Resolution of dielectric function in TDHF

[tomasz.w]

How to control the energy resolution of dielectric function in TDHF calculations? I'm always getting 1000 values, regardless of NEDOS, EMIN, EMAX set in the INCAR.

[alexey.tal]

Dear tomasz.w,

Thank you for your question.

If you are using ALGO=TDHF with the default solver, i.e., the exact diagonalization algorithm, the NEDOS should set the number of grid points in the dielectric function. Could you please provide your input and output files as explained in the guidelines?

Best wishes,
Alexey

[tomasz.w]

Dear Alexey,

Thanks for your answer. Files attached.

Best,
Tomasz

[alexey.tal]

Dear Tomasz,

I see that you set NEDOS=1000 and you get exactly 1000 values in the dielectric function in vasprun.xml, which is the expected result. If you would like to limit the maximum energy in the dielectric function you can set OMEGAMAX. Be aware that OMEGAMAX also excludes the transitions beyond a given energy in the BSE kernel. The minimum energy in the dielectric function is always 0.

[tomasz.w]

Dear Alexey,

I also got 1000 values of dielectric function when I did not set NEDOS (in that case NEDOS=301 was reported in OUTCAR) - see attachment.

Thanks for your hint with OMEGAMAX.

Is there a fast way to recalculate just the dielectric function, without re-doing the heavy diagonalization of BSE matrix?

Best,
Tomasz

[alexey.tal]

You are right. It is is not documented on the wiki, but in BSE/TDHF the minimum number of grid points in the dielectric function is 1000, so you can only increase the number of grid points.

Is there a fast way to recalculate just the dielectric function, without re-doing the heavy diagonalization of BSE matrix?

Unfortunately, it is not possible to recalculate the dielectric function without redoing the BSE calculation.
But in VASP 6.5.0 it is possible to use the Lanzcos iterative diagonalization algorithm by setting IBSE=3, which is a lot faster than the exact diagonalization.

Also, you can find more tips on how to optimize a BSE calculation in VASP 6.5.0 here.

[tomasz.w]

Dear Alexey,

Thanks for your hints. Indeed, I was able to increase the resolution by increasing NEDOS.

Unfortunately, I haven't got access to version 6.5 yet, so I'm still using 6.4.3.

Another question (let me know if I should start a new topic instead):
I'm comparing the same system in FM and AFM phases (ISPIN=2, no spin-orbit, no mixing of spin channels). Identical POSCAR, KPOINTS, POTCAR, INCAR - changing MAGMOM only (as attached previously). PBE, model hybrid (DSH) and TDHF calculations converged; bandgaps and magnetic moments are correct. BSE spectrum (imaginary part of epsilon) of AFM looks fine (similar to literature), but for FM it's wrong, comparing to AFM and literature (attached below)
1) Intensities are too low
2) Too many peaks
3) Finite off-diagonal components of epsilon emerged in TDHF (they are zero at DSH level)

Looks like symmetry breaking. Any ideas?

Best,
Tomasz

[alexey.tal]

Dear Tomasz,

Could you please include the output files for these calculations?

Best wishes,
Alexey

[tomasz.w]

Files attached below

[alexey.tal]

Thank you for providing the files.
Could you also show the results of the ground state calculation (input and output files and stdout)? So far, I don't really see anything wrong with the input.

[tomasz.w]

Files attached below