Geophysical Research Letters

Mantle wehrlite from Hess Deep as a crystal cumulate from an ultra-depleted primary melt in East Pacific Rise



[1] A piece of wehrlite, containing about 20 volume % of clinopyroxene, was found in harzburgite in a drill core of ODP Leg 147 from Hess Deep near the East Pacific Rise. Minerals are as refractory (olivine, Fo90.7; spinel, Cr# = 0.52) as those in the harzburgite. The clinopyroxene is depleted in REE (rare earth elements), being similar in chondrite-normalized pattern for middle to heavy REE to that of the harzburgite. The calculated melt in equilibrium with the clinopyroxene is as depleted as ultra-depleted MORB melts ever documented. The wehrlite is a cumulate from an in-situ segregated ultra-depleted MORB melt. Dimension of individual masses of the ultra-depleted MORB involved may be at least as large as a few tens of centimeters. The Hess Deep wehrlite is different from the lower crustal wehrlite from the ocean floor or some ophiolites, which is either a cumulate or a peridotite/melt reaction product.

1. Introduction

[2] The mid-ocean ridge basalts (MORB) released from the depleted mantle produce the oceanic crust. Special attention has been paid to behavior of the primary MORB melt beneath the mid-ocean ridge since disequilibria have been described between the erupted MORB and the mantle peridotite [e.g., Johnson et al., 1990]. There has not been direct evidence to indicate how the primary melt evolves to MORB beneath the spreading ridge [e.g., Kelemen et al., 1997], because deep-seated rocks that are precipitated from the primary melt have not been discovered from the ocean floor. The primary melt in equilibrium with the depleted mantle has been available only as minute inclusions within phenocryst minerals of MORB [e.g., Sobolev and Shimizu, 1993; Kamenetsky, 1996; Danyushevsky et al., 2003]. Very few documents have referred to direct linkage between abyssal refractory residual peridotite and its primary melt counterpart [cf. Koga et al., 2001]. Problems on the behavior of the primary MORB melt have not been fully unraveled at the upper mantle condition. We found a wehrlite within a drill core of ODP (Ocean Drilling Program) Leg 147 from Hess Deep near East Pacific Rise (EPR). The wehrlite may be a crystal cumulate, and will provide us with unrivaled information on the behavior of incipient melts beneath a fast spreading mid-ocean ridge. It may fill the missing link between the primary melt and MORB.

2. Geological Background

[3] This rock was found as a short interval in a mantle peridotite section of drill cores (site 895C) during Leg 147 of ODP [Arai and Matsukage, 1996]. Seven holes were drilled at Site 894 and six holes at Site 895 on Hess Deep around a propagating tip of the Cocos-Nazca spreading center, near East Pacific Rise, during ODP Leg 147 [Gillis et al., 1993]. Rocks of ophiolitic assemblage were discovered from Hess Deep [Francheteau et al., 1992], and deep-seated rocks were also extensively studied for dredged samples [Girardeau and Francheteau, 1993; Hekinian et al., 1993]. ODP Leg 147 (1992 to 1993) was an epoch-making cruise to drill for the first time through the deep-seated rocks exposed on the ocean floor of fast-spreading mid-ocean ridge origin. Gabbroic rocks of the lower crust and ultramafic rocks of the Moho transition zone to upper mantle were mainly obtained at Site 894 and Site 895, respectively [Gillis et al., 1993]. Harzburgites are dominant over dunites in some holes and vice versa in others at Site 895 [Gillis et al., 1993].

3. Sample Descriptions

[4] The wehrlite was found as Piece 22 of the core #147-895C-3R-1, which is harzburgite-dominant. Harzburgites (olivine + orthopyroxene + chromian spinel ± clinopyroxene) from Hess Deep are the most depleted in magmatic components of all mantle peridotites ever documented from the ocean floor [Dick and Bullen, 1984; Arai, 1994; Dick and Natland, 1996; Hellebrand et al., 2001], and can be a refractory residue for ultra-depleted melts. Dunites (olivine + chromian spinel) and troctolites (olivine + plagioclase + clinopyroxene + chromian spinel) are a reaction product between the primary MORB and mantle harzburgite at the shallowest mantle [Dick and Natland, 1996; Arai and Matsukage, 1996].

[5] The wehrlite was initially described as harzburgite [Gillis et al., 1993], but was later re-examined and identified as wehrlite [Arai and Matsukage, 1996]. The wehrlite is unclear in relation with the surroundings due to low recovery rate but seems to be located within harzburgite lithology [Arai and Matsukage, 1996]. The wehrlite portion may be around 10 cm in dimension in the core [Gillis et al., 1993; Arai and Matsukage, 1996]. The wehrlite sporadically contains small amount (<1%) of bastite serpentinite possibly pseudomorphous after orthopyroxene, suggesting the boundary with harzburgite is gradual. It is mainly composed of olivine and clinopyroxene (around 20 volume %) with a small amount of chromian spinel (Figure 1). The clinopyroxene is less anhedral and coarser, up to a few mm across, than in the harzburgite (Figure 1). It is altered along cleavage or parting planes to various extents. The olivine is turbid with many trails of melt inclusions. The chromian spinel is opaque and subhedral. Plagioclase is absent in the wehrlite.

