Methods for constructing an optimized overall model of nano-infrared signals and methods and devices for their detection.

Abstract

This application provides a method for constructing an optimized overall model for nano-infrared signals, as well as a detection method and apparatus. The method includes: constructing a local electromagnetic field model based on the sample's environment and a defined illumination path; wherein the local electromagnetic field model is calculated using a two-dimensional axisymmetric structure; establishing corresponding photodynamic sub-models for various photodynamic forces generated by the sample, and forming a multi-physics coupling model through multiple photodynamic sub-models; training the model using an MLP network as the basic architecture to obtain a deep neural network model; and constructing the overall optimized model for nano-infrared signals using the local electromagnetic field model, the multi-physics coupling model, and the deep neural network model. This application, by incorporating the local electromagnetic field model, the multi-physics coupling model, and the deep neural network model into the overall optimized model for nano-infrared signals, can systematically improve sample detection performance and enhance the detection sensitivity of the nano-infrared system.

Description

Technical Field

[0001] This application relates to the field of infrared detection, and more specifically, to a method for constructing an optimized overall model of nano-infrared signals and a detection method and apparatus. Background Technology

[0002] Nanoscale infrared technology combines infrared spectroscopy with scanning probe microscopy. It measures the absorption of infrared light by a sample using a nanoscale probe, thereby obtaining high spatial resolution infrared images and spectra of the sample surface. Nanoscale infrared detection enables the correlated imaging of multimodal physicochemical information, including morphology, chemistry, mechanics, and electricity, at the nanoscale, improving spatial resolution to below 10 nm, demonstrating significant advantages and promising prospects in important surface and interface research. However, current nanoscale infrared detection techniques suffer from relatively poor detection sensitivity, especially in high-density media such as water. Summary of the Invention

[0003] In view of this, the purpose of this application is to provide a method for constructing an optimized overall model of nano-infrared signals and a detection method and apparatus, which can improve the sensitivity of sample detection.

[0004] In a first aspect, embodiments of this application provide a method for constructing a comprehensive optimization model for nano-infrared signals, comprising: constructing a local electromagnetic field model based on the sample's environment and a set illumination path; wherein the local electromagnetic field model is calculated using a two-dimensional axisymmetric structure; establishing corresponding photodynamic sub-models for the various photodynamic forces generated by the sample, and forming a multi-physics coupling model through multiple photodynamic sub-models; training the model using an MLP network as the basic architecture to obtain a deep neural network model; and using the local electromagnetic field model, the multi-physics coupling model, and the deep neural network model together to constitute the comprehensive optimization model for nano-infrared signals.

[0005] In the above implementation process, the local electromagnetic field model is used to optimize parameters such as sample detection parameters, probe detection parameters, and pulse detection parameters; the multi-physics coupling model is used to optimize the pulse excitation method and improve sample detection sensitivity; and the deep neural network is used to extract the effective infrared signal from the probe motion curve in real time, improving the signal-to-noise ratio of the sample spectrum. By setting the local electromagnetic field model, multi-physics coupling model, and deep neural network model in the overall optimization model of nano-infrared signals, the performance of sample detection can be systematically improved, and the sensitivity of the overall optimization model of nano-infrared signals to sample detection can be enhanced.

[0006] In one embodiment, constructing a local electromagnetic field model based on the sample's environment and a set illumination path includes: calculating the background field without a needle tip under polarized light excitation according to Fresnel's equations; calculating the scattering field with a needle tip present based on a first target geometric model, material parameters, and the background field, and determining the sum of the background field and the scattering field; wherein the first target geometric model is a two-dimensional axisymmetric model of the complete needle-sample structure; wherein the sum of the background field and the scattering field is an enhanced electric field, and the enhanced electric field is configured to enhance the signal; the square of the enhanced electric field divided by the square of the background field yields the infrared enhancement factor.

[0007] In the above implementation process, compared with three-dimensional models, two-dimensional models have the advantages of simple structure, easy calculation, and fast computational efficiency. When creating a local electromagnetic field model, setting the first target geometric model as a two-dimensional model can improve the computational efficiency of the local electromagnetic field model. In addition, when creating the local electromagnetic field, by setting and changing the material parameters and the electromagnetic wave frequency domain physical field, the infrared enhancement factor output by the local electromagnetic field can be used to determine the optimal sample detection parameters, probe detection parameters, and other parameters, thereby optimizing the sample detection parameters, probe parameters, and other parameters, and improving the sensitivity of the local electromagnetic field model.

[0008] In one embodiment, the photoforce includes photogradient force, photoacoustic force, and photothermal force; the photoforce sub-model includes a photogradient force sub-model, a photoacoustic force sub-model, and a photothermal force sub-model; establishing corresponding photoforce sub-models for the various photoforces generated by the sample includes: modeling the photogradient force based on an electric dipole model to obtain the photogradient force sub-model; modeling the photoacoustic force through COMSOL acoustic simulation to obtain the photoacoustic force sub-model; and modeling the thermal expansion force based on a Hertz contact model to obtain the photothermal force sub-model.

[0009] In the above implementation process, based on the characteristics of various optical forces of the sample, the corresponding model creation method is selected to create the corresponding optical force sub-model, thereby constructing a more accurate optical force sub-model and improving the accuracy of various optical force calculations of the sample.

[0010] In one embodiment, modeling the optical gradient force based on the electric dipole model to obtain the optical gradient force quantum model includes: calculating the background field without a needle tip under polarized light excitation according to Fresnel equations; calculating the scattering field with a needle tip present based on a second target geometric model, material parameters, and the background field, and determining the sum of the background field and the scattering field; wherein the second target geometric model is a three-dimensional model of the complete needle-sample structure; and wherein the sum of the background field and the scattering field is the enhanced electric field.

[0011] In the above implementation process, the electromagnetic field model based on the two-dimensional axisymmetric structure cannot perform complete tip integration calculations. Modeling the optical gradient force using an electric dipole model, and thus obtaining an optical gradient force sub-model, can reduce the solution accuracy, decrease the computational load, and improve the computational efficiency of the optical gradient force sub-model. Furthermore, compared to the two-dimensional model, the three-dimensional model has higher computational accuracy; by replacing the geometric model in the optical gradient force sub-model with a three-dimensional model, the computational accuracy of the optical gradient force sub-model can be improved.

[0012] In one embodiment, the step of modeling the photoacoustic force using COMSOL acoustic simulation to obtain the photoacoustic force sub-model includes: constructing the photoacoustic force sub-model by distributing all physical fields in the overall space of COMSOL; wherein the photoacoustic force sub-model is configured to determine the photoacoustic force by integrating the sound pressure of the entire surface of the probe; wherein the overall spatial distribution of all physical fields from the outside to the inside is as follows: a perfectly matched layer, a transient pressure acoustic physical field, and a transient thermo-viscous acoustic physical field, and the three-dimensional structure of the probe and the sample is set within the transient thermo-viscous acoustic physical field.

[0013] In the above implementation process, when creating the photoacoustic sub-model, all physical fields are distributed in COMSOL to simulate the actual physical field of the sample. By using the actual physical field of the sample to simulate the photoacoustic sub-model, the accuracy of the photoacoustic sub-model can be improved.

[0014] In one embodiment, the step of modeling the thermal expansion force based on the Hertz contact model to obtain the photothermal sub-model includes: determining the planar and spherical structures of the Hertz contact model according to the probe tip and the sample; using the Hertz contact model composed of the planar and spherical structures as the photothermal sub-model; wherein the photothermal sub-model is configured to calculate the thermal expansion force based on the transient thermal expansion distribution.

[0015] In the above implementation process, since photothermal force arises when the probe tip contacts the sample, the sample's photothermal expansion causes an upward repulsive force on the probe tip, and the Hertz contact model is used to describe the deformation and stress distribution of the two elastic bodies near the contact point, constructing the photothermal sub-model using the Hertz model can improve the accuracy of the photothermal sub-model.

[0016] In one embodiment, training a deep neural network model based on an MLP network architecture includes: preprocessing the force curves of the collected samples to construct a dataset; using the noisy signals in the dataset as input to the MLP network, using the resonant frequency band integral values ​​of the clean signals in the dataset as output to the MLP network, and training the MLP network to obtain the deep neural network model.

[0017] In the above implementation process, since MLP networks have advantages such as simple structure and flexible design, and the network structure can be adjusted according to the needs of specific problems, training deep neural network models using MLP networks as the architecture can simplify the structure of deep neural network models, improve the computational efficiency of deep neural network models, and increase the flexibility of the deep neural network models.

