Experimental constraints on the damp peridotite solidus and oceanic mantle potential temperature

See allHide authors and affiliations

Science  03 Mar 2017:
Vol. 355, Issue 6328, pp. 942-945
DOI: 10.1126/science.aaj2165

Turning up the mantle temperature

The temperature at which Earth's mantle begins to melt is a long-standing question in geology. Sarafian et al. present a clever set of experiments to determine the impact of small amounts of water on the melting temperature of mantle rock (see the Perspective by Asimow). This allowed them to reinterpret geophysical observations of melting in the mantle and revise estimates of mantle temperature upward. A hotter mantle has a multitude of implications for mantle melting and geodynamic processes.

Science, this issue p. 942; see also p. 908


Decompression of hot mantle rock upwelling beneath oceanic spreading centers causes it to exceed the melting point (solidus), producing magmas that ascend to form basaltic crust ~6 to 7 kilometers thick. The oceanic upper mantle contains ~50 to 200 micrograms per gram of water (H2O) dissolved in nominally anhydrous minerals, which—relative to its low concentration—has a disproportionate effect on the solidus that has not been quantified experimentally. Here, we present results from an experimental determination of the peridotite solidus containing known amounts of dissolved hydrogen. Our data reveal that the H2O-undersaturated peridotite solidus is hotter than previously thought. Reconciling geophysical observations of the melting regime beneath the East Pacific Rise with our experimental results requires that existing estimates for the oceanic upper mantle potential temperature be adjusted upward by about 60°C.

New ocean crust is formed by basaltic magmatism along 65,000 km of divergent plate boundaries—known as mid-ocean ridges—that are characterized by elevated topography, shallow seismicity, and high heat flow (1). Mid-ocean ridge basalts (MORBs) are produced by decompression partial melting of mantle rock as it ascends adiabatically beneath the ridge (24). The depth at which mantle rock (e.g., peridotite, pyroxenite, or eclogite) begins to melt is controlled by a combination of mantle potential temperature and the solidus (5). The potential temperature of any parcel of mantle is the temperature that it would have upon ascending adiabatically to the surface without undergoing partial melting and has important implications for mantle dynamics. A recent estimate of the range of potential temperatures in the oceanic upper mantle is 1314° to 1464°C (6). It is generally accepted that the bulk of the upper mantle source of MORBs consists of peridotite, so knowledge of its solidus combined with geophysical observations indicating the presence of partial melt can be used to constrain potential temperature. However, the peridotite solidus is sensitive to the presence of small amounts of hydrogen dissolved in the nominally anhydrous minerals (NAMs) that comprise the rock. In H2O-undersaturated (“damp”) conditions, the melting temperature of peridotite is dramatically lower at any given pressure relative to the anhydrous solidus, thereby increasing the depth at which partial melting begins for any potential temperature (7). Partial melting that begins beneath a ridge at the “damp” solidus involves a larger volume of peridotite and produces a higher mean pressure of melting combined with lower mean extent of melting (8). The bulk oceanic upper mantle is estimated to contain 50 to 200 μg/g H2O (9, 10), but its influence on the onset of peridotite melting has proven challenging to quantify experimentally.

An experimental determination of the “damp” peridotite solidus requires controlling and characterizing the amount of H2O dissolved in NAMs. Despite meticulous drying protocols, a small amount of H2O is present in most high-pressure melting experiments due to adsorption onto starting materials (usually powders) (11). We are generally hindered from measuring the concentrations of H2O in NAMs that grow during partial melting experiments conducted near the peridotite solidus because of small grain sizes (12). Here, we developed an experimental approach that overcomes this difficulty and used it to quantify the solidus temperature for peridotite containing 140 μg/g dissolved H2O at 1.5 GPa near the center of the spinel lherzolite stability field (13). Our approach takes advantage of the consistent low concentration of adsorbed H2O, rapid H diffusion through olivine (14, 15), and strong partitioning of H2O into silicate melt (16, 17).

To determine the “damp” peridotite solidus, we conducted partial melting experiments at 1.5 GPa using a piston cylinder device (18). We added ~300 μm in diameter spheres of San Carlos olivine in 5:95 proportion by weight to a synthetic depleted MORB mantle (DMM) peridotite (19) (table S1). The rapid diffusivity of H in olivine (14, 15) allows a sphere 500 μm in diameter to diffusively equilibrate with the H2O fugacity imposed by the fine-grained peridotite matrix within a few hours (fig. S1) (13). This allowed the San Carlos spheres in our experiments to act as hygrometers with a H content that we analyzed using secondary ion mass spectrometry (SIMS) (Fig. 1 and table S2). Once we measured the H concentration of the spheres, we calculated the bulk H2O content of the peridotite using experimentally determined intermineral partition coefficients (12, 13, 16, 17, 20). All San Carlos olivine spheres that we measured within each capsule contained the same H2O content and are homogenous (see table S2), indicating experimental equilibrium.

Fig. 1 Backscattered electron image of experimentally produced peridotite mineral grains and a San Carlos olivine sphere.

