Machine learning-driven interface engineering for enhanced microwave absorption in MXene films

Abstract

The micro-nanostructure architecture for microwave absorption materials is widely considered an effective approach to enhance the properties of materials, providing unlimited design space. However, the structure-function black box limits the design and preparation of microwave-absorbing materials through traditional trial and error methods, characterized by a time-consuming cycle between microscopic material modification and macroscopic performance measurement. Here, we present a novel machine learning-based approach to predict electromagnetic parameters of materials with excellent microwave absorption properties. Through introducing air, a “five-layer” films structure, composed of lamellar MXene, hollow spherical MXene, lamellar MXene, hollow spherical MXene, and lamellar MXene, is designed, which exhibit greatly enhanced the microwave absorption performance compared with pure layered MXene films. Our results demonstrate that the precise tuning of the electromagnetic parameters and the moderate improvement of the impedance matching achieves within this composite, greatly enhancing the dielectric loss capability of the films. Owing to the microstructural characteristics, the films shows the minimum reflection loss (RL_min) of −48.15 dB and the maximum effective absorption bandwidth (EAB_max) of 5.84 GHz. In addition, when the angle between the incident wave and the plane normal is −60° < θ < +60°, the radar cross section (RCS) can be reduced by 25.73 dB m² when the MXene films with a 2.5 mm layer is covered with the PEC substrate, successfully demonstrating its practical application capability. This machine learning-guided material synthesis approach significantly shortens the experimental time and offers a highway to accelerate the development and industrialization of high-performance microwave absorption materials.