Dear LAMMPS mailing list and developers,
We have recently run into a problem with the newly setup composite system composed of 15651 atoms with the following dimensions
-22.46 22.46 xlo xhi
-21.66 21.66 ylo yhi
-40.21 44.91 zlo zhi
The pertinent forcefield details specified in in.input
units real
dimension 3 # 3 is also the default in LAMMPS
boundary p p p # p:periodic
atom_style full
pair_style hybrid lj/cut/coul/long 12.0 buck/coul/long 12
The initial setup message when starting the run includes this line:
master list distance cutoff = 151.5
The system behavior appear to indicate that the cutoff distance is really that large (it runs much slower than the larger rhodopsin system in benchmarks, the neighbor list size has to be increased from its default value).
I have searched the mailing list (and LAMMPS manual) but can't seem to find anything else rather than "pair_style" that would affect cutoff radius, to say nothing about setting it to almost twice the largest system dimension.
Would really appreciate any helpful hints on getting to the root of our problem.
Regards
Tatiana Kuznetsova
Department of Physics and Technology, University of Bergen, Norway
Dear LAMMPS mailing list and developers,
We have recently run into a problem with the newly setup composite
system composed of 15651 atoms with the following dimensions
-22.46 22.46 xlo xhi
-21.66 21.66 ylo yhi
-40.21 44.91 zlo zhi
The pertinent forcefield details specified in in.input
units real
dimension 3 # 3 is also the default in LAMMPS
boundary p p p # p:periodic
atom_style full
pair_style hybrid lj/cut/coul/long 12.0 buck/coul/long 12
The initial setup message when starting the run includes this line:
master list distance cutoff = 151.5
The system behavior appear to indicate that the cutoff distance is
really that large (it runs much slower than the larger rhodopsin system
in benchmarks, the neighbor list size has to be increased from its
default value).
I have searched the mailing list (and LAMMPS manual) but can't seem to
find anything else rather than "pair_style" that would affect cutoff
radius, to say nothing about setting it to almost twice the largest
system dimension.
there are more things that can have an impact like pair_coeff and
neighbor list settings
you also need to tell us whether this issue shows up with the very
latest version of LAMMPS.
Would really appreciate any helpful hints on getting to the root of our
problem.
the only way to say something for certain, is if you make a fully
functional input deck available.
axel.
The issue appears when using the 15 May 2015 version of LAMMPS. Please find attached the log file that shows the version and the cutoff distance as well as files necessary to recreate the problem.
Tatiana K.
data.HemCalcWatCO2_Sim_2 (2.96 MB)
HemCalcWatCO2.fflammps (10.4 KB)
in.input (4.13 KB)
output (2.42 KB)
This line in your coeff file:
pair_coeff 8 10 lj/cut/coul/long 0.0000000 3.407 149 #Fill O1 - Zeroed
sets the LJ cutoff to 149 Angs for that type pair.
Add a skin of 2.5 and you get 151.5.
Steve
Oops. Deleted the "149" holdover and left "page" in neigh_modify on its default setting, an immediate speed boosted.
Really appreciate your help.
Tatiana K.