[Pw_forum] Interpretation of vc-relax / pi-pi-stack

Kolb, Brian kolbba5 at wfu.edu
Tue Aug 2 15:21:35 CEST 2011

Hi Guntram,

In general, the results of an optimization will differ from experiment.
This is especially true with vdW-DF, which has been shown to give binding
distances that are too large in general.  That being said, I agree that the
difference between your computed result and experiment seems excessive.
Assuming all your atomic coordinates and cell parameters are correct, one
potential trouble I see in your input file is the "conv_thr" parameter.
1e-5 is quite a poor convergence criterion, especially for a variable-cell
relaxation run.  Forces and stresses tend to be harder to converge than
energies so you often need very low energy convergence thresholds for
relaxation runs.  I generally don't run any relaxation calculation without a
cutoff of 1e-8 or better.  Without seeing your output I can't know if this
will help but it's good practice in any case.

If these are ionic systems you're doing, as it seems they are, you might
need to do spin-polarization calculations.  Unfortunately, there is no
rigorous way to do spin polarization at present using vdW-DF.  Just as a
check you could try a spin polarized run with the Grimme corrections
(london=.true. in the &system section).

If you could attach an output file, I might be able to tell you a bit more.

Hope this helps,
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20110802/8aa983f1/attachment.html>

More information about the users mailing list