Diffuse porous melt flow and melt-rock reaction in the mantle lithosphere at a slow-spreading ridge: A structural petrology and LA-ICP-MS study of the Othris Peridotite Massif (Greece)

Authors


Abstract

[1] Harzburgites and plagioclase-peridotites from the Othris Peridotite Massif in Central Greece preserve microstructural and petrological evidence for interaction with a melt which became saturated in orthopyroxene while migrating by km-scale diffuse porous flow through the thermal boundary layer (TBL) and the base of the thermal lithosphere. The melt precipitated orthopyroxene, and eventually also plagioclase and clinopyroxene within the peridotites. Major and trace element geochemistry suggests that the melt was a depleted melt, i.e., a melt fraction from the melting column underneath a spreading centre produced by shallow melting of refractory peridotites. We see no evidence for the presence of boninitic melts. We argue that the melts in Othris migrated by diffuse porous flow as they crystallised orthopyroxenes and were therefore inherently unable to create their own high-permeability melt channels. We propose that depleted melt fractions can remain isolated from deeper melt fractions, possibly already aggregated into a MORB-like magma, because they migrate by different mechanisms through the TBL and the lithosphere.

1. Introduction

1.1. Channeled Versus Diffuse Porous Flow

[2] It is widely accepted that Mid-Ocean Ridge Basalts (MORB) are aggregated melts consisting of melt fractions produced at different depths in the mantle underneath ocean ridges [e.g., Johnson et al., 1990]. The way in which melts are extracted from the mantle is still debated. Primitive MORB is not in equilibrium with harzburgite at low pressures, suggesting that melts produced at depth can travel to the surface without reacting and equilibrating with the shallow mantle [Stolper, 1980; Elthon and Scarfe, 1984; Johnson et al., 1990; Kelemen et al., 1997]. This has led to the view that mantle melts migrate predominantly by porous flow in dunitic conduits (Figures 1a and 1b). Support for this view has been found in the mantle section of the Oman Ophiolite, where it is shown that dunites in the mantle section are in equilibrium with MORB, whereas the enclosing harzburgites are not [Kelemen et al., 1995b]. Cumulate minerals which crystallised just below the crust-mantle boundary, in the Moho Transition Zone, in Oman have MORB-like trace element compositions, suggesting that MORB melts are already fully aggregated before they reach crustal levels underneath spreading centres [Benoit et al., 1996; Koga et al., 2001]. Moreover, preservation of excess 230Th in young MORB at modern ridges, produced at deep levels by melting in the presence of garnet, shows that deep melt fractions are rapidly transported to the surface [Spiegelman and Elliott, 1993]. This requires that melt flow is confined to spatially restricted high-permeability channels [Kelemen et al., 1997].

Figure 1.

Models of mantle melt migration and melt aggregation underneath mid-ocean ridges. (a) Melt migration by channeled porous flow through dunite conduits. Melt fractions produced by decompression partial melting are collected in the main dunite channel network and are effectively aggregated into MORB-like magma before they reach the upper crustal melt lens [after Kelemen et al., 1995a, 1995b]. (b) Melt migration by channeled porous flow through dunite conduits. In this case, however, the dunite channels are separate and unconnected and melts are only aggregated in the upper melt lens. (c) Melt migration by diffuse porous flow and aggregation of melt fractions in upper crustal melt lens. In this scenario, shallow peridotites and percolating melts should attain trace element equilibrium. (d) Dual mode model. Shallowly produced “depleted” melt fractions travel by diffuse porous flow and remain isolated from deep melt fractions migrating through dunite channels. In this scenario, shallow peridotites are in equilibrium with low-pressure, depleted melt fractions only. This model probably applies to mid-ocean ridges with a thick thermal boundary layer and is favored for Othris.

[3] On the other hand, Niu [1997] argued that abyssal peridotites contain evidence for olivine precipitation during reactive, diffuse porous flow of melt. Furthermore, depleted to ultra-depleted compositions of melt inclusions in olivine crystals in MORB from the Mid-Atlantic Ridge [Sobolev and Shimizu, 1993] and of cumulate minerals in oceanic gabbros from the Atlantic [Ross and Elthon, 1993; Coogan et al., 2000], show that melt fractions produced by low-pressure partial melting (in the spinel stability field) of already depleted peridotites can travel upward isolated from aggregated melts produced at higher pressures. This suggests that, if the bulk of Mid-Ocean Ridge magmas travel through dunite channels, at least some depleted melt fractions produced at shallow levels remain outside of these channels. Such melt fractions may travel through the shallow mantle by diffuse porous flow (Figure 1c). Indeed, some olivine-hosted melt inclusions have trace element patterns which suggest that the melts reacted with depleted peridotites during diffuse porous flow through the shallow mantle [Kelemen et al., 1997; Shimizu, 1998].

