Inverse-design of non-Hermitian potentials for light management

In this paper, we propose a general inverse-design strategy based on genetic algorithm optimization to achieve 'on demand' manipulation of light in one-dimensional (1D) and two-dimensional (2D) non-Hermitian systems. The optimization process faithfully creates non-Hermitian potentials from any given arbitrary real (or imaginary) permittivity distribution for the desired frequency selective and broadband asymmetric response in 1D multilayer structures. As a demonstration in 2D, we design periodic and aperiodic complex permittivity spatial distributions to create "sink-type" concentrators of light around a desired area. The proposed inverse-design approach to generate non-Hermitian potentials represents an alternative to the Hilbert Transform (HT) generalizing the Kramers Kronig relations in space, additionally being selective in spectrum.