Dielectric function beyond IPA

[atul_pandey]

Dear Sir,

Which tag should be used instead of LOPTICS = .TRUE. for the calculation of frequency-dependent dielectric function for semiconductor and metallic materials using GW or hybrid functional? There is mention to use this tag for insulator only. I have attached the screenshot below.

[martin.schlipf]

For GW, you do not need to set the LOPTICS tag to get the dielectric function. I would recommend against using hybrid functionals for metals because their Hartree-Fock contribution tends to open a gap.

If you find that LOPTICS does not work, we have a couple of other methods that you can use to compute dielectric properties.

[atul_pandey]

It means we should not use hybrid functional for metal in case of frequency-dependent dielectric function.
Can I use the above tag for semiconductor material?

System = SiC unoccupied orbitals
NBANDS = 512 ! maybe even larger
ALGO = Exact
NELM = 1 ! since we are already converged stop after one step
ISMEAR = 0 ; SIGMA = 0.05
LHFCALC = .TRUE. ; AEXX = 0.25 ; HFSCREEN = 0.3
LOPTICS = .TRUE. # for insulators

[martin.schlipf]

There may be some exceptions that I am not aware of but in most cases you should not use hybrid functionals for metals. The physical intuition is that for long distances the HF fraction should approach 1/eps, i.e. vanishing for metals. Perhaps with dielectric dependent you could also describe metals but hybrids are also 10-100x more expensive so if you can describe them well with local functionals you strongly prefer using them.

To the second part of the question. Yes, I believe hybrid functionals with LOPTICS should work for semiconductors and insulators. Please let me know if you run into any problems. Just keep in mind the much larger computational cost compared to local functionals.