Introduction
Internal melt figures, first reported by Reference TyndallJohn Tyndall (1858), form in ice and other substances (Reference McLaurin and WhalleyMcLaurin and Whalley, 1988) when they melt internally. In the lab-oratory, internal melt figures have been formed in ice by infrared irradiation, dielectric heating and rapid adiabatic compression (Reference Kaiser and MagunKaiser and Magun, 1964; Reference Baumann, Bilgram and KänzigBaumann and others, 1984). Melting occurs internally by conduction of heat to nucleation sites from the surrounding superheated ice. For slow rates of energy input, the figures form in the shape of flat disks or hexagonal plates with a vapor bubble present, whereas for rapid energy input figures are highly dendritic. Compression melt fractures (Reference Knight and KnightKnight and Knight, 1972), i.e. flat liquid-filled disks without a vapor bubble, can also form by rapid energy input. Here, we report clear observations of dendritic melt figures and compression melt fractures formed by adiabatic compression and we note the possible importance of this method to the study of deep-focus earth-quakes.
Apparatus
A high-pressure vessel (Reference Gagnon, Kiefte, Clouter and WhalleyGagnon and others, 1988) was used hydrostatically to pressurize small samples (~ 150 mm3) cut from single crystals and bicrystals of ice grown from filtered de-ionized and degassed water. A microscope and video camera were used to record the behavior of the specimens through one of the windows (2.2 mm diameter) of the pressure vessel. A hand-pump fitted with a Heise pressure gauge was used to pressurize the vessel up to 2 kbar in ~300 bar increments at rates up to ~ 1.2 kbar s−1 per stroke.
Observations
During pressurization of a typical mono-crystalline specimen, tiny surface scratches and asperities from the sample preparation quickly melted when the phase-boundary pressure was reached. Etch pits, which also quickly melted away, were sometimes observed on the surface as the critical pressure was attained. Further application of pressure resulted in nucleation of dendritic fern-like melt figures at the surface that grew into the ice at a rate proportional to the rate of pressure application (Fig. 1). Figures were stable in size when pressure was static though the faceted and sharp-edged surface features became rounded as heat conducted from the surrounding ice and resulted in a slight thickening of the figure just after each pressure increment with the hand-pump. Relieving the pressure reversed the growth process and caused the figures to shrink at a rate proportional to the rate of decompression. Advancing dendrites were very thin (~20 μm). They grew primarily parallel to the basal plane and they would often have offshoots in planes normal to the basal plane (Fig. 2) and occasionally at other angles to the basal plane.
Pressurization of bicrystals with air bubbles (< 40 μm) at the grain boundary exhibited a similar behavior. Melting at the boundary would occur at the critical pressure and liquid would fill the air-bubble cavities. As the pressure increased, dendritic figures would then nucleate from the pockets of liquid and air, and grow into the ice primarily in the basal plane of each crystal. The same behavior was exhibited by a plane of tiny air bubbles (< 40 μm) in a single crystal, produced by the passage of a thin warm wire.
Decompression caused the melt figures to shrink and disappear, sometimes leaving behind many tiny (< 2μm) spherical water inclusions (also noted by Reference Baumann, Bilgram and KänzigBaumann and others (1984)), resulting from entrapment of pressurized liquid water in the refreezing process. These outlined the shapes that figures had prior to decompression (Fig. 3). Oil inclusions, much larger in size, were present in the ice in the areas where the figures opened at the surface of the sample. Re-application of pressure caused figures to re-appear at the same sites but their shapes were altered due to disruptions in the crystal lattice from the previous growth and refreezing.
Disk-shaped, liquid-filled compression melt fractures were also created by the adiabatic pressurization. In contrast to the dendritic figures, these nucleated in the crystal matrix when a certain degree of superheating had been reached and rapidly grew to a maximum size within the time span of one video frame (1/30s), independently of the rate of pressurization. These thin (< 20 μm) fractures then healed at the periphery to some extent, thereby shrinking and thickening slightly to a stable smaller size within the span of a few video frames. Once formed, the fractures grew again along the healed plane upon the further application of, and in proportion to, pressure until dendritic growth initiated at the periphery when its former maximum size was reached at the end of the healed zone (Fig. 4). Compression-melt fractures shrank in proportion to the rate of decompression, sometimes leaving behind planes of tiny liquid-water inclusions (as did the dendritic figures) and oil inclusions if the fracture had propagated to the surface of the specimen and permitted the entry of pressure fluid.
Conclusions
The observation of internal melt figures, resulting from pressurization of ice, is important to the study of ice/structure impacts and indentation where pressures are high and change rapidly. Many other crystals such as gallium, germanium and silicon have melts that are denser than the solid phase and should also develop internal melt figures as a result of pressurization near the melting temperature. Refreezing of surface melt during decompression was found to be a useful method for controlled dendritic growth of ice from liquid. This could be a useful tool in the study of snowflake growth. There are striking similarities between the present observations and the nucleation and growth of spinel (possibly superplastic) inclusions in olivine which cause faulting during mechanical loading of metastable Mg2GeO4 (Reference Burnley, Green and PriorBurnley and others, 1991). It has been postulated that the same mechanism operating in down-going slabs of lithosphere causes deep-focus earthquakes (Reference Burnley, Green and PriorBurnley and others, 1991). The ice I → II transformation in pressurized samples of ice jacketed with soft indium, which yields a replica of the transformation on the sample’s surface after testing, has been used as an analogue for geophysical processes related to deep-focus earthquakes (Reference Kirby, Durham and SternKirby and others, 1991). The present study suggests that the transparent property of ice could be a powerful additional tool for dynamically visualizing the role of phase transformations in deep-focus earthquakes by observing the ice I → liquid transformation, and possibly the ice I → II and ice I → III solid-solid transformations, in single crystals and at grain boundaries under various loading conditions.
Acknowledgements
We thank Dr E. Whalley for helpful discussions and encouragement at the beginning of this project.
The accuracy of references in the text and in this list is the responsibility of the authors, to whom queries should be addressed.