Data-driven analysis of dynamical heterogeneity in polymer melts near surfaces

Polymer melts near solid substrates exhibit heterogeneous dynamics, which has gained significant attention in recent decades. In this study, we propose a data -driven method, referred to as unsupervised learning and environment similarity (ULES), to identify the particles that significantly contribute to the dynamic heterogeneity (DH). This method involves clustering the dynamic features of particles obtained from molecular dynamics simulations and categorizing their dynamic patterns. By employing ULES, we can effectively differentiate and characterize the dominant particles ("slow"particles) responsible for DH. Additionally, we investigate the dynamic disparities between the confined polymer system and the bulk system by quantifying the environmental similarity, thereby discerning the adsorption, transition, and bulk regions. Furthermore, we employ the ULES method to explore the influence of interface interactions and molecular weight on DH. Through our study, we unveil that those "slow"particles play a key role in driving DH in the confined system. Moreover, we establish a correlation between DH and interface interactions, particularly within the adsorption region. Conversely, in the transition region and the bulk region, the molecular weight gradually emerges as the dominant factor influencing the dynamics of the system. Our proposed ULES method introduces a novel approach for analyzing the dynamic environment of complex systems and comprehending the dynamic properties of individual particles. By combining unsupervised learning techniques and environment similarity analysis, we gain valuable insights into the intricate dynamics of polymer systems near solid substrates. This research not only advances our understanding of DH but also provides valuable guidance for designing and optimizing polymer -based materials and processes in various industrial applications.