Figure 1.

Photomicrograph of the wehrlite from Hess Deep, EPR. Plane-polarized light. Clinopyroxene has been altered to various degrees, but unaltered parts are available (e.g., the central right grain).

4. Mineral Chemistry

[6] Major-element compositions of minerals were determined with an electron microprobe (JXA8800, JEOL) at Kanazawa University. Analytical conditions were 20 kV accelerating voltage, 20 nA probe current and 2 to 3 μm probe diameter for minerals. Ferrous and ferric iron contents in chromian spinel were calculated based on spinel stoichiometry. Minerals (olivine, clinopyroxene and chromian spinel) in the wehrlite sample (Table 1) are almost equivalent in major-element chemistry to those in harzburgites from Hess Deep [Arai and Matsukage, 1996; Dick and Natland, 1996]. The clinopyroxene is clearly more depleted in incompatible minor elements (Table 1) than those in dunites, troctolites and olivine gabbros [Arai and Matsukage, 1996].

Table 1. Selected Microprobe Analyses of Minerals in Wehrlite From EPRa
  • a

    Fe2+ and Fe3+ were calculated assuming spinel stoichiometry. FeO*, total iron as FeO; nd, not detected; Mg#, Mg/(Mg + Fe2+) atomic ratio; Cr#, Cr/(Cr + Al) atomic ratio; YCr, YAl and YFe, fractions of Cr, Al and Fe3+, respectively, to (Cr + Al + Fe3+).

Cr# 0.522 
Y Cr 0.545 
Y Al 0.411 
Y Fe 0.044 

[7] Trace-element characteristics of the clinopyroxene were determined by laser-ablation ICP-MS (Agilent 7500 with GeoLas Excimer laser system) at Kanazawa University [Morishita et al., 2005] (Table 2). The energy of laser ablation was 5 Hz and 8 J/cm2. The NIST 612 standard glass was used for calibration with Si as an internal standard. We used a 100-μm laser beam for analysis to collect sufficient signals for trace elements, but tried to exclude cleavage or parting planes. The clinopyroxene from this wehrlite exhibits very low contents of rare-earth elements (REE) (Figure 2). We analyzed cores and rims of the clinopyroxene grains to reveal chemical homogeneity in REE contents. The wehrlite clinopyroxene is far poorer in REE than clinopyroxenes in dunite-troctolite-gabbro associated with harzburgites from Hess Deep [Dick and Natland, 1996]. It shows a spoon-like chondrite-normalized pattern with a sharp decrease from heavy to middle REE and flat light REE contents (Figure 2). This pattern is similar to those of harzburgite clinopyroxenes for heavy to middle REE abundances (Figure 2). The slight enrichment of light REE in the wehrlite clinopyroxene is apparently not related with sea-floor alteration [e.g., Hellebrand et al., 2001], because it is not dependent on presence or absence of parting or cleavage planes, along which highly mobile elements were possibly incorporated during alteration. Metasomatic enrichment by percolated melt is possible [e.g., Johnson et al., 1990], but is not consistent with the depleted REE patterns of clinopyroxene in the Hess Deep harzburgite [Dick and Natland, 1996].

Figure 2.

Chondrite-normalized REE patterns for clinopyroxenes from the wehrlite and calculated equilibrium melts. Partition coefficients of clinopyroxene/melt are after Hart and Dunn [1993]. Related materials are also plotted for comparison. Chondrite abundances are after Anders and Grevesse [1989]. The wehrlite clinopyroxenes are almost comparable in middle to heavy REE to depleted clinopyroxenes in harzburgite from Hess Deep (HD), EPR [Dick and Natland, 1996]. The calculated melts in equilibrium with the wehrlite clinopyroxenes are almost similar in middle to heavy REE contents to ultra-depleted melt (UDM) inclusions in olivine from the Mid-Atlantic ridge (MAR) [Sobolev and Shimizu, 1993] as well as from the Siqueiros Transform Fault, EPR [Danyushevsky et al., 2003]. They are far more depleted than normal to magnesian MORB from EPR, Hess Deep [Allan et al., 1996] and the Siqueiros Transform Fault [Perfit et al., 1996].

Table 2. Trace-Element Contents of Clinopyroxenes in the Wehrlite Drilled From Hess Deep, EPRa
ElementsHdp 7Hdp 8Hdp 9Hdp 10Hdp 11
  • a

    Determined with La-ICP-MS; nd, not detected. Unit is ppm.


