Near-infrared band MIM structure perfect absorber and design method thereof
By employing square and circular ring structures with composite resonant layers in a near-infrared MIM structure perfect absorber, combined with automated simulation using Python scripts and deep learning models, the absorption bandwidth was broadened, solving the problem of limited spectral response range and absorption performance in existing technologies, and achieving rapid and efficient design optimization.
Patent Information
- Application Number
- CN202511501839.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-21
- Publication Date
- 2025-12-16
AI Technical Summary
Existing near-infrared perfect absorbers have limited spectral response range and overall absorption performance because their top resonant layer structure is mostly a single-shaped resonant unit.
A near-infrared MIM structure perfect absorber is designed, which uses a composite resonant layer consisting of periodically arranged coaxial square rings and circular rings to broaden the absorption bandwidth through electromagnetic field coupling. Rapid spectral prediction is achieved through automated simulation using Python scripts and deep learning model training.
This improved the spectral response range and overall absorption performance of the absorber, solved the problem of insufficient bandwidth in traditional single resonant structures, and enabled rapid and efficient design optimization.
Smart Images
Figure CN121142702A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of nanophotonics and optoelectronic metamaterials devices, and in particular to a near-infrared MIM structure perfect absorber and its design method. Background Technology
[0002] A near-infrared perfect absorber is a functional device that can almost completely absorb incident light energy in the near-infrared band with almost no reflection or transmission. It has important application value in fields such as infrared detection, thermal imaging, spectral analysis and optical energy utilization.
[0003] With the development of micro-nano fabrication technology, metamaterials and metasurfaces have provided a new paradigm for absorber design. Metasurfaces are a class of artificial periodic structures composed of subwavelength units. By controlling the geometric parameters and arrangement of these units, precise control over the amplitude, phase, polarization, and other properties of light can be achieved at ultrathin scales. Among them, the metal-insulator-metal (MIM) structure has become the mainstream design scheme due to its simple structure, ease of fabrication, and excellent performance. In this structure, the bottom metal layer blocks transmitted light; the top metal pattern structure acts as a resonator to excite localized surface plasmon resonance coupled with incident light; and the middle dielectric layer provides the resonance space and enhances the localization effect of the light field.
[0004] However, existing near-infrared perfect absorbers have limited spectral response range and overall absorption performance because their top resonant layer structure is mostly a single-shaped resonant unit (such as square, circular or cross-shaped). Summary of the Invention
[0005] The purpose of this invention is to provide a near-infrared band MIM structure perfect absorber and its design method, which solves the problem that existing near-infrared band perfect absorbers have limited spectral response range and overall absorption performance due to the fact that their top resonant layer structure is mostly a single-shaped resonant unit.
[0006] To achieve the above objectives, the present invention provides a near-infrared band MIM structure perfect absorber, the near-infrared band MIM structure perfect absorber comprising a bottom metal reflective layer, a dielectric layer and a composite resonant layer, wherein the dielectric layer is disposed on the upper surface of the bottom metal reflective layer, and the composite resonant layer is disposed on the upper surface of the dielectric layer, the composite resonant layer being composed of periodically arranged unit structures, each unit structure comprising a coaxially arranged square ring and a circular ring, the square ring and the circular ring being symmetrically distributed;
[0007] The bottom metal reflective layer is used to block transmitted light;
[0008] The dielectric layer is used to achieve resonant coupling;
[0009] The square ring is used to enhance absorption in the short-wavelength band;
[0010] The ring is used to enhance absorption in the long wavelength range;
[0011] The geometric parameters of both the square ring and the circular ring are adjustable design parameters.
[0012] The bottom metal reflective layer and the composite resonant layer are both made of chromium.
[0013] The material of the dielectric layer is aluminum oxide.
[0014] The thickness of the bottom metal reflective layer is set to 150 nm, and the thickness of the dielectric layer is adjustable.
[0015] This invention also provides a design method for a near-infrared band MIM structure perfect absorber, applied to the near-infrared band MIM structure perfect absorber as described above, comprising the following steps:
[0016] Structural modeling and parameter setting: The API interface of the electromagnetic simulation software is called through the script program to automatically generate the metasurface structure containing the composite resonant layer, and the range of geometric parameters of the composite resonant layer is set.
[0017] Cyclic Simulation and Spectrum Extraction: Simulation is automatically executed through scripts to perform cyclic simulations on different combinations of geometric parameters, obtain the corresponding reflection and absorption spectra, and automatically extract and store the optical response data;
[0018] Dataset construction: Different combinations of geometric parameters are combined with corresponding spectral data to form a training dataset;
[0019] Model training: Input the dataset into the designed deep learning network architecture, use mean squared error as the loss function, and train it through optimization algorithms until convergence;
[0020] Spectral prediction: By using a trained model and inputting new parameter combinations, the corresponding spectrum can be output.
