Machine-Learning-Assisted Accurate Prediction of Molecular Optical Properties upon Aggregation

Area of Science:

• Materials Science
• Computational Chemistry
• Machine Learning

Background:

• Molecules in practical applications often exist in aggregates.
• Predicting molecular performance in aggregates, such as aggregation-induced emission (AIE) or aggregation-caused quenching (ACQ), is crucial.
• Existing methods for predicting these properties can be limited.

Purpose of the Study:

• To develop a machine learning (ML) model for fast and accurate prediction of AIE/ACQ properties.
• To establish a structure-property relationship for molecules in aggregated states.
• To accelerate the discovery of new solid-state optical materials.

Main Methods:

• Established a database of AIE/ACQ molecules from literature.
• Employed a multi-modal approach integrating various molecular descriptors.
• Utilized dimensionality reduction for feature extraction and synthesized multi-modal features.
• Developed and compared several state-of-the-art ML methods.
• Implemented an ensemble strategy to combine predictions from multiple models.

Main Results:

• The ML models successfully built structure-property relationships.
• The ensemble strategy demonstrated superior prediction accuracy.
• Validation with three newly designed molecules showed reasonable consistency between predictions and experimental outcomes.
• Demonstrated the capability of ML in predicting properties of molecules in aggregated states.

Conclusions:

• Machine learning is a powerful tool for predicting molecular properties in the aggregated state.
• The proposed multi-modal ensemble approach effectively predicts AIE/ACQ properties.
• This work accelerates the development of advanced solid-state optical materials.