vasp 5 GW cpu number problem.

[metosa]

Hi,

I'm trying to reproduce graphene nanoribbon bandgap results of
PRL 99, 186801 (2007) from S. G. Louie.

A graphene nanoribbon having 10 carbon atoms in its unit cell
has ~ 0.40 eV band gap in DFT. S. G. Louie corrects this gap to
~ 1.80 with G0W0.

In http://139.179.65.2/~topsakal/vasp-GW/p ... oribbon_5/
i have tried to reproduce these results with vasp 5.2.

In http://139.179.65.2/~topsakal/vasp-GW/p ... n_5/step1/
i have obtained a close dft band gap ( 0.3697 eV ). Due to computational
problems. I have reduced vacuum and kpoint. The gap occurs at gamma point
and the band structure is like this:
http://139.179.65.2/~topsakal/vasp-GW/p ... /bands.eps

In the next step i run G0W0 calculation with 16 CPU in this folder
http://139.179.65.2/~topsakal/vasp-GW/p ... 0W0-16CPU/ .
The gap increased to 1.2663 eV.

The problem is that when i run the same calculation with 32 CPU i get
a different band gap ( 1.0432 ). What is wrong with me ? Can anybody explain this ?

More problematic situation also occurs for a wider nanoribbon.
At this time band structures totally wrong and changes with cpu number.

I'm putting all my makefile , input and output files to
http://139.179.65.2/~topsakal/vasp-GW/problem1/ . Each folder has a file
EGAPxxx giving the band gap value according to KPOINTS given.

Note: I have neraly reproduced bulk Si and BN results with G0W0, GW0, GW in
PHYSICAL REVIEW B 75, 235102 (2007). I'm also putting them to
http://139.179.65.2/~topsakal/vasp-GW/p ... d_results/

Thank you very much.

Mehmet Topsakal.

--
* * * * * * * * * * * * * * * * * * * * * * * * * *
* Mehmet Topsakal *
* Ph.D. Student *
* UNAM-Institute of Materials Science and *
* Nanotechnology. Bilkent University. *
* Room: UNAM-106. 06800 Bilkent, Ankara/Türkiye *
* Tel: +90 312 290 3527 Fax: +90 312 266 4365 *
* Web: www.unam.bilkent.edu.tr/~topsakal *
* E-mail: topsakal@nano.org.tr *
* metokal@gmail.com *
* UNAM : www.nano.org.tr *
* * * * * * * * * * * * * * * * * * * * * * * * * *

[yasin]

I think you have to change NPAR and LPLANE tag.