Kerr rotation spectrum

[MarkusMeinert]

Dear users and developers,

I try to compute Kerr rotation spectra of cubic materials using the optical routines in vasp. Naturally, I need to include spin-orbit coupling to find non-zero off-diagonal elements of the dielectric tensor. However, it seems that I also have to suppress symmetry to obtain the off-diagonal elements more or less properly. Here's my INCAR for fcc Nickel.

Code: Select all

SYSTEM = Ni fcc

# 1. Ground State
   ENCUT   = 400.00 eV  # energy cut-off for the calculation
   PREC    = Normal     # Normal precision
   ISPIN   = 2
   NPAR    = 4
   NEDOS   = 2000
   LSORBIT = .TRUE.
   SAXIS   = 0.0 0.0 1.0
   MAGMOM  = 0.0 0.0 1.0
   LORBIT  = 11

   ISMEAR  = -5

   ISYM    = 0
   NBANDS  = 40

# 2. exact diagonalization of virtual orbitals

#   ALGO    = Exact

# 3. optics calculation: independent particles, no LF


#   IALGO   = 2          # read wavefunctions and energies
#   CSHIFT  = 0.5
#   LOPTICS = .TRUE.

Of course, due to the suppressed symmetry the calculation slows down tremensously, in particular for the exact diagonalization step to obtain the virtual orbitals properly.

My question is: Is there a smarter way to do this?

Regards
Markus Meinert

[admin]

Unfortunately the slowing down is the most severe for cubic materials.
May be you can save the computer time by using a smaller k-point set
or setting the parallelization over k-points?

[MarkusMeinert]

Actually, I figured out that the off-diagonals of epsilon are still nonsense, basically just broadened numerical noise. It seems, something fundamental is missing. Is VASP capable of calculating the off-diagonals of epsilon at all? I don't find anything on this problem in the literature. All Kerr spectra I've found were calculated either with an FLAPW code or YAMBO.

[admin]

Vasp calculates off-diagonals of epsilon. The calculation of the Kerr rotation angle is not implemented.