[Pw_forum] GaAs surface with fractional charges H Zval=1.25, very large estimated scf accuracy

[Paolo Giannozzi]

On May 10, 2012, at 12:57 , Magdalena Birowska wrote:

> I found strange behaviour of first cycle of scf convergence.

it is quite normal to have a large error in the first iteration
and an even larger one in the second for surfaces. Typically
the scf error decays quite quickly afterwards. In this case
it is rather slow, although it seem to converge nicely at the end

> When I use H=1, I don't see such a behaviour.

you might start with H=1 and continue with H=1.25

>                etot_conv_thr = 1.0E-5

default (very tight)

>                forc_conv_thr = 1.0D-6

too tight: relaxation will never converge

>                       smearing  =  'fd' ,

very inefficient, should not be normally used

>                       degauss  =  0.005 ,

likely too small to have effect

>              conv_thr  =  1.0d-8

in structural relaxation it is usually convenient to start with a larger
threshold, 1.0d-6 or so; it will be automatically reduced during  
relaxation

>              diagonalization   =   'cg'

don't use this unless 'david' fails

P.
---
Paolo Giannozzi, Dept of Chemistry&Physics&Environment,
Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
Phone +39-0432-558216, fax +39-0432-558222