The very first step in the complex function of the retinal proteins is the absorption of light which is followed by anisomerization of the chromophore within the protein. The high speed, efficiency and selectivity of the process are directly related to the optimized molecular arrangement of the binding pocket. Different factors have so far limited the theoretical description of theses systems. First, the reaction occurs in an excited electronic state for which quantum chemical calculations provide accurate solutions only at high computational cost. Second, the dynamics in the retinal proteins are believed to happen on strongly coupled potential energy surfaces. This makes the use of advanced nonadiabatic molecular dynamics schemes mandatory in order to obtain realistic values for reaction time scales and quantum yields. Finally, the complexity of the protein interaction with the chromophore requires the inclusion of a large part of the binding pocket in structural models for the active site, which involves a high computational effort. We have developed methods to cope with these specific problems and are using them to analyze the early stages of the bR and Rh photocycles. These calculations will provide structural information on intermediates in the photocycle and lead to a detailed picture of the complex retinal photoreaction.

Our first implementations of highly efficient semi-empirical excited state nonadiabatic methodology were based on TD-DFTB, an approximation to TDDFT, and included a linear response treatment of excited states with analytical gradients and excited state density matrix (Z-vector equation), Ehrenfest nonadiabatic time propagation, non-adiabatic coupling vectors, and a Tully surface-hopping algorithm. At the same time it was discovered that TDDFT has principle problems describing intra-molecular charge transfer states and predicts a qualitatively wrong excited state potential energy surface (PES) of polyenes and protonated Schiff base retinals. Beside HF exchange, a balanced incorporation of dynamic and non-dynamic correlation is required to obtain realistic PES and response of optical properties to geometric distortions and electric fields.

An alternative method, OM2/MRCI, was found suitable in extended benchmarks and was successfully employed in TP7. Nonadiabatic coupling vector, conical intersection optimization schemes, and a Tully surface hopping scheme for this method have been implemented and are developed further in the group of W. Thiel (MPI Mülheim). After the evaluation of excited-state dynamics, and gas-phase simulations, the approach will be used in a QMMM setup to simulate the photoisomerization dynamics of the retinal chromophore in Rh and bR and the formation of the K intermediate of bR and the batho state of Rh. In parallel, the excited state PES and influence of the protein binding pocket on the energetics (regio-selectivity, mutations, modified chromophores) will be studied.

Further investigations will address the structure and decay of transient species (comparison with experiments), intermediates of non-reactive reactions, and the fluorescence decay kinetics in solution.