[0018] Secondly, embodiments of this application also provide a nano-infrared detection method, comprising: applying a nano-infrared signal optimization model to the first aspect above, or any possible implementation of the first aspect, wherein the nano-infrared signal optimization model includes a local electromagnetic field model, a multi-physics coupling model, and a deep neural network model; the method comprising: adjusting sample detection parameters, probe detection parameters, and pulse detection parameters through the local electromagnetic field model during sample detection by the nano-infrared detection device, and outputting the force curve of the sample in real time; calculating the photoelectric force values ​​corresponding to various photoelectric forces of the sample through the multi-physics coupling model; inputting the force curve into the deep neural network model, and demodulating a high signal-to-noise ratio infrared signal through the deep neural network model; wherein the infrared signal is configured to detect the performance of the sample.

[0019] In the above implementation process, during nano-infrared detection, the probe detection parameters, sample detection parameters, and pulse detection parameters used for sample detection are first adjusted using a local electromagnetic field model. This generates an optimal infrared enhancement factor, optimizing the probe, sample, and pulse parameters. Then, various photomechanical forces are calculated using a multiphysics coupling model to select an appropriate pulse excitation method to suppress photomechanical interference from non-chemical signals, thereby improving sample detection sensitivity. Finally, a deep neural network model extracts the effective infrared signal from the force curve, achieving the reconstruction and restoration of high-dimensional feature information into a clean signal, thus improving the signal-to-noise ratio of the sample spectrum.

[0020] In one embodiment, adjusting the sample detection parameters, probe detection parameters, and pulse detection parameters through the local electromagnetic field model includes: inputting the probe parameters and sample parameters of the probe and the sample into the local electromagnetic field model to obtain an infrared enhancement factor; if the infrared enhancement factor is not the strongest infrared enhancement factor, replacing the probe parameters and the sample parameters, and re-inputting the replaced probe parameters and the sample parameters into the local electromagnetic field model until the infrared enhancement factor output by the local electromagnetic field model is the strongest infrared enhancement factor; wherein, the probe parameters corresponding to the strongest infrared enhancement factor are the probe detection parameters, and the sample parameters corresponding to the strongest infrared enhancement factor are the sample detection parameters; and / or if the infrared enhancement factor is not the strongest infrared enhancement factor, changing the infrared pulse angle of the local electromagnetic field model, and outputting the infrared enhancement factor through the local electromagnetic field model after changing the infrared pulse angle, until the infrared enhancement factor output by the local electromagnetic field model is the strongest infrared enhancement factor; wherein, the infrared pulse angle corresponding to the strongest infrared enhancement factor is the pulse detection parameter.

[0021] In the above implementation process, based on the adjusted sample parameters, probe parameters, or pulse detection parameters, and the corresponding output infrared enhancement factor, the sample parameters, probe parameters, or pulse detection parameters corresponding to the strongest infrared enhancement factor are determined as the sample detection parameters, probe detection parameters, or pulse detection parameters, thereby optimizing the sample parameters, probe parameters, or pulse detection parameters and improving the accuracy of nano-infrared detection.

[0022] In one embodiment, the photoforce includes photogradient force, photoacoustic force, and photothermal force; the calculation of the photoforce values ​​corresponding to the various photoforces of the sample through the multiphysics coupling model includes: inputting the electric and magnetic fields of the tip-sample structure at various points in space into the Maxwell stress tensor formula to determine the corresponding Maxwell stress tensor; calculating the photogradient force using the Maxwell stress tensor and Gauss's law; calculating the total absorbed power of the material based on the electric field distribution generated in space by the tip-sample structure calculated by the local electromagnetic field model and the corresponding material absorption power formula; using the total absorbed power as the heat source term in the sample heat transfer equation to determine the transient thermal expansion distribution; wherein the transient thermal expansion distribution is used to calculate the photoacoustic force and photothermal force.

[0023] In the above implementation process, when calculating various optical forces of the sample, the calculation is performed according to the corresponding calculation method for each optical force, which can improve the accuracy of various optical force calculations.

[0024] Thirdly, embodiments of this application also provide a nano-infrared detection device, comprising: a laser emitting unit, a probe, and a data processing unit; the laser emitting unit is used to emit laser light to a sample; the probe is used to measure the absorption of infrared light by the sample; the data processing unit is used to detect the performance of the sample according to the method of the first aspect above, or any possible implementation of the first aspect, or the second aspect above, or any possible implementation of the second aspect.

[0025] Fourthly, embodiments of this application also provide a data processing unit, including: a processor and a memory, wherein the memory stores machine-readable instructions executable by the processor, and when the data processing unit is running, the machine-readable instructions are executed by the processor to perform the steps of the method described in the first aspect above, or any possible implementation of the first aspect.

[0026] Fifthly, embodiments of this application also provide a computer-readable storage medium storing a computer program that, when executed by a processor, performs the steps of the method described in the first aspect above, or any possible implementation of the first aspect, or the second aspect above, or any possible implementation of the second aspect.

[0027] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, specific embodiments are described below in conjunction with the accompanying drawings. Attached Figure Description

[0028] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 This is a schematic diagram of the nano-infrared detection device provided in the embodiments of this application;

[0030] Figure 2 A block diagram of the data processing unit provided in the embodiments of this application;

[0031] Figure 3 A flowchart illustrating the method for constructing a nano-infrared signal optimization overall model provided in this application embodiment;

[0032] Figure 4 A schematic diagram of the first target geometric model of the local electromagnetic field model provided in the embodiments of this application;

[0033] Figure 5This is a schematic diagram of the overall optimization model for nano-infrared signals provided in the embodiments of this application;

[0034] Figure 6 The needle tip and sample provided in the embodiments of this application are approximately electric dipoles.

[0035] Figure 7 This is an example diagram illustrating the overall spatial distribution of all physical fields provided in the embodiments of this application.

[0036] Figure 8 This is a schematic diagram of the contact between the needle tip and the sample provided in an embodiment of this application;

[0037] Figure 9 A schematic diagram of a deep neural network model provided in an embodiment of this application;

[0038] Figure 10 A flowchart of the nano-infrared detection method provided in the embodiments of this application;

[0039] Figure 11 A schematic diagram of the functional modules of the nano-infrared signal optimization overall model construction device provided in the embodiments of this application;

[0040] Figure 12 This is a schematic diagram of the functional modules of the nano-infrared detection device provided in the embodiments of this application.

[0041] Figure descriptions: 100-Data processing unit, 111-Memory, 113-Processor, 200-Probe, 300-Laser emission unit, 400-Sample, 801-Construction module, 802-Establishment module, 803-Training module, 804-Construction module, 901-Adjustment module, 902-Calculation module, 903-Demodulation module. Detailed Implementation

[0042] The technical solutions in the embodiments of this application will now be described with reference to the accompanying drawings.

[0043] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0044] In recent years, the fields of nanomaterials, solar cells, lithium-ion batteries, and semiconductor materials have developed rapidly. The microscopic features that affect the performance of these materials are usually at the nanometer level. Nano-infrared technology has broken through the optical diffraction limit and has become an important characterization tool for studying the chemical properties of these nanomaterials.

[0045] The inventors of this application have discovered through long-term research that the sensitivity of nano-infrared detection is affected by the laser itself, background noise, and the environment, resulting in poor sensitivity. Especially in critical solid-liquid interface research, the detection sensitivity, spatial resolution, and detection speed of nano-infrared detection technology are still insufficient to meet the requirements of detecting rapid reaction processes due to the limitations of the solution environment.

[0046] In view of this, this application proposes a method for constructing a comprehensive optimization model for nano-infrared signals. A local electromagnetic field model is used to optimize parameters such as sample detection parameters, probe detection parameters, and pulse detection parameters. A multi-physics coupling model is used to optimize the pulse excitation method and improve sample detection sensitivity. A deep neural network is used to extract the effective infrared signal from the probe motion curve in real time, improving the signal-to-noise ratio of the sample spectrum. By setting the local electromagnetic field model, multi-physics coupling model, and deep neural network model in the comprehensive optimization model for nano-infrared signals, the performance of sample detection can be systematically improved, and the sensitivity of the comprehensive optimization model for nano-infrared signals to sample detection can be enhanced.

[0047] To facilitate understanding of this embodiment, a nano-infrared detection device for implementing the nano-infrared signal optimization model construction method and / or nano-infrared detection method disclosed in this application will first be described in detail.

[0048] like Figure 1 The diagram shown is a schematic of a nano-infrared detection device provided in an embodiment of this application, including: a laser emitting unit 300, a probe 200, and a data processing unit 100.

