Water Solubility in Aluminous Orthopyroxene and the Origin of Earth's Asthenosphere

See allHide authors and affiliations

Science  19 Jan 2007:
Vol. 315, Issue 5810, pp. 364-368
DOI: 10.1126/science.1135422


Plate tectonics is based on the concept of rigid lithosphere plates sliding on a mechanically weak asthenosphere. Many models assume that the weakness of the asthenosphere is related to the presence of small amounts of hydrous melts. However, the mechanism that may cause melting in the asthenosphere is not well understood. We show that the asthenosphere coincides with a zone where the water solubility in mantle minerals has a pronounced minimum. The minimum is due to a sharp decrease of water solubility in aluminous orthopyroxene with depth, whereas the water solubility in olivine continuously increases with pressure. Melting in the asthenosphere may therefore be related not to volatile enrichment but to a minimum in water solubility, which causes excess water to form a hydrous silicate melt.

Earth's asthenosphere is often assumed to roughly coincide with the low-velocity zone, a layer of reduced seismic velocities and increased attenuation of seismic waves. The low-velocity zone usually begins at a depth of 60 to 80 km below the oceans and ends around 220 km (1). Below continental shields, the upper boundary is depressed to 150 km. The seismic characteristics of the low-velocity zone could be easily explained by the presence of a small fraction of partial melt as intergranular film (1, 2). Some water is required to generate such a partial melt in the mantle, as mantle temperatures at the relevant depths are below the dry melting point of peridotite but above the water-saturated solidus (3, 4). Originally, it was believed that the top of the low-velocity zone corresponds to the stability limit of hydrous phases such as phlogopite or hornblende (5, 6). However, this is unlikely because the solubility of water (and of alkalis) in nominally anhydrous mantle minerals (79) is so high that separate hydrous phases such as amphiboles and phlogopite are not stable in an upper mantle of pyrolite composition.

If the low-velocity zone were due to partial melting, the existence of a lower boundary would be even more difficult to understand, as the geotherm remains above the water-saturated solidus with increasing depth. Moreover, it is unclear whether low degrees of partial melt in the mantle would form an intergranular film (10), as dry basaltic melts do not wet mantle minerals and therefore tend to form isolated pockets. Accordingly, alternative models have been proposed. These models (10, 11) are based on the observation that water dissolved in mantle minerals such as olivine reduces both the strength of the mineral and the seismic velocities. The boundary between lithosphere and asthenosphere may then correspond to a boundary in intracrystalline water content, with the asthenosphere being water-rich, whereas the oceanic lithosphere is depleted in water as a result of the melt extraction at mid-ocean ridges. The presence of partial melt in the asthenosphere is not required in these models. However, they cannot explain the existence of a low-velocity zone below continental shields, as the mechanism of magma production below continents is very different from that prevailing at mid-ocean ridges.

All models of the low-velocity zone depend on the presence of water and its solubility in mantle minerals. The main constituents of the upper mantle are olivine and orthopyroxene (enstatite). The water solubility in both olivine and Al-free orthopyroxene is quite comparable and increases with pressure and temperature (79, 12). However, aluminum is known to greatly enhance water solubility in orthopyroxene, and at high Al contents, water in orthopyroxene may dominate the water budget in the mantle. Throughout Earth's upper mantle, olivine and orthopyroxene usually coexist with small amounts of an aluminous phase such as spinel or garnet. Therefore, we experimentally studied the solubility of water in aluminous MgSiO3 enstatite in equilibrium with spinel or garnet (i.e., under conditions of aluminum saturation).

Experiments were carried out in an endloaded piston-cylinder apparatus. Mixtures of Mg(OH)2, Al(OH)3, and SiO2 were sealed in platinum-rhodium (Pt0.95Rh0.05) capsules together with about 20 weight % of liquid water. The stoichiometry of the starting mixture was chosen to correspond to aluminous orthopyroxene plus small amounts of olivine and spinel or garnet. In some experiments, homogeneous mixtures of the starting chemicals were used. In most experiments, however, alternating layers with low and high silica content were introduced into the capsule to reduce nucleation rates so as to grow larger crystals. In addition, some experiments were carried out with amorphous gels and liquid water as starting material. Experiments were run at 15 to 35 kbar and 800° to 1100°C for a few days. After the experiments, the capsules were opened. Capsules that did not release excess water after the experiment were discarded. Perfectly clear and inclusion-free single crystals of orthopyroxene were picked from the charges, optically oriented, and doubly polished. Polarized infrared spectra (Fig. 1) were measured with a Bruker IFS 125 Fourier-transform infrared spectrometer coupled with an IRscope I microscope (tungsten source, CaF2 beam splitter, narrow-band mercury cadmium telluride detector, Al strip polarizer on KRS 5 substrate). Water contents were calculated from the infrared data with the use of the extinction coefficients of Bell et al. (13). Chemical analyses were carried out with a JEOL 8900 RL electron microprobe (15 kV, 15 nA, 120 s counting time per spot, focused beam).

