Dear Mark and Oscar, 

Thanks for your help!
I checked the velocity distribution and the velocity at different simulation frames. As you said the initial velocities are all set to zero and at the next step are already changed, I also guess by the random thermal noise.

Thus, my idea of simulating the same system with different random generator number for thermal noise (ld_seed) might account for a small variation of the "initial" conditions? 

loris...


In message <8c3f24a40903110611j4f69c75emb0f5eef0484de783@mail.gmail.com> Discussion list for GROMACS users <gmx-users@gromacs.org> writes:
> >
> > Thus I am a bit confused about it!
> >>
> >
> > I'm confused about the manual comment about only being meaningful with the
> > MD integrator, but we 'll have to wait until someone who knows something
> > about LD can post!
> 
> 
> For LD, the velocities will (should) get rescaled by the random thermal
> noise, that is- by the Langivin thermostat.
> With MD, the thermostat works like friction (positive or negative friction).
> So you see, that for LD, an initial velocity of zero should next be changed
> by the thermostat, while for MD the rescaling shouldn't work for zero.
> 
> --Omer Markovitch.
> 
> Koby Levy research group,
> Weizmann Institute of Science.
> http://www.weizmann.ac.il/sb/faculty_pages/Levy/