A method and device for acquiring material information based on near-field optical testing

Abstract

The present application belongs to the technical field of near-field optical testing, and specifically discloses a method and device for obtaining material information based on near-field optical testing, the method comprising: introducing pump light to focus below the needle tip of a near-field optical microscope and introducing detection light to illuminate below the needle tip, wherein a sample to be tested is arranged below the needle tip; under different pump-detection delays, a set of spectra of the sample to be tested is measured by a near-field optical microscope; a set of spectra of the sample to be tested is input into a deep learning model to obtain material information output by the deep learning model. By combining the means of measuring the spectra of the sample to be tested under different pump-detection delays with the means of using a deep learning model for prediction, the present application can ensure the efficiency of the test while ensuring the accuracy of the prediction, and realize the efficient and accurate acquisition of material information of the sample to be tested using near-field optical signals.

Description

Technical Field

[0001] The present application belongs to the technical field of near-field optical testing, and more specifically, relates to a method and device for acquiring material information based on near-field optical testing. Background Art

[0002] Scanning Near-field Optical Microscopy (SNOM) is a high-resolution optical microscopy technique that breaks through the diffraction limit of traditional optical microscopes and can achieve a resolution far beyond that of ordinary optical microscopes, reaching the nanometer level. Near-field optical microscopy uses the near-field effect, that is, when the needle tip (or probe) is very close to the sample surface (the distance is less than the wavelength), the near-field optical signal on the sample surface can be detected. By detecting near-field optical signals of different frequencies / wavelengths, near-field microscopy can perform spectral detection of samples at the nanoscale, and has important application potential in material identification, defect detection, molecular sensing, etc. How to use near-field optical signals to efficiently and accurately obtain material information of the sample being tested is a technical problem that needs to be solved urgently in this field. Summary of the invention

[0003] In view of the current research status of the existing technology, the purpose of this application is to use deep learning to assist in the efficient and accurate acquisition of material information of the sample being tested by a near-field optical microscope.

[0004] To achieve the above objectives, in a first aspect, the present application provides a method for obtaining material information based on near-field optical testing, comprising:

[0005] Introducing pump light to focus below the needle tip of the near-field optical microscope and introducing detection light to illuminate below the needle tip, and a sample to be measured is arranged below the needle tip;

[0006] A set of spectra of the sample under test is measured by a near-field optical microscope under different pump-probe delays. The pump-probe delay is used to indicate the time interval between the moment when the pump light is introduced and the moment when the probe light is introduced. A set of spectra includes multiple spectra, and different spectra in a set of spectra correspond to different pump-probe delays.

[0007] A set of spectra of the tested samples is input into the deep learning model to obtain the material information output by the deep learning model. The deep learning model is used to predict the material information of the tested samples based on a set of spectra of the tested samples.

[0008] In one possible implementation, the deep learning model is obtained through the following steps:

[0009] Acquire a spectral data set, the spectral data set including multiple groups of spectral samples and labels corresponding to each group of spectral samples, any group of spectral samples is acquired under multiple pump-probe delays during a near-field optical test, and the label is used to indicate material information represented by the corresponding group of spectral samples;

[0010] Based on multiple groups of spectral samples and labels corresponding to each group of spectral samples, an initial deep learning model is trained in a supervised manner to obtain a trained deep learning model.

[0011] In a possible implementation, before training an initial deep learning model in a supervised manner based on multiple groups of spectral samples and labels corresponding to each group of spectral samples and obtaining the trained deep learning model, the method further includes:

[0012] According to the pump-probe delay, the spectrum samples in the spectrum data set are divided into corresponding spectrum sample sets, and different spectrum sample sets correspond to different pump-probe delays;

[0013] Determine the mean and variance of each set of spectral samples;

[0014] Based on the mean value and variance of each spectral sample set, the spectral samples in each spectral sample set are standardized to obtain a standardized spectral data set.

[0015] In a possible implementation manner, obtaining a spectral dataset includes:

[0016] Acquire multiple groups of simulation spectra through simulation;

[0017] White noise and windowing are performed on multiple groups of simulated spectra to obtain multiple groups of spectrum samples.

[0018] In one possible implementation, when the sample to be tested is a low-dimensional material, the sample to be tested is disposed on a substrate, and the low-dimensional material is a material having a thickness of nanometers; when the sample to be tested is a bulk material, the sample to be tested is disposed under the needle tip as a substrate, and the bulk material is a material having a thickness exceeding nanometers.

