Connecting Structural Characteristics and Material Properties in Phase-Separating Polymer Solutions: Phase-Field...

Area of Science:

• Materials Science
• Polymer Science
• Computational Materials Science

Background:

• Morphology formed during phase separation critically influences material properties, such as in functional membranes.
• Predicting this morphology is challenging due to complex molecular interactions.
• The Cahn-Hilliard equation with Flory-Huggins free energy models polymer solution phase separation.

Purpose of the Study:

• To analyze parameter influence on morphological evolution in a 2D polymer solution.
• To develop a physics-informed neural network (PINN) for inverse parameter prediction.
• To determine the key parameter governing phase transition and domain growth.

Main Methods:

• Systematic sensitivity analysis of parameters (volume fraction, mobility, polymerization, surface tension, Flory-Huggins interaction) on morphological evolution.
• Development of coupled feedforward neural networks (PINNs) to represent phase-field equations.
• Inverse prediction of parameters using PINNs and renormalization with characteristic domain size change.

Main Results:

• Polymer-solvent affinity (Flory-Huggins interaction parameter) is identified as the most influential parameter.
• PINNs accurately determine unknown parameters from just two morphological snapshots.
• Parameter errors within a certain tolerance do not significantly impact morphology during domain growth.

Conclusions:

• Physics-informed neural networks offer an efficient method for inverse design in phase-separating systems.
• Accurate prediction of polymer-solvent affinity is achievable with reduced computational load.
• The approach simplifies inverse design for specific material property requirements.