Machine Learning–Assisted Performance Prediction of Graphene–Silicon Twin-Port Band-Notched Wideband Antenna in THz Domain

A 2-port graphene-silicon-built THz antenna is planned and examined in this communication. Aperture feeding and air cavities in silicon dielectric are used to achieve wideband features i.e. 2.35–3.8 THz. Three unique features of designed antenna are: (i) circular metallic rings added with printed line to notched the spectrum part, i.e., 2.75–3.15 THz; (ii) Deep neural network and Random Forest (RF) are applied to guess |S11| parameter of the designed radiator; and (iii) presence of DGS reduces the coupling below − 25 dB. Graphene layering assists in providing frequency tunability qualities to the planned antenna. The optimized design is validated using both CST and HFSS electromagnetic simulators, confirming that the proposed antenna operates effectively within the 2.2–2.65 THz and 3.3–3.7 THz bands. The antenna exhibits stable radiation characteristics along with favorable multi-port performance metrics, demonstrating its suitability for THz-based mobile communication systems.