Property Prediction of Bio-Derived Block Copolymer Thermoplastic Elastomers Using Graph Kernel Methods.

Area of Science:

• Polymer Science
• Materials Science
• Computational Chemistry

Background:

• Growing plastic pollution and greenhouse gas emissions necessitate novel bio-based polymers.
• Current iterative methods for discovering new materials are often slow.
• Machine learning (ML) offers a promising approach to accelerate materials discovery.

Purpose of the Study:

• To introduce PolyAGM, a ML algorithm for predicting polymer properties.
• To identify structural motifs responsible for material properties.
• To expedite the discovery of diverse bio-based polymers.

Main Methods:

• Developed PolyAGM, a ML algorithm utilizing graph kernel methods.
• Employed a "fingerprinting" technique to convert polymer graph representations into numerical vectors.
• Graphs encoded atomic, bonding, sequencing, chain length, and stereochemical information.

Main Results:

• PolyAGM accurately predicts thermal and mechanical properties of block copolymers.
• Predictions showed good agreement with experimental measurements.
• Identified key structural motifs influencing polymer properties.

Conclusions:

• PolyAGM effectively predicts properties of bio-derived ABA-block polymer thermoplastic elastomers.
• The general fingerprinting technique is applicable to other material science fields.
• ML significantly expedites the discovery and property identification of new bio-based polymers.