Results and Discussion

The starting material for our experiments was a mixture of amorphous SiO₂ and goethite (α-FeOOH) in the molar ratio of 4:1, with goethite serving as a source of water. About 2 wt % Pt black powder was added to the mixture for infrared laser absorption in laser-heating experiment (Methods has more details). The sample assembly consisted of the SiO₂–goethite mixture sandwiched between 2 layers of amorphous SiO₂ (Methods). The sample was first compressed to ∼30 GPa (Run 1) at room temperature. We used both ruby and Pt as internal pressure standards (Methods has details). The sample was then heated using a double-sided laser system and held at target temperatures for ∼10 min each: 1,240 K at 32.5 GPa (27 GPa), 1,450 K at 33.1 GPa (34 GPa), and 1,775 K at 34 GPa (37 GPa), where Pt pressures are in parentheses. In situ X-ray diffraction (XRD) spectra were acquired at each pressure and temperature condition and also, after quenching to ambient temperature. Subsequently, the sample was compressed to 39 GPa (Run 2) at room temperature and then heated in a different location to temperatures of 1,310 K at 48.3 GPa (37 GPa), 1,410 K at 49 GPa (52 GPa), and 1,835 K at 52 GPa (58 GPa). The diffraction data show that our heated sample was dominated by stishovite with a minor amount of Fe₂O₃ and FeOOH in the high-pressure ε phase (Fig. 1). At high pressure and 300 K, we find that sharp diffraction lines agree well with the expected diffraction pattern of anhydrous stishovite (31), except that they are systematically shifted to higher d spacings. Unit cell parameters acquired at 300 K and high pressure are summarized in Fig. 2 and Table 1. The much larger volumes of stishovite synthesized under hydrous conditions relative to volumes of anhydrous stishovite calculated from its equation of state at the same pressures (29, 31, 32) provide primary evidence for considerable hydration.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

XRD patterns at 34 and 52 GPa (ruby pressures) and 300 K after laser heating. At the center of the heated area, the majority of diffraction peaks belong to stishovite. The transition from stishovite to CaCl₂-type SiO₂ was not observed in our experiment. Data were acquired using an X-ray wavelength of 0.4066 Å.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Measured stishovite unit cell volumes at high pressure and 300 K. Solid squares (Run 1) and solid circles (Run 2) show pressure based on the ruby pressure calibration [the error bar based on stress anisotropy (51)], whereas open squares and open circles show pressure based on the Pt pressure calibration (the error bar of ±2 GPa). The solid black curve shows the equation of state for anhydrous stishovite of Andrault et al. (31). The solid green curve shows the equation of state of stishovite with 3.2 wt % H₂O from Nisr et al. (34). The dashed green line is an estimated equation of state for stishovite with 10 wt % H₂O on the basis of a linear extrapolation of the bulk moduli for the anhydrous stishovite and stishovite with 3.2% H₂O (31, 34). The decompression data are shown in SI Appendix, Fig. S3.

Spektor et al. (18, 33) and Nisr et al. (30) reported on the synthesis of hydrous Al-free stishovite samples grown at 9 and 10 GPa under hydrothermal, low-temperature conditions (350 °C to 550 °C). The water content of their samples was calibrated based on thermogravimetric analysis and secondary ion mass spectrometry, with reported water contents of up to 3.2 wt % H₂O. These authors suggest H incorporation into SiO₂ through an octahedral “hydrogarnet”-type defect mechanism with hydroxyl groups around a silicon vacancy, equivalent to direct cation substitution of the form 4H⁺ → Si⁴⁺ (18, 30). Spektor et al. (33) provided a calibration for the water content of stishovite on the basis of the observed volume change relative to anhydrous stishovite in samples quenched to 1 bar and 300 K. To make an initial estimate of the water content in our samples, we assume that the slope of the observed linear relationship at 1 bar is constant with pressure and use the equation of state of Andrault et al. (31) to determine the volume of anhydrous stishovite at high pressure. On this basis, the changes in unit cell volume observed in our study would indicate about 4.5 to 10 wt % H₂O in stishovite at high pressure (Table 1). However, this approach may not be reliable if the compression behaviors of hydrous and anhydrous stishovite are different. Nisr et al. (34) measured the equation of state of hydrous stishovite with ∼3.2 wt % H₂O, which is shown relative to the equation of state of anhydrous stishovite on Fig. 2. The higher compressibility of hydrous stishovite indicates that the change in volume at higher pressures with the addition of H₂O is expected to be less than at 1 bar as shown in Fig. 2. Thus, our estimates based on the 1-bar data would be minimal. On the basis of a linear extrapolation of the difference in observed bulk moduli from these studies (31, 34) to higher water contents, our samples could contain as much as 10 wt % or more H₂O (Fig. 2).