[4] Good evidence for non-channellized porous melt flow through the mantle is provided by plagioclase-peridotites from ocean ridges [Dick, 1989; Cannat et al., 1992; Elthon, 1992; Tartarotti et al., 1995; Seyler and Bonatti, 1997; Kelemen et al., 1997; Constantin, 1999; Tartarotti et al., 2002] and from ophiolites [Rampone et al., 1997; Batanova et al., 1998]. Such plagioclase-peridotites have resulted from impregnation of depleted peridotites with melt, which fractionally crystallised plagioclase, pyroxenes, and olivine. This paper presents the results of an integrated structural and petrologic study of the ophiolitic Othris Peridotite Massif in Central Greece. The Othris massif contains a significant volume of plagioclase-peridotites, in addition to plagioclase-free harzburgites [Menzies, 1973; Dijkstra et al., 2001; Barth et al., 2003]. These plagioclase-peridotites are similar to abyssal peridotites, and have probably resulted from refertilization of previously depleted peridotites by an impregnating melt [Dijkstra et al., 2001; Barth et al., 2003]. We conclude that diffuse porous flow of a melt with a composition similar to that of depleted MORB through relatively cold, lithospheric harzburgites (Figure 1d) led to extensive melt-rock reaction and fractional crystallization of plagioclase, orthopyroxene, and clinopyroxene. Importantly, the evidence for diffuse porous melt flow is not limited to the plagioclase-peridotites. This study provides constraints on the processes that take place underneath ocean ridges (in particular toward segment ends) in the base of the thermal lithosphere, at levels which are not normally sampled by dredging or drilling at modern ocean ridges. Moreover, the size and the coherency of the massif allows us to make inferences about the scale and the evolution of these processes, which could not have been made by studies of samples dredged from the ocean floor.

1.2. Thermal, Rheological, and Microstructural Lithosphere at Mid-Ocean Ridges

[5] Throughout this paper we use the term “thermal lithosphere” for the layer in which the effects of conductive cooling dominate over advection of heat from below (Figure 2a). The thermal lithosphere is not the same as the thermal boundary layer, which is the region in the mantle where the temperature gradient deviates from the adiabatic gradient at depth (Figure 2a). Partial melting of peridotites will be largely suppressed within the TBL, as the mantle cools below the peridotite solidus somewhere in the TBL. In fact, in some petrological studies of mantle melting at mid-ocean ridges the base of the TBL is equated with the level at which melting stops [Niu, 1997; Niu and Hékinian, 1997]. At segment centers of fast spreading ridges where geotherms are steep, the TBL and thermal lithosphere nearly coincide and are very thin, mainly confined to the crust. However, at segment ends of fast- and slow-spreading ridges, and possibly at segment centres of very slow-spreading ridges as well, the TBL and thermal lithosphere are relatively thick and their base will reach into the mantle (Figure 2a). At such ridges a cold mantle lid exists, and it is argued in this paper that this mantle lid has an important effect on melt migration.

Figure 2.

Thermal and rheological structure of suboceanic mantle such as can be found at segment-ends of slow-spreading ridges, where a relatively thick lithosphere and thermal boundary layer (TBL) may exist. Temperature profile is constrained by occurrence of micro-seismicity at temperatures <750°C downto a depth of 6–10 km at the Mid-Atlantic Ridge [Toomey et al., 1985]. (a) Following White [1988], the thermal lithosphere is defined as the layer in which the effect of conductive heat-loss to the surface dominates over advection of heat from below. The TBL is the layer in which the local geotherm deviates from the adiabatic geotherm (base of TBL taken as the level at which T = 0.95 · Tadiabatic). Partial melting will cease in the basal part of the TBL, where the temperature becomes lower than the (depleted) peridotite solidus. (b) Peridotite rheologies for given geotherm at flow stresses of 1,2, 10 and 25 MPa, calculated using the dry olivine dislocation creep flow law of Chopra and Patterson [1984]. Typical mantle convection strain rate shown by vertical dashed line (i.e., a velocity gradient of 5 cm/yr distributed over a layer of 300 km thick). Microstructures produced by olivine deformation at stresses above ∼10 MPa are generally classified as “lithospheric.” Comparison with Figure 2a suggests that they are limited to the thermal boundary layer or the base of the thermal lithosphere.