[0021] The script program is a Python script, and the electromagnetic simulation software is Lumerical FDTD.
[0022] This invention discloses a near-infrared MIM structure perfect absorber and its design method, comprising a bottom metal reflective layer, a dielectric layer, and a composite resonant layer. The composite resonant layer is composed of periodically arranged unit structures, each of which includes a coaxially arranged square ring and a circular ring, symmetrically distributed. When incident light irradiates the composite resonant layer, the square ring and the circular ring respectively excite localized surface plasmon resonances in different wavelength bands. Furthermore, since the square ring enhances absorption in the short-wavelength band and the circular ring enhances absorption in the long-wavelength band, the resonance peaks of the short-wavelength and long-wavelength bands superimpose through electromagnetic field coupling, thereby broadening the absorption bandwidth of the near-infrared MIM structure perfect absorber. This technical solution, through the interaction of two square rings and a circular ring of different shapes, improves the spectral response range and overall absorption performance of the resonant structure in the absorber, solving the technical problem of insufficient absorption bandwidth in traditional single resonant structures. Attached Figure Description
[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0024] Figure 1 This is a schematic diagram of the near-infrared band MIM structure perfect absorber provided by the present invention.
[0025] Figure 2 This is a simulation diagram showing the absorption spectra obtained by the near-infrared MIM structure perfect absorber provided by the present invention under different parameters.
[0026] Figure 3 This is a comparison of simulation results of the absorption spectra obtained by the near-infrared MIM structure perfect absorber provided by the present invention and the perfect absorber with a single-shaped top resonant layer in the prior art under different parameters.
[0027] Figure 4 This is a flowchart illustrating the design steps of the near-infrared band MIM structure perfect absorber provided by the present invention.
[0028] Figure 5 This is a loss curve generated during the training process of the model provided by this invention.
[0029] Figure 6 This is a schematic diagram comparing the simulated spectrum and the model predicted spectrum during the verification of the design method of the near-infrared band MIM structure perfect absorber provided by this invention.
[0030] Figure 7 This is a comparison chart showing the changes in absorption curves when the side length of the square ring is a variable during the design process of the near-infrared band MIM structure perfect absorber provided by this invention.
[0031] 101-Bottom metal reflective layer, 102-Dielectric layer, 103-Square ring, 104-Circular ring. Detailed Implementation
[0032] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.
[0033] Please see Figures 1 to 3 This invention provides a near-infrared band MIM structure perfect absorber, which includes a bottom metal reflective layer 101, a dielectric layer 102, and a composite resonant layer. The dielectric layer 102 is disposed on the upper surface of the bottom metal reflective layer 101, and the composite resonant layer is disposed on the upper surface of the dielectric layer 102. The composite resonant layer is composed of periodically arranged unit structures. Each unit structure includes a square ring 103 and a circular ring 104 arranged coaxially, and the square ring 103 and the circular ring 104 are symmetrically distributed.
[0034] The bottom metal reflective layer 101 is used to block transmitted light;
[0035] The dielectric layer 102 is used to achieve resonant coupling;
[0036] The square ring 103 is used to enhance absorption in the short-wave band;
[0037] The ring 104 is used to enhance absorption in the long wavelength range;
[0038] The geometric parameters of both the square ring 103 and the circular ring are adjustable design parameters.
[0039] In this embodiment, when incident light irradiates the composite resonant layer, the square ring 103 and the circular ring 104 respectively excite localized surface plasmon resonances in different wavelength bands. Furthermore, since the square ring 103 enhances absorption in the short-wavelength band and the circular ring 104 enhances absorption in the long-wavelength band, the resonance peaks of the short-wavelength and long-wavelength bands superimpose through electromagnetic field coupling, thereby broadening the absorption bandwidth of the near-infrared MIM structure perfect absorber. By employing this technical solution, the interaction of the two differently shaped square rings 103 and 104 improves the spectral response range and overall absorption performance of the resonant structure in the absorber, solving the technical problem of insufficient absorption bandwidth in traditional single resonant structures.
[0040] Furthermore, the geometric parameters (such as outer diameter, width, and height) of the square ring 103 and the circular ring 104 are all adjustable design parameters. The design parameters are automatically modified through code scripts for cyclic simulation, and then a deep learning model is used for rapid prediction and optimization.
[0041] Furthermore, both the bottom metal reflective layer 101 and the composite resonant layer are made of chromium (Cr).
[0042] The material of the dielectric layer 102 is aluminum oxide (Al2O3).
[0043] Furthermore, the thickness of the bottom metal reflective layer 101 is set to 150 nm, and the thickness of the dielectric layer 102 is adjustable. Figure 1 The thickness of the dielectric layer 102 described herein is 70 nm.
