Abstract
Body-centered cubic Fe-V alloys and their derived steels are extensively utilized in industry due to their outstanding mechanical properties, which include high strength, hardness, wear resistance, and improved impact toughness at elevated temperatures. Recent research has increasingly focused on the investigation of short-range order (SRO) in these alloys, particularly after the observation that the Cowley SRO parameter in a related Fe-Cr system changes with concentration. At low chromium concentrations, SRO is negative, indicating a propensity for short-range atomic ordering. Conversely, it becomes positive above 10 atomic percent, suggesting a tendency for segregation. Analyzing SRO in disordered alloys using molecular dynamics simulations requires the careful selection of highly accurate interatomic potentials to accurately represent the system’s behavior. This study compares interatomic potentials derived from two prominent machine learning techniques: the Moment Tensor Potential (MTP) and the Deep Potential method (DeePMD). These potentials are applied to the Fe-V system across a vanadium concentration range of 0 – 25 atomic percent. Their performance is evaluated through error and efficiency analyses, concentrating on their ability to predict material properties such as the Cowley parameter, equilibrium volume, and lattice constants. This assessment aims to determine the effectiveness of these potentials in capturing the characteristics of Fe-V alloys, thereby providing a foundation for further investigations into their behavior.
Funding
1. the Russian Science Foundation - project No. 25-22-20062.