The ABINIT Users Mailing List ( CLOSED )

[matthieu verstraete <matthieu.jean.verstraete@gmail.com>]

On Sat, Mar 14, 2009 at 4:39 PM, Pascal Boulet <pascal.boulet@univ-provence.fr> wrote:

Hello Matthieu,

Thank you for your answer. I understand the problem. Joe also suggested to use a smaller q-point grid for the phonon band, so that the GS k-points grid is finer than the one for phonons.

finer is not enough: the grids have to be commensurable. abinit checks quite precisely that the wf it has on hand are at the precise k-points it needs (down to the tolerance reported in your error message)
 

I suppose that calculating the wfn at each k+q points is more time-consuming than using the interpolation procedure (?).

GS calculations are usually quite cheap. For the k+q calculations you can do a non-SCF one, so it's even quicker. If you have strange qpoints (incommensurable ones typically), it is usually more efficient to do point-by-point k+q calculations, but as long as they are simple enough you can just do an initial GS with e.g. 12x12x12 k-points - I think it should be able to automatically deal with doing the phonons with a coarser grid. However note that your GS grid is a FCC-adapted grid with 4 shifts, so it is not simply "cubic" (in reduced coordinates) - you still have to make sure that your q points are contained in it: if you simply generated them with a single 0 0 0 shift it should be ok...

The interpolation procedure you mention is probably that in anaddb: interpolating over q-points. This is fine, and will concern only how dense your q-point mesh is. Actually interpolating the WF wrt k is very delicate, and best done with Wannier functions. This is coming for bands and other quantities using the Wannier90 package.

Matthieu

Regards
Pascal

matthieu verstraete wrote:

Hello Pascal,

you are asking for a number of qpoints, which are at "random" places in the Brillouin Zone. As long as the q-points are _differences_ between 2 k-points in your ground state grid, you are fine, but for the 1/6 and 1/12 type q-points, your ground state grid is not enough (because (1/12,0,0) is much finer than the difference between the k-points. The DFPT equations require the wavefunctions at both k and k+q (the latter is not among your GS k-points) so what you have to do is calculate the wavefunctions explicitly at k+q in your case. This gives you a file with suffix WFQ, and is also described in the tutorial on phonons at q/=0. Then for the phonon run you tell abinit which dataset (eg N) calculated the WFQ with the input variable getwfq N. If getwfq is not specified, abinit tries to find all the WF in the normal WFK file, and fails in your case.

The absolute shift of your k grid is not important, as long as q is the difference between 2 k-points your WFK file will be enough.

Matthieu

On Fri, Mar 13, 2009 at 6:15 PM, Pascal Boulet <pascal.boulet@univ-provence.fr> wrote:

DS 4 is failing (first q-point:  1/12    0.0000  0.0000).


Josef Zwanziger wrote:

Which dataset is it failing on? the DS2 one for DDK, or the DS3 one
for the Gamma phonon?

On Fri, Mar 13, 2009 at 1:30 PM, Pascal Boulet
<pascal.boulet@univ-provence.fr> wrote:
  

Dear Abinit users,

I have a problem to compute phonon bands using the reponse function
formalism. I try to follow the tutorial for that but Abinit (v5.4.4) stops
when starting the calculation at q-points other than the gamma one.

First, here's my input file (which is largely inspired from the tutorial):
    

ndtset 15
# DS 1 : ground state self-consistency
 getwfk1   0            # Cancel default
 kptopt1   1            # Automatic generation of k points
  nqpt1   0            # Cancel default
 tolvrs1   1.0d-18      # SCF stopping criterion (modify default)
 rfphon1   0            # Cancel default

# DS 2 : Response function calculation of d/dk wave function
  iscf2   -3         # Need this non-self-consistent option for d/dk
 kptopt2   2          # Modify default to use time-reversal symmetry
 rfphon2   0          # Cancel default
 rfelfd2   2          # Calculate d/dk wave function only
 tolvrs2   0.0        # Cancel default for d/dk
 tolwfr2   1.0d-18    # Use wave function residual criterion instead

# DS 3 : Response function calculation of Q=0 phonons and electric field
pert.
 getddk3   2          # d/dk wave functions from last dataset
 kptopt3   2          # Modify default to use time-reversal symmetry
 rfelfd3   3          # Electric-field perturbation response only

# DS 4-15 : Finite-wave-vector phonon calculations
 getwfk   1          # Use GS wave functions from dataset 1
 kptopt   3          # Need full k-point set for finite-Q response
 rfphon   1          # Do phonon response
 rfatpol   1 3        # Treat displacements of all atoms
  rfdir   1 1 1      # Do all directions (symmetry will be used)
 tolvrs   1.0d-12    # This default is active for sets 3-15

