Interstitial diffusion is a key process that influences phase stability and irradiation response in concentration solid solution alloys (CSAs) under nonequilibrium conditions. In this work, we study atomic transport interstitial-mediated diffusion in Fe-Ni CSAs by combining machine learning (ML) and kinetic Monte Carlo (KMC). Specifically, the ML model is trained on a dataset of migration energy barriers generated via nudged elastic band calculations based on a well-validated embedded-atom method potential. This trained ML model is then used to accurately and efficiently predict interstitial migration barriers on the fly during simulations. Using this tool, we identified that the interstitial-mediated sluggish diffusion occurs only when the reduction in the tracer correlation factor $f_{\mathrm{tr}}$ outweighs the increase in jump frequency ν. Unlike molecular dynamics, the ML-KMC tool provides energy-barrier information for both actual and potential migration paths during long-term diffusion, offering new insights into the underlying mechanisms of Fe-Ni CSAs. Specifically, energy barrier differences between correlated migration patterns collaboratively form a ‘‘route selector’’ that favors the migration of slower-diffusing components during dumbbell diffusion. This preference strengthens the correlation effect (decreasing $f_{\mathrm{tr}}$) and suppresses the increase in ν as the fast-diffusing component increases, resulting in interstitial-mediated sluggish diffusion. The current findings can be generalized to explain interstitial-mediated sluggish diffusion in other CSA systems.

Close

PrevNext

View figure in article

• Point defects
• Self-diffusion
• Alloys
• Diffusion quantum Monte Carlo
• Machine learning