CN113868965A - Design method and system of black phosphorus wave absorber - Google Patents
Design method and system of black phosphorus wave absorber Download PDFInfo
- Publication number
- CN113868965A CN113868965A CN202111451892.3A CN202111451892A CN113868965A CN 113868965 A CN113868965 A CN 113868965A CN 202111451892 A CN202111451892 A CN 202111451892A CN 113868965 A CN113868965 A CN 113868965A
- Authority
- CN
- China
- Prior art keywords
- black phosphorus
- absorption spectrum
- training
- preset
- wave absorber
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- OAICVXFJPJFONN-UHFFFAOYSA-N Phosphorus Chemical compound [P] OAICVXFJPJFONN-UHFFFAOYSA-N 0.000 title claims abstract description 102
- 239000006096 absorbing agent Substances 0.000 title claims abstract description 78
- 238000000034 method Methods 0.000 title claims abstract description 50
- 238000013461 design Methods 0.000 title claims abstract description 23
- 238000012549 training Methods 0.000 claims abstract description 83
- 238000000862 absorption spectrum Methods 0.000 claims abstract description 78
- 238000013528 artificial neural network Methods 0.000 claims abstract description 41
- 238000004088 simulation Methods 0.000 claims abstract description 17
- 230000003287 optical effect Effects 0.000 claims abstract description 5
- 239000010410 layer Substances 0.000 claims description 53
- 238000012360 testing method Methods 0.000 claims description 27
- 229910052581 Si3N4 Inorganic materials 0.000 claims description 15
- HQVNEWCFYHHQES-UHFFFAOYSA-N silicon nitride Chemical compound N12[Si]34N5[Si]62N3[Si]51N64 HQVNEWCFYHHQES-UHFFFAOYSA-N 0.000 claims description 15
- 239000000463 material Substances 0.000 claims description 13
- 239000011159 matrix material Substances 0.000 claims description 13
- 210000002569 neuron Anatomy 0.000 claims description 12
- 238000004422 calculation algorithm Methods 0.000 claims description 10
- 230000003044 adaptive effect Effects 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 7
- 238000003860 storage Methods 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 5
- 230000004044 response Effects 0.000 claims description 4
- 239000002131 composite material Substances 0.000 claims description 3
- 239000002356 single layer Substances 0.000 claims description 3
- 230000010287 polarization Effects 0.000 description 8
- 239000011358 absorbing material Substances 0.000 description 5
- 238000010521 absorption reaction Methods 0.000 description 5
- 238000010276 construction Methods 0.000 description 4
- 238000009826 distribution Methods 0.000 description 4
- 238000011160 research Methods 0.000 description 4
- 238000005070 sampling Methods 0.000 description 4
- 238000001228 spectrum Methods 0.000 description 4
- 230000015556 catabolic process Effects 0.000 description 3
- 238000006731 degradation reaction Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000003595 spectral effect Effects 0.000 description 3
- 239000000758 substrate Substances 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 239000002245 particle Substances 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- XUIMIQQOPSSXEZ-UHFFFAOYSA-N Silicon Chemical compound [Si] XUIMIQQOPSSXEZ-UHFFFAOYSA-N 0.000 description 1
- 230000002238 attenuated effect Effects 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 239000013078 crystal Substances 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000000407 epitaxy Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 238000010030 laminating Methods 0.000 description 1
- 238000003475 lamination Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 239000002184 metal Substances 0.000 description 1
- -1 metamaterials Substances 0.000 description 1
- 239000002086 nanomaterial Substances 0.000 description 1
- 239000002105 nanoparticle Substances 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 125000004437 phosphorous atom Chemical group 0.000 description 1
- 229910052698 phosphorus Inorganic materials 0.000 description 1
- 239000011574 phosphorus Substances 0.000 description 1
- 239000004038 photonic crystal Substances 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 229910052710 silicon Inorganic materials 0.000 description 1
- 239000010703 silicon Substances 0.000 description 1
- 230000009897 systematic effect Effects 0.000 description 1
- 239000010409 thin film Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The embodiment of the invention provides a method and a system for designing a black phosphorus wave absorber, belonging to the technical field of inverse design of optical devices. The method comprises the following steps: obtaining a target absorption spectrum training sample in a pre-constructed special residual error neural network according to a preset simulation scene and preset structural parameters of a black phosphorus wave absorber; obtaining a prediction model by using the training sample; acquiring a demand absorption spectrum, training the demand absorption spectrum by using the prediction model, and outputting a target structure parameter; and designing a black phosphorus wave absorber based on the target structure parameters. According to the scheme, the prediction model is built in the pre-built special residual error neural network, and then the black phosphorus wave absorber structure parameter prediction meeting the user requirements is carried out based on the prediction model, so that the accuracy of the black phosphorus wave absorber structure parameter prediction is greatly improved.
