Lithography-free subwavelength metacoatings for high thermal radiation background camouflage empowered by deep neural network.

The long wavelength infrared (LWIR) range (8-14 µm) is crucial for thermal radiation detection, necessitating effective camouflage against advanced infrared technologies. Conventional camouflage approaches often rely on complicated photonic structures, facing significant implementation challenges. This study introduces a novel polarization-insensitive and angle-robust metacoating emitter for LWIR camouflage, inversely designed through a deep neural network (DNN) framework. The DNN framework facilitates the automatic optimization of the metacoating's structural and material parameters. The resulting emitter achieves an average emissivity of 0.96 covering the LWIR range and a low emissivity of 0.25 in the other mid-infrared (MIR) region. Enhanced electromagnetic wave localization and energy dissipation, driven by high-lossy materials like bismuth and titanium, contribute to these properties. Infrared imaging confirms the emitter's superior camouflage performance, maintain effectiveness at incident angle up to 70° while exhibiting strong polarization independence. This inverse-designed metacoating demonstrates significant potential to advance infrared camouflage technology, providing robust countermeasures against modern, wide-angle, and polarization-sensitive detection systems.