Accuracy of TDDFT for spin-polarized systems

[guyohad]

Dear All,

According to the VASP wiki page about BSE type calculations (https://www.vasp.at/wiki/index.php/Beth ... lculations):

THDF/BSE calculations can be performed for non-spin-polarized, spin-polarized, and noncollinear cases, as well as the case with spin-orbit coupling. There is, however, one caveat. The local exchange-correlation kernel is approximated by the density-density part only. This makes predictions for spin-polarized systems less accurate than for non-spin-polarized systems.

I wonder if there is a more quantitative estimation of how less accurate TDHF calculations for spin-polarized systems are. Say, if I look at a certain peak in the imaginary part of the dielectric function at some energy, what will be the uncertainty in peak position (in eV), assuming that all parameters of the calculation are highly converged?

I tried answering this by calculating the optical absorption spectrum of ZnO with and without spin polarization, but the results were identical. I imagine that the case may be different for systems that naturally polarize their spin (e.g. antiferromagnetic systems) but since such systems cannot be described without spin polarization, it is hard to test.

I appreciate any insights about this,
Sincerely,
Guy

[alexey.tal]

Hi,

If you would like to include the local exchange-correlation kernel in your TDHF calculation, make sure to switch on the LFXC tag.
Usually, the contribution of the local xc kernel is quite small for the optical absorption of semiconductors (for example see Fig. 4 in A. Tal, Phys. Rev. Research 2, 032019 (2020)).
However, it is hard to give an estimate for a particular system. You can test the effect of the density-density xc kernel by switching on and off LFXC to get an idea of how significant the contribution of the local kernel is for your system.