Description
Technical Field
The invention relates to the technical field of inverse design of optical devices, in particular to a design method and a design system of a black phosphorus wave absorber.
Background
Based on the material characteristics of black phosphorus, the black phosphorus is an ideal material of the wave absorber substrate, and the wave absorber formed by alternately laminating the black phosphorus and silicon nitride is very wide in actual measurement and construction application. However, because black phosphorus has anisotropy, the absorption spectrum of TE polarization and the absorption spectrum of TM polarization have a significant difference, that is, the absorption spectra of different directions have a significant difference. Therefore, it is very difficult to predict the structural parameters as required. The existing methods for predicting the structural parameters of the black phosphorus wave absorber have the problems of low prediction precision and low efficiency. Accordingly, there is a need for a method of designing a black phosphorus absorber that can solve the above problems.
Disclosure of Invention
The embodiment of the invention aims to provide a method and a system for designing a black phosphorus wave absorber, which are used for at least solving the problems of low prediction precision and low efficiency of the conventional method for predicting the structural parameters of the black phosphorus wave absorber.
In order to achieve the above object, a first aspect of the present invention provides a method for designing a black phosphorus absorber, the method comprising: obtaining a target absorption spectrum training sample in a pre-constructed special residual error neural network according to a preset simulation scene and preset structural parameters of a black phosphorus wave absorber; obtaining a prediction model by using the training sample; acquiring a demand absorption spectrum, training the demand absorption spectrum by using the prediction model, and outputting target structure parameters of the black phosphorus wave absorber; and designing a black phosphorus wave absorber based on the target structure parameters.
Optionally, the simulation scenario includes: simulating incident angle, wavelength and oblique light admittance of incident light.
Optionally, the black phosphorus absorber includes: and the infinite width black phosphorus layer and the silicon nitride layer are alternately paved.
Optionally, the preset structural parameters of the black phosphorus wave absorber include: the thickness of the black phosphorus layer; wherein, the thickness of each black phosphorus layer is the same and is a preset fixed value; the target structure parameters of the black phosphorus wave absorber comprise: a thickness of the silicon nitride layer; wherein the thickness of each silicon nitride layer is the same or different.
Optionally, the method further includes: pre-constructing a dedicated residual neural network, comprising: adding a BN layer in front of each hidden layer of the traditional residual error neural network; each BN layer is normalized to carry out standardization of a corresponding hidden layer; wherein, the normalized output of the BN layer is as follows:
wherein,andfor adaptive parameters, i is the ith neuron of the BN layer;the mean value of the current input batch sample of the neuron i is obtained;the variance of the current input batch sample of the neuron i;is the response of neuron i.
Optionally, the obtaining a target absorption spectrum training sample in a pre-constructed dedicated residual neural network includes: acquiring absorption spectrum samples under each preset structure parameter by preset feature matrix algorithm based on the same simulation scene of the preset structure parameters of the black phosphorus wave absorbers; wherein the absorption spectrum sample is in a far infrared band; and integrating all absorption spectrum samples with preset structural parameters to form target absorption spectrum training sample data.
Optionally, the absorption spectrum samples under each preset structural parameter are obtained through a preset feature matrix algorithm: the calculation formula of the absorption spectrum sample is as follows:
wherein, X and Y are optical admittance from the first neural network layer to the air, and the expression is Z = X/Y; the preset feature matrix algorithm is as follows:
wherein M isjThe characteristic matrix of the j layer;the phase factor of incident light is expressed as:
wherein N isjIs the composite refractive index; djIs a single layer of material thickness;is the angle of the incident light;is the wavelength of the incident light; n isjFor guiding incident light.
Optionally, the target absorption spectrum training sample includes: training and testing sets; the training set is used for predictive model training; the test set is used for predictive model testing.
Optionally, the obtaining a prediction model by using the training samples includes: performing model training by using the training set as training data to obtain an initial prediction model; and performing initial prediction model test by using the test set as test data, judging whether the performance of the initial prediction model meets the expectation, and outputting the prediction model under the condition that the performance meets the expectation.
