Machine Learning Driven Calculation of Hydrogen Bonding Inside Densely Grafted Polyelectrolyte Brush Layers

Recent all-atom molecular dynamic (MD) simulations have revealed behavior and properties of the counterions and the water molecules supported by the polyelectrolyte (PE) brushes. Such efforts, however, are limited by the fact that very often a generic definition is used to describe the properties of water and ions inside the brush layer. For example, while defining the water-water hydrogen bonds (HBs) inside the brush layer, it is often disregarded that these HBs will behave differently between locations outside and inside the brush layer since water connectivity is severely disrupted inside the brush layer. Here we address this gap and present the use of an unsupervised machine learning (ML) approach, which is based on clustering algorithm and uses the all-atom MD simulation generated equilibrium coordinates of the water molecules as input, to predict the water-water HBs inside cationic and anionic PE brush layers. Our calculations enable us to (1) compare the clusters formed inside and outside the brush layer and identify the corresponding disruption of the hydrogen bonding inside the brush layer and (2) to quantify the possible PE-brush-induced confinement-driven changes to the average "hydrogen – acceptor-oxygen – donor-oxygen" angle that defines the HBs.