Identifying structural signatures of shear banding in polymer nanopillars

Recent studies (Schoenholz, 2016) have used machine learning methods to identify a field, ''softness," that characterizes the local structure around particles and is highly correlated with particle-level dynamics in glassy materials. While softness is a good predictor of local particle rearrangements, its ability to predict large length and timescale phenomena such as material failure remains unknown. Using machine learning methods, we identify mesoscale defects that lead to shear banding in confined polymer pillars well below the glass transition temperature. We successfully apply these methods to pillars of diameters of 12.5, 25, 50, and 100 particle diameters. Our results show that the primary structural features responsible for shear banding on this scale are small, particle sized fluctuations in the pillar diameter. Interestingly, we note the importance of mean softness in identifying shear band planes grows as a function of pillar diameter suggesting these features play a large role in material failure for bulk-like systems. We find that shear band planes are softer in the pillar’s interior but less soft on their boundaries than other planes prior to deformation.