Optionally, the obtaining a required absorption spectrum, training the required absorption spectrum by using the prediction model, and outputting target structure parameters of the black phosphorus absorber includes: acquiring a required absorption spectrum input by a user; and taking the required absorption spectrum as input data, carrying out the prediction model training, and integrating the output thickness parameter of each silicon nitride layer as a target structure parameter of the black phosphorus wave absorber.
The second aspect of the present invention provides a black phosphorus wave absorber design system, including: the device comprises a presetting unit, a control unit and a control unit, wherein the presetting unit is used for simulating scene presetting and structural parameter presetting of a black phosphorus wave absorber; a processing unit to: obtaining a target absorption spectrum training sample in a pre-constructed special residual error neural network according to a preset simulation scene and preset structural parameters of a black phosphorus wave absorber; and obtaining a prediction model using the training samples; the acquisition unit is used for acquiring a required absorption spectrum; the training unit is used for training the required absorption spectrum by using the prediction model and outputting target structure parameters of the black phosphorus wave absorber; and the execution unit is used for designing the black phosphorus wave absorber based on the target structure parameters.
In another aspect, the present invention provides a computer-readable storage medium, having instructions stored thereon, which, when executed on a computer, cause the computer to perform the above-mentioned black phosphorus absorber designing method.
The scheme of the invention innovates the neural network, obtains a large amount of corresponding relations between the absorption spectra and the structural parameters under the preset simulated scene, and the corresponding relations imply the characteristic rules between the absorption spectra and the structural parameters. And training based on the characteristic rule to obtain a prediction model, wherein the prediction model can output corresponding prediction structure parameters according to the required absorption spectrum performance of a user, and the prediction accuracy of the black phosphorus absorber structure parameters is greatly improved.
Additional features and advantages of embodiments of the invention will be set forth in the detailed description which follows.
Drawings
The accompanying drawings, which are included to provide a further understanding of the embodiments of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the embodiments of the invention without limiting the embodiments of the invention. In the drawings:
FIG. 1 is a flow chart of the steps of a method for designing a black phosphorus absorber according to an embodiment of the present invention;
FIG. 2 is a flowchart of obtaining a training sample in a method for designing a black phosphorus absorber according to an embodiment of the present invention;
fig. 3 is a system configuration diagram of a black phosphorus absorber design system according to an embodiment of the present invention.
Description of the reference numerals
10-a preset unit; 20-a processing unit; 30-a collection unit; 40-a training unit;
50-execution unit.
Detailed Description
The following detailed description of embodiments of the invention refers to the accompanying drawings. It should be understood that the detailed description and specific examples, while indicating the present invention, are given by way of illustration and explanation only, not limitation.
Fig. 3 is a system configuration diagram of a black phosphorus absorber design system according to an embodiment of the present invention. As shown in fig. 3, an embodiment of the present invention provides a black phosphorus absorber design system, including: the presetting unit 10 is used for simulating scene presetting and structural parameter presetting of a black phosphorus wave absorber; a processing unit 20 for: obtaining a target absorption spectrum training sample in a pre-constructed special residual error neural network according to a preset simulation scene and preset structural parameters of a black phosphorus wave absorber; and obtaining a prediction model using the training samples; the acquisition unit 30 is used for acquiring a required absorption spectrum; the training unit 40 is used for training the required absorption spectrum by using the prediction model and outputting target structure parameters of the black phosphorus wave absorber; and the execution unit 50 is used for designing the black phosphorus wave absorber based on the target structure parameters.
Fig. 1 is a flowchart of a method for designing a black phosphorus absorber according to an embodiment of the present invention. As shown in fig. 1, an embodiment of the present invention provides a method for designing a black phosphorus absorber, where the method includes: obtaining a target absorption spectrum training sample in a pre-constructed special residual error neural network according to a preset simulation scene and preset structural parameters of a black phosphorus wave absorber; obtaining a prediction model by using the training sample; acquiring a demand absorption spectrum, training the demand absorption spectrum by using the prediction model, and outputting target structure parameters of the black phosphorus wave absorber; and designing a black phosphorus wave absorber based on the target structure parameters.
Step S10: presetting a simulation scene and presetting structural parameters of a black phosphorus wave absorber.
