[gmx-users] Strange energy minimization problem with large Martini box

[Jochen Hub]

Hi all,

we are having a strange problem with an energy minimization of a large 
box of MARTINI octane. This seems like a Gromacs problem (and not a 
Martini problem), that's why I ask here.

When creating a larger octane box from a small equilibrated octane box 
with genbox, such as:

gmx genconf -f small.gro -nbox 5 5 2 -o large.pdb

Then, starting an MD directly from large.pdb works fine, as it should, 
since small.gro was equilibrated.

However, when first doing an energy minimization on large.pdb, then the 
follow-up MD stops in error due to huge box scalings. This also happens 
with even larger boxes, such as when using -nbox 6 6 2, -nbox 7 7 2, etc.

When creating a smaller box, such as:

gmx genconf -f small.gro -nbox 3 3 2 -o not-so-large.pdb

(or with -nbox 3 3 2) then no problem occurs after the minimiation. 
Also, when creating only one layer in Z-direction , such as -nbox 5 5 1 
or -nbox 10 10 1, all works fine.

Finally: the Fmax in the unstable energy minimization are reasonably low 
(10E+3 or 10E+2), but also seem a bit unstable, with occational 
increases to ~10E4. This typically means that there is an issue with the 
energy minimization.

This occurred in 2016.3 and 5.13, in single and double precision, with 
steep and cg integrator. So seems to be a general problem.

Did anyone see something similar?

Many thanks,
Jochen

-- 
---------------------------------------------------
Dr. Jochen Hub
Computational Molecular Biophysics Group
Institute for Microbiology and Genetics
Georg-August-University of Göttingen
Justus-von-Liebig-Weg 11, 37077 Göttingen, Germany.
Phone: +49-551-39-14189
http://cmb.bio.uni-goettingen.de/
---------------------------------------------------