Deep Learning Design of Orbital Angular Momentum Generators by Leaky-Wave Holograms

This letter introduces a pioneering approach for the design of holographic leaky-wave antennas, effectively generating electromagnetic waves carrying orbital angular momentum by synergizing flat optics and deep learning (DL) techniques. In this study, we employ the DL technique to tune the constituent parameters of the hologram surface impedance, uncovering a framework adept at controlling radiation pattern characteristics. By utilizing a total of 77 000 generated datasets, we deduce the optimal values for each parameter, resulting in the attainment of the desired radiation pattern. Importantly, the utilization of DL not only accelerates the process but also yields outcomes of greater accuracy compared to manual parameter adjustments and conventional optimization techniques.