This research, (supported in part by NSF Grant #0800718 ($259,269, with PI Huang)), focuses on the behavior of interfaces on the nanoscale and the development of multiscale structure-property relationships which are motivated by atomistic simulations. The primary thrust of the work performed in the CI-TRAIN project is to use molecular dynamics (MD) to identify the mechanisms associated with dislocation nucleation in nanocrystalline (NC) metallic materials.
Of particular interest is the role of impurities or dopants on grain boundary behavior, an important aspect of material behavior which is currently unresolved. For example, recent work by Spearot et al. has employed atomistic simulations to study the influence of dopant atom concentration on the mechanical properties of NC materials. MD simulations indicate that Shockley partial dislocations are nucleated from the grain boundaries and propagate across the grain interiors (leaving a trailing stacking fault in their wake), in agreement with previous MD simulations in the literature. Interestingly, for certain interfaces, grain boundary sources for dislocation nucleation appear to correlate visually with the presence of Sb dopants at the grain boundaries.
The proposed CI-TRAIN parallel visualization capabilities with appropriate storage facilities for visualization purposes will enable atomistic simulation to study mechanisms associated with failure of nanocrystalline materials with larger grain sizes (closer to typical experimental nanocrystalline samples) and more realistic grain size and disorientation distributions. Ultimately, the aim is to provide an understanding of the inverse Hall-Petch response in nanocrystalline materials with dopants or impurities distributed along the grain boundaries.
Current computational facilities are capable of performing simulations on tens of millions of atoms within the MD framework; however, future advancement of this research is limited by (1) the ability to store data required for resolving images of the atomistic output and (2) the ability to render three-dimensional models of the atomistic output in a parallel visualization environment. Accordingly, atomistic visualization is restricted only to smaller models. Data sets generated for larger atomic models are in the range of 1TB in size, demanding the need for an integrated computation and visualization system.
Investigator: Spearot (UAF)