- © 2011 American Mineralogist
The 3.65 Å phase, a member of the family of dense hydrous magnesium silicates (DHMS) in the system MgO-SiO2-H2O, was synthesized in a 77 h multi-anvil press experiment at conditions of 10 GPa and 425 °C by using a gel of composition MgSiO3 plus water in excess as starting materials. From our multi-methodical study including SEM, TEM, EMP, IR, and Raman analysis, we determined the composition of the 3.65 Å phase to be MgSi(OH)6. Powder XRD combined with Rietveld refinement revealed the 3.65 Å phase to be isostructural with δ-Al(OH)3. The 3.65 Å phase can be described as a hydrous A-site vacant perovskite with probably long-range random distribution of Si and Mg at octahedral sites. Locally, some ordering of Mg and Si might exist, as indicated from the spectroscopic measurements. The 3.65 Å phase represents the second DHMS with exclusively sixfold-coordinated Si, the other being phase D. The 3.65 Å phase is stable at pressures above about 9.0 GPa and temperatures below 500 °C. This limited P-T stability together with its high water content makes the 3.65 Å phase an unrealistic mantle component. If at all, it might only occur under hydrous conditions in the coldest parts of deeply and extremely fast subducted old oceanic slabs.
- 3.65 Å phase
- dense hydrous magnesium silicates
- high-pressure synthesis
- crystal structure
- octahedral-coordinated Si
Since the first synthesis report of the 10 Å phase by Sclar et al. (1965) and of the alphabetical phases A, B, and C by Ringwood and Major (1967), six further dense hydrous magnesium silicates (DHMS; phase D, E, F, G, shy B, 3.65 Å phase) have been synthesized and numerous investigations have focused on their P-T stabilities and properties [for reviews see e.g., Prewitt and Finger (1992), Mysen et al. (1998), Kawamoto (2006)]. Except 10 Å phase, which was discovered as inclusions of kimberlitic olivine (Khisina et al. 2001), none of the other DHMS have been found in nature and are only known from high-pressure syntheses. Nevertheless, they are discussed to be important hosts and carriers of water under hydrous conditions of the Earth’s mantle and in subduction zones. If present at mantle depths, DHMS would have significant effects on mantle properties, e.g., elastic wave velocities, electrical conductivity, and their dehydration within descending slabs might be associated with water-induced and water-influenced processes, e.g., melt generation, metasomatism, and seismicity. Therefore, their study is central to the understanding of the Earth’s deep water cycle.
The 3.65 Å phase, named after the d-value of its prominent X-ray reflection, is the least-characterized phase of the DHMS family. Sclar and Morzenti (1971) first described the synthesis of the 3.65 Å phase in the system MgO-SiO2-H2O (MSH) at pressures above 9.0 GPa and temperatures below 500 °C and inferred from observed phase assemblages a Mg2Si□(OH)8 composition. They indexed the X-ray pattern based on a hexagonal cell with a = 14.72 Å and c = 3.655 Å and suggested a structure equal to brucite with three-fourths of the octahedral sites in alternate layers occupied by magnesium and silicon ions in the ratio 2:1 and one-fourth octahedral sites vacant. Employing energy-dispersive (EDX) analyses in the electron microscope, Rice et al. (1989) determined a Mg/Si ratio of 1.077 and suggested the composition Mg1.4Si1.3□1.3(OH)8 again derived from brucite Mg4(OH)8 by the substitution mechanism Mg−2Si□. In the course of determining the upper P-T stability of the 10 Å phase, the 3.65 Å phase was recently found as a breakdown product of the 10 Å phase above about 9.5 GPa and temperatures below 500 °C (Pawley et al. 2011). Neither composition nor structure of the 3.65 Å phase was determined by Pawley et al. (2011).
The proposed composition and structure of the 3.65 Å phase needs revision for the following reasons. (1) The two studies by Sclar and Morzenti (1971) and Rice et al. (1989) represent only short conference abstracts without following full publications. (2) According to Prewitt and Finger (1992), the extremely high OH content and the fact that all the Si should be in octahedral coordination is quite unusual for the relatively low synthesis pressure. Here we report on the synthesis of the 3.65 Å phase and detailed results of a multi-methodical study with respect to its composition, structure, and some further properties.
