Prediction of Surface Free Energy of Polymer Surfaces by Machine Learning Modeling of Chemical Structure–Property Relationships

The wettability of liquids on polymer surfaces is an
important
issue in basic science and applications, and the ability of polymers
to repel both water and oil with low surface free energy is considered
useful for developing antifouling low-adhesion surfaces. Understanding
the quantitative structure–property relationship for the wettability
of polymer materials is essential for designing such polymer surfaces.
Machine learning modeling can quantitatively analyze factors strongly
correlated to the target properties of materials. In this study, a
machine learning model for contact angle values for water and diiodomethane
and surface free energy is presented to clarify the quantitative structure–property
relationship for the wettability. This strategy not only predicts
contact angles of the two types of liquids but also makes surface
free energy predictable as a thermodynamic parameter from the molecular
structure of polymers, where the surface free energy can be calculated
using a theoretical model, the Owens–Wendt equation. A series
of polymer brush surfaces with diverse chemical structures, including
polymers with alkyl, silicone-like, and fluorinated side chains, were
synthesized by surface-initiated atom transfer radical polymerization,
and their wettability and surface free energies were evaluated. A
data set was constructed using values for 72 types of polymer brush
surfaces and bulk polymers collected from both our experiments and
literature sources. The decision tree-based models were suitable to
predict the water contact angle, diiodomethane contact angle, and
surface free energy. The trained decision tree-based models were interpreted
using Shapley additive explanation analysis and utilized to simulate
the effect of the number of CH2 and CF2 groups
in the side chains of hypothetical polymers. This machine learning
model is expected to aid in designing water- and oil-repellent polymers
and help with directly predicting the surface free energy from the
chemical structure.