1 Introduction

The transportation process of water into the deep mantle is important for understanding the structure and dynamics and the chemical evolution of the Earth because the physical and chemical properties such as melting phase relations, rheology, electrical and thermal conductivity of minerals and rocks in water-bearing systems can be significantly different from those of the dry system (e.g., Thompson 1992; Wang et al. 2006; Kohlstedt 2006; Marzotto et al. 2020). The water solubility of nominally anhydrous minerals and stability of hydrous minerals are therefore essential to clarify the water transport process in the mantle. The discovery of water-rich minerals such as hydrous ringwoodite, superhydrous phase B, and hydrous phase egg as diamond inclusions suggest that water is transported at least into the mantle transition zone (Wirth et al. 2007; Pearson et al. 2014; Tschauner et al. 2018). On the other hand, water transport and its distribution in the lower mantle are poorly understood. Oceanic island basalt contains more water than mid-ocean ridge basalt, suggesting water-rich sources in the lower mantle (Hirschmann 2006). However, major minerals in lower-mantle peridotitic assemblages such as bridgmanite and ferropericlase have generally limited water solubilities less than 0.1 wt% (e.g., Bolfan-Casanova et al. 2002; Fu et al. 2019).

High-pressure silica polymorphs of stishovite and CaCl2-type phase are major constituents of basaltic crusts, sediments, and continental crustal rocks in the lower mantle (e.g., Irifune et al. 1994; Hirose et al. 1999; Ishii et al. 2012, 2019a, b, 2022a). Recent phase relation studies in SiO2–H2O systems reported that these silica phases can accommodate water at the weight percent levels (Lin et al. 2020; Nisr et al. 2020) although other studies showed limited water contents up to several hundred ppm (Litasov et al. 2007; Purevjav et al. 2024). Other studies examined their water storage capacities in systems containing Al2O3, an important component in the silica-rich rocks mentioned above, and found that the water incorporation mechanism is different from that in an Al-free system and their water contents increase with temperature. The increases are due to water being stored as the AlOOH component, and the mixing entropy increasing its solubility with temperature (Litasov et al. 2007; Ishii et al. 2022b). Ishii et al. (2022b) showed that the water content in CaCl2-type phase reaches 1.1 wt% at a temperature of 1900 °C, which is even higher than the average mantle (Katsura 2022). The stability of CaCl2-type phase at pressures up to ~ 120 GPa (Murakami et al. 2003) allows the silica-enriched rocks that make up the upper layers of the subducting slab to transport water into the deep lower mantle.

Although aluminous silicas are important water hosts in the lower mantle, subducting slabs can have lower temperatures than the average mantle (Thompson 1992; Ganguly et al. 2009), which may limit their water solubility. An aluminous hydrous mineral, hydrous phase δ with compositions of (Al, Fe, Si, Mg)OOH, is considered to be another lower-mantle water carrier because it is stable at pressures down to the bottom of the lower mantle (e.g., Sano et al. 2008; Duan et al. 2018; Strozewski et al. 2023). A recent phase relation study of a hydrous basalt system reported that stishovite coexists with hydrous phase δ at 25–26 GPa, the pressures at the top of the lower mantle, and at relatively low temperatures of 1000–1200 °C (Liu et al. 2019). Further studies reported that hydrous phase δ is stable at relatively low temperatures up to 1400 °C and pressures of 24–28 GPa in SiO2–Al2O3–H2O and MgSiO3–Al2O3–H2O systems (Ishii et al. 2022b, c). Although the presence of hydrous phase δ increases the water storage capacity in the silica-enriched rocks, it could limit the AlOOH component and therefore the water storage capacities in the silica minerals. Thus, the solubilities of the AlOOH component in the silica phases coexisting with hydrous phase δ are needed to evaluate the bulk water storage capacity in the silica-enriched rocks.

In this study, we determined the water solubility in stishovite coexisting with hydrous phase δ in a SiO2–Al2O3–H2O system under water-undersaturated conditions using a multi-anvil press in combination with electron probe microanalysis and Fourier transform infrared spectroscopy (FTIR) after recovery.

