Massively Parallel Simulation of Vacancy-Jog Interaction with ParSplice
Dallas Trinkle, University of Illinois at Urbana-Champaign
Usage Details
Daniel Katz, Danny Perez, Dallas Trinkle, Yadu Babuji, Joshua Vita, Lauren Smith, Zhuozhao Li, Anna WoodardWe plan to use the massively parallel program ParSplice—running on top of the molecular dynamics code LAMMPS—to simulate long time trajectories for a vacancy diffusing near a dislocation jog in nickel. Unlike most multiscaling approaches, ParSplice is designed specifically to simulate large times rather than large spatial domains; moreover, it shows excellent parallel scaling to large numbers of cores. The simulation we propose will quantify the rate of reaction for a vacancy and a dislocation jog in nickel: a fundamental atomic-scale process for the climb of a dislocation, which in turn controls creep (high temperature time-dependent strain due to a time-independent load). Creep is the primary design criterion in materials for jet turbine blades, and nickel is the primary constituent in nickel-based superalloys for such applications. Despite the crucial nature of this work, the long simulation times needed have been prohibitive to computational study. Now, with ParSplice and access to a very large number of cores on Blue Waters, we will be able to achieve the appropriate time scales to study how vacancies and dislocation jogs interact.