A few questions regarding a simulation of solid-liquid and melting temperature‏

LAMMPS LAMMPS Mailing List Mirror
1

Hi,

I want to use the solid-liquid interface technique to determine the melting temperature of a

LJ liquid.

I want to do this by having a large simulation box, with the liquid at the bottom

and the solid at the top, visually inspecting the movement of the interface,

Using the idea from J. Morris and X. Song, J. Chem. Phys. 116, 9352 (2002).

My problem is, the second read_data command doesn’t add the atoms to the specified region.

Could you please point to what am I doing wrong?

in the google drive link are the two dump files in compressed form and the practice

script i’m using “in.Tmelting_snapshot”.

https://drive.google.com/open?id=0B4_1VwCLMtkcREVrckJYalZuLUk

I want to point out, the script isn’t the full simulation script,only the part where I use read_dump

and afterwards the dump command to visualize my system.

My second question is about the example of melting temperature of aluminium in the

scripts and howto section:

http://lammps.sandia.gov/user/script4/in.two_phase_simulation

In this script, the author deletes atoms from the simulation box and my question is why?

Thank you,

2

this sounds quite complicated.

i would just create your entire lennard-jones system using the “create_atoms” command at the expected density around the melting point.
then define two groups with each representing half of the system.
then you can run time integration and thermal equilibration on each group separately while_alternating_ a couple times between them until you finally run the entire system with the same thermostat/integrator.

this way you bypass all the complications from having to combine to separately and independently equilibrated system.

3

Hi Axel,

Thank you for for the answer, I will address it with my adviser.

However, my “problem” lies in the fact that i’ll need to quench and re-heat one of the groups,

whereas now, I have all this data already.

Is there a way to use read_dump twice to construct this vertical set-up? I “feel” that i’m missing something

Thank you,

4

Hi Axel,

Thank you for for the answer, I will address it with my adviser.

However, my "problem" lies in the fact that i'll need to quench and
re-heat one of the groups,

whereas now, I have all this data already.

​which is mostly useless, since by placing two independent systems
side-by-side, you'll have two "incompatible" interfaces, which you have to
"heal out"​ to become. you cannot put them too close together, or else
you'll have close contacts and atoms with high potential energy flying
through your system. and if you place them further apart, you'll insert a
vacuum and will have to deal with a little "implosion". in both cases,
you'll have to deal with strong systematic sound waves being generated and
those must be completely dissipated before starting a production
simulation. doing the melting and quenching part for the liquid atoms,
while having the solid atoms already in place (but not moving them), and
equilibrating the solid atoms (while not moving the liquid ones and
preserving their velocities), and switching a few times around will cleanly
avoid this. so at best your saving a little bit of time by using a
"pre-melted" liquid component.
please also, note, that you can employ replication to speed up the
setup/equilibration process: by starting out with a system much smaller in
x- and y- direction but the final size in z-direction, you'll save a lot of
time for the basic procedure, and then use "replicate n n 1" before the
final equilibration part, where you only need to "heal out" the small
overstructuring from the replication.

Is there a way to use read_dump twice to construct this vertical set-up? I
"feel" that i'm missing something

​read_dump is not a good option to build a system, but it has an option to
add particles to a simulation. read the documentation! it doesn't solve the
"incompatible surfaces" issue.

however, if you are struggling with your overall approach, it is a very bad
idea to do it with a production size simulation. this is something that is
true for any kind of computational approach. you've got to learn and
understand the principles of what you are doing first. switching to a
smaller test setup allows you to quickly explore your options, and avoids
wasting time on ineffective approaches.

axel.
​

5

Thank you very much,

the picture is clear now