Fig. 5: Final potential (TS-f) prediction of material properties with size-converged SRO.
From: Capturing short-range order in high-entropy alloys with machine learning potentials

a Relative error of WC parameters with system size (εsize), measured similarly to Eq. (2) but against the largest size prediction. Simulations with <2000 atoms are not converged with respect to system size. b Temperature dependence of WC parameters (Eq. (1)). c Temperature dependence of stacking-fault energy (γsf) for a random solid solution and a solid solution in thermal equilibrium (i.e., with appropriate SRO as obtained through Monte Carlo simulations). The “DFT-size” data point was evaluated with TS-f, but using a system with only 180 atoms, which is the typical size of DFT calculations (as indicated in Fig. 5a). d Temperature dependence of the fcc-hcp phase stability. See text for a full discussion of the effect of SRO and the phase transition at low temperatures leading to the kink at 600 K for solid solutions in thermal equilibrium.