temperature seems to be not the same for all atoms in NVT

Dear LAMMPS users,

attached is input and restart files for a NVT run. watching trajectory in VMD I see there is high mobility in a region but rest of atoms in other regions stay still. when atoms of first region get to a sort of “sub-equilibrium” stops fluctuating and other atoms which have been frozen (or very low mobility) start to interact. this only happens in high lj epsilon/low.

  • is this weird behavior because LAMMPS keeps the average T of whole system constant but not T of each atom? if not, where does this effect come from?
  • does this affect the final equilibrium state of the molecule?

I’d appreciate sharing experience if anybody else had came across to similar effect.

Cheers,

Hamed.

in_restart.dat (573 Bytes)

restart.md (392 KB)

Dear LAMMPS users,

attached is input and restart files for a NVT run. watching trajectory in
VMD I see there is high mobility in a region but rest of atoms in other
regions stay still. when atoms of first region get to a sort of
"sub-equilibrium" stops fluctuating and other atoms which have been frozen
(or very low mobility) start to interact. this only happens in high lj
epsilon/low.

- is this weird behavior because LAMMPS keeps the average T of whole system
constant but not T of each atom? if not, where does this effect come from?

there is no such thing as a per atom temperature. temperature is a
macroscopic property. we do compute temperature in MD simulations
based on the assumption of equipartitioning of the kinetic energy and
that averaging over time is equivalent to averaging over volume.

- does this affect the final equilibrium state of the molecule?

that is hard to say, if i visualized the bond structure of your
system, it looks *very* strange. like as if your choice of parameters
and model is not very physical.

I'd appreciate sharing experience if anybody else had came across to similar
effect.

it usually happens when simulating a bad model leading to atoms
getting too close and the system then exploding.
MD is "garbage in, garbage out". so before worrying about a problem
with LAMMPS, you first have to worry about a problem of your general
input and model.

axel.