Determining glass transition in all-atom acrylic polymeric melt simulations using machine learning.

Area of Science:

• Polymer Science
• Materials Science
• Computational Chemistry

Background:

• Polymer functionality is often dictated by their glass transition temperature (Tg).
• Accurate Tg determination in simulations typically relies on macroscopic property changes, sensitive to fitting protocols.
• Previous work introduced a data-driven method using microscopic details from coarse-grained models.

Purpose of the Study:

• To demonstrate the generalizability of a previously proposed data-driven Tg determination method.
• To apply and validate the approach on an atomistic model of acrylic polymers.
• To elucidate the specific roles of side chain and backbone structures in influencing Tg.

Main Methods:

• Utilized dimensionality reduction techniques.
• Employed clustering algorithms for data analysis.
• Applied the data-driven method to atomistic molecular dynamics simulation data of acrylic polymers.

Main Results:

• Successfully demonstrated the generality of the data-driven Tg determination approach across different methods.
• Validated the method's applicability to atomistic polymer models.
• Identified and quantified the distinct contributions of side chain and backbone residues to the glass transition temperature.

Conclusions:

• The proposed data-driven method offers a robust alternative for determining Tg from simulation data.
• The approach effectively leverages microscopic details for precise Tg calculation.
• Side chain and backbone dynamics play explicit, separable roles in the glass transition behavior of acrylic polymers.