Data-Driven Modeling and Design of Sustainable High Tg Polymers.

Area of Science:

• Polymer Science
• Materials Science
• Machine Learning

Background:

• Predicting the glass transition temperature (Tg) is crucial for designing polymers with specific thermal properties.
• Traditional methods for Tg prediction are often time-consuming and may not capture complex structure-property relationships.

Purpose of the Study:

• To develop a rapid and robust machine learning (ML) methodology for predicting polymer glass transition temperature (Tg).
• To identify key chemical features influencing Tg for the design of sustainable high-temperature polymers.

Main Methods:

• Developed a comprehensive feature set integrating various aspects of polymer chemistry.
• Employed ML techniques to correlate chemical features with Tg and build a high-throughput predictive model.
• Utilized non-linear manifold transformations to capture complex relationships between polymer topology and Tg.

Main Results:

• Identified key chemical descriptors impacting Tg, including rotational degrees of freedom and a steric hindrance-based backbone index.
• Achieved robust Tg predictions across different data types, validated by experimental testing of novel polymer chemistries.
• Demonstrated that polymer rigidity and interaction dynamics are critical for tuning Tg.

Conclusions:

• Machine learning models can accurately predict Tg, indicating that topological descriptors contain the essential information.
• The complex, non-linear relationships between polymer structure and Tg necessitate advanced ML approaches over traditional regression.
• This ML methodology enables rapid optimization of polymer chemistries for targeted high-temperature applications.