A multi-dimensional raman phonon spectrum peak characteristic analysis method and a detection device

Abstract

The application discloses a multi-dimensional Raman phonon spectrum peak feature analysis method and a detection device, belongs to the field of spectral microscopic technology, and solves the technical problems that the existing single-dimensional Raman phonon spectrum peak feature analysis method is influenced by local heat accumulation of two-dimensional materials, has insufficient detection precision, and is difficult to analyze the complex mapping relationship between multi-dimensional spectrum peak features and microscopic structure parameters. The device comprises a front-end optical path, a rear-end optical path and a sample positioning unit, and realizes ultra-narrow linewidth excitation and low wave number detection through volume phase holographic filtering. The method extracts multi-dimensional features of Raman phonon spectrum peaks, obtains weighted sampling probability in combination with a plurality of characteristic importance measurements, constructs a feature weighted integrated decision tree classifier, adopts Bayesian optimization to obtain optimal hyperparameters, and outputs microscopic structure parameters through majority voting. The application is suitable for scenes of detecting microscopic structure parameters such as the number of layers, stacking mode, stress distribution, defect concentration and phase change behavior of two-dimensional materials.

Description

Technical Field

[0001] This invention belongs to the field of spectroscopic microscopy technology, and mainly relates to a multidimensional Raman phonon spectrum peak characteristic analysis method and detection device for detecting the microstructure parameters of materials. Background Technology

[0002] The minimum feature size of silicon-based integrated circuit chips has approached the quantum tunneling critical size of silicon materials, and Moore's Law is nearing its physical limit. As the size of traditional silicon-based transistors is scaled down to the nanoscale, their short-channel effect is becoming increasingly severe. Conventional technological innovations, such as changing circuit structures and shrinking device and interconnect sizes, are no longer sufficient to further increase their integration density. There is an urgent need to find new materials that can be used to manufacture transistors.

[0003] The International Devices and Systems Roadmap organization identifies two-dimensional (2D) materials as the most competitive and promising new transistor materials for continuing Moore's Law in next-generation semiconductor devices, according to its technology roadmap. Since the microstructural parameters of 2D materials (including but not limited to the number of layers, torsion angle, stacking type, stress distribution, defect concentration, and phase transition behavior) directly affect the intrinsic properties of the materials (including but not limited to band gap and band structure), they significantly alter the key performance indicators of the fabricated 2D semiconductor transistors (including but not limited to carrier mobility and on / off ratio), ultimately profoundly impacting the optoelectronic performance and reliability of semiconductor devices. Therefore, high-precision detection of the microstructural parameters of 2D materials is crucial for ensuring device performance and achieving controllable material fabrication and application.

[0004] Currently, Raman spectroscopy microscopy has become a mainstream non-destructive technique for detecting the microstructural parameters of two-dimensional materials. This is because two-dimensional materials have low wavenumbers (≤ 10 ... ), medium to high wave number (> There is an inherent correlation between Raman phonon spectral peak characteristics (including but not limited to peak position, peak position difference, peak area / peak intensity, peak intensity ratio, peak height, and peak width) and their microstructural parameters. Traditional Raman spectroscopy analysis methods detect the microstructural parameters of two-dimensional materials by analyzing only a single dimension of Raman phonon spectral peak characteristics. This involves constructing relationships such as "peak position - microstructural parameter," "peak intensity - microstructural parameter," or "peak width - microstructural parameter" as the basis for detecting the microstructural parameters of two-dimensional materials. However, in actual microstructural parameter detection, due to the atomic-level layer thickness, high photothermal conversion efficiency, and limited thermal diffusion of two-dimensional materials, local thermal accumulation effects can occur. These local thermal accumulation effects significantly affect the Raman phonon spectral peak characteristics related to the material's microstructural parameters. Therefore, analyzing only a single dimension of Raman phonon spectral peak characteristics is insufficient to meet the requirements for high-precision detection of material microstructural parameters.