Fig. 1.

Polarized infrared spectra of two aluminum-saturated enstatite crystals synthesized at 1100°C and 15 kbar. Water contents (in ppm) correspond to the absorbances measured parallel to the three crystallographic axes. Bulk water contents are obtained by adding these values. Numbers in parentheses denote percentage of total water observed in each of the three directions of polarization. The spectra shown here and in Fig. 2 were obtained from the raw data by subtracting a linear baseline defined by the points at 3800 cm–1 and 2800 cm–1. The two weak features between 2800 cm–1 and 3000 cm–1 could be due to organic surface contamination. However, they persisted after repolishing and they may therefore be intrinsic to the sample.

The water contents in the aluminous pyroxenes are strikingly high, reaching values close to 1 weight % at low pressures and temperatures (Table 1 and Fig. 2). Water solubilities clearly decrease with both pressure and temperature, opposite to the behavior observed for olivine and Al-free enstatite. The high water solubilities are correlated with anomalously high Al contents in the pyroxenes, which are much higher than those predicted from existing thermodynamic models and experimental calibrations (1416). However, in previous studies, only a few experiments were carried out at the low pressures and temperatures where we observe high Al and water contents, and the water fugacity was probably not carefully controlled in all experiments. The high water contents appear to be intrinsic to the pyroxenes. The presence of foreign phases in the crystals is unlikely, because the infrared bands are strongly polarized (Fig. 1) and measurements were taken only on perfectly clear and optically inclusion-free crystals. To rule out any impurities at the submicroscopic level, we investigated several orthopyroxene crystals by transmission electron microscopy. The structure of the pyroxene crystals (Fig. 3) is undisturbed without any foreign phases or linear and planar defects. The high water contents are therefore definitively due to OH point defects in the structure.

Fig. 2.

Polarized infrared spectra (electrical field vector E parallel to the crystallographic c axis) and total water contents of aluminum-saturated enstatite at 800° to 1100°C and 15 to 35 kbar. Numbers are average water contents derived from the bulk water contents of all samples synthesized at a given pressure and temperature; n is the number of measurements. Bulk water contents were calculated as explained in Fig. 1. The water content of the sample synthesized at 25 kbar and 800°C, where only two crystals could be measured, is probably somewhat below the real water solubility under these conditions.

Fig. 3.

Representative high-resolution transmission electron microscopy image taken along the [010] direction and corresponding selected area electron diffraction pattern (inset; a* and c* axes are shown) of an aluminum-saturated enstatite with 0.75 weight % water. The sample was synthesized at 15 kbar and 900°C. No evidence for foreign phases or planar defects can be detected in this image.

Table 1.

Experimental results on water solubility in aluminum-saturated enstatite. All experiments were carried out using an oxide-hydroxide mixture as starting material, except as noted. Number of infrared measurements refers to the numbers of different spots measured, usually on different crystals. Water contents (in ppm by weight) were calculated from the infrared data according to two different extinction coefficients, from Bell et al. (13) and from Paterson (28). Errors are one standard deviation. Water contents in atoms H per 106 Si can be obtained by multiplying by a factor of 22.3. All data reported here for a given pressure and temperature were included in the calibration of the thermodynamic model of water solubility. Solid phases detected refer to those phases that could be directly observed by x-ray diffraction or Raman spectroscopy. However, all the samples must have contained some aluminous phase, probably spinel or pyrope, as the aluminum content of the orthopyroxene was always several weight % lower than the aluminum content of the starting material. A fluid phase was always present during the experiments, as all samples released considerable amounts of water upon opening of the capsules. No evidence for melting was seen, as the run products were usually loose powders without interstitial glass. Some fluffy material and isolated glass beads probably represent material precipitated from the fluid upon quenching. En, enstatite; Sp, spinel; Ol, olivine; Prp, pyrope; Crn, corundum; Ky, kyanite; Sr, sapphirine; Prl, pyrophyllite. Al contents in enstatite were sometimes slightly inhomogeneous. The numbers and standard deviations given were usually derived from more than 100 individual analyses of different crystals in the charge. The molar Mg/Si ratio in all samples is equal to 1 within the error of the measurement.

