Segmentation fault dependency on minimization tolerance

I am trying to use a structure containing graphene sheets, carbon nanotubes, and methane (attached as gCNT10,10_l2_2by2_methane.lammps) to study how surface roughness (i.e. the presence of CNTs) affects heat transfer. Unfortunately, I am getting a segmentation fault during equilibration. Interestingly, I found that when the segmentation fault occurs depends on the tolerance I set for minimization, which by itself might make sense (a more minimized structure should be more stable and less likely to crash), but for me it is the reverse. That is, if I use a higher tolerance (i.e. 1e-4 and 1e-5 instead of 1e-6 and 1e-7), the simulation runs for longer (error at 1.04 million steps instead of 484000 steps) before getting a segmentation fault. I don’t know how this makes sense physically, and, as such, do not know how to proceed. I have attached the input script I am using as nve.lammps. I am using the 1-July C++ version of LAMMPS.

gCNT10,10_l2_2by2_methane.lammps (313 KB)

nve.lammps (867 Bytes)

I am trying to use a structure containing graphene sheets, carbon nanotubes,
and methane (attached as gCNT10,10_l2_2by2_methane.lammps) to study how
surface roughness (i.e. the presence of CNTs) affects heat transfer.
Unfortunately, I am getting a segmentation fault during equilibration.
Interestingly, I found that when the segmentation fault occurs depends on
the tolerance I set for minimization, which by itself might make sense (a
more minimized structure should be more stable and less likely to crash),
but for me it is the reverse. That is, if I use a higher tolerance (i.e.
1e-4 and 1e-5 instead of 1e-6 and 1e-7), the simulation runs for longer
(error at 1.04 million steps instead of 484000 steps) before getting a

if you run for half a million steps or more, there is no more impact
from a minimization of your system. the reason for the segmentation
fault has to be a different one, i would look at the thermodynamic
data of your simulation and visualize its trajectory to see what is
going wrong. this looks more like you have bad simulation parameters
or a bad physics model.

axel.

Dr. Kohlmeyer,

The trajectory shows nothing untoward as far as I can tell (i.e., nothing getting too close, shooting apart, etc.). Also, while I understand that the effect of minimization becomes smaller as I pass 500K steps, my system shows that there is still some effect as the exact point where I get the segmentation fault changes as I change the parameters for minimization. The parameters that I am using are the ones found in CH.airebo in the LAMMPS directory.