Interpreting Molecular Descriptors for Glass Transition Temperature Prediction and Design of Polyimides

Area of Science:

• Materials Science
• Polymer Chemistry
• Computational Chemistry

Background:

• Rational design of polyimides (PIs) with targeted glass transition temperature (Tg) is critical for microelectronics.
• Existing data-driven models often lack physical interpretability, especially with limited data.

Purpose of the Study:

• Develop a highly interpretable Quantitative Structure-Property Relationship (QSPR) model for accurate PI Tg prediction.
• Provide physicochemical insights into descriptor influence on Tg.
• Establish molecular design guidelines for PIs.

Main Methods:

• Employed Genetic Algorithm combined with Multiple Linear Regression (GA-MLR) to identify optimal molecular descriptors.
• Utilized a curated dataset of polyimides.
• Validated model performance through statistical tests and free volume theory.

Main Results:

• Identified seven key molecular descriptors for accurate Tg prediction.
• Demonstrated robust predictive performance and generalization ability.
• Physicochemically interpreted descriptors, linking them to fractional free volume and chain packing/intermolecular interactions.

Conclusions:

• Established a reliable and transparent computational tool for PI development.
• Provided mechanistic understanding of descriptor influence on Tg.
• Generated clear molecular design guidelines for achieving desired Tg and processability in PIs.