FENE bond and relaxation of 2d CG polymer

LAMMPS LAMMPS Mailing List Mirror
1

Dear LAMMPS users,

I know that my question is not directly related to LAMMPS itself,
i'm sorry if I take your time reading that. But I also believe
that somebody else can also learn from answers to it.

I'm building a relaxed 2D configuration of FENE polymer.
my initial configuration is something like (LJ units, periodic bc):
---- ---- ---- ----
   ---- ----- ---- ----
---- ---- ---- ----
and so on

the distance between bonded atoms is 0.97 (roughly equilibrium of FENE bond),
the distance between polymers (both in X and Y) is 2^(1/6) (equilibrium of LJ-12-6)

some part of my input script is :

pair_style lj/cut 2.5
...
velocity all create 1.0 1231
fix 1 all nve
fix 2 all langevin 1.0 1.0 1.0 904297
fix 3 all enforce2d
...
timestep 0.0001

If I use:
pair_coeff 1 1 1.0 2.0 2.5

I have "FENE bond too long" warning followed by Error at some point.

But if I decrease cut-off distance for LJ to, say, 1.0:
pair_coeff 1 1 1.0 1.0 2.5
(which is basically only repulsive LJ since 1.0<2^(1/6))

then everything is ok, BUT it's not what I really want.
Because when I try to relax pressure with finite temperature
(do NPT with ...iso 0 0 ${pdump} drag 2 )
obviously the whole system flies away
because there are no attractive interactions between chains.

I tried lowering integration timestep (it's quite low already),
but it does not seem to help.

I just don't see an explanation why attractive part of LJ, which is
substantially weaker than FENE, can lead to configurations where FENE
bond is too long?

Of course it should also depend on initial velocity and temperature,
I tried to lower it but it seems it does not have big influence on the problem.

Kind regards,
Denis Davydov

2

Dear LAMMPS users,

I know that my question is not directly related to LAMMPS itself,
i'm sorry if I take your time reading that. But I also believe
that somebody else can also learn from answers to it.

I'm building a relaxed 2D configuration of FENE polymer.
my initial configuration is something like (LJ units, periodic bc):
---- ---- ---- ----
---- ----- ---- ----
---- ---- ---- ----
and so on

the distance between bonded atoms is 0.97 (roughly equilibrium of FENE
bond),
the distance between polymers (both in X and Y) is 2^(1/6) (equilibrium
of LJ-12-6)

some part of my input script is :

why don't you provide the _full_ input.
this would take away so much guesswork...

pair_style lj/cut 2.5
...
velocity all create 1.0 1231
fix 1 all nve
fix 2 all langevin 1.0 1.0 1.0 904297
fix 3 all enforce2d
...
timestep 0.0001

If I use:
pair_coeff 1 1 1.0 2.0 2.5

I have "FENE bond too long" warning followed by Error at some point.

But if I decrease cut-off distance for LJ to, say, 1.0:
pair_coeff 1 1 1.0 1.0 2.5
(which is basically only repulsive LJ since 1.0<2^(1/6))

then everything is ok, BUT it's not what I really want.
Because when I try to relax pressure with finite temperature
(do NPT with ...iso 0 0 ${pdump} drag 2 )
obviously the whole system flies away
because there are no attractive interactions between chains.

I tried lowering integration timestep (it's quite low already),
but it does not seem to help.

I just don't see an explanation why attractive part of LJ, which is
substantially weaker than FENE, can lead to configurations where FENE
bond is too long?

are you using the correct settings for "special bonds", a.k.a. exclusions?
there is a "special_bonds fene" shortcut, IIRC.

cheers,
    axel.

3

Dear Dr. Kohlmeyer,

Thank you for your prompt answer and a will to help.
I attach the LAMMPS input file, data file and visualization from VMD.

Initial micro-structure is so, that the distances between the closest atoms is the following:
0.97 (bonded atoms, FENE)
2^(1/6) (distance between adjacent chains with the same Y-position; LJ minimum)

distances between atoms in adjacent chains with different Y-position:
1.6574
2^(1/6) ~ 1.12
1.0180

(these three can be easily calculated using simple trigonometry).
So in order to cover all 3 and have some attractive forces for later NPT run
i try to use lj/cut of 2.0

I also minimize (relax) initial micro-structure before MD.

p/s/ and yes, i do use special bonds to exclude double counting of LJ between adjacent atoms:
special_bonds fene

Kind regards,
Denis Davydov

in.CG2LJ (5.3 KB)

initMSr1r2minLJr3.jpg

data.2dPolyM500N20Lx196Ly40.zip (105 KB)

4

If the special bonds setting is correct then bad dynamics
can lead to stretched FENE bonds and that error.

If you get it quickly, then you have a bad intitial config,
leading to stretched bonds.

If you get it slowly, then you are not doing dynamics correctly,
or are imposing some condition that puts energy into
the system.

Steve