Molecular Simulation and Machine Learning Analysis on Glass Transition Temperature (Tg) Variability in Polymer Thin Films

Extensive efforts, spanning both experimental and simulation methods, have been devoted to understanding the underlying physics of the Glass Transition Temperature (Tg) in polymer thin films. Despite these endeavors, a definitive picture of thin film glass transition behavior remains elusive, largely due to challenges in separating out the effect of substrate and discrepancies arising from different methodologies. Seeking to clarify the conditions under which Tg of polymer films starts to deviate from the bulk values, and to shed lights on the intrinsic mechanisms of glass transition droppage, we employed molecular dynamics simulations on both supported and freestanding films. We meticulously tracked the mobility of individual chains across temperature variations, and by combining a machine learning approach, we enhanced our ability to pinpoint the transition from glass to liquid state. Moreover, we assess the distinct contributions from different film regions, especially those proximal to interfaces versus the film’s center, to delineate their impact on the overall Tg.