[0049] The laser emitting unit 300 is a device capable of generating infrared radiation, used to emit laser light towards the sample 400. For example, the laser emitting unit 300 can be a light-induced force nano-infrared microscope, a scattering near-field optical microscope, etc., and the laser emitting unit 300 can be selected according to the actual situation.

[0050] The probe 200 here is a tool that combines atomic force microscopy and infrared spectroscopy. The probe 200 is used to measure the absorption of infrared light by the sample 400.

[0051] Specifically, in the nano-infrared detection process, the tip of an atomic force microscope is used as the detector for the infrared spectrum. Infrared laser light is focused onto the tip of probe 200. When the laser light irradiates the surface of sample 400, the molecules in sample 400 absorb infrared light of a specific wavelength, causing local thermal expansion. The tip of probe 200 detects this minute change in thermal expansion and converts it into an electrical signal, thereby analyzing the infrared spectral information of sample 400.

[0052] The data processing unit 100 described above is used to construct an optimized overall model of nano-infrared signals according to the nano-infrared signal optimization overall model construction method in the embodiments of this application and / or to detect the performance of sample 400 according to the nano-infrared detection method in the embodiments of this application.

[0053] In one embodiment, the data processing unit 100 is connected to the probe 200 for data communication or interaction.

[0054] Optionally, the data processing unit 100 can be a web server, database server, personal computer (PC), tablet computer, smartphone, personal digital assistant (PDA), microcontroller, programmable controller, etc. The data processing unit 100 can be selected according to the actual situation.

[0055] To facilitate understanding of this embodiment, the data processing unit 100 that performs the nano-infrared signal optimization model construction method and / or nano-infrared detection method disclosed in this application embodiment will be described in detail below.

[0056] like Figure 2 The diagram shown is a block illustration of a data processing unit. The data processing unit 100 may include a memory 111 and a processor 113. Those skilled in the art will understand that... Figure 2 The structure shown is for illustrative purposes only and does not limit the structure of the data processing unit 100. For example, the data processing unit 100 may also include components that are larger than... Figure 2 The more or fewer components shown, or having the same Figure 2 The different configurations shown.

[0057] The aforementioned memory 111 and processor 113 are electrically connected to each other directly or indirectly to enable data transmission or interaction. For example, these components can be electrically connected to each other via one or more communication buses or signal lines. The aforementioned processor 113 is used to execute executable modules stored in the memory.

[0058] The memory 111 can be, but is not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), etc. The memory 111 stores programs, and the processor 113 executes these programs upon receiving execution instructions. The methods executed by the data processing unit 100, as defined in any embodiment of this application, can be applied to or implemented by the processor 113.

[0059] The aforementioned processor 113 may be an integrated circuit chip with signal processing capabilities. The processor 113 may be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it may also be a digital signal processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor may be a microprocessor or any conventional processor.

[0060] The data processing unit 100 in this embodiment can be used to execute various steps in the various methods provided in the embodiments of this application. The implementation process of the nano-infrared signal optimization overall model construction method and the nano-infrared detection method is described in detail below through several embodiments.

[0061] Please see Figure 3 This is a flowchart of the method for constructing the overall model for optimizing nano-infrared signals provided in the embodiments of this application. The following will describe... Figure 3 The specific process shown will be explained in detail.

[0062] Step S201: Construct a local electromagnetic field model based on the sample's environment and the set illumination path.

[0063] The local electromagnetic field model is calculated using a two-dimensional axisymmetric structure.

[0064] The sample here refers to the substance to be tested, such as solid samples, liquid samples, nanoparticles, etc. The sample can be selected according to the actual situation.

[0065] Optionally, the lighting path can be set as a top-side lighting path, a bottom lighting path, etc., and the lighting path can be selected according to the actual experimental environment.

[0066] For example, if the experimental environment is an air environment, the lighting path can be set to top-side lighting. If the experimental environment is a liquid environment, the lighting path can be set to bottom lighting.

[0067] In one embodiment, the local electromagnetic field model can be established using COMSOL Multiphysics, a simulation software based on the finite element method.

[0068] It should be understood that, in order to improve the computational speed of the model, such as Figure 4 As shown, when creating a local electromagnetic field model, a two-dimensional axisymmetric structure can be used instead of a three-dimensional model.

[0069] For example, if Figure 4 The sample detection environment for the two-dimensional axisymmetric structure shown is an air environment. In two-dimensional cylindrical coordinates (z, r), after rotation around the z-axis, it can be regarded as: cylinder (z≤0nm, set as polydimethylsiloxane material, i.e. the sample being tested), cylinder (z≥0.5nm, air material), and cone (z≥0.5nm, set as gold material, i.e. the probe tip).

[0070] The above-mentioned local electromagnetic field model configuration involves adjusting parameters such as sample detection parameters, probe detection parameters, and pulse detection parameters.

[0071] Step S202: Establish corresponding photodynamic sub-models for the various photodynamic forces generated by the sample, and form a multi-physics coupling model through multiple photodynamic sub-models.

[0072] The optical force here refers to the force generated when a laser interacts with a sample. Examples include optical gradient force, photoacoustic force, and photothermal force, which can be adjusted according to the actual situation.

[0073] It should be understood that the generation methods, principles, and meanings of each type of luminous force may differ, and consequently, their calculation methods may also differ. Therefore, for each type of luminous force, a corresponding luminous force sub-model can be established to calculate the corresponding luminous force index.

[0074] For example, a corresponding optical gradient force sub-model can be established for optical gradient force, a corresponding optical acoustic force sub-model can be established for optical acoustic force, and a corresponding optical thermal force sub-model can be established for optical thermal force, etc. The optical force sub-model can be adjusted according to the optical force required in practice.

[0075] Step S203: Train the model using the MLP network as the basic architecture to obtain a deep neural network model.

[0076] The MLP network is a multilayer perceptual network consisting of an input layer, hidden layers, and an output layer. Each neuron in an MLP network receives the output from the neuron in the previous layer, performs a nonlinear transformation through an activation function, and then passes the result to the neuron in the next layer.

[0077] The deep neural network here can be trained using the force curve data of the sample. This force curve is a curve showing the real-time change in the reactive force generated by the sample during the detection process.

[0078] Step S204: The overall optimization model for nano-infrared signals is constructed by combining the local electromagnetic field model, the multi-physics coupling model, and the deep neural network model.

[0079] Among them, such as Figure 5 As shown, the local electromagnetic field model as the first part, the multi-physics coupling model as the second part, and the deep neural network model as the third part together constitute the overall optimization model for nano-infrared signals.

[0080] The local electromagnetic field model is used to calculate and optimize the probe material structure and sample material. The multiphysics coupling model is used to optimize the pulse excitation method by modulating the waveform to create photoacoustic-mechanical interference consistent with non-chemical signals. The deep neural network is used to extract the effective infrared signal from the probe motion curve in real time.

[0081] In the above implementation process, the local electromagnetic field model is used to optimize parameters such as sample detection parameters, probe detection parameters, and pulse detection parameters; the multi-physics coupling model is used to optimize the pulse excitation method and improve sample detection sensitivity; and the deep neural network is used to extract the effective infrared signal from the probe motion curve in real time, improving the signal-to-noise ratio of the sample spectrum. By setting the local electromagnetic field model, multi-physics coupling model, and deep neural network model in the overall optimization model of nano-infrared signals, the performance of sample detection can be systematically improved, and the sensitivity of the overall optimization model of nano-infrared signals to sample detection can be enhanced.

[0082] In one possible implementation, step S201 includes: calculating the background field without a needle tip under polarized light excitation according to the Fresnel equation; calculating the scattering field with a needle tip present based on the first target geometric model, material parameters and background field, and determining the sum of the background field and the scattering field.

[0083] The needle tip here refers to the tip of the probe.

[0084] In one embodiment, the background field can be calculated through the following steps:

[0085] Considering the reflection and refraction of polarized plane waves at a plane interface, the transverse wave number can be expressed using the incident angle θ as follows:

[0086]

[0087] The magnitude k of the longitudinal wave vector in the incident medium and the projected medium z1 k z2 It can be represented as follows:

[0088]

[0089] Where k1 and k2 are the wave numbers in the incident medium and the projected medium, respectively.

[0090]

[0091] Where, k x and k y These are the wave vector components in the x and y directions at the interface, respectively, where ω is the electromagnetic wave angular frequency, c is the speed of light, and k is the wave vector component in the x and y directions. || ε1 is the transverse wave number, ε2 is the dielectric constant of the incident medium, μ1 is the permeability constant of the incident medium, and μ2 is the permeability constant of the transmitted medium.

