TDDFT Core states
Just Got Here
5:44:27 AM PDT - Wed, Apr 4th 2012
Dear Community,

i'd like to calculate approximate xray-pre-edge absorption. I would like to freeze some of the orbitals, so that only core orbitals of interest are allowed to be excited to the manifold of virtuals. With FREEZE i cannot freeze the upper occupuied orbitals., however.

is there a possibility to do this, other than described in the documentation in the FREEZE Section of TDDFT ?
Is there an explicit Core-Excited-State Module ... the list of new features statet so ..

Thank you very much in advance for your answer.
Best regards
Alexander Kulesza

Forum Regular
10:17:39 PM PDT - Wed, Apr 4th 2012
Hi Alexander,

This can be done with the ecut keyword in the TDDFT block. This keyword enables restricted excitation window TDDFT (REW-TDDFT). This is an approach best suited for core excitations. By specifying this keyword only excitations from occupied states below the energy cutoff will be considered.

For example,

echo
start h2o_core

memory 1000 mb

geometry units au
O 0       0        0

H 0       1.430   -1.107

H 0      -1.430   -1.107
end


basis
 O library 6-31g*

 H library 6-31g*
end

dft
 xc beckehandh
end

tddft
 ecut -10

 nroots 10

 notriplet
end
task tddft

Here only excitations from occupied orbitals below -10 Hartrees will be cosidered.

Let me know if there are any issues.

Best regards,
-Niri
niri.govind@pnnl.gov

Just Got Here
5:14:30 PM PDT - Thu, Apr 5th 2012
ECUT refs
Hi Niri,
 I was just looking at this keyword - are there any references for the REW-TDDFT method, and how well DFT works in practice for such high-energy excitations?  I'm hoping to use this for the Fe m-edge (around 55 eV).  

 Also, is there any reason I couldn't use this to predict excited-state x-ray absorption by using VECTORS SWAP?  For example, suppose I want to predict what the core excitation would be after photoexciting a molecule, which happens to be just a HOMO-LUMO transition (say from orbital 22 to 23).  Could I then do something like:

vectors
 input ground_state_calc.movecs

 swap 22 23
end

td
 ecut -1.5

 nroots 20
end

task tddft energy

I'd expect to see one or more negative excitations, as I've seen when I accidentally converge to an excited state, but apart from is there anything wrong with this approach?

Thanks,
Josh

Forum Regular
11:10:59 PM PDT - Thu, Apr 5th 2012
Hi Josh,

I haven't published the REW-TDDFT approach yet. I am working on a few systems that I plan to submit very soon. Indeed, I implemented this to study pre-edge and NEXAFS core spectra. For the deep core states, I have found that CIS is identical to full-TDDFT. This is also cheaper. The full-TDDFT is more accurate for the valence excitations.

In terms of your case you don't need to swap orbitals if you use my windowing approach. For the m-edge, I recommend running a ground state calculation first to get an idea of the relevant orbitals and eigenvalues in that energy range and then use that energy as a cutoff. Give it a try and let me know how it goes.

Let me know if there are any issues.

Best regards,
-Niri
niri.govind@pnnl.gov

Just Got Here
4:41:30 AM PDT - Tue, Apr 10th 2012
Dear Niri,
first, thank you for your reply. I obvously missed that point in the Documentation. The approach is definitely the feature I was lking for. However, i tried to calculate your example getting the following error 
 tddft_input: unknown directive        0
 ------------------------------------------------------------------------
 ------------------------------------------------------------------------
  current input line :
    17:  ecut -10


I guess i have an outdated version. I currently use 
   nwchem branch = 6.0

I will recompile 6.1 and will try again. Maybe i will also try to use KS orbitals which i precalculated with a hole (1. calculöation with fractional occupation in core states according to 1electron hole). and then run the tddft core excitation calculation with KS orbitals with the hoile (guess-only) but electron configuration without hole. I will post if i will succeed. Please let me know what you think about this. Anyway, thanks for your help !
Best regards
Alexander

Forum Regular
2:42:20 PM PDT - Tue, Apr 10th 2012
Hi Alexander,

Yes. You need 6.1 for that feature. You can definitely use vectors with a hole.
However, the point of TDDFT core excitations is to be able to capture the relaxation because the electron-hole response is factored in via the response part of TDDFT. If you have a core-hole state to start with, you already have some response included. Give it a try.

Best regards,
-Niri


Forum >> NWChem's corner >> NWChem functionality