Error: Too many atom sorting bins

LAMMPS LAMMPS General Discussion
,
1

Hello everyone,

I am running a reactive Nitrogen gas simulation using Reaxff potential. The initial condition is such that I have a very dilute system (density=1.25*10^-4 kg/m^3) and about a million molecules. Due to this my initial domain size is quite big (cube of length 26499 Angstroms). When I run my simulation using default settings for neighbor generation, I get an error

Error: Too many atom sorting bins

I tried increasing bin size by using the command

neigh_modify binsize 1000.0

where 1000.0 is in distance units which means 1000.0 Angstroms for my choice of units. But I still get the same error of too many bins.

Any help in this regard will be highly appreciated.

Thanks a lot,
Gunjan

2

The bin size for sorting is set with the atom_modify command — LAMMPS documentation

3

Hello Alex,
Thanks a lot for your reply.
I tried the atom_modify command command as well. I used something like this:
atom_modify sort 1000 1000.0

But I still see the same error.
ERROR on proc 0: Too many neighbor bins (src/nbin_standard.cpp:184)

I tried different binsize parameters but that doesn’t seem to affect the error.
Thank you so much for your help.

4

Didn’t you notice that this and the previously reported one are two different error messages?

5

Please note that if you ever expect that this system will shrink significantly, you will run into significant problems:

  • your simulation speed will decrease significantly as the number of neighbors is growing. Since you have a fixed cutoff and thus a fixed volume the number of pairs of neighbors will grow O(N**2) with the density. … and that is only for the pair wise terms. The cost of the angle and dihedral terms will grow by an even higher order.
  • the memory management of the ReaxFF implementation in LAMMPS is assuming that the number of neighbors does not change much. So if your density changes you may need to (completely) re-initialize the pair style and the simulation to update the estimates. The KOKKOS version (can be used with CPU only, if needed) is more robust, but not immune either.