Neural network interatomic potential-driven analysis of phase stability in Ti–V alloys at the atomistic scale

Abstract

The evolution of the phase in titanium–vanadium (Ti–V) alloys is critical for their mechanical properties, particularly in aerospace and biomedical applications. This study employs a Rapid Artificial Neural Network (RANN) potential to model the phase evolution at the atomistic level, demonstrating a high degree of consistency with experimental observations, unlike the Modified Embedded Atom Method (MEAM), which fails to capture this phase transformation accurately. RANN simulations replicate key phenomena such as the nucleation of precipitates at interfaces and accurate lattice orientations, enhancing our understanding of phase stability and transformation kinetics. The findings affirm that RANN potentials can significantly improve the prediction accuracy of complex material behaviors, offering a powerful tool for designing advanced materials with tailored properties such as solute effect in various stacking fault energies. This approach not only bridges the gap between theoretical predictions and empirical data but also sets a new direction for future research in materials science, emphasizing the integration of machine learning techniques in the development and optimization of new alloys.