My Community

Hi there,

I would like to run some TD-DDH calculations, but as a first step want to fit the DDH parameters to my materials using the dielectric function (as was done in this paper DOI: 10.1103/PhysRevMaterials.2.073803). Is the only method to compute the dielectric function via RPA, or can I get this function from just running a calculation with LOPTICS=.TRUE.? My understanding is I need to plot the dielectric constant as a function of G, but the output of LOPTICS=.TRUE. gives me the dielectric as a function of energy. Perhaps there is something obvious I am overlooking, or I have a big misunderstanding. This is not exactly my area of expertise, but I guess I am wondering if I can get an approximation of these parameters without running an expensive RPA calculation, as was done in the TD-DDH tutorial posted on the VASP wiki.

Thanks for any help or guidance!
Jake

Dear Jake,

The answer to your question depends on the approximation to the dielectric function that you want to take. If you use LOPTICS=.TRUE. during a ground-state calculation with extra, empty states, you will obtain the independent-particle macroscopic dielectric function, which is independent of G and has no electron-hole interaction taken into account. To perform a TD-DDH calculation you will need to provide VASP with \epsilon^{-1}_\infty and \mu parameters in order to model the G-dependent dielectric function (see LMODELHF for more).

Keep in mind that the model only includes the diagonal components of the dielectric function, which is normally a tensor in (G,G'). So there are no elements with G different from G'.

In our page for LMODELHF we provide some information of the value to use for a list of materials. Of course, if you have a reference value of those parameters for the material you want to study you can just use them in the INCAR.

Kind regards,
Pedro

Dear Pedro,

Thank you for your detailed reply, this has helped me to identify some gaps in my understanding. I have also come across some methods of parameterizing these functional parameters using DFPT to obtain the macroscopic dielectric constant (and thereby determine AEXX), and using an effective Thomas-Fermi screening parameter to approximate \mu (HFSCREEN) (DOI: 10.1038/s41524-022-00869-6). I guess this bypasses the requirement for an RPA calculation (which I understand now is necessary to parameterize according to the methodology in the previous paper I posted), but presents other issues, particularly for small-gap semiconductors.

Have a nice day!
Jake