Machine learning-assisted molecular design and efficiency prediction for high-performance organic photovoltaic...

Area of Science:

• Materials Science
• Organic Electronics
• Computational Chemistry

Background:

• Developing high-performance organic photovoltaics (OPVs) requires efficient methods for identifying suitable donor materials.
• Establishing structure-property relationships is crucial for predictive material design, but experimental synthesis and testing are time-consuming.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting photovoltaic properties of organic donor materials based on their chemical structures.
• To enable rapid screening of potential OPV materials, thereby accelerating the discovery process.

Main Methods:

• Compilation of a database of over 1700 donor materials from existing literature.
• Exploration of various molecular representations (images, ASCII strings, descriptors, fingerprints) as input for supervised ML algorithms.
• Training and evaluation of ML models to establish structure-property relationships.

Main Results:

• Machine learning models, particularly those utilizing molecular fingerprints exceeding 1000 bits, achieved high prediction accuracy for photovoltaic properties.
• Screening of 10 newly designed donor materials demonstrated good consistency between ML predictions and experimental outcomes.
• Validation confirmed the reliability of the ML approach for prescreening OPV materials.

Conclusions:

• Machine learning provides a powerful and efficient tool for prescreening organic photovoltaic materials.
• The established structure-property relationships can significantly accelerate the development of advanced OPV technologies.