[QE-users] Why does the difference between VASP and QE, even though they are BOMD theory
tone.kokalj at ijs.si
Wed Apr 11 11:52:44 CEST 2018
On Wed, 2018-04-11 at 11:19 +0530, Venkataramana Imandi wrote:
> Dear QE users,
> Simulation cell parameters are same in the VASP and QE. In both
> cases, simulation started from the same input structure.
> I am using 6.2.1 version of QE.
> QE: I used the BFGS for geometry optimization, Davidson
> diagonalization for electronic scf and Pt(Pt.pbe-nd-rrkjus.UPF),
> VASP: I used RMM-DIIS for geometry optimization, Davidson
> diagaonalization or cg for electronic scf and Pt(PW91-ultrasoft),
> For information,
> I obtained all O_ad(O-adsorption) sites(top,bridge,hcp and fcc) on
> pt-surface in the VASP. In VASP, I found interlayer spacing(3rd and
> 2nd layer difference) varied +0.02 to +0.03 angstroms compared to
> initial structure.
> But in QE, I obtained all O_ad sites except bridge site on Pt-
Bridge site is likely not a local minimum but a saddle point between
the two hollow sites: if the forces there are very small the optimizer
will not move the ion. This means that some small "variation" may cause
in one case the atom to be move and in the other to stay where it is.
Do a NEB calculation connecting two hollow sites via the bridge site
and you will get your answer.
If you just want the adsorption energy at the bridge site (for
comparison with other sites) then constrain the O atom laterally, i.e.
use the "0 0 1" if_pos flags.
> In QE, I found interlayer spacing(3rd and 2nd layer difference)
> varied >+0.05 angstroms more compared to initital structure.
Note that interlayer relaxations are very "soft" and you will need to
tighten the forc_conv_thr parameter: the default value of 1d-3 is
usually too large, reduce it to something like 1d-4.
Best regards, Tone
J. Stefan Institute, Jamova 39, 1000 Ljubljana, Slovenia
(tel: +386-1-477-3523 // fax: +386-1-251-9385)
More information about the users