[0019] In a possible implementation, the needle tip and / or the substrate has a pumping effect.

[0020] In a second aspect, the present application provides a material information acquisition device based on near-field optical testing, comprising:

[0021] A pump and probe light introduction module is used to introduce pump light to focus below the needle tip of the near-field optical microscope and introduce probe light to irradiate below the needle tip, and a sample to be measured is arranged below the needle tip;

[0022] A spectrum acquisition module is used to measure a set of spectra of the sample under test by a near-field optical microscope under different pump-probe delays, wherein the pump-probe delay is used to indicate the time interval between the moment of introducing the pump light and the moment of introducing the probe light, wherein a set of spectra includes multiple spectra, and different spectra in a set of spectra correspond to different pump-probe delays;

[0023] The material information acquisition module is used to input a set of spectra of the tested samples into the deep learning model to obtain the material information output by the deep learning model. The deep learning model is used to predict the material information of the tested samples based on a set of spectra of the tested samples.

[0024] In a third aspect, the present application provides an electronic device comprising: at least one memory for storing programs; and at least one processor for executing the programs stored in the memory. When the program stored in the memory is executed, the processor is used to execute the method described in the first aspect or any possible implementation of the first aspect.

[0025] In a fourth aspect, the present application provides a computer-readable storage medium, which stores a computer program. When the computer program runs on a processor, the processor executes the method described in the first aspect or any possible implementation of the first aspect.

[0026] In a fifth aspect, the present application provides a computer program product. When the computer program product runs on a processor, the processor executes the method described in the first aspect or any possible implementation of the first aspect.

[0027] In general, the above technical solutions conceived by this application have the following beneficial effects compared with the prior art:

[0028] Under the action of pump light in the band corresponding to the material properties of the sample under test, non-equilibrium carriers will be generated in the system composed of the needle tip, the sample under test and the substrate, exciting photo-induced polaritons, so that a response significantly different from the equilibrium state (without pump light) can be obtained in the detection light band. The measured spectrum can characterize the material properties of the sample under test. At the same time, by using the time delay between the pump light and the detection light, the spectrum generated by the photo-induced polaritons at different delays can be detected (the above response will change with the time delay between the pump light and the detection light), and multiple delay-related spectra characterizing the properties of the sample can be obtained. On the one hand, a set of time-delay-related spectra (including multiple spectra) can be obtained in one test to ensure test efficiency. On the other hand, the measured spectrum contains information on the time dimension of polaritons, and the characterization is more comprehensive, providing data support for the accurate prediction of the deep learning model. Then, a set of spectra obtained by the near-field optical test is input into the deep learning model, and the fitting ability of the deep learning model is used to extract and summarize features from multiple spectra, which can accurately predict the material information of the sample under test. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 It is a flow chart of a method for acquiring material information based on near-field optical testing provided in an embodiment of the present application;

[0030] Figure 2 is a schematic diagram of a scene of a near-field optical test provided in an embodiment of the present application;

[0031] Figure 3 is a schematic diagram of the spectrum under different delays provided in the embodiments of the present application;

[0032] Figure 4 It is a schematic diagram of the process of predicting the presence or absence of thin layer molecules using the deep learning model provided in the embodiment of the present application;

[0033] Figure 5 It is one of the schematic diagrams of the prediction results provided in the embodiments of the present application;

[0034] Figure 6 This is the second schematic diagram of the prediction results provided in the embodiment of the present application;

[0035] Figure 7 is a schematic structural diagram of a material information acquisition device based on near-field optical testing provided in an embodiment of the present application;

[0036] Figure 8 It is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0037] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0038] The term "and / or" in this article is a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. The symbol " / " in this article indicates that the associated objects are in an or relationship, for example, A / B means A or B.

[0039] The terms "first" and "second" in the specification and claims herein are used to distinguish different objects rather than to describe a specific order of objects. For example, a first delay and a second delay are used to distinguish different delays rather than to describe a specific order of delays.

[0040] In the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0041] In the description of the embodiments of the present application, unless otherwise specified, "multiple" means two or more than two. For example, multiple processing units refer to two or more processing units, etc.; multiple elements refer to two or more elements, etc.

[0042] The embodiments of the present application are described below in conjunction with the drawings in the embodiments of the present application.

[0043] Figure 1 is a flow chart of a method for obtaining material information based on near-field optical testing provided in an embodiment of the present application, such as Figure 1 As shown, the method includes the following steps S101, S102 and S103.