[0092] Correspondingly, the Fresnel reflection coefficient ρ and the Fresnel transmission coefficient τ at the interface can be defined as follows:

[0093]

[0094]

[0095] The background field E1 of the incident medium and the background field E2 of the transmitting medium can then be obtained:

[0096] E1 = E in +ρE in ;

[0097] E2=τE in

[0098] Among them, E in This represents the electric field of the incident wave.

[0099] The sum of the background field and the scattered field can be determined through the following steps:

[0100] Step 1: Establish the first target geometric model;

[0101] Step 2: Set material parameters;

[0102] Step 3: Define the frequency domain physical field of electromagnetic waves;

[0103] Step 4: Grid generation.

[0104] The first target geometric model is a two-dimensional axisymmetric model of the complete needle tip-sample structure.

[0105] The material parameters here include probe parameters and sample parameters. For example, the probe material (gold, platinum, iron, silicon, etc.), the sample material (silicon, silica, polydimethylsiloxane, etc.), and the sample thickness. These material parameters can be selected according to the actual situation.

[0106] In one embodiment, defining the electromagnetic wave frequency domain physical field may include defining the pulse incident angle.

[0107] For example, the laser pulse incident parameters can be set as follows: wavelength of 9 μm and polar angle θ of 60°.

[0108] The sum of the background field and the scattered field constitutes the enhanced electric field. This enhanced electric field has a higher field strength than the background field, and is configured to enhance the signal.

[0109] The infrared enhancement factor is calculated by dividing the square of the enhanced electric field by the square of the background field.

[0110] It should be understood that when selecting sample detection parameters and probe detection parameters using a local electromagnetic field model, the material parameters can be changed in the second step described above, and the infrared enhancement factor can be determined based on the local electromagnetic field model after the material parameters have been changed. When the infrared enhancement factor determined by the local electromagnetic field model meets the requirements, the material parameters corresponding to the required infrared enhancement factor are the sample detection parameters and probe detection parameters needed for the current sample detection.

[0111] When determining the pulse incident angle using a local electromagnetic field model, the pulse incident angle can be changed in the third step described above, and the corresponding infrared enhancement factor can be determined based on the local electromagnetic field model. When the infrared enhancement factor determined by the local electromagnetic field model meets the requirements, the pulse incident angle corresponding to the required infrared enhancement factor is the pulse incident angle required for the current sample detection.

[0112] In the above implementation process, compared with three-dimensional models, two-dimensional models have the advantages of simple structure, easy calculation, and fast computational efficiency. When creating a local electromagnetic field model, setting the first target geometric model as a two-dimensional model can improve the computational efficiency of the local electromagnetic field model. In addition, when creating the local electromagnetic field, by setting and changing the material parameters and the electromagnetic wave frequency domain physical field, the infrared enhancement factor output by the local electromagnetic field can be used to determine the optimal sample detection parameters, probe detection parameters, and other parameters, thereby optimizing the sample detection parameters, probe parameters, and other parameters, and improving the sensitivity of the local electromagnetic field model.

[0113] In one possible implementation, step S202 includes: modeling the optical gradient force based on the electric dipole model to obtain an optical gradient force sub-model; modeling the photoacoustic force through COMSOL acoustic simulation to obtain a photoacoustic force sub-model; and modeling the thermal expansion force based on the Hertz contact model to obtain a photothermal force sub-model.

[0114] It should be understood that when light enters the tip-sample gap, a series of physical processes occur in the sample: First, the local electromagnetic field is enhanced, dominated by the lightning rod effect, which significantly increases the intensity of the infrared signal. Simultaneously, an optical gradient force occurs, i.e., the attractive force between the tip and the sample. After continuous illumination for a period of time, the sample absorbs a certain amount of infrared light, leading to thermal expansion (the scale of thermal expansion depends on the sample and light intensity). Since the illumination time is the period of contact between the tip and the sample, the photothermal expansion of the sample causes an upward repulsive force on the probe tip, i.e., a photothermal (contact) force. Simultaneously with the thermal expansion of the sample, according to the photoacoustic effect, rapid volume expansion and contraction radiate sound waves outward. When these sound waves reach the surface of the probe cantilever beam, a pressure difference is generated on the probe cantilever surface, causing cantilever movement, i.e., a photoacoustic force.

[0115] The optical gradient force, photoacoustic force, and photothermal force here are used to reflect the different states of the sample at different times when the sample is irradiated by laser.

[0116] Understandably, when the probe tip is irradiated by a laser, the charge oscillations in the tip material will induce polarization at the light-driven frequency, approximating an oscillating dipole. This dipole will be subjected to an electromagnetic force in the continuously irradiated light field. Simultaneously, the sample will also be induced to polarize under the focused light field. Its presence affects the local field near the induced dipole at the tip, thus influencing the electromagnetic force applied to the tip. Therefore, as... Figure 6 As shown, the needle tip and sample can be approximated as two electric dipoles. That is, an optical gradient force quantum model can be created based on the electric dipole model.

[0117] The COMSOL acoustic simulation described here is a method of acoustic simulation analysis based on COMSOL Multiphysics software. Because COMSOL Multiphysics software has multi-field coupling analysis capabilities, it can achieve accurate numerical simulations.

[0118] When constructing a photoacoustic sub-model using COMSOL acoustic simulation, it can be achieved by describing the overall spatial distribution of all physical fields.

[0119] The Hertz contact model described above is a physical model that describes the contact relationship between two objects. It is used to calculate the elastic deformation and contact pressure distribution generated when two curved surfaces come into contact with each other under the action of external forces.

[0120] In the above implementation process, based on the characteristics of various optical forces of the sample, the corresponding model creation method is selected to create the corresponding optical force sub-model, thereby constructing a more accurate optical force sub-model and improving the accuracy of various optical force calculations of the sample.

[0121] In one possible implementation, the optical gradient force is modeled based on the electric dipole model to obtain the optical gradient force sub-model, which includes: calculating the background field without a needle tip under polarized light excitation according to Fresnel equations; calculating the scattering field with a needle tip based on the second target geometric model, material parameters and background field, and determining the sum of the background field and the scattering field.

[0122] The background field here is calculated in the same way as the background field in the creation of the local electromagnetic field model mentioned above.

[0123] If the optical gradient force model is constructed after the local electromagnetic field model, then the background field does not need to be calculated separately according to the Fresnel equations when constructing the optical gradient force model. If the local electromagnetic field model and the optical gradient force model are constructed simultaneously (or in parallel), then the background field needs to be calculated according to the Fresnel equations when constructing the optical gradient force model.

[0124] It should be understood that since the electromagnetic field model based on the construction of two-dimensional axisymmetric structures cannot perform complete needle tip integration calculations, the modeling method using electric dipole approximation can reduce the solution accuracy to a certain extent, thereby reducing the amount of computation.

[0125] The sum of the background field and the scattered field can be determined through the following steps:

[0126] Step 1: Establish the geometric model of the second objective;

[0127] Step 2: Set material parameters;

[0128] Step 3: Define the frequency domain physical field of electromagnetic waves;

[0129] Step 4: Grid generation.

[0130] The second target geometric model is a three-dimensional model of the complete needle tip-sample structure.

[0131] In the above implementation process, the electromagnetic field model based on the two-dimensional axisymmetric structure cannot perform complete tip integration calculations. Modeling the optical gradient force using an electric dipole model, and thus obtaining an optical gradient force sub-model, can reduce the solution accuracy, decrease the computational load, and improve the computational efficiency of the optical gradient force sub-model. Furthermore, compared to the two-dimensional model, the three-dimensional model has higher computational accuracy; by replacing the geometric model in the optical gradient force sub-model with a three-dimensional model, the computational accuracy of the optical gradient force sub-model can be improved.

[0132] In one possible implementation, photoacoustic force modeling is performed using COMSOL acoustic simulation to obtain a photoacoustic force sub-model, which includes: constructing the photoacoustic force sub-model by distributing all physical fields in the overall space of COMSOL.

[0133] Among them, such as Figure 7 As shown, the overall spatial distribution of all physical fields from the outside to the inside is as follows: perfect matching layer, transient pressure acoustic physical field and transient thermoviscous acoustic physical field. The three-dimensional structure of the probe and sample is set in the transient thermoviscous acoustic physical field.

[0134] Figure 7 The right side shows the three-dimensional structure of the probe and sample, situated within a transient thermoviscous acoustic physics field. The polymethyl methacrylate and substrate are 600 nanometers thick (visibility can be enhanced by magnification up to five times), with the gold-plated probe positioned on top of them.

