[lammps-users] why volumn keeps increase

Dear all

I am studying a argon/copper heat transfer case with boundary condition p p s with ensemble NVE.

When the temperature of my system is not large, I got very good results. But when my system’s temperature is larger, like hot wall 140k, cold wall 120k. My distance lz in z direction keeps increasing. I understand when temperature is higher, and I used NVE so the pressure will be larger. How come the larger pressure makes my volumn increase? Although I used boundary condition s in z direction, I also used NVE. My system’s volumn shouldn’t change.

Can anybody tell me why this happen?

ziyuan

Post your input script. We can't tell you much without it.
Matt

Quoting shiziyuan <[email protected]...>:

Dear all

I am studying a argon/copper heat transfer case with boundary condition p p
s with ensemble NVE.

When the temperature of my system is not large, I got very good results. But
when my system's temperature is larger, like hot wall 140k, cold wall 120k.
My distance lz in z direction keeps increasing. I understand when
temperature is higher, and I used NVE so the pressure will be larger. How
come the larger pressure makes my volumn increase? Although I used boundary
condition s in z direction, I also used NVE. My system's volumn shouldn't
change.

you are misunderstanding the meaning of NVE in the special
context of your boundary conditions. you use "p p s", i.e. shrinkwrapped
boundaries in z directions. this means a particle does not enter
back into the cell as its periodic image, but the cell is automatically
enlarged. it is effectively like having an infinite size box in z direction.

fix nve in this context only means that it does not even try to change
the box size (unlike npt). if you don't want the box to grow, you have
to use either a "p p p" boundary or "p p f" with a reflecting wall or
use a lj wall (if you choose the cutoff carefully, it will only be
reflective, too).

cheers,
    axel.

Dear Axel,

Thank you for your reply, I understand that I need to use p p f. But my system is that I have hot wall and cold wall on z direction and liquid is in between. My wall has thermal oscillation, so if I use f boundary condition. I will have an error “didn’t assign atoms position correctly”.

So do you think how I can deal with this?

ziyuan

Dear Axel,

Thank you for your reply, I understand that I need to use p p f. But my

not really. it is only required for reflecting walls.

system is that I have hot wall and cold wall on z direction and liquid is in
between. My wall has thermal oscillation, so if I use f boundary condition.
I will have an error "didn't assign atoms position correctly".

that would be an issue of your initial coordinates. you may need to shift
them or adjust the box ?lo and ?hi values so that all particles are inside
the box.

i also have a hunch of what your real problem might be: you have to time
integrate your walls, what do you do to keep them in place? do you use
something like fix spring or fix spring/self? otherwise the atoms between
the walls will keep bumping into them and transfer some momentum, but
there is nothing that pushes back.

So do you think how I can deal with this?

without seeing your input and data file, people are restricted to guessing.

cheers,
    axel.

Hi Axel,

I used eam potential to deal with my wall’s oscillation. And my wall interacts with my liquid using LJ potential. But I didn’t keep them in place. Because I thought, if the force is assigned correctly, the solid will not move. Like in reality, there is a silver container, inside is liquid. No matter what the temperature is, the walls will not move.

So I think maybe that’s the problem. MD to me is quite new. I used to do continuum mechanics. I am sorry that I asked very silly questions.

Below is my input data:

units metal
boundary p p s
atom_style full
read_data data.mysilverar

pair_style hybrid eam lj/cut 10.0
pair_coeff 2 2 lj/cut 0.0103 3.405
pair_coeff 1 1 eam Ag_u3.eam
pair_coeff 1 2 lj/cut 0.06013 2.9895

group argon type 2
group hotwall id <= 486
group coldwall subtract all argon hotwall

timestep 0.004
minimize 0.0 1.0e-8 5000 5000

velocity all create 100.0 3265486 mom yes rot yes dist gaussian
fix 2 all nvt 100.0 100.0 0.01
run 1000000
unfix 2

fix 3 hotwall langevin 140 140 0.01 8766790
fix 4 coldwall langevin 120 120 0.01 6757890
fix 5 all nve
run 1000000

ziyuan

You are using fix nve all, which means all atoms
will move, assuming there are forces on them.

Steve

Hi Axel,

I used eam potential to deal with my wall's oscillation. And my wall
interacts with my liquid using LJ potential. But I didn't keep them in
place. Because I thought, if the force is assigned correctly, the solid will
not move. Like in reality, there is a silver container, inside is liquid. No
matter what the temperature is, the walls will not move.

So I think maybe that's the problem. MD to me is quite new. I used to do
continuum mechanics. I am sorry that I asked very silly questions.

well, you can either use some silver atoms representing the container's
outside (with a couple of moveable atom layers) by not including them into
any time integrating fix (nve/nve) or put a weak restraining potential
via fix spring
on each of the walls, or use fix spring/self on the wall atoms. i would try
the last option first as that should preserve the structure of the wall.

cheers,
   axel.

Thanks a lot. I really appreciate.

ziyuan