Parallelization of Carbon Nanotube Based Composites
Computational simulation is expected to play a vital role in the nanotechnology revolution. Molecular dynamics (MD) has been used traditionally by physicists and chemists to predict physical and chemical properties of small atomic systems. However, at present MD is finding increasing use also in the...
| Other Authors: | |
|---|---|
| Format: | Others |
| Language: | English English |
| Published: |
Florida State University
|
| Subjects: | |
| Online Access: | http://purl.flvc.org/fsu/fd/FSU_migr_etd-2846 |
| Summary: | Computational simulation is expected to play a vital role in the nanotechnology revolution. Molecular dynamics (MD) has been used traditionally by physicists and chemists to predict physical and chemical properties of small atomic systems. However, at present MD is finding increasing use also in the area of medicine, pharmacy and other engineering applications. Despite their important role, MD computations are severely restricted by the spatial and temporal scales of simulations. Thus improving efficiency of computations will have a widespread impact. This thesis describes the methods used to achieve effective spatial parallelization of a MD code that is based on a multi-body bond order potential. The material system studied here is carbon nanotubes (CNTs) with and without functional attachments. We discuss the scientific and computational issues in the development and implementation of parallelization algorithms when the domain needs to be discretized with fine granularity. Specific issues in terms of neighbor-list computation, communication reduction, and cache awareness are delineated, with corresponding benefit in terms of speed up. Important practical problems relevant to CNT based composites are studied using the above methodology, and the results are used to illustrate the effectiveness of various strategies used in the method. Our implementation achieves efficient parallelization at a finer granularity compared to published works on CNTs with complex configurations. Though we have parallelized a specific simulation of CNT based composites, the algorithms developed in this study should be applicable to other systems involving multi body potential. === A Thesis submitted to the Department of Computer Science in partial fulfillment of the requirements for the degree of Master of Science. === Summer Semester, 2004. === July 12, 2004. === Carbon Nanotube, Parallelization, Molecular Dynamics, Nanotechnology, Communication, MPI === Includes bibliographical references. === Ashok Srinivasan, Professor Directing Thesis; Namas Chandra, Committee Member; Lois Wright Hawkes, Committee Member. |
|---|