[0135] The calculation of the heat transfer equations here takes place in the transient pressure acoustic physics field and the transient thermoviscous acoustic physics field, coupled with the pressure acoustic and thermoviscous acoustic physics fields.

[0136] The aforementioned photoacoustic force sub-model is configured to determine the photoacoustic force by integrating the sound pressure across the entire surface using an integral probe.

[0137] In one embodiment, during the modeling of photoacoustic force using COMSOL acoustic simulation, the transient propagation process of sound waves in space can be simulated by solving the acoustic wave equation. After simulating the transient propagation process of sound waves in space, the photoacoustic force can be determined by integrating the sound pressure across the entire surface of the probe.

[0138] In the above implementation process, when creating the photoacoustic sub-model, all physical fields are distributed in COMSOL to simulate the actual physical field of the sample. By using the actual physical field of the sample to simulate the photoacoustic sub-model, the accuracy of the photoacoustic sub-model can be improved.

[0139] In one possible implementation, a photothermal sub-model is obtained by modeling thermal expansion force based on the Hertz contact model, including: determining the planar and spherical structures of the Hertz contact model according to the probe tip and the sample; and using the Hertz contact model composed of the planar and spherical structures as the photothermal sub-model.

[0140] The photothermal sub-model is configured to calculate thermal expansion force based on transient thermal expansion distribution.

[0141] In the Hertz contact model, the plane is the plane in contact between the sample and the probe tip, and the spherical structure is the part in contact between the probe tip and the sample.

[0142] Understandably, such as Figure 8 As shown, the probe tip typically has a very small radius of curvature, which allows it to be considered a spherical structure when in contact with the sample. The sample surface is relatively flat on a microscale and therefore can be considered a plane in the Hertz model.

[0143] In one embodiment, the plane can be considered as an infinitely large plane.

[0144] The Hertz contact model described above is used to describe the deformation and stress distribution of two elastic bodies near the contact point.

[0145] In the above implementation process, since photothermal force arises when the probe tip contacts the sample, the sample's photothermal expansion causes an upward repulsive force on the probe tip, and the Hertz contact model is used to describe the deformation and stress distribution of the two elastic bodies near the contact point, constructing the photothermal sub-model using the Hertz model can improve the accuracy of the photothermal sub-model.

[0146] In one possible implementation, step S203 includes: preprocessing the force curves of the collected samples to construct a dataset; using the noisy signal in the dataset as the input of the MLP network, using the resonant frequency band integral value of the clean signal in the dataset as the output of the MLP network, and training the MLP network to obtain a deep neural network model.

[0147] Among them, the force curve is the curve of the force on the generated reaction sample during the nano-infrared detection process.

[0148] The preprocessing of the force curve here can include background fitting, force curve subtraction, and other processing methods. The preprocessing method of the force curve can be selected according to the actual situation.

[0149] Optionally, the dataset may include experimental test sets, simulation training sets, simulation test sets, etc., and the dataset can be selected according to the actual situation.

[0150] In one embodiment, the dataset can be constructed through the following steps: First, the acquired force curves are preprocessed to obtain the original oscillation signal, which serves as the experimental test set. A Fast Fourier Transform is performed on the oscillation signal, and the expected variance of the noise in the non-resonant frequency band is calculated. The noise at the data points of the oscillation signal is then calculated to obtain the noise variance in the experiment. Based on the calculated noise variance, Gaussian white noise is simulated, and simultaneously, a sine wave with the same frequency as the experimental data is simulated. After Hanning window filtering, a clean simulation signal is generated. The Gaussian white noise is then added to the clean simulation signal to generate a noisy simulation signal. Simulation signals with the same amplitude are grouped together to generate a dataset. After repeatedly increasing the sine wave amplitude, the noisy simulation signal and the clean simulation signal are finally generated as the simulation training set for the model. Following the above process, the sine wave amplitude is multiplied by the spectral data to generate a set of noisy simulation signals as the simulation test set.

[0151] For example, the 500 collected force curves were preprocessed (with a sampling frequency of 25 MHz) to obtain 500 oscillation signals containing the infrared absorption information of the sample, with 1664 data points, forming a 500×1664 original oscillation signal as the experimental test set. A Fast Fourier Transform was performed on the oscillation signals, and the expected variance of the non-resonant frequency noise (infrared signals in the resonant frequency band) was calculated. Where N is the total amount of data, x i Let μ be the value of the i-th noise (and μ be the average value of the overall noise). Calculate the noise from 829,500 data points across 500 oscillating signals to obtain the noise variance σ in the experiment. 2 It is 0.0322. Based on the calculated noise variance σ... 2 Simulated Gaussian white noise and a sine wave with the same frequency as the experimental data were used. After Hanning window filtering, a clean simulation signal was generated. The two simulations were then combined to generate a noisy simulation signal. The dataset used grouped 1000 simulation signals with the same amplitude to generate a 1000×1664 dataset. After increasing the sine wave amplitude ten times, a 10000×1664 noisy simulation signal and a 10000×1664 clean simulation signal were generated as the model's training set. Following the same procedure, the sine wave amplitude was multiplied by the spectral data to generate a 500×1664 noisy simulation signal as the simulation test set.

[0152] The simulation training set is used to train the MLP network, and the simulation test set is used to test the accuracy of the trained simulation test set.

[0153] It should be understood that the force curve signal of the acquired sample is usually a time-domain signal, while model training requires a frequency-domain signal. Therefore, before model training, the time-domain signal needs to be converted into a frequency-domain signal.

[0154] For example, the noisy time-domain signal y(t) = x(t) + n(t) can be converted to the frequency domain using a fast Fourier transform, allowing the calculation of its spectrum Y(k) + Y(k). * )=X(k * )+N(k * )+X(k)+N(k), where k * Let represent the contact resonant frequency range, k represent the non-resonant frequency range, x(t) be the clean signal, n(t) be the noise signal, and Y(k) be the value of the noise signal in the non-resonant frequency range. * X(k) represents the noise signal value in the contact resonant frequency range, and X(k) represents the clean signal value in the non-resonant frequency range. * N(k) represents the value of the clean signal in the contact resonant frequency range, and N(k) represents the conjugate complex number of the noise signal in the non-resonant frequency range. * ) is the conjugate complex number of the noise signal in the contact resonant frequency range.

[0155] In one embodiment, the deep neural network model described above is trained using a mean squared error loss function. The method further includes updating the model parameters of the neural network model during training iterations using the Adam optimization algorithm.

[0156] The mean squared error loss function can be defined as follows: Among them, L MSE f(x) represents the mean squared error loss function, where n is the number of samples. i ) represents the i-th sample x i The predicted value, y i This represents the actual value of the i-th sample.

[0157] The Adam optimization algorithm here is an efficient gradient descent optimization algorithm that combines the advantages of Momentum-SGD and RMSProp. It can adaptively adjust the learning rate and take into account the momentum of the gradient to accelerate convergence.

[0158] Optionally, the model parameters that can be optimized may include parameters such as training period, batch size, initial learning rate, and learning rate decay factor. The model parameters to be optimized can be selected according to the actual situation.

[0159] For example, if the training set contains 10,000 signal samples, the deep neural network model can be trained for 300 epochs with a batch size of 1,000, resulting in 300,000 model parameter updates. During training, the model parameters are optimized using the Adam optimization algorithm. As shown in Table 1, the loss function curve of the MLP network converges rapidly, reaching its minimum mean square error after 300 training epochs, indicating that the MLP network performs well under these parameter settings. The other model parameters of the MLP network at this point are shown in Table 1.

[0160] Table 1:

[0161]

[0162] like Figure 9 As shown, the structure of the deep neural network model described above can be specifically as follows: it includes 5 hidden layers, with 1664 neurons in the input layer, 512 neurons in the first hidden layer, 256 neurons in the second hidden layer, 64 neurons in the third hidden layer, 16 neurons in the fourth hidden layer, and 1 neuron in the output layer.

[0163] The structure of the deep neural network model presented here is merely illustrative, and the structure of this deep network model can be adjusted according to actual circumstances.

[0164] In the above implementation process, since MLP networks have advantages such as simple structure and flexible design, and the network structure can be adjusted according to the needs of specific problems, training deep neural network models using MLP networks as the architecture can simplify the structure of deep neural network models, improve the computational efficiency of deep neural network models, and increase the flexibility of the deep neural network models.

[0165] Please see Figure 10 This is a flowchart of the nano-infrared detection method provided in the embodiments of this application. This nano-infrared detection method is applied to the overall optimization model of nano-infrared signals in the above embodiments. The following will discuss... Figure 9 The specific process shown will be explained in detail.

