Machine learning glass transition temperature of polymers.

Area of Science:

• Polymer Science
• Computational Chemistry
• Materials Science

Background:

• The glass transition temperature (Tg) is a critical thermophysical property of polymers.
• Experimental determination of Tg can be time-consuming and resource-intensive.
• Data-driven modeling offers a promising avenue for rapid and robust Tg prediction.

Purpose of the Study:

• To develop a predictive model for polymer glass transition temperature (Tg).
• To utilize molecular descriptors and machine learning for accurate Tg estimations.
• To provide a fast and cost-effective alternative to experimental Tg determination.

Main Methods:

• Employed Gaussian process regression (GPR) as the core modeling technique.
• Utilized molecular traceless quadrupole moment and average hexadecapole moment as key descriptors.
• Trained and validated the model on a dataset of 60 polymer samples with known Tg values.

Main Results:

• The GPR model accurately predicted Tg for polymers across a range of 194 K to 440 K.
• The model demonstrated high accuracy and stability in its Tg estimations.
• The developed method offers rapid and low-cost predictions of Tg.

Conclusions:

• Molecular moments are effective descriptors for predicting polymer Tg.
• Gaussian process regression provides a powerful tool for thermophysical property prediction.
• This data-driven approach significantly enhances the efficiency of Tg determination in polymer research.