Machine learning-aided analysis for complex local structure of liquid crystal polymers.

Area of Science:

• Materials Science
• Computational Chemistry
• Polymer Science

Background:

• Understanding mesoscopic structures is crucial for advanced material development.
• Molecular dynamics simulations offer microscopic insights but analyzing complex structures is challenging.
• Liquid crystal polymers (LCPs) are of interest for sensors and soft actuators due to their unique structures.

Purpose of the Study:

• To develop a Machine Learning-aided Local Structure Analyzer (ML-LSA) for classifying complex mesoscopic molecular structures.
• To apply ML-LSA to liquid crystal polymer (LCP) systems for accurate structural classification.
• To identify optimal order parameters for distinguishing mesogenic structures.

Main Methods:

• Developed a Machine Learning-aided Local Structure Analyzer (ML-LSA).
• Trained a machine learning (ML) model on computationally inexpensive monodomain LCP trajectories.
• Applied the ML-LSA to large, complex quenched LCP structures.

Main Results:

• The ML model accurately distinguished nematic- and smectic-like monodomain LCP structures.
• ML-LSA successfully classified complex local structures in quenched LCPs as nematic- or smectic-like.
• The study identified a key order parameter for differentiating these mesogenic structures.

Conclusions:

• ML-LSA enables automatic and systematic analysis of mesogenic structures without prior knowledge.
• The developed tool overcomes the challenge of manually determining order parameters for complex structures.
• This approach facilitates the optimization of advanced materials, including LCPs for sensors and actuators.