Quantifying uncertainty in molecular dynamics simulations of grain boundary migration

Abstract

Molecular dynamics simulations of simple bicrystal systems have been much used as a tool to explore how the migration of grain boundaries varies with their structure and with experimental conditions. In order to permit the exploration of a large parameter space, many studies are forced to rely on a small number of simulations (often a single simulation) for each configuration. The motion of a grain boundary is inherently statistical and any variability in the measured grain boundary velocity should be taken into account in subsequent analysis of trends in grain boundary mobility. Here we present the results of large numbers of simulations of equivalent boundaries, which show that this variability can be large, particularly when small systems are simulated. We show how a bootstrap resampling approach can be used to characterise the statistical uncertainty in boundary velocity using the information present in a single simulation. We show that the approach is robust across a variety of system sizes, temperatures and driving force strengths and types, and provides a good order-of-magnitude measure of the population standard deviation across multiple equivalent simulations.

Keywords::

• molecular dynamics
• grain boundary mobility
• bootstrap resampling

Acknowledgements

C.P.R. was funded by a Postdoctoral Fellowship of the Alexander von Humboldt Foundation.

Notes

^1. Current address: Dalton Nuclear Institute, University of Manchester, Oxford Road, Manchester M13 9PL, UK.