Novel two-dimensional photonic crystals designed by evolutionary algorithms

摘要

We use evolutionary algorithms to search the space of two-dimensional photonic crystals with maximum bandgaps for a given index contrast. The unit-cells of the photonic crystal are represented as bitmaps which are either directly encoded or generated through bottom-up and top-down construction trees. The fitness criterion rewards for partial band gaps and is evaluated by solving for the photonic crystals bands using an eigensolver in a planewave basis. Starting from random patterns and with no prior knowledge, the process discovered a number of novel photonic crystals, some with bandgaps that are 12.5% larger than any previously reported human-designed crystals.