[gmx-users] neighborsearching in semiisotropic systems

[Till Rudack]

Dear users,

During protein simulations occurred the range checking error of the 
neighborsearching. I checked the comments on this mistake in the user 
list and then checked my system on any hints of an exploding system but 
it seems to be stable. In order to be sure I cut down the system to a 
small peptide consisting of only ten aminoacids and used the ffG43a1 
forcefield.


I performed 10ns simulations of this small system in a cubic 
spc-waterbox with different pressure parameters and tried single and 
double precision. I got the following results:


For

pcoupltype = semiisotropic

tau_p = 0.5 0.5

compressibility = 4.5E-5 4.5E-5

ref_p = 1 1


with double precision occurred the mistake:


-------------------------------------------------------

Program mdrun_d.mpi, VERSION 3.3.1

Source code file: nsgrid.c, line: 226


Range checking error:

Explanation: During neighborsearching, we assign each particle to a grid

based on its coordinates. If your system contains collisions or parameter

errors that give particles very high velocities you might end up with some

coordinates being +-Infinity or NaN (not-a-number). Obviously, we cannot

put these on a grid, so this is usually where we detect those errors.

Make sure your system is properly energy-minimized and that the potential

energy seems reasonable before trying again.


Variable ci has value 284. It should have been within [ 0 .. 200 ]

Please report this to the mailing list (gmx-users at gromacs.org 
<mailto:gmx-users at gromacs.org>)


Then, I changed tau_p to 5.0 and the mistake did not occurce anymore 
during the 10 ns. With single precision the mistake never occurres. I 
also tested double precision with pcoupltype = isotropic and this also 
works.

Next, I wanted to reproduce the mistake so I took the same tpr file of 
the crashed simulation with tau_p = 0.5 and started the simulation 
again. The mistake occurred again but not at the same timestep. The 
first run crashed after step 1558630 and the second run after step 
2597500 and also the ci variable is not exactly the same (Variable ci 
has value 283. It should have been within [ 0 .. 200 ]).


I think there might be a problem between the neighborsearching and the 
semiisotropic pressurcoupling together with double precision. I am quite 
sure that the peptide is not the problem. Maybe there is a problem with 
a water molecule. I would be glad if someone can give me an explanation 
why this mistake is not reproducable at the same timestep and why it 
occures with tau_p=0.5 and not with tau_p=5.0.


Thanks


Till