Question on "soft potential to remove atom overlap"

Dear members,

I implemented soft potential to remove atoms overlap as suggested by a member.

( Followed member suggestion to look up “micelle” simulation script )

The NVE simulation using soft potential gives appropriate temperature, pressure and energy. But when I

switch to the desired potential, the pressure and energy becomes infinite instantly.

  1. So the overlapping atoms are not pushed apart ?

  2. Could anyone please suggest why and how to over come it ?

Below I briefly describe problem setup and simulation parameters that can help in diagnosis.

( As suggested in lammps manual, I tried “fix nve/limit” and “fix viscous”

also – but not yet successful )

Assuming you have a consistent force field, what I typically do to remove any remaining overlaps is to run with a small time step (1e-4 fs) for a few steps (1000) and check the energy. Then slowly increase the time step in between 1000 step runs. I haven’t done this in lammps yet but it seems to work for MD in general if you have an appropriate force field. During these runs you can use velocity rescaling to keep things behaving. You would do this with the force field you are interested in, not the soft force field you are using to remove overlaps.

-s-

Dear Salomon,

Thank you for the suggestion. I will test it.

regards

shankar

I have had success with the method Salomon describes in solids for circumstances when I do not want to remove atoms.

However, If you still are having difficulties, I wonder if there is an easier way to create liquid argon(as I think was stated many subject lines ago). If it were me, I would try creating argon atoms on a regular lattice with the desired density (optionally you could randomize those positions with displace_atoms random), create velocity and perform a simulation at the desired temperature.

Dear Zachary,

Thank you for the suggestion.

Initially, I started with creating argon on lattice.

I , however, could not get the exact density by arbitrarily changing the lattice size.

I am trying to reproduce the results of a published paper for that density,

domain size, and number of argon atoms – and so do not want to

adjust the domain size.

So I switched to random atom creation with exact number of atoms in the paper.

I first ran a simple NVT simulation of argon and got the appropriate

equilibrium pressure at a chosen temperature.

In this simulation, I used random argon atom creation, then energy minimization to remove atom overlap and it worked fine.

Now I am trying to do argon in equilibrium around cnt ( published paper reference).

I am now having problems with energy minimization over argon only, keeping

cnt atoms fixed.

Axel. K made the following suggestions and I am implementing it – not yet successfully, though.

  1. used "fix setforce " to keep cnt fixed. But minimization fails on the first

iteration.

  1. To fix (1), soft potential was suggested and the experience is documented in this

email thread.

  1. Tired “fix nve/limit” and “fix viscous” also to drain energy and move atoms apart.

It did not work. Have to take another look.

This is my first MD/VMD/LAMMPS learning problem – doubtless, these issues

are probably common for beginners – and I appreciate the help I am getting
from the members, and the email archives.

regards

shankar

Initially, I started with creating argon on lattice.

I , however, could not get the exact density by arbitrarily changing the lattice size.

I am trying to reproduce the results of a published paper for that density,

domain size, and number of argon atoms – and so do not want to

adjust the domain size.

If you go back to that old script, you could count all argon atoms you have, and try using delete_atoms porosity to get to the number you want.

This is my first MD/VMD/LAMMPS learning problem – doubtless, these issues

are probably common for beginners – and I appreciate the help I am getting
from the members, and the email archives.

Causing “explosions” with overlaps is certainly part of the learning experience. I’m not sure the overlaps are particularly useful in this case.

Dear Zachary

why don't you set back a little before you get completely distracted.

i would recommend you go back and start from scratch and produce the
most minimal input that models what you want and post it here for
discussion. it is more likely that you are not using the recommended
approaches correctly, than that they don't work. since there are
several suggestions and this is a rather common problem, this would be
a good opportunity to work out how to apply each strategy to the same
problem and then collect the results as a tutorial page for the LAMMPS
homepage. you should also state in more details what the requirements
for the resulting system should be.
then everybody that suggested a method can help you to get it working
and you can write a summary and put it all together. in the end, you
will have learned what you need (and then some) and future LAMMPS
users can benefit from your experience.

axel.

Dear Axel

I will soon provide the essential details of the problem.
( Also information on the paper I am trying to reproduce and also my code.)

To clarify to members, I do not want to convey an impression that the
suggestions/methods do not work. Not at all.

On the contrary, I have not yet been able to make one of them work.

Perhaps my usage is improper ; or the problem is elsewhere and not overlap at all.

regards

shankar

No point in making people’s heads spin thinking about your problem if you don’t show the full input script.

I’m impressed by the goodwill of all the users who provided you with advice assuming that indeed the problem
is a “physical” one.

Just put together a test deck that keep failing and send it. Otherwise you would have cried wolf way too many times…

Carlos