A Design of Silicon Based Horn Antenna Optimized Using Bayesian Optimization

In this work, a 60-GHz silicon-based step-profiled horn antenna is designed and optimized using Bayesian Optimization (BO). Specifically, the designed antenna is fed through a substrate integrated waveguide (SIW). The electromagnetic wave is transformed upward by a transition structure which is also used for impedance matching between the silicon substrate and air. To achieve better impedance matching and broadside radiation gain, a step-profiled horn is mounted on the transition by stacked silicon layers with air cavities fenced by vias. Before optimization, the impedance bandwidth of the antenna is 0. To effectively optimize this antenna that has 10 parameters to optimize, a well-designed objective function is proposed and then BO is adopted to find the optimal result. Within 90 times iteration, an optimized result satisfying an impedance bandwidth of 10.4% is obtained. At the same time, the simulated realized broadside gain is between 5.9 to 6.9 dB in the corresponding bandwidth.