[0044] Step S101 , introducing pump light to focus below the needle tip of the near-field optical microscope and introducing detection light to illuminate below the needle tip, and a sample to be measured is arranged below the needle tip.

[0045] Step S102, under different pump-probe delays, a group of spectra of the sample under test are measured by a near-field optical microscope (under one delay, one spectrum can be obtained, and by changing the delay through an ultrafast near-field experiment, a group of spectra can be obtained). The pump-probe delay is used to indicate the time interval between the moment of introducing the pump light and the moment of introducing the probe light. A group of spectra includes multiple spectra, and different spectra in a group of spectra correspond to different pump-probe delays.

[0046] Specifically, by utilizing the time delay between the pump light and the detection light, the spectra generated by the photoinduced polaritons at different delays can be detected, thereby obtaining a large amount of delay-related spectral data characterizing the properties of the sample.

[0047] Moreover, the spectral data not only contains the information on the evolution of polaritons at different delays, but also contains the information on the evolution of the coupling between polaritons and molecular characteristic vibrations at different delays, which can comprehensively describe the material information from the time dimension.

[0048] Therefore, near-field pump-probe can obtain multiple (e.g., hundreds) of delay-related data through a single test, and the spectral data contains information on the time dimension of polaritons, providing a more comprehensive characterization.

[0049] It should be noted that the pump light and probe light used in this application are not limited to any specific optical band. They only need to match the material properties to achieve a good pumping effect, that is, the pump light can cause the material to generate non-equilibrium carriers, and the probe light can measure an optical response that is significantly different from the equilibrium state (when no pump light is added). This application does not limit the test conditions such as the needle tip vibration frequency, amplitude, and spectral order.

[0050] Step S103, input a set of spectra of the tested sample into the deep learning model to obtain material information output by the deep learning model. The deep learning model is used to predict the material information of the tested sample based on the set of spectra of the tested sample.

[0051] Exemplarily, the material information of the sample under test may be the presence or absence of the sample under test (for example, if the sample under test is a low-dimensional material, determine whether the low-dimensional material exists on the substrate), the carrier concentration of the sample under test, or the dielectric constant of the sample under test, etc.

[0052] Specifically, the spectral data (with or without preprocessing) obtained from near-field optical testing is used as input data for the deep learning algorithm to predict the required material information, such as the presence or absence of organic molecules and the carrier concentration of semiconductor materials.

[0053] Inputting large amounts of comprehensive spectral data into a deep learning model can fully utilize the deep learning model's ability to extract and summarize features from large amounts of complex data, making the deep learning model more robust and ensuring the accuracy of the deep learning model in predicting material information.

[0054] The specific network structure used by the deep learning model can be selected according to the complexity of the training data set and the difficulty of the prediction task. For example, for the classification task of predicting whether there are organic molecules in the sample, the complexity of the prediction task is relatively low, and the deep learning model can be a multi-layer perceptron. For example, for the task of predicting the dielectric constant of the sample based on spectral data, the complexity of the prediction task is relatively high, and the deep learning model can be a convolutional neural network.

[0055] Here, the working principle of the above method is described in detail. Under the action of pump light in the band corresponding to the material properties of the sample under test, non-equilibrium carriers will be generated in the system composed of the needle tip, the sample under test and the substrate, and photo-induced polaritons will be excited, so that a response significantly different from the equilibrium state (without pump light) can be obtained in the detection light band. The measured spectrum can characterize the material properties of the sample under test. At the same time, by using the time delay between the pump light and the detection light, the spectrum generated by the photo-induced polaritons at different delays can be detected (the above response will change with the time delay between the pump light and the detection light), and multiple delay-related spectra characterizing the properties of the sample can be obtained. On the one hand, a set of time-delay-related spectra (including multiple spectra) can be obtained in one test to ensure the test efficiency. On the other hand, the measured spectrum contains information on the time dimension of polaritons, and the characterization is more comprehensive, providing data support for the accurate prediction of the deep learning model. Then, a set of spectra obtained by the near-field optical test is input into the deep learning model, and the fitting ability of the deep learning model is used to extract and summarize features from multiple spectra, which can accurately predict the material information of the sample under test.

[0056] Therefore, by combining the technical means of measuring the spectrum of the sample under different pump-probe delays with the technical means of using deep learning models for prediction, the efficiency of the test can be guaranteed while ensuring the accuracy of the prediction, and the material information of the sample under test can be efficiently and accurately obtained using near-field optical signals.

