Accelerated Modeling of Interfacial Phases in the Ni-Bi System with Machine Learning Interatomic Potential
24 Pages Posted: 27 Jan 2020 Publication Status: Accepted
High-performance modeling of interfacial phases is a challenge because of the low scalability of first-principle methods. Here we present a data-driven approach based on using the machine learning potential to address this problem. The developed model quantitatively reproduces the formation energy of Bi films on selected Ni grain boundaries. This scheme allows us to model arbitrary grain boundaries, preserving chemical accuracy of the reference method. The suitability of the interatomic potential is also confirmed by the construction of a grain boundary phase diagram. This approach opens the door for the accelerated study of the full configurational space of interfacial phases.
Keywords: grain boundaries, grain boundary segregation, phase diagram, density functional theory (DFT), machine learning
Suggested Citation: Suggested Citation