The synthesis of the 3.65 Å phase was performed in two multi-anvil press experiments (Table 1⇓). We used MgO(+5%Cr2O3) octahedra with an edge length of 14 mm as pressure media. The WC anvils had a truncation edge length of 8 mm. Temperature was measured using W5%Re-W26%Re (type C) thermocouples with a precision of about ±10 °C. Pressure was calibrated by press-load experiments on several room- and high-temperature phase transitions. The estimated pressure uncertainty is about ±0.3 GPa. During heating, the Walker-type module was rotated through 360° with a rate of 5°/s to prevent separation of fluid and solid components (Schmidt and Ulmer 2004). A detailed description of the press and the experimental method is given by Watenphul et al. (2009).
The first synthesis experiment (no. MA290) was performed at 10 GPa, 400 °C, and 114 h run duration. We used synthetic brucite, Mg(OH)2, and quartz as solid starting materials, which were mixed in the Mg/Si ratio of 1.077, according to the proposed composition of the 3.65 Å phase after Rice et al. (1989). About 5 mg of this solid starting material was placed together with about 1 mg of water into a Pt capsule of 3 mm length and 2 mm diameter.
The second synthesis run (no. MA304) was performed at 10 GPa, 425 °C, and 77 h run duration. Based on the results of the electron microscope EDX analyses on no. MA290 yielding a Mg/Si ratio of 1 (Table 1⇑) we used for no. MA304 a gel with corresponding composition as solid starting material. The gel was prepared after the method of Hamilton and Henderson (1968). As determined by X-ray fluorescence spectroscopy the gel had an Mg/Si ratio of 1 and water content of 3.3 wt%. For no. MA304 about 5 mg of the gel and 40 wt% of water were added to the Pt capsule.
After quenching of the experiments and slow decompression, the recovered capsules were cleaned, checked for leakage by weighing and then opened.
Scanning electron microscopy (SEM).
The SEM used was a Zeiss SMT Ultra 55 Plus, equipped with an UltraDry detector (Thermo Fisher Scientific) for energy-dispersive spectroscopic (EDS) analyses, which were performed at an accelerating voltage of 20 kV and a beam current of 1 nA.
Transmission electron microscopy (TEM).
The run products were slightly compressed and embedded in epoxy. TEM foils of approximately 0.150 μm thickness were cut out of the epoxy using a focused ion beam (FIB) technique (FEI FIB 200 TEM) with a Ga-ion source operated with an acceleration voltage of 30 kV (Wirth 2004). The TEM investigations were performed with a FEI Tecnai G2 F20 X-Twin transmission electron microscope with a FEG electron emitter. An energy-dispersive X-ray analyzer (EDAX) with ultrathin window was used for chemical analysis. The chemical composition was measured in scanning transmission (STEM) mode scanning the electron beam within a window. The size of the window was adapted to the individual crystal size and was always as large as possible. This procedure avoids mass loss during the measurement. Acquisition time was 60 s. We applied the KMg and KSi factors of 1.034 and 1.000 from the TIA software package to quantify the results. The foil was tilted at 15° toward the X-ray detector during measurement. The analytical error is ~5%.
Electron microprobe (EMP).
EMP analyses were obtained with the JEOL thermal field emission type (FEG) electron probe JXA-8500F (HYPERPROBE). The analytical conditions included an acceleration voltage of 15 kV, a beam current of 1–2 nA, and a beam diameter of 1 μm. On-peak counting times were 20 s and background on both sides of the peak were 10 s. As standard we used well-defined synthetic Fe-bearing ringwoodite.
Infrared (IR) spectroscopy.
