Machine Learning‐Directed Fast and High‐Throughput Acquisition of High‐Efficiency Microwave Absorbents From Infinite Design Space

Abstract Intrinsic composition properties and extrinsic micro‐/nano‐structural effects constitute the infinite design space of microwave absorption (MA) materials wherein the high‐efficiency performance is expected to advance stealth and anti‐interference technologies. However, restricted to the black box of physical mechanisms, discovering those materials too often relies on the traditional trial‐and‐error methods, falling into the time‐consuming loop between material modification and performance measurement. Herein, an unprecedented machine learning‐based forecasting system (MLFS) is constructed to directly predict the process conditions of carbonyl iron/ferrosoferric oxide hybrids with enhanced MA performance. The high‐throughput screening and inverse projection based on pattern recognition recommend a series of excellent MA materials with the highest performance correlation coefficient up to 0.9844. After manual selection from this set, the enhancement of maximum absorption efficiency and bandwidth of the optimal hybrid reach 207% and 360% in comparison with the original database. The standardized MLFS procedure immensely shortens the research cycle to a few weeks compared to several months of the manual orthogonal experiment. This is believed to be an expressway for accelerating the discovery of high‐performance MA materials and their industrialization.