Ultra-fast Simulation and Inverse Design of Metallic Antennas

We report a new approach for rapid evaluation and inverse design of metallic antennas. The technique enables electromagnetic solution of candidate antenna structures in less than a second per design without making any approximations, achieving multiple orders of magnitude speedup compared to standard full-wave approaches. The method involves a completely parallelized precomputation stage, which has been GPU-accelerated, used to compute a basis set of numerical Green's functions, which can then be used to reconstruct the field solution for any antenna in the design space by solving an inexpensive linear system. We couple this fast evaluation approach with a directed binary search algorithm and use it to inverse design new, high-gain antennas. We conclude by presenting a practical design of an ultra-wideband planar metallic antenna on ground backed dielectric substrate which achieves 50% fractional bandwidth at a 30 GHz center frequency and greater than 6.8 dB gain across the whole range.