[0005] Therefore, a pressing technical problem that needs to be solved by those skilled in the art is how to simultaneously combine the multi-dimensional Raman phonon spectral peak feature information of materials, analyze the complex mapping relationship between the multi-dimensional Raman phonon spectral peak feature information of materials and microstructural parameters, and effectively improve the detection accuracy of Raman spectroscopy microscopy system for microstructural parameters of two-dimensional materials. Summary of the Invention

[0006] This invention provides a method and detection device for multidimensional Raman phonon spectral peak feature analysis. Its purpose is to address the technical problems of existing single-dimensional Raman phonon spectral peak feature analysis methods, which are affected by local thermal accumulation in two-dimensional materials, resulting in insufficient detection accuracy and difficulty in resolving the complex mapping relationship between multidimensional spectral peak features and microstructural parameters. Furthermore, this invention addresses how to simultaneously combine multidimensional Raman phonon spectral peak feature information of materials to resolve the complex mapping relationship between the multidimensional Raman phonon spectral peak feature information and microstructural parameters, effectively improving the detection accuracy of Raman spectroscopy microscopy systems for material microstructural parameters.

[0007] In a first aspect, the purpose of this invention is to provide a multidimensional Raman phonon spectrum peak feature detection device, comprising: a front-end optical path processing unit, a rear-end optical path processing unit, and a sample positioning unit;

[0008] The front-end optical path processing unit is used to generate and modulate excitation light, and is provided sequentially along the optical axis as follows: a laser generation system, a collimation and beam expansion system, a first polarization control system, a volume phase holographic bandpass filter system, a beam angle control system, and a volume phase holographic notch filter system.

[0009] The back-end optical path processing unit is used to process Raman spectral signals and obtain multi-dimensional Raman phonon spectral peak characteristic information of materials. It is provided in sequence along the optical axis as follows: beam noise reduction system, second polarization control system, achromatic beam focusing system, spectral dispersion system, and area array detection system.

[0010] The sample positioning unit is used to fix the sample and adjust its position, including a three-dimensional displacement stage;

[0011] All optical elements are arranged along the same optical axis, and all lenses are perpendicular to the optical axis;

[0012] The beam angle control system is used to incident a laser beam onto a volume phase holographic notch filter system, wherein the laser beam is a laser beam filtered by the volume phase holographic bandpass filter system, and the incident angle satisfies Bragg's law.

[0013] The reflective optical path of the volume phase holographic notch filter system is sequentially arranged with a microscope objective and a sample; the sample is fixed on the three-dimensional displacement stage and located at the front focal plane of the microscope objective; the Raman spectral signal generated by the sample is transmitted to the area array detection system via the microscope objective, the transmission end of the volume phase holographic notch filter system, the beam noise reduction system, the second polarization control system, the achromatic beam focusing system, and the spectral dispersion system.

[0014] Furthermore, a preferred embodiment is provided: the volume phase holographic bandpass filter system is used to reduce the noise of the laser beam after it has been processed by the collimation and beam expansion system and the first polarization control system, so as to obtain an ultra-narrow linewidth monochromatic excitation beam with a spectral linewidth of less than 1 MHz.

[0015] Furthermore, a preferred embodiment is provided: the volume phase holographic notch filter system is used to notch filter the beam signal collected by the microscope objective.

