Natural systems like photosynthetic complexes have shown that scaffolded molecular networks are capable of efficient electronic processes. While artificial systems typically have much lower efficiencies, natural examples suggest that their functional bottlenecks can be mitigated by finely controlling their structural and energetic properties. To meet this need, the structures of modular molecular networks can be controlled precisely using DNA scaffolds. However, to leverage this control, the optimization rules must be well understood. Here, we consider a 14-site donor-bridge-acceptor network. When the donor molecules are excited, energy cascades energetically downhill through the bridge sites to the acceptors. The energy-transport efficiency changes in distinct ways when particular bridge sites are removed from the system, sometimes leading to improvements in the transport efficiency, leading to the question of what role each of the identical bridge sites have in this system. This question is difficult to probe in transient spectra, because the excitonic transition energies are so similar. Using machine-learning methods to analyze experimental FRET measurements (Figure 1), as well as molecular-dynamics and energy-transport simulations, the roles of the sites are analyzed. Furthermore, efficient system configurations are predicted by the machine-learning model and confirmed experimentally.

Machine-learning methods for understanding energy-transport in molecular networks

Fall 2022 abstracts, posters, and presentations will launch in mid-September.

ACS Fall 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.