The dashed white line marks the edge of the San Carlos olivine sphere. The black box illustrates the approximate size of a NanoSIMS analysis showing that the peridotite is too fine grained for measurement, but the olivine sphere is large enough to conduct multiple measurements and determine the H2O content of the experiment. Brightness is proportional to mean atomic number.

The initial H2O content of the San Carlos olivine spheres was 6.7 ± 0.1 μg/g (n = 3) (21) (all uncertainties reported as 2 SE). Our analyses of the postexperiment spheres indicate an average H2O concentration of 32 ± 3 μg/g (n = 7) at temperatures below 1320°C (Fig. 2). We interpreted these experiments to be below the solidus because they did not contain any discernible partial melt. We used mass balance calculations based on the major element compositions of the phases (data S1) and the H2O content of the olivine spheres to determine that the bulk peridotite had a H2O content of 140 ± 20 μg/g H2O (n = 5) (table S3). Water is incompatible in the residual solid during peridotite partial melting with a bulk partition coefficient similar to that of the light rare-earth element Ce (10, 17). Therefore, upon crossing the “damp” solidus, H2O should be strongly concentrated into the melt, and the amount of H2O in the olivine spheres is expected to drop. We see this in spheres from experiments conducted above 1320°C, where H2O decreases systematically with increasing temperature (Fig. 2). We interpret these experiments to be above the solidus because they contain partial melt. We conclude from our experiments that the 140 μg/g H2O peridotite solidus is between 1310° and 1330°C (i.e., ~1320°C). For comparison, the parameterization of Hirschmann (11) of the nominally anhydrous solidus (referred to as nominally anhydrous because it is based on experiments with unknown H2O contents) places the initiation of melting at 1.5 GPa and 1308°C (Fig. 2). Our experimentally derived solidus of 1320°C at 1.5 GPa is in agreement with the nominally anhydrous solidus, given the ±10°C temperature uncertainty typically associated with piston cylinder experiments. This level of agreement suggests that ~140 μg/g is a reasonable estimate for the bulk H2O content of previous experiments (Fig. 3). A cryoscopic calculation (12) based on our experimentally determined 140 μg/g H2O peridotite solidus places the 1.5 GPa truly anhydrous (0 μg/g H2O) peridotite solidus at ~1370°C (Fig. 2) (13).

Fig. 2 Average H2O content of San Carlos olivine spheres with temperature at 1.5 GPa.

Subsolidus experiments contain on average 32 ± 3 μg/g H2O, and experiments above the solidus show a general trend of decreasing H2O content with increasing temperature. Error bars are 2 SE in H2O content and ±10°C in temperature. The nominally anhydrous solidus (solid line) (11), the 140 μg/g H2O solidus determined experimentally in this study (dashed line), and the calculated 0 μg/g H2O (truly anhydrous) solidus (dotted line) are shown.

Fig. 3 Comparison of our experimental results with the nominally anhydrous solidus parameterized from earlier partial melting experiments.

The nominally anhydrous solidus (bold line) was parameterized based on the presence (inverted triangles) and absence (filled triangles) of melt in experiments containing unknown concentrations of H2O (11). Experiments from this study conducted at 1.5 GPa and containing 140 μg/g H2O (gray squares are experiments that contain melt; black squares are melt-free experiments) show a strong agreement with the nominally anhydrous solidus, indicating that it is actually damp. Figure adapted from (11).

Magnetotelluric (MT) observations indicate the presence of a highly conductive region at 20- to 90-km depth beneath the ridge crest along the 8° to 11°N segment of the East Pacific Rise (EPR), which has been interpreted as the portion of the ascending mantle undergoing substantial partial melting in response to passive upwelling. The conductivity of partial melt is orders of magnitude higher than that of residual peridotite, so that MT is a sensitive tool for identifying the presence of melt in the oceanic upper mantle (22). If, as in previous studies, we assume that the nominally anhydrous peridotite solidus represents the 0 μg/g H2O solidus, an adiabat with a potential temperature of 1350°C would cross the anhydrous solidus at 70-km depth (11), ~20 km from the base of the observed conductive region. The presence of 200 μg/g H2O, thought to be in the MORB source, would lower this (assumed 0 μg/g H2O) peridotite solidus such that melting at the same potential temperature would begin at 86-km depth (20), near the base of the conductive anomaly (Fig. 4, A and B). This suggests that the main region of partial melting under the EPR identified by the mantle conductivity structure occurs when the mantle adiabat crosses the damp solidus.

Fig. 4 Reconciling magnetotelluric observations with the experimental results.