[6] We can also define a rheological lithosphere, that is the layer in the mantle where temperatures are low such that deformation by olivine creep at stresses associated with mantle convection (i.e., a few MPa) is effectively arrested. In this definition, the rheological lithosphere is only effectively rigid with respect to mantle convection in the asthenosphere. Under higher (“tectonic”) stress and strain rate conditions even part of the lower lithosphere can deform. Such conditions are relevant for tectonic stretching of the lithosphere to accommodate extension during a-magmatic stages of seafloor spreading at magma-starved ridges, for mantle lithosphere in the vicinity of a transform fault which is undergoing strike-slip or transtensional deformation, or for a fragment of oceanic lithosphere which is under compression or transpression at the start of intraoceanic thrusting.

[7] The terms lithosphere and asthenosphere are also used in the context of olivine microstructures in deformed mantle peridotites [Nicolas, 1986; Ceuleneer et al., 1988; Ildefonse et al., 1998a, 1998b]. Briefly, two relevant classes of peridotite microstructures can be recognized in abyssal peridotites and ophiolitic mantle rocks. First, coarse-grained equant to porphyroclastic microstructures showing evidence for fast recovery, are formed by low stress (<5–10 MPa), high temperature deformation, most likely during mantle upwelling or corner flow underneath spreading centers. Such microstructures are often referred to as “asthenospheric” microstructures [Nicolas, 1986; Ceuleneer et al., 1988; Ildefonse et al., 1998a, 1998b]. Second, “lithospheric”, fine-grained porphyroclastic microstructures showing evidence for less effective recovery, are formed by higher stress (>10 MPa) and lower temperature deformation [Nicolas, 1986; Ceuleneer et al., 1988; Ildefonse et al., 1998a, 1998b]. Lithospheric microstructures probably represent high (“tectonic”) stress and strain rate deformation of mantle lithosphere. It should be noted that although stresses can be adequately estimated from deformation microstructures through empirically calibrated relationships between stress and recrystallized grain size [Van der Wal et al., 1993], deformation temperatures are much less well-constrained. The evidence for fast and effective recovery in “asthenospheric” microstructures, sometimes associated with evidence for melting, suggest deformation temperatures >1250°C for such microstructures [Nicolas, 1986]. According to experimentally derived olivine flow laws, however, deformation at low strain rates (approaching those of mantle convection) and stresses of 1–10 MPa can occur at temperatures <1200°C, even as low as 1000°C (Figure 2b), suggesting that “asthenospheric” microstructures can be formed in the TBL or even in the thermal lithosphere. It is clear though that relatively high-stress “lithospheric”: olivine microstructures must represent deformation in the TBL or thermal mantle lithosphere.

1.3. Othris Peridotite Massif

[8] The Othris Peridotite Massif is considered to be the mantle section of the strongly dismembered Othris Ophiolite, a fragment of oceanic lithosphere from the Mesozoic Neotethys ocean [Hynes et al., 1972; Menzies, 1973; Menzies and Allen, 1974; Smith, 1993; Rassios and Smith, 2000]. The Othris Peridotite Massif forms the uppermost and most western tectonic unit of an internally thrusted passive margin sequence in the Othris Mountains [Smith et al., 1975; Ferrière, 1985]. More ophiolitic rocks, including peridotites, occur further east in the Othris Mountains [Smith et al., 1975; Rassios and Konstantopoulou, 1993; Rassios and Smith, 2000]. The Othris Peridotite Massif consists of three separate peridotite bodies, Katáchloron, Fournos Kaïtsa, and Mega Isoma (Figure 3, inset). Because of the relatively fertile compositions of peridotites in the Othris Massif, Menzies and Allen [1974] concluded that the Othris Ophiolite originated in a rift-like setting. On the basis of a geochemical study of pyroxenes in peridotites from the Vourinos and Pindos ophiolites of central Greece and the Bulqiza ophiolite in Albania, and of orthopyroxene in one sample from the Eretria chromite mine in eastern Othris, Bizimis et al. [2000] concluded that all these ophiolites originated in a supra-subduction zone environment. However, based on a structural and petrological study, we argued that the Othris Peridotite Massif resembles sub-oceanic mantle from a near-transform fault region of a slow-spreading, Atlantic-type mid-ocean ridge [Dijkstra et al., 2001]. The results of our geochemical study in the present paper support this view.

Figure 3.

