Strategy to Screen Donor and Acceptor Pairs for Organic Solar Cells Through Machine Learning

Area of Science:

• Materials Science
• Computational Chemistry
• Renewable Energy

Background:

• Developing efficient and stable solar cell materials is crucial for renewable energy advancement.
• Traditional methods for screening donor and acceptor molecules are time-consuming and resource-intensive.

Purpose of the Study:

• To employ machine learning (ML) for optimizing donor and acceptor molecules in solar cells.
• To identify high-performance solar cell materials based on predicted device efficiency and chemical stability.

Main Methods:

• Evaluated approximately 42 machine learning models, including random forest (RF) regression, light gradient boosting machine (LGBM), and Nu support vector regression (NuSVR).
• Utilized chemical reactivity parameters and synthetic accessibility for molecule screening.
• Predicted key device properties: photoconversion efficiency (PCE), short-circuit current (Jsc), open-circuit voltage (Voc), and charge transfer (ΔN).

Main Results:

• Identified RF regression, LGBM, and NuSVR as the best-performing ML models.
• Achieved high R-squared values, indicating a close resemblance between predicted and actual device properties.
• Successfully screened molecules based on chemical reactivity and synthetic accessibility for improved PCE.

Conclusions:

• The developed ML framework efficiently identifies stable and high-performance donor and acceptor molecules for solar cells.
• This approach significantly reduces the time and resources required for material discovery.
• The study demonstrates the potential of ML in accelerating the development of advanced solar cell technologies.