[gmx-users] Shift and Switch are obsolete!?

[Berk Hess]

>
>GMX manual recommends rlist to be 0.1 to 0.3 nm larger than rcoulomb.  But
>I got it from grompp:
>
>ERROR: rcoulomb must be >=3D rlist
>ERROR: rvdw must be >=3D rlist

This is a known bug, which has been introduced in Gromacs version 3.1.
It will be fixed in the next release.

>Please, one question more.  If I use pure cut-off,  I still do not follow
>why rlist must be larger or equal to rcoulomb.  How does GMX couple with
>an atom at the exactly distance of cut-off, and atoms beyond?  One cannot
>have discontinuity!  So, is a shift function implicit with cut-off?

One does have discontinuities, which is pretty bad. Using shifted potentials
or PME solves this. Also reaction-field is much better than a normal 
cut-off.
The rcoulomb and rlist with a normal cut-off scheme are not really the 
cut-off
for the potential, but the cut-off for the long range neighbor list,
or the short-range neighborlist when they are equal to rlist.
There should be no discontinuity in the potential itself, as neutral charge
groups, which are effectively dipoles, could turn into real charges.
By using a cut-off for the neighborlist(s) a system with neutral charge 
groups
one only has discontinuities on the dipole-dipole interaction level.

Berk.



_________________________________________________________________
Chat with friends online, try MSN Messenger: http://messenger.msn.com