Water-carbon dioxide solid phase equilibria at pressures above 4 GPa

A solid phase in the mixed water-carbon dioxide system, previously identified as carbonic acid, was observed in the high-pressure diamond-anvil cell. The pressure-temperature paths of both its melting and peritectic curves were measured, beginning at 4.4 GPa and 165 °C (where it exists in a quadruple equilibrium, together with an aqueous fluid and the ices H2O(VII) and CO2(I)) and proceeding to higher pressures and temperatures. Single-crystal X-ray diffraction revealed a triclinic crystal with unit cell parameters (at 6.5 GPa and 20 °C) of a = 5.88 Å, b = 6.59 Å, c = 6.99 Å, α = 88.7°, β = 79.7°, and γ = 67.7°. Raman spectra exhibit a major line at ~1080 cm−1 and lattice modes below 300 cm−1.


Methods
Water was loaded along with a bubble of air into a modified Merrill-Bassett diamond-anvil cell (DAC). After sealing, the DAC was subsequently immersed in liquid carbon dioxide at 10 °C and 58 bar, briefly re-opened to let CO 2 displace the bubble, and re-sealed. Gaskets were made of rhenium.
The DAC was placed in an oven with windows in front and back. Temperatures were measured to 1 °C with type K thermocouples located next to the diamonds. Pressure was determined from the Raman-scattered light from one or more pieces of cubic boron nitride (cBN) included in the load. Scattering was induced by 20 mW of 488 nm laser light, focused through the front window by a microscope objective of 0.28 numerical aperture, back-scattered radiation being collected through the same objective. The light was sent to a 0.3 m monochromator with an 1800 lines/mm grating and dispersed onto a CCD camera. To maintain the precision of the pressure measurements, the grating of the monochromator used for this purpose was kept at a fixed position. The frequency, v, of the cBN transverse optical phonon was assumed to vary with temperature and pressure as 11,12 : c P  cT c T  c   1  2  3  2  4  5  2  6 with c 1 = 3.303 cm −1 · GPa −1 , c 2 = 1.85 × 10 −4 cm −1 · GPa −1 · K −1 , c 3 = −9.72 × 10 −3 cm −1 · GPa −2 , c 4 = 5.28 × 10 −3 cm −1 · K −1 , c 5 = −2.94 × 10 −5 cm −1 · K −2 , and c 6 = 1054.96 cm −1 . The precision of pressure measurements was usually 0.1 GPa although, for certain pressures and temperatures, the cBN line was partially overlapped by the (weaker) sample lines seen at 1040 cm −1 (Fig. 5) and the precision thereby degraded. Since this occurred only infrequently, for thick loads coupled with smaller pieces of cBN, these data could be discarded.
When Raman spectra of the water-carbon dioxide mixture were to be recorded a (100% reflective) pick-off mirror was inserted into the path of the collected light so as to direct the Raman scattering into a second monochromator (0.25 m, 1200 lines/mm). The slits of this monochromator were set to give a resolution of 5 cm −1 (full width at half maximum). The time required to accumulate a suitable spectrum was typically 10-20 minutes.
X-ray diffraction data were taken on line 12.2.2 of the Advanced Light Source (Berkeley) with a monochromatic beam of 25 keV. The beam had a cross-section of 20 μm × 20 μm, allowing it to be focused on single crystals previously identified as the S3 phase through Raman spectroscopy. The diamond supports allowed scattering to be observed over a circular aperture of ±35° around the center-line of the DAC. Data were taken at room temperature.
Phase transitions were determined by first bringing the DAC to a desired starting pressure, then slowly raising the temperature while monitoring changes in pressure. As is usual in the quasi-isochoric DAC, the initiation or completion of a phase transition causes an observable discontinuity in the slope of the pressure-temperature trace (inset, Fig. 1). After completion of a transition, the pressure could be increased and another run performed with the same load. In Fig. 1, points of phase transition observed over several different runs, and three different DAC loads, are collected in a single plot along with lower pressure data from two previous studies.   solid CO 2 . Temperature-composition plots of these phase boundaries are shown in Fig. 3, with the four univariant (three phase) lines A-D appearing as points at the selected pressures.