SampleT (°C)P (kbar)Duration (hours)Number of infrared measurementsWater content (ppm)Solid phases detectedAl2O3 content in enstatite (weight %)
Bell et al. (View inline)Paterson (View inline)
En63 800 15 168 2 8420 ± 750 6280 ± 235 En, Crn 9.2 ± 0.99
En59/4 900 15 70 6 3710 ± 795 2810 ± 575 En, Ol, Sp, Sr 10.05 ± 0.85
En59/1 900 15 168 5 8380 ± 3030 6730 ± 2280 En, Crn 8.99 ± 1.5
En59 900 15 168 14 8290 ± 2650 6720 ± 2210 En, Crn 11.9 ± 2.22
En60 1000 15 120 6 3110 ± 600 2440 ± 465 En 8.28 ± 2.2
En47View inline 1100 15 120 5 2460 ± 960 1860 ± 780 En, Ol, Sp 6.7 ± 1.33
En47/1 1100 15 120 4 1730 ± 450 1340 ± 325 En 8.77 ± 2.16
En47/2 1100 15 168 2 1590 ± 185 1290 ± 255 En, Sp, (Ol) 8.96 ± 1.46
En86 800 25 168 2 4670 ± 655 3590 ± 600 En, Ky 6.67 ± 2.44
En85 900 25 168 4 6400 ± 1330 5040 ± 885 En 5.57 ± 1.56
En84 1100 25 120 2 1420 ± 115 1080 ± 190 En, Ol 5.4 ± 0.84
En87View inline 1000 35 120 2 1680 ± 490 1140 ± 340 En, Prp 1.73 ± 0.21
En87/2 1000 35 120 6 2370 ± 450 1640 ± 435 En, Ol, Prp 1.21 ± 0.61
En70View inline 1100 35 168 1 1230 960 En, Ol, Prp 1.75 ± 0.65
En90/1 1100 35 120 3 1500 ± 305 1150 ± 280 En, Ol, Prp, Prl 1.57 ± 0.25
  • View inline* Synthesized from gels.

  • Electron microprobe analyses suggest that most of the water is dissolved by the coupled substitution of Al3+ + H+ for Si4+ and by the substitution of Al3+ + H+ for 2 Mg2+. Both substitutions appear to occur with roughly equal abundance; that is, Al is distributed about equally among tetrahedral and octahedral sites, irrespective of water content. Both substitution mechanisms imply a molar 1:1 ratio between Al and H. This is consistent with the observation that the “excess” of aluminum in the orthopyroxenes relative to existing calibrations (1416) roughly equals the water content, if both Al2O3 and H2O are expressed in molar fractions. The substitution mechanism was confirmed by a single-crystal x-ray diffraction structure refinement of one aluminous pyroxene containing 7500 ppm (by weight) of water. The structure refinement yielded 5% vacancies on one of the Mg sites [M2 (17)], consistent with H+ substituting for Mg2+ and a significantly enlarged polyhedral volume of one of the Si sites [Si2 (17)] of 1.6575 Å 3, consistent with a substitution of Al3+ + H+ for Si4+. Structurally, the decrease in H and Al contents with increasing pressure results from the pressure destabilization of tetrahedral Al.

    The systematic variations in water content with pressure and temperature observed in this study (Fig. 2) suggest that the water contents represent true equilibrium solubility. This is also supported by the observation that runs with different starting materials (oxide mixture and gels) yield similar results. To describe the water solubility in orthopyroxene coexisting with olivine and an aluminous phase as a function of pressure, temperature, and water fugacity, we calibrated a model that describes the water solubility in aluminous enstatite as the sum of the water solubility in Al-free enstatite and the water solubility coupled to aluminum. The water solubility in Al-free enstatite was previously calibrated (8, 9) and can be expressed as Math(1) Math where A = 0.01354 ppm/bar, fH2O is water fugacity [calculated using the equation of state of (18)], ΔH1bar = –4563 J/mol, ΔV solid = 12.1 cm3/mol, R is the gas constant, P is pressure, and T is absolute temperature. The additional water solubility due to Al can then be described by Math(2) Math

