How to calculate dielectric function in microwave (GhZ) range

[askhetan]

Dear VASP team and community,

I tried this this tutorial for calculating the ionic contributions to the frequency dependent dielectric function of NaCl:
https://www.vasp.at/wiki/index.php/Ioni ... on_of_NaCl

.. and things work as expected. I was wondering if you could please tell me what keyword modifications I can make to calculate the dielectric properties (especially the imaginary part) on much finer frequency grid and in the lower range. For e.g. the shown graph (diel.ion.dat) has values of the real and imaginary part in the Thz range and I am interested in the values in the microwave range i.e. 300 MHz to 300 GHz. The grid spacing currently is 0.05 ThZ which is too high for my intended purpose. Any information on how to proceed will be very helpful.

Thanks in advance

[martin.schlipf]

The relevant flags to control the frequency grid are NEDOS and OMEGAMAX. Note that this grid is for the integrals in the Kramers-Kronig transformation (see LOPTICS). So be careful that integral is still correct when you change the boundaries.

[askhetan]

Thanks for your reply. I had the time to execute a combination of different keywords regarding NEDOS, OMEGAMAX, EDIFF, KPOINTS, etc.
No matter what NEDOS (2000, 5000 or 20000) I use, the number of grid points in the ionic contribution has always 2000 points and there is no change in the imag.dat file that contains the imaginary part of the dielectric response as a function of frequency. Notably, the gridspacing varies only because i changed EDIFF from 10E-05 to 10E-08.

I have attached a zip folder, which contains folders for two cases (EDIFF-10E-5 and EDIFF-10E-8), in each folder there are INCAR_NEDOS, OUTCAR_NEDOS and imag_NEDOS files where NEDOS is either 2000 or 5000. OMEGAMAX = 0.00124 eV (=300 MHz is upper limit of Microwave range) in all cases.

So the question is again - is it possible to change the grid spacing and maximum frequency such that the ionic contribution to the dielectric response can be calculated in the microvave range (f = 1.24 micro-eV to 1.24 milli-eV ). ?
Thanks very much for your help in advance.

[martin.schlipf]

Sorry, I had missed that you were interested in the ionic dielectric function. I checked the code and the grid is indeed hard coded in the current version of VASP. So the only option to change the grid is to go into the source code (finite_diff.F, subroutine: EPSILON_ION_OMEGA) and change NOMEGA and DOMEGA and recompile. Unfortunately you need to recompile the code for every different setting you choose.

[askhetan]

thank you very much! that is already very helpful to know. I will get back after doing this and if I am successful, I'll let the community know here. Perhaps this can be fixed in a future version because understanding dielectric response in microwave range is becoming extremely important for catalysis and I believe many people will benefit from this.