Depending on the initial photoexcitation of the donor or acceptor phase, different efficiencies of charge generation can be observed in organic solar cells. We investigate the origin of this dichotomy by simulations based on quantum-quantum-classical embedded GW-Bethe-Salpeter equation of conversion dynamics from localized to charge-transfer (CT) excitations at the interface of a diketopyrrolopyrrole (DPP) polymer and fullerene. Specifically, we determine the excitonic energy levels, their electronic couplings, and the reorganization energies for the respective conversion processes within Marcus theory. Our calculations yield a variety of CT-type excitations of different characters with the lowest integer CT excitations of relevance for charge generation separated by 0.30 eV. Further analysis reveals that the activation barrier for conversion to the lowest CT state is significantly higher (0.25 eV) for the polymer LE than for the fullerene LE (0.05 eV), leading to a preferred population of the higher, less strongly bound CT state from the photoexcited donor. From a population dynamics model, we find that, indeed, on the time scale of one picosecond after the respective excitation, the donor excitation leads to the formation of a CT excitation with on average 0.16-0.27 eV lower electron-hole binding energy, providing a pathway to faster charge separation.
Sundaram et al. (2026) studied this question.