negative stress and boundary conditions

Hello LAMMPS users,
I have used the LAMMPS to compress some cylinder metal pillars. However at a very low strain (0 ~ 0.015), i got a negative stress which makes no sense. Has anyone seen this problem before and how to solve this problem.
The other problem is when i used p s s for the boundary, because the s condition will encompass all the atoms inside the simulation box. So the Y and Z direction expanded so much. Is this problem related to a bad dynamics such as bad EAM potential or too small sample dimensions.
Thanks

What is the stress with no strain, e.g. at time = 0.0?
There is no conceptual problem with negative stress. It
just means the system wants to shrink in size.

Steve

Hello,
The tress at 0 strain is around -1GPa. I have added the negative to all calculation to make the value positive. So actually at very low strain, LAMMPS gives a positive value which means the system wants to expand in size. However, i am doing a compression test. It should not be that way.
Thanks.

Junhua Gu

You need to relax your system before doing compression test (put it to the energy minimum given by the particular interatomic potential).

Oleg

16:59, 10 декабря 2014 г., Junhua Gu <[email protected]…4193…>:

Hello,
The tress at 0 strain is around -1GPa. I have added the negative to all
calculation to make the value positive. So actually at very low strain,
LAMMPS gives a positive value which means the system wants to expand in
size. However, i am doing a compression test. It should not be that way.

please keep in mind that when you don't get the result you expect,
there are *at least* three possible explanations:
1) your expectation is wrong, i.e. you didn't fully understand what it
is that you are calculating
2) your methodology is wrong, i.e. the procedure you have scripted in
your input is not correct for your needs
3) there is a bug in LAMMPS. in this case, however, you need to
provide convincing proof beyond "i don't get the number i expect" that
allows to quickly reproduce the issue (there are plenty of examples
how to do it well in the mailing list archives).

since you are doing something that is apparently a common operation,
the chances are high, that cases 1) or 2) apply here.

axel.