Flexible Design of Rotating Bi-anisotropic Media Chirality Response Using Deep Learning

Bi-anisotropic structures are vital in modern optics and electromagnetics due to their unique electromagnetic properties, especially in double-layer rotating transmission structures that induce chiral responses. This article introduces a deep learning method for predicting the rotation angle of chiral structures using a forward spectral prediction network and a reverse rotation angle prediction network. The forward network accelerates the prediction of transmission coefficients for left- and right-handed waves, while the reverse network predicts rotation angles and modes from these coefficients. Both networks demonstrate high accuracy and efficiency, aiding the exploration of circularly polarized wave behaviors in transmission systems.