ИСТИНА

Various carotenoids play a number of important roles in photosynthesis including absorption of light in blue range of visible spectrum followed by excitation energy transfer to chlorophylls and dissipation of excessive energy which prevents photodamage of photosynthetic apparatus caused by highly reactive species formed under excessive light illumination. Two excited states of carotenoids are most likely responsible for these processes: the optically bright 1Bu+ state and the dark 2Ag- state, respectively. Adiabatic relaxation of both excited states is accompanied with a significant change of bond length alternation (BLA) in the conjugated π-system which is responsible for pronounced vibrational structure in absorption spectra and impedes assessment of quality of computational results usually relying on vertical transitions for rather large molecules. Excitation energies of adiabatic and vertical transitions were modeled for a series of polyenes with the number of conjugated double bonds ranging from 8 to 13 which corresponds to the majority of natural carotenoids. Excited state geometries were optimized using TDDFT/CAMB3LYP for 1Bu+ and MCSCF in the entire space of π-orbitals, energies of excited states were computed using DMRGSCF with NEVPT2 correction accounting for dynamic correlation. This approach provided qualitative agreement with experimental data for 1Ag- → 2Ag- excitation energies and good agreement for 1Ag- → 1Bu+. The difference between vertical and adiabatic excitation energies for 1Ag- → 2Ag- transition is approximately -1.1 eV which is due to the strong impact of BLA on energy levels of carotenoids. The aforementioned agreement with experimental data shows that DFT/MRCI and AM1/CASCI methods widely used for modeling of non-photochemical quenching significantly underestimate excitation energies since they produce vertical energies which are close to the experimental ones. While in this work the unsubstituted linear polyenes were studied (thus the effect of substituents has been neglected), this approach can be relatively easily extended to natural polyenes provided that two computational bottlenecks are addressed: geometry optimization of the 2Ag- state requiring seminumerical MCSCF gradients and limited performance of NEVPT2 in current implementation.