Generative Deep Learning-Based Efficient Design of Organic Molecules with Tailored Properties.

Area of Science:

• Materials Science
• Computational Chemistry
• Artificial Intelligence

Background:

• Designing molecules with specific properties is crucial for advancing various scientific fields.
• Deep learning (DL) offers a powerful approach to accelerate materials discovery by analyzing molecular structure-property relationships.

Purpose of the Study:

• To develop a generative deep learning (Gen-DL) model capable of designing molecules with targeted optical properties across different solvents.
• To demonstrate the model's ability to incorporate key chemical design principles and generate practically useful molecules.

Main Methods:

• Trained a Gen-DL model on an experimental database (DB_exp) of 71,424 molecule/solvent pairs.
• The model was designed to predict and generate molecules with specific optical characteristics, including absorption/emission peaks, bandwidth, extinction coefficient, and photoluminescence (PL) quantum yield and lifetime.

Main Results:

• The Gen-DL model successfully generated molecules with specified optical properties.
• The model demonstrated an understanding of conjugation effects, Stokes shifts, and solvent effects in molecular design.
• Generated molecules were suitable for real-world applications, including organic photovoltaics (OPVs), organic light-emitting diodes (OLEDs), and bioimaging dyes.

Conclusions:

• The developed Gen-DL model is a promising tool for accelerating the discovery and design of novel molecules with tailored properties.
• This AI-driven approach can significantly benefit research areas requiring advanced materials with specific optical functionalities.