egrad density difference (ed.plt)

[Helph]

Dear Turbomole users,
(before my question I give some context)

I am performing two-photon absorption calculations of an organic molecule. Now I would like to visualize the electron density difference and I have found in the manual that with egrad it is possible to do it.

1) So you would have to add in the control file the keywords:
$pointval
$soes
a    1
$exopt
a  1
(where 1 is the excitation you are interested to get) and after run egrad > egrad.out
with this I would get two files, ed.plt and td.plt. Being ed.plt the file that contains the information about density difference.


But also, if you check the escf.out file there is information about the the dominant contributions for each excitation. However, in the escf.out you would be able to see each transition separately along with the percentage in contribution (e.g. 1st excitation: HOMO --> LUMO 90%). If one plot the orbitals of those transitions, with the followong command:
$pointval mo 354-355 fmt=cub
(where 354 in HOMO and 355 LUMO)

one would also get the electron density difference.

My questions:
1) is there any difference between those two methods?
2)I am not interested in optimizing the structure of an excited state. Is it possible to use egrad without optimizing the structure of the excited state?

I've attached an image with both methods. There one can see that the electron density diff looks similar, but not the same.
Thanks

[christof.haettig]

If a single one-electron transition dominates the excitation from the ground to the excited state, you can indeed construct the so-called orbital-unrelaxed difference density from the orbitals.
The egrad program should plot the orbital-relaxed difference density, which is usually a better approximation.
You can do that without doing a geometry optimization, just calling egrad once for the current geometry. The gradient that it computes in addition to the orbital-relaxed density will not cost that much extra. The most costly step is the solution of the Z-vector equations that is needed for the density.

Constructing the density from the orbitals becomes complicated when several orbital transitions contribute significantly since also the phases of the orbitals matter.

[Helph]

Thank you very much for taking the time to answer 

It's now clear that when dealing with multiple transitions, only visualizing orbitals is not enough. Egrad proves to be more accurate as it considers all contributions involved in a specific excitation.