Specifically, the wave absorber is a material that absorbs radio waves without reflecting the radio waves, and is used for preventing ghost images of television radio waves, anechoic chambers for radio wave experiments, and the like. The wave absorber is widely used in the test field, and has different requirements on the electric wave absorption performance of the wave absorber corresponding to different test requirements. Black phosphorus is a black, metallic crystal formed by the conversion of white phosphorus at very high pressures and temperatures. Black phosphorus is the least reactive among the phosphorus allotropes and is not pyrophoric in air. Black phosphorus is a semiconductor characterized by a density of 2.70g/cm and a hardness of 2 by thin film epitaxy. Its lattice is composed of diatomic layers, each layer being composed of a tortuous chain of phosphorus atoms. In the chains, the P-P bond angle is 90 degrees, the phosphorus-phosphorus bond distance is 2.17 angstroms, black phosphorus is stable in air, the applicability to any material also wraps the physical and mechanical properties of the material, and for the effectiveness, a common method is to manufacture the wave-absorbing material by adding effective particle components into different base materials, so the research of the manufacturing technology of the wave-absorbing material is actually to research the relationship between the types and the density of filling particles and the frequency band of using electromagnetic waves, and is assisted by the research of the structure and the base material and the comprehensive experimental technology of the absorption energy efficiency consideration in narrow-frequency and wide-frequency applications. Based on the structural characteristics of the black phosphorus material, the black phosphorus material is an ideal material for the wave absorber substrate. Based on the alternate lamination of black phosphorus and silicon nitride, when electromagnetic waves enter the surface of the wave-absorbing material, refraction and reflection occur. The incident part of the electromagnetic waves are reflected by the bottom metal plate, and then emergent waves are formed from the surface of the wave-absorbing material, and the propagation direction is not changed. Because the thickness of the wave-absorbing material is a quarter wavelength, the phase difference between the outgoing wave and the incoming wave is exactly 180 degrees, and the interference principle of the waves tells us that the outgoing wave and the incoming wave are completely cancelled at the moment, so that the total reflected wave is greatly attenuated. However, because of the anisotropy of black phosphorus, there is a significant difference between the absorption spectrum of TE polarization and the absorption spectrum of TM polarization, i.e., the absorption spectra planned in different directions are stored in significant difference. Therefore, the difficulty of predicting the structural parameters according to the requirements is very high, and even if the existing neural network algorithm is utilized, the actual prediction precision is very low, so that the error of designing the black phosphorus metamaterial photon structure by the traditional neural network is increased.
Based on the problems, the scheme of the invention innovates the neural network, then designs the structural parameter prediction method of the black phosphorus wave absorber, and carries out the structural design of the black phosphorus wave absorber on the premise of ensuring the prediction precision. The method mainly includes the steps that a large number of corresponding relations between absorption spectrums and structural parameters are obtained under a preset simulated scene, and the corresponding relations imply characteristic rules between the absorption spectrums and the structural parameters. And training based on the characteristic rule to obtain a prediction model, wherein the prediction model can output corresponding prediction structure parameters according to the required absorption spectrum performance of the user. The scheme of the invention is fundamentally realized by the improved residual error neural network, which effectively improves the prediction precision of the structural parameters.
Based on the above, the scheme of the present invention preferably needs to obtain the characteristic rule thereof, i.e. the corresponding relationship between a large number of absorption spectra and structural parameters. First, a black phosphorus absorber structure is provided, which is composed of infinite-width alternating layers of black phosphorus and silicon nitride, and the black phosphorus layer is used as a substrate and has a fixed thickness, i.e., the thickness of each black phosphorus layer is the same, for example, the thickness of all black phosphorus layers is set to 0.35 nm. The core purpose of the scheme is then to predict the thickness of each silicon nitride layer, denoted as S = [ t1, t 2.. tn ], where t1-tn represents the thickness of the structure from the bottom to the top silicon nitride layer. In order to follow the principle of the control variable method and avoid interference of other factors on the prediction result, all corresponding relations are performed in the same simulation environment, namely a standard environment needs to be simulated, and all corresponding relations need to be performed in the standard environment. Preferably, the standard environment is preset to simulate the incident light as 85 ° incident angle, S/P polarization. The larger the number of training samples is, the more closely the model obtained by training is to the actual situation, so in order to improve the model accuracy, it is preferable to preset a large number of structural parameters and obtain a corresponding number of absorption spectra. For example, if 50000 structural parameters are preset, 50000 absorption spectra are correspondingly obtained.
Step S20: training samples are obtained in a pre-constructed dedicated residual neural network with respect to the target absorption spectrum.
