Machine learning in prediction of MXenes-based metasurface absorber for maximizing solar spectral absorption

Abstract

The broadband absorption of the solar spectrum, which is central to solar cell technology. Despite more than a decade of development in this field, current structures obtained using conventional materials and artificial design methods still struggle to meet realistic needs. MXenes has been shown to be one of the most promising solar-absorbing materials due to its good spectral selectivity. The optical properties of MXenes depend mainly on its terminal groups and in this paper the absorption properties of different MXenes-based metasurface absorbers (MMA) in the solar spectrum are systematically investigated. Random forest regression models are constructed separately for MMAs of different terminal groups, and machine learning (ML) is used to complete the whole process from parameter selection to structure optimization (the error is only 0.02%). The simulation results show that the MMAO (Ti3C2O2) absorbs 93.19% in the solar spectrum and emits only 1.21% in the MIR band (5000 – 13000 nm), which is a near-perfect solar absorber. The combination of MXenes and metasurfaces is feasible and effective, and the combination of ML can achieve high accuracy prediction of MMA structural parameters, which provides a new idea for the material selection and design methods of solar–thermal energy collecting and manipulation.