Ultrahydrous stishovite from high-pressure hydrothermal treatment of SiO₂

Abstract

Stishovite (SiO₂ with the rutile structure and octahedrally coordinated silicon) is an important high-pressure mineral. It has previously been considered to be essentially anhydrous. In this study, hydrothermal treatment of silica glass and coesite at 350–550 °C near 10 GPa produces stishovite with significant amounts of H₂O in its structure. A combination of methodologies (X-ray diffraction, thermal analysis, oxide melt solution calorimetry, secondary ion mass spectrometry, infrared and nuclear magnetic resonance spectroscopy) indicate the presence of 1.3 ± 0.2 wt % H₂O and NMR suggests that the primary mechanism for the H₂O uptake is a direct hydrogarnet-like substitution of 4H⁺ for Si⁴⁺, with the protons clustered as hydroxyls around a silicon vacancy. This substitution is accompanied by a substantial volume decrease for the system (SiO₂ + H₂O), although the stishovite expands slightly, and it is only slightly unfavorable in energy. Stishovite could thus be a host for H₂O at convergent plate boundaries, and in other relatively cool high-pressure environments.

Silica is an archetypical system for Earth and materials science. Its complex polymorphism remains the subject of continuing study (1–3). At ambient to moderate pressures (up to 9 GPa), all forms of silica are built up of SiO₄ tetrahedra, with coesite the highest pressure polymorph of this type. At higher pressures, dense forms containing SiO₆ octahedra occur. Stishovite with the tetragonal rutile structure is stable between 9 and 50 GPa (4). Coesite and stishovite are believed to occur in silica rich parts of subducted oceanic slabs and crustal fragments in the Earth’s mantle (5). They are considered nominally anhydrous minerals (6), although stishovite may contain small amounts of hydroxyl in conjunction with Al incorporation (7, 8).

There are large kinetic barriers associated with transformations among silica polymorphs (9). Silica glass can be irreversibly densified above 15 GPa without transformation to coesite or stishovite (10, 11). Temperatures above approximately 1,000 °C were required to constrain the coesite-stishovite equilibrium phase boundary (12). If kinetic barriers could be lowered, new, intermediate, high-pressure forms of silica may become accessible (13, 14). Additionally, low temperatures can afford nanostructured forms of high-pressure silica phases (15). To explore routes for lowering kinetic barriers, we investigated a high-pressure hydrothermal environment. Experiments at 10 GPa on silica glass–water or coesite–water mixtures at various temperatures (see SI Text for details) produced stishovite containing unprecedented amounts of structural water.

Results and Discussion

The evolution of products during 8-h experiments using glass as starting material was as follows (Fig. S1). At 250 °C, the sample remained essentially amorphous; at 300 °C, coesite with about 13% stishovite coexisted with some residual glass; and at 350 °C, stishovite with about 11% coesite was obtained. Products at 400–550 °C were coesite-free stishovite.

Fig. 1 presents optical and scanning electron microscopy images for selected samples. Fig. 1A shows the polarized light micrograph of a partially transformed larger particle consisting of a coesite rim of about 20-μm thickness and a center of strained glass. The coesite rim shows undulatory extinction, suggesting it is composed of oriented micrometer-sized domains that grow inward from the surface. Stishovite occurs as fine, 0.5 to 1-μm-sized, euhedral tabular crystals. In the 350 °C product from glass starting material, stishovite crystals are peculiarly intergrown (Fig. 1B). The overall texture is most likely a result of the microstructure of coesite domains obtained initially from the hydrothermal transformation of the glass. Samples prepared at higher temperatures consist of homogeneous stishovite crystals (Fig. 1C). Previously intergrown crystals are now largely separated with the former intergrowth contacts clearly visible (Fig. 1D). In contrast, when using coesite starting material, stishovite crystals appear well sintered into agglomerates that have the shape and size of the original coesite particles (Fig. 1E). We conclude that the hydrothermal transformation of silica glass into stishovite proceeds via coesite, and that the microstructure of coesite determines the size and shape of stishovite crystals.

