optical properties of metallic systems

[IBRAHIM]

Dear All

I would like to calculate the optical properties of metallic clusters and nanoparticles using the independent-particle approximation, TDDFT+U, and TD-hybrid functionals.

1- For the independent-particle approximation, is it possible to use LOPTICS = .TRUE. for metallic systems?

2- For time-dependent DFT+U and hybrid functionals, according to the VASP wiki, the TD-hybrid functional calculation involves two steps:

First step: ground-state calculation

Second step: optical absorption calculation

- For metallic systems, should LOPTICS = .TRUE. be used in the first step? or avoided as in GW calculations.
- Also, in the second step, what about NBANDSO(V) for metallic systems?
- Can I use the same procedure for TDDFT+U as well, or should there be some changes?

Many thanks in advance,
Ibrahim

[alexey.tal]

Dear Ibrahim,

Thank you for your question.
If your system has a band gap you can use the Casida TDDFT and TDHF algorithms in VASP (ALGO=TDHF see wiki). However, if you have partial occupation of states this algorithm should not be used as it doesn't account properly for the intraband transitions.

As far as I know, small nanoclusters (up to several hundred atoms) have a finite band gap, so they don't have partially occupied states. If this is the case you can use the TDHF algorithm for IP, TDDFT, and TDHF calculations. You can do a TDDFT+U calculation, but the Hubbard U will only be included in the ground-state calculation, i.e., only the PBE xc kernel will be included in the response-function calculation.

The most efficient way to calculation the IP spectrum is via LOPTICS. This approach also allows to account for the intraband transitions.
The dielectric function in metallic systems, i.e., systems with no band gap, is dominated by the intraband transitions (or Drude term) in the low-frequency range. This term can be included in the dielectric function with LOPTICS via the WPLASMAI tag.

Do you actually have a band gap in your nanoclusters?

Best,
Alexey

[IBRAHIM]

Dear Alexey,

Thank you for the detailed reply.

Yes, I observe a band gap in some systems; however, partial occupations appear in others, which I will try to avoid by adjusting the SIGMA value.

Regarding the IP approximation, As mentioned that LOPTICS can be used for metallic systems, and the Drude term can be included via the WPLASMAI tag.
- Could you please explain how to determine the appropriate value to use for this tag?
- Also, one more question:
Can VASP account for solvent effects when calculating optical properties using the solvation model of VASPsol?

Best regards,
Ibrahim

[alexey.tal]

Yes, I observe a band gap in some systems; however, partial occupations appear in others, which I will try to avoid by adjusting the SIGMA value.

That is a good idea. The gap might be rather small and in order to avoid partial occupations one should use a small smearing parameter, e.g., SIGMA=0.01.

WPLASMAI accounts for the scattering effects or relaxation time, thus it is usually determined by fitting the spectrum to experiment. The values of WPLASMAI can vary typically between 0.1 and 0.7 eV, depending on the material. You can see the values used previously for Fe, Co and Ni: Delin et al., Phys. Rev. B, 14105-14114 (1999).

Can VASP account for solvent effects when calculating optical properties using the solvation model of VASPsol?

I don't have experience using such models. However, the implicit effect should be present as the solvent influences the electronic structure, which will be reflected in the calculated dielectric function. Whether such effects play a major role I don't know. More importantly, the absorption by the solvent molecules will not be included in the dielectric function in such a model.

[IBRAHIM]

Thank you very much