Ensemble learning enables accurate and interpretable identification of hydrogen storage in 2D materials beyond MXenes

Hydrogen storage is a critical enabler for the reliable hydrogen energy utilization, with 2D materials-particularly MXenes-emerging as promising candidates due to their intrinsic solid-state hydrogen storage capabilities. However, the rapid, accurate, and rational prediction of hydrogen storage properties in 2D materials remains a significant challenge, hindering the advancement of solid-state hydrogen storage technologies. Here, we present an advanced ensemble learning framework that leverages high-quality datasets derived from first-principles calculations and experimental results to enable precise and efficient predictions of hydrogen storage properties in MXenes. By establishing 21 physically meaningful descriptors and selecting 4 optimal machine learning models from diverse algorithms, we develop a robust ensemble model. The reliability and accuracy of this approach are rigorously validated using Ti2CT2 MXenes as a case study, with results demonstrating excellent agreement with theoretical and experimental data. Feature importance analysis reveals that the ionic radius of X element is the most critical descriptor, highlighting the interpretability of the ensemble learning framework. Furthermore, we extend the application of this method to predict the hydrogen storage properties of borophene, a promising one beyond MXenes. This work not only introduces a powerful computational tool for accelerating the discovery of next-generation hydrogen storage materials but also provides valuable insights into the underlying physicochemical mechanisms governing hydrogen storage in 2D systems.