The phase diagram of water/ice
and a new metastable phase of ice

The ice phase diagram is extremely rich, comprising 11 confirmed crystalline phases, in which the water molecules link through hydrogen bonds to form tetrahedral frameworks. The structures and stabilities of many of these phases have been established definitively by means of neutron powder-diffraction in collaboration with Prof. Werner Kuhs, University of Göttingen, Germany. This work is of importance to a large interdisciplinary group of researchers interested in the hydrogen bond, while the versatility of the water molecule in forming so many different structures is relevant to its biological importance.

15 years ago, little was known about the structures of ice under pressure, most information in the literature being obtained from samples 'recovered' to ambient pressure at liquid nitrogen temperatures. In the early 1980s, the high-pressure facilities at ILL, together with the Rietveld powder refinement technique, had developed to a stage where the structures of most of these phases could be studied under their conditions of stability. So began our collaboration which has, through neutron work both at ILL and ISIS, not only sorted out a good part of the phase diagram (Fig. 1), but has also led to new, unexpected results of wide significance.

The earliest work using clamped high-pressure cells found that ice VIII was antiferroelectrically ordered with non-bonded oxygen-oxygen contacts shorter than the bonded distances. Although this might seem a strange result, it confirmed beautifully other ideas which stressed the importance of non-bonded repulsions in determining hydrogen bonded structures in general. In addition to the expected hydrogen disorder, ice VII was shown to have oxygen disorder. Again, this was initially unexpected, though it has been found more generally since, e.g. in the disordered medium pressure phase ice VI. The lower pressure part of the phase diagram was probed using the He-gas cell. This raised many experimental difficulties: ice V was almost impossible to form and ice III was very difficult. The problem was resolved when it was discovered that the helium gas was stabilising a previously unknown He-hydrate with a water molecule topology identical to that of ice II - another unexpected result and new structure.

The work on ice III and IX revealed that, in contrast to earlier assumptions, both ices are partially (dis)ordered. Very recently, full sets of data on ices III and V under various conditions of temperature and pressure have been obtained, thus ending a long-standing uncertainty about H-ordering in these phases. Ice III and ice V both show partial ordering in the 20 to 30% range even close to the melting point.

Figure 1: The phase diagram of water/ice. Inset: The medium pressure range showing the melting curves of metastable ices IV and XII.

Figure 2: The H-bond framework of tetragonal ice XII viewed down the c-axis. The spacegroup is I42d, lattice constants are a = 8.304 Å and c = 4.024 Å.

However, the phase diagram is still not fully understood. On several occasions during the last 15 years, powder lines have been seen that could not be identified with any known ice or clathrate phase. As we have pinned down with increasing precision the preparation conditions for these phases, we have begun to back them into a corner. The first success has been ice XII, a totally new structure that we have found within the stability region of ice V and which was prepared by crystallisation from the liquid phase. The topology of ice XII is unlike any of the known ice phases, and contains a mixture of 5 and 7 membered rings. The inset in Fig. 1 shows the tentative stability region and Fig. 2 the structure clearly exhibiting the 5 membered rings organised to form channels along the unique axis.


Figure 3: The H-bond framework of rhombohedral ice IV showing the auto-clathrate arrangement with H-bonds passing through the centre of 6 membered rings.

Another metastable phase of ice, ice IV (Fig. 3), discovered in the 1930s by Bridgman, was obtained in situ in our experiments for the first time by following a slightly different preparation recipe. The density of ice V (1.402 for D2O) is smaller than the densities of ice IV and XII (1.436 and 1.437 resp.), which are quite similar to each other. Both ice IV and XII are fully hydrogen disordered, while ice V is partially ordered as mentioned above. On the other hand, differences occur in the degree of hydrogen-bond bending: compared to ice V and XII the structure of ice IV shows distinctly smaller bending, yet exhibits interpenetration of the H-bond framework, a phenomenon sometimes referred to as auto-clathration. Clearly there are two ways of increasing the density in water structures: additional hydrogen bond bending as in ice V and XII or hydrogen bond interpenetration as in ice IV and also in the next highest pressure phase, ice VI. At present, ice XII is the densest known phase of the water substance without interpenetration. Yet in all these structures the non-bonded repulsive constraints are active and confirmed by our neutron results.


Two lines of further research are now developing from these findings. First, the enhanced richness of the ice phase-diagram in the medium pressure range is an excellent demonstration of the versatility of the water molecule that enables the building of a variety of hydrogen-bonded structures sometimes in very close competition for occupying the same region of p-T space. The seemingly delicate balance of enthalpic and entropic contributions to the total energy will thus allow us to test very critically the viability of water potential functions used widely in computer calculations in chemical and biomolecular systems. Secondly, the fact that we have formed metastable phases directly from the liquid makes the water system now an excellent candidate for studies of metastability, including both thermodynamic and kinetic aspects of phase formation.

This work has also demonstrated that neutron diffraction is by far the best method to explore the phase diagram of ice as it allows the detection of topological phase changes and any sudden or continuous changes in H-ordering. At the same time, information on expansivities and compressibilities is obtained which gives us further quantitative information on the details of the chemically and biologically important water-water intermolecular interaction.