[0044] Simulation results of the near-infrared MIM structure perfect absorber:
[0045] Lumerical FDTD simulations were used, with the wavelength range set to 800–2200 nm. Absorption spectra were obtained under different parameter combinations, for example... Figure 2 As shown. Based on the results of the optimal structure in the database, calculations show that the average absorption rate of the combined resonance of the square ring 103 and the circular ring 104 is greater than 94% in the target wavelength band. Figure 3 It is evident that the absorption peak coverage of the combined resonance is significantly wider than that of a single square ring or a single circular ring 104, and the absorption effect is superior.
[0046] Please see Figures 4 to 7 The present invention also provides a design method for a near-infrared band MIM structure perfect absorber, applied to the near-infrared band MIM structure perfect absorber as described above, comprising the following steps:
[0047] 1. Dataset Construction: By calling the Lumerical API interface through Python scripts, automatic parameter traversal and batch simulation are achieved to obtain reflectance spectrum datasets corresponding to different geometric parameters.
[0048] 2. Model Training: Construct a suitable neural network as the forward prediction model, taking structural parameters (such as R1, R2, S1, S2, h, etc.) as input and outputting the corresponding spectral response. Use mean squared error as the loss function and train through backpropagation until convergence.
[0049] 3. Prediction and Optimization: Input different parameter combinations into the trained positive prediction model, quickly output spectral results, and compare and verify with simulation results.
[0050] In this embodiment, traditional metasurface design methods require a separate model to be built and FDTD simulations to obtain the corresponding spectral response for each parameter combination. Due to the high dimensionality of the parameters and the large number of combinations, this approach often requires significant computational resources and time, resulting in low design efficiency and difficulty in meeting the rapid optimization needs of complex structures.
[0051] The specific working principle of the design method for the near-infrared band MIM structure perfect absorber in this technical solution is as follows:
[0052] 1. Automated data generation
[0053] By using a Python script to call the Lumerical FDTD API, automatic parameter iteration and modeling are achieved. The script can complete large-scale simulation tasks unattended and automatically extracts reflectance and absorption spectra after each simulation, thereby generating a large-scale training dataset containing structural parameters and spectral data. This process significantly reduces manual intervention and improves data acquisition efficiency and consistency.
[0054] 2. Deep learning model training
[0055] A forward prediction model is trained using the constructed dataset. This model takes structural parameters as input and spectral curves as output, continuously optimizing network weights through backpropagation to ultimately achieve a high-precision parameter-spectral mapping. After training, the model possesses fast inference capabilities.
[0056] 3. Rapid spectral prediction
[0057] In the application phase, simply inputting the new structural parameters into the trained forward prediction model will output the predicted spectrum within milliseconds. Compared to traditional methods, there is no need to rerun the FDTD simulation, thus significantly reducing computational resource consumption and design cycle time.
[0058] 4. Key Mechanisms
[0059] The core mechanism of this design method lies in utilizing the nonlinear mapping capabilities of deep learning to transform complex electromagnetic field simulations into a direct relationship between parameters and spectra. This approach achieves a shift from "simulation-driven" to "data-driven," fundamentally changing the traditional working mode that relies on successive simulations.
[0060] Based on the above principles, this design method effectively solves the problems of low design efficiency, cumbersome manual operation, and high computational cost in the existing technology, and finally realizes fast, efficient, and low-cost metasurface design and optimization.
[0061] Furthermore, such as Figure 5 The image shows the loss curves generated during model training. Train Loss represents the training loss, and Val Loss represents the test loss.
[0062] The graph shows that the designed model converges well on both the training and validation sets (both on the order of 1e-05), and the val_loss curve and the train_loss curve are close, indicating that there is no obvious overfitting.
[0063] Experimental data:
[0064] By comparing the simulated spectrum with the model's predicted spectrum, the accuracy of the forward model design is further demonstrated intuitively. Figure 6 As shown, the dashed line represents the predicted value, which is realized as the true value. The predicted value and the true value highly overlap, and the average error is less than 5%, which verifies the effectiveness and reliability of this method.
[0065] By using a deep learning model to predict the structure, a single simulation can be predicted in just 0.3 seconds, a process that would take an average of one minute to calculate using traditional methods.
[0066] Under fixed external environmental conditions, the geometric parameters of the square ring 103 and the circular ring 104 in the composite resonant layer of this design method are adjustable, mainly including the radius and width of the circular ring 104, the side length, width, and height of the square ring 103. By adjusting the above parameters, the spectral response characteristics of the device can be significantly changed (when the structural geometric parameters change, the localized surface plasmon (LSP) resonance conditions change accordingly. For the square ring 103, changes in its side length and width will change the distribution path of the ring current, thereby causing the short-wavelength resonance peak to shift; for the circular ring 104, its radius and thickness determine the effective length of the plasma oscillation, thereby affecting the long-wavelength resonance conditions; the coupling effect between the square ring 103 and the circular ring 104 allows the resonances of the two bands to superimpose or even influence each other, thereby achieving continuous tunability of the absorption peak within a certain range).
