Ultra-broadband metamaterial absorber for near-infrared and mid-infrared applications optimized via deep learning.

Metamaterials, with their unique subwavelength-scale structures, enable exceptional control over electromagnetic properties, making them ideal for advanced optical devices. This study introduces a novel seven-layer metamaterial absorber, to our knowledge, designed for ultra-broadband absorption across the near-infrared to mid-infrared spectrum (2.3-7.5 µm). Comprising alternating titanium (Ti) and gallium arsenide (GaAs) layers, the absorber achieves an average absorptance of 97.8% and a peak absorptance of 99.8%. A deep neural network (DNN) optimizes structural parameters, ensuring high performance. The absorber's absorption mechanism, analyzed through electromagnetic field distributions, reveals contributions from localized surface plasmon resonance (LSPR), propagating surface plasmon resonance (PSPR), inter-ring coupling, and Fabry-Pérot resonances. The design exhibits robust performance, with insensitivity to incident and polarization angles up to 60° and 90°, respectively. Comparative analysis with recent infrared absorbers highlights its superior bandwidth and absorptance, positioning it as a promising candidate for applications in solar energy systems and infrared stealth technology.