[gmx-users] energy group exclusions - are forces calculated?

Michael Brunsteiner mbx0009 at yahoo.com
Sat Nov 29 04:40:54 CET 2008


Short question: if I define energy group exclusions
between, say, group A and B - will members of group
A and B be completely invisible to each other, or
is simply their interaction energy not calculated or
recorded in the energy output?

long version: I'd like to set up a simulation in which for
two (NOT constrained) groups of atoms not only their interaction
energies are not calculated but also these two groups would be
entirely invisible to each other. I assumed energy group exclusions
would help me there but I did a test with a water box (using gmx 4.0.0)
where i defined two energy groups w1 and w2 each comprising one
half of all molecules. when i then say:

energygrps               = w1 w2
energygrp_excl           = w1 w2

and run a short (40 ps) simulation (300 K, ~1 atm, NO PME) i find
in the resulting trajectory that no pair of waters from group w1 and w2,
respectively ever come closer to each other than they would do in
a normal simulation of the same system, but the interaction energy
in ener.edr between w1 and w2 is zero - which seems to imply that
waters in w1 and w2 actually do "feel" each other, or that only their
interaction energies are not calculated, but the forces are.

If, on the other hand side, i simply concatenate and renumber the
waters in two equally sized water boxes that were independently
equilibrated, defining each one of the original systems as
one excluded energy group, i.e., when I start with a lot
of mutually excluded waters overlapping, then the simulation 
immediately blows up, although the overlapping waters are actually not
supposed to see each other at all.

Here energygrp_excl does not seem to work in the way i'd like it
to do - so is there any other work-around to get what i want, i.e., two
systems in the same box, at the same time, each of them behaving
is if the other one wasn't there at all?

thanks for any suggestions!


More information about the gromacs.org_gmx-users mailing list