Simplified maps and cross-sections through the Katáchloron and Fournos Kaïtsa bodies in the Othris Peridotite Massif. Insets show the position of the massif in Greece and the position of the two studies bodies. Outline of dunite bodies partly after A. Rassios (personal communication, 1995). Foliation trajectories in the peridotite tectonites are shown by thick lines. Stippled rocks are plagioclase-peridotites which contain lenses of plagioclase ± orthopyroxene ± clinopyroxene and various generations of gabbroic dykes. These lenses and dykes are the (cumulate) products of fractional crystallization of an impregnating, depleted melt (see text for discussion). Location of sample GOF1 taken for trace element analysis is indicated (southern tip of Fournos Kaïtsa).

2. Structural Petrology of Othris Peridotites

2.1. Large-Scale Structure

[9] We have mapped and studied two peridotite bodies in detail (Figure 3). The structure of the first (Katáchloron) comprises a 1 km-wide, N-S-striking, steeply east dipping peridotite mylonite shear zone, juxtaposing two blocks of peridotite tectonites with steep to moderately steep foliations. The western block consists of clinopyroxene-poor harzburgites with large cross-cutting, 100 m-scale dunite bodies (Figures 3 and 4a). The eastern block of the Katáchloron body consists of clinopyroxene-poor harzburgites with few dunitic bands and irregular dunitic patches (Figures 5a and 5b). The fine-grained harzburgites grade up- and eastward into clinopyroxene-rich harzburgites and plagioclase-harzburgites and -lherzolites. The plagioclase-in boundary makes a large angle with the tectonite foliations. The plagioclase-peridotites contain generally concordant mm-scale lenses of plagioclase ± orthopyroxene ± clinopyroxene, occasionally with small amounts of amphibole. They further contain plagioclase-rich bands (Figure 6a), generally concordant olivine-gabbro dykes (Figure 6b) and discordant olivine-gabbro and gabbro-norite (generally pegmatitic, Figures 6c and 6d) dykes. The second peridotite body (Fournos Kaïtsa) is in structure and composition very similar to the eastern block of the Katáchloron body, comprising harzburgites grading up- and eastward into plagioclase-peridotites with abundant gabbroic dykes.

Figure 4.

Selected macroscopic and microscopic features of coarse-grained harzburgite tectonites from the Katáchloron body, west of large shear zone (all traced from photographs). (a) Close-up of margin of large cross-cutting dunite bodies (see map in Figure 3). (b)–(c) Remains of heavily corroded orthopyroxene clasts. Dashed lines are subgrain walls in olivine crystals.

Figure 5.

Selected macroscopic and microscopic features of fine-grained harzburgite tectonites from the Katáchloron body (east of shear zone) and the western part of the Fournos Kaïtsa body (all traced from photographs). (a) Isoclinally folded dunite layer with sharp boundaries in harzburgites (Katáchloron). (b) Irregular small replacive dunite patch (Katáchloron). (c) Flaser-like orthopyroxene grains in very fine-grained rims around orthopyroxene porphyroclasts which have the same crystallographic orientation as the adjacent clast. These are interpreted as the remains of incompletely dissolved orthopyroxene clasts [Dijkstra et al., 2002]. (d)–(g) Interstitial orthopyroxene, sometimes along low-angle (sub-) grain boundaries in recrystallizing olivine clasts (f–g) in olivine-rich domains. Olivine next to such interstitial orthopyroxene grains are occasionally seen to develop facet-like straight crystal faces.

Figure 6.

Selected macroscopic and microscopic features of plagioclase-bearing tectonites from the Katáchloron body (east of shear zone) and the eastern part of the Fournos Kaïtsa body (all traced from photographs). (a)–(d) Plagioclase-rich bands and concordant and discordant gabbroic dykes (Figure 6a, Katáchloron; Figures 6a–6d, Fournos Kaïtsa). (e)–(g) Irregular orthopyroxene porphyroclasts, interpreted as corroded clasts. (h) Irregular clinopyroxene crystals, probably also corroded porphyroclasts. (i) Plagioclase aggregate which crystallised in a fine-grained olivine-orthopyroxene domain. (j) Plagioclase aggregate, associated with orthopyroxene, which crystallized in a coarser olivine domain. (k)–(m) Close-ups of plagioclase ± orthopyroxene ± clinopyroxene lenses in sample used for chemical analysis. Plagioclase is frequently altered in this sample.

2.2. Microstructures

[10] We have recognised three classes of peridotite microstructures [Dijkstra et al., 2001, 2002].

[11] 1. Coarse tectonites which locally preserve low-stress, “asthenospheric” microstructures with recrystallized olivine grain sizes >0.5 mm; application of the paleopiezometer of Van der Wal et al. [1993] yields deformation stresses <13 MPa. Most of the coarse tectonites are, however, partly recrystallized to a smaller grain size, indicating an imprint of higher stress, ‘lithospheric’ deformation. Coarse tectonites are mainly confined to the western block in the Katáchloron area. They are also found locally within the plagioclase-bearing peridotites in the Fournos Kaïtsa area.

