A metastable limit for compressed liquid water

Abstract

The transformation of liquid water to solid ice is typically a slow process. To cool a sample below the melting point requires some time, as does nucleation from the metastable liquid 1 , so freezing usually occurs over many seconds 2. Freezing conditions can be created much more quickly using isentropic compression techniques, which provide insight into the limiting timescales of the phase transition. Here, we show that water rapidly freezes without a nucleator under sufficient compression, establishing a practical limit for the metastable liquid phase. Above 7 GPa, compressed water completely transforms to a high-pressure phase within a few nanoseconds. The consistent observation of freezing with different samples and container materials suggests that the transition nucleates homogeneously. The observation of complete freezing on these timescales further implies that the liquid reaches a hypercooled state 3. Computational studies suggest that freezing can occur on 0.1-1 ns timescales, although for water such simulations require a highly confined geometry 4 and/or strong electric fields 5,6. Unconfined simulations of supercooled water 7 indicate that freezing is possible on 100 ns timescales, many times faster than experimental observations. Simply cooling a liquid on that timescale is challenging: 10 7 −10 10 K s −1 cooling rates can be achieved by spraying droplets into a cryogen 8 , but it is difficult to carry out real-time measurements. The fastest real-time observation of freezing in expansion-cooled water clusters occurred on 10−30 μs timescales 9 , leaving a 2-3 decade gap between the experimental and computational studies of freezing. Adiabatic compression is an alternative route to solidification, even though liquids become hotter in the process. Temperature increase can be mitigated by using isentropic (rather than single shock wave) compression techniques, yielding the coldest possible adiabatic state. As shown in Fig. 1, isentropic compression of liquid water crosses the melt line between 2 and 3 GPa (T ≈ 400 K). Although compression freezing involves a different portion of the phase diagram than cooling (ice VII (ref. 10) rather than ice Ih), freezing conditions are created very quickly, providing insight into the limiting phase-transition timescales. When liquid water is isentropically compressed above 2 GPa in the presence of a quartz or fused-silica window, freezing will be observed over 10-100 ns timescales 11,12. The phase transition quickens with increasing pressure, but only in the presence of a silica window. Even at 5 GPa, where the liquid is nearly 70 K below the equilibrium melt line, no freezing is observed during compression within sapphire windows (≈800 ns experiment duration). Solidification is characterized by two basic events: the onset and the completion of freezing. The onset of freezing is defined by the time needed to create freezing conditions (whether by cooling

Melting dynamics of ice in the mesoscopic regime

“…Thermoreflectance (22,23) does not contribute to our signal because of the absence of highly reflecting interfaces, and its transient signal would, however, be concluded in ∼1 ns. The transmission loss due to scattering at ice/water interfaces has been used to detect the freezing of water into ice VII under multiple-shock experiments (6,24), using white light. Here, with monochromatic probes, we can estimate the size and concentration of the scattering domains evolving after the pulse.…”

“…The molecular mechanisms of melting and the characteristic timescales of the process are not well elucidated, as for any other phase transition. In fact, great effort is presently being made with state of the art experimental techniques to understand the intrinsic kinetics of structural transformations, particularly under dynamic compression, by which phase-transition boundaries up to extreme pressures (P) and temperatures (T) can be accessed (1)(2)(3)(4)(5)(6). Melting is the least hindered of phase transitions, requiring the disruption of the crystalline order and the achievement of the local structure of the liquid phase.…”

High pressure phase transformation in iron under fast compression

“…As new methods are being developed and new experimental regimes are being explored, fresh puzzles however continue to emerge on the behavior of materials undergoing phase transitions under conditions of very rapid compression. Recently, the quasi-isentropic uniaxial loading of liquid water across its ice VII phase boundary [18,19] yielded dynamic features resembling Van der Waals loops, which as of yet are not fully understood from a fundamental point of view. In this paper we present fast compression experimental results on the solid -solid, α to ǫ phase transformation of iron which, surprisingly, exhibit similar characteristics with the ones observed in the liquid to solid rapid quasi-isentropic quench of water.…”

“…Common approaches rely on simple representations of the phase transformation rates, e.g. depending linearly on the difference between the chemical potentials of the two phases [28], which reproduce some of the dynamic behavior observed in shock experiments but cannot explain more complex experimental features such as the negative acceleration loops recorded in these or the water experiments [18,19]. We adopt here the phenomenological but physical picture proposed by Kolmogorov and others [3], where well defined domains of the growing phase increase their size at the expense of the parent phase through the motion of an infinitely thin interface.…”

Ramp wave loading experiments driven by heavy ion beams: A feasibility study

“…Unlike a shock wave experiment where a single point on a shock adiabat is obtained, in RWL experiments, a continuous set of data points is recorded and the solid state of a sample is ensured up to high pressures. The RWL method was also shown to be a more sensitive tool for studying the dynamics of the ultrafast structural phase transformations than the shock-wave based methods (Smith et al, 2008;Bastea et al, 2005;Dolan et al, 2007). RWL has been demonstrated with different drivers such as magnetic pulse loading using high-current pulsed power generators with typical loading times of 100 ns (Reisman et al, 2000;Hayes, 2001;Cauble et al, 2002;Hayes et al, 2004;Rothman et al, 2005;Davis, 2006), gas guns (Chhabildas & Barker, 1986) and high explosives (Barnes et al, 1974), with graded density impactors (1 ms) and highpower lasers (10 ms) (Lorenz et al, 2004(Lorenz et al, , 2005Swift & Johnson, 2005;Smith et al, 2007).…”