The ABINIT Users Mailing List ( CLOSED )

[Quan Phung Manh <manhquank9@gmail.com>]

Dear Adam.
I'm wondering about your nstep=30. Is it too small?
With your huge system, it would take much more time to converge the SCF cycle. 
If you use the same number of k-points. I think it would take you about: 64^3/8^3=512 days ??? to optimize your supercell (Am I right?).

2009/5/21 Adam Sorini <asorini@gmail.com>

Thank you for your quick response. If I understand correctly, you are saying that to keep the same density of k-points I should change the value from my undoped cell (acell=x) input file

ngkpt 6 6 6

to a value for doped supercell (acell=2*x) input file

ngkpt 3 3 3

?

Also, I have included my doped cell input file for reference. Any comments/criticism for improvement of the input file would be much appreciated. The parameters such as ecut and ngkpt were obtained by converging the eight atom unit cell calculation w.r.t. the total energy. The pseudopotentials I used are pspnc files available on the ABINIT website (I also attempted to use HGH pseudopotential with fewer electrons, but I was not able to see convergence using those other psps). Thank you in advance for useful comments. Cheers,

Adam

# Hg1201: supercell optimization with a dopant oxygen

prtdos 1 # print the DOS

#Definition of occupation numbers
occopt 4 # metallic occupation w "cold smearing"
tsmear 0.05 # smearing parameter, not really temperature at all
#fband 0.5 #can change nband here, will give more unoccbands
# needed to increase from default get rid of warnings
fband 0.8

#Definition of the unit cell in Bohr,
#
#acell 7.31627 7.31627 17.9505       
acell 14.63254 14.63254 35.901      

rprim  1.0  0.0  0.0   # scaled by acell
       0.0  1.0  0.0
       0.0  0.0  1.0

#Optimization of the reduced coords and cell volume
optcell 1 # fix the cell size
ionmov  2 # an algorithm for m.d. of red coords
ntime  20 # number of times steps for m.d.
dilatmx 1.05 # not needed if optcell=0
ecutsm  0.5 # not needed if optcell=0

#Definition of the atom types
ntypat 4          # There are 4 types of atom, Hg, Ba, Cu, O
znucl 80 56 29 8  # The keyword "znucl" refers to the atomic number of the
                  # possible types of atom. The pseudopotentials
                  # mentioned in the "files" file must correspond
                  # to the types of atom. Here, types are
                  # Hg=80, Ba=56, Cu=29, O=8
                        

#Definition of the atoms
#natom 64        
natom 65           # There 65=64+1 dopant atoms per cell
# dopant number 65, oxygen
#natom 8
typat 1 4 2 3 4 4 2 4  1 4 2 3 4 4 2 4  1 4 2 3 4 4 2 4  1 4 2 3 4 4 2 4  1 4 2 3 4 4 2 4  1 4 2 3 4 4 2 4  1 4 2 3 4 4 2 4  1 4 2 3 4 4 2 4 4 
#specify types like 1 2 3 4 as above, whatever.