In particular, it is known from the above that the implementation of the solution of the present invention requires a residual neural network innovation at all to ensure the improvement of the final prediction accuracy. Therefore, when training sample acquisition is performed, the dedicated residual neural network construction is performed first, and then subsequent method steps are performed on the constructed residual neural network. Specifically, as shown in fig. 2, the method includes the following steps:
step S201: and constructing a special residual error neural network.
Specifically, theoretically, the data collected after the entity test is more accurate, but because a large number of data samples are needed, and the difficulty of the structural design of the black phosphorus wave absorber is large, if a large number of experimental samples are directly manufactured, the structural parameters and the experimental data can not be guaranteed to be completely accurate, and a large amount of resources can be wasted. Aiming at the characteristics of Neural Networks, a simulation test is carried out, a large number of training samples can be quickly obtained without entity construction, and great benefits are provided for model construction. The neural network is an arithmetic mathematical model which imitates the behavior characteristics of the animal neural network and performs distributed parallel information processing. The neural network adopts nonlinear conversion to abstract the model for data processing and calculation, breaks through the fields of computer vision, speech recognition, text recognition, data mining and the like continuously, is successfully applied to physical researches except computer science, such as frequency spectrum prediction, antenna design, multilayer nano particles and the like, and forms a revolutionary and effective method. The basic idea is that a predefined training model is used for predicting the current problem, a better prediction effect is achieved by continuously iteratively updating the weight parameters, and the back propagation algorithm is suitable for a data set with a complicated data structure and numerous parameters. Neural networks have been widely used in the design and prediction of electromagnetism to solve such problems with higher precision and efficiency, such as filters, sensors, and super-surface structures. For more complex photonic structure designs such as photonic crystals, metamaterials, silicon photonic devices, etc., still in the beginning stage, systematic deep learning solutions have not been formed due to the complexity of performance and the flexibility of design.
Neural networks are usually trained using a first order gradient method, with two branches: random Gradient Descent (SGD) methods such as Momentum random Gradient method (Momentum-SGD), neisserov accelerated Gradient method (NAG), and adaptive learning rate methods such as Adam, adarad, AdaBound, AdamW, RMSProp, and the like. However, compared with text and image recognition, the sampling space for nano-structure parameter prediction is relatively small, a large number of samples are difficult to obtain, the prediction accuracy is greatly influenced, and the common neural network has the defects that the common neural network is trapped in a local optimal solution and a test set depends on a training model.
The main contribution of the residual neural network is to discover a Degradation phenomenon (Degradation) and invent a Shortcut connection (Shortcut connection) aiming at the Degradation phenomenon, so that the problem of difficult neural network training with excessive depth is greatly eliminated, and the training efficiency is also improved. But the generalization ability is low, and the absorption spectrum of TE polarization and the absorption spectrum of TM polarization have obvious difference due to the anisotropy of black phosphorus, so that the method has great limitation in a small sampling space sample. Therefore, the residual neural network cannot be directly used for training sample simulation in the present application. Based on this, preferably, in order to effectively improve the generalization capability of the network and overcome the limitation of small sampling space samples, an Adaptive BN layer is added before each hidden layer of the residual neural network. The essence of the neural network learning process is to learn data distribution, and once the distribution of training data is different from that of test data, the generalization capability of the network is greatly reduced; on the other hand, once the distribution of each batch of training data is different (the batch gradient is decreased), the network needs to learn to adapt to the different distribution in each iteration, so that the training speed of the network is greatly reduced. Therefore, the BN layer is normalized according to the current domain for each hidden layer to ensure that each layer receives similarly distributed data, for example, when m samples are given to the ith neuron of the BN layer, the mean μ i and the variance σ i can be expressed as follows:
ni is the amount of storage of the current batch sample of neuron i in the past iteration, and μ and σ 2 represent the mean and variance, respectively, of the current input batch sample of neuron i. The output of the Adaptive BN layer is represented by:
where γ i and β i are adaptive parameters, xi (k) is the response of neuron i. The algorithm can be regarded as the adaptive combination of identification mapping and batch normalization, no additional parameters are needed, the calculation amount is small, and in the calculation process, gamma i and beta i are dynamically changed, so that the importance of BN in reducing errors is adjusted.
Step S202: training samples are obtained for the target absorption spectrum.