[0016] Secondly, the purpose of this invention is to propose a method for multidimensional Raman phonon spectrum peak characteristic analysis. This method is implemented based on a multidimensional Raman phonon spectrum peak characteristic detection device described in any one or more of the above-mentioned schemes, and includes the following steps:

[0017] S1: Obtain the Raman spectrum of the material, extract the Raman phonon peak feature information of the material, and combine it with the microstructure parameters of the material to form the original sample dataset;

[0018] S2: Based on at least two feature importance measurement methods, calculate the feature importance measurement factor for the detection of corresponding microstructure parameters for each Raman phonon spectrum peak feature information of the material;

[0019] S3: Normalize each feature importance metric factor corresponding to each Raman phonon spectral peak feature information, then sum them by weight and take the average value to obtain the weighted average feature importance metric factor for each type of Raman phonon spectral peak feature information;

[0020] S4: Calculate the corresponding weighted average random sampling probability distribution value based on the weighted average feature importance measure factor of each Raman phonon spectrum peak feature information;

[0021] S5: Perform multiple uniform random samplings with replacement on the original sample dataset to generate multiple decision trees;

[0022] S6: When each decision tree splits at a node, it performs feature-weighted average random sampling based on the weighted average random sampling probability distribution value of the feature information of each Raman phonon spectrum peak.

[0023] S7: Based on Bayesian optimization theory of tree structure estimator, construct an optimization function with the objective of identifying the accuracy of material microstructure parameters, and solve to obtain the global optimal hyperparameter combination of multiple decision trees;

[0024] S8: Based on the aforementioned globally optimal hyperparameter combination, construct a feature-weighted average random sampling integrated decision tree classifier;

[0025] S9: The multi-dimensional Raman phonon spectrum peak feature information of the acquired material is synchronously imported into the classifier. The voting results of each decision tree are counted. Based on the principle that the majority is better than the minority, the result with the most votes is taken as the final detection result of the material's microstructure parameters.

[0026] Furthermore, a preferred embodiment is provided: the original sample dataset is composed of at least two of the Raman phonon spectrum peak feature information and the microstructure parameters of the material; the Raman phonon spectrum peak feature information includes peak position, peak position difference, peak intensity, peak intensity ratio, peak height, peak height ratio, peak width, and peak width ratio.

[0027] Furthermore, a preferred embodiment is provided: the feature-weighted average random sampling integrated decision tree classifier analyzes multi-dimensional Raman phonon spectrum peak feature information in the low wavenumber range and / or medium-to-high wavenumber range of the material; wherein, the multi-dimensional Raman phonon spectrum peak feature information in the low wavenumber range is characterized by wavenumbers not greater than... Multidimensional Raman phonon spectrum peak characteristics, especially in the mid-to-high wavenumber range where the wavenumber is greater than... Multidimensional Raman phonon spectrum peak characteristic information.

[0028] Furthermore, a preferred embodiment is provided: the weighted average feature importance metric factor is used to comprehensively evaluate the contribution of each Raman phonon spectral peak feature information to the detection of the corresponding microstructure parameters.

[0029] Furthermore, a preferred solution is provided: In S6, the higher the contribution of the Raman phonon spectrum peak feature information of the corresponding microstructure parameter detection to the detection of the corresponding microstructure parameter, the greater the probability of the Raman phonon spectrum peak feature information being selected when splitting the decision tree node.

[0030] Thirdly, the present invention aims to provide a computer device, the computer device including a memory and a processor, the memory storing a computer program, and when the processor runs the computer program stored in the memory, the processor executes a multidimensional Raman phonon spectrum peak characteristic analysis method according to any one or more of the above-described schemes.

[0031] Fourthly, the present invention aims to provide a computer-readable storage medium for storing a computer program that executes a multidimensional Raman phonon spectrum peak characteristic analysis method described in any one or more of the above-described schemes.

[0032] Compared with the prior art, the advantages of the present invention are:

[0033] This invention replaces the single-dimensional Raman peak feature analysis in traditional Raman spectroscopy microscopy with simultaneous multi-dimensional Raman peak feature analysis. It uses weighted average random sampling probabilities obtained based on multiple feature importance measurement methods to construct an integrated decision tree classifier based on feature-weighted average random sampling, strengthening the dominant role of important features. Combining random sampling with replacement and feature-weighted average random sampling ensures the diversity of the decision tree. Furthermore, it replaces traditional grid search and random search optimization with Bayesian optimization based on a tree structure estimator, using the accuracy of microstructure parameter identification as the optimization objective. It adaptively searches for the globally optimal hyperparameter combination, improving the classifier's tuning efficiency, generalization, and robustness. This invention can simultaneously combine multi-dimensional Raman peak features in low and medium-high wavenumber ranges, analyzing the complex mapping relationship between peak features and microstructure parameters, significantly improving the system's detection accuracy of material microstructure parameters.

