[gmx-users] Energy minimization of a simple model without PBC

[Ivan Fedyanin]

Hello,

I'm totally new to gromacs and I've just started to make some simple
things with simple models to expand them later.
One of the models I concern is simply an isolated n-decane molecule in
a large box without PBC. However, I start from quite bad geometry
(e.g. bond lengths ca. 1.2-1.6A, angles down to 85 deg.) to check if
energy minimization leads to the acceptable structure. Also I use
self-constructed topology and force field files with ff parameters
taken from papers describing my (much lager) molecule of interest
(phospholipide). However, the latter parameters are quite close to
G43a2.

The first question is about SHAKE algorithm and energy minimization.
If I use steepest descend together with SHAKE algorithm to constrain
bonds, I obtain many errors such as

---
Inner product between old and new vector <= 0.0!
constraint #0 atoms 1 and 2
    i     mi      j     mj      before       after   should be
    1   1.00      2   1.00     0.15300     0.17099     0.15300
    2   1.00      3   1.00     0.15299     0.17963     0.15300
    3   1.00      4   1.00     0.15300     0.73464     0.15300
...
---
Thus the geometry produced by shake algorithm is just a set of
unconnected atoms. However, everything is almost fine with lincs, at
least constrained bonds and unconstrained angles and torsions are very
close to the ideal values.
So the question is: does shake algorithm suit for energy minimisation
or one should always use lincs?

The second question is about threshold values for energy minimization.
It's written in the gromacs manual that minimization stops if Fmax <
emtol, or nsteps is exceeded. But almost always I get the following
output:
---
Steepest Descents converged to machine precision in ... steps,
but did not reach the requested Fmax < 1.
---
I guess that it means that the change in energy is too small to
continue minimization, but what is the possible reason for that, is it
too bad or not for the isolated molecule optimization?

Thanks in advance,
Ivan