Specifically, in the constructed special residual error neural network, the absorption spectrum of each preset wave absorber structural parameter in a preset scene is simulated by taking a preset simulation scene and a preset black phosphorus wave absorber structural parameter as references. For example, using an 85 ° incident angle, S/P polarization as an example, 50000 absorption spectrum samples based on specific structural parameters are calculated as a training set by using a feature matrix method, and an additional 1000 samples are calculated as a test set, each sample being in the far infrared band, and the absorption spectra of 256 sampling points are calculated for the wavelength range S1-S256. The target absorption spectrum based on the intermediate infrared band black phosphorus metamaterial is obtained by adopting a characteristic matrix method, and the characteristic matrix of the j layer is shown as the following formula:
wherein N isjIs the composite refractive index; djIs a single layer of material thickness;is the angle of the incident light;is the wavelength of the incident light; n isjThe target spectrum is calculated for the incident light admittance from the first BP layer to air, Z = X/Y, as shown in the following equation:
therefore, the calculation formula for the absorption spectrum is:
based on the calculation formula, the incident light phase factor is calculated by taking the incident light information under the preset simulation condition and each preset structural parameter information as known data, and then the absorption spectrum corresponding to the corresponding structural parameter is obtained step by step. Each preset structure parameter and the corresponding absorption spectrum form a corresponding relation, and if 50000 structure parameters are preset, 50000 absorption spectrum samples are correspondingly obtained. Preferably, in order to improve the training accuracy of the model, the absorption spectrum comprises a training set and a test set; the training set is used for predictive model training; the test set is used for the predictive model test. The model training is carried out through a large number of training sets, and then a part of new training samples are used as a testing set to carry out the model testing, so that the model accuracy is ensured.
Step S30: and obtaining a prediction model by using the training sample.
Specifically, in the constructed special residual error neural network, the training set is used as input data to perform model training, and after the training of all the training set data is completed, an initial prediction model is obtained. And (3) carrying out the initial prediction model test by using a preset test set, judging whether the working result of the initial prediction model accords with the expectation, if not, judging that the initial prediction model cannot meet the use requirement, adding training sample data to carry out model retraining, knowing that the obtained prediction model accords with the expectation, and outputting an accurate prediction model. The prediction model is represented as:
wherein HiInput data for the model, which is the user's demand spectral response; o isjIs the output of the model, which represents the optimal structural parameters to meet the current demand spectrum. Based on the model, the structure parameter can be predicted according to the absorption spectrum requirement of a user.
Step S40: and acquiring a user demand absorption spectrum, and outputting target structure parameters of the black phosphorus wave absorber according to the demand absorption spectrum and the prediction model.
Specifically, after model training is completed, a user can predict the structural parameters of the black phosphorus wave absorber with design according to the model. Firstly, a user judges the wave absorbing characteristics required by the dark room according to the self test or use requirements and converts the wave absorbing characteristics into corresponding spectrum absorption performance. And then the spectral absorption performance is taken as the self required absorption spectrum and is input into a prediction model obtained by training. The training model outputs corresponding target structure parameters according to the self execution process, the target structure parameters are the optimal matching structure parameters of the preset required absorption spectrum, and the black phosphorus wave absorber designed based on the target structure parameters can meet the spectral absorption requirement of a user.
Step S50: and designing a black phosphorus wave absorber based on the target structure parameters.
Specifically, after target structure parameters are obtained, the target structure parameters are analyzed, that is, thickness information of each silicon nitride layer is obtained, an upgrading scheme that the black phosphorus layers with fixed thicknesses and the silicon nitride layers with various output thicknesses are alternately stacked is generated by combining preset thickness information of the black phosphorus layers, and black phosphorus absorber design is performed based on the design scheme. Preferably, according to different thicknesses of the black phosphorus layer, the needed prediction models are different, and a user presets different simulation scenes and the thicknesses of the black phosphorus layer according to the needs of the user, trains a special prediction model and ensures the prediction accuracy.
Embodiments of the present invention also provide a computer-readable storage medium, which stores instructions that, when executed on a computer, cause the computer to execute the above-mentioned black phosphorus absorber design method.
