Neural Network Based Modeling of Grain Boundary Complexions Localized in Simple Symmetric Tilt Boundaries Σ3 (111) and Σ5 (210)

A method is proposed for the neural network based analysis of the existence and stability of grain boundary complexions formed at high-symmetry tilt boundaries Σ3 (111) and Σ5 (210) in a polycrystalline Ni(Bi) solid solution. This method is based on the use of reference interparticle interaction potentials constructed within the framework of the density functional theory in combination with the structural capabilities of an artificial two-level self-learning neural network. The absolute error in determining potential energy by the neurosystem analysis is 0.012 eV/atom. The values of the formation enthalpy of grain boundary complexions for Σ3 and Σ5 boundaries are in rather good agreement with the published results of simulating this system and experimental data.