[0034] This invention is applicable to scenarios involving the detection of microstructural parameters such as the number of layers, stacking method, stress distribution, defect concentration, and phase transition behavior of two-dimensional materials, and can also be extended to the field of multidimensional Raman phonon spectrum feature analysis of other nanomaterials. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0036] Figure 1 This is a schematic diagram of the multidimensional Raman phonon spectrum peak feature detection device according to a specific embodiment of the present invention;

[0037] Figure 2 This is a schematic diagram of the multidimensional Raman phonon spectrum peak characteristic analysis method described in Embodiment 2 of the present invention;

[0038] Among them: 1-Laser generation system, 2-Collimation and beam expansion system, 3-First polarization control system, 4-Volume phase holographic bandpass filter system, 5-Beam angle control system, 6-Volume phase holographic notch filter system, 7-Microscopic objective lens, 8-Sample, 9-Three-dimensional displacement stage, 10-Beam noise reduction system, 11-Second polarization control system, 12-Achromatic beam focusing system, 13-Spectral dispersion system, 14-Area array detection system. Detailed Implementation

[0039] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application can also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of this application with unnecessary detail.

[0040] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0041] Many specific details are set forth in the following description in order to provide a full understanding of this application. However, this application may also be implemented in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the spirit of this application. Therefore, this application is not limited to the specific embodiments disclosed below.

[0042] Implementation Method 1

[0043] The multidimensional Raman phonon spectrum peak feature detection device described in this embodiment is as follows: Figure 1 As shown.

[0044] The device includes: a front-end optical path processing unit, a rear-end optical path processing unit, and a sample positioning unit;

[0045] The front-end optical path processing unit is used to generate and modulate excitation light, and is provided in sequence along the optical axis as follows: laser generation system 1, collimation and beam expansion system 2, first polarization control system 3, volume phase holographic bandpass filter system 4, beam angle control system 5, and volume phase holographic notch filter system 6.

[0046] The back-end optical path processing unit is used to process Raman spectral signals and obtain multi-dimensional Raman phonon spectral peak characteristic information of materials. It is provided in sequence along the optical axis as follows: beam noise reduction system 10, second polarization control system 11, achromatic beam focusing system 12, spectral dispersion system 13, and area array detection system 14.

[0047] The sample positioning unit is used to fix the sample 8 and adjust its position, including a three-dimensional displacement stage 9;

[0048] All optical elements are arranged along the optical axis, that is, the center of the optical surface of all optical elements coincides with the optical axis formed by the center beam of the emitted laser and Raman spectral signal of the laser generation system 1, and all lenses are perpendicular to the optical axis.

[0049] The beam angle control system 5 is used to direct the laser beam filtered by the volume phase holographic bandpass filter system 4 into the volume phase holographic notch filter system 6 at an angle that satisfies Bragg's law.

[0050] The reflective end optical path of the volume phase holographic notch filter system 6 is sequentially arranged with a microscope objective 7 and a sample 8; the sample 8 is fixed on the three-dimensional displacement stage 9 and located at the front focal plane of the microscope objective 7; the Raman spectral signal generated by the sample 8 is transmitted to the area array detection system 14 via the microscope objective 7, the transmission end of the volume phase holographic notch filter system 6, the beam noise reduction system 10, the second polarization control system 11, the achromatic beam focusing system 12, and the spectral dispersion system 13.

