Excited state energy - TDDFT/PCM

Dear users and developers!

I am trying to perform a geometry optimization at the TDDFT level of theory, using non-linear state specific pcm approximation to model the solvent effects.

The input goes as follows:

0 1
O 1.0851107 -1.4723113 0.0000000
C -0.0881814 -0.6867215 0.0000000
C -1.2522510 -1.5374500 0.0000000
H -2.2842915 -1.1889138 0.0000000
C -0.7529743 -2.8518344 0.0000000
H -1.3257357 -3.7810222 0.0000000
C 0.6500541 -2.7764116 0.0000000
H 1.4327640 -3.5344305 0.0000000
O -1.0851107 1.4723113 0.0000000
C 0.0881814 0.6867215 0.0000000
C 1.2522510 1.5374500 0.0000000
H 2.2842915 1.1889138 0.0000000
C 0.7529743 2.8518344 0.0000000
H 1.3257357 3.7810222 0.0000000
C -0.6500541 2.7764116 0.0000000
H -1.4327640 3.5344305 0.0000000


jobtype opt
method b3lyp5
basis cc-pVDZ
purecart 11
gen_scfman false
scf_final_print true
solvent_method pcm

StateSpecific perturb

dielectric 37.5
opticaldielectric 1.806874

The problem that I have encountered lies in the final energy listed in the output, after the geometry has converged, corresponding to the ground state energy instead of the excited state one. When PCM is not included, the final energy rightly refers to the excited state that has been the target state for the gradient calculations. Once the pcm is added, the energy switches to the ground state.

I could (and I think I can) compute the excited state energy using the ground state one and the excitation energy, but I would rather not do it If I don’t have to. Also, I am not certain that the gradient
is calculated for the excited state energy instead of the ground state one - is it?

Thank you for any assistance here.

Forgot to mention: I am using Q-Chem 6.0.

Couple of comments:
(1) There are no gradients yet for ptSS. You can optimize on an excited state using TDDFT + equilibrium PCM and then compute single-point energies with the noneq PCM, I believe that is what this input file will do.
(2) State-switching (a.k.a. root-flipping) is always a potential problem in excited-state optimizations and can be more so if you are using a state-specific PCM approach. What is the energy gap between the ground state and first excited state at the end of the optimization? Monitoring this energy gap along the whole excited-state optimization pathway is always good practice.

I was wondering if ptSS (or any cLR formalism) was important for geom opts for my own work, and this was the best study I could find. They concluded that SS approaches are insignificant but I’m really not happy with their analysis… who on earth states errors in +/- nm???

(I have no opinions on their conclusions without doing my own analysis though)

LOL, nm. The answer to your (perhaps rhetorical) question is: experimental chemists. And Gaussian TDDFT will output excitation energies in nm, which I refuse to do because I don’t want to do long division to put them in useful units, and energy is more fundamental in QM (e.g., energy gap law, energy denominators in perturbation theory… there’s a reason those things are in energy not wavelength units). I have for years offered my experimental collaborators that my group will stop using atomic units in public if they will stop using wavelength in public (both have their uses but not in the “mixed company” of experiment and theory). So far, no takers.

It is true that ptSS and cLR are essentially the same models. Q-Chem doesn’t have gradients for ptSS yet (there is someone working on it but I’m not prepared to commit to an ETA). I maintain that geometry changes are likely small.

1 Like

Thank You for your reply and clarification for which version of PCM the gradients are really computed.
My question, however, referred mainly to the strange (in my opinion) fact that Q-Chem in the TDDFT calculation that included solvent (PCM) gives the ground state energies as the final energy values instead of the target excited state ones. As the excited state I am investigating is a well defined, mainly HOMO-LUMO singlet state, I do not think that there isn’t any root flipping in the process of geometry optimization. Especially that in the TD-DFT section of the output (in every step of the geometry optimization process) the excitation energy is positive and properly large (over 1.5 eV).

I suspect that this is some minor bug in the code, especially that when no PCM is used, the final energy properly correspond to the singlet excited state energy. Once the PCM in switched on, the final energies begin to refer to the ground state ones. Isn’t it queer?

Best wishes!

Not sure that I follow. Can you provide a bit of the output file that illustrates what you mean?