[gmx-developers] long range interactions depend on bonds ?

[Michael Brunsteiner]

Dear all,

I saw some quite unexpected results from Gromacs
runs today, and would appreciate if somebody could
tell whats going on here ...

I do MD with a system containing three energygroups, A, B and C.
One of the groups, A, contains only frozen atoms.
To get the long range contributions to the interaction
energies A-B, B-C, etc I use tpbconv, setting up six reruns 
with the charges set to zero for each group and each
combination of two groups (as previously suggested in this list)
this seems to work fine.

For some reason I repeated a couple of these simulations
and due to some mistake I ended up with a number of bonds
bewteen atoms in group A being defined in the topology file 
(In the first set of runs these bonds were not there
which shouldn't make a difference since all these atoms are 
frozen anyway)
what I see is the following:
the long range contribution ELR to the interaction energies
depends on whether these bonds are there or not, and most
notably, not only the ELR for A-A interactions but also those
for A-B and A-C depend on the bonding between atoms in A.

I this supposed to be like that ? 

Electrostatic short range interactions between two bonded atoms are 
excluded, but why would the bond affect also the long range interactions ?
And even stranger why should it affect long range interaction energies
between two entirely different groups (i.e. molecules) ???

thanks in advance for any ideas!

cheers,
Michael





	
		
__________________________________
Do you Yahoo!?
New and Improved Yahoo! Mail - 100MB free storage!
http://promotions.yahoo.com/new_mail