Extreme electromagnetic wave absorption properties of graphene through machine learning

This paper introduces a comprehensive design framework for graphene-based absorbing materials that integrates experimental fabrication, machine learning prediction and intelligent optimisation. This framework aims to explore the optimal absorption performance achievable by single-component graphene under the physical limit of a 1 mm thickness, overcoming the bottleneck faced by single-layer absorbing materials in balancing the multiple requirements of ‘high absorption, wide bandwidth and thin thickness’. Based on experimental data, this paper constructs an integrated prediction model combining the Maxwell-Garnett (MG) empirical equation and Chain-RF, and introduces SHAP and PDP interpretability analyses to elucidate the underlying physical logic from a data-driven perspective. During the performance optimisation phase, based on transmission line theory calculations, a systematic comparison was conducted between the Ant Colony Optimisation (ABC) and Genetic Algorithm (GA) in terms of their optimisation capabilities within a multi-dimensional non-linear parameter space. The results confirmed that the ABC algorithm possesses stronger global search capabilities, successfully identifying the optimal gradient structure parameters and achieving a breakthrough in performance characterised by minimal reflection loss (−63.76 dB) and an ultra-wide effective absorption bandwidth (12 GHz). The integrated intelligent design framework established in this study provides a novel approach to the customisation of high-performance electromagnetic wave bsorbing materials.