Efficient algorithms for the simulation of the non-adiabatic exciton transfer dynamics in subcomplexes of Photosystem II

• Applicants: Professor Dr. Marcus Elstner; Professor Dr. Ulrich Kleinekathöfer
• Subject Area: Theoretical Chemistry: Electronic Structure, Dynamics, Simulation ; Theoretical Chemistry: Molecules, Materials, Surfaces
• Term: since 2018
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 397706021

The first step of photosynthesis, which occurs in plants, algae and certain bacteria, is the harvesting of sunlight. During the initial funding period, a non-adiabatic molecular dynamics (NAMD) scheme was developed to study the transport of excitation energy in light-harvesting complexes. The efficiency of this method is based on the use of a special parametrized density functional tight binding (DFTB) version for the ground state dynamics and the range separated version of DFTB for excited states properties. To further increase the efficiency, the training of neural networks is used in each time-critical step, allowing the simulation of exciton dynamics in extended systems on the picosecond time scale. This procedure lifts some of the limitations inherent to earlier approaches, wherein time-independent Frenkel Hamiltonians are determined and the impact of the surrounding environment is characterized by spectral densities. In addition, the effects of the excitations of the pigment molecules are described at the molecular level. In the present scheme, a time-dependent Frenkel Hamiltonian is evaluated “on the fly” during molecular dynamics simulations and subsequently used to propagate the electronic Schrödinger equation. A remarkable acceleration in computational speed is achieved by training neural networks, which effectively replace the time-consuming quantum computations. A major goal of this proposal is to improve the NAMD scheme developed during the initial funding period by incorporating charge-transfer states and carotenoid molecules into the dynamic electronic description. This will enable the study of exciton transfer between carotenoid and chlorophyll molecules. Further technical steps include the explicit treatment of excited states forces and an automatic parametrization of the neural networks. With regard to the systems to be studied, the focus will now be on specific components of the plant PSII system, namely the supercomplexes LHCII-CP24-CP29 and LHCII-CP26-CP43. Given the considerable scale of these systems, our initial approach will be to benchmark the enhancements to the NAMD scheme using the CP24 complex, with a view to subsequently extending the calculations to the larger supercomplexes. In order to establish a direct correlation with experimental data, we will compute time-independent and time-dependent spectroscopic signals with the existing NAMD scheme. These measurement methods will include at least linear and transient absorption as well as two-dimensional spectroscopy. These experimental techniques have recently been applied to LHCII-CP24-CP29 and LHCII-CP26-CP43. Therefore, a direct comparison between the NAMD approach and the experimental results is planned for the second funding period.

• DFG Programme: Research Grants