Machine learning and evolutionary algorithm studies of graphene metamaterials for optimized plasmon-induced transparency

• DOI: 10.1364/oe.389231
• Publication Date: 2020-04-12T23:00:07Z
• Authors: Tian Zhang; Qi Liu; Yihang Dan; Shuai Yu; Xu Han; Jian Dai; Kun Xu
• Keywords: 0103 physical sciences; FOS: Physical sciences; Physics - Applied Physics; Applied Physics (physics.app-ph); 02 engineering and technology; 0210 nano-technology; 01 natural sciences; Physics - Optics; Optics (physics.optics)

Machine learning and optimization algorithms have been widely applied in the design for photonics devices. We briefly review recent progress of this field research show data-driven applications, including spectrum prediction, inverse performance optimization, novel graphene metamaterials (GMs). The structure GMs is well-designed to achieve wideband plasmon induced transparency (PIT) effect, which can be theoretically demonstrated by using transfer matrix method. Some traditional machine algorithms, k nearest neighbour, decision tree, random forest artificial neural networks, are utilized equivalently substitute numerical simulation forward prediction complete GMs. calculated results demonstrate that all effective has advantages terms accuracy training speed. Moreover, evolutionary single-objective (genetic algorithm) multi-objective (NSGA-II), used steep transmission characteristics PIT effect synthetically taking many different metrics into consideration. maximum difference between peaks dips optimized reaches 0.97. In comparison previous works, we provide a guidance intelligent devices based on reference selection simple problems.