IR spectra were taken in the OH-stretching region to analyze the water incorporation and in the MIR region down to 500 cm−1. Due to the high absorption, very thin samples (2–3 μm) are required for water-rich samples in both spectral regions. Therefore, we made a thin film of the 3.65 Å phase from no. MA304 by pressing a small amount of the sample in a piston-cylinder type diamond-anvil cell (DAC). Depending on the diamond culet size pressing by hand without gasket results in 2–5 GPa on the sample. In run no. MA304, the 3.65 Å phase occurred together with small amounts of the 10 Å phase (Table 1⇑). Therefore, prior to this procedure, we separated the 10 Å phase from the sample by handpicking under a binocular microscope. The 10 Å phase was easily recognized optically due to its platy morphology and larger crystal size (see later, Fig. 1b⇓). After pressure release and removal of the cylinder part of the DAC, the film normally sticks on the piston diamond and is ready for measurements. After measurements the film was fixed on a fine needle and the thickness was determined to be 2 ± 0.2 μm by analysis of video images using the program OPUS by Bruker. IR measurements were carried out with a Vertex 80 v FTIR spectrometer connected to a Hyperion microscope. We used a Globar light source, a KBr beam splitter, and a MCT detector. Spectra were taken in the spectral range 4000–500 cm−1 with a resolution of 2 cm−1. The measured spectra were averaged over 256 scans.
The same film was used for recording Raman spectra with a Horiba Jobin-Yvon Labram HR 800 UV-VIS spectrometer (grating 1800 grooves/mm) in backscattering configuration. The spectrometer was equipped with a Peltier-cooled CCD detector, a coherent water-cooled argon laser, an Olympus optical microscope and a long working distance 40× objective. For sample excitation, we used the 514 nm Ar+ line and a laser power of 320 mW. The confocal pinhole of 100 μm allowed us to measure with a spectral resolution of about 1 cm−1. Raman spectra of the thin film were taken in the spectral range 900–100 cm−1 with an acquisition time of 50 s and 3 accumulations. For IR and Raman spectra, the band positions and integrated intensities were obtained using the PeakFit software by Jandel Scientific.
Powder X-ray diffraction (XRD).
XRD patterns were recorded in transmission mode, using a fully automated STOE Stadi P diffractometer, equipped with a primary monochromator and a 7° wide position sensitive detector. Intensities were recorded in the 2Θ range 5–125° for CuKα1-radiation in 0.1° intervals. The normal-focus Cu X-ray tube was operated at 40 kV and 40 mA, using a take-off angle of 6°. Counting times were selected to yield a maximum of 4200 counts, resulting in 10 s per detector step. For quantitative phase analyses and determination of structural parameter, Rietveld analyses were performed using the GSAS software package (Larson and Von Dreele 1987).
The first synthesis experiment no. MA290 (Table 1⇑) resulted in brucite and stishovite as main components and contained the 3.65 Å phase only in smaller fraction (9 wt% as determined using the derived structure of the 3.65 Å phase from the XRD analysis of no. MA304, see later). The additional occurrence of coesite in traces might be due to metastable formation of coesite in the stability field of stishovite at low temperatures (Yagi and Akimoto 1976). EDX analyses using the TEM (Table 1⇑) gave Mg/Si ratios close to 1 measuring several grains of the 3.65 Å phase. Therefore, the second experiment no. MA304 (Table 1⇑) was performed using a gel of MgSiO3 composition plus excess water. This experiment resulted in yield of the 3.65 Å phase of about 90 wt% together with 10 Å phase and an unknown phase (see later). In the following we will focus on this sample (no. MA304).
SEM pictures of the product phases from run no. MA304 are shown in Figures 1a–1b⇑. Crystals of the 3.65 Å phase are mostly of rounded shape with badly developed morphology and have typically sizes of 1–2 μm. Less often, idiomorphic crystals with sizes up to 3 μm were observed, whose crystal morphology resembles the pseudo-cubic structure of perovskite (Fig. 1c⇑). The marked crystal in Figure 1a⇑ has a Mg/Si ratio of 1 as determined by the EDX system of the SEM. Together with the 3.65 Å phase, up to 50 μm large crystals of the 10 Å phase are visible (Fig. 1b⇑), which are easily distinguishable from the 3.65 Å phase due to the typical platy morphology of a layer silicate. The Mg/Si ratio of the 10 Å phase was determined to 0.75.