[0051] The volume phase holographic bandpass filter system 4 is used to reduce the noise of the beam after it has been processed by the collimation and beam expansion system 2 and the first polarization control system 3, so as to obtain an ultra-narrow linewidth monochromatic excitation beam with a spectral linewidth of less than 1 MHz.

[0052] The volume phase holographic notch filter system 6 is used to notch filter the beam signal collected by the microscope objective (7) so as to achieve a minimum detectable Raman phonon peak of less than or equal to 9 wavenumbers.

[0053] The multidimensional Raman phonon spectral peak feature detection device described in this embodiment can simultaneously detect low wavenumber ranges (≤ ), medium to high wavenumber range (> The Raman spectroscopy detection capability of the instrument can simultaneously detect the characteristic information of low-wavenumber and medium-high-wavenumber Raman spectral peaks in two-dimensional materials.

[0054] Implementation Method 2

[0055] A method for analyzing the characteristics of multidimensional Raman phonon spectral peaks, the method being implemented based on a multidimensional Raman phonon spectral peak detection device as described in Embodiment 1, includes the following steps:

[0056] S1: The Raman spectrum of the material is obtained using the area array detection system 14. The complete Raman phonon spectrum peak feature information of the material is extracted and combined with the microstructure parameters of the material to form the original sample dataset. That is, the original sample dataset can be composed of the peak position A1, peak position difference B1, peak intensity C1, peak intensity ratio D1, and peak width E1 of a Raman phonon mode in the low wavenumber range, the peak position A2, peak position difference B2, peak intensity C2, peak intensity ratio D2, and peak width E2 of a Raman phonon mode in the high wavenumber range, and the number of material layers L. N sub-sample datasets are formed by random sampling without replacement from the original sample dataset.

[0057] S2: Based on multiple feature importance measurement methods, the feature importance measurement factors for each Raman phonon spectrum peak feature information of the material in layer number detection are calculated for each subsample set. This is calculated using the Shapley Additive Explanations method based on game theory, resulting in the following set of feature importance measurement factors for a Raman phonon mode in the low wavenumber range: peak position A1, peak position difference B1, peak intensity C1, peak intensity ratio D1, and peak width E1. The set of characteristic importance metrics for a Raman phonon mode in the high wavenumber range, including peak position A2, peak position difference B2, peak intensity C2, peak intensity ratio D2, and peak width E2, are as follows: Based on the Mutual Information method of information theory, the set of characteristic importance metrics for a Raman phonon mode in the low wavenumber range, namely peak position A1, peak position difference B1, peak intensity C1, peak intensity ratio D1, and peak width E1, is calculated as follows: The set of characteristic importance metrics for a Raman phonon mode in the high wavenumber range, including peak position A2, peak position difference B2, peak intensity C2, peak intensity ratio D2, and peak width E2, are as follows: ;

[0058] S3: Normalize the mean of each feature importance metric factor for each Raman phonon spectral peak feature information of the material, and then sum and average the weighted values ​​to obtain the weighted average feature importance metric factor for each Raman phonon spectral peak feature information. That is, use min-max normalization to obtain the normalized feature importance metric factors for peak position A1, peak position difference B1, peak intensity C1, peak intensity ratio D1, and peak width E1 of a Raman phonon mode in the low wavenumber range, which are as follows: The normalized characteristic importance measures of peak position A2, peak position difference B2, peak intensity C2, peak intensity ratio D2, and peak width E2 for a Raman phonon mode in the high wavenumber range are as follows: , The weighted average feature importance measures of peak position A1, peak position difference B1, peak intensity C1, peak intensity ratio D1, and peak width E1 for a Raman phonon mode in the low wavenumber range are as follows:

[0059]

[0060] The weighted average feature importance measures of peak position A2, peak position difference B2, peak intensity C2, peak intensity ratio D2, and peak width E2 for a Raman phonon mode in the high wavenumber range are as follows:

[0061] .

