[gmx-developers] Unusual virtual interaction sites

Mark Abraham Mark.Abraham at anu.edu.au
Mon May 9 15:19:36 CEST 2011


On 8/05/2011 6:30 AM, Lee-Ping wrote:
> I did think about that.  My reasoning is a bit roundabout - my force
> field parameterization strategy is to match the MM energies/forces to ab
> initio calculations.  Thus, in the end there needs to be a force on the
> physical atom.

OK you want only one non-virtual atom, and there needs to be at least 
three non-virtual non-co-linear atoms to create the context for 
constructing such virtual sites. So we are at an impasse.

I think you'd need to use Chris's no-LJ atoms, and find out how to 
project the forces mdrun computes onto the central atom for your 
parametrization. The maths won't be too bad because of your symmetry. 
You should be able to solve a simple 2D case by hand. There will be 
torque, however, and I dunno how you're going to get that from ab initio 
calcs.

You can write the forces to a .trr file with nstfout and get access to 
them with g_traj, among other options. Also, learn from the lessons of 
biomolecular force fields, and try to parametrize on some "condensed 
phase" data as well as ab initio gas-phase data.

> When one has virtual sites, the forces are propagated from the virtual
> site onto the host atom.  But when there are constraints, I don't think
> this is the case.

Correct.

> - Lee-Ping
>
> On Sat, 2011-05-07 at 15:29 -0400, chris.neale at utoronto.ca wrote:
>> why do they need to be virtual sites? you could do this with regular
>> atoms with zero LJ terms and define the geometry by a set of
>> constraints.
>>
>> Quoting Lee-Ping<leeping at MIT.EDU>:
>>
>>> Dear all,
>>>
>>> I've been randomly thinking of a new type of model for elemental
>>> materials and alloys in which individual atoms can be described using
>>> several virtual sites.  For example, consider a metal ion with six
>>> virtual sites that allows it to coordinate water in an octahedral
>>> geometry.  It would allow for bonds to be broken and formed without
>>> explicit topology definitions or bond-order potentials.
>>>
>>> I am wondering if it's possible at all to have such a force field with
>>> Gromacs.  It does seem like all of the virtual sites are defined with
>>> respect to 2 atoms or more, which is reasonable - conceptually I don't
>>> see how I can uniquely attach a virtual site to only one atom.  Has
>>> anyone considered this problem before?
>>>
>>> Thanks!
>>>
>>> - Lee-Ping
>>>
>>>
>>>
>>> --
>>> gmx-developers mailing list
>>> gmx-developers at gromacs.org
>>> http://lists.gromacs.org/mailman/listinfo/gmx-developers
>>> Please don't post (un)subscribe requests to the list. Use the
>>> www interface or send it to gmx-developers-request at gromacs.org.
>>>
>>
>>
>




More information about the gromacs.org_gmx-developers mailing list