[Wien] runtime for lapw2 -qtl -p

Peter Blaha pblaha at theochem.tuwien.ac.at
Thu Dec 18 12:03:25 CET 2008


For your original cell (with 16 atoms) you do NOT need a 10x10x10 mesh, because this
cell is much larger than fcc-Cu (16 times larger, thus your mesh can be
16 times smaller)!
In addition it is a semiconductor, thus you do NOT have any effects due to partially
occupied bands (Fermi surface effects), which usually also means reducing the mesh by an
order of magnitude.

In essence: I'm pretty sure that your scf calculation can be done with A SINGLE k-point
(maybe use 2-4 if you have more processors anyway), while for a DOS you should use a bit
larger mesh (use something like 16 k-points if you have 16 processors), otherwise you
get a very "spiky" DOS (or no DOS for a single k-point, because the tetrahedron
integration does not capture the set of delta-functions coming from a single k-point).

Ludwig, Christian schrieb:
> My supercell is only 2x2x2, the complicated unit cells makes for 128 atoms total. So as I understand it, my 108 k-point are effectively worth 864 k-points, which is not terribly much. And I am dealing with a semiconductor, no idea how that influences a reasonable number of k-points.
> 
> Thanks for your comments.
> Cheers,
> Christian
> ________________________________________
> Von: wien-bounces at zeus.theochem.tuwien.ac.at [wien-bounces at zeus.theochem.tuwien.ac.at] im Auftrag von Peter Blaha [pblaha at theochem.tuwien.ac.at]
> Gesendet: Donnerstag, 18. Dezember 2008 11:17
> An: A Mailing list for WIEN2k users
> Betreff: Re: [Wien] runtime for lapw2 -qtl -p
> 
> Yes, x lapw2 -qtl -p runs actually on only one processor. It also does a lot of
> I/O, thus it is for bigger cases quite slow (but you do it only once and it blocks
> only one processor).
> 
> On the other hand: why are you using for a 128 atom supercell 108 k-points ?
> 
> If you do fcc Cu a "poor" k-mesh is 10x10x10, a good one 20x20x20.
> A 5x5x5 supercell of Cu (125 atoms/cell) would thus need a 2x2x2 (or 4x4x4)
> mesh, which even without symmetry (but plus inversion) would lead to only
> "a few" k-points in the BZ.
> 
> If you have an insulator, this k-mesh can be reduced by another order of magnitude.
> 
> Use a mesh which is in a reasonable range and optimally compatible with your 16
> processors.
> 
> 
> Ludwig, Christian schrieb:
>> Hello,
>>
>> I am calculating a 128 atom supercell with a random distribution of a constituent, so no symmetry operations. The SCF converged in 75 hours (22 iterations, 108 k-points, 16 procs) and now I am running lapw2 -qtl -p to get the DOS. Since this is not parallel it takes an awful lot of time. Over 50 hours and still running. Now I cannot help but wonder: Is it normal, that this part takes as long as the total SCF convergence or is there something wrong?
>>
>> Kind regards,
>> Christian
>> _______________________________________________
>> Wien mailing list
>> Wien at zeus.theochem.tuwien.ac.at
>> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> 
> --
> 
>                                        P.Blaha
> --------------------------------------------------------------------------
> Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
> Phone: +43-1-58801-15671             FAX: +43-1-58801-15698
> Email: blaha at theochem.tuwien.ac.at    WWW: http://info.tuwien.ac.at/theochem/
> --------------------------------------------------------------------------
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien

-- 

                                       P.Blaha
--------------------------------------------------------------------------
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-15671             FAX: +43-1-58801-15698
Email: blaha at theochem.tuwien.ac.at    WWW: http://info.tuwien.ac.at/theochem/
--------------------------------------------------------------------------


More information about the Wien mailing list