[0057] In one possible implementation, the deep learning model is obtained through the following steps:

[0058] A spectral dataset is obtained, where the spectral dataset includes multiple groups of spectral samples and labels corresponding to each group of spectral samples. Any group of spectral samples is obtained under multiple pump-probe delays during a near-field optical test (one spectral sample can be obtained under one pump-probe delay, and multiple spectral samples can be obtained under multiple pump-probe delays, and these multiple spectral samples constitute a group of spectral samples). The label is used to indicate the material information represented by the corresponding group of spectral samples. Based on the multiple groups of spectral samples and the labels corresponding to each group of spectral samples, an initial deep learning model is trained in a supervised manner to obtain a trained deep learning model.

[0059] It can be understood that, compared with the traditional method of fitting spectral data with a needle tip model in the prior art, the method of training a deep learning model based on multiple groups of spectral samples provided in the present application can, on the one hand, avoid the cumbersome fitting process of the traditional method (the traditional method requires manual parameter adjustment and fitting of each data line), and on the other hand, obtain features that are difficult to obtain with the traditional method. For example, when the thickness of the material to be tested is reduced to a monolayer (that is, when the material to be tested is a low-dimensional material), the spectral detection sensitivity of SNOM will be significantly weakened. When the effective signal is relatively weak and is masked by the detection noise, it is difficult for the traditional fitting method to effectively analyze the effective signal. However, through the training method provided in the present application, the deep learning model can be used to extract and summarize features from a large amount of complex data. The model has a powerful ability, so that the trained deep learning model can still effectively analyze the effective signal when the effective signal is relatively weak, thereby effectively detecting the weak near-field optical signals of ultra-thin trace samples and efficiently and accurately obtaining material information therefrom.

[0060] Therefore, the method provided in this application for training a deep learning model based on multiple groups of spectral samples saves manpower and the conclusions are more objective and accurate.

[0061] Optionally, before the initial deep learning model is trained in a supervised manner based on multiple groups of spectral samples and labels corresponding to each group of spectral samples, and before the trained deep learning model is obtained, the method further includes: dividing the spectral samples in the spectral data set into corresponding spectral sample sets according to pump-probe delay, and different spectral sample sets correspond to different pump-probe delays; determining the mean and variance of each spectral sample set; and standardizing the spectral samples in each spectral sample set based on the mean and variance of each spectral sample set (subtracting the mean of the data in each set from the data points in the set, and then dividing the result by the variance of the data in the set), obtaining each standardized spectral sample set, and combining each standardized spectral sample set to obtain a standardized spectral data set.

[0062] It is understandable that standardizing data is a preprocessing method, which refers to the technology of converting data into a distribution with a mean of 0 and a standard deviation of 1. According to the characteristics of the data set, in the preprocessing standardization process, the data points in each set are subtracted from the mean value of the data in the set, and then divided by the variance. This processing method is different from the conventional standardization method (the conventional method uses the mean and variance of the entire data set). Each set of data is standardized separately, and the calculation uses the mean and variance of each set, not the mean and variance of the entire data set.

[0063] In principle, the concentration of photo-induced carriers excited at each time delay is different, resulting in differences in the linear shape and intensity of the optical signal generated at different time delays. By standardizing each set of data separately, the features can be maximized, making the model better at learning, accelerating convergence and improving the generalization ability of the model.

[0064] Optionally, the obtaining of the spectral data set includes: obtaining a plurality of groups of simulated spectra through simulation; and adding white noise and performing windowing processing on the plurality of groups of simulated spectra to obtain a plurality of groups of spectral samples.

[0065] For the windowing process, specifically, a window function may be selected, and the window function (such as a rectangular window, a Hanning window, a Hamming window, etc.) may be multiplied by the original data.

[0066] In one possible implementation, when the sample to be tested is a low-dimensional material, the sample to be tested is arranged on a substrate (the sample to be tested and the substrate are both arranged below the needle tip), and the low-dimensional material is a material with a thickness of nanometers; when the sample to be tested is a bulk material, the sample to be tested is arranged as a substrate below the needle tip, and the bulk material is a material with a thickness exceeding nanometers.

