Problem by increasing timestep

Dear LAMMPS-Users,

I am trying to simulate a non-equilibrium molecular dynamic simulation. I made a simulation model as attached in conjunction with this email. At first I have set a simulation to get the equilibrate state in reservoir with 0.02 fs timestep. After that I connected a pore element to the pool region and applied the “fix addforce” command aimed at implementing an external force on water molecules. When I want to increase the timestep I got “nan” error for pressure and energy at timestep=2 fs and a weird temperature at timestep= 0.2 fs and maybe a reasonable at timestep=0.02 fs as following

timestep=2 fs

Step Temp PotEng TotEng Press Volume
100 250.68264 -12777.883 -10096.048 4771.9805 45000
1000 -nan -nan -nan -nan 45000
1100 -nan -nan -nan -nan 45000

timestep=0.2 fs

Step Temp PotEng TotEng Press Volume
100 250.68264 -12777.883 -10096.048 4771.9805 45000
1000 2.0061574e-267 -12726.568 -12726.568 2477.494 45000
1100 0 -12743.755 -12743.755 2456.5811 45000

timestep=0.02 fs

Step Temp PotEng TotEng Press Volume
100 250.68264 -12777.883 -10096.048 4771.9805 45000
1000 296.31565 -9936.5383 -6766.5158 15341.201 45000
1100 300.96602 -10322.804 -7103.0315 13559.941 45000

Does it mean that I need to use the ts= 0.02 fs for my model even when applying an external force or I have to find something of which leads to these issues?

Whit my poor knowledge on MD I think the ts=0.02 fs may be appropriate to get the equilibrate state but I do not know whether it is logical or not for non-equilibrium simulation because referring to some relevant published papers I found that 2 fs as timestep used by the authors for non-equilibrium simulation. Please take a look at my input script at the end of this email.

Screenshot from 2015-02-17 13:23:14.png

Dear LAMMPS-Users,

I am trying to simulate a non-equilibrium molecular dynamic simulation. I
made a simulation model as attached in conjunction with this email. At first
I have set a simulation to get the equilibrate state in reservoir with 0.02
fs timestep. After that I connected a pore element to the pool region and
applied the "fix addforce" command aimed at implementing an external force
on water molecules. When I want to increase the timestep I got "nan" error
for pressure and energy at timestep=2 fs and a weird temperature at
timestep= 0.2 fs and maybe a reasonable at timestep=0.02 fs as following

no. *all* simulations are obviously crap.

didn't you see a warning about atoms being time integrated multiple times?
also, i am not surprised that the time integration chokes on you with
as ridiculously short a time constant as you are using.when you crank
up the time step this much.
MD is "garbage in, garbage out" and it is pointless to discuss minor
issues as the length of a time step, when you have much bigger issues
to resolve.
doesn't the massive pressure worry, too?

the best MD code is helpless against careless and inattentive users.

axel.

Dear Axel and Users,

Thank you very much for your reply. Yes, you’re right. I got the "warning: One or more atoms are time integrated more than once ". Actually, I thought that this warning not to be serious.

I am attempting to find out that from where the problem comes.

Thanks in advance

MajiD

Dear Axel and Users,

Thank you very much for your reply. Yes, you're right. I got the "warning:
One or more atoms are time integrated more than once ". Actually, I thought
that this warning not to be serious.

how did you determine this? it has been discussed on this mailing list
so many times that even a minimalist web search would have told you.

I am attempting to find out that from where the problem comes.

this particular one comes from time integrating all atoms, while also
using fix move to immobilize the carbon atom.
immobilizing atoms is also a subject of many recurring inquiries on
this very mailing list and it is recommended time and again, to simply
exclude the immobilized atoms from time integration and only - if from
some reason this is not possible - initialize their velocities to zero
and use fix setforce to remove the forces. but this will most
certainly mess up your temperature computation, since you can compute
a meaningful temperature only from the mobile atoms. so you have to
take care of that as well. this is all pretty standard stuff and has
been discussed to death. please make an effort to not repeat issues
that have been discussed and solved so many, many times, or by the
time you have a really difficult and new problem, you will be ignored
for having worn out any good will on already solved problems.

axel.

Dear Axel,

Sorry for bothering you and thanks for your valuable time. I did not consider that warning because I asked it from a wrong person. He told me it is not very important. I also did not attempt to check it out myself by referring to mailing lists. It was my own mistake.
Yes, as you did favor and explain it completely I could find that the problem comes from the “fix move” and “fix nvt” on “all atoms” just by a short searching on mailing list (http://lammps.sandia.gov/threads/msg21878.html).

Sorry again for my trivial question

Thanks in advance

MajiD
.