Laser & Optoelectronics Progress, Volume. 61, Issue 23, 2300003(2024)
Advances in Artificial Intelligence for Design and Optimization of Terahertz Metamaterials
Figures & Tables(20)
![Schematic structure of a cross-shaped flexible terahertz metamaterial filter[50]](/Images/icon/loading.gif){kind=icon}
Fig. 2. Schematic structure of a cross-shaped flexible terahertz metamaterial filter[50]
Fig. 5. Heat map of the Adj-R2S of the ERT model[85]. Adj-R2S of ERT model using various values of (a)‒(c) substrate thickness, (d)‒(f) periodic dimension, (g)‒(i) incident angle, and (j)‒(l) polarization angle under different test cases of 30%, 40%, and 50%
Fig. 6. Schematic diagram of the process of combining ML to predict metasurface structural parameters[86]
Fig. 7. 1-bit random coded HMM based on the combination of MOPSO and Python-CST co-simulation[89]
Fig. 8. Intelligent real-time terahertz beamforming scheme based on self-adaptive deep reinforcement learning models[90]
Fig. 11. GA-based multi-objective optimization prediction metasurface patterns and electromagnetic responses[97]
Fig. 13. Dielectric metasurfaces composed of pixelated unit cells[102]. (a) A pixelated metasurface unit cell made of silicon; (b) binary matrix description of the unit cell pattern
Table 1. Absorption rates and performance specifications of single-band absorbers, dual-band absorbers, broadband absorbers, and multi-band absorbers
View tableTable 1. Absorption rates and performance specifications of single-band absorbers, dual-band absorbers, broadband absorbers, and multi-band absorbers
Function Frequency band Absorptivity Performance indicators Reference Single-band 0.9628 THz ‒ S=800 GHz/RIU， Q=2407
FWHM is 0.4 GHz
［ 36 ］Dual-band 0.89 THz
1.36 THz
>99% Q1=203.2
Q2=121.4
［ 37 ］Four-band 2.82 THz
6.44 THz
7.84 THz
8.6 THz
90.95%
97.25%
91%
99.6%
S1=881.35 GHz/RIU
S2=439.65 GHz/RIU
S3=2664.46 GHz/RIU
S4=4501.19 GHz/RIU
［ 38 ］Ultra-broadband 0.1‒16 THz >95% Angle of incidence ranges is about 80° ［ 39 ］
Table 2. Modulation objects of metamaterial modulators and schematic structure
View tableTable 2. Modulation objects of metamaterial modulators and schematic structure
Modulation object Variable Structure Structural diagram Reference Amplitude Temperature Metamaterial ［41］ Beam Coding sequence Metasurface ［42］ Frequency/phase/amplitude Position of the silicon wafers，Fermi energy level Metasurface ［43］ Amplitude Fermi energy level Metasurface ［44］ Phase Fermi energy level Metasurface ［45］
Table 3. Comparison of Important Sensor Metrics
View tableTable 3. Comparison of Important Sensor Metrics
Structural diagram Frequency band Performance indicator Analyte Reference 0.635 THz Q=79.26， S=91.5 GHz/RIU Glucose solution ［51］ ‒ S1=105.3 GHz/RIU， FOM1 is 2.92S2=122.5 GHz/RIU， FOM2 is 4.36 Bovine serum albumin solutions ［52］ 0.88 THz S=0.0446 dL/mglimit of detection is 1.64 mg/mL Glucose solution ［53］ 6.29‒7.512 THz S=2.538 THz/RIU， Q=24.22 ‒ ［54］ 0.59 THz1.07 THz1.34 THz S1=93.5 THz/RIUS2=173.0 THz/RIUS3=182.4 THz/RIU Lactose solutions ［55］ 3.62 THz3.814 THz Q=51.7， S=3 THz/RIUQ=1411.11， S=3.59 THz/RIU Methane， chloroform ［56］
Table 4. Specific functions of the three-function integrated metamaterial absorber
View tableTable 4. Specific functions of the three-function integrated metamaterial absorber
Patterned VO2 layers Unpatterned VO2 layers Frequency band Function Absorptivity Metallic Metallic 2.7‒6.2 THz Low-frequency
broadband absorber
>90% Insulator Metallic 5.8‒7.6 THz High-frequency broadband absorber >90% Insulator Insulator ‒ Multi-band absorber ‒
Table 5. Comparison of structures and functions of tunable multifunction devices
View tableTable 5. Comparison of structures and functions of tunable multifunction devices
