[gmx-users] Neighbor searching artifacts in permittivity when using shift

Szilárd Páll pall.szilard at gmail.com
Thu Dec 1 16:45:46 CET 2016


Unless you are actually not using the Verlet scheme (or you have set
the tolerance to -1), rlist has no effect, it is not used in any way
as it is calculated automatically for the given system and settings.

Hence, your observation that
>  The permittivity was even stronger influenced by variations of rlist.
is suspicious.

Secondly, do look at the parameters recommended by Peter K. above, and
if you want to tweak the verlet buffer, I'd rather recommend using the
tolerance for that.


On Wed, Nov 23, 2016 at 1:19 PM, Julian Michalowsky
<j.michalowsky at gmail.com> wrote:
> Hi,
> I am using gromacs 5.1.0 with the Martini force field. When trying to
> reproduce some data regarding the 2010 Martini polarizable water model, I
> noticed that it is affected by the values of rlist and nstlist. In the
> reference work, the following parameters are used:
> cutoff-scheme=group
> nstlist=5
> rlist=1.2
> vdwtype=shift
> rvdw=1.2
> rvdw-switch=0.9
> coulombtype=shift
> rcoulomb=1.2
> rcoulomb-switch=0.0
> Along with this, Berendsen weak coupling schemes were used. I myself on the
> other hand use the v-rescale thermostat and parrinello-rahman barostat, but
> I can reproduce the reference data if I use the respective values for rlist
> and nstlist.
> What I did: In one set of simulations, I varied rlist=1.20, 1.22,...,1.40
> and kept nstlist=1. In another set, I set rlist=1.20 and varied
> nstlist=1,2,...,10. For each simulation, I calculated the permittivity
> using gmx dipoles. Simulations are each 100ns NPT production runs and
> sufficiently equilibrated; boxes contain 2797 polarizable water beads (and
> nothing else).
> What I expected: No variation in the first set of simulations, as the
> smallest value for rlist is still >= the largest interaction cutoff, and I
> update the neighbor lists every time step (nstlist=1). I expect some
> variations when changing the value of nstlist, though, and keeping
> rlist=rvdw=rcoulomb.
> What I found: Variations of the measured permittivity in both simulation
> sets (monotonous trends). The permittivity was even stronger influenced by
> variations of rlist.
> What I am wondering about is why rlist influences the data, despite rlist
>>= rvdw and nstlist=1, and if this should be the case. If so, there is some
> conceptual thing I don't quite understand about how the group neighbor list
> implementation in Gromacs works, and I kindly ask for an explanation. If
> not, then I guess there is something going on that shouldn't be, or I made
> a different mistake. In any case, I require some assistance on this one.
> On a sidenote: I know that the 'shift' statement is deprecated, but it
> worked for me using different Gromacs versions, including 5.0.5 and 5.1.0.
> Trying to use the potential modifier force-switch instead results in
> immediate system crashes after minimization, so they are probably not 100%
> interchangable (which is what I thought they were supposed to be).
> Thanks a lot in advance for your help.
> Kind Regards,
> Julian Michalowsky
> --
> Gromacs Users mailing list
> * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before posting!
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or send a mail to gmx-users-request at gromacs.org.

More information about the gromacs.org_gmx-users mailing list