ABSTRACT Creating a stretchable microwave metasurface capable of maintaining stable performance without relying on active tunable components is crucial for low‐cost, large‐scale, conformal, and shape‐adaptive electromagnetic wave manipulation. Current design paradigms leveraging simulations and theoretical predictions with deterministic metasurface geometries suffer from significantly high computational overhead to achieve this goal. Here, we present a stretchable metasurface (SM) that maintains consistent beam‐steering performance under external mechanical loading, designed through a machine learning‐based algorithm guided by simulation and experimental datasets of SM's electromagnetic responses to the deformation. The SM with hierarchical I‐shaped meta‐atoms exhibits robust beam‐steering performance (angular deviation ≤ 2°) under tensile strain up to 20%, enabling stable wireless communication in the 7.75–8.25 GHz band with an error vector magnitude variation of ≤ 1.5 dB. Additionally, an SM capable of consistently reducing the radar cross‐section by over 10 dB across 7.5–8.5 GHz is also realized with the proposed design approach.
Li et al. (Fri,) studied this question.