fixing wall temperature

LAMMPS LAMMPS Mailing List Mirror
1

Dear all
As we know Lammps doesn’t have any thermal wall model. So if one wants to have a wall with a constant temperature he should use atomistic walls.
I want to calculate thermal conductivity of a liquid confined in two solid walls (up wall and down wall).
I want to use NEMD method so i have to set the upper and lower walls in two different temperature (by temp/rescale command). then i have to calculate the energy that “fix temp/rescale” added to (or absorbed from) walls . then calculate heat flux. and calculate thermal conductivity by dividing heat flux to temperature gradient. I did the work but the result was not correct. I think the main reason could be raised from modelling walls.
I used atomistic walls, LJ potential between wall atoms and a spring with “fix spring/self” command. The reason of using an spring is to hold wall atoms about their original position. But as we know in real world there is no spring attached to atoms and the energy of springs are not real. So i think using springs cases false “flux calculation” and false “thermal conductivity” calculation.
If my gusse is true then how can i model a wall with constant temperature with Lammps while there is no “thermal wall” model provided in Lammps?

regards
mahdi sahebi

2

Dear all
As we know Lammps doesn't have any thermal wall model. So if one wants to

what does a "thermal wall" do?
it is fairly straightforward to add
new features to LAMMPS, you
can always go ahead and implement
what is missing.

have a wall with a constant temperature he should use atomistic walls.

on the microscopic scale there is no "constant" temperature.
in fact, "temperature" itself is not so well defined. in computer
simulations, we generally assume equipartitioning, but that is
not rigorous.

I want to calculate thermal conductivity of a liquid confined in two solid
walls (up wall and down wall).
I want to use NEMD method so i have to set the upper and lower walls in two
different temperature (by temp/rescale command). then i have to calculate
the energy that "fix temp/rescale" added to (or absorbed from) walls . then

proper temperature control and "fix temp/rescale" are two different things.

also, aren't you computing the thermal conductivity of your wall - liquid - wall
set up and not just the liquid this way, i.e. isn't your calculation tainted
by how well the wall material and geometry transfers kinetic energy to
your liquid?

calculate heat flux. and calculate thermal conductivity by dividing heat
flux to temperature gradient. I did the work but the result was not correct.
I think the main reason could be raised from modelling walls.
I used atomistic walls, LJ potential between wall atoms and a spring with
"fix spring/self" command. The reason of using an spring is to hold wall

why do you need fix spring/self? either your wall remains a wall
or it doesn't. if it does not, you don't have a good description of
its material.

atoms about their original position. But as we know in real world there is
no spring attached to atoms and the energy of springs are not real. So i
think using springs cases false "flux calculation" and false "thermal
conductivity" calculation.

i think your problem is much more on a conceptual basis.

what is it exactly that you want to compute?

why is it so important to include the wall material?

If my gusse is true then how can i model a wall with constant temperature
with Lammps while there is no "thermal wall" model provided in Lammps?

if you want a true microscopic representation of your
system, you should use proper dimensions and proper
potentials and then there should be no need for fix
spring/self. if you want to have an abstract model,
why would you need a "thermal wall" (whatever that is)?

axel.