questions about CHGCAR from TD-DFT

[awang95]

Hi,

I'm interested in getting the excited state charge density for various systems, but am confused whether the CHGCAR that I get from calculations with LOPTICS = TRUE and subsequently ALGO = TDHF actually corresponds to the charge density of the excited state. I've been getting isosurfaces that look pretty much identical to the ground state static calculation, which makes me wonder if charge density outputs are simply the ones corresponding to the ground state. If so, is there any way to get the charge density for a particular excited (i.e. S1, S2) state? For example, the charge density (and also total system energy if possible) when the electron is excited to the LUMO, LUMO + n...etc. Attached are the 2 INCAR files I'm using.

I've tried doing a static calculation using ISMEAR = -2 with FERWE/FERDO tags and LDIAG = FALSE to "fix" electron occupancies in a higher level orbital, but had the validity of the method questioned. My understanding is that it's like delta-SCF - is that true? Also, would there be any papers that you could share with respect to the usage or documentation of ISMEAR = -2 combined with FERWE/FERDO? Alternatively, would it work if I simply do the same thing but with a non-SCF calculation (i.e. NELM = 1) to get the charge density?

My goal in the end is to obtain the charge density difference between ground and excited states. Ideally, I would also like to get transition densities but not really sure if it's even possible - if there are any pointers for that it would also be appreciated. I have been able to get absorption spectra and oscillator strengths, but I hope to go deeper into individual transitions and analyze how the electron distribution shifts for specific excitations. Would there be any tags in the INCAR (including hidden features) that could accomplish that?

thanks!

[henrique_miranda]

OK, I think you asked a related question on another thread:
https://www.vasp.at/forum/viewtopic.php?f=4&t=19263

The CHGCAR you get from LOPTICS=.TRUE. is the exact same one that you get from the ground-state calculations.
This tag simply activated the computation of the optical spectra as a sum-over state (see documentation and references therein):
https://www.vasp.at/wiki/index.php/LOPTICS

If you want to get a CHGCAR for the excited state then you need to set ISMEAR=-2 and FERWE/FERDO.
The results you get would be like delta-SCF. I don't know which paper I could suggest to you related to this topic.
Note that to do non-SCF calculation you should set ICHARG=11. By setting NELM=1 you tell VASP to do a single electronic minimization which might lead to an unconverged calculation.

As for your last question, my suggestion would be the same as above: try using ISMEAR=-2 and FERWE/FERDO.
I don't know what other tags could help.

[awang95]

ok that makes sense; thanks a lot!