss spectrum, the background exhibits an exponential decay characteristic. Based on this characteristic, a power-law algorithm is used to subtract the background, and the calculation formula is as follows: I=a·x b energy (1) Where a and b are the fitting parameters, x energy This refers to the energy value of the fitted region. To avoid measurement errors, the fitted signal is defined as the average value of a spectral image. During batch processing, the energy range of 50 eV to 100 eV before the peak is selected as the background fitted region. The fitted signal is defined as the average value of all points within [x, y] of each spectral image. Simultaneously, considering the signal difference at each point, the magnification parameter c is used. i Optimize: I=I acquire -c i ·a·x b energy (2) Among them, I acquire I is the measured value, and I is the signal after background subtraction.

2. The batch information processing and imaging method for electron energy loss spectrum according to claim 1, characterized in that, The specific implementation of the logarithmic ratio thickness calculation is as follows: The thickness is calculated using the logarithmic ratio method based on the integral intensity of the signal in the low loss spectrum. The thickness is a multiple of the inelastic mean free path. The calculation method is as follows: t / µ1ln(I t / I0) (3) Where I0 is the sum of zero-loss peak counts, I t The sum of the total spectral counts, i.e., the zero-loss peak value, is added to the extrapolated 50 eV loss component. μ is the mean free path of inelastic scattering, and t is the sample thickness. Using the relative thickness, the thickness-corrected intensity is calculated as follows: I high-loss =I·e^(-t / μ) (4) Among them, I high-loss This is the corrected high-energy loss spectrum.

3. The batch information processing and imaging method for electron energy loss spectrum according to claim 2, characterized in that, The Fourier ratio deconvolution for demultiplexing is specifically implemented as follows: The energy loss function (ELF) for all points in the [x,y] region of each spectral image is calculated using a two-dimensional Fourier transform. The high-energy loss spectrum is then transformed from the energy domain to the frequency domain using a Fourier transform. In the frequency domain, the Fourier transform of the high-energy loss spectrum is divided by the energy loss function for deconvolution. The deconvolutioned frequency domain data is then subjected to an inverse Fourier transform to convert it back to the energy domain, eliminating the influence of multiple scattering effects on the high-energy loss spectrum. The calculation formula is as follows: I corr =IFT((FT(I high-loss )) / FT(I low-loss )) (5) Where FT represents the Fourier transform, I low-loss For zero-loss spectrum, IFT stands for Inverse Fourier Transform, I corr This represents the corrected signal strength.

4. The batch information processing and imaging method for electron energy loss spectrum according to claim 3, characterized in that, The multi-order Gaussian peak background fitting is specifically implemented as follows: Based on the baseline of the elemental peaks in the material fitted by the correction signal, the fitting region is selected as the extended region of the pre-peak energy value x1 and the post-peak energy value x2. The baseline fitting adopts a third-order Gaussian model: I g =a1·e^((-((x energy -b1) / c1)^2))+a2·e^((-((x energy -b2) / c2)^2))…+a3·e^((-((x energy -b3) / c3)^2)) (6) Among them, I g As the baseline value, a i ,b i ,c i The fitting parameters are i = 1, 2, 3.

5. The batch information processing and imaging method for electron energy loss spectra according to claim 4, characterized in that, The peak integral is specifically implemented as follows: Calculate the peak integral at each point [x, y] in each spectral image. This value represents the material information feature value at the corresponding location. Image the feature values ​​of each image to obtain a batch of two-dimensional images; the integral value is: Intensity=∑(I corr -I g ) (7)。 6. A batch information processing and imaging system for electron energy loss spectra, characterized in that, It includes a memory, a processor, and computer program instructions stored in the memory and executable by the processor, which, when executed by the processor, enable the implementation of the steps of the method as described in any one of claims 1-5.

Images