[gmx-users] On robustness of settle algorithm

[Qikai Li]

I've done some tests on a box of water.
First I set up a box filled with water, use the default PBC (xyz), and
run  for 100,000 MD steps without any problem for both FLEXIBLE and
settle (spt.itp).

Then, I enlarge the box along one direction, for example, z direction by
around 1~2 nm, at the same time, freeze the bottom water layer (about 1
nm thick), and the boundary condition is as before.

The water model is of spc, when settle algorithm is used,  the mdrun
almost always aborts with the following messages on screen:

++++++++++++++++++++++++++++++++++++++++++++++++++++
Back Off! I just backed up traj.trr to ./#traj.trr.1#
Wrote pdb files with previous and current coordinates
Wrote pdb files with previous and current coordinates
Wrote pdb files with previous and current coordinates
Wrote pdb files with previous and current coordinates
Wrote pdb files with previous and current coordinates
Wrote pdb files with previous and current coordinates
Wrote pdb files with previous and current coordinates
Wrote pdb files with previous and current coordinates
Wrote pdb files with previous and current coordinates

Back Off! I just backed up step3596.pdb to ./#step3596.pdb.1#
Wrote pdb files with previous and current coordinates
Wrote pdb files with previous and current coordinates
Wrote pdb files with previous and current coordinates
Wrote pdb files with previous and current coordinates
Wrote pdb files with previous and current coordinates
Wrote pdb files with previous and current coordinates
Wrote pdb files with previous and current coordinates
Wrote pdb files with previous and current coordinates

Back Off! I just backed up step3888.pdb to ./#step3888.pdb.1#
Wrote pdb files with previous and current coordinates
Wrote pdb files with previous and current coordinates
Wrote pdb files with previous and current coordinates

Back Off! I just backed up step3929.pdb to ./#step3929.pdb.1#
Wrote pdb files with previous and current coordinates

Back Off! I just backed up step3930.pdb to ./#step3930.pdb.1#
Wrote pdb files with previous and current coordinates

Back Off! I just backed up step3931.pdb to ./#step3931.pdb.1#
Wrote pdb files with previous and current coordinates

Back Off! I just backed up step3932.pdb to ./#step3932.pdb.1#
Wrote pdb files with previous and current coordinates
Fatal error: ci = -2147483648 should be in 0 .. 5599 [FILE nsgrid.c,
LINE 218]
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++


and the md.log file with the following messages:

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
        LJ (SR)        LJ (LR)   Coulomb (SR)   Coulomb (LR)
Potential     7.80686e+05   -1.01402e+03   -7.31035e+05   -7.84496e+04
-2.98123e+04
    Kinetic En.   Total Energy    Temperature Pressure (bar)
    9.08501e+04    6.10378e+04    3.09634e+02    6.27565e+04


t = 1.675 ps: Water molecule starting at atom 36247 can not be settled.
Check for bad contacts and/or reduce the timestep.Wrote pdb files with
previous and current coordinates

t = 1.676 ps: Water molecule starting at atom 36247 can not be settled.
Check for bad contacts and/or reduce the timestep.Wrote pdb files with
previous and current coordinates
Large VCM(group rest):      0.00147,     -0.00212,      0.00208,
ekin-cm:  1.65710e+00

t = 1.677 ps: Water molecule starting at atom 36247 can not be settled.
Check for bad contacts and/or reduce the timestep.Wrote pdb files with
previous and current coordinates
..................
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

However,  when I add a define line  as follows in the md.mdp control file:

define   =  -DFLEXIBLE

then, mdrun can be finished with any error/warning message.

I've also tried to trace into the mdrun with -g option, and found that
the problem is related to a block of assembly code.
I don't know if someone had a similar experience.

Thanks,
Qikai Li