Sampling rare events in hybrid statistical physics quantum mechanical studies of hydrophobic interactions in liquids: A gpu implementation

Abstract

A method for sampling rare events throughout hybrid statistical physics quantum mechanical studies of hydrophobic interactions in liquid phases was analyzed and implemented on GPU architecture. The rare events considered in the present case are related to formation of the so-called dangling X-H bonds of small polar solvent molecules around notably non-polar solute molecules. The approach is based on detecting various populations of polar solvent molecules which reside in the first solvation shell around the non-polar solute and noting the differences between such populations. In particular, the search for structurally different populations of solvent molecules around the solute is started by detecting the differences in vibrational spectroscopic characteristics of the X-H stretching frequencies, which could be computed by various approaches, such as, e.g. within the Stark shift approximation. The results show x25 overall speedup of the GPU implementation compared to the sequential implementation.