[12] 2. Fine tectonites with relatively high-stress “lithospheric” microstructures are the dominant peridotite-type in Othris. The recrystallised olivine grain size in the fine-grained tectonites is 0.1–0.6 mm, suggesting deformation stresses of 10–40 MPa using the above paleopiezometer. Fine tectonites are found in the eastern block in the Katáchloron area, and in most of the Fournos Kaïtsa area. In our interpretation the fine tectonites represent transtensional deformation at a ridge-transform intersection [Dijkstra et al., 2001; Dijkstra, 2001].

[13] 3. Mylonites and adjacent proto-mylonitic tectonites cross-cut tectonites in both the Katáchloron and Fournos Kaïtsa areas. They may represent structures related to emplacement of the Othris ophiolite [Dijkstra, 2001].

2.3. Coarse Tectonites and Evidence for Reaction With an opx-Undersaturated Melt

[14] Harzburgites of the western block in the Katáchloron area with coarse-grained porphyroclastic microstructures often contain orthopyroxene (and occasionally clinopyroxene) porphyroclasts with strongly irregular outlines caused by numerous embayments of mainly olivine (Figure 4b), together with clusters of small, interstitial orthopyroxene crystals which all have the same crystallographic orientation (Figure 4c). Similar irregular orthopyroxene clasts also occur in the domains of coarse-grained porphyroclastic peridotites in the Fournos Kaïtsa area (Figures 6e–6g). Such orthopyroxene crystals are interpreted as “corroded” clasts, produced by orthopyroxene dissolution coupled with olivine crystallization, or by an incongruent melting reaction:

display math

The presence of chrome-spinel clusters with remnants of orthopyroxene crystals in their cores within the large dunites in the Katáchloron area, as well as the presence of foliated harzburgite enclaves, suggests a replacive origin by the same reaction rather than a cumulate origin for these dunites.

2.4. Fine Tectonites and Evidence for Reaction With Melt Close to opx-Saturation

[15] The majority of the peridotites of the eastern block in the Katáchloron area and in the Fournos Kaïtsa area contain orthopyroxene clasts with different textures. Within the fine-grained porphyroclastic tectonites, small (10–100 μm), flaser-like orthopyroxene crystals occur mixed together with olivine in very fine-grained rims around orthopyroxene porphyroclasts which contain some olivine embayments (Figure 5c). These rims are best explained by incomplete replacement of the orthopyroxene clasts by olivine due to a reaction with a melt [Dijkstra et al., 2002]. In addition small irregular, interstitial orthopyroxene crystals are found between olivine crystals in the olivine-rich domains in the tectonites (Figures 5d–5g). Occasionally, olivine crystals next to such interstitial orthopyroxene crystals are seen to develop straight crystal faces (Figure 5d), whose orientations are consistent with the orientation of olivine faces that develop by olivine growth next to melt pockets [Dijkstra et al., 2002]. This suggests that orthopyroxene locally precipitated from a melt within the olivine domains, as a result of either fractional crystallization or replacement of olivine by orthopyroxene. The fine-grained tectonites thus show a net reaction of the form:

display math

i.e., a reaction which was locally melt-producing and locally melt-consuming, suggesting that the melt was at or just below orthopyroxene saturation. Importantly, small orthopyroxene crystals in the olivine matrix are found along low-angle grain boundaries in olivine aggregates which are completely surrounded by higher-angle grain boundaries (Figures 5f and 5g). The formation of such orthopyroxene crystals can only be explained by precipitation of orthopyroxene or transformation of olivine into orthopyroxene along low-angle or subgrain boundaries during dynamic recrystallization of olivine in the fine-grained porphyroclastic tectonites. This allows us to link the deformation conditions directly to the conditions during this stage of melt-rock reaction.

2.5. Plagioclase-Peridotites: Fractional Crystallization of plag + opx + cpx

[16] Plagioclase ± orthopyroxene ± clinopyroxene lenses (Figures 6i–6l) in the clinopyroxene-rich harzburgites and plagioclase-peridotites clearly crystallized from a melt [Menzies, 1973; Dijkstra et al., 2001; Barth et al., 2003]. Immediately adjacent to the plagioclase-peridotites there is a zone in which there is evidence for crystallization of clinopyroxene in lenses and veinlets, without plagioclase. In the plagioclase-peridotites, we also note many irregular clinopyroxene crystals, which may be relic clinopyroxene clasts corroded by melt (Figure 6h). The plagioclase-bearing lenses are spatially associated and locally contiguous with concordant and discordant gabbro dykes. The plagioclase lenses are often concordant to the foliation in the fine-grained porphyroclastic tectonites, but they are generally only weakly deformed, suggesting that they crystallised during the last stages of deformation. The lenses represent the stage when, as a result of cooling, the melts became saturated in plagioclase, orthopyroxene, and clinopyroxene.