[0062] S4: Based on the weighted average feature importance metric of each Raman phonon spectral peak feature information of the material, calculate the weighted average random sampling probability distribution value of each Raman phonon spectral peak feature information, that is, sort the weighted average feature importance metric of each Raman phonon spectral peak feature information in descending order. If:

[0063] ,

[0064] make ,

[0065] Find the maximum value of k that satisfies: ,

[0066] Calculate the threshold: ,

[0067] The weighted average random sampling probability distribution values ​​of peak position A1, peak position difference B1, peak intensity C1, peak intensity ratio D1, and peak width E1 for a Raman phonon mode in the low wavenumber range are as follows: , The weighted average random sampling probability distribution values ​​of peak position A2, peak position difference B2, peak intensity C2, peak intensity ratio D2, and peak width E2 of a Raman phonon mode in the high wavenumber range are as follows: , ;

[0068] S5: Perform multiple uniform random samplings with replacement on the original sample dataset to generate multiple decision trees;

[0069] S6: Feature selection when splitting each decision tree node is based on the weighted average random sampling probability distribution value of each Raman phonon spectrum peak feature information;

[0070] S7: Based on Bayesian optimization theory of tree-structured Parsons estimators, an optimization function is constructed with the material layer number identification accuracy as the objective. That is, the objective optimization function can be expressed as:

[0071] ,

[0072] Where θ represents the hyperparameter vector (number of decision trees, maximum depth, number of features sampled by node splitting, minimum number of samples a node can split into, maximum number of leaf nodes, etc.), and Θ represents the search space. This represents a decision tree classifier. It is a k-fold cross-validation set. By solving the objective optimization function, the globally optimal hyperparameter combination of multiple decision trees is obtained.

[0073] S8: Based on the obtained global optimal hyperparameter combination, construct a feature-weighted average random sampling integrated decision tree classifier;

[0074] S9: The multi-dimensional Raman phonon spectrum peak feature information of each Raman phonon mode detected by the system, such as peak position, peak position difference, peak intensity, peak intensity ratio, and peak width, is synchronously imported into the constructed feature-weighted average random sampling integrated decision tree classifier. The voting results of each decision tree are counted. Based on the principle of "majority is better than minority", the result with the most votes is regarded as the final detection result of the material layer number.

[0075] In this embodiment, the original sample dataset is composed of five feature information of the material’s Raman phonon spectrum peaks, namely peak position, peak position difference, peak intensity, peak intensity ratio, and peak width, together with the number of material layers, in order to construct a feature-weighted average random sampling integrated decision tree classifier.

[0076] In this embodiment, the constructed feature-weighted average random sampling integrated decision tree classifier can simultaneously combine multi-dimensional Raman phonon spectrum peak feature information in the low wavenumber range and the medium-high wavenumber range of the material to analyze the complex mapping relationship between the multi-dimensional Raman phonon spectrum peak feature information and the number of layers of the material.

[0077] In this embodiment, the influence of each Raman phonon peak feature information of the material on the layer number detection can be comprehensively evaluated by using a weighted average feature importance measurement factor obtained based on two feature importance measurement methods.

[0078] In this embodiment, for feature selection when splitting each decision tree node, a weighted average random sampling is performed based on the weighted average random sampling probability distribution value of each Raman phonon spectrum peak feature information of the material. Raman phonon spectrum peak feature information that has a greater impact on the detection of the material layer number will be selected more likely when splitting the decision tree node, so as to improve the influence of this type of Raman phonon spectrum peak feature information on the final detection result.

[0079] In this embodiment, uniform random sampling of sample data and weighted average random sampling of sample features are used, which not only ensures the diversity of decision trees, but also strengthens the dominant role of Raman phonon spectrum peak feature information with high weighted average feature importance measurement factor in decision trees.