Those skilled in the art will appreciate that all or part of the steps in the method for implementing the above embodiments may be implemented by a program, which is stored in a storage medium and includes several instructions to enable a single chip, a chip, or a processor (processor) to execute all or part of the steps in the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
While the embodiments of the present invention have been described in detail with reference to the accompanying drawings, the embodiments of the present invention are not limited to the details of the above embodiments, and various simple modifications can be made to the technical solution of the embodiments of the present invention within the technical idea of the embodiments of the present invention, and the simple modifications are within the scope of the embodiments of the present invention. It should be noted that the various features described in the above embodiments may be combined in any suitable manner without departing from the scope of the invention. In order to avoid unnecessary repetition, the embodiments of the present invention will not be described separately for the various possible combinations.
In addition, any combination of the various embodiments of the present invention is also possible, and the same should be considered as disclosed in the embodiments of the present invention as long as it does not depart from the spirit of the embodiments of the present invention.
Claims (12)
1. A method for designing a black phosphorus absorber, the method comprising: obtaining a target absorption spectrum training sample in a pre-constructed special residual error neural network according to a preset simulation scene and preset structural parameters of a black phosphorus wave absorber; obtaining a prediction model by using the training sample; acquiring a demand absorption spectrum, training the demand absorption spectrum by using the prediction model, and outputting target structure parameters of the black phosphorus wave absorber; and designing a black phosphorus wave absorber based on the target structure parameters.
2. The method of claim 1, wherein simulating the scene comprises: simulating incident angle, wavelength and oblique light admittance of incident light.
3. The method of claim 1, wherein the black phosphorus absorber comprises: and the infinite width black phosphorus layer and the silicon nitride layer are alternately paved.
4. The method according to claim 1, wherein the preset structural parameters of the black phosphorus absorber comprise: the thickness of the black phosphorus layer; wherein, the thickness of each black phosphorus layer is the same and is a preset fixed value; the target structure parameters of the black phosphorus wave absorber comprise: a thickness of the silicon nitride layer; wherein the thickness of each silicon nitride layer is the same or different.
5. The method of claim 1, further comprising:
pre-constructing a dedicated residual neural network, comprising: adding a BN layer in front of each hidden layer of the traditional residual error neural network; each BN layer is normalized to carry out standardization of a corresponding hidden layer; wherein, the normalized output of the BN layer is as follows:
6. The method of claim 1, wherein obtaining target absorption spectrum training samples in a pre-constructed dedicated residual neural network comprises: acquiring absorption spectrum samples under each preset structure parameter by preset feature matrix algorithm based on the same simulation scene of the preset structure parameters of the black phosphorus wave absorbers; wherein the absorption spectrum sample is in a far infrared band; and integrating all absorption spectrum samples with preset structural parameters to form target absorption spectrum training sample data.
7. The method according to claim 6, wherein the absorption spectrum samples under each preset structure parameter are obtained through a preset feature matrix algorithm, wherein the preset feature matrix algorithm comprises the following steps: the calculation formula of the absorption spectrum sample is as follows:
wherein, X and Y are optical admittance from the first neural network layer to the air, and the expression is Z = X/Y; the preset feature matrix algorithm is as follows:
8. The method of claim 7, wherein the target absorption spectrum training sample comprises: training and testing sets; the training set is used for predictive model training; the test set is used for predictive model testing.
9. The method of claim 8, wherein the obtaining a predictive model using the training samples comprises: performing model training by using the training set as training data to obtain an initial prediction model; and performing initial prediction model test by using the test set as test data, judging whether the performance of the initial prediction model meets the expectation, and outputting the prediction model under the condition that the performance meets the expectation.
10. The method of claim 4, wherein the obtaining a desired absorption spectrum, training the desired absorption spectrum with the predictive model, and outputting target structural parameters of a black phosphorus absorber comprises: acquiring a required absorption spectrum input by a user; and taking the required absorption spectrum as input data, carrying out the prediction model training, and integrating the output thickness parameter of each silicon nitride layer as a target structure parameter of the black phosphorus wave absorber.
11. A black phosphorus absorber design system, the system comprising: the device comprises a presetting unit, a control unit and a control unit, wherein the presetting unit is used for simulating scene presetting and structural parameter presetting of a black phosphorus wave absorber; a processing unit to: obtaining a target absorption spectrum training sample in a pre-constructed special residual error neural network according to a preset simulation scene and preset structural parameters of a black phosphorus wave absorber; and obtaining a prediction model using the training samples; the acquisition unit is used for acquiring a required absorption spectrum; the training unit is used for training the required absorption spectrum by using the prediction model and outputting target structure parameters of the black phosphorus wave absorber; and the execution unit is used for designing the black phosphorus wave absorber based on the target structure parameters.