    Note, however, that in Eq. 2 water fugacity enters as a square-root term (19) because the coupled substitution with Al yields isolated OH groups, unlike the OH pairs that result from the substitution of 2 H+ for Mg2+ in pure enstatite (8, 9). A least-squares fit of all experimental data to Eq. 2 yielded AAl = 0.042 ppm/bar0.5, ΔH1barAl = –79,685 J/mol, and ΔV solidAl= 11.3 cm3/mol. The total water solubility in orthopyroxene coexisting with olivine and either spinel or pyrope can now be calculated at any pressure and temperature by adding the results from Eqs. 1 and 2. This is consistent with observations from previous studies that the water solubility coupled to Al and the water solubility in Al-free enstatite are due to different and independent defects, with the bulk water solubility being the sum of the individual defect solubilities (8, 20). Only pressure and temperature are required to calculate the equilibrium water content in the Al-saturated orthopyroxene. This is because according to the phase rule, in a four-component system (MgO-Al2O3-SiO2-H2O) the coexistence of four phases (orthopyroxene, olivine, aluminous phase, and fluid) only leaves two degrees of freedom. Therefore, if pressure and temperature are given, all compositional variables in the system are determined.

    Bulk mantle water solubility has a pronounced minimum (Fig. 4) between 150 and 200 km depth, coinciding with the location of the seismic low-velocity zone (shaded) below continental shields. The minimum is due to the sharp decrease of water solubility in aluminous orthopyroxene with temperature and also with pressure, whereas water solubility in olivine continuously increases with pressure and temperature (7, 12). As shown in Fig. 4, at a bulk water content of about 800 ppm, the mantle in the low-velocity zone would be oversaturated with water (i.e., the water activity would equal 1). However, as the geotherm at this depth is located above the water-saturated peridotite solidus (3, 4) of about 1000°C, a hydrous melt will form in the presence of sufficient amounts of water. Because the temperature of the geotherm is far above the water-saturated solidus under these conditions, a water activity around 0.1 is probably sufficient to induce melting (21). This water activity would imply that a few hundred ppm of water are sufficient to generate a small fraction of hydrous melt in the asthenosphere. Such water contents are to be expected in the upper mantle (2224). If the same calculation is carried out for a hotter oceanic geotherm (Fig. 4, right panel), the upper boundary of the zone of minimum water solubility is lifted to a depth of only 60 to 80 km, consistent with the position of the low-velocity zone below oceans. Moreover, this behavior also provides a straightforward explanation for the seismic observation that the top of the low-velocity zone is very sharp and well defined, whereas the lower boundary is more diffuse and difficult to locate (2, 25). As the water solubility in mantle minerals sharply increases with decreasing depth, the fraction of partial melt in equilibrium with these minerals will also sharply decrease at the asthenosphere-lithosphere boundary. On the other hand, toward the lower boundary of the asthenosphere, the decrease in melt fraction will be more gradual, reflecting the gradual increase of water solubility in olivine and orthopyroxene.

    Fig. 4.

    Water solubility (in ppm by weight) in upper-mantle minerals as a function of depth for a typical continental shield and oceanic geotherm (26). The typical position of the low-velocity zone (LVZ) below continental shields and below oceans is shaded in gray. Water solubility in olivine is according to Kohlstedt et al. (7). Water solubility in aluminum-saturated enstatite was calculated from Eqs. 1 and 2. Recently (27), it was suggested that the infrared extinction coefficient of water in olivine may be considerably smaller than previously thought. If this new calibration were applied, the water solubility in olivine would increase by a factor of about 2.5. This would somewhat sharpen the minimum in the bulk water solubility curves shown above and move them to higher water contents. It would not, however, change the general shape of the curves or the position of the inferred boundaries of the asthenosphere. The effect of using different experimental data for water solubility in olivine as well as the effect of changing the ratio of orthopyroxene to olivine is further discussed in (21).

    Our results therefore support the concept that the low-velocity zone may be related to partial melting (1, 2, 6). However, even in the absence of melting, the partitioning of water between olivine and orthopyroxene would strongly depend on depth. The high water solubilities in aluminous orthopyroxene at low pressure and temperature will effectively “dry out” olivine, and this may also contribute to a stiffening of the lithosphere. In any case, however, our results imply that the existence of an asthenosphere—and therefore of plate tectonics as we know it—is possible only in a planet with a water-bearing mantle.

    Supporting Online Material

    SOM Text

    Figs. S1 and S2


    References and Notes

    View Abstract

    Stay Connected to Science

    Navigate This Article