Fig. 1.

Photomicrograph of a partially glass-coesite transformed particle obtained at 300 °C in plane-polarized light (A, Left) and in cross-polarized light (A, Right). SEM images of stishovite crystals obtained at 350 °C (B) and 450 °C (C) using glass starting material (D) close-up of Fig. 1C. (E) SEM image of stishovite crystals obtained at 450 °C from coesite.

The powder X-ray diffraction (PXRD) patterns of hydrothermally formed stishovite show that reflections are shifted to lower Bragg angles (compared to dry stishovite, ref. 16), indicating a slightly larger unit cell volume (Fig. 2A, Table 1, and Table S1). Whereas the c lattice parameter remains largely unaffected, the a parameter increases by almost 0.5%. With increasing synthesis temperature, the a lattice parameter of anhydrous stishovite is approached. These changes, outside experimental error, strongly suggest a structural role for H₂O.

Fig. 2.

Experimental evidence for structural water in hydrous stishovite. (A) Shape and location of the 220 reflection in the PXRD pattern for various stishovite samples, (B) TGA (solid) and DSC (dotted) traces for dry stishovite (black) and hydrous stishovite 450-G (red). (C and D) IR spectra for dry stishovite (black) and hydrous stishovites 450-G (blue) and 450-Co (red). (C) Low energy IR with Si-O vibrations. The assigned wavenumbers refer to the three E_u modes. (D) High-energy IR with O-H vibrations. (E) ²⁹Si MAS-NMR spectra for dry stishovite and hydrous stishovite 450-G. The arrow marks the second Si site. ²⁹Si CP-MAS spectrum for 450-G (from top to bottom). (F) ¹H MAS-NMR spectra for 450-G, dry stishovite, and 450-Co (from top to bottom).

Table 1.

Preparation conditions, unit cell parameters, water content after TGA and SIMS and values for the enthalpies of drop solution (ΔH_ds) for various stishovite samples

Sample	Preparation (T, P, t, starting mat.)	a, Å	c, Å	V, Å3	Water content n in SiO2·nH2O	ΔHds, kJ per mol of SiO2.nH2O
Dry	1,000 °C, 10 GPa, 5 h, glass	4.1779	2.6657	46.53	0.007 (TGA)	2.86
						2.51
550-G	550 °C, 10 GPa, 8 h, glass, hy	4.1831	2.6647	46.63	0.030 (TGA)	—
450-Co	450 °C, 10 GPa, 8 h, coesite, hy	4.1895	2.6654	46.78	0.051 (TGA)	2.40
						3.02
450-G	450 °C, 10 GPa, 8 h, glass, hy	4.1953	2.6650	46.90	0.056 (TGA)	−1.31
					0.044 (SIMS)	−1.89

Estimated standard deviations for A and C are 0.002 Å or less; hy, hydrothermal.

Secondary ion mass spectrometry (SIMS) was performed on the sample obtained from glass at 450 °C (i.e., sample 450-G, see Table 1). Trace elements (B, Al, Na, Mg) did not exceed 60-wt-ppm and the H₂O content was 1.3( ± 0.1) wt %. Selected samples were subjected to thermal analysis (Fig. 2B and Table 1). Dry stishovite transforms exothermically into a glass at 550 °C. The small weight loss below 300 °C in all samples represents surface water. Hydrous stishovite 450-G decomposes at a lower temperature, 500 °C. Its decomposition is associated with a weight loss of 1.4 wt %, in agreement with the SIMS result, and we consider 1.3 ± 0.2 wt % to be the water content of 450-G. The presence of H₂O in stishovite in excess of 1 wt % is about three orders of magnitude higher than previously seen for Al-free stishovite and about one order of magnitude higher than observed for Al-bearing stishovite (7, 8).

