Functional monomer design for synthetically accessible polymers.

Area of Science:

• Polymer Science
• Materials Science
• Computational Chemistry

Background:

• Machine learning (ML) shows promise for predicting polymer structure-property relationships.
• Data scarcity is a significant challenge limiting ML's practical use in polymer science.
• Developing comprehensive monomer property data is crucial for advancing polymer design.

Purpose of the Study:

• To create the first extensive database of monomer-level properties for millions of synthetically accessible polymers.
• To address the obstacle of data sparsity in machine learning applications for polymer science.
• To facilitate functional monomer design and accelerate the discovery of novel polymeric materials.

Main Methods:

• Integrated quantum chemistry calculations with active learning to generate monomer properties.
• Developed a comprehensive database covering approximately 12 million synthetically accessible polymers.
• Benchmarked monomer-level property descriptors against advanced computational predictions and experimental data.

Main Results:

• Demonstrated the relevance of monomer-level properties for effective polymer design.
• Revealed weak correlations among many monomer properties, indicating design flexibility.
• Showcased simultaneous optimization of multiple physical properties through monomer selection.

Conclusions:

• The developed database provides a powerful resource for machine learning-driven polymer design.
• Weak correlations between monomer properties offer significant freedom for optimizing material characteristics.
• This work paves the way for creating novel, synthetically accessible polymers with tailored functionalities.