[0067] like Figure 7 As shown (Red: R1=90nm, R2=90nm, S1=160nm, S2=160nm, h=220nm; blue: R1=90nm, R2=90nm, S1=80nm, S2=80nm, h=220nm), in a specific experiment, taking the side length of the square ring 103 as an example, with other parameters fixed, when the outer side length of the square ring 103 is adjusted from 240 nm to 320 nm, the simulation results show that the position of the main absorption peak shifts.
[0068] Simulation results show that, unlike traditional fixed structures, this design utilizes the adjustable parameters of the composite structure to achieve flexible shifting of the absorption peak in the visible and near-infrared bands. This mechanism allows users to dynamically adjust the absorption peak position by changing the structural geometry according to actual application needs, meeting the application requirements of different scenarios such as multi-band sensing, optical communication, and infrared detection.
[0069] Therefore, this embodiment effectively solves the technical problems of fixed absorption spectra and lack of tunability in existing metasurface absorbers, and further improves the application flexibility and adaptability of the device.
[0070] In summary, in the near-infrared MIM structure perfect absorber provided by this technical solution, the composite resonant layer is composed of periodically arranged unit structures. Each unit structure includes a coaxially arranged square ring 103 and a circular ring 104. The bottom metal reflective layer 101 and the composite resonant layer are both made of chromium (Cr). The thickness of the bottom metal reflective layer 101 is designed to be 150 nm to ensure complete blocking of transmitted light. The dielectric layer 102 is made of aluminum oxide (Al2O3) with adjustable thickness and is used to generate local electromagnetic resonance. Multi-resonant coupling broadens the absorption bandwidth, solving the problem of limited bandwidth of a single resonant in existing technologies.
[0071] The design method of the near-infrared band MIM structure perfect absorber provided in this technical solution improves the efficiency of dataset construction and spectral prediction by using Python + Lumerical API for automated modeling and deep learning for positive prediction, thus solving the problem of low design efficiency. At the same time, the absorption peak is dynamically adjustable by adjusting geometric parameters, which solves the shortcomings of traditional structures that are not adjustable.
[0072] The above description discloses only one preferred embodiment of the present invention, and should not be construed as limiting the scope of the present invention. Those skilled in the art will understand that all or part of the processes of the above embodiments can be implemented, and equivalent changes made in accordance with the claims of the present invention are still within the scope of the invention.
Claims
1. A near-infrared band MIM structure perfect absorber, characterized in that, It includes a bottom metal reflective layer, a dielectric layer, and a composite resonant layer. The dielectric layer is disposed on the upper surface of the bottom metal reflective layer, and the composite resonant layer is disposed on the upper surface of the dielectric layer. The composite resonant layer is composed of periodically arranged unit structures. Each unit structure includes a square ring and a circular ring arranged coaxially, and the square ring and the circular ring are symmetrically distributed. The bottom metal reflective layer is used to block transmitted light; The dielectric layer is used to achieve resonant coupling; The square ring is used to enhance absorption in the short-wavelength band; The ring is used to enhance absorption in the long wavelength range; The geometric parameters of both the square ring and the circular ring are adjustable design parameters.
2. The near-infrared band MIM structure perfect absorber as described in claim 1, characterized in that, Both the bottom metal reflective layer and the composite resonant layer are made of chromium. The material of the dielectric layer is aluminum oxide.
3. The near-infrared band MIM structure perfect absorber as described in claim 2, characterized in that, The thickness of the bottom metal reflective layer is set to 150 nm, and the thickness of the dielectric layer is adjustable.
4. A design method for a near-infrared band MIM structure perfect absorber, applied to the near-infrared band MIM structure perfect absorber as described in claim 1, characterized in that, Includes the following steps: Structural modeling and parameter setting: The API interface of the electromagnetic simulation software is called through the script program to automatically generate the metasurface structure containing the composite resonant layer, and the range of geometric parameters of the composite resonant layer is set. Cyclic Simulation and Spectrum Extraction: Simulation is automatically executed through scripts to perform cyclic simulations on different combinations of geometric parameters, obtain the corresponding reflection and absorption spectra, and automatically extract and store the optical response data; Dataset construction: Different combinations of geometric parameters are combined with corresponding spectral data to form a training dataset; Model training: Input the dataset into the designed deep learning network architecture, use mean squared error as the loss function, and train it through optimization algorithms until convergence; Spectral prediction: By using a trained model and inputting new parameter combinations, the corresponding spectrum can be output.
5. The design method of the near-infrared band MIM structure perfect absorber as described in claim 4, characterized in that, The script program is a Python script, and the electromagnetic simulation software is Lumerical FDTD.