Fourier Filtering of MD Trajectories

We apply Fourier filtering to separate the different time scales in MD simulations. By filtering out certain frequency regimes from the Fourier spectrum of the MD trajectory or the corresponding time-dependent energy, atomic events that happen in the surviving frequency window can be isolated. This allows detection and analysis of infrequent events as well as finding atoms that form rigid groups (apart from thermal oscillations) and can be substituted by single “meso particles” for mesoscale simulations. The following examples filter out high-frequent oscillations, but preserve the events that “live” on a time scale of 2 ps or slower.

Some examples:

· Self-interstitial diffusion in silicon

Si self-interstitial diffusion, ab-initio MD simulation, 1300 K (click on image to play movie).	Si self-interstitial diffusion, 40 ps Fourier-filtered ab-initio MD simulation, 1300 K, 0.5 THz frequency cutoff (click on image to play movie). “Breathing” of the system triggering hops can be seen.

· Contact formation between a DNA fragment and a carbon nanotube

DNA fragment (thymine, guanine + sugar backbone) between two (5,0) carbon nanotubes, unfiltered (“top” view; click on image to play movie).	DNA fragment (Thymine, guanine and sugar backbone) between two (5,0) carbon nanotubes, filtered with 0.5 THz frequency cutoff (“front” view; click on image to play movie). Rigid groups and “floppy” joints can be identified.

· Impact of 2 nm Cu nanoparticle on Zr surface
(0.5 THz cutoff frequency)

Impact of a spherical Cu nanoparticle (2 nm diameter) on a Zr slab, filtered with 0.5 THz frequency cutoff. Classical MD with EAM potentials. Total simulated time: 33.6 ps. Click on image to play movie.

* W. Windl et al. (to be published).