Mechanisms of Strength, Thermal Stability, and Doping Effects in Metal-Doped Bilayer Borophene for Energy Storage Applications Using a DFT-Trained Deep-Learning Potential

Abstract

Bilayer borophene exhibits exceptional in-plane properties; however, its failure mechanisms and thermal robustness at realistic scales remain unclear, which limits its potential for applications in energy storage devices. We develop a deep-learning potential (DLP) trained on density-functional-theory (DFT) data (implemented with the DeepMD framework) and use it to investigate the coupled mechanical-thermal response of metal-doped bilayer borophene for energy storage applications. Focusing on two topologies (p6m2 and p6mmm) with substitutional Al and Ru, we find that p6m2 exhibits higher in-plane stiffness and strength than p6mmm, but reduced ductility due to earlier bond localization. Al doping increases stiffness while promoting brittle failure, whereas Ru preserves ductility and enhances thermal stability. Quantitative agreement between DFT and the DLP for energies, forces, and virials enables large-cell molecular dynamics simulations that resolve bond rearrangement and fracture processes beyond tractable DFT sizes. Using a transparent Lindemann-criterion analysis with a defined detection window and threshold, we quantify a higher structural transition temperature (STT) for p6mmm than for p6m2; Al decreases the STT temperature, while Ru increases it. Coordination-number and bond-angle tracking rationalize these trends, revealing topology-and dopant-specific mechanisms at interlayer "pillar" motifs. The validated workflow-DFT-trained DLP, explicit validation metrics, and mechanistic molecular dynamics at scale-provides transferable guidelines for modelling and designing robust, doped 2D boron architectures for energy storage applications.