[0080] In this embodiment, an optimization function is constructed with the material layer number recognition accuracy as the objective. The function consists of five hyperparameters: the number of decision trees, the maximum depth, the number of features sampled by node splitting, the minimum number of samples that a node can split, and the maximum number of leaf nodes. Based on the Bayesian optimization theory of the tree structure Parsons estimator, the globally optimal hyperparameter combination of multiple decision trees is achieved to improve the generalization and robustness of the feature-weighted average random sampling integrated decision tree classifier.

[0081] In this embodiment, the single-dimensional Raman peak feature analysis in traditional Raman spectroscopy microscopy is replaced with simultaneous multi-dimensional Raman peak feature analysis. A weighted average random sampling probability is obtained based on two feature importance measurement methods, and a feature-weighted average random sampling integrated decision tree classifier is constructed to strengthen the dominant role of important features. Combining random sampling with replacement and feature-weighted average random sampling ensures the diversity of the decision tree. Furthermore, traditional grid search and random search optimization are replaced with Bayesian optimization based on a tree-structured Parsons estimator. With the material layer number identification accuracy as the optimization objective, the system adaptively searches for the globally optimal hyperparameter combination, improving the classifier's tuning efficiency, generalization, and robustness. This invention can simultaneously combine multi-dimensional Raman peak features in low and medium-high wavenumber ranges, analyze the complex mapping relationship between peak features and layer number, and significantly improve the system's layer number detection accuracy.

[0082] This invention is applicable to scenarios involving the detection of microstructural parameters such as the number of layers, stacking method, stress distribution, defect concentration, and phase transition behavior of two-dimensional materials, and can also be extended to the field of multidimensional Raman phonon spectrum feature analysis of other nanomaterials.

[0083] It is understood that the present invention has been described through some embodiments, and those skilled in the art will recognize that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of the invention. Furthermore, under the teachings of the present invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of the present invention.

Claims

1. A multidimensional Raman phonon spectrum peak feature detection device, characterized in that, include: Front-end optical path processing unit, back-end optical path processing unit, and sample positioning unit; The front-end optical path processing unit is used to generate and modulate excitation light, and is provided in sequence along the optical axis as follows: laser generation system (1), collimation and beam expansion system (2), first polarization control system (3), volume phase holographic bandpass filter system (4), beam angle control system (5), and volume phase holographic notch filter system (6). The back-end optical path processing unit is used to process Raman spectral signals and obtain multi-dimensional Raman phonon spectral peak characteristic information of materials. It is provided in sequence along the optical axis as follows: beam noise reduction system (10), second polarization control system (11), achromatic beam focusing system (12), spectral dispersion system (13), and area array detection system (14). The sample positioning unit is used to fix the sample (8) and adjust its position, including a three-dimensional displacement stage (9). All optical elements are arranged along the same optical axis, and all lenses are perpendicular to the optical axis; The beam angle control system (5) is used to incident the laser beam onto the volume phase holographic notch filter system (6), wherein the laser beam is a laser beam filtered by the volume phase holographic bandpass filter system (4), and the incident angle satisfies Bragg's law. The reflective end optical path of the volume phase holographic notch filter system (6) is sequentially arranged with a microscope objective (7) and a sample (8); the sample (8) is fixed on the three-dimensional displacement stage (9) and located at the front focal plane of the microscope objective (7); the Raman spectral signal generated by the sample (8) is transmitted to the area array detection system (14) through the microscope objective (7), the transmission end of the volume phase holographic notch filter system (6), the beam noise reduction system (10), the second polarization control system (11), the achromatic beam focusing system (12), and the spectral dispersion system (13).

2. The multidimensional Raman phonon spectrum peak feature detection device according to claim 1, characterized in that, The volume phase holographic bandpass filter system (4) is used to reduce the noise of the laser beam after it has been processed by the collimation and beam expansion system (2) and the first polarization control system (3) in order to obtain an ultra-narrow linewidth monochromatic excitation beam with a spectral linewidth of less than 1 MHz.