5. Melt in Equilibrium With the Hess Deep Wehrlite

[8] The mineral chemical characteristics strongly indicate the wehrlite formed from a melt in equilibrium with the most refractory harzburgite ever documented from mid-ocean ridge environments [Johnson et al., 1990; Dick and Natland, 1996] (Figure 2). The calculated equilibrium melt compositions using proposed partition coefficients [Hart and Dunn, 1993] are far more depleted in REE than ordinary MORB, indicating an involvement of a depleted melt in this wehrlite genesis. The calculated melt is almost similar to an ultra-depleted melt trapped by olivine phenocryst in MORB [Sobolev and Shimizu, 1993] for middle to heavy REE (Figure 2). It is also more depleted than a primitive picritic melt [Perfit et al., 1996] as well as glass inclusions in its olivine [Danyushevsky et al., 2003] from the Siqueiros Transform Fault, EPR (Figure 2). The relative enrichment of light REE in the melt involved in the wehrlite genesis (Figure 2) is due to reaction with wall harzburgite [e.g., Kelemen et al., 1997; Shimizu, 1998]. Shimizu [1998] reported olivine-hosted melt inclusions with variable light REE patterns but with less various heavy REE levels in a single lava from the FAMOUS area of the Mid-Atlantic Ridge. This would be equivalent to ultra-depleted melts with variable degrees of light REE enrichment: from light-REE-depleted [Sobolev and Shimizu, 1993] to light-REE-enriched from Hess Deep (this study) (Figure 2).

6. Origin of the Hess Deep Wehrlite

[9] The ultra-depleted melt may have been the last, shallowest incremental partial melt formed around the top of mantle diapir beneath EPR [Johnson et al., 1990]. Ordinary MORB evolve from primary incremental melts by mixing [Johnson et al., 1990; Koga et al., 2001] and/or through interaction with peridotite [Arai and Matsukage, 1996; Dick and Natland, 1996], during ascent from the partial melting zone. The ordinary MORB can be in equilibrium with dunite-troctolite [Kelemen et al., 1995; Arai, 2005], interpreted as a melt/harzburgite reaction at the shallowest mantle [Arai and Matsukage, 1996; Dick and Natland, 1996]. We interpret that the wehrlite was precipitated at higher pressures [Herzberg, 2004] from an ultra-depleted MORB almost in-situ segregated from residual harzburgite around the depth of final release of partial melt beneath EPR. In this case, the melt was almost in equilibrium with harzburgite because of absence of dunite around the boundary and presence of orthopyroxene within wehrlite. The presence of a small amount of clinopyroxene in the depleted harzburgite [e.g., Arai and Matsukage, 1996; Dick and Natland, 1996] indicates that the highest-degree melting may be around “clinopyroxene-out” beneath a mid-ocean ridge. For such a depleted melt, we expect diopsidic clinopyroxenes to crystallize after olivine at this condition, taking the pressure-dependent shift of liquidus boundary lines in the peridotite system into account [Kushiro, 1969]. Since the only cumulate (wehrlite) has been left within the harzburgite, an open-system crystallization process probably occurred.

7. Other Implications

[10] The size of the wehrlite piece (about 10 cm) in the drill core may suggest the dimension of the mass of a batch of the primary depleted melt segregated within the residual peridotite to be at least several tens of centimeters, although caution should be taken [Koga et al., 2001]. Alternatively but less probably, larger wehrlite masses of deep origin, if any, have been replaced with dunite and related rocks and obliterated during vigorous intrusion of successive melts after uplifting to shallower levels beneath the mid-ocean ridge [Arai and Matsukage, 1996; Dick and Natland, 1996].

[11] Wehrlitic rocks can be also formed at shallower levels through crystal accumulation or reaction between peridotite and MORB. Plagioclase wehrlite was found from Hess Deep [Girardeau and Francheteau, 1993]. Wehrlitic lithology was documented from a gabbro-dominant part of the Atlantis Massif, one of the oceanic core complexes of the Atlantic [Ildefonse et al., 2006]. These wehrlites, however, contain plagioclase, and may be primitive cumulates at low pressures [Ildefonse et al., 2006] or harzburgite/melt reaction products [Girardeau and Francheteau, 1993]. They are distinct from the mantle wehrlite from Hess Deep both in mode of occurrence and in lithology/petrography. The Hess Deep wehrlite reported here is also distinct from the lower crustal wehrlite [Benn et al., 1988] in the southern Oman ophiolite (Figure 2) [Koga et al., 2001]. The clinopyroxene-rich (up to 19 volume %) varieties of the latter tend to contain plagioclase (up to 17 volume %) [Koga et al., 2001], which is absent in the former.


[12] We greatly acknowledge K. Matsukage, E. Isobe, A. Tamura, Y. Ishida, and T. Morishita for discussion and help in analysis. We are grateful to K. Gillis, C. Mével, J. Allan, and other shipboard scientists of Leg 147 of ODP for collaboration on board. We thank K. T. M. Johnson and C. Herzberg for their constructive comments.