Results and Discussion
A photomicrograph of crystals of the new phase, grown out of a water-rich fluid along curve "A", is presented in Fig. 4. In the course of this run, both crystals "a" and "b" were at first clear; crystal "b" subsequently darkened in patches, presumably by trapping a solution which precipitated a micro-crystalline aggregate of the H 2 O(VII)-S3 eutectic. Allowing such a "darkened" crystal to sit for a few hours leads to lightening along the rims which are in contact with solution, as dissolution and re-growth take place. In our experiments, euhedral crystals of S3 could be grown from the fluid phase repeatedly and reversibly. In contrast to the report of Wang et al., our observations clearly extend into the regime of thermodynamic stability of the new phase.
Single-crystal X-ray diffraction of S3 revealed a triclinic system with unit cell parameters (at 6.5 GPa and 20 °C) of a = 5.88 Å, b = 6.59 Å, c = 6.99 Å, α = 88.7°, β = 79.7°, γ = 67.7°, and a volume of 246.5 Å 3 . Note that possible crystal structures resulting from the calculations of refs 6 and 7 are all orthorhombic or monoclinic. So far, it has not been possible to experimentally determine the exact crystal structure. In samples with large fractions of melt, S3 crystals could be seen to sink, while H 2 O(VII) crystals rose. For S3 to be denser than H 2 O(VII) at the conditions of growth (1.72 g·cm −3 ) 13 , and assuming its formula is indeed H 2 CO 3 , there must be at least 5 molecules per unit cell for a density of 2.10 g·cm −3 . If the crystal consists of pairs of hydrogen-bonded molecules, then an even number per unit cell would suggest six molecules for a density of 2.52 g·cm −3 .
In our experiments, S3 was initially noticed when raising the temperature caused the solid contents of the cell to turn black in transmitted light. In reflected light, however, the contents were seen to be white, indicating the cause of the opacity to be the production of many small crystals rather than the absorption of light; this is in agreement with the observations of Wang et al. The transition between S3 and a solid mixture of water and CO 2 is sluggish, preventing a clear determination of the position of the equilibrium boundary; the dashed line in Fig. 2 is a reminder that such a boundary must exist, connecting to the quadruple point, Q. When heating at pressures above 4 GPa, S3 can be seen to grow in as temperatures reach roughly 150-160 °C (filled triangles in Fig. 2); however, this is clearly due to having surpassed a kinetic barrier, as the reaction is not reversible (subsequent reduction of the temperature to 20 °C does not result in the disappearance of S3). Wang et al. report that S3 decomposes at room temperature somewhat below 2.4 GPa, and we have similarly observed decomposition to occur at 85 °C between 3.8 and 3.3 GPa. This, again, may be the result of overcoming a kinetic barrier, as the metastable extension of curve B (Fig. 2) is approached and production of fluid (with immediate freezing into the solid phases of water and CO 2 ) becomes possible.
When initially loaded (before heating for the first time) our samples produced Raman signals (Fig. 5, trace a) which matched exactly those known for CO 2 (I) 14 , with no other lines in the monitored range below 1500 cm −1 . The appearance of the distinctive S3 Raman features (described below) is associated with a decrease in the intensity of the solid CO 2 contributions (up to a complete disappearance for water-rich loads). Decomposition of S3 (when pressure was lowered) was evidenced by a disappearance of the S3 features and a reappearance of those of solid CO 2 .
The Raman spectrum of the new phase is distinguished by a strong, characteristic line located at 1084 cm −1 (at 7.6 GPa), with a pressure derivative of ~5 cm −1 · GPa −1 (Fig. 5, traces d and e), and a weaker line at ~820 cm −1 . Strong Raman scattering also occurs below 300 cm −1 , presumably from lattice modes. The relative intensities of the lines vary in strength with the orientation of the crystal probed. As noted by Wang et al., the spectrum bears a reasonable resemblance to that of carbonic acid produced as a cryogenic surface deposit (trace b) 4, 15 . The weaker line at ~820 cm −1 , visible in the current crystal spectra, was not observed by Wang et al. (trace c) and, conversely, two small peaks at 640 and 680 cm −1 , previously noted, were not observed from our samples, perhaps due to the orientations of the crystals. Lines shift systematically with pressure but were not affected significantly by differences in temperature over the range explored in our experiments.