2 Methods/experimental

2.1 Starting materials

To measure water content by FTIR, it is necessary to grow stishovite crystals at least to 50 microns due to the limited spatial resolution of conventional FTIR spectrometers. However, it is very difficult to grow stishovite crystals coexisting with hydrous phase δ because the synthesis temperature is relatively low and the coexisting phase hampers grain growth. For this reason, the water contents of stishovite coexisting with hydrous phase δ were determined using pre-synthesized stishovite single crystals annealed at a target pressure and temperature in stishovite-hydrous phase δ matrix.

Our preliminary experiments showed that the alumina content in stishovite coexisting with hydrous phase δ is 0.68–0.81 wt% at the same conditions as this study (Ishii et al. 2022b). Based on this result, single crystals of aluminous stishovite were synthesized at 22 GPa and 1900 °C for 2 h using a powdered mixture of reagent-grade SiO2 quartz and Al(OH)3 gibbsite at a molar ratio of 99:1 including ~ 0.85 wt% Al2O3. The recovered stishovite single crystals have dimensions up to 200–400 μm and an alumina content of 0.64(3) wt% as determined by electron microprobe analysis (Table 2 and Additional file 1: Table S1). The lower alumina content in stishovite may be due to presence of hydrous Al–Si melt (Litasov et al. 2007). The water content was estimated to be 994(44) ppm as an average value using 10 randomly oriented crystals by unpolarized FTIR based on Paterson’s (1982) calibration. Single crystal X-ray diffraction of a recovered crystal gave the lattice parameters of a = 4.18077(7) Å, c = 2.66658(13) Å, and the unit cell volume of V = 46.6086(27) Å3. Further details on the analytical methods are provided in the section below. As for the matrix to be employed in the annealing experiments, this was obtained by mixing reagent-grade SiO2 quartz and α-AlOOH boehmite, which was then converted to hydrous phase δ after high pressure–temperature annealing.

2.2 High pressure–temperature experiments

All high-pressure experiments were performed using Kawai-type multi-anvil presses installed at the Bayerisches Geoinstitut, University of Bayreuth. We conducted a pre-synthesis experiment of aluminous stishovite starting crystals at 22 GPa and 1900 °C for 2 h using a 10-MN press with a split-sphere guide block. The pressure calibration was reported in Keppler and Frost (2005). The cell assembly with 4 mm truncation tungsten carbide (WC) anvils and a 10 mm edge octahedron reported in Kawazoe et al. (2017) was adopted. The powdered starting sample was packed in a welded Pt capsule.

The water partitioning experiments between stishovite and hydrous phase δ were carried out at pressures of 24 and 28 GPa and temperatures of 1000–1200 °C for 24 h. The 10-MN press was also used for experiments at 24 GPa, in which the pressure estimation by Keppler and Frost (2005) was also adopted. Experiments at 28 GPa were conducted using the 15-MN press with an Osugi-type guide block, IRIS-15 (Ishii et al. 2016, 2019b). Generated pressures against press load were determined in Liu et al. (2017). A similar cell assembly design to that reported by Liu et al. (2020) was adopted, except for the sample chamber (see Fig. 1) at both conditions. A 7 mm Cr2O3-doped MgO pressure medium was combined with tungsten carbide (WC) anvils with 3 mm truncated edge lengths. A cylindrical LaCrO3 heater was inserted into the center of the octahedron. LaCrO3 plugs were placed at both ends of the heater. Some single crystals of the pre-synthesized aluminous stishovite were placed in a welded gold capsule, in which the single crystals were separated from each other by the SiO2–AlOOH powdered mixture. The sample capsule was placed at the center of the heater, from which it was electrically insulated using an MgO sleeve and two plugs. Sample temperatures were measured at the surface of the capsule using a D-type thermocouple. Ceramic components of the cell assembly (LaCrO3 and MgO parts and pressure media) were fired at 1000 °C for at least 3 h. The sample was first compressed to a desired press load at room temperature and then heated to the target temperature at a rate of 100 °C/min. The target conditions were kept for 24 h with temperature fluctuations of less than ± 5 °C. After keeping the target pressure–temperature conditions for the desired time, the temperature was rapidly decreased to room temperature by shutting off the electric power supply to the heater, and then the cell assembly was slowly decompressed to ambient pressure for 15 h.

