Periodic MD

[Axel]

On Sep 30, 5:51 am, Fawzi Mohamed <fa... at gmx.ch> wrote:
> On Sep 27, 2008, at 3:21 PM, ilya wrote:
>
>
>
> > Dear users and developers,
>
> > as the doc says CP2K_INPUT / FORCE_EVAL / SUBSYS / CELL / PERIODIC
> > keyword "only applies to the generation of the pair lists" and as i
> > understand it doesn't deal with motion. So when I run my MD sim for
> > several hundreds of ps at high temperature the water molecules just
> > fly away from the center even when everything PERIODIC is set to XYZ.
> > I didn't find a method to constrain a system to the box. For example
> > some apps can do something like this: when one molecule crosses the
> > border it appears on the other side of the box. So may I constrain the
> > system so it will not evaporate?
>
> your system does not evaporate if it is periodic, the periodic copies
> interact as they should.
>
> It is true that the coordinates are not wrapped in the initial cell,
> but this is done on purpose:
>
> 1) when everything is periodic putting a particle in the central cell
> or in one of its periodic copies in the other cells is exactly the
> same thing
>
> 2) wrapping the coordinates is easy (just do a modulo 1 in reduced
> units)

to add to this:
if you write trajectories in .dcd format, it is as easy as
calling "pbc wrap" in VMD to get the molecules wrapped back
into the principal unit cell. there are several ways to do this
(by atom, by residue, by chain) and thus it is a good idea to
store coordinates "unwrapped" so that one can wrap them back
in the preferred way. unwrapping a previously wrapped trajectory
is far more error prone and complicated. for details see:
http://www.ks.uiuc.edu/Research/vmd/plugins/pbctools/

the .dcd format is needed as it stores the unit cell information,
otherwise you have to use "pbc set" to set the box manually.
heers,
   axel.

>
> 3) non wrapping make the trajectory of a particle continuos in the
> normal space (no need of berry phase or something like it to remove
> the apparent jumps), this is very useful for example if you want to
> calculate diffusion.
>
> Fawzi
>
> > Is CP2K_INPUT / MOTION / FLEXIBLE_PARTITIONING a way to constrain the
> > system (as I understand we can construct a spherical constrain around
> > a CENTRAL_ATOM using it??) ??
>
> > Thanks a lot,
> > Ilya.