In recent years, a number of computational studies have claimed that there exist "charge-transfer like" (CT-like) excitations, such as certain singlet pi-pi* transitions in organic chromophores that no spectroscopist would assign as CT. It has so far not been demonstrated convincingly what exactly renders such an excitation CT-like, with the exception that TDDFT with standard functionals dramatically underestimates their energy and long-range corrected (LC) hybrid functionals - especially when ’tuned’ - produce TDDFT energies that are close to benchmark data, usually from CC2 calculations. This behavior mirrors that of the well-known breakdown of TDDFT for CT-proper. However, according to our analysis there are no further similarities and improvements from LC functionals are obtained for the wrong reasons. Further, for molecules with "CT-like" behavior CC2 is not accurate enough to reveal when TDDFT performs well and when it doesn’t. We examine various criteria by which to quantify physical CT. Assuming that CCSD(T)/TZVP benchmark data generated for our study are sufficiently accurate, it follows that there are no known "CT-like" excitations. The problem is differential correlation between the ground and the excited states, which TDDFT cannot describe well enough for "CT-like" systems.
Université Paul Sabatier - Bat. 3R1b4 - 118 route de Narbonne 31062 Toulouse Cedex 09, France