Data-Driven and Multiscale Modeling of DNA-Templated Dye Aggregates.

Area of Science:

• Materials Science
• Computational Chemistry
• Nanotechnology

Background:

• Dye aggregates are crucial for excitonic applications like biomedical imaging and organic photovoltaics.
• Large transition dipole moments (μ) are essential for optimizing coupling within dye aggregates.
• Identifying dyes with high extinction coefficients (ε > 150,000 M−1cm−1) is key, but current methods are slow.

Purpose of the Study:

• To develop a high-throughput computational workflow for screening dyes with desirable excitonic properties.
• To identify novel dyes with large extinction coefficients (ε) and transition dipole moments (μ) for excitonic applications.

Main Methods:

• Established a computational workflow integrating machine learning (ML), density functional theory (DFT), time-dependent (TD-DFT), and molecular dynamics (MD).
• Developed ML models (Classifier and Regressor) trained on 8802 dyes, achieving 97% accuracy and R2 > 0.9.
• Predicted ε for over 18,000 dyes and further screened top candidates using DFT/TD-DFT.

Main Results:

• ML models accurately predicted dye properties, enabling rapid screening of a large molecular dataset.
• Identified 15 promising dyes with large μ relative to the Cy5 reference dye.
• MD simulations accurately reproduced experimental results for dye dimers.

Conclusions:

• The developed computational workflow is effective for identifying dyes with large μ for excitonic applications.
• This approach accelerates the discovery and development of novel dyes for advanced technological uses.