[gmx-users] Water molecules cannot be settled, why?

[Justin A. Lemkul]

On 6/4/12 3:53 PM, John Ladasky wrote:
> Recently I have started to get crash messages in my molecular dynamics runs
> which look like this:
>
> Reading file exp37b-prep.tpr, VERSION 4.5.4 (single precision)
> Making 1D domain decomposition 5 x 1 x 1
> starting mdrun 'Protein in water'
> 500000 steps, 1000.0 ps.
>
> step 81240: Water molecule starting at atom 45191 can not be settled.
> Check for bad contacts and/or reduce the timestep if appropriate.
> Wrote pdb files with previous and current coordinates
> [j-l:02837] *** Process received signal ***
> [j-l:02837] Signal: Segmentation fault (11)
> [j-l:02837] Signal code: Address not mapped (1)
> [j-l:02837] Failing at address: 0x422112b0
> [j-l:02837] [ 0] /lib/x86_64-linux-gnu/libpthread.so.0(+0x10060) [0x7fb79a50d060]
> [j-l:02837] [ 1] /usr/lib/libgmx_mpi.openmpi.so.6(+0x1da5a1) [0x7fb79aba55a1]
> [j-l:02837] *** End of error message ***
> --------------------------------------------------------------------------
> mpirun noticed that process rank 4 with PID 2837 on node j-l exited on signal 11
> (Segmentation fault).
> --------------------------------------------------------------------------
>
> The surprising thing to me is that I'm getting the water molecule settling error
> after a very large number of simulation steps. As you can see, this occurred on
> step 81,240. In another related simulation, it takes over 100,000 steps before
> an error occurs. Any time that I have set up a simulation incorrectly in the
> past, I always get an error quickly, usually a LINCS error.
>
> If I rerun the exact same simulation, the problem is reproducible. My current
> project compares four conformations of a partially-folded protein. Only two of
> these four conformations produce a water settling error within the time frame I
> am testing (500,000 steps, representing 1.0 ns of real time).
>
> Recently I have been trying to avoid the -deuterate option in pdb2gmx, thinking
> that I might be introducing biases into my simulations by using it. However, am
> I correct in thinking that the -deuterate option does not affect the hydrogen
> atoms of the solvent that I add using genbox and spc216.gro? So, switching back
> to using -deuterate should not affect the motion of solvent hydrogens, at least,
> not directly.
>

Your topology should tell you this, though I doubt that it's a cause of 
anything.  The fact that the water molecule cannot satisfy the SETTLE algorithm 
does not necessarily indicate that water is a problem - other clashes and 
particles moving rapidly across the system can slam into an unsuspecting water 
molecule and cause issues.

> I would appreciate any guidance you may have. Thanks!

We would need to see an .mdp file to be able to provide any help.  MD is 
chaotic, and there is no guarantee that surviving the first few (hundred, 
thousand...) steps that the simulation will be stable.  Have you watched the 
trajectory to see if anything weird is going on?  What is your minimization and 
equilibration protocol?

Other general advice for troubleshooting can be found at 
http://www.gromacs.org/Documentation/Terminology/Blowing_Up#Diagnosing_an_Unstable_System.

-Justin

-- 
========================================

Justin A. Lemkul, Ph.D.
Research Scientist
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin

========================================