Low-Temperature Phenomena in Aqueous Systems

Abstract

Water, the most abundant liquid on earth, is also one of the least understood. Theexistence of a liquid-liquid critical point (LLCP) in the deeply-supercooled regime is a leading explanation for many of them. The molecular mechanisms underlying these behaviors occur at length and time scales which are often inaccessible experimentally. This work examines them with large-scale, atomistic molecular dynamics simulations. It explores how microscopic properties such as bond flexibility and polarizability in water and amino acid sequences of proteins influence such macroscopic properties as ice crystallization and the LLCP. An LLCP is rigorously located in WAIL, a water model parameterized using ab initiocalculations only, and incorporating realistic bond flexibility and polarizability. The existence of a critical point in WAIL provides strong support to the view that the LLCP is a robust feature in the free energy landscape of supercooled water. Previous models shown to contain an LLCP did not permit bond flexion or polarization despite their known importance. Classical nucleation theory is used to compute the homogeneous nucleation rateof ice Ih in the TIP4P/Ice model at conditions ranging from ambient to the vicinity of the LLCP. Supercooling was found to be the dominant influence on nucleation rate, but at high supercoolings the Widom line causes the appearance of a locus of maxima with regard to pressure. The Widom line affects nucleation rates primarily through the ice-liquid surface tension. Recent advances in protein structure prediction have made possible a largerreference dataset and more general genetic algorithm for optimization of antifreeze proteins (AFPs) than that used by Kozuch et al. A neural network trained on the expanded AFP data set and used to optimize four AFPs predicts significant increases in thermal hysteresis. The binding surface is also demonstrated. Many questions remain about supercooled water. The LLCP can only bedefinitively shown with experiments. The quantitative description of nucleation in TIP4P/Ice contains approximations and empiricisms. Synthesizing and purifying proteins with high predicted thermal hysteresis and incorporating them into the reference dataset would more clearly define what is possible with AFPs, as would incorporating data on ice-nucleating proteins.