The ABINIT Users Mailing List ( CLOSED )

["Matteo Giantomassi" <Matteo.Giantomassi@uclouvain.be>]

There was a bug in the KSS generation in parallel when nsppol=2.
Is has been solved, and the bug fix will be included in 5.7.4
Meanwhile, I'm sending to you the patch.
It should be applied to 15common/outkss.F90 working on top of v5.7.3.
Other (older) versions of outkss should be modified by hand
but it should not be so difficult.

Regards
Matteo Giantomassi



> Dear abinit developers and users:
>
>     A error message kept showing up in my gw calculation, and it says:
> -------------------------------------------------------------------------
>  Opening file for KS structure output: gw_rougho_DS2_KSS
>  number of Gamma centered plane waves 3261
>  number of Gamma centered shells    354
>  number of bands   3261
>  maximum angular momentum components    4
>  number of symmetry operations        6 (without inversion)
>    *   g( 1)=  0  0  0
>    *   g( 2)= -2 -2 -2
>    *   g( 3)=  2  2  2
>    *   g( 4)= -2  0  0
>    *   g( 5)=  2  0  0
>    *   g( 6)=  0  2  0
>    *   g( 7)=  0 -2  0
>    *   g( 8)=  0  0 -2
>    *   g(3255)= 22 28 24
>    *   g(3256)= 28 22 24
>    *   g(3257)=-28-22-24
>    *   g(3258)=-24-22-28
>    *   g(3259)= 24 22 28
>    *   g(3260)= 22 24 28
>    *   g(3261)=-22-24-28
> -P-0000
> -P-0000  k-point   1   SPIN UP:
>
>  outkss: BUG -
>   The set of g vectors is inconsistent ! Check source.
>   Program does not stop but _KSS file will not be created...
>   Action : contact ABINIT group.
> ------------------------------------------------------------
>   The system is AFII type NiO, and the job is to roughly generate KSS
> file.
> That is to say, the job includes two dataset. One is usual scf, another is
> screening. I planned to do convergence study after this job and finally
> compute
> gw energy. The input file for this job is listed below. I hope someone
> could
> give me some advices.
> ------------------------------------------------------------
> ndtset         3
>
> # Dataset1: usual self-consistent ground-state calculation
> # Definition of the k-point grid
> kptopt1    1
> nkpt1  570
> ngkpt1         18 18 18
> nshiftk1   1
> shiftk1   0.5 0.5 0.5
>
> # Definition of the SCF procedure
> toldfe1  1.0d-8
> prtden1  1   # Print out density
>
> # Dataset2: calculation of kss file
> # Definition of k-points
> kptopt2  0            # K-points will be provided
> nkpt2  1              # Take only 1 k-point:
> kpt2   0.0  0.0  0.0  # the Gamma point
> istwfk2  1            # Option needed for special k-points like Gamma
> # Definition of the SCF procedure
> iscf2  -2             # Non self-consistent calculation
> getden2  -1           # Read previous density file
> tolwfr2  1.0d-10      # Still get it converged
> nband2         17
>
> # Definition of parameters for the calculation of the kss file
> nbandkss2 -1      # Number of bands in KSS file (the maximum possible)
>
> # Dataset3: Calculation of the screening (epsilon^-1 matrix)
> optdriver3  3
> getkss3     -1
> nband3            25
> ecutwfn3    6
> ecuteps3    7
> ppmfrq3    16.7 eV
>
> #definition of the unit cell
> acell  9.7351316228E+00  9.7351316228E+00  9.7351316228E+00 Bohr
> rprim  8.1659649670E-01  4.0814835635E-01  4.0814835635E-01
>        4.0814835635E-01  8.1659649670E-01  4.0814835635E-01
>        4.0814835635E-01  4.0814835635E-01  8.1659649670E-01
>
> #definition of the atom types
> ntypat        2
> znucl 28 8
> natom 4         # There are four atoms
> typat  2*1 2*2
> xred  0.0000000000E+00  0.0000000000E+00  0.0000000000E+00
>       5.0000000000E-01  5.0000000000E-01  5.0000000000E-01
>       2.5000000000E-01  2.5000000000E-01  2.5000000000E-01
>       7.5000000000E-01  7.5000000000E-01  7.5000000000E-01
> nsppol               2
> spinat          0.0000000000E+00  0.0000000000E+00 -2.0000000000E+00
>         0.0000000000E+00  0.0000000000E+00  2.0000000000E+00
>         0.0000000000E+00  0.0000000000E+00  0.0000000000E+00
>         0.0000000000E+00  0.0000000000E+00  0.0000000000E+00
> ecut 42
> # Use only symmorphic operations
> symmorphi 0
>
> nstep 100
> diemac 4.0
> -------------------------------------------------------
>
> Best regards,
> Aaron
>
--- outkss.F90	2009-04-16 01:30:10.000000000 +0200
+++ new_outkss.F90	2009-04-16 01:31:03.000000000 +0200
@@ -506,13 +506,14 @@
  allocate(nbasek(nkpt),gbasek(3,mpw,nkpt),cnormk(mpw,nkpt),gcurr(3,mpw))
  nbasek(:)=0 ; cnormk(:,:)=zero ; gbasek(:,:,:)=0
 !
-!Loop over k-points, in case of k-point parallelism each processor 
-!takes care of its own set of g-vectors. In this case, indeed, kg 
-!contains only mkmem G-spheres, where mkmem<nkpt.
+! === Find irreducible G-vectors at each k ===
+! * In case of k // each processor takes care of its own set of g-vectors. 
+!   as kg contains only mkmem G-spheres, where mkmem<nkpt.
+! * Parallelized over the cpus dealing with the spin-up component.
  do ikpt=1,nkpt
 
   if (MPI_enreg%paral_compil_kpt==1) then
-   if (MINVAL(ABS(MPI_enreg%proc_distrb(ikpt,1,1:nsppol)-MPI_enreg%me))/=0) CYCLE
+   if (MPI_enreg%proc_distrb(ikpt,1,1)/=MPI_enreg%me) CYCLE
   end if
 
   if (mkmem==0) then
@@ -535,6 +536,7 @@
  end do ! k-points
 
  if (MPI_enreg%paral_compil_kpt==1) then
+  call leave_test(MPI_enreg)
   call xsum_mpi(nbasek,spaceComm,ierr)
   call xsum_mpi(cnormk,spaceComm,ierr)
   call xsum_mpi(gbasek,spaceComm,ierr)