[0166] Step S301: During the sample detection process of the nano-infrared detection device, the sample detection parameters, probe detection parameters and pulse detection parameters are adjusted by the local electromagnetic field model, and the force curve of the sample is output in real time.

[0167] It should be understood that material parameters such as sample parameters and probe parameters are among the factors in creating a local electromagnetic field model. Different sample parameters and probe parameters may result in different local electromagnetic field models. Accordingly, under the same pulse detection parameters, the infrared enhancement factor output by this local electromagnetic field may differ. Similarly, a local electromagnetic field with the same material parameters may also have different infrared enhancement factors output under different pulse detection parameters.

[0168] In the process of sample detection using a nano-infrared detection device, by changing the sample parameters and probe parameters of the local electromagnetic field model, and outputting the corresponding infrared enhancement factor, the sample detection parameters and probe detection parameters corresponding to the required infrared enhancement factor can be determined. Furthermore, by changing the pulse detection parameters and then determining whether the infrared enhancement factor output by the local electromagnetic field is the required infrared enhancement factor, the corresponding pulse detection parameters can be determined.

[0169] In one embodiment, during the process of detecting a sample using a nano-infrared detection device, the local electromagnetic field model can be changed by altering the geometric structure, thereby determining the corresponding geometric detection structure.

[0170] Understandably, when the nano-infrared detection device is activated, the laser emitting unit emits a laser beam towards the sample. Under the influence of the laser, the sample generates a corresponding photodynamic force, and the data processing unit generates a force curve based on this force. The local electromagnetic field model is used to adjust the sample detection parameters, probe detection parameters, and pulse detection parameters. These parameters affect the photodynamic force generated by the sample, thus influencing the force curve. In other words, the force curve is adjusted by modifying the material parameters and / or pulse parameters during the nano-infrared detection process using the local electromagnetic field model.

[0171] Step S302: Calculate the optical force values ​​corresponding to various optical forces of the sample using a multiphysics coupling model.

[0172] The optical force here can include optical gradient force, photoacoustic force, and photothermal force, etc., and the optical force can be selected according to the actual situation.

[0173] The aforementioned multiphysics coupling model is formed by combining multiple photodynamic sub-models. Each photodynamic sub-model is used to calculate the corresponding photodynamic force.

[0174] In one embodiment, the multiphysics coupling model can also be optimized for pulse excitation mode.

[0175] Understandably, the sound sources of photoacoustic waves are the thermal expansion caused by the absorption of infrared light by the sample material, and the volume expansion caused by the absorption of infrared light by water molecules in the medium environment. The infrared absorption by the medium environment is the source of interference with the sample signal. Since the density ρ, the Green's Eisen parameter Γ, and the sound velocity c of the medium environment are determined by the physical properties of the material, based on the formula... The rate of change of absorbed power over time can be used to measure the material's absorption power. To change the size of the interfering sound source.

[0176] According to the formula Q(t)=∏(t)*P abs The pulse waveform ∏(t) can be adjusted to control... Infrared pulses with a trapezoidal power distribution typically undergo four phases within a single pulse cycle: the rising edge, the plateau phase, the falling edge, and the thermal phase. This can be expressed by the following formula:

[0177]

[0178] Where P0 is the set infrared pulse power, t is the time, and t edge For the rising edge phase time, t plateau This refers to the plateau phase.

[0179] The above reasoning process shows that while keeping the total pulse power ∫Pdt constant, increasing the rise time t... edge The parameter settings of the pulse waveform function ∏(t) can be changed.

[0180] Step S303: Input the force curve into the deep neural network model, and demodulate the high signal-to-noise ratio infrared signal through the deep neural network model.

[0181] The infrared signal is configured to detect the performance of the sample.

[0182] The deep neural network here is used to achieve high signal-to-noise ratio (SNR) demodulation of sample signals. The main evaluation metrics for the deep neural network are standard deviation and signal-to-noise ratio.

[0183] In one embodiment, the standard deviation can be expressed by the following formula:

[0184]

[0185] The signal-to-noise ratio can be expressed by the following formula:

[0186]

[0187] Where N is the total number of points, x i Let be the i-th infrared intensity value, and μ be the average value of the infrared intensity.

[0188] The standard deviation mentioned above can quantify the range of variation in measurement data to assess the stability of the signal, and thus directly reflect the noise level.

[0189] In the above implementation process, during nano-infrared detection, the probe detection parameters, sample detection parameters, and pulse detection parameters used for sample detection are first adjusted using a local electromagnetic field model. This generates an optimal infrared enhancement factor, optimizing the probe, sample, and pulse parameters. Then, various photomechanical forces are calculated using a multiphysics coupling model to select an appropriate pulse excitation method to suppress photomechanical interference from non-chemical signals, thereby improving sample detection sensitivity. Finally, a deep neural network model extracts the effective infrared signal from the force curve, achieving the reconstruction and restoration of high-dimensional feature information into a clean signal, thus improving the signal-to-noise ratio of the sample spectrum.

[0190] In one possible implementation, step S301 includes: inputting probe parameters and sample parameters into a local electromagnetic field model to obtain an infrared enhancement factor; if the infrared enhancement factor is not the strongest infrared enhancement factor, changing the probe parameters and sample parameters, and re-inputting the changed probe parameters and sample parameters into the local electromagnetic field model until the infrared enhancement factor output by the local electromagnetic field model is the strongest infrared enhancement factor; and / or if the infrared enhancement factor is not the strongest infrared enhancement factor, changing the infrared pulse angle of the local electromagnetic field model, and outputting the infrared enhancement factor through the local electromagnetic field model after changing the infrared pulse angle, until the infrared enhancement factor output by the local electromagnetic field model is the strongest infrared enhancement factor.

[0191] Among them, the probe parameter corresponding to the strongest infrared enhancement factor is the probe detection parameter, the sample parameter corresponding to the strongest infrared enhancement factor is the sample detection parameter, and the infrared pulse angle corresponding to the strongest infrared enhancement factor is the pulse detection parameter.

[0192] The sample detection parameters, probe detection parameters, and pulse detection parameters can be adjusted simultaneously or partially. The adjustment of these parameters can be selected according to the situation.

[0193] For example, the probe structure can be optimized by changing the radius of curvature of the probe tip (e.g., 1 nm, 5 nm, 15 nm, 25 nm, etc.) and determining which size of the tip radius of curvature can output the highest infrared enhancement factor by using the controlled variable method.

[0194] By changing the material settings of the probe tip (e.g., gold, platinum, iron, silicon, etc.) and using the controlled variable method to determine which material tip can output the highest infrared enhancement factor, the probe material can be optimized.

[0195] By changing the material settings of the sample (e.g., silicon, silica, polydimethylsiloxane, etc.) and using the controlled variable method to determine which substrate material can achieve the highest infrared enhancement factor, the substrate material can be optimized.

[0196] By changing the incident angle of the infrared pulse (e.g., the polar angle), the incident angle can be freely defined, thereby outputting the strongest infrared enhancement factor and completing parameter optimization.

[0197] By changing the height of the first target geometric model (where the height z ranges from 0 to 50 nm), the sample thickness can be freely defined to output the strongest infrared enhancement factor, thus completing parameter optimization.

[0198] In the above implementation process, based on the adjusted sample parameters, probe parameters, or pulse detection parameters, and the corresponding output infrared enhancement factor, the sample parameters, probe parameters, or pulse detection parameters corresponding to the strongest infrared enhancement factor are determined as the sample detection parameters, probe detection parameters, or pulse detection parameters, thereby optimizing the sample parameters, probe parameters, or pulse detection parameters and improving the accuracy of nano-infrared detection.

[0199] In one possible implementation, step S302 includes: inputting the electric and magnetic fields of the tip-sample structure at various points in space into the Maxwell stress tensor formula to determine the corresponding Maxwell stress tensor; calculating the optical gradient force using the Maxwell stress tensor and Gauss's law; calculating the total absorbed power of the material based on the electric field distribution generated in space by the tip-sample structure calculated by the local electromagnetic field model and the corresponding material absorbed power formula; determining the transient thermal expansion distribution by using the total absorbed power as the heat source term in the sample heat transfer equation; and calculating the photoacoustic force and photothermal force using the transient thermal expansion distribution.

[0200] The Maxwell stress tensor formula here can be expressed by the following formula:

[0201]

[0202] Among them, T ij Let δ be the Maxwell stress tensor. ij It is the Kronecker symbol, E i and E j Let B be the component of the electric field intensity in the i-th and j-th directions. i and B j Let μi be the component of magnetic flux density in the i-th and j-th directions, μ0 be the permeability of free space, and μ0 be the permittivity of free space.