When using the ruby pressure calibration, where the ruby was off center to the heated area, the results apparently indicate that the water contents in stishovite are very high (e.g., 10 wt % H₂O or more) but decrease systematically with temperature. However, when using the Pt pressure calibration, where the Pt is in the same location as the X-ray spot, the data fan out nearly parallel to a compression curve, possibly indicating synthesis of stishovite with a relatively constant but still very high water content (Table 1). The remarkably high water content predicted for our stishovite samples based on the volume change may be explicable when considering that other rutile-structured high-pressure hydrous phases, like δ-AlOOH and Phase H (MgSiH₂O₄), contain ∼15 wt % H₂O in their structures. We note that Phase H becomes stable at ∼35 to 40 GPa in the MgO–SiO₂–H₂O system at the temperatures in our study (13). Whether or not the stishovite we observe is a fully hydrated stoichiometric phase or a high but limited solid solution remains an open question that we cannot address further with our data.

The extremely high amounts of water estimated for the stishovite samples in this study based on their expanded unit cell volumes are much higher than observed in any previous study (summarized in SI Appendix, Table S1). For example, the stishovite samples synthesized at ∼9 to 10 GPa and 723 K by Spektor et al. (18) (∼1.3 wt %) and Nisr et al. (30) (∼3.2 wt %) and at 15 GPa and 1,500 °C by Bolfan-Casanova et al. (14) (72 ppm) in the pure SiO₂ system have much lower water contents, especially at high temperature. Previous results have indicated that the incorporation of alumina (<2.3 wt % Al) in stishovite promotes the substitution of water in stishovite through the cation substitution mechanism (Al³⁺ + H⁺ → Si⁴⁺), which includes the Al-bearing stishovites synthesized at 10 GPa and 1,200 °C by Pawley et al. (35) (82 ppm for 1.51 wt % Al₂O₃), 15 GPa and 1,400 °C by Chung and Kagi (36) (844 ppm for 1.32 wt % Al₂O₃), ∼32 GPa and 2,577 °C by Panero et al. (37) (∼460 ppm for 3 wt % Al₂O₃), and 20 GPa and 1,400 °C by Litasov et al. (17) (3,010 ppm for 4.4 wt % Al₂O₃).

Stishovite is not expected to be a stable phase in peridotite bulk compositions either in ambient mantle or in the mantle portion of subducted lithosphere (14, 38, 39). However, in subducted basaltic compositions, like midocean ridge basalt (MORB), stishovite becomes a modally abundant phase at transition zone depths where it coexists with majorite garnet. At lower mantle conditions, stishovite comprises ∼23 modal % in MORB, coexisting with aluminous bridgmanite (∼22 modal %), Na–Al phase phase (∼35 modal %), and Ca–ferrite-structured phase (∼20 modal %) (17). Among these nominally anhydrous phases in basaltic compositions at lower mantle pressures, stishovite stands out as a phase that can potentially contain a large amount of water while remaining stable along the mantle geotherm. In contrast, bridgmanite can accommodate perhaps 2 orders of magnitude less water (∼1,000 ppm water) (40), and there are no data on the water storage capacities of the Na–Al phase and Ca–ferrite phases.

Dense hydrous phases can also potentially be stable in basaltic compositions, like Phase D in the transition zone and upper lower mantle or Phase H in the midlower mantle to the base of the mantle (15). However, these phases are not stable along a mantle geotherm and would only stabilize at lower temperatures in cold slabs after the nominally anhydrous phases, like stishovite in oceanic crust, exceed their storage capacity (13). Because of its potential to store exceptionally large amounts of water at high temperatures, stishovite is potentially the most important host of water in basaltic compositions in the transition zone and lower mantle.

