Deep learning based inverse design of metasurface absorber for maximizing solar spectral absorption

•Metasurface solar spectral absorber obtained using deep learning inverse design.•AM 1.5 solar spectrum-weighted absorptivity reached 97.75%•Insensitive to light polarization and incident angle (70°) The contemporary global challenges of energy crisis and global warming demand efficient utilization of solar thermal energy resources. In this study, a novel approach is employed wherein micro-nano-structures are meticulously designed in both the top and bottom layers of a bilayer quadrangular prismatic metasurface absorber (BQPMA). This method results in a remarkable weighted absorption of solar radiation in the extensive 280–4000 nm band range, attaining an impressive 97.75%. The innovative design process incorporates inverse design coupled with deep learning techniques, enabling rapid and precise selection of structural parameters with a minimal error of merely 0.06%. The extensive broadband absorption observed within the solar spectral range can be primarily attributed to the light trapping and moth-eye effects facilitated by Si3N4 and Ti quadrangular prismatic structures. Notably, the proposed BQPMA exhibits robust ultra-broadband absorption across a wide range of incidence angles, achieving approximately 92% absorption even at 70°. The proposed BQPMA device is ideal for solar thermal applications (such as solar thermal systems, thermal photovoltaics and thermoelectronic devices, etc.) due to the high temperature resistance of materials such as titanium nitride and its excellent spectral properties.