[0203] The aforementioned optical gradient force can be calculated using the following formula:

[0204]

[0205] in, Let be the Maxwell stress tensor.

[0206] In one embodiment, the aforementioned local electromagnetic field model can be used to calculate the electric field distribution of the tip-sample structure in space.

[0207] In this multiphysics coupling model, the obtained electric field magnitude can be substituted into the material's absorbed power formula, and the absorbed power distribution of the material can be integrated over time to obtain the total absorbed power. By treating the total absorbed power as the heat source term in the sample heat transfer equation, the sample temperature distribution and thermal expansion distribution can be solved, thus obtaining the transient thermal expansion distribution.

[0208] The formula for the power absorbed by the material can be:

[0209]

[0210] The total absorbed power can be expressed by the following formula:

[0211] Q(t)=∏(t)*P abs ;

[0212] The heat transfer equation for the sample can be expressed as follows:

[0213]

[0214] Where ε0 is the dielectric constant of the material, c is the speed of light, n is the complex refractive index of the sample, λ is the incident wavelength, and V abs P is the volume of light absorbed by the material. abs Let ∏(t) be the power absorbed by the material, and k be the pulse waveform function. eff Let ρ be the effective thermal conductivity of the material, C be the density, V be the heat capacity, T be the volume, Re[n] be the real part of the complex refractive index, and Im[n] be the imaginary part of the complex refractive index.

[0215] The photoacoustic force and photothermal force here are calculated based on the transient thermal expansion distribution.

[0216] In one embodiment, photoacoustic force can simulate the transient propagation of sound waves in space by solving the acoustic wave equation. Accordingly, its expression can be:

[0217]

[0218] Where p is the sound pressure, c is the sound velocity in different fluids, S1 is the sound source generated by the thermal expansion of the sample, S2 is the sound source generated by the thermal expansion of water molecules, and c is the sound velocity.

[0219] In one embodiment,

[0220] Where ρ is the density of the medium, u is the displacement component at the interface (i.e., the transient thermal expansion distribution), c is the speed of sound, Q(t) is the total absorbed power, β is the coefficient of thermal expansion, and Γ is the Green's parameter.

[0221] It should be understood that the propagation of transient sound waves generates a pressure gradient in space, where the acoustic pressure gradient on the probe surface causes the cantilever to oscillate, i.e., the photoacoustic force F. pa After simulating the transient propagation of sound waves in space, the sound pressure across the entire surface of the integrating probe can be used for quantification. The corresponding expression is as follows:

[0222] F pa =∫ probe pds;

[0223] In one embodiment, photothermal power can be expressed by the following formula:

[0224]

[0225] Among them, E * R is the equivalent elastic modulus, R is the composite radius of curvature, and δ is the indentation depth (i.e., the transient thermal expansion distribution).

[0226] The equivalent elastic modulus and the combined radius of curvature mentioned above can be expressed by the following formulas:

[0227]

[0228]

[0229] Where v1, E1, and R1 are Poisson's ratio, elastic modulus, and radius of curvature of the needle tip material, respectively, and v2, E2, and R2 are Poisson's ratio, elastic modulus, and radius of curvature of the sample material, respectively.

[0230] In the above implementation process, when calculating various optical forces of the sample, the calculation is performed according to the corresponding calculation method for each optical force, which can improve the accuracy of various optical force calculations.

[0231] Based on the same application concept, this application also provides a nano-infrared signal optimization model construction device corresponding to the nano-infrared signal optimization model construction method. Since the principle of the device in this application is similar to that of the aforementioned nano-infrared signal optimization model construction method, the implementation of the device in this application can refer to the description in the above-mentioned method embodiments, and the repeated parts will not be described again.

[0232] Please see Figure 11This is a functional module diagram of the nano-infrared signal optimization overall model construction device provided in this application embodiment. Each module in the nano-infrared signal optimization overall model construction device in this embodiment is used to execute the various steps in the above method embodiments. The nano-infrared signal optimization overall model construction device includes a construction module 801, an establishment module 802, a training module 803, and a composition module 804. Wherein,

[0233] The construction module 801 is used to construct a local electromagnetic field model based on the sample's environment and the set illumination path; wherein, the local electromagnetic field model is calculated through a two-dimensional axisymmetric structure.

[0234] The module 802 is used to establish corresponding photodynamic sub-models for the various photodynamic forces generated by the sample, and to form a multiphysics coupling model through multiple photodynamic sub-models.

[0235] The training module 803 is used to train models based on MLP networks to obtain deep neural network models.

[0236] The constitutive module 804 is used to construct the overall optimization model of the nano-infrared signal by combining the local electromagnetic field model, the multi-physics coupling model and the deep neural network model.

[0237] In one possible implementation, the construction module 801 is specifically used for: calculating the background field without a needle tip under polarized light excitation according to the Fresnel equation; calculating the scattering field with a needle tip present based on a first target geometric model, material parameters, and the background field, and determining the sum of the background field and the scattering field; wherein the first target geometric model is a two-dimensional axisymmetric model of a complete needle-sample structure; wherein the sum of the background field and the scattering field is an enhancement electric field, and the enhancement electric field is configured to enhance the signal; the square of the enhancement electric field divided by the square of the background field yields the infrared enhancement factor.

[0238] In one possible implementation, module 802 is specifically used for: modeling the optical gradient force based on the electric dipole model to obtain the optical gradient force sub-model; modeling the photoacoustic force through COMSOL acoustic simulation to obtain the photoacoustic force sub-model; and modeling the thermal expansion force based on the Hertz contact model to obtain the photothermal force sub-model.

[0239] In one possible implementation, module 802 is specifically used to: calculate the background field without a needle tip under polarized light excitation according to the Fresnel equation; calculate the scattering field with a needle tip present based on the second target geometric model, material parameters, and the background field, and determine the sum of the background field and the scattering field; wherein the second target geometric model is a three-dimensional model of the complete needle tip-sample structure; wherein the sum of the background field and the scattering field is the enhanced electric field.

[0240] In one possible implementation, module 802 is specifically used to: construct the photoacoustic force sub-model by distributing all physical fields in the overall space of COMSOL; wherein the photoacoustic force sub-model is configured to determine the photoacoustic force by integrating the sound pressure of the entire surface of the probe; wherein the overall spatial distribution of all physical fields from the outside to the inside is as follows: a perfectly matched layer, a transient pressure acoustic physical field, and a transient thermo-viscous acoustic physical field, and the three-dimensional structure of the probe and the sample is set within the transient thermo-viscous acoustic physical field.

[0241] In one possible implementation, module 802 is specifically used to: determine the planar and spherical structures of the Hertz contact model based on the probe tip and the sample; use the Hertz contact model composed of the planar and spherical structures as the photothermal sub-model; wherein the photothermal sub-model is configured to calculate the thermal expansion force based on the transient thermal expansion distribution.

[0242] In one possible implementation, the training module 803 is specifically used to: preprocess the force curves of the collected samples to construct a dataset; use the noisy signals in the dataset as input to the MLP network, use the resonant frequency band integral value of the clean signals in the dataset as output to the MLP network, and train the MLP network to obtain the deep neural network model.

[0243] Please see Figure 12 This is a functional module diagram of the nano-infrared detection device provided in this application embodiment. Each module in the nano-infrared signal optimization overall model construction device in this embodiment is used to execute the various steps in the above method embodiments. The nano-infrared detection device includes an adjustment module 901, a calculation module 902, and a demodulation module 903. Wherein,

[0244] The adjustment module 901 is used to adjust the sample detection parameters, probe detection parameters, and pulse detection parameters through the local electromagnetic field model during the sample detection process of the nano-infrared detection device, and output the force curve of the sample in real time.

[0245] The calculation module 902 is used to calculate the optical force values ​​corresponding to various optical forces of the sample through the multiphysics coupling model.

[0246] The demodulation module 903 is used to input the force curve into a deep neural network model and demodulate the high signal-to-noise ratio infrared signal through the deep neural network model; wherein the infrared signal is configured to detect the performance of the sample.

[0247] In one possible implementation, the adjustment module 901 is specifically configured to: input the probe parameters of the probe and the sample parameters of the sample into the local electromagnetic field model to obtain an infrared enhancement factor; if the infrared enhancement factor is not the strongest infrared enhancement factor, replace the probe parameters and the sample parameters, and re-input the replaced probe parameters and the sample parameters into the local electromagnetic field model until the infrared enhancement factor output by the local electromagnetic field model is the strongest infrared enhancement factor; wherein, the probe parameters corresponding to the strongest infrared enhancement factor are the probe detection parameters, and the sample parameters corresponding to the strongest infrared enhancement factor are the sample detection parameters; and / or if the infrared enhancement factor is not the strongest infrared enhancement factor, change the infrared pulse angle of the local electromagnetic field model, and output the infrared enhancement factor through the local electromagnetic field model after changing the infrared pulse angle, until the infrared enhancement factor output by the local electromagnetic field model is the strongest infrared enhancement factor; wherein, the infrared pulse angle corresponding to the strongest infrared enhancement factor is the pulse detection parameter.

