correct use of FERWE and LOPTICS

[call_me_Al]

hi All,

Does anyone know the correct recipe for using your own occupancies using FERWE and the calculating the optical properties? I have tried several approaches - standard DFT, FERWE with ISTART = 1, and the ALGO = exact with LOPTICS. in three steps. but everything seems to give me a nonsense spectrum.

what is the correct way to use FERWE with LOPTICS?

thanks for your input
Al

[call_me_Al]

just to clarify, if I do the groundstate with standarad smearing methods and then LOPTICS the spectrum it looks fine,

but if I use FERWE with the same occupations as the standard DFT for every K-point, and run LOPTICS it's a completely different (non realistic) spectrum

[call_me_Al]

so after a lot of playing around, its seems whenever you define your own occupations via FERWE and ISMEAR = -2, the LOPTICS run seems to give an erroneous spectrum.

Using:

ALGO = Exact
NBANDS = 128
EDIFF = 1e-8

ISMEAR = -2;
FERWE = "the same occupancies as the groundstate with small smearing"

LOPTICS = .TRUE.

gives nothing like the spectrum using standard smearing methods

can this be done or am I missing an obvious step?

any input appreciated
Al

[call_me_Al]

update:

I think I have found a cheap workaround for this issue.

Just to clarify what I am trying to achieve: I want to use the groundstate orbitals of standard DFT (including higher energy orbitals with ALGO = EXACT) and then calculate the optical properties with a specific occupancy of these states. In other words what will the spectrum look like if I move an electron from one state to another. The states will be the ones defined by the groundstate charge density, and wont be self consistent with the smearing, I just want to see the effect of Pauli blocking on the optical properties.

So, with this in mind if I:
1) take the standard DFT (with ALGO = EXACT) and standard 'cold' smearing (say ISMEAR = 0; SIGMA = 0.2)
2) then do the LOPTICS = .TRUE. run with a smearing of my choice but with NELM = 0 it seems to work

this is basically stopping any attempts to self-consistently converge the density and corresponding WAVECAR, it just uses the occupancies in the groundstate one electron wavefunctions.

does this seem in any way reasonable, or is it just plain incorrect?

thanks
Al