[Pw_forum] Obtaining SCF convergence for Relax Calculation

Paolo Giannozzi giannozz at democritos.it
Fri Nov 18 14:11:44 CET 2011


On Wed, 2011-11-16 at 10:14 -0500, Michael Mrozik wrote:

> is there a place/table/listing where the actual determined/calculated
> values for U are maintained apart from just buried in the literature?

not that I know

>  The input description of pwscf indicates that I need to include
> values for all atoms in  my system

where is this written? You have to include U for all atoms that need it,
typically only transition metals (but not always)

>    starting_magnetization(1) = 1,
>    starting_magnetization(2) = -1,

too much: this will assume that all Fe states are fully polarized.
Not that it will make any difference though

>                  occupations = 'fixed',

you should use smearing: with fixed occupancies you may easily
run into convergence problems if you system is not an insulator
(sometimes, even if it is)

>                  mixing_mode = 'local-TF',

this is mostly useful for surface. Is this one?

>              diagonalization = 'cg',

typically much slower than the default ('david')

>                  startingpot = 'random' ,
>                  startingwfc = 'random', 

I would leave the default values for these two

P.

-- 
Paolo Giannozzi, IOM-Democritos and University of Udine, Italy





More information about the users mailing list