Abstract
High-entropy alloys (HEAs) are a new class of metallic materials that demonstrate potentially very useful functional and structural properties. Sluggish diffusion, one of the core effects responsible for their exotic properties, has been intensively debated. Here, we demonstrate that a combination of machine learning (ML) and kinetic Monte Carlo (kMC) can uncover the complicated links between the rough potential energy landscape (PEL) and atomic transport in HEAs. The ML model accurately represents the local environment dependence of PEL, and the developed ML-kMC allows us to reach the timescale required to reveal how composition-dependent PEL governs self-diffusion in HEAs. We further delineate a species-resolved analytical diffusion model that can capture essential features of self-diffusion in arbitrary alloy composition and temperature in HEAs. This work elucidates the governing mechanism for sluggish diffusion in HEAs, which enables efficient and accurate manipulation of diffusion properties in HEAs by tailoring alloy composition and corresponding PEL.
