[gmx-users] Bad energy conservation.

[Ram'on Garc'ia Fern'andez]

On Thu, Dec 05, 2002 at 11:45:02AM +0100, Anton Feenstra wrote:
> Not to my knowledge. 
> rlist is the distance for inclusion in the neighborlist, and the
> corresponding interactions, that are outside rvdw/rcoul will be 
> calculated each nstlist steps.

No.
The reason for needed rlist > rvdw is that otherwise there will be
errors because the neighbour list is not updated, that is, there are
atoms that should be in the neighbour list that aren't. This can be
compensated by making the neighbour list slightly larger than needed.
In this way, if during the steps when then neighbour list is not updated,
an atom suddenly enters into the cutoff, it is likely that this atom
was not too far from the cutoff the last time the list is computed, and
so making rlist larger that rvdw will ensure that this atom is included.

In fact, Gromacs manual advises rlist > rvdw when vdwtype = shift, but then
grompp does not allow it.

Ramon