[0067] Optionally, the tip and / or substrate has a pumping effect. It is understood that when the experimental conditions such as pump light, probe light, etc. match the material properties, good pumping results can be achieved. Near-field optical testing can adopt any of the following three experimental setups:

[0068] (1) The tip material is a material that can be pumped well under experimental conditions, and the base material does not have a pumping effect;

[0069] (2) The tip material does not have a pumping effect, but the substrate is a material that can be pumped well;

[0070] (3) Both the tip material and the substrate material can be pumped well under experimental conditions.

[0071] The specific experimental setup to be used can be determined based on the experimental conditions and the properties of the sample being tested. For example, if the sample being tested is a bulk material, it can be used as a substrate, and then the experimental setup to be used can be selected based on whether it can be well pumped; if the sample being tested is a low-dimensional material, it can be placed on a pumpable or non-pumpable substrate, and then the experimental setup to be used can be selected based on the properties of the substrate.

[0072] At least one of the tip and the substrate has a pumping effect. If both the tip and the substrate have a pumping effect, a nanoscale optical cavity is naturally formed between the tip and the substrate, allowing light to be localized between the tip and the substrate to a greater extent, thereby collecting more accurate high-resolution sample information.

[0073] Exemplarily, a silicon needle tip is a commonly used near-field needle tip, which is made of silicon. It is known that silicon is a typical indirect bandgap material. The pump light is set to a femtosecond laser with a central wavelength of 800nm, which can excite photoinduced carriers on the surface of the needle tip. The detection light is set to a broadband pulse laser of 0.8-7.2THz, which can detect the light response changes caused by the photoinduced carriers that are significantly different from those when not pumped. In practical applications, the selection of the detection band also needs to consider the material of the sample being tested. For example, if the sample being tested is a low-dimensional material Bi2Se3, which is known to have an obvious polariton response at 2-3THz, the detection light band can be set in the terahertz band.

[0074] Therefore, when experimental conditions such as pump light and probe light match the material properties, good experimental results can be achieved. The pump light needs to take into account the material properties of the needle tip and the substrate, and pumping can be performed when the band gap conditions of the material are met; the probe light needs to consider both the sample under test and the pump material. The sample under test has a characteristic response in the detection band, and the pump material has an obvious response in the detection band after being pumped.

[0075] The material information acquisition method based on near-field optical testing provided by the present application is explained below through an example.

[0076] The near-field optical test equipment used in this example is a scattering near-field optical microscope. The pump light is a laser with a central wavelength of 780nm, and the infrared light (as the detection light) is a pulsed light of 8μm-16μm. The near-field needle tip used in the test is a platinum needle with a needle tip vibration frequency of 270kHz and an amplitude of 40nm. The spectral signals mentioned below are all second-order near-field signals.

[0077] Figure 2 is a schematic diagram of a near-field optical test scenario provided in an embodiment of the present application, such as Figure 2 As shown, the introduced pump light is focused below the needle tip of the near-field optical microscope and the introduced detection light is irradiated below the needle tip. A sample to be tested is arranged below the needle tip. The sample to be tested is a thin layer of polymer polyethylene terephthalate (PET) covered on an indium arsenide (InAs) substrate.

[0078] The detection light is focused by a parabolic mirror to the bottom of the atomic force microscope (AFM) tip working in tapping mode. After the near-field light field on the surface of the sample is scattered to the far field by the tip, the scattered light is reflected back to the Michelson interferometer optical path through the parabolic mirror and finally collected by the mercury cadmium telluride detector, thus constructing a nano pump-probe spectroscopy system.

[0079] Figure 3 is a schematic diagram of the spectrum under different delays provided in the embodiment of the present application, such as Figure 3 As shown, the near-field pump-probe technique is used at different pump-probe delays (such as Figure 3The first delay, second delay and third delay in the PET of different thicknesses (such as Figure 3 Samples with 1nm thickness, 3nm thickness and 6nm thickness) were tested, and the first delay was smaller than the second delay, and the second delay was smaller than the third delay.

[0080] When the PET thickness is less than 3nm, the weak molecular signal is masked by noise, and almost only the spectral changes at different delays generated by InAs polaritons can be observed. When the PET thickness is less than 3nm, it is impossible to determine whether there are molecules on the substrate using traditional methods, and the infrared molecular peak signal is very unclear ( Figure 3 middle vertical line position).

