Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

MO: Fachverband Molekülphysik

MO 3: Electronic I

MO 3.4: Vortrag

Montag, 14. März 2022, 17:30–17:45, MO-H5

Unsupervised learning as a key tool to explore elements of the efficiency of PS1 in an QM/MM approach — •Ferdinand Kiss, Sebastian Reiter, and Regina de Vivie-Riedle — Department of Chemistry, LMU Munich, Germany

Modern photovoltaic materials can be seen as biomimetics of photosynthesis in photoautotrophic organisms. Photosystem I (PS1) has one of the highest conversion efficiencies of 88%, from absorbed quanta to the reduction of NADP⁺. A deeper understanding of the effects of structural relations and electrostatic influences on the site energies, low-lying charge transfer states and absorption profiles of photoactive components of the PS1 promises to yield the answer to its outstanding efficiency. We developed an automated protocol for data extraction and processing from MD simulations by unsupervised machine learning. On this basis we set up electronic structure investigations in a QM/MM approach. Our maxim of a bias-free, dimensionality reduced and thus computational affordable approach to QM/MM studies aim towards a post-classical description of processes in large complex systems. With the developed tools at hand, we were able to rationalize relevant structural parameters in the 288 chlorophylls of the PS1 trimer. Furthermore, we were able to approximate electrostatic embedding in different pockets within the PS1 with minimal computational cost. The protocol as mentioned above and its results will guide the understanding of photosynthesis. The insights will help in the development of novel artificial photosynthesis designs.