The differential scanning calorimetry (DSC) curves show strongly exothermic decomposition of stishovite to glass, consistent with previous thermodynamic studies (17). There are some differences in the shape and area of the peaks between the hydrous and dry samples, but the heat effects are not readily quantified. A more accurate thermochemical approach utilizes high-temperature oxide melt drop solution calorimetry. Using an appropriate thermochemical cycle (Table S2), one can calculate the enthalpy of the reaction, at ambient temperature:

[1]

The calculated enthalpies of formation for samples 450-G and 450-Co are 7.3 and 3.0 kJ/mol, respectively. Assuming an error of ± 0.2 wt % for the water content, the values above will change by ± 0.4 kJ/mol. The two materials have similar water contents according to thermogravimetric analysis (TGA), but appear to have somewhat different enthalpies. Differences between the two hydrous stishovites became already apparent in the analysis of their morphologies (see Fig. 1 D and E) and PXRD patterns (see Fig. 2A). Further exploration of possible structural and energetic differences among differently prepared samples is the subject of future studies. The salient point of the present findings is that the incorporation of water into the stishovite has only a small energetic penalty. Although the stishovite expands slightly with the incorporation of water (Table 1), the volume change for reaction (1) is strongly negative (-0.7 cm³/mol at ambient conditions, and estimated to be -0.4 cm³/mol at the conditions of synthesis). Thus the pV term (integral of ΔVdp from 1 atm to high pressure) at room temperature and 10 GPa is estimated to be about -7 kJ/mol (values at higher temperature depend on the equation of state of water but will be roughly similar) and therefore can overcome the destabilizing enthalpy at atmospheric pressure. The contribution of entropy, TΔS, cannot be readily constrained, but it generally makes hydration less favorable with increasing temperature.

To shed further light into the nature of the incorporated water, spectroscopic investigations were performed on samples 450-G and 450-Co. IR spectra of hydrous and dry stishovite are profoundly different (Fig. 2C). In the hydrous material, bands below 1,000 cm^-1 corresponding to Si-O vibrations are slightly red shifted (perhaps reflecting increased Si-O distances). There is an intense sharp band at approximately 1,420 cm^-1, whose origin is unknown. Three broad bands between 2,500 and 3,500 cm^-1 may suggest O-H stretching (Fig. 2D). Whereas the location of the low-frequency band (near 2,650 cm^-1) is virtually identical, for the 450-Co sample the intermediate band (around 2,900 cm^-1) seems to split and the high-frequency band occurs at a lower wavenumber (3,322 vs. 3,389 cm^-1) with considerably higher intensity. The high-energy region of the IR spectrum of hydrous stishovite appears very different from that of nominally dry stishovite and Al-bearing stishovite with up to approximately 0.3 wt % H₂O. The IR spectra of the latter two are similar and characterized by an intense, broad, and anisotropic band at 3,111–3,134 cm^-1 (8, 18). We conclude that these pronounced changes in the IR spectra of hydrous stishovite also point to a unique structural role for water.

²⁹Si-magic angle spinning (MAS) NMR experiments were performed on dry stishovite and on the hydrous samples 450-G and 450-Co (Fig. 2E). The latter two spectra are virtually identical. The octahedral Si site in stishovite has a shift of -191.1 ppm (19, 20), which is also the dominant resonance for the hydrous samples. However, hydrous stishovite shows a second peak at -188.6 ppm. A cross polarization (CP) ¹H → ²⁹Si experiment utilizing polarization from the proton shows that the minor peak at -188.6 ppm is considerably enhanced over the resonance at -191.1 ppm, indicating that this Si site is associated with a proton environment.

