Thank you very much  for very accurate advices.
*
 **Alexander Vozny wrote: also hope you have 2 pseudoH per As*

Do you  think that I should also optimize the position of Hydroden atoms
or I shoudn't  care about it?
*
Secondly, what surface reconstruction do you use?*

I don't use any  reconstruction, I just optimize atoms on the surface.
Optimazed surface is for me  equal to reconstructed surface, am I right?
Is it possible to get reconstcruced surface after optimization or it depends
of the recontruction - I mean if reconstruction is far away from starting
geometry is it possible not to get proper results.


*600Ry is way too much! This is the main reason of your slow calculation.
300Ry should be enough.*

But I checked that for bulk calculation, cell containing 64 GaAs atoms,
Meshcutoff= 500 Ry is efficient to get accurate calculations.

Cheers,
Magdalena,

2010/7/23 Alexander Vozny <[email protected]>

> I would add:
>
> 1. H should be 0.75 if you passivate As (and 1.25 if you passivate Ga)
> otherwise you lack several electrons and this will contribute to surface
> states.
> (I also hope you have 2 pseudoH per As)
>
> 2. Secondly, what surface reconstruction do you use?
> Most likely you will have a lot of surface states (maybe even a metallic
> surface) because of Mn (and depending on surface reconstruction), so that
> your results would be very sensitive to the k-point sampling (despite it is
> a huge unit cell, you may need a 25A k-grid cutoff rather than 12A that you
> use) and thus a very slow SCF convergence.
> So yes, using a 1300K electronic temperature was a right choice.
>
> 3. 10 lattice constants in z-direction is way too much (which means
> (10-3.25)*5.76A=39A of vacuum) - this is not a problem for strictly
> localized orbitals, but for the grid it is still a considerable effort.
> ~15A of vacuum should be typically enough, especially if you turn on the
> SlabDipoleCorrection
>
> 4. 600Ry is way too much! This is the main reason of your slow calculation.
> 300Ry should be enough.
> If you really want a better grid try using GridCellSampling (even the
> shortest example from the manual would work fine)
>
> Alex.
>

Responder a