Population analysis for TD-DFT excited states

[vvallet]

Dear colleagues,

Does anyone know how to perform population analysis for the TD-DFT excited states? I don't know how one can retrieve the excited state densities and ask $pop to use this one.

Any hints would be appreciated.

Valérie Vallet

[uwe]

Dear Valérie,

if you add $pop to the control file and then run egrad, you will get the Mulliken population analysis of the excited state you have chosen to calculate the excited state gradient for. Running dscf or grad with $pop will result in the population analysis of the ground state.

Regards,

Uwe

[martijn]

I have a related question. When doing population analysis on an excited state, can one only obtain charges or also spin populations (e.g. for triplet excited states)?

Thanks in advance,

Martijn

[Arnim]

Hello Martijn,

wouldn't it be more meaningful to visualize the MOs of the dominant contributions from an excitation than to rely on these population analysises, which are questionable already in the ground state?

Best,
Arnim

[christof.haettig]

The escf (and also the ricc2) code calculates for triplet excitations the spin component which has the same S_z eigenvalue as the ground state. Therefore, one does (starting from a closed-shell ground state) obtain exactly the  same densities for alpha spin and beta spin and the spin density can not obtained by simply taking the difference...

Christof

[martijn]

Hi,

Thanks. I'am well aware of the deficiencies of population analysis schemes in general and Mulliken's method in particular but I'm curious to see if some symmetry breaking occurs in the excited state and in an deltaSCF set-up spin population analysis has proven to be a quick way of testing this (take two atoms that should be symmetrically equivalent based on the geometry and check if they have the same/similar population). Guess, I could in principle use Mulliken charges (instead of spin populations) for the same purpose.

Chrisof, do the charges that turbomole reports for an excited state suffer from a similar problem? Guess they shouldn't as they should depend only on the total density and not spin densities but would like to be sure. Also, I'm not sure from your reply if you can extract (alpha and beta spin) orbitals for the excited state. Is that possible?

Cheers,

Martijn

[christof.haettig]

What is the reference state from which you start? Is it a closed-shell ground state?

Christof

[martijn]

Yep, the reference state is the closed shell groundstate of the cluster we study. We look at both singlet and triplet excitations (we're actually mostly interested in the former, triplets are mostly studied to compare with delta-SCF calculations).

Cheers,

Martijn

[christof.haettig]

It that case there will be not spin-contamination in the excited state from the TDDFT calculation. Of cause, you don't get by TDDFT really a wavefunction for the excited state, but within the limits you interprete you excitation amplitudes as descriptors of a wavefunction you get  in this case spin (S^2) eigenfunctions. I.e. for singlet excited state the spin-density is zero everwhere, for triplet states the difference density between ground and excited state might be used as approximation for the spin-density.

Christof