[Pw_forum] Problem with high accuracy and parallel ph.x (Mohammad Saghayezhian)

[Andrea Dal Corso]

Which is the error message, if there is one. Note that the phonon code
at a low symmetry q point requires much more k points than the pw.x code.

Andrea


Quoting Mohammad Saghayezhian <m.saghayezhian at gmail.com>:

> Dear Andrea,
>
> Thanks for reply.
>
> The nkp in my calculation is far less than 40000. As I know, the symmetry
> reduces nkp,
> therefore for a mesh of 30*30*28 the max nkp is 25200, So if the problem
> rooted in max nkp parameter,
> above mesh should not make any problem, there are some
> tests that I did for checking the valid nkp and cut-off energy:
> Cut-off(Ry) , Kpoints: 33, 22*22*18 (OK)
>                                 35, 25*25*22 (OK)
>                                 35, 30*30*26 ()
>                                 38, 22*22*18 (OK)
>                                 38, 30*30*26 (Failed)
>                                 40, 30*30*26 (Failed)
> All of above number kpoints are lower that nkp max (40000). I also run the
> calculation
> without npool option to provide higher memory for large cut-off. but still I
> wonder why
> the last two calculation did not worked!
>
> Could it be due to optimization deficiency in compilation of ph.x files?
> Please take a look at this links:
> http://www.democritos.it/pipermail/pw_forum/2005-February/002070.html
>  http://www.democritos.it/pipermail/pw_forum/2005-February/002074.html
> Thanks a lot
>
>
>
> On Thu, Fri, 01 Jul 2011 10:01:48 +0200,  Andrea Dal Corso wrote:
>> Please note that the maximum number of k points is still a parameter in
>> QE (presently set to 40000), so to make phonon calculations with a very
>> large mesh of k points you have to increase the value of npk in
>> Modules/parameters.f90.
>
>
>> HTH,
>
>> Andrea
>
>
>
> On Thu, 2011-06-30 at 11:58 +0330, Mohammad Saghayezhian wrote:
>> Dear All,
>>
>> When I perform the scf calculation with high kp (45 45 45) and cut-off
>> (42 Ry), the parallel execution ph.x fails and stops (q-points in
>> ph.in: 4 4 3). but lowering the kp (15 15 15) and cut-off (22 Ry)
>> solves the problem and the parallel ph.x goes on without any problem.
>> Does this mean that I can not calculate phonon spectra with this
>> accuracy?
>> Also, the calculations are for metals.
> --
> -------------------------------------------------------------------
> Mohammad Saghayezhian
> Computational Condensed Matter Research Lab
> Physics Department, Isfahan University of Technology, Isfahan, Iran
>
> Tel lab: +98 311 391 3731     Fax Office: +98311 391 3746
> -------------------------------------------------------------------
>



----------------------------------------------------------------
   SISSA Webmail https://webmail.sissa.it/
   Powered by Horde http://www.horde.org/