Fig. 1
figure 1

A schematic drawing of the cell assembly used for the water partitioning experiments

Full size image

2.3 Analysis of recovered samples

The phases present in the recovered matrix parts were confirmed using a micro-focused X-ray diffractometer with a micro-focus source (IμS) of Co–Kα (Bruker, D8 Discover) in combination with a two-dimensional solid-state detector (VÅNTEC-500).

Single crystals of the starting stishovite were hand-picked from the recovered Pt-capsule and observed under a polarizing microscope. To estimate density of the single crystal stishovite for calculations of their water content based on their FTIR data, an inclusion-free grain with ~ 50 μm in diameter displaying sharp optical extinction was then selected and mounted on glass needles for single-crystal X-ray diffraction measurements. The unit-cell lattice parameters of the single crystal were determined using a Eulerian-geometry Huber single-crystal diffractometer equipped with a MoKα sealed tube operated at 40 kV and 50 mA and a point detector. The software program SINGLE (Angel and Finger 2011) was used to drive the diffractometer and collect diffraction profiles of eight unique reflections, each centered using the 8-position method (King and Finger 1979). The symmetry-constrained lattice parameters were then calculated via a vector least-square refinement.

Recovered crystals were double-side polished to 70–176 μm in thickness for unpolarized FTIR spectroscopy. The FTIR measurements were performed using a Bruker IFS 120 high-resolution spectrometer in combination with a Bruker IR microscope and collected using several randomly oriented crystals. A baseline correction was applied by subtracting from each raw spectrum using a spline function that was fitted outside the OH bands wavenumber region. The water contents in recovered aluminous stishovite were calculated based on the empirical calibration of Paterson (1982):

$$\begin{array}{*{20}c} {C_{{{\text{H}}_{2} {\text{O}}}} = 3 \times \frac{{X_{i} }}{150}\int {\frac{K\left( \upsilon \right)}{{\left( {3780 - v} \right)}}{\text{d}}v} } \\ \end{array}$$
(1)

where CH2O is the water content (wt ppm), K(ν) is the absorption coefficient (cm−1) at a given wave number ν, and Xi is a density factor given as Xi = 18/2d × 106, where d is the mineral density (g/l) determined by single-crystal X-ray diffraction and chemical-composition analysis. The absorption coefficients were integrated in the wavenumber range of 2400–3700 cm−1.

The chemical compositions of the crystals were measured using an electron probe microanalyzer with wavelength-dispersive spectrometers (JEOL, JXA-8200) at an accelerating voltage and probe current of 15 kV and 15 nA, respectively. Standard materials were quartz (SiO2) and synthetic corundum (Al2O3) for Si and Al, respectively.

3 Results and discussion

3.1 Water content of pre-synthesized starting aluminous stishovite

Unpolarized FTIR spectra of the pre-synthesized starting stishovite single crystals are shown in Fig. 2. The spectrum of the starting stishovite shows main OH-absorption bands at 2667, 3115, 3240, 3318, and 3403 cm−1. These observations agree with those of stishovites reported by Litasov et al. (2007). Although it is known that stishovite is highly anisotropic (Litasov et al. 2007; Ishii et al. 2022b), we observed limited differences in the unpolarized FTIR profiles of 10 crystals with random crystal orientation, resulting in water contents of 940–1059 ppm. Thus, the average water content of 993 ppm was obtained with one standard deviation of the mean of 44 ppm. The aluminum and hydrogen contents of the starting stishovite were estimated to be 0.75(4) mol% and 0.66(3) mol%, respectively, indicating that the Al/H ratio is close to 1. Previous studies reported Al/H ratios between 2 and 7 (Litasov et al. 2007; Zhang et al. 2022; Ishii et al. 2022b), indicating a high concentration of oxygen vacancy. Thus, the present starting stishovite has a higher Al/H ratio than previously reported. This is probably because our stishovite was synthesized at a higher temperature of 1900 °C than those in previous studies (1200–1800 °C) (Litasov et al. 2007; Ishii et al. 2022b). Ishii et al. (2022b) reported a positive temperature dependence of water content in aluminous silica phases coupled with an increase in the alumina concentration close to or under alumina saturated conditions. On the other hand, since our stishovite with the limited alumina content was synthesized quite far from the alumina saturated conditions, the present experimental condition is not the same as the previous conditions with much more alumina. However, it would be possible to incorporate more hydrogen by a possible positive temperature dependence of the AlOOH content instead of oxygen vacancy formation. This hypothesis is supported by a previous ab initio calculation study (Panero and Stixrude 2004).