Using the EDX-system of the TEM, we analyzed the Mg/Si ratio of 25 individual grains of the 3.65 Å phase of run no. MA304. This procedure resulted in a mean Mg/Si ratio of 0.98 ± 0.06 (Table 1⇑), which differs significantly from the compositions given by Sclar and Morzenti (1971) and agrees with the Mg/Si ratio given by Rice et al. (1989) within 2σ uncertainty. The 3.65 Å phase was extremely unstable under the electron beam of the TEM and immediately became amorphous even with a defocused electron beam. Therefore, no electron diffraction of the 3.65 Å phase could be performed and no structural information is available from the TEM.
Forty-three individual EMP analyses of the 3.65 Å phase of run no. MA304 resulted in a mean Mg/Si ratio of 1.02 ± 0.06 (Table 1⇑) and mean analytical sum of 66.5 ± 3.2. The 3.65 Å phase was extremely unstable and dehydrated during the measurements. Beside the 3.65 Å phase and the 10 Å phase no other phase was detected.
IR and Raman spectra of sample no. MA304 are shown in Figures 2a–2c⇓. Characteristic bands of the 10 Å phase (Kleppe and Jephcoat 2006) were not observed in the spectra, indicating successful separation of the 10 Å phase from the sample. However, we cannot rule out that some of the (weak) bands belong to the unknown phase which was observed in the XRD pattern of sample no. MA304. In the range of the OH-stretching frequencies four distinct, broad bands at 3216, 3321, 3418, and 3460 cm−1 with at least two shoulders at 3188 and 3394 cm−1, and one weak band at 3575 cm−1 are visible (Fig. 2a⇓). The water content of the 3.65 Å phase was calculated from the integrated intensities Ai of the OH-bands measured in the spectral range 3800–2800 cm−1 (Fig. 2a⇓) following the Beer-Lambert law with the equation: c (wt% H2O) = Ai (cm−1) × 1.8/[t (cm) × ρ (g/cm3) × εi (cm−2 per mol H2O/L)], with t = sample thickness, ρ = density of the 3.65 Å phase, εi = integrated molar absorption coefficient taken from Libowitzky and Rossman (1997). The water content was determined to be 34 ± 3 wt%.
In the range of lattice vibrations (Fig. 2b⇑), we observe three principal groups of bands at around 685, 830, and in the range 1300–1100 cm−1. The Raman spectrum (Fig. 2c⇑) exhibits a strong, broad band at 681 cm−1 and a strong, sharp band at 269 cm−1 besides some weak bands. A symmetry analysis based on the assumed space group Pnam (point group mmm) for the derived 3.65 Å phase structure (see below) calculates the number and symmetry of expected IR and Raman modes: 10 Ag (Raman) + 10 B1g (Raman) + 8 B2g (Raman) + 8 B3g (Raman) + 11 Au (inactive) + 10 B1u (IR) + 12 B2u (IR) + 12 B3u (IR). Thus, there are 36 modes expected in the Raman and 34 modes in the IR spectra. As commonly seen, the observed number of bands in our IR and Raman spectra is much smaller than the number of calculated bands.
Taking into account that the Mg/Si ratio is 1 and that the 3.65 Å phase is OH-bearing as determined from IR analyses, a possible composition of the 3.65 Å phase is MgSi(OH)6. With this information and after much trial and error, we were able to solve the structure of the 3.65 Å phase by taking the structural model of δ-Al(OD)3 determined by Komatsu et al. (2007) as a starting point for refinement of the XRD powder diagram of no. MA304 (Fig. 3⇓). The hydrous analog δ-Al(OH)3 is a high-pressure phase and was first synthesized at about 6.1 GPa and 410 °C by Dachille and Gigl (1983). Komatsu et al. (2007) synthesized δ-Al(OD)3 at 18 GPa and 700 °C and solved its structure in orthorhombic space group Pnam with Al in sixfold coordination. In our refinement the octahedral Al of δ-Al(OD)3 was replaced by Mg and Si according to the substitution 2Al = Mg + Si, and assuming random distribution of these two cations sharing the M site equally. Furthermore, atomic coordinations of deuterium after Komatsu et al. (2007) were taken as H-positions of the 3.65 Å phase. Accordingly, four non-equivalent proton sites exist, each with site occupancies of 0.5. For the 10 Å phase, structural data were taken from Comodi et al. (2005).
