A hierarchical clustering method of hydrogen bond networks in liquid water undergoing shear flow

Many properties of water, such as turbulent flow, are closely related to water clusters, whereas how water clusters form and transform in bulk water remains unclear. A hierarchical clustering method is introduced to search out water clusters in hydrogen bonded network based on modified Louvain algorithm of graph community. Hydrogen bonds, rings and fragments are considered as 1st-, 2nd-, and 3rd-level structures, respectively. The distribution, dynamics and structural characteristics of 4th- and 5th-level clusters undergoing non-shear- and shear-driven flow are also analyzed at various temperatures. At low temperatures, nearly 50% of water molecules are included in clusters. Over 60% of clusters remain unchanged between neighboring configurations. Obvious collective translational motion of clusters is observed. The topological difference for clusters is elucidated between the inner layer, which favors 6-membered rings, and the external surface layer, which contains more 5-membered rings. Temperature and shearing can not only accelerate the transformation or destruction of clusters at all levels but also change cluster structures. The assembly of large clusters can be used to discretize continuous liquid water to elucidate the properties of liquid water.