Speedup Techniques for Molecular Dynamics Simulations of the Interaction of Acoustic Waves and Nanomaterials
HA Bennett and AC Zander and BS Cazzolato and DM Huang, 21ST INTERNATIONAL CONGRESS ON MODELLING AND SIMULATION (MODSIM2015), 655-661 (2015).
Nanomaterials are seen to have great potential for use in the area of sound absorption. However, direct inspection of the interactions between acoustic waves and nanomaterials is not feasible due to the short time and length-scales involved. Molecular dynamics simulations can assist in improving understanding of the mechanisms involved in this process, but they have limitations that must be overcome to make their use viable. The primary limitation is that molecular dynamics is computationally expensive, making the time-scales over which results can be obtained very short. This, in turn, makes the acoustic frequencies that can be examined extremely high. In the current work, the use of a simplified force field, multiple time-stepping, and an analytical description of the sound source producing the acoustic waves are investigated as methods to improve the speed of a model that simulates acoustic wave and nanomaterial interactions, as speedup directly translates into increased feasibility of longer time-scales (lower acoustic frequencies) and larger domains. The speedup and accuracy of these techniques are determined through benchmarking against existing computational results for the interaction of a carbon nanotube with a 2.57 GHz acoustic wave propagating through argon gas. Significant speedup is obtained using these techniques: replacing the oscillating atomistic wall in the benchmark case with the analytical oscillating wall produces a speedup factor of 1.3; using the simpler Dreiding force field for the carbon nanotube instead of the benchmark case's REBO potential results in a speedup factor of 3.6; and exchanging the Velocity Verlet time integrator in the benchmark case with an rRESPA multiple time-step integrator along with using the Dreiding force field leads to a speedup factor of approximately 39. Combining all of these techniques further increases the speedup, resulting in a speedup factor of approximately 50 compared with the benchmark. The error introduced into the numerical results is no greater than 6%, suggesting these speedup techniques are appropriate for molecular dynamics simulations of acoustic wave and nanomaterial interactions.
Return to Publications page