Glass Transition Prediction of Binary Copolymers Across Large Chemical Spaces Using Machine Learning and...

Area of Science:

• Polymer Science
• Materials Informatics
• Computational Chemistry

Background:

• The glass transition temperature (Tg) is critical for polymer performance but difficult to predict rapidly across many chemical structures.
• Predicting Tg in binary copolymers is challenging due to complex structure-property relationships.

Purpose of the Study:

• Develop a machine learning (ML) framework for high-throughput prediction of Tg in binary copolymers.
• Enable rapid exploration of large chemical spaces for polymer design.

Main Methods:

• Trained ML models on experimental datasets of homopolymers and copolymers.
• Integrated structural descriptors and molar composition ratios.
• Utilized graph-based algorithms and molecular dynamics (MD) simulations for validation.

Main Results:

• Achieved high predictive accuracy for Tg with RMSE of ~14K and R² of ~0.98.
• Generated and predicted Tg for ~148,000 binary copolymer compositions.
• Identified optimal monomer pairings for targeted applications, including elastomers.

Conclusions:

• The ML framework enables accelerated discovery and multi-property optimization of functional copolymers.
• This integrated approach combines experimental data, ML, and physics-based validation.
• Facilitates the design of polymers with tailored thermal stability for specific applications.