With these input data, the quantitative phase analyses of the powder XRD pattern of run no. MA304 resulted in 90 wt% of the 3.65 Å phase and 10 wt% of the 10 Å phase (Table 1⇑). The quality of the refinement statistics, wRp = 0.093, χ2 = 1.348, DWd (Durbin-Watson) = 1.148 is good except for the occurrence of two reflections at d = 4.234 and 4.192 Å of an unknown phase, which could not be assigned to the 10 Å phase. These two reflections do not occur in the XRD spectrum of run no. MA290, which might either be due to the small amount of the 3.65 Å phase in this sample or perhaps indicating that these two reflections also do not belong to the 3.65 Å phase. None of the known MSH-phases has two prominent reflections at d ~ 4.2 Å.
In comparison to δ-Al(OD)3, the cell dimensions of the 3.65 Å phase are all significantly enlarged (Table 2⇓). All reflections given by Kamatsu et al. (2007) for δ-Al(OD)3 are also present in the XRD pattern of the 3.65 Å phase but are occurring at larger d-values and with similar relative intensities (Table 3⇓). No indication of any deviation from space group Pnam could be detected. The H-positions of the 3.65 Å phase could not be determined reliably from our powder XRD Rietveld refinement because of the low-atomic scattering factors of H. We therefore present in Table 4⇓ only preliminary H-positions originally taken from Komatsu et al. (2007).
The structure of the 3.65 Å phase consists of corner-sharing (MgSi)(OH)6 octahedral sheets (Figs. 4a and 4b⇓). The structure shows strong similarities to the structure of perovskite, CaTiO3 (Figs. 4c and 4d⇓), however, without occupation of the large Ca-position. In analogy with δ-Al(OD)3, the O-H dipoles of the four H-positions are probably arranged within these empty A-site positions in the 3.65 Å phase.
Recently, Matsui et al. (2011) determined ordering of the H-positions and a symmetry reduction to P212121 in δ-Al(OH)3 by performing neutron powder diffraction and ab initio calculations. In space group P212121 three different H-positions exist. We refined the 3.65 Å phase of run no. MA304 in space group P212121 and found the new cell dimensions a = 5.1899(4), b = 7.3302(5), c = 5.1132(3) Å, and V = 194.52 Å3 (refinement statistics: wRp = 0.087, χ2 = 1.184, DWd = 1.310) without refinement of the H-positions. In P212121, the two unknown reflections could now be assigned to the 3.65 Å phase, however, with significant difference between measured and calculated intensities.
The slightly better refinement statistics for P212121 compared to Pnam structure are not a clear indication for preferring space group P212121 to Pnam, because of less structural variables in the higher symmetric structure. The observed four main OH-bands in the IR-spectrum of the 3.65 Å phase (Fig. 2a⇑) are in better agreement with space group Pnam, which requires four different H-positions. At this stage, we cannot make a clear decision concerning the correct space group of the 3.65 Å phase. For clarification of the proton positions and thus for the space group, a crystal structure analysis of deuterated 3.65 Å phase using neutron diffraction is needed, which, however, is beyond the scope of this study.
