"Bond/angle/dihedral extent > half of periodic box length" due to improper setting of images in input file

Hi,

I was trying to run a simulation using files from a DL_POLY simulation. Basically I converted the DL_POLY CONFIG file to a PDB file. Then with this PDB file and CHARMM force field file the LAMMPS input files was generated using ch2lmp. However, while running the equilibration, I found that the initial energy and temperature of the system was extremely high, and I have to use very small time step to get the system equilibrated. This does not look normal because the initial configuration I was using has already been equilibrated.

While looking at the ‘log.lammps’ file, I could find the following warnings.

WARNING: Inconsistent image flags (…/domain.cpp:594)
WARNING: Bond/angle/dihedral extent > half of periodic box length (…/domain.cpp:667)

Combining with the fact that in the initial stage of my equilibration the bond energy was very high, I think this might have something to do with improper bonding between atoms. After doing some search in the archive of this mailing list, I guess the warning message arises from the wrong image information in my data file. Since my input data file was generated from ch2lmp, no information about the image was given. Therefore the image number of all atoms would be initiated as zero at the beginning of the simulation, resulting in very long bond connecting two atoms in the same molecule if they were distributed on the two sides of the box, respectively.

So I’m wondering if there is anyway that I could resolve this issue by simply modifying my input parameters, or I must add image information to the data file?

I’ve also pasted my input file here, and thank you very much for your help.

Shule

Created by charmm2lammps v1.8.1 on Thu Aug 29 15:42:35 CDT 2013

units real
neighbor 2.0 bin
neigh_modify every 1

boundary p p p
atom_style full
bond_style harmonic
angle_style harmonic
dihedral_style charmm
pair_style lj/cut/coul/long 14.0
pair_modify mix arithmetic
kspace_style ewald 1e-5

read_data cyl10-1.2_assign.data

group C1 type 15
fix 3 C1 setforce 0.0 0.0 0.0

fix 1 all nvt temp 298.0 298.0 100

thermo 100
thermo_style multi
timestep 0.001

restart 100 cyl10-1.2_assign.restart1 cyl10-1.2_assign.restart2
dump 1000 all custom 100 cyl10-1.2.trj id type q x y z vx vy vz fx fy fz

run 1000000

Hi,

I was trying to run a simulation using files from a DL_POLY simulation.
Basically I converted the DL_POLY CONFIG file to a PDB file. Then with this

and with this step you removed all the useful information about the
bonded interactions, since PDB files don't store them.

PDB file and CHARMM force field file the LAMMPS input files was generated
using ch2lmp. However, while running the equilibration, I found that the

and then try to generate them again with a tool that can only work
because it makes certain assumptions about the input PDB data, that
may or may not be true for your system.

initial energy and temperature of the system was extremely high, and I have
to use very small time step to get the system equilibrated. This does not
look normal because the initial configuration I was using has already been
equilibrated.

which clearly indicates, that the conversion process did not work
correctly. you did not say, which force field you have used with
DL_POLY? you also didn't specify whether you were able to use the
exact same parameters in LAMMPS.

While looking at the 'log.lammps' file, I could find the following warnings.

WARNING: Inconsistent image flags (../domain.cpp:594)
WARNING: Bond/angle/dihedral extent > half of periodic box length
(../domain.cpp:667)

Combining with the fact that in the initial stage of my equilibration the
bond energy was very high, I think this might have something to do with
improper bonding between atoms. After doing some search in the archive of
this mailing list, I guess the warning message arises from the wrong image

there may be other problems as well. you need to narrow them down.
rather than converting the entire system, you should first work with a
few single molecules and manually confirm that the conversion is
correct or not.

information in my data file. Since my input data file was generated from
ch2lmp, no information about the image was given. Therefore the image number
of all atoms would be initiated as zero at the beginning of the simulation,
resulting in very long bond connecting two atoms in the same molecule if
they were distributed on the two sides of the box, respectively.

So I'm wondering if there is anyway that I could resolve this issue by
simply modifying my input parameters, or I must add image information to the
data file?

the main problem is in the entire approach of removing information and
then trying to recreate it based on heuristics. this almost never
works perfectly.
MD is a "garbage in, garbage out" method, that can run pretty bad
simulations without telling you that your results are nonsense. thus
particular scrutiny has to be applied already to the entire process of
preparing the system and placing blind faith in some conversion tool
is not exactly that.

axel.