High-Performance Polyimide Design: Machine Learning-Accelerated Multiobjective Optimization

Area of Science:

• Materials Science
• Polymer Chemistry
• Computational Materials Science

Background:

• Advanced polyimides (PIs) are crucial for flexible electronics, foldable displays, and aerospace due to their transparency, mechanical robustness, and thermal stability.
• Simultaneously optimizing these properties in PIs presents a significant design challenge.

Purpose of the Study:

• To develop a data-driven multiobjective optimization framework for systematic polyimide design.
• To accelerate the discovery of multifunctional polymers with tailored properties for next-generation applications.

Main Methods:

• Utilized large language models for high-quality data extraction.
• Employed machine learning models for performance prediction and structure-property relationship analysis.
• Validated predictions through experimental synthesis and evaluation of five PI films.

Main Results:

• Successfully screened over 20,000 potential polyimide formulations.
• Developed PI films (PI-a, PI-b, PI-e) exceeding commercial benchmarks.
• Achieved high glass transition temperatures (336-376°C), tensile strengths (207-324 MPa), and transmittance (>89.6%).

Conclusions:

• The data-driven framework enables rapid design and optimization of multifunctional polymers.
• This approach effectively deciphers complex structure-property relationships, particularly those influenced by charge transfer complexes.
• The framework is broadly applicable for designing advanced polymers for flexible electronics and energy storage.