2.6. Mineral Chemistry of Mineral Phases Associated With Melt

[17] We obtained detailed major element, rare earth element (REE) and other trace element analyses from clinopyroxene, plagioclase, and some orthopyroxene from an impregnated plagioclase-peridotite sample (GOF1) using and electron microprobe and a laser-ablation induced-coupled plasma-mass spectrometer (LA-ICP-MS). This new data set (Table 1) complements mineral trace element data from plagioclase-bearing and plagioclase-free peridotites from the Othris Massif reported in Barth et al. [2003].

Table 1. Trace Element Concentrations (ppm) in Sample GOF1 Measured by Laser-Ablation ICP-MSa
TypebClinopyroxene analyses
G19-2G19-3G19-4G19-5G19-6G19-7G19-8Z23-1Z23-2Z23-3Y21-1Y21-2G15-1G15-2D1-1D1-2R26-1R26-2S30-1S30-2
clast-rim clast-core clast-core clast-coreclast-rimclast-rimclast-rimclast-rimclast-rimclast-coreinterstitinterstitinterstitinterstitpocketpocketinterstitinterstitpocketpocket
  • a

    Also shown are analytical errors (1σ) for the first three analysis of each mineral, and the average composition of the NIST 612 standard glass [Pearce et al., 1997] as measured during analysis.

  • b

    Type of crystal: “clast” means rounded porphyroclast, possibly residual in character; “interstit” means interstitual crystal with irregular outlines; “pocket” means crystal which is part of a plag ± cpx ± opx cumulate aggregate or in contact with plag.

Rb0.70.10.390.07<0.100.070.520.160.180.32<0.120.150.250.210.540.30.3<0.200.36<0.20.5<0.2<0.2
Sr0.430.050.690.060.480.050.540.490.550.390.490.450.50.580.40.10.20.850.650.720.390.860.4
Y18.11.219.41.317.31.213.816.215.316.419.721.222.418.719.114.617.921.319.722.419.614.518.7
Zr8.80.68.50.68.90.67.59.27.97.99.210.31010.19.77.19.010.39.79.99.85.89.0
Nb<0.030.010.030.01<0.0080.0050.0090.04<0.0050.013<0.0080.0180.014<0.0110.042<0.015<0.0240.0180.04<0.0<0.040.08<0.015
Ba0.070.030.20.050.060.040.550.480.610.630.030.018<0.060.13<0.030.40.50.160.5<0.070.350.150.3
La<0.0030.00090.0280.0070.0320.0080.0150.0080.0150.0140.0230.0590.030.0270.0350.060.020.0210.0170.0380.0170.018<0.006
Ce0.420.030.310.020.260.020.360.360.40.280.440.390.410.40.370.540.280.370.340.320.310.190.41
Pr0.20.020.170.020.180.020.160.160.130.150.190.250.220.170.20.10.330.150.190.170.220.090.19
Nd1.850.21.750.11.750.21.651.81.51.452.152.32.31.752.01.351.62.152.01.62.051.31.65
Sm1.350.21.250.21.50.21.21.40.951.31.61.651.71.351.341.051.61.71.31.191.350.541.3
Eu0.440.040.460.040.380.040.340.540.360.370.590.520.570.490.450.250.510.620.590.630.630.370.41
Gd2.10.21.90.21.90.21.91.61.82.22.52.42.62.42.42.12.62.62.12.41.922.1
Tb0.470.040.470.040.420.040.370.40.320.410.50.540.550.440.460.320.360.50.460.480.460.310.39
Dy3.150.32.90.33.00.32.52.92.83.03.74.04.053.353.553.352.853.553.653.953.22.353.25
Ho0.660.050.640.050.560.050.650.550.550.590.850.790.810.750.770.70.690.810.780.970.70.480.69
Er1.80.11.70.12.00.21.41.71.81.82.22.22.32.32.321.92.22.42.42.21.61.9
Tm0.240.020.340.030.280.030.230.290.160.230.30.310.320.270.250.270.230.290.350.290.230.310.26
Yb1.450.21.750.21.450.21.61.651.51.61.952.02.051.951.92.352.252.051.81.951.951.21.7
Lu0.210.020.240.020.280.030.160.260.180.120.250.260.270.290.250.180.210.230.290.310.230.210.31
Hf0.640.080.620.080.360.060.440.670.560.520.540.650.630.630.70.590.590.750.640.760.880.410.62
Ta<0.0030.0008<0.0010.001<0.0020.00060.0020.004<0.0020.0160.0050.0070.0110.0060.0120.0040.0040.0060.0090.0170.012<0.0090.005
Pb0.140.020.090.020.130.020.090.080.160.080.0560.0280.0410.0550.050.310.090.0260.10.130.060.030.06
Th0.0370.0060.0060.0020.0240.0050.0120.0020.0020.0090.0140.0160.025<0.0030.0110.024<0.0010.0160.0030.0060.003<0.0040.005
U0.0020.0010.410.030.0060.0020.0060.1170.0050.0070.0090.00570.008<0.002<0.0010.002<0.0010.0030.0050.028<0.0020.008<0.003
 