Fig. 2
figure 2

Representative unpolarized FTIR spectra of recovered stishovite single crystals

Full size image

3.2 Water contents of aluminous stishovite coexisting with δ-AlOOH

We conducted three water partitioning experiments at 24–28 GPa and 1000–1200 °C (Table 1). Figure 3 shows a representative cross section of a recovered sample capsule including a stishovite single crystal and the matrix part. Powder X-ray diffraction and electron microprobe analysis confirmed that the matrix portion became stishovite with ~ 0.7 wt% Al2O3 and hydrous phase δ with ~ 4.0 wt% SiO2 after annealing, showing negligible temperature dependence of the major compositions in the present temperature range. The annealed crystals show the same OH-absorption bands as those of the starting stishovite (Fig. 2). On the other hand, the OH bands of the annealed crystals have lower integrated intensities than those of the starting crystal, indicating a decrease in the water content during annealing. The water contents of aluminous stishovite after annealing have been estimated to be 456–535 ppm corresponding to 0.30–0.35 mol% of hydrogen (Table 2). Thus, the Al/H ratio became ~ 2, which is close to the trend previously reported (Fig. 4) (Litasov et al. 2007; Zhang et al. 2022; Ishii et al. 2022b). We confirmed that the recovered single crystals were chemically homogeneous based on EPMA and FTIR analyses at different places (Table 2 and Additional file 1: Table S1). The annealing time in this study would be enough to reach equilibrium for hydrogen based on previously determined H diffusivities of minerals such as olivine, its polymorphs, and TiO2 rutile that has the same structure as stishovite (Ingrin and Blanchard 2006; Chakraborty 2010). We note that alumina contents in stishovite single crystals were consistent within error before/after the partitioning experiments (Table 1, 2 and Additional file 1: Table S1). Alumina contents of stishovite in the matrix part (~ 0.7 wt%) appear to be slightly higher than those of the single crystals (0.59–0.67 wt%). While these values are often in agreement with each other within their mutual uncertainties, it is also possible that Al may have not completely reached chemical equilibrium during the experiments. In the latter case, for an Al/H ratio of 2 (Fig. 4), the water content would increase by only a few tens ppm at the Al content in the matrix part, which is anyway within two standard deviations of the analytical error on the H2O concentration. Thus, such a small difference in alumina content does not change our conclusions. This result suggests strong water partitioning into hydrous phase δ and is consistent with water partitioning between nominally anhydrous minerals and hydrous minerals in other systems such as bridgmanite and hydrous phase δ/D and olivine polymorphs and hydrous minerals (Ishii and Ohtani 2021, Ishii et al. 2022c).

Table 1 Experimental conditions and average water contents of recovered stishovite
Full size table
Table 2 Water and alumina contents of recovered stishovites
Full size table
Fig. 3
figure 3

A representative photo of recovered stishovite after the water partitioning experiment (H5324)

Full size image
Fig. 4
figure 4

H+ contents of stishovite as a function of Al3+ content. ST, starting stishovite pre-synthesized; AA, stishovite coexisting with hydrous phase δ after annealing. Some of stishovites in previous studies coexist with liquid phase or hydrous phase egg. Errors of the data from this study are within the size of the symbol

Full size image

3.3 Water transport in the lower mantle

Here, we discuss water transport in the mantle. Down to the mantle transition zone, subducted slabs transport water mainly as dense hydrous magnesium silicates such as phase A, superhydrous phase B, and phase D (Litasov and Ohtani 2003; Liu et al. 2019). On the other hand, in the lower mantle, a solid solution between hydrous phase δ and H may crystallize in basaltic and peridotitic layers when the slab temperature is relatively lower than the average mantle (Liu et al. 2019; Nishi et al. 2015; Ishii et al. 2022c). As mentioned earlier, although the nominally anhydrous minerals in the lower mantle, such as bridgmanite and ferropericlase, are almost dry even under water-saturated conditions (e.g., Bolfan-Casanova et al. 2002; Fu et al. 2019), the presence of hydrous phase δ further facilitates dehydration of these minerals and stishovite (Ishii et al. 2022c; this study) and can destabilize alkali element- and alumina-rich minerals, such as the aluminous phases of calcium ferrite phase and hexagonal phase (Ishii et al. 2023), due to the strong alumina and water partitioning to hydrous phase δ.