Nevertheless, there are various lines of evidence indicating that we have determined the correct composition and the general structure of the 3.65 Å phase. (1) EDX analyses using the SEM and TEM and EMP analyses resulted in Mg/Si ratio close to 1. (2) The water-content determined by IR (34 ± 3 wt%) and the difference of the sum to 100 of the EMP analyses (33.5 ± 3.2 wt%) fit perfectly to the 35 wt% H2O for the proposed MgSi(OH)6 composition. Interestingly, already Finger and Hazen (2000) proposed that a phase of MgSi(OH)6 composition might be an important component of the Earth’s deep interior. (3) By analogy with the 3.65 Å phase, the IR spectrum of δ-Al(OH)3 shows four distinct OH-bands in the range of the OH-stretching frequencies (Xue and Kanzaki 2007), which might correspond to the four individual H-positions in these two Pnam structures. (4) Idiomorphic crystals of the 3.65 Å phase have a “perovskite” morphology; the structure of the 3.65 Å phase shows similarities to a perovskite structure and can be described as a hydrous A-site vacant perovskite. (5) The structure of the 3.65 Å phase was derived from δ-Al(OH)3 and δ-Al(OD)3, which are both high-pressure phases. And (6), XRD-refinement statistics of our refinement in space group Pnam on the sample containing 90 wt% of the 3.65 Å phase are good, even with the occurrence of two non-identified XRD-reflections of an unknown phase. Given the phase assemblage of the product of run no. MA304 (3.65 Å phase, 10 Å phase), and considering that a gel of MgSiO3 composition was used as starting material, an Mg/Si ratio > 1 might be suggested for the unknown phase. However, from the EMP analyses we did not detect any additional phase besides the 3.65 Å phase and the 10 Å phase. It is conceivable that the unknown phase formed crystals too small to be detected by EMP.
The H-positions of the 3.65 Å phase could not be refined from our XRD Rietveld analysis. Therefore, from the given structural data (Tables 2⇑ and 4⇑) no realistic O-H distances can be derived for the 3.65 Å phase. Nevertheless, the refined O-(H)…O distances in space group Pnam are comparable to the O-(D)…O distances of δ-Al(OD)3 (Komatsu et al. 2007): O1-H1…O1 and O1-H2…O1 in the 3.65 Å phase are 2.940 Å compared to 2.920 Å in δ-Al(OD)3, O2-H4…O2 and O2-H3…O2 distances in the 3.65 Å phase are 2.740 and 2.736 Å, respectively, compared to 2.738 and 2.766 Å, respectively, in δ-Al(OD)3. Taking the general correlation of O-H stretching wavenumbers and O-H…O bond lengths (Libowitzky 1999) into account, we can assign the two IR bands at higher wavenumber (3460 and 3418 cm−1) to H located between the larger O1…O1 distance and thus to O1-H1 and O1-H2 stretching vibrations, and the two bands at lower wavenumber (3321 and 3216 cm−1) can be assigned to H located between the shorter O2…O2 distances. The OH-bands are broad and they show at least two distinct shoulders, which perhaps indicate a local Mg-Si ordering. Local ordering would lead to three different environments (Mg,Mg-O-H), (Mg,Si-O-H), and (Si,Si-O-H), which differ in their electronegativity. As there is an anticorrelation of the energy of an OH-band with the electronegativity of the cations bonded to the hydroxyl ions (e.g., Hawthorne 1981) one could expect three subbands. The one associated with (Mg,Mg-O-H) would be at higher wavenumbers than the one attributed to (Si,Si-O-H). Thus, taking the most prominent asymmetric band at 3216 cm−1, this could be resolved in three bands: one at 3240 cm−1 assigned to (Mg,Mg-O-H), one at 3216 cm−1 assigned to (Mg,Si-O-H) and one at 3188 cm−1 assigned to (Si,Si-O-H).
Vibrational spectra of phases with Si in octahedral coordination have several features in common (e.g., Williams et al. 1987; Cynn et al. 1996): bands in the range of 1200–1000 cm−1 are due to asymmetric O-Si-O bending, those in the range of 800–750 cm−1 can be assigned to asymmetric stretching modes and the bands in the range 700–500 cm−1 to deformational vibrations of SiO6 octahedra. Thus, our IR spectrum (Fig. 2b⇑) is dominated by motions of Si-O in octahedral coordination. Also in this spectral region the bands are broad, indicating local ordering and we assume that the corresponding MgO6 motions are only slightly separated from the SiO6 motions. Considering only the mass difference, which is admittedly a simple model, the corresponding bands would be separated by 20 cm−1 only. In the Raman spectrum (Fig. 2c⇑), the broad peak at 681 cm−1 is also assigned to octahedral stretching vibrations. Problematic is the assignment of the sharp band at 269 cm−1. For other DHMS (e.g., Cynn et al. 1996) bands around this position have been assigned to translations of octahedral Mg. However, the octahedra refined for the 3.65 Å phase structure are positioned on an inversion center, and therefore Raman active vibrations cannot involve translation of these atoms. Therefore, the presence of this prominent band and its sharpness might suggest reduction of the site symmetry due to local ordering.
