The ABINIT Users Mailing List ( CLOSED )

[ธนูสิทธิ์ บุรินทร์ประโคน <thanusit@kku.ac.th>]

Dear Matteo Giantomassi

Thank you very much for your reply. I will try as you suggested on the KSS file generation.

Since the k-grid for screening calculation is the same as the mesh used to generate the KSS file, does it mean that Abinit will take the k-grid directly from the KSS file without any need to input separately the k-grid for the screening run?

Kind Regards,

Thanusit Burinprakhon


On May 13, 2008 10:12:58 PM ICT, "Matteo Giantomassi" <Matteo.Giantomassi@uclouvain.be> wrote:

> > Dear All
> >
> > I've tried to perform the GW corrections for GaP band structue. My
> > platform is Fedora 8.0 running on an Intel Core 2 Quad CPU with 4 Gb
> > of RAM. I firstly calculated the ground state density and the KSS file,
> > by using the parallelized Abinit 5.4.4(MPI version, prepared for a
> > x86_64_linux) and the Abinit 5.4.4(sequential version, prepared for a
> > x86_64_
> > linux), respectively. Both calculations were successfully completed.
> >
> > However, when I carried out the screening calculation using the
> > parallelized
> > Abinit 5.4.4 with the pre-calculated KSS file, the program seemed to
> > terminate
> > within the screening main loop at the start of the "cchi0" calculation of
> > the
> > q point number 2. The error message at the end of the log file is as
> > follows.
> > The
> > complete input and log files are listed below. (sorry if it's very
> > lengthy)
> >
> > ....................................................................
> > cchi0 : enter
> >
> > use symmetries = 0
> > use time reversal = F
> > spectral method = 0
> > calculating chi0(q,omega,G,G")
> >
> >
> > cchi0 : reading wavefunctions from file
> > trying to allocate wavefunctions in real space
> > allocated
> > cchi0 : real space wavefunctions are now in memory
> >
> > calculation status ( 865 to be completed) :
> > cchi0 : ik = 1 / 865 is = 1 done by processor 0
> > cchi0 : ik = 2 / 865 is = 1 done by processor 0
> >
> > cchi0 : ERROR -
> > kp=k-q-G0 not found
> > ik = 3 k = 0.166667 0.166667 0.083333
> > k-q = 0.250000 0.333333 0.083333
> > -P-0000
> > -P-0000 leave_new : decision taken to exit ...
> > ......................................................................
> >
>
>
> Dear Thanusit Burinprakhon
>
> The difference between k-q and k should be a reciprocal lattice vector.
> In your case the first two reduced components of k and k-q
> do not cancel each other. I think the problem is in the k-mesh
> used to generate the KSS file.
>
> Can you check if the errors shows up also if the k-mesh is specified
> using the "standard" procedure, namely using ngkpt and nshiftk and shiftk.
> You might try the values used in the GW tutorial for silicon:
>
> # Dataset2: calculation of kss file
> # Definition of k-points
>
> ngkpt2 4 4 4 (or 6 6 6)
> nshiftk2 4
> shiftk2 0.0 0.0 0.0 # This grid contains the Gamma point
> 0.0 0.5 0.5
> 0.5 0.0 0.5
> 0.5 0.5 0.0
>
> You have to run these tests only to check if abinit finds
> all the points it needs, so you can safely decrease
> ecutwfn, ecuteps and .nbands during the tests.
>
> >
> > I also have another question on the screening issue. Does Abinit need the
> > k-point
> > grid for screening calculations?
>
> Of course, the k-grid is the same as the mesh used to generate the KSS file.
> The dielectric matrix is calculated for a set of q-points that are given
> by all the possible difference between two k-points (taking into account
> symmetries)
>
> Hope this helps
> Regards
> Matteo Giantomassi