Conference Presentation (Invited)

FZJ-2015-07861

http://join2-wiki.gsi.de/foswiki/pub/Main/Artwork/join2_logo100x88.png Parallel Molecular Dynamics for Materials Science

Sutmann, G. (Corresponding author)FZJ*

2015

Abstract: Materials science simulations often require large scale particle simulations for taking into account shape transformations, development of defect structures or cracks, which strongly depend on the stress distribution within the material. The simulation has to take into account the time dependence of suhc transformations which often result in a non-homogeneous development of computational work within the material. A static domain decomposition on a parallel computer is therefore at risk to degrade efficiency of the runtime behavior of the simulation. In the talk, a load-balancing method, based on non-cartesian domains, is shown which keeps track on time dependent changes in the load and provides a higher efficiency for systems with complex geometry or time dependent shape. Examples will be shown for nanopillar compression and nano-particle surface coatings.

---

Contributing Institute(s):

1. Institute for Advanced Simulation (IAS)
2. Jülich Supercomputing Center (JSC)

Research Program(s):

1. 511 - Computational Science and Mathematical Methods (POF3-511) (POF3-511)
2. Simulation and Data Laboratory Complex Particle Systems (SDLCPS)

Appears in the scientific report 2015

---

Database coverage:
No Authors Fulltext

---