Nucleation of Ice in Large Water Clusters: Experiment and Simulation

Summary

Experimental studies of water in greatly confined spaces carried out at the University of Michigan are reviewed. In particular, measurements of rates of homogeneous nucleation of ice in large clusters of water probed by electron diffraction are discussed. Nucleation rates were astronomically higher than any previously observed in the laboratory. Measurements of rates permit inferences to be drawn about interfacial free energies of the ice-water boundary. Diffraction patterns also show that the phase of ice formed when supercooling is deep is the metastable cubic ice. This is because the interfacial free energy for the cubic ice boundary is lower than that for the stable hexagonal phase. Moreover, it is shown that very finely divided water can be cooled substantially below the temperature at which bulk water has been proposed to freeze catastrophically. Possible reasons for small drops avoiding such a critical point are proposed. Molecular dynamics simulations of large, crystalline and deeply supercooled liquid clusters were carried out with a variety of potential functions. They indicated that, despite the disorder found in the surface layers of the crystalline clusters, this disorder was not responsible for the nonideal profiles of the Bragg reflections seen in experiments. Simulations show promise in the field of nucleation. Fully realistic simulations of the freezing of water would be much more enlightening than the traditional nucleation experiments because of the detailed accounts of the underlying cooperative molecular motions they would afford. Such simulations have proven to be elusive, partly because of the enormous demands on computer times involved. Even with advances in computer technology showing signs of overcoming that obstacle, it is not clear that a suitable interaction potential function is available for the purpose. Steps that may be necessary to resolve the problem are discussed briefly.