Remarkably, besides phase D (Yang et al. 1997), the 3.65 Å phase represents the second DHMS with exclusively sixfold-coordinated Si. This is somewhat surprising, because the synthesis pressure of the 3.65 Å phase is only about half the pressure of the phase D synthesis (Yang et al. 1997). Furthermore, phase E and phase B, both with synthesis pressures similar to the 3.65 Å phase, have Si in tetrahedral and mixed (Si and Si) coordination (Kudoh et al. 1989; Finger et al. 1989), respectively. The reason for this inconsistency might result from differences in the octahedral Mg, Si distribution, which is ordered for Phase B, D, E, and not unambiguously clarified for the 3.65 Å phase. We assume long-range random distibution of Mg and Si at octahedral sites, because symmetry reduction due to ordering leads to the occurrence of additional XRD reflections, which were not observed in the XRD pattern.
Combining the phase relations given by Pawley et al. (2011), the syntheses conditions of the 3.65 Å phase from Sclar and Morzenti (1971) and Rice et al. (1989), and knowledge of its correct MgSi(OH)6 composition from this study, we constructed a Schreinemaker’s diagram of reactions involving the 10 Å phase of the composition Mg3Si4O10(OH)2·H2O, brucite, stishovite, high-clinoenstatite 3.65 Å phase, and water (Fig. 5⇓). Accordingly, the 3.65 Å phase is a stable phase above about 9.0 GPa and decomposes at temperatures above about 500 °C due to the reaction 3.65 Å phase = high-clinoenstatite + water. This limited P-T stability together with its high water content of 35 wt% H2O makes the 3.65 Å phase an unrealistic phase in the Earth’s mantle. If at all, the 3.65 Å phase might only exist under hydrous conditions in very limited areas, i.e., in the coldest parts of old and extremely fast subducted slabs, e.g., Tonga (Kirby et al. 1996) as shown in Figure 6⇓. The first formation of the 3.65 Å phase within inner parts of the slab during deep subduction is due to reaction 10 Å phase + brucite + water = 3.65 Å phase at around 9.0 GPa (Fig. 5⇓). With increasing pressure, further 3.65 Å phase-forming reactions are 10 Å phase + brucite = 3.65 Å phase + high-clinoenstatite, 10 Å phase + water = 3.65 Å phase + stishovite, 10 Å phase = 3.65 Å phase + high-clinoenstatite + stishovite. According to Pawley et al. (2011), the latter reaction represents the upper pressure stability of the 10 Å phase.
They also observed formation of phase A and phase E in an experiment in the MSH-system at 9.5 GPa and 450 °C. However, Sclar and Morzenti (1971) and Rice et al. (1989) did not observe such additional DHMS phases, nor did phase A and phase E form in the experiments of our study. Nevertheless, it cannot be ruled out that the phase relations presented in Figure 5⇑ have to be modified by considering further MSH phases, which are stable within the P-T stability field of the 3.65 Å phase.
The P-T stability field of the 3.65 Å phase will perhaps be decreased to lower pressures and/or increased to higher temperatures by additional components, e.g., Al, Fe2+, Fe3+ substituting for Mg or Si in the structure of the 3.65 Å phase, thus making it a potentially more important receptacle for water transport in cold subduction zones. Furthermore, future experiments on testing a possible solid solution between δ-Al(OH)3 and MgSi(OH)6 according to the exchange mechanism 2Al = Mg + Si are strongly required.
The authors are grateful to J. Pohlenz for preparation of the two multi-anvil experiments, A. Schreiber for producing the FIB foils, R. Naumann for determination of the gel composition, H. Kemnitz and D. Rhede for helping during the SEM and EMP analyses, and H.-P. Nabein for sample preparation and XRD measurements. Helpful reviews by A. Pawley (Manchester) and an anonymous reviewer are gratefully acknowledged.
Manuscript handled by Lars Ehm
- Manuscript Received January 10, 2011.
- Manuscript Accepted April 4, 2011.