Leveraging Machine Learning and DFT for Identification and Analysis of Dyes for Use in DNA-Templated Dye Aggregates

Frenkel exciton delocalization has been observed in natural and synthetic organic dye aggregates. The phenomenon of exciton delocalization is useful for a variety of applications, including light harvesting, medical imaging, and photodynamic therapy. Additionally, the excitonic properties of organic dye aggregates show promise for use in quantum information systems (QIS). Machine Learning (ML) can be used to quickly screen large numbers of dye monomers for advantageous properties, such as high transition dipole. Density Functional Theory (DFT) can then be used to confirm and further probe the structure-property relationships of target dye families. We trained a Random Forest classifier and regressor model to identify several molecules with high extinction coefficient. Specific structure-property relationships of bacteriochlorin dyes were identified using DFT and confirmed experimentally.