ISSN:
1573-4900
Keywords:
High pressure
;
Linear scaling
;
Methane
;
Molecular dynamics
;
Polyethylene
;
Tight-binding
Source:
Springer Online Journal Archives 1860-2000
Topics:
Mechanical Engineering, Materials Science, Production Engineering, Mining and Metallurgy, Traffic Engineering, Precision Mechanics
Notes:
Abstract A molecular dynamics program has been written that is based on a quantum mechanical (tight-binding) treatment of the valence electrons. A new algorithmic approach to the solution of the tight-binding equations has been employed that (i) naturally leads to a very efficient parallel implementation; and (ii) is O(N), where the computational effort scales linearly with respect to the number of atoms N. Both very high single node performance as well as significant parallel speedup are obtained on the Silicon Graphics Origin 2000, IBM RS/6000 SP, and Intel TFLOPS parallel computers. Polymer simulations of size up to C3072H6250 (18 538 valence electrons) were included in the benchmark calculations. A parallel speedup of 400, relative to a single processor, was obtained using 768 processors on the TFLOPS computer. Sustained molecular dynamics simulations of the dissociation of a dense methane fluid and of stress–strain in a large hydrocarbon polymer are presented. The dissociation of methane into elemental carbon and molecular hydrogen is studied for fixed volume and eight different temperatures using a 128-molecule (1024 valence electrons) simulation cell and trajectories of length up to 6.6 ps (13 200 time steps). The nature and structure of the final dissociation products are probed with pair correlation function, cluster, and nearest-neighbor analyses. These results are compared with shock-compression experiments, chemical equilibria calculations, and an ab initio molecular dynamics simulation. In the second application, a calculation of the stress–strain curve for an amorphous simulation cell of polyethylene (single-chain C1000H2002, 6002 valence electrons) is described, where a trajectory of length 12 ps (12 000 timesteps) was generated.
Type of Medium:
Electronic Resource
URL:
http://dx.doi.org/10.1023/A:1008637113920
Permalink