mBSE with metallic systems

[hong_tang1]

Dear VASP team,

Does the model BSE (mBSE ) "ALGO = TDHF ; IBSE = 0 ; LHARTREE = .True. ;LADDER = .TRUE. ; LFXC = .True. ; LMODELHF = .TRUE. " work for metallic systems in the most recent version of VASP?

Thanks!

Best,
Hong

[manuel_engel1]

Dear Hong,

Thank you for posting on the VASP forum. BSE calculations (model or not) are not well defined for metallic systems. I am not sure what happens if you still run the calculation but it's physically not very meaningful. I hope that I was able to answer your question.

[pedro_melo]

Dear hong_tang1,

I will add to what my colleague wrote.

In principle yes, you can use the model BSE on a metallic system. However, there are several points you should consider before proceeding with the calculation:

1 - The model dielectric function is only valid for cubic systems. If you wish to use a system with a different symmetry, it will not provide you with the right results.
2 - Please make sure that the HFSCREEN and AEXX are well suited for your material. We provide some values for HFSCREEN in our wiki, while AEXX can be extracted from the dielectric function.
3 - We advise you to use the new driver, which you can choose with IBSE=2. Other solvers are available, so please check the wiki and choose what is most convenient for your case.
4 - You will likely need a very dense k-point grid, since the bands of interest are near the Fermi level.

Finally, and likely the most important, for metals you need to provide the imaginary frequency that is added to the Drude contribution to the dielectric function. This is done with WPLASMAI. Otherwise results will be unphysical.

I suggest that you first try to use an LOPTICS calculation. This will give you the optical absorption at the level of non-interacting electrons and holes. If this is already good enough when compared to experimental data (it that exists), it means that excitonic effects are not needed.

Kind regards,
Pedro

[hong_tang1]

Dear Pedro and Manuel,

Thanks so much for helping!

1. As I understand, for metals, in mBSE calculations, I use "KPOINT_BSE = 13 0 0 0 ", where 13 indicates the 13th q point for the mBSE calculation. So, since the momentum transfer q is not zero, all the intra-band and inter-band transitions are included, so, there is no need to using plasmon model or Drudal term. Is this correct?

2. for metals, in mBSE calculations, I use "KPOINT_BSE = 1 0 0 0 ", where 1 indicates the Gamma point (usually I use Gamma centered zero shift k-grid). So, the Drudal term has to be included in order to account for the intra-band effects. Please let me know how to input the plasmon frequency in this case. How to determine the plasmon frequency for a metallic system? VASP wiki for the WPLASMAI is very limited information. Please give us more information on this.

3. The other question is related to the model dielectric function used in mBSE. I can use ALGO = CHI to calculate the dielectric function vs. q. The calculated 1/epsilon are listed in the vasprun.xml file. My question is for each q, it only has one number of 1/epsilon. epsilon is a tensor, so, how to get just one number from a tensor, which has several elements. What kind of "average" is done here? Plese help.

Appreciate your help very much!

Best,
Hong