Efficient Inverse Design of Large-Area Metasurfaces for Incoherent Light

Abstract

Incoherent light is ubiquitous, yet designing optical devices that can handle its random nature is very challenging, since directly averaging over many incoherent incident beams can require a huge number of scattering calculations. We show how to instead solve this problem with a reciprocity technique that leads to 3 orders of magnitude speedup: one Maxwell solve (using any numerical technique) instead of thousands. This improvement enables us to perform efficient inverse design, large-scale optimization of metasurfaces for applications such as light collimators and concentrators. We show the impact of the angular distribution of incident light on the resulting performance and show especially promising designs for the case of “annular” beams distributed only over nonzero angles.

Topological Learning for the Classification of Disorder: An Application to the Design of Metasurfaces

“…The promising applications of disordered metasurfaces have led to more recent effort to tailor disorder for specific, desired optical properties, − for example, using inverse design methods , based on machine learning , or via topology optimization. , Indeed, by combining disorder engineering and topology optimization, one can build metasurfaces with selective light polarization conversion, while minimizing the in-plane phase fluctuation . Such methods can directly generate optimized disordered patterns, but they can be time-consuming and computationally expensive to implement.…”

“…For example, a correlation length, induced either by a minimum distance between the nanostructures or by some stealthy hyperuniformity properties, helps to create metasurfaces with larger absorption bands 28 or broader diffusive properties 19 or prevent light trapping between nanostructures for more efficient light extraction. 30 The promising applications of disordered metasurfaces have led to more recent effort to tailor disorder for specific, desired optical properties, 31−34 for example, using inverse design methods 35,36 based on machine learning 37,38 or via topology optimization. 39,40 and topology optimization, one can build metasurfaces with selective light polarization conversion, while minimizing the inplane phase fluctuation.…”