Function Variable Structural diagram Reference Dual-band/trip-band/four-band absorption VO2 ［64］ Broadband absorption/polarization conversion VO2 ［65］ Absorption/transmission/reflection Graphene， VO2 ［66］ Absorption/polarization conversion Graphene ［67］ Ultra-broadband absorption/polarization conversion GST ［68］ Broadband/multi-band absorption VO2 ［69］
Table 6. Comparison of algorithms used in metamaterial devices with device functionality
View tableTable 6. Comparison of algorithms used in metamaterial devices with device functionality
Algorithm Optimization object Function of device Significance Reference ERT ML Absorber Proposed efficient multi-band MMA design with machine learning behavior prediction can be used for sensing applications. ［ 85 ］RF Absorber Combination of machine learning and photonic device design to predict absorption bandwidths is feasible and effective and provides new ways to design complex systems related to the propagation of absorbed， reflected， and transmitted electromagnetic waves. ［ 86 ］［ 87 ］PIML Nanoresonator Inverse design method holds immense promise for designing terahertz nanophotonic devices for next-generation communication technologies and molecular sensing applications ［ 95 ］GA EA Absorber Results may provide guidance for the automated design studies of ultra-broadband tunable terahertz metasurface absorbers. ［ 88 ］MOPSO Absorber/polarization converter This work provides a solution to design low RCS metasurfaces， specifically for the dual-band requirement， with great potential in shape stealth in the THz regime. ［ 89 ］PSO Modulator Results suggest a feasibility to build the terahertz EIT effect in the metasurface through an optimization algorithm of inverse design. Furthermore， this method can be further utilized to design multifunctional and high-performance terahertz devices. ［ 96 ］CNN， GA Filters/modulators Multi-objective optimization method improves the design diversity of the metasurface and increases the design efficiency. ［ 97 ］DRL DL Beamforming A promising approach is provided for real-time terahertz beamforming in MIMO systems for next-generation wireless communications， and similar schemes may be useful for terahertz imaging and sensing applications. ［ 90 ］TCGN Chiral wavefront control Both simulation and experimental results show that this method can effectively design the metasurface structure on-demand through specific targets， and the prediction results are diverse. ［ 104 ］PNN Collimator An efficient and reliable solution for designing complex meta-atomic structures in high-performance optical device implementations. ［ 91 ］DNN Absorber/polarization converter There is great potential for energy absorbers， high-speed intelligent detection devices and optoelectronics in the terahertz frequency range. ［ 79 ，92 ］ANN Absorber This innovation holds significant promise for applications in security detection， and object stealth. ［ 100 ‒101 ］CNN Wavefront shaping Results prove the effectiveness of the inverse model in predicting the generic nonintuitive patterns， offering a powerful and general tool for multifunctional beam manipulation. ［ 102 ］
Tools
Get Citation
Copy Citation Text
Hongyi Ge, Yuwei Bu, Yuying Jiang, Xiaodi Ji, Keke Jia, Xuyang Wu, Yuan Zhang, Yujie Zhang, Qingcheng Sun, Shun Wang. Advances in Artificial Intelligence for Design and Optimization of Terahertz Metamaterials[J]. Laser & Optoelectronics Progress, 2024, 61(23): 2300003
Paper Information
Category: Reviews
Received: Mar. 20, 2024
Accepted: Apr. 3, 2024
Published Online: Nov. 27, 2024
The Author Email: Yuying Jiang (jiangyuying11@163.com), Yuan Zhang (zy_haut@163.com)
CSTR:32186.14.LOP240937
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20