[0081] In response to this situation, the near-field spectrum is used as input data and a neural network method is used to predict whether there are PET molecules on the InAs substrate. Figure 4 is a schematic diagram of the process of predicting the presence or absence of thin layer molecules using the deep learning model provided in the embodiment of the present application, such as Figure 4 As shown in the figure, a multilayer perceptron is used as the neural network in this example. The input data is the normalized near-field spectrum, and the output is the prediction result of whether there is a thin layer or not, that is, the percentage of molecules with (P0) and without (P1).

[0082] The training set is obtained by adding white noise and windowing to 20,000 sets of simulated spectral data. The “addition of white noise and windowing” makes the simulated data as close to the experimental data as possible. Among them, for a set of simulated spectral data, it can be a set of simulated spectral data consisting of multiple simulated spectra with sample information obtained under different pump and detection time delays during a simulation test.

[0083] Figure 5 is one of the schematic diagrams of the prediction results provided in the embodiments of the present application, Figure 6 This is the second schematic diagram of the prediction results provided in the embodiment of the present application. Figure 5 This is the predicted result when a 1nm thick organic molecule (PET) is placed under the needle tip. Figure 6 This is the prediction result when there is no organic molecule (PET) under the needle tip. After training, the multilayer perceptron can effectively predict the presence or absence of molecules. It should be noted that the neural network type, training set generation and processing method (simulation, experiment or other methods), and input spectrum preprocessing method are not limited to this example.

[0084] The material information acquisition device based on near-field optical testing provided in the present application is described below. The material information acquisition device based on near-field optical testing described below and the material information acquisition method based on near-field optical testing described above can be referenced to each other.

[0085] Figure 7is a schematic diagram of the structure of a material information acquisition device based on near-field optical testing provided in an embodiment of the present application, such as Figure 7 As shown, the device comprises: a pump-probe light introduction module 10, a spectrum acquisition module 20 and a material information acquisition module 30. Wherein:

[0086] A pump and probe light introduction module 10 is used to introduce pump light to focus below the needle tip of the near-field optical microscope and introduce probe light to irradiate below the needle tip, and a sample to be measured is arranged below the needle tip;

[0087] A spectrum acquisition module 20 is used to measure a set of spectra of the sample under test by a near-field optical microscope under different pump-probe delays, wherein the pump-probe delay is used to indicate the time interval between the moment of introducing the pump light and the moment of introducing the probe light, wherein a set of spectra includes a plurality of spectra, and different spectra in a set of spectra correspond to different pump-probe delays;

[0088] The material information acquisition module 30 is used to input a set of spectra of the tested sample into the deep learning model to obtain the material information output by the deep learning model. The deep learning model is used to predict the material information of the tested sample based on a set of spectra of the tested sample.

[0089] It can be understood that the detailed functional implementation of each of the above-mentioned units / modules can be found in the introduction of the aforementioned method embodiment, and will not be repeated here.

[0090] It should be understood that the above-mentioned device is used to execute the method in the above-mentioned embodiment. The implementation principle and technical effect of the corresponding program module in the device are similar to those described in the above-mentioned method. The working process of the device can refer to the corresponding process in the above-mentioned method, which will not be repeated here.

[0091] Based on the method in the above embodiment, an embodiment of the present application provides an electronic device, Figure 8 is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application, such as Figure 8 As shown, the electronic device may include: a processor (Processor) 810, a communication interface (Communications Interface) 820, a memory (Memory) 830 and a communication bus 840, wherein the processor 810, the communication interface 820, and the memory 830 communicate with each other through the communication bus 840. The processor 810 can call the logic instructions in the memory 830 to execute the method in the above embodiment.

[0092] In addition, the logic instructions in the above-mentioned memory 830 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application.

[0093] Based on the method in the above embodiment, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program runs on a processor, the processor executes the method in the above embodiment.

[0094] Based on the method in the above embodiment, an embodiment of the present application provides a computer program product. When the computer program product runs on a processor, the processor executes the method in the above embodiment.

[0095] It is understandable that the processor in the embodiment of the present application may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. The general-purpose processor may be a microprocessor or any conventional processor.

[0096] The method steps in the embodiments of the present application can be implemented by hardware or by a processor executing software instructions. The software instructions can be composed of corresponding software modules, and the software modules can be stored in random access memory (RAM), flash memory, read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, mobile hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an ASIC.

[0097] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented by software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer instructions may be transmitted from a website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)), etc.

[0098] It should be understood that the various numerical numbers involved in the embodiments of the present application are only used for the convenience of description and are not used to limit the scope of the embodiments of the present application.