Recent experimental and theoretical studies suggested that the water solubility of davemaoite is up to a few weight percent, although the unquenchability prevents a definitive conclusion (Chen et al. 2020; Shim et al. 2022). Chen et al. (2020) demonstrated that davemaoite may be hydrous even when coexisting with hydrous phase δ based on the crystal structure change in davemaoite from cubic to tetragonal and the volume reduction relative to the dry sample by a laser-heated diamond anvil cell. The strong alumina and water partitioning to hydrous phase δ against bridgmanite and stishovite suggests that davemaoite is also alumina- and water-depleted at least when coexisting with hydrous phase δ, although further experiments are needed to investigate the water solubility of davemaoite in the CaSiO3-AlOOH system.

Therefore, we suggest that a main water carrier would be hydrous phase δ-H solid solution in a relatively low temperature slab. Because slab temperature should be gradually increasing, hydrous phases are likely to decompose at some depths when the slab temperature approaches that of the average lower mantle (Katsura 2022). After dehydration, the released water will react with aluminous silica phases, resulting in the water carrier being replaced by hydrous aluminous silicas such as stishovite and CaCl2-type phase due to their positive temperature dependence of alumina and water contents (Ishii et al. 2022b).

3.4 Seismic scatterers in the lower mantle

Seismological studies found ~ 10 km thick seismic scattering bodies with lower shear-wave velocities and higher densities than the surrounding mantle between 700 and 1900 km depths (e.g., Kaneshima 2019 and references therein). These observations have been explained by the second-order ferroelastic transition from stishovite to CaCl2-type phase in subducted basaltic crusts because the transition pressure decreases with increasing alumina and water content (e.g., Lakshtanov et al. 2007; Zhang et al. 2022; Ishii et al. 2022b; Criniti et al. 2023). Thus, the deepest seismic scatterers around 1900 km depth were interpreted by the transition in dry and almost pure SiO2 system (e.g., Nomura et al. 2010; Fischer et al. 2018), implying that subducted basaltic crusts at this depth could be dry because stishovite in a dry basaltic crust is alumina depleted (Irifune et al. 1994; Ishii et al. 2019a, b, 2022a). The present study suggests that even low-temperature hydrous basaltic crust with hydrous phase δ may be compatible with these observations as alumina and water will be strongly partitioned into hydrous phase δ. We note that stishovite coexisting with hydrous phase δ has a small amount of alumina and water as described above and even such small amounts of alumina and water may affect the transition pressure. Since previous studies demonstrated a reduction in the transition pressure by H and Al incorporation in relatively high alumina- and water-bearing samples with more than 3 wt% and 0.25 wt%, respectively (e.g., Lakshtanov et al. 2007; Zhang et al. 2022; Criniti et al. 2023), further studies are needed to clarify the effect of low H and Al incorporation.

4 Conclusions

We have determined water contents of aluminous stishovite coexisting with hydrous phase δ at 24–28 GPa and 1000–1200 °C using a Kawai-type multi-anvil press in combination with FTIR measurements of recovered stishovites at ambient pressure. Aluminous stishovite coexisting with hydrous phase δ is depleted in alumina (< 1 wt%) and has ~ 500 ppm water, showing a strong water and alumina partitioning to hydrous phase δ. The present results in addition to previous results of water solubility of nominally anhydrous lower-mantle minerals suggest that hydrous phase δ is a main water carrier in the lower mantle when a subducting slab is at relatively low temperatures compared with the average mantle. Once the subducting slab becomes warm enough to trigger the dehydration of hydrous minerals, aluminous silica minerals such as stishovite and CaCl2-type phase will become the main water carriers. Seismic scatterers observed around 1900 km depth may be explained by the transition of Al, H-poor stishovite to CaCl2-type phase coexisting with hydrous phase δ at low temperatures.