Machine Learning Potential for Modeling Fracture in 2D High-Entropy VTiCrMoC3 MXenes

Two-dimensional (2D) transition-metal carbides, nitrides, and carbonitrides, known as MXenes, have obtained significant attention due to their unique combination of mechanical, thermal, and electronic properties. While MXenes have traditionally been composed of single transition metals, recent advancements have enabled the synthesis of multi-principal element high-entropy MXenes, such as VTiCrMoC3Tx. Despite their promising potential, the mechanical characteristics of high-entropy MXenes remain largely unexplored. In this work, we present a versatile machine learning interatomic potential (MLIP) potential capable of modeling both regular and high-entropy MXenes, including the complex mechanical behavior and fracture properties of M4X3 (M = V, Ti, Cr, Mo; X = C). Our ML potential accurately reproduces a wide range of material properties, including lattice and elastic constants, free surface energies, vacancy formation and migration energies, generalized stacking fault energies (GSFE), and short-range order (SRO) effects, as validated through first-principles calculations. These fundamental properties serve as a foundation for modeling stress-strain curves and mechanical response under various loading conditions. Additionally, the potential provides accurate descriptions of defect-related phenomena, including dislocation motion, plasticity, and diffusion, which are critical to understanding material deformation in operation conditions. It also enables detailed modeling of crack propagation, revealing how crack initiation and growth are influenced by SRO effects. By extending the scale and time limitations of first-principles methods, this ML potential serves as a powerful tool for exploring the interplay between SRO and mechanical strengthening, offering valuable insights for designing both conventional and high-entropy MXenes in advanced material applications.