Data enhanced iterative few-sample learning algorithm-based inverse design of 2D programmable chiral metamaterials

Abstract A data enhanced iterative few-sample (DEIFS) algorithm is proposed to achieve the accurate and efficient inverse design of multi-shaped 2D chiral metamaterials. Specifically, three categories diffractive structures with different geometrical parameters, including widths, separation spaces, bridge lengths, gold lengths are studied utilising both conventional rigorous coupled wave analysis (RCWA) approach DEIFS algorithm, former assisting training process for latter. The can be divided into two main stages, namely enhancement iterations. Firstly, some “pseudo data” generated by a forward prediction network that efficiently predict circular dichroism (CD) response metamaterials reinforce dataset after necessary denoising. Then, uses CD spectra predictions parameters smaller errors iteratively values remaining parameters. Meanwhile, according impact geometric on chiroptical response, new functionality added interpret experimental results from perspective data, improving interpretability DEIFS. In this way, replaces time-consuming optimization faster simpler achieves whose amount at least one orders magnitude less than most previous deep learning methods, reducing dependence simulated spectra. Furthermore, fast multiple shaped allows light manipulation, demonstrating excellent potentials in applications optical coding information processing. This work belongs first attempts thoroughly characterize flexibility, interpretability, generalization ability studying various effects accelerating hypersensitive photonic devices.