(A) The conductivity structure of the Northern East Pacific Rise. Warm colors show the melting regime under the ridge. Black lines show the nominally anhydrous solidus (dot-dashed) and 200 μg/g H2O solidus (dashed) of previous studies (left side) (11, 12), corresponding to a mantle potential temperature (Tp) of 1350°C. White lines show the 0 μg/g H2O (dot-dashed) and 450 μg/g H2O solidi (dashed) from this study (right side) at the same Tp. Red lines are the 0 μg/g H2O (dot-dashed) and 200μg/g H2O solidi (dashed) from this study, corresponding to a mantle Tp of 1410°C. From (22); reprinted by permission from Macmillan Publishers Ltd., copyright 2013, (B) The solidi and conductivity structure with depth beneath the ridge from (A), and the corresponding 1350° (bold black) and 1410°C (bold dashed black) mantle adiabats. Black lines are the nominally anhydrous (solid) and corresponding 200 μg/g H2O (dashed) solidi from previous studies (11, 12). Red lines are the 0 μg/g H2O (solid), and corresponding 200 (dashed) and 450 μg/g H2O (dot-dashed) solidi based on this study. The kink in the solidi at 2.8 GPa represents the spinel to garnet transition. The adiabat-solidi intersections illustrate depth where melting begins.

Our new experimental results can be used to reevaluate this potential temperature estimate. If we cryoscopically calculate the 0 μg/g H2O peridotite solidus on the basis of results from our experiments containing ~140 μg/g H2O (Fig. 2), 0 μg/g H2O peridotite ascending along a 1350°C potential temperature adiabat would begin to melt at a depth of 52 km beneath the EPR, well within the most conductive (highest melt fraction) region indicated by the MT data (22). To lower the solidus temperature such that the initiation of partial melting is at 86-km depth along the lower edge of the observed region of partial melting requires ~450 μg/g H2O dissolved in the oceanic upper mantle (Fig. 4, A and B). Although there is evidence for an enriched mantle component in the Pacific upper mantle containing up to ~660 μg/g H2O (23), MORB glasses from the 9° to 10°N portion of the EPR indicate a mantle source containing ~130 μg/g H2O (24). Therefore, reconciling our experimental results with the MT observations for a potential temperature of 1350°C requires unrealistically high H2O contents for the depleted oceanic upper mantle if partial melts formed beneath the ridge are quantitatively aggregated to the ridge axis (10).

Our study suggests that the temperature of the 0 μg/g H2O peridotite solidus is ~60°C higher at 1.5 GPa than previously thought, so that current estimates of potential temperature and H2O content of the oceanic upper mantle are unable to explain the geophysical observations along this portion of the EPR. To match the conductivity anomaly given the solidus determined in this study, the mantle adiabatic potential temperature beneath the northern EPR must be 1410°C (Fig. 4, A and B) (13). This increase in mantle potential temperature is based on the assumption that deep melting in the observed melt region from MT imaging is due solely to mantle peridotite crossing the “damp” solidus. If an enriched material, such as pyroxenite, is a major mantle component, the deeper melting signal in the MT data could be a result of the enriched material melting (23), and the mantle potential temperature required to match the observed conductivity would be lower. The depth at which pyroxenite partial melting begins (and ends) is strongly dependent on composition, with some pyroxenites melting at depths equal to or shallower than peridotite. Further, for a given combination of crustal thickness and potential temperature, there is a trade-off between the amount of pyroxenite in the oceanic mantle and its composition (25). Therefore, it is difficult to assess the potential contribution of pyroxenite to the observed deep melting signal.

More broadly, our experimental results indicate that mantle potential temperatures along all ocean spreading centers are hotter than existing estimates. The highest temperatures found at the Reykjanes Ridge may potentially require an increase up to 1530°C (6). A hotter interior has important implications for key mantle dynamic parameters such as viscosity (26). Higher mantle potential temperatures reduce mantle viscosity and may provide the weakening of the asthenosphere necessary to explain plate motion at the lithosphere-asthenosphere boundary (LAB), the boundary between the rigid tectonic plate and the convecting mantle. The LAB structure is known to be complex, containing discontinuities associated with the seismic low-velocity zone and the Gutenberg discontinuity, making it difficult to discern the global mechanism controlling plate behavior. Two mechanisms often invoked are H2O dissolved in the peridotite (7) or the presence of partial melt (27) in the upper asthenosphere. However, although additional mechanisms may be at work at the local scale, an increase in mantle potential temperature based on our experimental findings may show that global plate motion has a simple thermal origin as recent basin-wide seismic waveform studies (28) and localized conductivity measurements (29) in the Pacific suggest.

Supplementary Materials

Materials and Methods

Fig. S1

Tables S1 to S3

Data S1

References (3032)

References and Notes

  1. Materials and methods are available as supplementary materials.
Acknowledgments: Funding for this research was provided by the National Science Foundation, grant number OCE-1459649, and Woods Hole Oceanographic Institution Deep Ocean Exploration Institute. E.H.H. thanks the Carnegie Institution for Science and the Deep Carbon Observatory for support. We thank K. Key for use of the magnetotelluric data. We are grateful to three anonymous reviewers for providing constructive suggestions that improved the paper. The data reported in this study can be found in the supplementary materials. E.S. and G.A.G. initiated the project and designed the experiments. E. Sarafian performed the experiments. E.S. and A.R.S. performed and E.H.H. oversaw the SIMS analyses. All authors contributed to developing the ideas and writing the manuscript.
View Abstract

Navigate This Article