Accurate Fe–Ni–Al deep learning potential for investigating NiAl precipitation and its strengthening effect in bcc-Fe

Fe Ni-based alloys represent a critical system for developing advanced high-strength toughness specialty alloys. The addition of the cost-effective element Al facilitates the precipitation of nanoscale B2-NiAl, enhancing strength while promoting material lightweighting. To elucidate the mechanism of Ni and Al migration and clustering at the atomic scale, we developed a deep neural network potential (DNNP) for the Fe Ni Al system. This potential demonstrates exceptional accuracy, with root-mean-square errors of 8.61 meV/atom for energy and 65.43 meV/Å for force, outperforming existing empirical potentials in predicting lattice parameters and elastic properties. Hybrid MD/MC simulations at 773 K reveal the continuous formation and growth of Ni Al clusters, with radial distribution function and Warren Cowley short-range order analyses confirming progressive B2 ordering. Uniaxial tensile simulations reveal that Ni Al clusters result in heterogeneous distributions of stress and strain, causing premature nonlinear elastic deviation of the samples. Plastic deformation initiates via twin boundary sliding, followed by the activation of 1/2 < 111 > and < 100 > dislocation slip systems.