[0099] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A method for obtaining material information based on near-field optical testing, characterized in that: include: Introducing pump light to focus below the needle tip of the near-field optical microscope and introducing detection light to illuminate below the needle tip, wherein a sample to be measured is arranged below the needle tip; Under different pump-probe delays, a group of spectra of the sample under test is measured by the near-field optical microscope, wherein the pump-probe delay is used to indicate the time interval between the moment of introducing the pump light and the moment of introducing the probe light, a group of spectra includes a plurality of spectra, and different spectra in a group of spectra correspond to different pump-probe delays; Inputting a set of spectra of the sample to be tested into a deep learning model to obtain material information output by the deep learning model, wherein the deep learning model is used to predict the material information of the sample to be tested based on the set of spectra of the sample to be tested; The deep learning model is obtained through the following steps: Acquire a spectral data set, wherein the spectral data set includes multiple groups of spectral samples and labels corresponding to each group of spectral samples, wherein any group of spectral samples is acquired under multiple pump-probe delays during a near-field optical test, and the labels are used to indicate material information represented by the corresponding group of spectral samples; Based on multiple groups of spectral samples and labels corresponding to each group of spectral samples, an initial deep learning model is trained in a supervised manner to obtain a trained deep learning model.

2. The material information acquisition method based on near-field optical testing according to claim 1, characterized in that: Before an initial deep learning model is trained in a supervised manner based on multiple groups of spectral samples and labels corresponding to each group of spectral samples and obtaining the trained deep learning model, the method further includes: According to the pump-probe delay, the spectrum samples in the spectrum data set are divided into corresponding spectrum sample sets, and different spectrum sample sets correspond to different pump-probe delays; Determine the mean and variance of each set of spectral samples; Based on the average value and variance of each spectral sample set, the spectral samples in each spectral sample set are standardized to obtain a standardized spectral data set.

3. The material information acquisition method based on near-field optical testing according to claim 1 is characterized in that: The step of acquiring a spectral data set comprises: Acquire multiple groups of simulation spectra through simulation; White noise and windowing are performed on the multiple groups of simulated spectra to obtain the multiple groups of spectrum samples.

4. The material information acquisition method based on near-field optical testing according to any one of claims 1 to 3, characterized in that: When the sample to be tested is a low-dimensional material, the sample to be tested is arranged on a substrate, and the low-dimensional material is a material with a thickness of nanometers; when the sample to be tested is a bulk material, the sample to be tested is arranged as a substrate under the needle tip, and the bulk material is a material with a thickness exceeding nanometers.

5. The material information acquisition method based on near-field optical testing according to claim 4 is characterized in that: The needle tip and / or the substrate has a pumping effect.

6. A material information acquisition device based on near-field optical testing, characterized in that: include: A pump and probe light introduction module, used to introduce pump light to focus below the needle tip of the near-field optical microscope and introduce probe light to irradiate below the needle tip, and a sample to be measured is arranged below the needle tip; A spectrum acquisition module, used for measuring a set of spectra of the sample under test by the near-field optical microscope under different pump-probe delays, wherein the pump-probe delay is used to indicate the time interval between the moment of introducing the pump light and the moment of introducing the probe light, wherein a set of spectra includes a plurality of spectra, and different spectra in a set of spectra correspond to different pump-probe delays; A material information acquisition module, used to input a set of spectra of the sample to be tested into a deep learning model to obtain material information output by the deep learning model, wherein the deep learning model is used to predict the material information of the sample to be tested based on the set of spectra of the sample to be tested; The deep learning model is obtained through the following steps: Acquire a spectral data set, wherein the spectral data set includes multiple groups of spectral samples and labels corresponding to each group of spectral samples, wherein any group of spectral samples is acquired under multiple pump-probe delays during a near-field optical test, and the labels are used to indicate material information represented by the corresponding group of spectral samples; Based on multiple groups of spectral samples and labels corresponding to each group of spectral samples, an initial deep learning model is trained in a supervised manner to obtain a trained deep learning model.

7. An electronic device, characterized in that: include: at least one memory for storing a computer program; At least one processor is used to execute the program stored in the memory. When the program stored in the memory is executed, the processor is used to execute the method according to any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program runs on a processor, the processor is caused to execute the method according to any one of claims 1 to 5.

9. A computer program product, characterized in that When the computer program product runs on a processor, the processor is caused to execute the method according to any one of claims 1 to 5.

Images