The ¹H-MAS NMR spectrum of hydrous stishovite shows three groups of protons with chemical shifts of 10.5, 4.7, and 1 ppm (Fig. 2F). The resonance around 1 ppm is also observed for dry stishovite and is known from zeolites (21) and attributed to surface hydroxyl species (22, 23). With respect to the resonances at 10.5 and 4.7 ppm, the two hydrous samples show noticeable differences. The peak at 4.7 ppm is much more intense for 450-Co, and when compared to the 10.5 ppm resonance shows considerable motional narrowing. The broadening of the 10.5 ppm resonance is mainly attributed to strong proton–proton dipolar coupling which is consistent with a continuous narrowing of this resonance when higher MAS frequencies are applied. However, there are also heterogeneous contributions to the line width of the 10.5 ppm resonance, indicating chemical shift dispersion. ¹H/²⁹Si heteronuclear correlation experiments show that only the protons at 10.5 ppm are correlated with silicon (Fig. S2). The nature of the protons at 4.7 ppm, particularly present in the 450-Co sample that has sintered particles, is not yet clear. We hypothesize that the observed differences in the IR spectra, PXRD patterns, and enthalpies of formation between 450-G and 450-Co samples relate to those protons. Further experiments are necessary to clarify this phenomenon.

A 2D double quantum/single quantum (DQ/SQ) NMR correlation spectrum was collected for sample 450-G (Fig. S3) and shows a strong DQ signal on the diagonal for the 10.5 ppm resonance, indicating a clustering of these protons. DQ spinning sideband patterns were collected for this proton environment with two different excitation periods (Fig. S4). The H–H dipolar couplings thus obtained are 7.0 and 4.8 kHz, which correspond to proton–proton distances of 2.6 and 2.9 Å, respectively. We conclude that the protons at 10.5 ppm are clustered within the structure at a distance < 3 Å. Without other metals substituting for Si (e.g., Al), we suggest the mechanism of hydrogen incorporation is the hydrogarnet defect where a cluster of four hydroxyl groups ([OH^-]₄) replaces an entity . Although hydrogarnet defects have been established for tetrahedrally coordinated Si in grossular garnets (18), they are rare for minerals outside the garnet group and have never been reported in structures containing octahedrally coordinated Si.

Tetrahedral defects in hydrogarnets have been characterized structurally for silicon-free end members [e.g., Ca₃Al₂(OH)₁₂, Ba₃In₂(OH)₁₂] (24, 25). The four O atoms around a silicon vacancy are terminated as hydroxyl. The short hydroxyl O-H (0.91 Å) and long next nearest O…H distances (> 2.5 Å) are compatible with weak hydrogen bonding. We conjecture that the unique octahedral defect in hydrous stishovite represents a more complex bonding situation. If, analogous to the tetrahedral defect, four O atoms are terminated as hydroxyl, the two remaining ones will become formally underbonded and act as acceptors for strong hydrogen bonds. This situation is illustrated schematically in Fig. 3 where the known arrangement of protons in the tetrahedral defect has been imposed on the octahedral geometry, while observing reasonable interatomic distances for O-H, Si-H, and H-H (see SI Text for details). The ¹H NMR and IR spectra, however, suggest multiple H environments and thus a low symmetry. O-H stretching frequencies at 2,650 and 2,900 cm^-1 point to strong hydrogen bonding (18). Further insight into the structure and bonding situation of the proposed octahedral defect may be obtained from additional spectroscopic investigation, including deuterated samples, and computational modeling.

Fig. 3.

A simple, high-symmetry model of the octahedral hydrogarnet defect based on the established proton arrangement of the tetrahedral defect (see SI Text for details). Si, O, and H atoms are represented as light-gray, large dark-gray, and small dark-gray circles, respectively. (A) Unit cell enclosing a defect. (B) Si atoms surrounding a defect and underbonded O atoms involved in strong hydrogen bonding (thick, gray lines).

