Water is not a simple compound — it exhibits many anomalous physical behaviours that defy adequate explanation. Any fresh information on the structure of water in its various condensed forms is therefore welcome. Writing in Nature, Tulk et al.1 report a study of water under high pressure. They find that it passes through a sequence of crystalline phases rather than forming an amorphous solid, as had been reported by previous studies.
The melting point of ordinary crystalline ice decreases with increasing pressure. This observation inspired a landmark study in 1984, which sought to determine whether such ice would ‘melt’ when compressed at low temperatures to form a solid that has a disordered molecular structure resembling that of liquid water2. Indeed, the study showed that ice compressed at 77 kelvin collapses into a dense form known as high-density amorphous (HDA) ice, which can be recovered at low temperatures under ambient pressure. Remarkably, when heated at ambient pressure, HDA ice transforms into a low-density amorphous (LDA) form instead of reverting to its original crystalline state3.
Measurements made under conditions of successive compression and decompression have shown that the change in volume associated with the interconversion between HDA ice and LDA ice is discontinuous, and that the transition between these two forms of ice is reversible and does not seem to involve the formation of any intermediate phases4. The observations suggest that the interconversion might belong to a class of process known as thermodynamically first-order transitions. If so, this could have important consequences for the phase diagram of water, which relates the temperatures and pressures at which thermodynamically distinct phases of water occur.
The details of the phase diagram of water are not yet fully understood. One possibility is that the boundary between the HDA ice and LDA ice phases extends into, and terminates in, a region of the diagram where water is supercooled (a phase in which water is liquid, despite being below its freezing point). The end of the boundary is known as a critical point. Above the critical point, water would be a mixture of two distinct liquids that have different densities. A feature of this ‘two-liquid’ model is that compressed ice would form two amorphous solid phases of very different densities that are related to the two liquid waters5. Intense experimental and computational efforts have been made to find evidence to support the two-liquid model, including proof of the existence of different amorphous phases in compressed ice.
In general, when a crystalline solid is compressed under ‘hydrostatic’ conditions that allow thermodynamic equilibrium to be reached, it is expected to transform into another crystalline phase. The formation, instead, of a metastable amorphous phase suggests that an energy barrier has inhibited the transformation of the solid into the second crystalline structure. Such a barrier can be breached if the solid is compressed slowly, which gives time for the structure to relax and for thermodynamic equilibrium to be attained6. In this scenario, the formation of the amorphous phase would be described as a kinetic effect, because it depends on the amount of time that is given for a transition to occur.
Ice and the minerals α-quartz and berlinite were the archetypal examples of crystalline solids that become amorphous under pressure. The latter two compounds, however, are now known to transform into crystalline structures when compressed under uniform (isotropic) pressure7,8. The pressure-transmitting media used to compress α-quartz and berlinite isotropically are incompatible with water at high pressures; no other suitable pressure transmitter has been available. One study9 generated quasi-hydrostatic conditions using a pressure chamber known as a double-sided dynamic diamond anvil cell, and observed the effects on water both with and without a silicone-based pressure-transmitting medium. Ice was seen to change from one crystal form to another before its amorphization, but observations under true hydrostatic conditions were still desirable.
Tulk et al. have finally observed the anticipated crystal–crystal transformation of compressed ice under hydrostatic conditions, using neutron diffraction. The authors encapsulated deuterated water in a gasket, which was cooled using boiling liquid nitrogen and placed inside a compression apparatus known as a Paris–Edinburgh press. (Deuterated water contains a heavy isotope of hydrogen, rather than the most abundant isotope, and was used to enhance neutron scattering.) The sample was then compressed in incremental steps, and allowed to rest for an hour at each pressure.
The resulting diffraction patterns show that HDA ice did not form. Instead, the sample sequentially transformed into a crystalline phase known as ice IXʹ at 3–7 kilobar, then another crystalline phase (ice XVʹ) above 10 kbar, and finally into a mixture of ice XVʹ and a third crystal form (ice VIIIʹ) above 30 kbar (Fig. 1). These low-temperature phases correspond, respectively, to crystalline phases known as ice III, ice VI and ice VII, which are observed at ambient temperature.
Tulk and colleagues’ experiments unequivocally show that, under equilibrium conditions, a ‘normal’ sequence of crystalline transformations occurs in ice. The HDA ice that was observed in previous studies therefore formed as a result of kinetic effects. Tulk et al. were able to observe the crystal–crystal transformations in part because they used a larger sample size than did previous experiments, but also because the long periods between the compressions in their experiments enabled thermodynamic equilibrium to be established, and reduced non-uniformity in the stress that was exerted on the samples during compression.
So, is HDA ice similar in structure to liquid water? There are many lines of evidence to consider. A previously reported analysis of the thermodynamics of ice amorphization showed that the observed pressure at which HDA ice forms from normal ice at low temperatures is much higher than the pressure at which normal ice would be expected to melt at the same temperatures10. HDA ice has been found to transform into an even denser amorphous form when heated11, which shows that it is a kinetic product of rapid compression. And a neutron- and X-ray diffraction study of ice decompressed at constant temperatures found that the transformation of HDA ice into LDA ice involves several intermediate amorphous forms, and is therefore not a first-order process12. Computational studies of the molecular dynamics of ice at constant pressure have also been used to reproduce the main characteristics and behaviours of amorphous ice6. A theoretical analysis of data from one study suggests that the transformation of ice into HDA ice is not thermodynamically driven, but is instead the result of a type of instability (known as a mechanical or elastic instability) that occurs in certain solids13. This result has been confirmed by experiment14.
The above evidence, taken together with Tulk and colleagues’ studies, shows that the structure of HDA ice is not related to that of liquid water. It is a transitional phase that lies between the phases of ice that contain single networks of molecules that are connected by hydrogen bonds (normal ice and ice IXʹ) and the ice phases that consist of interpenetrating hydrogen-bond networks (ice XVʹ and ice VIIIʹ). The substantial reconstruction that is needed to convert single networks into interpenetrating ones requires a large energy barrier to be overcome. At low temperatures, compressed ice therefore initially settles into an amorphous state, but can eventually overcome the energy barrier to form ice VIIIʹ. Thus, the structure of HDA ice is most likely to be a distorted form similar to that of ice XVʹ.
In light of Tulk and co-workers’ results, the assumption that the existence of two amorphous ices that have different densities (HDA ice and LDA ice) supports the two-liquid model of water must be reconsidered. Determining where the boundary between LDA ice and HDA ice fits into water’s phase diagram, and whether it extends into and terminates in a ‘no-man’s land’ region of the diagram that has not yet been accessed experimentally15, and in which only crystalline ices are expected to form, will be the next experimental challenge.
Nature 569, 495-496 (2019)