tsmear question - convergence

kello · Post by **kello** » Fri Oct 14, 2011 10:53 pm

Hello:

I have been using ABINIT for a while and one of the important aspects that the tutorial suggests is to check the convergence of tsmear. Well, I did it but I don't know if the values are correct or not. I have a Ti-Pt phase with 17 atoms, triclinic structure. I found that a k-grid of 4x4x4 is good if I want to increase de size of the system, ecut=16 Ha. I create a loop to see the tsmear convergence, but the minimum energy is achieved at 0.07 tsmear (png file attached) which is not close to the values suggested by the tutorial "Default is 0.04 Ha. This should be OK for a free-electron metal like Al. For d-band metals, use 0.01 Ha."

am I doing something wrong, maybe I forgot to add an input variable?

I will greatly appreciate your help.

Karem

Code attached:

Code: Select all

# Ti4Pt3

ndtset 10
tsmear: 0.01   tsmear+ 0.01

#getwfk -1        # This is to speed up the calculation, by restarting
                  # from previous wavefunctions, transferred from the old 
                  # to the new k-points.

#Definition of occupation numbers
occopt 4          #metallic occupation of levels. Cold smearing of N. Marzari
#tsmear 0.07       #broadening of the occupation numbers

#Definition of the unit cell # triclinic
acell  15.1178 15.1178 9.448631
angdeg 88 89 120
spgroup 1

#Definition of the atom types
npsp   2           # 2 pseudopotentials
ntypat 2           # 2 type of atoms
znucl  22 78       # The keyword "znucl" refers to the atomic number of the 
                   # possible type(s) of atom. The pseudopotential(s) 
                   # mentioned in the "files" file must correspond
                   # to the type(s) of atom.
nband 200
                        
#Definition of the atoms
natom 16                # 16 atoms
typat 10*1 6*2         
xred   0.333333 0.666667 0.000 #Ti1
0.666667 0.333333 0.000
0.333333 0.666667 0.500 
0.666667 0.333333 0.500 
0.236000 0.000000 0.250
0.000000 0.236000 0.250 
0.236000 0.236000 0.750 
0.764000 0.000000 0.750 
0.000000 0.764000 0.750 
0.764000 0.764000 0.250 
0.599500 0.000000 0.250 
0.000000 0.599500 0.250 
0.599500 0.599500 0.750 
0.400500 0.000000 0.750 
0.000000 0.400500 0.750 
0.400500 0.400500 0.250 

#Definition of the k-point grid
ngkpt   4 4 4       # Definition of the different grids


#Exchange-correlation functional
ixc 1             # LDA Teter Pade parametrization

#Definition of the planewave basis set
ecut  16         # Maximal kinetic energy cut-off, in Hartree

#Definition of the SCF procedure
nstep  300       # Maximal number of SCF cycles
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) 
                 # This value is way too large for most realistic studies of materials

david.waroquiers · Post by **david.waroquiers** » Sat Oct 15, 2011 11:21 am

Hello,

In fact the combined tsmear/kpoints convergence is used in a different way. You do not want to find the tsmear that minimize the energy. You want to describe your system as close as possible to the fully converged one within a reasonable amount of time. For this you have to increase your k-point grid. The problem is that for metals, as some states are partially occupied, if the electronic distribution function is a step function, you will need MANY k-points to get something reasonably converged. The trick is to add a smearing (with the variable tsmear) to the distribution function so that you can reach convergence with respect to k-points more rapidly.

In brief, you should try to have your k-point grid (a very dense one) converged for a very low tsmear (fixed, something like 0.001 Ha or less) (which would correspond to the "real" step function case). This system is the "reference" one. Then you should compare the convergence of the k-point grid with other values of tsmear (higher tsmear : 0.01 0.02 0.04 ....). At k-point convergence, the properties with this tsmear should be within some tolerance of the properties obtained with the very low tsmear.

Usually you do this convergence study for your "test" case, for example the pure system. Then you use the k-point/tsmear parameters you found to use it for further studies (if you want to do GW calculations for example, or defects in a supercell -- in this case if you increase the size of the supercell by 2 in each direction for example, do not forget that the k-point grid can be decreased by 2 in each direction).

I hope it helps you

David

kello · Post by **kello** » Sat Oct 15, 2011 11:02 pm

David:

Thank you very much for your answer. Now, I understand better how the convergence study has to be done.

Karem

ABINIT Discussion Forums

tsmear question - convergence

tsmear question - convergence

Re: tsmear question - convergence

Re: tsmear question - convergence