TypebPlagioclase analyses Orthopyroxene analyses STD Glass
E9-11 σE9-21σD2-11σD2-2E9-3E9-4E9-5 H22-11σH22-21σH22-31σH22-4Q24-1Q24-2 NIST 6121σ
pocket pocket pocket pocketpocketpocketpocket pocket pocket pocket pocketpocketpocket  n = 20
Rb0.40.10.50.10.40.1<0.230.16<0.14<0.05 0.20.05<0.30.1<0.060.02<0.040.060.16 31.70.5
Sr7.30.37.10.37.60.36.97.277.2 0.490.031.710.090.490.030.340.550.93 76.11.1
Y<0.040.020.120.020.20.030.130.10.0730.14 1.70.11.90.21.90.21.71.41.1 38.11.1
Zr<0.040.02<0.070.030.150.03<0.060.0390.02<0.009 0.660.060.520.060.710.060.780.690.9 35.90.9
Nb0.0380.009<0.030.01<0.0180.008<0.027<0.019<0.028<0.005 <0.0060.0030.0170.005<0.0080.0030.0050.0070.057 38.10.8
Ba1.20.11.50.21.10.11.21.31.681.42 0.030.01<0.140.05<0.0140.007<0.0200.020.69 37.91.5
La0.0460.0070.0140.0050.0220.0050.0150.0390.0320.026 <0.0060.0020.020.005<0.0020.0008<0.002<0.0020.016 35.70.9
Ce0.190.020.20.020.140.010.240.1840.1370.134 <0.0020.0010.010.002<0.0020.0009<0.0020.0030.01 38.50.9
Pr0.0410.0060.0390.0060.0690.0080.0580.0220.0320.027 <0.0020.001<0.0070.002<0.0010.0004<0.001<0.001<0.002 37.20.8
Nd0.30.0040.320.0040.1850.0030.140.160.120.185 <0.020.0080.0450.0150.0130.004<0.0100.0060.054 35.31.0
Sm0.140.0040.0350.0020.0350.0020.010.0450.02050.06 0.00850.009<0.050.0150.0220.0060.0110.0050.02 36.71.0
Eu0.240.020.280.030.380.020.20.290.270.27 <0.0090.004<0.0060.0030.0080.002<0.0030.012<0.004 34.51.0
Gd0.060.020.040.020.030.020.11<0.030.0550.014 0.060.01<0.0170.0080.0490.0080.0550.0280.03 37.01.8
Tb0.0150.0040.0060.0020.0140.0010.0030.0040.0060.007 0.0130.0020.020.0040.0210.0020.0170.0120.022 35.91.0
Dy<0.0030.0020.020.0020.030.0010.0750.0540.02850.0325 0.1950.030.20.030.210.020.20.1950.14 35.91.2
Ho0.010.003<0.0040.0010.0320.0050.0050.006<0.0060.005 0.0570.0060.0660.0080.0660.0060.0610.0630.053 37.81.3
Er<0.0110.005<0.0140.005<0.0160.0090.013<0.010<0.0090.008 0.170.020.230.030.230.020.20.250.19 37.41.1
Tm0.00040.0005<0.0100.005<0.0030.002<0.005<0.004<0.0020.0026 0.0420.0050.0480.0070.0510.0050.0470.0360.04 37.51.2
Yb0.030.010.040.02<0.040.02<0.022<0.027<0.028<0.029 <0.0220.005<0.0230.005<0.0240.005<0.025<0.026<0.027 39.91.2
Lu<0.0030.001<0.0030.002<0.0040.0020.003<0.003<0.003<0.001 0.080.0080.090.010.0870.0080.0780.0610.069 37.61.3
Hf<0.020.007<0.0020.0080.00550.0080.0150.00450.0050.003 0.0590.010.0850.0150.0510.0080.0510.0270.052 34.71.2
Ta0.0050.002<0.0040.002<0.0090.003<0.005<0.004<0.006<0.001 0.00080.0004<0.0050.0020.00140.00050.00240.0011<0.002 39.81.0
Pb0.120.020.120.020.070.020.10.0840.0870.062 0.0620.0080.040.010.0250.0040.020.0650.14 39.01.1
Th<0.0020.001<0.0050.002<0.0040.0020.004<0.0030.003<0.0003 <0.0010.00050.0040.001<0.0010.00030.0005<0.001<0.002 37.21.0
U0.0040.001<0.0040.001<0.0020.00070.013<0.002<0.002<0.001 <0.0010.0005<0.0030.0009<0.0040.0002<0.004<0.0010.003 37.41.0