12. A computer readable storage medium having instructions stored thereon which, when executed on a computer, cause the computer to perform the black phosphorus absorber design method of any one of claims 1-10.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111451892.3A CN113868965B (en) | 2021-12-01 | 2021-12-01 | Design method and system of black phosphorus wave absorber |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111451892.3A CN113868965B (en) | 2021-12-01 | 2021-12-01 | Design method and system of black phosphorus wave absorber |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113868965A true CN113868965A (en) | 2021-12-31 |
CN113868965B CN113868965B (en) | 2022-02-22 |
Family
ID=78985540
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111451892.3A Active CN113868965B (en) | 2021-12-01 | 2021-12-01 | Design method and system of black phosphorus wave absorber |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113868965B (en) |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108196326A (en) * | 2018-03-28 | 2018-06-22 | 常州大学 | A kind of broadband wave absorbing device based on black phosphorus and super surface |
US20190278880A1 (en) * | 2018-03-12 | 2019-09-12 | Exxonmobil Research And Engineering Company | Hybrid computational materials fabrication |
-
2021
- 2021-12-01 CN CN202111451892.3A patent/CN113868965B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20190278880A1 (en) * | 2018-03-12 | 2019-09-12 | Exxonmobil Research And Engineering Company | Hybrid computational materials fabrication |
CN108196326A (en) * | 2018-03-28 | 2018-06-22 | 常州大学 | A kind of broadband wave absorbing device based on black phosphorus and super surface |
Non-Patent Citations (2)
Title |
---|
曲立志: "《基于各向异性黑磷超材料的可调谐吸波体的研究》", 《中国优秀博硕士学位论文全文数据库(硕士)》 * |
王彤灵: "《基于超材料的电磁诱导透明效应与吸波体器件研究》", 《中国优秀博硕士学位论文全文数据库(硕士)》 * |
Also Published As
Publication number | Publication date |
---|---|
CN113868965B (en) | 2022-02-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Donda et al. | Ultrathin acoustic absorbing metasurface based on deep learning approach | |
Naseri et al. | A generative machine learning-based approach for inverse design of multilayer metasurfaces | |
Jiménez et al. | Metadiffusers: Deep-subwavelength sound diffusers | |
Wu et al. | A reappraisal of the validity of the IEM model for backscattering from rough surfaces | |
CN114325886B (en) | Super-surface, design method and device thereof and electronic equipment | |
Liu et al. | Deep-learning-based acoustic metamaterial design for attenuating structure-borne noise in auditory frequency bands | |
Wiest et al. | Robust design of an asymmetrically absorbing Willis acoustic metasurface subject to manufacturing-induced dimensional variations | |
Li et al. | Unconditionally stable higher order CNAD-PML for left-handed materials | |
CN113158563B (en) | One-dimensional photonic crystal reverse design method based on neural network and Monte Carlo | |
CN113868965B (en) | Design method and system of black phosphorus wave absorber | |
Lazarov et al. | Topology optimization of pulse shaping filters using the Hilbert transform envelope extraction | |
Kumar et al. | Comparative analysis of machine learning algorithms on prediction of the sound absorption coefficient for reconfigurable acoustic meta-absorbers | |
US12013352B2 (en) | Space-time scattering network for inverse design and tomography | |
Seçgin et al. | Determination of effective medium parameters of an in-line cavity based acoustic metamaterials with visco-thermal losses | |
Demésy et al. | Dispersive perfectly matched layers and high-order absorbing boundary conditions for electromagnetic quasinormal modes | |
Qian et al. | Deep learning for the design of random coding Metasurfaces | |
CN111950198A (en) | Ground clutter simulation method based on neural network | |
Shyroki | Squeezing of open boundaries by Maxwell-consistent real coordinate transformation | |
Hamilton et al. | Space-time scattering network for electromagnetic inverse design and tomography | |
Guo et al. | Parametric modeling and deep learning-based forward and inverse design for acoustic metamaterial plates | |
Hospital-Bravo et al. | Numerical modeling of undersea acoustics using a partition of unity method with plane waves enrichment | |
CN118278296B (en) | Machine learning method and system for inverting wave-absorbing honeycomb electromagnetic parameters | |
Kojima et al. | Machine learning for design and optimization of photonic devices | |
CN116913436B (en) | Super-atom reverse design method based on LDM-PNN and particle swarm optimization | |
Kubytskyi et al. | Fast optimization of microwave absorbers |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |