Regression analysis for predicting the elasticity of liquid crystal elastomers

Area of Science:

• Materials Science
• Polymer Science
• Computational Materials Science

Background:

• Understanding structure-property relationships in soft materials like liquid crystal elastomers (LCEs) is challenging.
• Microscopic molecular details significantly influence macroscopic material properties, particularly deformations.

Purpose of the Study:

• To identify design variables in LCE molecular architectures that govern macroscopic deformations.
• To develop a predictive model for LCE stress-strain behavior based on molecular design.

Main Methods:

• Utilized machine learning (ML) regression analysis on a database of coarse-grained molecular dynamics simulations for LCEs.
• Trained a surrogate model using LCE molecular architecture descriptors and simulation conditions as input, and stress-strain curves as output.
• Validated the ML model's predictive performance on novel LCE molecular architectures.

Main Results:

• Identified specific molecular descriptors that significantly govern the stress-strain behavior of LCEs.
• The ML-generated surrogate model accurately predicted stress-strain curves for previously unseen LCE architectures.
• Predicted curves demonstrated strong agreement with results from direct molecular dynamics simulations.

Conclusions:

• Machine learning regression analysis is effective in uncovering key molecular design principles for LCEs.
• The developed ML scheme can accelerate the exploration and design of LCE materials.
• Predicting LCE deformations from molecular architecture using ML offers a powerful tool for materials discovery.