[Pw_forum] (no subject)
akohlmey at cmm.chem.upenn.edu
Mon Dec 3 16:30:21 CET 2007
On Dec 3, 2007 5:46 AM, lan haiping <lanhaiping at gmail.com> wrote:
> the convergence thread in your script is
> conv_thr = 1.d-3
> which is too large to give any result reliable.
> The default value is 1.0D-6..
to elaborate a little bit on this:
to converge a geometry well, you need accurate forces.
to get accurate forces you need a well converged wavefunction
(=> small conv_thr in &electrons). you also have to converge
your basis set (ecutwfc) with respect to forces. now the latter
depends on the pseudopotentials, that you are using and for
norm-conserving pseudopotentials for oxygen a cutoff of 60ry
is too small. for CP dynamics people have gotten away with 70ry
for a long time, but there is evidence that 85ry is much more
appropriate and for very accurate geometries you may need over
100ry. finally, as mentioned before, for norm-conserving pseudopotentials,
you should stick with the regular ecutwfc/ecutrho ratio of 4. this
is twice as important for geometry optimizations, since any other
grid ratio will introduce additional "ripples" in the 3d density grid
and thus make convergece more difficult.
hope this helps,
> Hai-Ping Lan
> Department of Electronics ,
> Peking University , Bejing, 100871
> lanhaiping at gmail.com, hplan at pku.edu.cn
> Pw_forum mailing list
> Pw_forum at pwscf.org
Axel Kohlmeyer akohlmey at cmm.chem.upenn.edu http://www.cmm.upenn.edu
Center for Molecular Modeling -- University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582, fax: 1-215-573-6233, office-tel: 1-215-898-5425
If you make something idiot-proof, the universe creates a better idiot.
More information about the users