The ABINIT Users Mailing List ( CLOSED )

["Anglade Pierre-Matthieu" <anglade@gmail.com>]

On 5/1/06, msmajdoub@uh.edu <msmajdoub@uh.edu> wrote:

Thank you for your advice!

Do you think that these values of the following parameters are reasonable?:
# Set 1 : Initial self-consistent run
iscf1 7
kptopt1 1
tolvrs1 1.0d-18 #need excellent convergence of GS quantities for RF runs

you are working with double precision floating point numbers in abinit so
tolvrs 1e-18 is almost meaning less (unless you have some non iee754 floating point numbers on your machine).
tolvrs1 1.0d-14 is already "perfect" convergence. I think you even not need that much convergence.
 

iprcel 49 # Model dielectric preconditioner
nstep 100 # Maxiumum number of SCF iterations
iscf 7 #SCF algorithm
ecut 20.0 # Maximum kinetic energy cutoff (Hartree)

I can't tell you anything about the ecut. Just test your absolute convergence on small systems. And keep the one that looks satisfactory for you.
 

ecutsm 0.5 # Smoothing energy needed for lattice paramete

For the k-point sampling, and number of bands what should I do?
How can I increase the sampling?
What if I don't specify the number of bands (nband)?
(valence of Boron and nitrogen are 3 for each)

1) k-point: you have to choose a mesh that gives almost the same absolut convergence as the one you get with your ecut. Obviously for such a big system it will be diffcult to test many meshes. So you can concentrate on the variable kptrlen (in your output):
kptrlen= 40 is a nice convergence
kptrlen=  60 is a good convergence
kptrlen= 80 is an almost perfect k-point convergence for metals. Usualy to get a nice convergence(in kpoint) for isolator you need smaller numbers.
To optimize you mesh  you can try to use the input variable <a href=""http://www.abinit.org/Infos_v5.1/input_variables/varfil.html#prtkpt ">prtkpt</a>.

2) As this is not a metal it is possible to avoid specifying the number of bands. Yet this is not recommended if you  use the rpa dielectric matrix preconditionner (iprcel 4x). Because for this you need to have some empty bands in your system. The formula to get the good number of bands is:
~(number of electrons)/2 + (Number of ions)/4
 

 

-- 
Pierre-Matthieu Anglade