[0248] In one possible implementation, the calculation module 902 is specifically used for: inputting the electric and magnetic fields of the tip-sample structure at various points in space into the Maxwell stress tensor formula to determine the corresponding Maxwell stress tensor; calculating the optical gradient force using the Maxwell stress tensor and Gauss's law; calculating the total absorbed power of the material based on the electric field distribution generated in space by the tip-sample structure calculated by the local electromagnetic field model and the corresponding material absorption power formula; using the total absorbed power as the heat source term in the sample heat transfer equation to determine the transient thermal expansion distribution; and calculating the photoacoustic force and photothermal force using the transient thermal expansion distribution.

[0249] Furthermore, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when run by a processor, executes the steps of the nano-infrared signal optimization overall model construction method and / or nano-infrared detection method described in the above method embodiments.

[0250] The computer program product of the nano-infrared signal optimization model construction method and / or nano-infrared detection method provided in the embodiments of this application includes a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the steps of the nano-infrared signal optimization model construction method and / or nano-infrared detection method described in the above method embodiments. For details, please refer to the above method embodiments, which will not be repeated here.

[0251] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0252] In addition, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0253] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks. It should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0254] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application. It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0255] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for constructing an optimized overall model for nano-infrared signals, characterized in that, include: A local electromagnetic field model is constructed based on the sample's environment and the set illumination path; wherein, the local electromagnetic field model is calculated using a two-dimensional axisymmetric structure; For the various optical forces generated by the sample, corresponding optical force sub-models are established, and a multi-physics coupling model is formed by multiple optical force sub-models. A deep neural network model is obtained by training the model based on the MLP network architecture. The local electromagnetic field model, the multiphysics coupling model, and the deep neural network model together constitute the overall optimization model for the nano-infrared signal. The step of constructing a local electromagnetic field model based on the sample's environment and the set illumination path includes: Calculate the background field without a needle tip under polarized light excitation based on Fresnel's equations; Based on the first target geometric model, material parameters, and the background field, the scattering field when the needle tip is present is calculated, and the sum of the background field and the scattering field is determined; wherein, the first target geometric model is a two-dimensional axisymmetric model of the complete structure of the needle tip-sample; Wherein, the sum of the background field and the scattered field is the enhanced electric field, and the enhanced electric field is configured to enhance the signal; the square of the enhanced electric field divided by the square of the background field yields the infrared enhancement factor; The optical force includes optical gradient force, photoacoustic force, and photothermal force; the optical force sub-model includes optical gradient force sub-model, photoacoustic force sub-model, and photothermal force sub-model. The establishment of corresponding photodynamic sub-models for the various photodynamic forces generated by the sample includes: The optical gradient force is modeled based on the electric dipole model, resulting in the optical gradient force sub-model. The photoacoustic force was modeled using COMSOL acoustic simulation to obtain the photoacoustic force sub-model; The photothermal sub-model is obtained by modeling thermal expansion force based on the Hertz contact model.

2. The method according to claim 1, characterized in that, The process of modeling the optical gradient force based on the electric dipole model to obtain the optical gradient force sub-model includes: Calculate the background field without a needle tip under polarized light excitation based on Fresnel's equations; Based on the second target geometric model, material parameters, and the background field, the scattering field when the needle tip is present is calculated, and the sum of the background field and the scattering field is determined; wherein, the second target geometric model is a three-dimensional model of the complete structure of the needle tip-sample; The sum of the background field and the scattering field is the enhanced electric field.

3. The method according to claim 1, characterized in that, The process of modeling the photoacoustic force using COMSOL acoustic simulation to obtain the photoacoustic force sub-model includes: The photoacoustic force sub-model is constructed by distributing all physical fields in the overall space of COMSOL; wherein the photoacoustic force sub-model is configured to determine the photoacoustic force by integrating the sound pressure of the entire surface of the probe. The overall spatial distribution of all physical fields from the outside to the inside is as follows: perfect matching layer, transient pressure acoustic physical field and transient thermo-viscous acoustic physical field, and the three-dimensional structure of the probe and sample is set in the transient thermo-viscous acoustic physical field.

4. The method according to claim 1, characterized in that, The photothermal sub-model obtained by modeling thermal expansion force based on the Hertz contact model includes: Based on the probe tip and the sample, the planar and spherical structures of the Hertz contact model were determined; The Hertz contact model formed by the plane and the sphere structure is used as the photothermal sub-model; The photothermal sub-model is configured to calculate thermal expansion force based on transient thermal expansion distribution.

5. The method according to claim 1, characterized in that, The process of training a deep neural network model based on an MLP network architecture includes: The force curves of the collected samples are preprocessed to construct a dataset; The noisy signal in the dataset is used as the input of the MLP network, and the resonant frequency band integral value of the clean signal in the dataset is used as the output of the MLP network. The MLP network is then trained to obtain the deep neural network model.

6. A nano-infrared detection method, characterized in that, The method applies to the overall optimization model for nano-infrared signals according to any one of claims 1-5, wherein the overall optimization model for nano-infrared signals includes a local electromagnetic field model, a multi-physics coupling model, and a deep neural network model, and comprises: During the sample detection process using the nano-infrared detection device, the sample detection parameters, probe detection parameters, and pulse detection parameters are adjusted using the local electromagnetic field model, and the force curve of the sample is output in real time. The optical force values ​​corresponding to various optical forces of the sample are calculated using the multiphysics coupling model. The force curve is input into a deep neural network model, and the high signal-to-noise ratio infrared signal is demodulated by the deep neural network model. The infrared signal is configured to detect the performance of the sample.

7. The method according to claim 6, characterized in that, The adjustment of sample detection parameters, probe detection parameters, and pulse detection parameters through the local electromagnetic field model includes: The probe parameters and sample parameters are input into the local electromagnetic field model to obtain the infrared enhancement factor. If the infrared enhancement factor is not the strongest infrared enhancement factor, the probe parameters and the sample parameters are replaced, and the replaced probe parameters and the sample parameters are re-input into the local electromagnetic field model until the infrared enhancement factor output by the local electromagnetic field model is the strongest infrared enhancement factor; wherein, the probe parameters corresponding to the strongest infrared enhancement factor are the probe detection parameters, and the sample parameters corresponding to the strongest infrared enhancement factor are the sample detection parameters; and / or If the infrared enhancement factor is not the strongest infrared enhancement factor, the infrared pulse angle of the local electromagnetic field model is changed, and the infrared enhancement factor is output by the local electromagnetic field model after changing the infrared pulse angle, until the infrared enhancement factor output by the local electromagnetic field model is the strongest infrared enhancement factor; wherein, the infrared pulse angle corresponding to the strongest infrared enhancement factor is the pulse detection parameter.

8. The method according to claim 6, characterized in that, in, The optical force includes optical gradient force, photoacoustic force, and photothermal force; The calculation of the optical force values ​​corresponding to various optical forces of the sample through the multiphysics coupling model includes: The electric and magnetic fields at various points in space of the tip-sample structure are input into the Maxwell stress tensor formula to determine the corresponding Maxwell stress tensor. The optical gradient force is calculated using the Maxwell stress tensor and Gauss's law. Based on the electric field distribution generated in space by the tip-sample structure calculated by the local electromagnetic field model and the corresponding material absorption power formula, the total absorbed power of the material is calculated. The total absorbed power is used as the heat source term in the sample heat transfer equation to determine the transient thermal expansion distribution. Photoacoustic force and photothermal force are calculated based on the transient thermal expansion distribution.

9. A nano-infrared detection device, characterized in that, include: Laser emitting unit, probe, and data processing unit; The laser emitting unit is used to emit laser light toward the sample; The probe is used to measure the absorption of infrared light by the sample; The data processing unit is used to construct the overall optimization model of the nano-infrared signal according to the overall optimization model construction method of any one of claims 1-5 and / or to detect the performance of the sample according to the nano-infrared detection method of any one of claims 6-8.

10. A computer program product, comprising a computer program, characterized in that, The computer program is executed by the processor to perform the steps of the method as described in any one of claims 1 to 5 and / or 6 to 8.

Images