[lammps-users] Bulk Teflon NPT Dynamics

I am trying to run dynamics of bulk Teflon (the unit cell contains 54 -20mers) at 300K.* The system was built in Cerius2 and equilibrated there using the Jang et al forcefield. I then converted the .bgf file and force field to LAMMPS compatible ‘in’ and ‘data’ files. These files had to be modified slightly (pair_style and dihedral_style). I am attaching below my e-mail the ‘in’ file and the header of the ‘data’ file.

I then compared the single point energies in Cerius2 and LAMMPS and they agreed well for bond and angle energies…

From Cerius2:

in.rtf (6.4 KB)

data_head.rtf (1.14 KB)

The error happens when atoms in a single bond or angle are too far
away from each other and a processor can't find them to compute
the bond or angle. If you set k = 0, then it's like the atoms aren't
bonded and nothing prevents them from drifting far apart. I don't
know why you would do this. If you really mean there to be
no bond, then you should use the delete_bonds command, not set
k = 0.


Hi Steve,

My understanding is that there are two types of k's.
First there is the k0 which relates tot the equilibrium bond length and comes with the force field (defined in the header of the data file)
Then there is the k of the restraint which eventually should be set to zero to allow the natural evolution of the system.
My mindset comes from NAMD where any kind of restraint you define upon your system should eventually be removed.
I have not used LAMMPS in a long time and from reading the manual it was unclear to me why k should be different then zero.
I don't understand why atoms don't remain bonded given that their equilibrium constant from the force field/data file is nonzero.
Using a nonzero value for k doesn't that alter the values in the force field and as such the total energy?
If a nonzero value for k needs to be used what should that be, how is that decided for each system in particular and ... what is the underlying reason.
I am a little confused...


Steve Plimpton wrote:

I thought the k was for bonds/angle and not the fix. You're correct,
you should be able to have a zero spring constant for the fix self/spring
without a problem. Thought it might make more sense to just "unfix"
the fix at that point.

However the error you are getting is due to the 3 atoms in one angle
flying too far apart. If you ran on one proc, this error might not
occur (since all atoms are still in the simulatoin box), but it means
something very unphysical is happening with that angle. I would
try to viz the system and see what is going on. Also, you need the
pair cutoff to be long enough to encompass the 3 atoms, but I presume
this is the case for a Cerius ff.