Problem about "soft" pair potential running a bead-spring model

Hi,
I created a lammps data file of bead-spring polymer chains using the chain tool, as the manual says, we need to run the system initially a "soft" pair potential to un-overlap it.  
However, when I run the "soft" pair potential, it seems that the  potential  energy become bigger and bigger, and the energy is very huge when I change the soft pair potential to lj pair potential, This is not normal.

Following is the input file, can somebody please see what's wrong!
Thanks,
Yuan

# FENE beadspring benchmark
units		lj
atom_style	bond
special_bonds   fene

read_data	data_ne.chain

neighbor	0.4 bin
neigh_modify	every 1 delay 1

bond_style      fene
bond_coeff	1 30.0 2.0 1.0 1.0

pair_style soft  2.5
pair_coeff * *   0.0  
variable prefactor equal ramp(0,30)
fix      1 all adapt 1 pair soft a * * v_prefactor

velocity  all create 1.0 2349852 rot yes dist gaussian

fix	   2 all langevin 1.0 1.0 10.0 904297 
fix	   3 all nve 

thermo          100
timestep	    0.006

dump		r1 all xyz 1000 relax.xyz     

run             100000

unfix 1
unfix 2
unfix 3

Don't know, have you tried visuzlizing the system to see
what is happening. Or computing an RDF to see if
you are in fact pushing beads off of each other?

Steve

Hi,

I created a lammps data file of bead-spring polymer chains using the chain
tool, as the manual says, we need to run the system initially a "soft" pair
potential to un-overlap it.

However, when I run the "soft" pair potential, it seems that the potential
energy become bigger and bigger, and the energy is very huge when I change
the soft pair potential to lj pair potential, This is not normal.

Following is the input file, can somebody please see what's wrong!

if you ramp up the potential, potential energy *has* to increase,
but you should probably reduce the time constant on fix langevin
to remove the excess kinetic energy faster. also, it may be a good
idea to reduce the length of the timestep (and run for more steps)
to make sure they rise in energy is not due to instability of the
integrator.

axel.

I have a similar script which is stable, but I agree it can be a nagging headache to tune. The only differences I see between our scripts that are apparent:

  1. We ramp from 0 to 10 for 10,000 time steps (rather than 0 to 30 for 100,000)

  2. In fix langevin, we use damp=sigma*sqrt(mass/epsilon); your damp value may be inconsistent with this unless you have very large particle masses which I can’t tell because it is not reported (see manual explanations of damp and gamma in fix langevin and fix viscous)

  3. After the reset to lj/cut, but before the main run, we run the minimize function with the following parameters,

min_style cg

minimize 0.00000001 0.000000001 10000 100000

and then reset the time.

Other than that, it looks the same. My guess is that the differences in 1) above are inconsequential, but 2) and 3) will help.

Fred

Hi,

I think the problem is that you set the cutoff distance for the ‘soft’ potential too large. Instead of 2.5, you should use 1.0 for the cutoff distance.
On the website (http://lammps.sandia.gov/doc/pair_soft.html), the ‘soft’ is a cosine function, which mean you have to set the cutoff distance of this potential closer to the equilibrium distance of your LJ potential you are going to use.

For soft potential, you can use:

pair_style soft 1.0
pair_coeff * * 0.0 1.0
variable prefactor1 equal ramp(0,60)
fix 1 all adapt 1 pair soft a * * v_prefactor1

For FENE bond, you can use:

bond_style fene
bond_coeff 1 30.0 1.5 1.0 1.0

These parameters work well for my simulation.

Cheers,

Thanks Steve, Axel and Ying Li for All your help!
I changed the cutoff distance of the ‘soft’ potential to 1.0, really very effective! But another problem occurs: I need a non-periodic system, so I have to change some codes in “chain.f” to generate a non-periodic system. When I run the ‘soft’ pair potential, the atoms will move in the non-periodic direction. I tried the method to construct bead-spring polymer thin film using 2 repulsive walls, But after the walls are removed, the atoms still will move in the non-periodicl direction.
Could somebody please tell me how to generate a data.chain file which can used in a non-period system? Why the atoms move towards the non-periodic direction during the ‘soft’ relaxing stage? (Following is the input file)
Thanks very much!