Harnessing DFT and machine learning for accurate optical gap prediction in conjugated polymers

Area of Science:

• Materials Science and Engineering
• Computational Chemistry
• Polymer Science

Background:

• Conjugated polymers (CPs) possess tunable electronic properties crucial for optoelectronic applications.
• Accurate prediction of the optical band gap (E_expgap) in CPs is a significant challenge.
• Existing methods struggle to correlate theoretical calculations with experimental optical band gaps.

Purpose of the Study:

• To develop a robust model for predicting the experimentally measured optical band gap (E_expgap) of conjugated polymers (CPs).
• To integrate density functional theory (DFT) calculations with a data-driven machine learning (ML) approach.
• To accelerate the design and development of high-performance CPs for photoelectronic applications.

Main Methods:

• Utilized 1096 data points of CPs, employing DFT to calculate the HOMO-LUMO gap (E_oligomergap) of modified oligomers.
• Developed ML models using E_oligomergap and molecular features of monomers to capture electronic properties.
• Trained and validated six ML models, with XGBoost-2 identified as the best performing model.

Main Results:

• Modified oligomers significantly improved the correlation between DFT-calculated and experimental band gaps (R² = 0.51 vs. 0.15 for monomers).
• The XGBoost-2 model achieved high accuracy (R² = 0.77, MAE = 0.065 eV) in predicting E_expgap, within experimental error.
• XGBoost-2 demonstrated excellent interpolation and extrapolation capabilities across diverse CP structures.

Conclusions:

• A novel strategy combining DFT and ML provides accurate and efficient prediction of CP optical band gaps.
• This integrated approach facilitates the accelerated design of advanced conjugated polymers.
• The developed model shows significant promise for future high-performance CPs in photoelectronic devices.