Davydov-type excitonic effects on the absorption spectra of parallel-stacked and herringbone aggregates of pentacene: Time-dependent density-functional theory and time-dependent density-functional tight binding

Exciton formation leads to J-bands in solid pentacene. Describing these exciton bands represents a challenge for both time-dependent (TD) density-functional theory (DFT) and for its semi-empirical analog, namely, for TD density-functional tight binding (DFTB) for three reasons: (i) solid pentacene and pentacene aggregates are bound only by van der Waals forces which are notoriously difficult to describe with DFT and DFTB, (ii) the proper description of the long-range coupling between molecules, needed to describe Davydov splitting, is not easy to include in TD-DFT with traditional functionals and in TD-DFTB, and (iii) mixing may occur between local and charge transfer excitons, which may, in turn, require special functionals. We assess how far TD-DFTB has progressed toward a correct description of this type of exciton by including both a dispersion correction for the ground state and a range-separated hybrid functional for the excited state and comparing the results against corresponding TD-CAM-B3LYP/CAM-B3LYP+D3 results. Analytic results for parallel-stacked ethylene are derived which go beyond Kasha's exciton model [M. Kasha, H. R. Rawls, and A. El-Bayoumi, Pure Appl. Chem. 11, 371 (1965)] in that we are able to make a clear distinction between charge transfer and energy transfer excitons. This is further confirmed when it is shown that range-separated hybrids have a markedly greater effect on charge-transfer excitons than on energy-transfer excitons in the case of parallel-stacked pentacenes. TD-DFT calculations with the CAM-B3LYP functional and TD-lc-DFT calculations lead to negligible excitonic corrections for the herringbone crystal structure, possibly because of an overcorrection of charge-transfer effects (CAM refers to Coulomb attenuated method). In this case, TD-DFT calculations with the B3LYP functional or TD-DFTB calculations parameterized to B3LYP give the best results for excitonic corrections for the herringbone crystal structure as judged from comparison with experimental spectra and with Bethe-Salpeter equation calculations from the literature.