Problem with super cell and primitive cell

[Jagdish Kumar]

I have done some calculations for Al in two different structures… One is primitive cell (with one atom) and conventional cell (with four atoms).. When I plot electronic DOS for both the systems in (States/Al/Hartree) it is found that they do NOT match with each other, though they should as final properties should be independent of the cell size.. The DOS in case of primitive cell matches with other calculations but not one calculated using conventional cell.. I am attaching the results of my calculations… Please have a look into the issue…
Here are my input files…
#############################################
spacegruop.in for primitive cell..
'Fm-3m'                              : hrmg
7.653061224 7.653061224 7.653061224  : a, b, c
90.0     90.0     90.0               : bc, ac, ab
1  1  1                              : ncell
.true.                               : primcell
1                                    : nspecies
'Al'                                 : spsymb, spfname
1                                    : nwpos
0.0000    0.0000    0.0000           : wpos
##############################################
spacegroup.in for conventional cell…
'Fm-3m'                              : hrmg
7.653061224 7.653061224 7.653061224  : a, b, c
90.0     90.0     90.0               : bc, ac, ab
1  1  1                              : ncell
.false.                              : primcell
1                                    : nspecies
'Al'                                 : spsymb, spfname
1                                    : nwpos
0.0000    0.0000    0.0000           : wpos
##############################################
elk.in for primitive cell

tasks
  1
10

spinpol
.false.

! this is the relative path to the species files
sppath
  '/home/anil/src/elk-1.2.15/species/'

avec
   3.826700000       3.826700000       0.000000000   
   3.826700000       0.000000000       3.826700000   
   0.000000000       3.826700000       3.826700000   

atoms
   1                                    : nspecies
'Al.in'                                : spfname
   1                                    : natoms; atpos, bfcmt below
    0.00000000    0.00000000    0.00000000    0.00000000  0.00000000  0.00000000

radkpt
80.00

autokpt
.true.

!ngridk
! 5 5 5 

dosmsum
.true.

dos
1000 200 0
-1.0 1.5

tfibs
.false.

!vkloff
!  0.5  0.5  0.5
#################################
elk.in for conventional cell

tasks
  1
10

spinpol
.false.
! this is the relative path to the species files
sppath
  '/home/anil/src/elk-1.2.15/species/'

avec
   7.653061224       0.000000000       0.000000000   
   0.000000000       7.653061224       0.000000000   
   0.000000000       0.000000000       7.653061224   

atoms
   1                                    : nspecies
'Al.in'                                 : spfname
   4                                    : natoms; atposl, bfcmt below
    0.00000000    0.00000000    0.00000000    0.00000000  0.00000000  0.00000000
    0.00000000    0.50000000    0.50000000    0.00000000  0.00000000  0.00000000
    0.50000000    0.00000000    0.50000000    0.00000000  0.00000000  0.00000000
    0.50000000    0.50000000    0.00000000    0.00000000  0.00000000  0.00000000

radkpt
80.00

autokpt
.true.

dos
1000 200 0
-1.0 1.5

!ngridk
! 5 5 5 

tfibs
.false.

[Marty Blaber]

Dear Jagadish?

The main difference is the number of empty states needs to be greater for the conventional cell. You must include 4 times as many empty states as the default to get the same maximum energy in the DOS. i.e.

nempty
  20

(the default is 5)

There are 2 minor points:

There are 4 times as many atoms, and the DOS is not normalized to the number of atoms. So you must divide the otuput DOS by 4.

The smoothing is set to zero by default, and more states with fewer k-points makes the integration jittery.  Making rK larger helps. Increasing the number of 3 point averages helps (via the dos block variable nsmdos), and so does increasing swidth.

Cheers,
Marty

[Marty Blaber]

Sorry that picture is so big..