Excited state geometry optimization Newbie question

[Jerry]

Hi all,
I have just started doing TDDFT excited state calculations with Turbomole and want to make sure I understand how things work.  To begin, I ran a vertical singlet excitation from the groundstate with
$scfinstab rpas
$soes a 5

and obtained the following in the spectrum file (IRREP a is the only IRREP for my molecule)

# Electronic excitation spectrum of TmoleXProject, IRREP a
# singlet excitations
#  excitation energy / eV  oscillator strength (length rep.)
# singlet excitations
 0.23182321032856E+01   0.10635053693431E-02
 0.27799040543984E+01   0.12966638141346E-02
 0.29734947988753E+01   0.35385871572430E+00
 0.31173464145357E+01   0.49196883080611E-02
 0.32445058998271E+01   0.43938827670142E-02

The result indicates that there are two excited singlet states (although weak) of lower energy than the strongest excited singlet.  A few questions on using these results to guide a geometry optimization:

1.  I'm trying to find the geometry of the first excited state that corresponds to relaxation from the 2.97 eV vertical excitation (3rd vertical excited singlet state), since that has the largest oscillator strength.  To do this, should I use:
$scfinstab rpas
$soes a 3

2.  Or, do I ignore the vertical excitation data and just use $ soes a 1 (and $scfinstab rpas) to obtain a the geometry optimized S1?

Of course, I will be running jobex -ex for the geometry optimizations (and used ridft for the vertical results)


Thank you for your help,
Jerry

[Hauke]

I'm trying to find the geometry of the first excited state that corresponds to relaxation from the 2.97 eV vertical excitation (3rd vertical excited singlet state)

I'm not an expert in the field either, but for me this seems to be two different goals (or how how do you know that the 3rd vertical excited state becomes the first excited state after optimization?).

If you want to optimize the structure of the first excited state but also include the higher states (to avoid problems with root flipping...)  I would recommend to use something like

$scfinstab rpas
$soes a 5
$exopt 1

See manual section "7.4.5 Excited State Geometry Optimizations". But I'm not 100% sure if this is what you want.

[Jerry]

Thanks Hauke.  I was hoping that the information from the vertical excitations might give me some guidance as to which state(s) to pursue.  But, as you said, this will change as the geometry is optimized.  It looks like it'll be research: optimize singlet states 1, 2, and 3, and evaluate them according to energy and oscillator strength to find the most important/relavent transitions.

Jerry