[gmx-users] Bug or Feature? Energy Minimization + Restraints + Polarization Possible Issue?

[Jones De Andrade]

Hi all.

We are trying something a bit really "off" on gromacs here and now: we 
want to calculate the optimum position (energetic minimum) for Drude 
Shell particles in a molecule (with longer that 1-4 intramolecular 
interactions) given conformation.

As such, we succeeded on modifying the force field to both place 
restraints on the atoms positions (easy) and adding the Drude Shell 
particles (somewhat more difficult). Basically, gmx grompp accepted the 
topology.

Then, we modified the .gro file to add the Drude Shell particles at 
random positions around the atoms. Again, seems that there are no 
issues.

The next step was to minimize the energy, and as such we tried to use 
the .mdp options steep and cg. Here is where things got strange: for 
some reason, no matter the polarizability value chosen, whenever the 
*atoms* position restraints (we did *not* explicitly placed any 
restraints on any shell) are activated, the shells are also affected and 
do not seem to move. We used several different values for polarizability 
and for the random shell positions, with the same effect.

That is a feature, the way gromacs minimization is intended to act 
under use of restraints in polarizable molecules, or not? Is there any 
way around this "issue"? We will really need highly (l-BFGS "level" 
preferably) optimized shell positions for this work, and we really did 
not expect this specific "roadblock". :)

Btw: we are using gromacs version 2018.4 just because it "was already 
available". If it is a known bug, we will happily upgrade asap.

Any suggestion will be really helpful.

Thanks a lot in advance,

Best Regards.

-- 
Jones de Andrade
(jdandrade at iq.ufrgs.br)
DFQ/IQ/UFRGS
Lattes: http://lattes.cnpq.br/6675936210583999
Orcid: https://orcid.org/0000-0003-3429-8119
ResearcherID: https://publons.com/researcher/AAC-5337-2019/