Machine learning–assisted design of a wideband Fe–SiO2–MXene metamaterial solar absorber for angle-insensitive thermal energy harvesting

The manuscript proposed an efficient broadband metamaterial-inspired multilayered solar absorber for efficient energy harvesting. Its structure is a periodic assembly of Fe, SiO₂, MXene, and Fe layers, thus facilitating the scalable fabrication. Finite element method (FEM) simulations are employed to evaluate and optimise the optical performance. The novel geometry and arrangement of the radiating elements result in peak absorptance exceeding 90% across 0.87–1.79 μm, 2.07–2.425 μm, and 3.045–3.295 μm. Parametric optimisation is conducted by varying resonator geometries (filled and solid), material combinations for resonator and ground layers (Au, Fe), and dielectric thicknesses to achieve optimal absorption characteristics. Angular stability is examined for both TE and TM polarisations over an incidence range of 0°–80°, demonstrating robust wide-angle performance. The TE and TM mode field are analysed at peak absorption wavelength. Furthermore, different machine learning models are implemented to analyse and predict absorption characteristics, validating the agreement between simulated and predicted results. The proposed design was compared and analysed with other similar works to identify performance improvements. The proposed design exhibits strong potential for broadband photothermal and solar-thermal energy harvesting applications.