xcart              # This keyword indicate that the location of the atoms
   0.0000000       0.0000000       0.0000000   
   0.0000000       0.0000000       3.7480645   
   3.6581349       3.6581349       5.3259134   
   0.0000000       0.0000000       8.9752502   
   3.6581349       0.0000000       8.9752502   
   0.0000000       3.6581349       8.9752502   
   3.6581349       3.6581349       12.624587   
   0.0000000       0.0000000       14.202435   
   7.3162699       0.0000000       0.0000000   
   7.3162699       0.0000000       3.7480645   
   10.974405       3.6581349       5.3259134   
   7.3162699       0.0000000       8.9752502   
   10.974405       0.0000000       8.9752502   
   7.3162699       3.6581349       8.9752502   
   10.974405       3.6581349       12.624587   
   7.3162699       0.0000000       14.202435   
   0.0000000       7.3162699       0.0000000   
   0.0000000       7.3162699       3.7480645   
   3.6581349       10.974405       5.3259134   
   0.0000000       7.3162699       8.9752502   
   3.6581349       7.3162699       8.9752502   
   0.0000000       10.974405       8.9752502   
   3.6581349       10.974405       12.624587   
   0.0000000       7.3162699       14.202435   
   7.3162699       7.3162699       0.0000000   
   7.3162699       7.3162699       3.7480645   
   10.974405       10.974405       5.3259134   
   7.3162699       7.3162699       8.9752502   
   10.974405       7.3162699       8.9752502   
   7.3162699       10.974405       8.9752502   
   10.974405       10.974405       12.624587   
   7.3162699       7.3162699       14.202435   
   0.0000000       0.0000000       17.950500   
   0.0000000       0.0000000       21.698565   
   3.6581349       3.6581349       23.276413   
   0.0000000       0.0000000       26.925751   
   3.6581349       0.0000000       26.925751   
   0.0000000       3.6581349       26.925751   
   3.6581349       3.6581349       30.575089   
   0.0000000       0.0000000       32.152935   
   7.3162699       0.0000000       17.950500   
   7.3162699       0.0000000       21.698565   
   10.974405       3.6581349       23.276413   
   7.3162699       0.0000000       26.925751   
   10.974405       0.0000000       26.925751   
   7.3162699       3.6581349       26.925751   
   10.974405       3.6581349       30.575089   
   7.3162699       0.0000000       32.152935   
   0.0000000       7.3162699       17.950500   
   0.0000000       7.3162699       21.698565   
   3.6581349       10.974405       23.276413   
   0.0000000       7.3162699       26.925751   
   3.6581349       7.3162699       26.925751   
   0.0000000       10.974405       26.925751   
   3.6581349       10.974405       30.575089   
   0.0000000       7.3162699       32.152935   
   7.3162699       7.3162699       17.950500   
   7.3162699       7.3162699       21.698565   
   10.974405       10.974405       23.276413   
   7.3162699       7.3162699       26.925751   
   10.974405       7.3162699       26.925751   
   7.3162699       10.974405       26.925751   
   10.974405       10.974405       30.575089   
   7.3162699       7.3162699       32.152935   
   3.658135        3.658135        0.0
#Definition of the planewave basis set
ecut 50
#ecut 1.0

#Exchange-correlation functional
ixc 1             # LDA Teter Pade parametrization
                  # maybe use gga instead... whatever
#Definition of the k-point grid
kptopt 1          # Option for the automatic generation of k points, taking
     
#ngkpt 1 1 1            # into account the symmetry
ngkpt 6 6 6       # This was used for single cell, Aldo says can reduce it...
#ngkpt ? ? ?       # 3 3 3?
nshiftk 1         # of the reciprocal space. For a FCC real space lattice,
                  # usually us nshiftk 4, but for tetragonal use 1.
                  # one shift is used :
shiftk 0.5 0.5 0.5

#Definition of the SCF procedure
#nstep 0          # Maximal number of SCF cycles
nstep 30          # Maximal number of SCF cycles, 30 is default
toldfe 1.0d-6     # Will stop when, twice in a row, the difference
                  # between two consecutive evaluations of total energy
                  # differ by less than toldfe (in Hartree)

On Wed, May 20, 2009 at 12:31 PM, Aldo Humberto Romero <aromero@qro.cinvestav.mx> wrote:

Hi Adam

it is very hard to answer your question without having more info
but at least I can tell you that by increasing the unit cell you
need to decrease the number of K-points accordingly.

regards

-aldo.


-> Hi forum,

->
->  I have a question regarding the length of time a certain supercell
-> calculation
-> will take that I was hoping someone could help me with. For starters, I
-> used
-> abinis to optimize the geometry of a unit cell (acell 7.3 7.3 17.9)
-> containing
-> eight atoms, and the calculation took about a day. Next, I created a
-> supercell
-> with twice the acell value and 65 (64 plus one dopant) atoms. This
-> calculation
-> is still running and I was wondering how long it will take, and/or how to
-> estimate in general the scaling with the number of atoms in a supercell
-> calculation. Both calculations have the same fixed ecut=50, and 18 kpts. I
-> can
-> also send the input files if that would be helpful. Thank you very much in
-> advance. Cheers,
->
-> Adam

->


--

Prof. Aldo Humberto Romero
CINVESTAV-Unidad Queretaro
Libramiento Norponiente 2000
CP 76230, Queretaro, QRO, Mexico
tel: 442 211 9909
fax: 442 211 9938

email: aromero@qro.cinvestav.mx
       aldorome@gmail.com
www: qro.cinvestav.mx/~aromero