Long-term atomistic finite-temperature substitutional diffusion

TL;DR

This paper introduces a novel multiscale simulation method combining statistical mechanics and transition state theory to accurately predict long-term atomic diffusion in alloys within feasible computational times.

Contribution

A new approach that bridges atomic vibration time scales to vacancy-mediated hops, enabling long-term diffusion simulations with atomic precision.

Findings

Successfully simulated bulk copper self-diffusion.

Predicted impurity concentrations in aluminum matching theoretical models.

Achieved realistic long-term diffusion predictions within hours of computation.

Abstract

Simulating long-term mass diffusion kinetics with atomic precision is important to predict chemical and mechanical properties of alloys over time scales of engineering interest in applications, including (but not limited to) alloy heat treatment, corrosion resistance, and hydrogen embrittlement. We present a new strategy to bridge from the time scale of atomic vibrations to that of vacancy-mediated atomic hops by a combination of statistical mechanics-based Gaussian phase packets (GPP) relaxation and a nudged elastic band (NEB)-facilitated harmonic transition state theory (H-TST) time update. We validate the approach by simulating bulk self-diffusion in copper and the segregation of vacancies and magnesium to a stacking fault and a symmetric tilt grain boundary in aluminum, modeled with an embedded atom method (EAM) potential. The method correctly predicts the kinetics in bulk copper and equilibrium impurity concentrations in aluminum, in agreement with the Langmuir-Mclean solution in the dilute limit. Notably, this technique can reach realistic diffusion time scales of days, weeks, and even years in a computational time of hours, demonstrating its capability to study the long-term chemo-thermo-mechanically coupled behavior of atomic ensembles.