[Pw_forum] The cell blowing up in MD simulation

[Tian Lan]

Hi all,

I am trying to run the MD simulation on Ag2O of cuprite structure, with a 2
by 2 by 2 supercell. I find it cannot converge. I just started to do MD on
QE, so I cannot figure out what is going on. Your help is appreciated very
much.

The scf calculation gives a good convergence, including the phonon
frequencies calculated by PH., I also run the vc-relax, it converged too
and gives an optimized lattice parameter very close to experimental data.
Then I used those parameters running MD, cp.x, however, even in the first
relaxation step with the constant volume, it is hard to converge. And only
if ortho_max is set to 1000, the orthonormalization cannot be fulfilled.
After the job is done, I find the final total stress exceeds 200 GPa. Then
I restart, with a random atom displacement, it does not help. I then run
the vc-cp, the cell keeps going larger, like blowing up. The temperature
cannot converge as well, and far from the target. I think it is not
surprising, since the initial stress in NVE is so high. I also changed the
cell back to the primitive one. It performs in the same way, so I think the
supercell is not the problem.

I am very confused about what is happening. It appears to be a
contradiction, I expected a good structure from scf calculation should more
or less agree with the MD run.

The basic setup is like this: I used ultrasoft potential
Ag.pbe-d-rrkjus.UPF and O.pbe-rrkjus.UPF. I followed most of default
settings, except, emass=700, and the emass_cutoff=3, nr1b=10, ecutwfc=30
(tried 20 to 70). The # of valence electrons are, 11 for Ag and 6 for O.

If you have experience about it, please give me a hint. I can provide more
information if needed.

Best,
Tian
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20120912/e0bf000f/attachment.html>