3. The multidimensional Raman phonon spectrum peak feature detection device according to claim 1, characterized in that, The volume phase holographic notch filter system (6) is used to notch filter the beam signal collected by the microscope objective (7).

4. A method for analyzing the peak characteristics of multidimensional Raman phonon spectra, characterized in that, The method is based on a multidimensional Raman phonon spectrum peak feature detection device as described in any one of claims 1-3, and includes the following steps: S1: Obtain the Raman spectrum of the material, extract the Raman phonon peak feature information of the material, and combine it with the microstructure parameters of the material to form the original sample dataset; S2: Based on at least two feature importance measurement methods, calculate the feature importance measurement factor for the detection of corresponding microstructure parameters for each Raman phonon spectrum peak feature information of the material; S3: Normalize each feature importance metric factor corresponding to each Raman phonon spectral peak feature information, then sum them by weight and take the average value to obtain the weighted average feature importance metric factor for each type of Raman phonon spectral peak feature information; S4: Calculate the corresponding weighted average random sampling probability distribution value based on the weighted average feature importance measure factor of each Raman phonon spectrum peak feature information; S5: Perform multiple uniform random samplings with replacement on the original sample dataset to generate multiple decision trees; S6: When each decision tree splits at a node, it performs feature-weighted average random sampling based on the weighted average random sampling probability distribution value of the feature information of each Raman phonon spectrum peak. S7: Based on Bayesian optimization theory of tree structure estimator, construct an optimization function with the objective of identifying the accuracy of material microstructure parameters, and solve to obtain the global optimal hyperparameter combination of multiple decision trees; S8: Based on the aforementioned globally optimal hyperparameter combination, construct a feature-weighted average random sampling integrated decision tree classifier; S9: The multi-dimensional Raman phonon spectrum peak feature information of the acquired material is synchronously imported into the classifier. The voting results of each decision tree are counted. Based on the principle that the majority is better than the minority, the result with the most votes is taken as the final detection result of the material's microstructure parameters.

5. The method for analyzing the peak characteristics of multidimensional Raman phonon spectra according to claim 4, characterized in that, The original sample dataset is composed of at least two of the Raman phonon spectrum peak feature information and the microstructure parameters of the material; the Raman phonon spectrum peak feature information includes peak position, peak position difference, peak intensity, peak intensity ratio, peak height, peak height ratio, peak width, and peak width ratio.

6. The method for analyzing the peak characteristics of multidimensional Raman phonon spectra according to claim 4, characterized in that, The feature-weighted average random sampling integrated decision tree classifier combines multi-dimensional Raman phonon spectrum peak feature information in the low wavenumber range and / or medium-high wavenumber range of the material for analysis; wherein, the multi-dimensional Raman phonon spectrum peak feature information in the low wavenumber range is characterized by wavenumbers not greater than... Multidimensional Raman phonon spectrum peak characteristics, especially in the mid-to-high wavenumber range where the wavenumber is greater than... Multidimensional Raman phonon spectrum peak characteristic information.

7. The method for analyzing the peak characteristics of multidimensional Raman phonon spectra according to claim 4, characterized in that, The weighted average feature importance metric is used to comprehensively evaluate the contribution of each Raman phonon spectral peak feature information to the detection of the corresponding microstructure parameters.

8. The method for analyzing the peak characteristics of multidimensional Raman phonon spectra according to claim 7, characterized in that, In step S6, the higher the contribution of Raman phonon spectrum peak feature information to the detection of corresponding microstructure parameters of the material, the greater the probability of Raman phonon spectrum peak feature information being selected when splitting the decision tree node.

9. A computer device, characterized in that, The computer device includes a memory and a processor. The memory stores a computer program. When the processor runs the computer program stored in the memory, the processor executes a multidimensional Raman phonon spectrum peak characteristic analysis method according to any one of claims 4-8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store a computer program that executes a multidimensional Raman phonon spectrum peak characteristic analysis method according to any one of claims 4-8.

Images