Broadband Terahertz Antenna Design with Defected Ground Structure and Parasitic Elements Using a Feedforward Neural Network

In this paper, a novel wideband terahertz antenna with an imperfect ground structure and parasitic elements has been proposed. The optimized antenna geometrical parameters are obtained using an artificial neural network which employs forward feedback propagation along with the Levenberg-Marquardt optimization algorithm including 10 neurons. The objective of this design is to increase the obtained return loss bandwidth of designed antenna. This demanded objective is obtained using imperfect ground structure and parasitic elements together. The antenna structure features a substrate with a dielectric constant of 9.4 and a loss tangent of 0.00002, designed within dimensions of 100 by 100 square micrometers. The antenna configuration is derived from the neural network implemented in MATLAB program using 143 training samples. As a result, the obtained performance bandwisth is 41% more than that of ordinary terahertz microstrip antenna. Furthermore, the obtained radiation patterns are investigated over operational frequency band showing a suitable performance.