[Q-e-developers] Plotting many KS orbitals in one pp.x run (and more)

Sun Jan 8 16:52:22 CET 2017

Hi Leopold, thank you for the patch. I just committed it, removed the
undocumented and presumably unused case plot_num=13,14,15 and related
variable "epsilon", fixed example05. The problem in example06 shows up only
in parallel execution and is an infamous "S matirx not positive definite",
for which no easy solution is available, I am afraid

Paolo

On Fri, Jan 6, 2017 at 11:11 PM, Leopold Talirz <leopold.talirz at gmail.com>
wrote:

> Dear Paolo,
>
> thanks for moving forward so quickly with the patch.
> I have now verified that the specific namelist behavior required for the
> patch is provided by the recent gfortran, ifort, pgf90 and nag compilers
> (although I have compiled full QE only with gfortran and ifort).
>
> While testing the examples provided with PP I stumbled upon a problem in
> my patch (namely, when pp.x is used with an empty INPUTPP section). The
> attached patch resolves the problem and also includes some fixes to the
> documentation (which I have finally been able to compile).
>
> While running the examples, I noticed two further things that are not
> related to PP:
>
> * example05:  run_example (line 190) wants to copy a file feo_af.occup
> that doesn't seem to be produced by pw.x. Perhaps this output was removed
> at some point?
> * example06:  run_example (line 177) stops with
>   running the band calculation for Fe with PAW and spin-orbit...[cli_0]:
> aborting job:
> application called MPI_Abort(MPI_COMM_WORLD, 1) - process 0
>
> Apart from this, the examples run through fine.
>
>
> Best,
> Leopold
>
>
>
>
>
> > On 11 Nov 2016, at 13:10 , Nicola Marzari <nicola.marzari at epfl.ch>
> wrote:
> >
> >
> >
> > Dear All,
> >
> >
> > slightly tangential to this discussion, but just wanted to remind/alert
> everyone
> > that Giovanni Pizzi (and especially his Japanese colleagues) have made a
> major
> > effort in devising standard paths in k-space for any structure, in ways
> that are
> > fully compliant with current crystallography standards (this was not the
> case
> > up to now).
> >
> > Great if we all could adhere to this framework - see page below for a
> link to
> > the paper, an online tool that provides all points, labels and paths
> given a
> > pwscf input, and the link to the github repository if you want to use
> directly
> > the python source:
> >
> > http://www.materialscloud.org/tools/seekpath/input_structure/
> >
> >
> >                               nicola
> >
> >
> >
> > On 11/11/2016 12:41, Pignedoli, Carlo wrote:
> >> Dear all,
> >> I understand it could be more complex, but since
> >> you are going to consider this useful modification
> >> I would spend an effort in trying to allow flexibility in the selection
> of k-points:
> >>
> >> normally I will plot k-points after a bandstructure calculation
> >> performed with several k-points (could easily be 100 in one directions)
> >>
> >> I iwl probably need to plot soem selected bands at selected k-points
> >>
> >> now a "stride option" like plot every 3 kpoint would not be enough
> >> because most likely it would exclude some high simmetry points...
> >>
> >> thus a format that allows to specify
> >>
> >> 1 5 10 15 20 25 34
> >> in a calculation where e.g. 1 is G and 34 is X
> >> would be extremely useful.
> >>
> >> Kind regards
> >>
> >> Carlo
> >>
> >>
> >>
> >>
> >> Dr. Carlo Antonio Pignedoli
> >> Empa
> >> Swiss Federal Laboratories for Materials Science and Technology
> >> Überlandstrasse 129
> >> 8600 Dübendorf
> >> Switzerland
> >> Tel +41 58 765 4206
> >> Fax +41 58 765 4031
> >> e-mail carlo.pignedoli at empa.ch
> >> www.surfaces.ch
> >>
> >>
> >>> -----Original Message-----
> >>> From: Leopold Talirz [mailto:leopold.talirz at gmail.com]
> >>> Sent: Thursday, November 10, 2016 12:55
> >>> To: tone.kokalj at ijs.si
> >>> Cc: General discussion list for Quantum ESPRESSO developers;
> Yakutovich,
> >>> Aliaksandr; Pignedoli, Carlo
> >>> Subject: Re: [Q-e-developers] Plotting many KS orbitals in one pp.x
> run (and
> >>> more)
> >>>
> >>> Dear Tone,
> >>>
> >>> I agree that Guido's suggestion is the best compromise.
> >>>
> >>> Regarding your example, I think one should keep requiring the user to
> use
> >>> the kpoint keyword as well (otherwise one might end up by lots of
> orbitals by
> >>> accident).
> >>> But I guess this was not the point you were making. Regarding input
> format,
> >>> the same code will allow both
> >>>
> >>> kband(1) = 21
> >>> kband(2) = 38
> >>> spin_component(1) = 0
> >>> spin_component(2) = 3
> >>>
> >>> and
> >>>
> >>> kband = 21 38
> >>> spin_component = 0 3
> >>>
> >>> and
> >>>
> >>> kband = 21
> >>> spin_component = 0
> >>>
> >>> (at least with gfortran and intel, I haven't tested other compilers so
> far).
> >>>
> >>> Best wishes,
> >>> Leopold
> >>>
> >>>
> >>>> On 10 Nov 2016, at 11:38 , Tone Kokalj <tone.kokalj at ijs.si> wrote:
> >>>>
> >>>> On Thu, 2016-11-10 at 11:20 +0000, Leopold Talirz wrote:
> >>>>> Dear Tone,
> >>>>>
> >>>>>> In addition to lsign, I suggest also to make the spin_component
> >>>>>> constant: if more components are needed it can be done with few pp.x
> >>>>>> calculations.
> >>>>>
> >>>>> From just the input file perspective I don't see a particular reason
> >>>>> to handle spin_component differently than kpoint and kband. Or am I
> >>>>> missing something?
> >>>>
> >>>> Indeed you are right; if somebody is interested in more components it
> >>>> would make sense to make the spin_component follow the Guido
> >>>> suggestion. To illustrate: I found this useful (Guido scheme)
> >>>>
> >>>> kband(1) = 21
> >>>> kband(2) = 38
> >>>> spin_component(1) = 0
> >>>> spin_component(2) = 3
> >>>>
> >>>> according to Guide scheme this should make the total and x,y,z
> >>>> components for bands from 21 to 38. But I found this too fine grained:
> >>>>
> >>>> kband = 21 22 23 24
> >>>> spin_component = 1 1 2 1
> >>>>
> >>>>
> >>>> Best regards, Tone
> >>>> --
> >>>> J. Stefan Institute, Jamova 39, 1000 Ljubljana, Slovenia
> >>>> (tel: +386-1-477-3523 // fax: +386-1-251-9385)
> >>
> >>
> >> _______________________________________________
> >> Q-e-developers mailing list
> >> Q-e-developers at qe-forge.org
> >> http://qe-forge.org/mailman/listinfo/q-e-developers
> >>
> >
> > --
> > ----------------------------------------------------------------------
> > Prof Nicola Marzari, Chair of Theory and Simulation of Materials, EPFL
> > Director, National Centre for Competence in Research NCCR MARVEL, EPFL
> > http://theossrv1.epfl.ch/Main/Contact http://nccr-marvel.ch/en/project
>
>
>

-- 
Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche,
Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
Phone +39-0432-558216, fax +39-0432-558222
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/developers/attachments/20170108/4c6f5abf/attachment.html>