Superradiant Broadband Magnetoelectric Arrays Empowered by Meta-Learning

Laws of electrodynamics constrain scattering cross sections of resonant objects. Nevertheless, a fundamental bound that expresses how large that scattering cross section can be is yet to be found. Approaches based on cascading multiple resonances permitted to push the scattering responses of subwavelength structures and to exceed existing estimators. In this context, the Chu-Harrington criterion is, potentially, the most commonly considered one. The superradiant empirical criterion, addressing scattering performances of near-field coupled resonator arrays, was subsequently developed to tighten existing estimates, setting a new bound that prompted efforts to find structures that exceed it. Here, we demonstrate that genetically designed superscattering structures, encompassing arrays of constructively interfering electric and magnetic dipoles, can build up high scatting cross sections exceeding those imposed by existing criteria in electromagnetic theory, including the superradiant. After undergoing thousands of evolutionary generations, iterating sizes, mutual orientations, and locations of resonators, the structures approach their heuristically maximized performance, which is unlikely to be obtained by a random distribution given more than a billion trials. As an additional practically valuable parameter, the scattering bandwidth also underwent optimization. We demonstrate that flat wavelength-comparable structures can have significant backscattering ( $\sim 4\times $ prevailing the superradiant criterion) alongside more than 40% fractional bandwidth, given an endfire excitation. The result demonstrates the fundamental perspective to untighten scattering cross section from bandwidth limitations. New capabilities of genetic optimization algorithms, equipped with fast computational tools and constrained by experimentally obtainable electromagnetic parameters, allow chasing well-accepted traditional criteria, demonstrating ever-seen electromagnetic performances.