Hello Jackson, providing this information will not help diagnose the underlying issue. Please have a look at other properties as well, density pressure, temperature … (as I suggested before)
This is again related to how to MD.
Without a clear idea of how the entire simulation is being set up it would be hard to provide any useful suggestion.
At this point I feel that this is not a lammps question.
Thanks for understanding
Good luck.
Arun
What is your density? It could actually be that your simulation is working fine, but the neighbour list just overflows due to the normal mechanics of the system?
What if you use neigh_modify to set the maximum number of neighbours to something higher? Like 20000 instead of the 2000? If your dynamics are still bad, even that one should overflow. If they aren’t (although I am still wondering why you are getting such a high number) then you might be alright.
Did you use some post-processing tool on a trajectory to see where the particles with lots of neighbours are? Maybe something fishy is going on but it’s just tricky to diagnose.
What is your density? It could actually be that your simulation is working
fine, but the neighbour list just overflows due to the normal mechanics of
the system?What if you use neigh_modify to set the maximum number of neighbours to
something higher? Like 20000 instead of the 2000? If your dynamics are
still bad, even that one should overflow. If they aren't (although I am
still wondering why you are getting such a high number) then you might be
alright.
with a cutoff radius of 15 \AA and an additional 2 \AA neighbor list skin,
there *will* be a large number of neighbors. keep in mind, that this number
scales O(r**3). the default setting assumes typical bulk atomic systems,
where the cutoff is more likely in the range of 10 to 12 \AA. LAMMPS will
output the average number of neighbors at the end of a run, so it is a good
idea to check after equilibration how close it is to the assumed maximum of
neighbors per atom. for example, with the bundled peptide example, the
average number of neighbors changes from around 350 to about 1000 by
increasing the cutoff from 10 to 15, which closely matches the ratio of
volume between a 12 \AA sphere and an 17 \AA sphere (2.85). if the
radiation damage will cause some significant structural changes, there may
be a significant increase of neighbors for some atoms. however, it is also
conceivable to push atoms into an unphysical geometry and thus cause
unphysical structural changes.
thus it is always a good idea to visualize a system to see, if there are
any anomalies.
axel.
It is true that with a 15.0 A cutoff and a 2.0 A skin, the number of neighbors per atom could be quite large. However, the default max size of 2000 neighbors is set large enough that it is rarely exceeded unless something unphysical is happening. Hence, using neigh_modify one 20,000 will probably eliminate the error, but it may not eliminate the underlying problem. If you are using a Buckingham potential or any potential with an exponential repulsion, there is a well-known problem with “nuclear fusion” in high-energy collisions. Once the first pair fuse, releasing energy, many more pairs can follow. This manifests itself by a rapidly rising temperature and rapidly decreasing potential energy. But if you are not printing out these quantities every timestep, the first thing you will see is the error message that you cite. The solution is to add in a strong repulsive core and carefully test it with the rest of your potential to make sure the total energy is well-behaved.
I increased neigh modify to 10,000 and error eliminated. Thanks!