Neural Network Topology-Based Terahertz Absorber Using Fractal Frequency Selective Surface

This work offers a new kind of polarization-insensitive, angular stable, tunable fractal frequency selective surface (FSS) based terahertz (THz) absorber. An equivalent circuit model (ECM) of the absorber is developed and impedance is obtained analytically. Further, an ECM-backed artificial neural network (ANN) approach is proposed and used to optimize the absorber to the desired frequency and bandwidth. The minimum mean square error between targeted and observed outputs is only
$1.86\times10$
^-6
. The maximum absorption of 99% is achieved at a central frequency of 1.5 THz. The tunability is accomplished by varying the chemical potential (
$\mu _{c}$
) of the graphene layer introduced in the proposed absorbing structure. The resonance frequency shifts to 1.66 THz from 1.39 THz with a change in
$\mu _{c}$
from 0.3eV to 0.6eV.The dual-band absorption peaks are noticed with fractal FSS based absorber within the operating frequency range of 0.1-4.0 THz. The maximum absorption of 0.99 and 0.98 at the resonant frequency of 1.0 THz and 3.1 THz with fractal FSS, respectively is achieved. Furthermore, the designed absorber works well for both perpendicular and parallel polarizations up to 60
⁰
angle of incidence. Moreover, the effect of structural parameters is also observed on the equivalent inductance and capacitance to understand the physics behind the structure. All the results obtained from ANN are verified with particle swarm optimization (PSO) and Whale optimization (WO) techniques. Good congruence among ANN, PSO, and WO results shows the effectiveness of the aforementioned technique. The possible applications of the proposed tunable absorber are THz imaging, sensing, detection, and stealth.