smearing tuning

[vunazarov]

Dear All,
We are trying to calculate frequency-dependent dielectric function for a system with the difference between HOMO and LUMO of 0.12 eV (obtained after bands were plotted).
For the imaginary part of the dielectric function, we obtained notable absorption below 0.12 eV. As we understand, VASP does not take into account exciton effects
when ALGO=normal, and absorption in the band gap arises due to the smearing. Therefore, we decreased SIGMA from 0.02 to 0.01, 0.005 and 0.001, but did not obtain significant difference in absorption at low energies.
What are we doing wrong?

A related question: can the smearing for the ground-state calculation and that for optics be tuned separately ?

Thank you very much in advance.

Best regards,

Vladimir Nazarov.

INCAR file:

Embedded_nc_bFeSi2_1nm
ICHARG = 0
ISMEAR = -1
SIGMA = 0.01
ALGO = normal
NSW = 0
PREC = Accurate
ISPIN = 2
IBRION = 2

LREAL = .TRUE.
NELM = 40
NPAR = 16

LASPH = .TRUE.

LOPTICS = .TRUE.

[admin]

please do a GW-calculation, following the instructions given in the manual.

[vunazarov]

Thank you for the response.

There are two points to it:

1. We cannot do the GW calculation technically because of our huge system (about 1000 atoms per unit cell).

2. Irrespective to GW, we would like to understand how smearing works in VASP for DFT (not for GW this time). In particular, whether the same smearing is used for the ground-state and optics calculations
and why in our calculation the decrease of the smearing does not affect the absorption in the band gap, which is unphysical.

Thank you very much for looking into this.

Best regards,
Vladimir Nazarov.