Predicting Young's Modulus of Linear Polyurethane and Polyurethane-Polyurea Elastomers: Bridging Length Scales with...

Area of Science:

• Polymer Science
• Materials Science
• Computational Chemistry

Background:

• Predicting complex polymer properties requires multiscale modeling, bridging molecular to bulk behavior.
• Polyurethanes exhibit inherent heterogeneities due to domain incompatibility, complicating property prediction.
• Machine learning for polymers faces challenges with small datasets and monomer diversity.

Purpose of the Study:

• To develop accurate machine learning models for predicting polyurethane and polyurethane-urea elastomer properties.
• To establish a link between monomer chemistry and interchain interactions for property prediction.
• To demonstrate the utility of machine learning as a materials design tool.

Main Methods:

• Utilized a dataset of 63 complex polyurethane elastomers.
• Employed machine learning by estimating interchain interactions from monomer chemistry.
• Trained models to predict Young's modulus based on these interchain features.

Main Results:

• Achieved accurate prediction of Young's modulus with an R-squared value of 0.91.
• Demonstrated successful application of the trained model to identify compositions for target moduli.
• Validated the methodology's effectiveness for complex elastomers with limited data.

Conclusions:

• Machine learning, focused on interchain interactions derived from monomer chemistry, enables accurate prediction of polyurethane properties.
• The developed methodology serves as a valuable design tool for tailoring material properties.
• This approach offers a pathway for modeling materials with limited datasets and understanding physicochemical forces.