A statistics-based study and machine-learning of stacking fault energies in HEAs.

Due to the chemical disorder in the multi-principal component alloy, the stacking fault energy (SFE) of high entropy alloy is greatly affected by the complex local elemental environment, thus invalidating the traditional SFE calculation approach. Herein, a novel strategy for localized stacking fault energy (LSFE) calculation is proposed, which can not only reasonably incorporate the local chemical fluctuation effect in HEAs but also allows to realize high-throughput SFE calculations. Based on statistical method, the quantitative probability distributions of SFEs in FCC and BCC HEAs are achieved, which are necessary for further study of dislocation motion and up-scale modeling of HEAs. Finally, the intrinsic correlation between the LSFE and the local composition inhomogeneity in HEAs is unprecedentedly established with the machine learning (ML) methods. By classifying the features, the main factors that affect the LSFE are revealed, which can provide significant guidance for the composition optimization of HEAs. [Display omitted] • A novel strategy is proposed for the localized stacking fault energy (LSFE) calculation in HEAs. • Based on statistical method, the quantitative probability distributions of SFEs in HEAs are achieved. • The intrinsic correlation between the LSFE and the local composition inhomogeneity is established using the ML method. • The bond-based features are classified to correlate the different levels of LSFEs, which is promising to guide the design of HEA. [ABSTRACT FROM AUTHOR]