Inverse Design of Tunable Infrared Metasurface Absorbers via a Conditional Wasserstein Generative Adversarial Network

ABSTRACT Narrowband perfect absorbers are interesting for spectrum sensing, molecular detection, and infrared imaging, However, their design is limited by inflexible intuitive‐iterative methods and multi‐objective optimization challenges. Here, we propose a deep learning‐enabled inverse‐design framework via conditional Wasserstein Generative Adversarial Network (WGAN). Our implementation uniquely integrates the conditional WGAN with a uniform sampling algorithm that generates diverse geometric shapes, together with a dual‐channel image encoding scheme that represents geometry and thickness. This combination enables the network to learn rich, viable structure distributions for target optical responses. By leveraging the conditional WGAN's ability to learn the distribution P(), we resolve the inherent “one‐to‐many” problem, generating diverse functional designs per input spectrum with exceptional spectral fidelity (resonance peak nm, MSE ). The model is robust under oblique illumination (–) by transfer learning fine‐tuning. Full‐wave simulations confirm hybrid plasmonic‐dielectric resonance supporting near‐perfect absorption and strong near‐field enhancement. This data‐driven paradigm transcends conventional parametric optimization, establishing a versatile platform for on‐demand inverse design of advanced photonic devices for sensing, spectroscopy, and optical signal processing.