Getting Low Intensity Excited States at Long Wavelength with TDDFT

Other than setting the number of requested excited states higher and setting the tdscf_coeff_cutoff lower, are there other (perhaps better) ways to pull out ‘low intensity’ excited states at long wavelength via TDDFT? In other words, I am assuming (based on experimental UV Vis) that I’m not getting to these because the oscillator strength is small especially versus the other excited states.

If the state isn’t found by TD-DFT, tdscf_coeff_cutoff won’t help you.
The algorithm will find all states from low to high energy, so increasing the number of requested states will find states at higher wavelength.
Small oscillator strengths are not a fundamental problem for finding the state.

There could be a case where you miss a low energy state because of a too small expansion space but the chance is higher that TD-DFT itself fails. E.g. for a dark state from double-excitations character. Or charge-transfer state.

The experimental low-lying dark state might also well be at a different (=wrong) position in TD-DFT.


With an excited state like →

Excited State    2 (1 A):   0.14172 au   321.50 nm f = 0.0002
  Sums of squares: Xssq =  1.001023e+00; Yssq =  1.022856e-03; Xssq - Yssq =  1.000000e+00
    46 -> 48   -0.711110 (50.568%)
    47 -> 49    0.699963 (48.995%)

Where you’re getting almost an even contribution from each (major) excitation, I wonder about if these states are convoluted such that they might correspond to two close-by peaks in the experimental spectra – ???

Perhaps the state could split when vibrational effects are considered, but I dont think that is typical.
Language wise you have to be careful with the word excitation and not mix orbital pair contributions with the molecular state.

It is more likely that the state you are searching just has the wrong energy and is somewhere else. The state ordering can be problem in TD-DFT, it is unfortunately not very black box to do spectra.
Did you try functionals with (very) different Fock exchange?

OK, this makes sense. Let me try some other functionals and do some comparison.

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.