The unexpected discovery of hydrous stishovite shows the potential of hydrothermal environments at gigapascal pressures for creating new materials. There have been very few reports on the application of such environments at pressures approaching 10 GPa (26), yet these conditions may be significant in Earth and planetary settings. For example, conditions of 450–550 °C and 9–10 GPa can occur during the subduction of old, cold oceanic crust (the Tohoku subduction zone, ref. 27), and subducted mid-ocean-ridge basalts (MORB) can form free silica phases at high pressures (28). Under such low-T, high-P conditions, continuous dehydration reactions of relatively low-P minerals (e.g., serpentine) will provide H₂O which could then react with coesite to form hydrous stishovite (27). Rocks exhumed by natural processes from these environments have been found to contain coesite, and some are suggested to contain relicts of stishovite (29). If such stishovite is indeed hydrous, such a reaction path may provide an additional mechanism for bringing crustal H₂O into the mantle. Chung and Kagi (30) found about 10 times higher H₂O solubility in MORB than Al-bearing stishovite. Although they claimed trivalent cations might be responsible, the discrepancy is still difficult to understand by trivalent cations alone. The water incorporation found in this study for pure stishovite may help to explain such large H₂O solubility in MORB. Mosenfelder (31) reported water solubility in coesite to be minor at 1,200 °C, about a factor of 50 less than that found in stishovite at much lower temperatures in the present work. Thus it is not known whether coesite can contain and transport significant H₂O under low-temperature hydrothermal conditions. In a relatively cold subduction regime, a number of hydrated lower pressure phases may persist, stably or metastably, into the stishovite stability field, with their decomposition on heating providing sources of H₂O. Hydrous stishovite may provide a “bridge” at pressures of 9–12 GPa for the transport of H₂O downward into the P–T regime of other high-pressure phases containing significant H₂O (e.g., wadsleyite and other high-pressure hydrous magnesium silicates). Because of the low temperatures involved, transport of water and its transfer among mineral assemblages is likely to be controlled by kinetics as well as thermodynamics.

Furthermore, the rich variety of P–T environments inferred for the ever-growing number of exoplanets may include rather cool high-pressure environments which, in the presence of water, may result in hydrothermal conditions similar to those studied here. Finally, the octahedral hydrogarnet defect should be explored as a general hydrogen storage mechanism in other nominally anhydrous silicates containing octahedral silicon in Earth and planetary interiors.

In conclusion, we have shown that high-pressure hydrothermal treatment of SiO₂ catalyzes the coesite–stishovite transition and produces hydrous forms of stishovite. At pressures near 10 GPa, stishovite is observed at temperatures below 350 °C, with significant yields above 400 °C. Unique to these results is the amount of water incorporated (> 1 wt %), the substitution mechanism via unprecedented octahedral hydrogarnet defects, and the very modest (< 10 kJ per mole of SiO₂) energetic destabilization associated with the observed H₂O in stishovite. There needs to be future detailed characterization of the solubility, structural state, and thermodynamics of water in stishovite as a function of temperature and pressure, and investigation of the possible occurrence of the hydrogarnet-like substitution in other minerals containing octahedral silicon (e.g., silicate garnet, perovskite, postperovskite phases).

Materials and Methods

In a typical synthesis, 55–65 mg of silica (glass or coesite with a particle size distribution of about 2–200 μm) and 25–35 mg water (i.e., the molar ratio was roughly 1∶1) were sealed in noble metal capsules. The capsules were pressurized in a multianvil device (32) and subsequently heated to a temperature between 250 and 550 °C at a rate of 20 °C/ min. After equilibrating samples at their target temperature for 8 h, the temperature was quenched and the pressure released over a period of 11 h. Recovered capsules were pinched open and the water removed by evaporation. The SI Text contains further details on the synthesis procedure and the analysis of products by PXRD, optical and scanning electron microscopy, SIMS, thermal analysis (TGA/DSC), drop solution calorimetry, IR and solid-state NMR spectroscopy. Additionally, parameters for the structure model presented in Fig. 3 are given.