[Pw_forum] K-point in phonon calculation

[程迎春]

在 2011年8月21日 上午4:36，Lo_wan_2005XW <lo_wan_2005 at hotmail.com>写道：

>  Dear developer and users:
>
> Exactly follow the userguide, I am trying to calculate the phonon.
>
> Firstly I make a self-consistent calculation with
>
> K_POINTS automatic
> 8 8 8   0  0  0
>
> After the calculation convergent, I start the ph.x calculation. But the calculation is extremely slow, thus I goto
> the output-file, and find a strange thing.
>
> For some q-qoint, I find;
> number of k points=  1024  Marzari-Vanderbilt smearing, width (Ry)=  0.0500
>
> 8*8*8=512. If you set the spin polarized scf calculation, I think the
number of total k points should multiply 2 again, so there are 1024 kpoints.


>
> I understand that phonon may break some symmetry, but how it can go to 1024-k-point when I use 8-8-8 mesh?
>
> I already use "start_q=X, last_q=X, recover=.true.," to separate q-point. I also use parallel to speed up the calculation.
> But it is still extremely slow. My system has only 12 atoms per unit cell, and it seems that some q-point
> (where it requires 1024 k-points) may need one month!
>
> I don't know if there is other method to speed up the calculation?
>
> 1,To make sure pwscf can find all symmetry operation in scf calculation.
2, To reduce the number of k points, ecut energy~
Hope it helps!

>
> Thanks!
>
> Xiangang Wan
>
>
>
>
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum
>
>


-- 
Y. C. Cheng
Department of Physics
Nanjing University
Nanjing 210093
P. R. China
Tel: 86-25-83592907
Email: yccheng.nju at gmail.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20110821/37fa470b/attachment.html>