Increasing Density of Dielectric Function

[aruth]

Hello I am calculating the dielectric function according to: https://cms.mpi.univie.ac.at/wiki/index.php/LOPTICS

On that page it appears that the energy step between the calculated dielectric function is based upon the energy step of the DOS which is set by NEDOS. I noticed that even though my default DOS had 301 points, my dielectric function had 200 points.

I tried increasing NEDOS to 2000 points but the dielectric function is still too coarse (200 points). Is there a way to decrease the energy step in the dielectric function? Is the dielectric function printed anywhere besides OUTCAR?

My INCAR:

Code: Select all

general:
System = MoSx2005U9
ISMEAR = 0 ! default 1: 0 GAUSSIAN, N>0 method of Methfessel-Paxton order N.
LORBIT= 10
NSW    = 0 !Number of ionic steps
LOPTICS= .TRUE.
NPAR= 4 !number of cores per band
LDAU = .TRUE.
LDAUTYPE=2 !This is the default
LDAUL = 2 -1 -1 -1 -1 -1
LDAUU = 9 0 0 0 0 0
LDAUJ = 0 0 0 0 0 0
LDAUPRINT = 2
LASPH = .TRUE.!Non-spherical PAW corrections
NBANDS = 1000 !Suggested by VASP wiki
NEDOS = 2000

[admin]

The dielectric function is written by
WRITE I,NEDOS,ISTEP
ISTEP=MIN(10,NEDOS/40)
I.e. when using NEDOS=2000 each 10th calcualted value is written producing
200 lines in OUTCAR
When using NEDOS=3000 you will get 300 lines.

[call_me_Al]

regarding this topic, can someone clarify the dependancy of the LOPTICS on the DOS? Does the LOPTICS routine actually use the DOSCAR, or does it integrate over the BZ itself with whatever smearing you have? Or does it just use the information in the WAVECAR and do the integration over the BZ with the ISMEAR and NEDOS you have defined?

thanks!

[gbal]

Hi,

You can take the dielectric function with number_of_lines=NEDOS from the vasprun.xml file.

The smearing process does affect the optics output not just by being a different method -- there are different routines for gaussian or other integration.