[gmx-developers] Beginning developing Gromacs

Jason de Joannis jdejoan at emory.edu
Wed May 2 17:20:51 CEST 2007


>> What I am attempting to do is to modify mdrun in such a way as to
>> modify the structure of a loaded molecule every n iteration steps. The
>> point is to load a molecule which is very simple, but incomplete, and
>> then, every n steps, reconstruct the molecule and calculate the
>> electrostatic field around in the (reconstructed) molecule.
>
> The way to do this is to stop mdrun, modify your structure and 
> topology off-line and then restart the calculation. Otherwise you 
> have to deal with a whole lot of crud maintaining topologies, pair 
> lists, exclusion lists and probably more. It won't be as simple as 
> "adding new atoms".
>

I agree. It is challenging to change the number of atoms in a molecule 
or to make serious topological changes on the fly. I have created ghost 
molecules for this purpose in my work which contain an entirely 
different topology (this was a serious time investment). Another 
possibility that has been suggested to me is to make use of the A and B 
topologies that are already devised for the free energy perturbation 
algorithm. When doing this kind of thing, not only does one have to 
worry about the change in force field but also maintenance of pair 
lists, coordinate transformations (whole vs inbox) and changing degrees 
of freedom. Other things to thing of include the effect on constraints 
and the virial.

>>
>> I would expect some structure to hold all information about the
>> simulated molecule (atom bonds, atom positions, atom masses, atom
>> velocities) at every time step, but was unable to find one. Is there
>> such a structure, and in that case, which structure is it? If not,
>> what approach should I choose in order to modify the molecule?
>
> t_state is the data type you think you're looking for.
>
Yes, t_state will give you the current config (positions & velocities). 
For the masses you can use t_atoms or t_mdatoms. For the bonds and 
other interactions you will need t_idef.


-- 
\ Jason de Joannis  \
\ Emory University  \
  \ jdejoan at emory.edu \
   \ 404-402-1332      \





More information about the gromacs.org_gmx-developers mailing list