Subducted oceanic crust is considered to be largely dehydrated in the upper mantle (20). However, water may be subducted into the transition zone in the peridotitic lithosphere in postserpentine dense hydrous phases, like Phase A, superhydrous Phase B, or Phase D (41). These phases may break down in the transition zone if subducting lithosphere founders and heats up. The discovery of hydrous ringwoodite as an inclusion in “superdeep” diamond (42) as well as boron-rich diamonds originating from around 700-km depth (43) indicates that at least locally water-enriched regions of diamond formation may be linked to slab dehydration. Furthermore, seismic velocity anomalies observed in the uppermost lower mantle have been ostensibly attributed to the melting caused by dehydration of hydrous ringwoodite when it breaks down to nominally anhydrous bridgmanite and ferropericlase (6) or alternatively, due to melting caused by breakdown of Phase D or Phase H (13). These observations indicating release of water from subducted lithosphere via dehydration or hydrous melting also suggest that stishovite, even if it were effectively anhydrous in oceanic crust that had previously dehydrated, could become an important carrier for water into the lower mantle on rehydration (Fig. 3).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Transporting water into the lower mantle by ultrahydrous SiO₂. Dehydration of basaltic crust produces melting in the mantle wedge beneath island arcs and effectively dehydrates the crust. However, colder slabs may transport water in subducted mantle lithosphere to transition zone depths where dehydration occurs. Infiltration of fluids or melts through the crust may rehydrate stishovite in the crust to form ultrahydrous stishovite, which can then be transported into the deep lower mantle.

Water or hydrous melts from dehydration of ringwoodite or dense hydrous silicates in peridotitic subducted lithosphere in the transition zone or the upper lower mantle may infiltrate subducted basaltic oceanic crust where they can be accommodated in hydrous stishovite. With further subduction into the lower mantle, stishovite would remain stable, even if the slab heats up to the mantle geotherm. Stishovite undergoes a transition to CaCl₂-structered phase at ∼70 GPa along a mantle geotherm, and it is not known if this phase can also accommodate abundant water (28). In addition, Lakshtanov et al. (5) suggested that the rutile–CaCl₂ transition pressure in stishovite with ∼5 wt % Al₂O₃ can be as low as ∼24 GPa. However, as the transition is second order resulting from a small distortion of the unit cell from tetragonal to orthorhombic with only a minor volume change (44), we predict it may also accommodate significant H through protonation.

As a modally abundant phase in subducted oceanic crust, stishovite has the potential to transport a significant quantity of water into the deep mantle. Assuming current values for the average subduction rate (5 cm y⁻¹), length of subduction zones (46,000 km), and crustal thickness (6 km) (37), we estimate that, for an abundance of 23% in a subducted MORB, 4.6 × 10²² kg of stishovite would be subducted into the deep mantle per year. Although we cannot constrain the water content in deeply subducted oceanic crust, with the addition of each 1 wt % H₂O to stishovite, subducted oceanic crust could transport ∼10¹¹ kg of water per year into the deep mantle. To put that quantity into perspective, current estimates for the water content of the mantle are about 0.75 oceans masses based on geochemical arguments (45) or about 10²¹ kg. This entire quantity of water could be transported to the mantle by stishovite over 4 billion y of subduction if the average stishovite water content was ∼2 wt %. If water in stishovite is partially released during the transportation, stishovite is requested to contain water more than ∼2 wt % at the start to reach the geochemical estimation (45).

In conclusion, our results indicate that stishovite can accommodate weight percent levels of water in its structure. As a modally abundant phase in subducted oceanic crust, stishovite is a potential carrier of significant quantities of water into the deep earth even at the high temperatures along a mantle geotherm. Our results have not yet quantified precisely the storage capacity of water in stishovite in oceanic crustal assemblages, and more work is needed to understand how water partitions among coexisting phases as well as the thermodynamic and physical properties of hydrous stishovite (e.g., seismic velocities) as a function of temperature, pressure, and minor element concentrations (Al³⁺, Fe³⁺).