[gmx-users] virtual sites set up for topology file

Mark Abraham Mark.Abraham at anu.edu.au
Fri May 6 14:22:07 CEST 2011

On 6/05/2011 10:17 PM, Gavin Melaugh wrote:
> Yeah I see your point about the types. With regard to the initial
> configuration state I would have assumed that gromacs knew the initial
> position of the virtual site  when I stated that it was to be at the
> COG  of the  3 atoms in  the [virtual_sitesn] directive.

Yes it does, but (IIRC) the implementation requires that each of the 
constructing atoms and the virtual atom(s) are distinct atoms. mdrun 
does the book-keeping to know which atoms are sensible at different 
stages of the integration. So the input coordinate file needs to have 
such atoms there, but I rather suspect their coordinates will get 
ignored - that's easily tested, of course.


> Justin A. Lemkul wrote:
>> Gavin Melaugh wrote:
>>> Justin
>>> I have tried this but I am now getting different errors. I take it that:
>>> I specify the virtual sites in the atomtypes directive as I have seen
>>> from examples?
>> Virtual sites are included in all the force fields already, but if you
>> want some custom name, then yes, include them in a new [atomtypes]
>> directive.  I see no reason to create three distinct, but identical,
>> types as you have.
>>> I index the virtual sites in the atoms directive in accordance with the
>>> rest of the molecule. atom numbers go from 1-228, therefore I label the
>>> 3 virtual sites 229 to231.
>>> The error I get now is
>>> Atom index (229) in virtual_sites3 out of bounds (1-228).
>>> This probably means that you have inserted topology section
>>> "virtual_sites3"
>>> in a part belonging to a different molecule than you intended to.
>>> In that case move the "virtual_sites3" section to the right molecule.
>>> Do I have to have the virtual sites in the gro file also? This doesn't
>>> make sense
>> Yes, you need their coordinates as part of the initial state.
>> -Justin

More information about the gromacs.org_gmx-users mailing list