Bug in grid_generator.F

[claudio]

Dear developers

I found a bug in grid_generator.F when symmetries are present.

Here the correct version

best
Claudio

[Davide Sangalli]

Ciao Claudio,
I had a look to the file.
The only difference compared to the previous version is that you erased some lines.
Is that correct ?

Best,
D.

Code: Select all

diff --git a/ypp/interpolate/grid_generator.F b/ypp/interpolate/grid_generator.F
index c7b1d56..49ae4ea 100644
--- a/ypp/interpolate/grid_generator.F
+++ b/ypp/interpolate/grid_generator.F
@@ -40,8 +40,6 @@ subroutine grid_generator(USER_k,units,grid_dim)
  !
  integer  :: i1,i2,i3,ic
  character(schlen) :: ch
- real(SP), allocatable :: k_bz(:,:)
- logical :: found
  !
  if(any(grid_dim(:)<1)) call error("Wrong grid dimension")
  !
@@ -82,47 +80,8 @@ subroutine grid_generator(USER_k,units,grid_dim)
   call k_ibz2bz(USER_k,'d',.TRUE.)
   call k_ibz2bz(USER_k,'a',.TRUE.)
   !
-  ! Map in the positive BZ between [0,1)
-  !
-  where(USER_k%ptbz(:,:)<-zero_dfl)
-    USER_k%ptbz(:,:)=1._SP+USER_k%ptbz(:,:)
-  end where
-  !
-  ! Reorder in such a way to have the IBZ point at the beginning
-  ! of the list
-  !
-  allocate(k_bz(USER_k%nbz,3))
-  !
-  k_bz=USER_k%ptbz
-  !
-  do ic=1,USER_k%nibz ! Copy the IBZ points in the BZ list
-    call c2a(v_in=USER_k%pt(ic,:),v_out=USER_k%ptbz(ic,:),mode="ki2a")
-    USER_k%ptbz(ic,1:3)=USER_k%ptbz(ic,1:3)-nint(USER_k%ptbz(ic,1:3))     
-    where(USER_k%ptbz(ic,:)<-zero_dfl)
-      USER_k%ptbz(ic,:)=1._SP+USER_k%ptbz(ic,:)
-    end where
-  enddo  
-  !
-  ! Add the remaining points
-  !
-  ic=USER_k%nibz+1
-  do i1=1,USER_k%nbz 
-    found=.false.
-    i2=1
-    do while(.not.found.and.i2<=USER_k%nibz)
-      if(v_is_zero(USER_k%ptbz(i2,:)-k_bz(i1,:))) found=.true.
-      i2=i2+1
-    enddo
-    !
-    if(.not.found) then
-     USER_k%ptbz(ic,:)=k_bz(i1,:)
-     ic=ic+1
-    endif
-  enddo
-  !
   do ic=1,USER_k%nbz
     call c2a(v_in=USER_k%ptbz(ic,:),mode="ka2"//units)
   enddo
- 
   !
 end subroutine grid_generator

[claudio]

Ciao Davide

yes, it is correct. I removed a part that reorder points in the BZ.
The problem is that if you change the BZ k-points order you break the relation between IBZ and BZ points.
Probably at the beginning this subroutine was working only without symmetries and it was fine,
but with symmetries it is a mess. Interpolation is done in the IBZ and then the mapping on BZ is wrong.
But if you remove the reorder (that is useless just aesthetic) everything is fine

best
Claudio