# q-points
 nqpt   1
 qpt2   0.0000  0.0000  0.0000
 qpt3   0.0000  0.0000  0.0000
 qpt4   1/12    0.0000  0.0000
 qpt5   1/6     0.0000  0.0000
 qpt6   1/4     0.0000  0.0000
 qpt7   1/3     0.0000  0.0000
 qpt8   5/12    0.0000  0.0000
 qpt9   1/2     0.0000  0.0000
 qpt10  1/12    1/12    0.0000
 qpt11  1/6     1/12    0.0000
 qpt12  1/4     1/12    0.0000
 qpt13  1/3     1/12    0.0000
 qpt14  5/12    1/12    0.0000
 qpt15  1/2     1/12    0.0000
# common variables
# cell definition
 acell  3*12.059281889
 rprim  0.0  0.5  0.5        0.5  0.0  0.5        0.5  0.5  0.0
# k-points
  ngkpt   3*4           # number of k points of Monkhorst-Pack grids
 nshiftk   4
 shiftk   0.5 0.5 0.5
          0.5 0.0 0.0
          0.0 0.5 0.0
          0.0 0.0 0.5

# dft procedure and accuracy
  ixc 11
 nstep 80
 ecut 120. Ry

# spin polarization and bands occupation
  nband   4
 nsppol   1
 nspinor   1
 nspden   1

# dielectric (for semiconductors)
 diemac 9.0

# atom and symmetry
 natom  3
 ntypat  2
 typat  1 2 2
 znucl  ... ...
  xred
 0.     0.     0.
 0.25   0.25   0.25
 0.75   0.75   0.75
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<

The q-points where generated with the following input file:
    

 kptopt   1
 ngkpt3   3*12          # number of k points of Monkhorst-Pack grids
 nshiftk   1
 shiftk   0.0 0.0 0.0
 prtvol   1
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<

I get 72 q-points. For CPU-time limitation reasons I start computing phonons
for the first 12 q-points only (first input file).

Now the error message I get is the following:
The file k444.o_DS1_WFK cannot be used to start the
present calculation. It was asked that the wavefunctions be accurate, but
at least one of the k points could not be generated from a symmetrical one.
dksqmax=    1.432570E-04
Action: check your wf file and k point input variables
(e.g. kptopt or shiftk might be wrong in the present dataset or the
preparatory one.


The "usual" grid of point for FCC cells is shifted:
 nshiftk   4
 shiftk   0.5 0.5 0.5     0.5 0.0 0.0     0.0 0.5 0.0      0.0 0.0 0.5
whereas the q-points where generated with:
 nshiftk   1
 shiftk   0.0  0.0  0.0

Is this the problem?


Thank you for your hep
Pascal


--
Dr. Pascal Boulet, Computational Chemist
University Aix-Marseille I
Laboratoire Chimie Provence, UMR6264
Centre Saint-Jerome, Bat. MADIREL
F-13397 MARSEILLE Cedex 20, France
Tel. +33 (0) 491 63 71 17 Fax. +33 (0) 491 63 71 11
courriel: pascal.boulet@univ-provence.fr
http://www.lc-provence.fr
http://allos.up.univ-mrs.fr/boulet


    

-- 
Dr. Pascal Boulet, Computational Chemist
University Aix-Marseille I
Laboratoire Chimie Provence, UMR6264
Centre Saint-Jerome, Bat. MADIREL
F-13397 MARSEILLE Cedex 20, France
Tel. +33 (0) 491 63 71 17 Fax. +33 (0) 491 63 71 11
courriel: pascal.boulet@univ-provence.fr
http://www.lc-provence.fr
http://allos.up.univ-mrs.fr/boulet
    

--
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Dr. Matthieu Verstraete

European Theoretical Spectroscopy Facility (ETSF) 
Dpto. Fisica de Materiales, 
U. del Pais Vasco,            
Centro Joxe Mari Korta, Av. de Tolosa, 72,   Phone: +34-943018393
E-20018 Donostia-San Sebastian, Spain        Fax  : +34-943018390

Mail : matthieu.jean.verstraete@gmail.com
http://www-users.york.ac.uk/~mjv500

-- 
Dr. Pascal Boulet, Computational Chemist
University Aix-Marseille I
Laboratoire Chimie Provence, UMR6264
Centre Saint-Jerome, Bat. MADIREL
F-13397 MARSEILLE Cedex 20, France
Tel. +33 (0) 491 63 71 17 Fax. +33 (0) 491 63 71 11
courriel: pascal.boulet@univ-provence.fr

http://www.lc-provence.fr
http://allos.up.univ-mrs.fr/boulet

-- 
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Dr. Matthieu Verstraete

European Theoretical Spectroscopy Facility (ETSF) 
Dpto. Fisica de Materiales, 
U. del Pais Vasco,            
Centro Joxe Mari Korta, Av. de Tolosa, 72,   Phone: +34-943018393
E-20018 Donostia-San Sebastian, Spain        Fax  : +34-943018390

Mail : matthieu.jean.verstraete@gmail.com
http://www-users.york.ac.uk/~mjv500