[Pw_forum] graphite cell optimization failed

[Lorenzo Paulatto]

On Wed, 23 Mar 2011 17:33:02 +0100, Eric Germaneau <germaneau at gucas.ac.cn>  
wrote:
> I'm trying to optimize unit cell parameter of graphite.
> The problem is the c axis gets too long.
> Starting from published data my initial parameters are a=4.59203 Bohr
> and c=12.513 Bohr.
> The final values are 4.664 and 15.829 respectively.
> I need someone to explain to me what I did wrong, I've attached my input
> file.

Dear Eric,
your input atomic positions are only specified to 4 significant digits:  
0.3333 is not really 1/3. Hence you are loosing 18 out of 24 symmetry  
operation: I would suggest increasing the precision to 6 or 7 digits to  
recover them. With more symmetry you will be able to afford more kpoints,  
as graphite can be tricky to converge a 4x4x4 grid can not insufficient. I  
also doubt that you need 80Ry to converge an ultrasoft pseudopotential,  
but I assume you did your tests.

Finally, your results seem to be in accord with other DFT calculation;  
considered you're using GGA they are actually quite good. In LDA the usual  
overbinding partially compensates for the lack of VdW interaction, while  
this effect does not occur for gradient-corrected functionals.

best regards


-- 
Lorenzo Paulatto (IdR)
IMPMC - CNRS UMR 7590 & Université P&M Curie
T23-C13/23-5e27 - 4 place Jussieu - 75252 Paris Cedex5
phone: +33 (0)144 27 5211
www:   http://www-int.impmc.upmc.fr/~paulatto/