Also shown in Fig. 5 is a spectrum from the fluid phase (trace f). A line appears at 1040 cm −1 , ~40 cm −1 red-shifted from the strong S3 line and broadened, as well as a weaker line at ~650 cm −1 . While it is tempting to consider these to be from dissolved molecules of the same species forming S3, it is also possible that at least the stronger of the two should be ascribed to the bicarbonate ion. A spectrum containing lines of both bicarbonate and carbonate 16 in a dilute aqueous solution is shown in trace (g); the spectral shift from the line in (f) may be due to the high concentration of CO 2 (~30 mole%) in the latter solution.
Should S3 be crystalline H 2 CO 3 , as suggested by Wang et al., our observations would then establish for this compound a regime of high-pressure thermodynamic stability, as predicted by simulations 6,7 . Furthermore, the wide pressure range under which S3 was observed by Wang et al.,up to 25 GPa, might now also be presumed to represent equilibrium conditions, in agreement with simulations (Saleh and Oganov predicted H 2 CO 3 to be stable as a solid between roughly 1 and 44 GPa at 27 °C, with a subsequent polymorph stable up to at least 400 GPa). It should be borne in mind that none of the simulated structures matches the triclinic system of our samples; however, discovery of the true crystal structure within the reportedly complex energy landscape of H 2 CO 3 7 may have been prevented by small errors in calculation. As well, the calculations of Saleh and Oganov at pressures less than 10 GPa used only 4 formula units of H 2 CO 3 , which, as shown by the combination of our x-ray and density data, is less than the number required for a primitive unit cell.

Conclusion
Experiments on mixed H 2 O-CO 2 samples up to 7.6 GPa evidence the existence of a new solid phase (S3). Above a quadruple point at 4.4 GPa and 165 °C, the new phase can be observed along both its melting and peritectic curves, distinct from the known melting curves of H 2 O and CO 2 ices at the same pressures; it is found to be in equilibrium with the fluid up to the highest investigated pressure. Raman spectra of S3 exhibit a strong line at ~1080 cm −1 and weaker one at ~820 cm −1 , in addition to a complex of lattice modes below 300 cm −1 . In the fluid, close to the melting curve of S3, another strong line appears at ~1040 cm −1 ; whether this is properly attributed to dissolved molecules of S3, corresponding to the 1080 cm −1 line of the solid, or rather to the bicarbonate ion, is not yet clear.
S3 crystallizes in a triclinic system with unit cell parameters (at 6.5 GPa and 20 °C) of a = 5.88 Å, b = 6.59 Å, c = 6.99 Å, α = 88.7°, β = 79.7° and γ = 67.7°. The optical behaviour, spectroscopic features and range of occurrence of S3 are consistent with those reported recently by Wang et al. for a new solid phase observed at room temperature between 2.4 and 25 GPa and tentatively identified as crystalline carbonic acid. While these experimental The sharp mode at ~1080 cm −1 is characteristic of the phase. Lattice modes are also prominent. (f) In solution, a line appears at ~1040 cm −1 , shifted ~40 cm −1 to the red and about twice as broad as the nearby S3 line; this may be associated with dissolved molecules of the S3 phase, but may otherwise be due to the bicarbonate ion. (g) Lines from bicarbonate and carbonate ions in a dilute aqueous solution 16 .
observations are consistent with calculations suggesting that solid H 2 CO 3 is thermodynamically stable at pressures of several to tens of GPa, the observed triclinic form was not predicted.
Our updated picture of the H 2 O-CO 2 system reveals an intricate succession of phase transitions likely to govern the behaviour of these species in the crust of terrestrial planets and the hydrosphere of icy worlds. Below 1 GPa, and at temperatures less than 21 °C, H 2 O and CO 2 associate preferentially as hydrates, with mutual fluid solubilities not exceeding a few percent. Above 4.4 GPa, and at higher temperatures, lies a domain of stability of the putative H 2 CO 3 , in a regime of high solubilities and evolving speciation. H 2 CO 3 may be stable at pressure to at least tens of GPa, covering a wide range of conditions in planetary interiors. Between 1 and 4.4 GPa, only pure H 2 O and CO 2 ices have been confirmed as thermodynamically stable solids, and the speciation and miscibility behaviour of the binary system remain largely unknown. Experimental investigation of this behaviour will be necessary to assess geochemical processes involving these two major planetary volatiles.