[CP2K:1076] COM motion

Teodoro Laino teodor... at gmail.com
Thu May 22 09:25:35 UTC 2008

Ciao Francesco,

On 22 May 2008, at 11:04, Francesco wrote:

> Hi,
> I am running cp2k on an isolated molecule (poisson_solver wavelet,
> etc). I started with the
> molecule centered in the box, but I've seen that it moves around the
> box. After 10000
> steps it even arrives close to the borders, and cp2k complains that
> the density hits the
> boundaries.  In the output I've seen that the COM has small
> velocities, so I think I have
> to freeze movements and rotations of the whole system.

The velocity of the COM is initially set to zero (unless you provide  
externally, in this case you must check that the velocity of COM is  
In principle, in an isolated system, there should not be external  
forces.. so the COM
should stay fixed in his position. Unfortunately numerical roundoff  
errors (mainly due
to the use of the grid) propagate and the one you see is the effect:  
the COM moves.

> In the cp2k input manual I've found the COMVEL_TOL keyword.
> If I leave the default value COMVEL_TOL 0.0, the COM seems to move. A
> typical step shows:
> Centre of mass motion (COM):
> x =      -0.0000002303
> y =       0.0000002945
> z =      -0.0000001305
> if I put COMVEL_TOL 1e-8, in the log file I have:
> Temperature scaled to requested temperature:     85.43 K  ->     85.43
> K
> MD| Old VCOM =    -0.0000002303    0.0000002945   -0.0000001305 a.u.
> MD| New VCOM =    0.0000000000    0.0000000000    0.0000000000 a.u
> 1) Is this the correct way to block the motion of COM ?
Yep.. that's correct.. value of zero does not apply anything (sounds  
strange but that's it..sigh!)..
value different than zero, it will zero the COM and the energy  
removed from the translation of
the center of mass is redistributed in the system.. so it's correct..

> 2) What about rotations ? Is ANGVEL_ZERO T ok ?
ANGVEL_ZERO influences only the angular velocity at the beginning..  
if set to F (default) the system
may have a starting angular velocity different from zero. if set to T  
it's removed.
No zeroing/rescaling is performed for the angular velocity during MD.


> Thanks
> Francesco
> >

More information about the CP2K-user mailing list