[gmx-users] Re: [gmx-developers] Re: long range interactions dependon bonds

[Berk Hess]

>well, I do use constraints (all bonds) but this does not seem to be the
>actual problem. I've got the output of two simulations, OLD and NEW.
>md.mdp, conf.gro and index.ndx are identical in OLD and NEW,
>ONLY difference: topol.top in OLD contains bond definitions between
>atoms in the frozen group A while topol.top in NEW does not (I double and
>triple checked that with diff)
>
>If I compare the content of the two md.log files I see that the
>only difference before the energy output starts is where the number
>of constraints is reported, then the energies at step zero:
>
>   OLD:
>           Angle Ryckaert-Bell.          LJ-14     Coulomb-14        LJ 
>(SR)
>     8.27019e+01    1.81476e+01    2.02982e+01   -1.49987e+01    
>1.19705e+04
>    Coulomb (SR)   Coulomb (LR)      Potential    Kinetic En.   Total 
>Energy
>    -7.81710e+04   -6.11122e+03   -7.22056e+04    1.26054e+04   
>-5.96002e+04
>   NEW:
>           Angle Ryckaert-Bell.          LJ-14     Coulomb-14        LJ 
>(SR)
>     8.27019e+01    1.81476e+01    2.02982e+01   -1.49987e+01    
>1.19705e+04
>    Coulomb (SR)   Coulomb (LR)      Potential    Kinetic En.   Total 
>Energy
>    -7.81710e+04   -5.40009e+03   -7.14945e+04    1.11621e+04   
>-6.03324e+04
>
>Mind that all energies are identical in A and B, ONLY the Coulomb (LR)
>values differ considerably. Later the other energy terms also start to
>differ somewhat (not surprisingly).
>This must mean that the declaration of bonds does something
>to the Coul-LR energy, why ?
>
>groetjes,
>Michael
>
>
>additional info:
>All the frozen atoms in my sims have a mass of 1.0e+12, resulting in
>identical positions of the frozen atoms in conf.gro and confout.gro.
>also are all bon-lengths in the frozen part at their equilibrium
>values, therefore Lincs has no probelm with the frozen atoms when
>bonds are defined.

There could still be small differences beyond the gro precision,
but these will never explain such a big difference in LR energy.

I saw that you use PME. Since the mesh part calculates the interaction
between all particles, the exclusions need to be subtracted. These
terms contribute to the LR energy term.
So it could be that you have extra exclusions due to the bonds.
I just saw you have an energy group exclusion for the frozen group.
This means you do not see the exclusion difference in the short range term,
because the contribution of the frozen group is not calculated at all.

This whole confusion is due to the energy group exclusions. The direct
particle-particle interacions can be exluded, but the ewald long-range
parts can not be seperated.
grompp and/or mdrun should give a warning for this when you use
energy group exclusions together with PME or Ewald.

Berk.

_________________________________________________________________
Play online games with your friends with MSN Messenger 
http://messenger.msn.nl/