Simulation of the surface segregation profile of composition spread alloy films (CSAFs) is important to investigate their impact on oxidative passivation in C r x F e y N i 1 − x − y ternary alloys. The use of density functional theory (DFT) to directly evaluate the potential energies of the slab configurations across the entire composition triangle is computationally too costly for practical simulations. In this work, we train a high-fidelity neural network (NN) based on DFT calculated data to serve as a surrogate model for predicting the potential energies of the fcc(110) slabs for a ternary C r x F e y N i 1 − x − y alloy, both in clean and oxygen-adsorbed conditions, to simulate realistic environments. The trained interatomic potential can accurately predict the C r x F e y N i 1 − x − y potential energies throughout the whole ternary space, with mean absolute errors of less than 4 meV/atom and it accurately reproduces surface energetics in both environments. In order to calculate equilibrium segregation profiles at temperatures of 573 and 773 K, we embedded the ML model into a series of Monte Carlo simulations spanning ternary composition space. Simulations revealed distinct surface segregation behaviors between clean and oxygen-covered surfaces. In CrFeNi CSAFs, segregation behavior depends critically on both composition and environmental exposure. Under oxygen-free conditions, Fe stays in bulk and Cr only reaches the surface once its bulk level exceeds about 20 at.%, with notable exceptions in Ni-rich and Fe-poor regimes. Introducing oxygen reverses these trends, where Fe becomes the primary surface segregating element when Cr is below ≈ 25 at.%, perfectly mirroring experimental data, and Ni is pushed out of surface sites. This study demonstrates a robust, efficient framework that accurately captures the interplay between clean and oxidizing environments when exploring segregation and oxidation phenomena in complex ternary CrFeNi CSAFs materials systems.
El-Asri et al. (Fri,) studied this question.