[18] Trace element concentrations were determined in situ for individual mineral phases using LA-ICP-MS at Utrecht University following techniques described in de Hoog et al. [2001] and Mason et al. [1999]. The sample was a polished thin section of ∼80 μm thickness. The system incorporates a 193 nm Ar-F Excimer laser ablation system (GeoLas, MicroLas and Lambda Physik) and a quadrupole ICP-MS with collision and reaction cell (Platform ICP, Micromass) in pulse counting mode. Ablation was performed at a fixed point on the sample surface at a fluence of 20 Jcm−2, a pulse repetition rate of 10 Hz and a typical crater diameter of 60–120 microns. Each analysis represents an ablated volume of <1 microgram. Calcium, previously determined by JEOL Superprobe 8600 electron microprobe analysis at the Department of Geology at the University of Leicester (wavelength dispersive system, 25 nA current, 15 kV accelerating voltage, ∼10 μm focused beam), was used as an internal standard element. Calibration was performed against NIST 612 glass using the normalizing coefficients of Pearce et al. [1997]. Accuracy for rare earth elements was typically better than 10% for homogeneous in-house reference clinopyroxene standards [Mason et al., 1999] The Jeol microprobe was also used to determine major element and some trace element concentrations in minerals in the sample used for laser ablation.

[19] Plagioclase in the studied sample is concentrated in lenses, and is associated with orthopyroxene and/or clinopyroxene (Figures 6i–6m). In places, it is quite strongly altered to a brownish, fine-grained, nearly isotropic alteration product, which is very common in Othris. However, many fresh, unaltered plagioclase crystals remain. In addition, we analyzed clinopyroxene crystals spatially associated with plagioclase, and some highly irregular clinopyroxene crystals which are probably corroded by melt (Figure 6h) and which are expected to be in trace element equilibrium with the melts that were present in these rocks. We were not able to distinguish between relic and cumulate clinopyroxene within the studied sample on the basis of major and trace element geochemistry; compositions of large, mm-sized clinopyroxene clasts overlap with those of clinopyroxene in plagioclase-bearing lenses. We have included some orthopyroxene crystals spatially associated with plagioclase and clinopyroxene in the analyses. As trace element analyses are limited to phases in the plagioclase-peridotites, the results give insights into the last stage of melt-rock interaction in Othris.

[20] Microprobe analyses show that both the clinopyroxene crystals in the melt-derived lenses and the strongly corroded clinopyroxene crystals are characterized by higher concentrations of Ti and Na than residual clinopyroxene porphyroclasts in the plagioclase-free harzburgites (Figure 7). Mg-numbers {i.e., 100 · Mg/(Mg + Fe)} of clinopyroxene in plagioclase-bearing lenses are 92.4 ± 0.4 (1σ); those of the irregular single crystals analyzed agree within error, 92.8 ± 0.7. Plagioclase crystals have anorthite contents of 85.7 ± 1.6 {i.e., 100 · Ca/(Ca + Fe + Na + K)}.

Figure 7.

Clinopyroxene (both cumulate and corroded) crystals in sample GOF1 are all enriched in Na and Ti with respect to residual clinopyroxene clasts in plagioclase-free peridotites in Othris. Fields for residual, “un-reacted” clinopyroxene and for cumulate and reacted clinopyroxene are based on previously collected data set [Dijkstra et al., 2001].

[21] The set of trace element analyses of clinopyroxene in sample GOF1 using LA-ICP-MS is relatively homogeneous (Figure 8a, Table 1). Clinopyroxene crystals are depleted in light REE (LREE) with respect to the middle and heavy REE (MREE, HREE). La has <0.2 times chondritic values, whereas the MREE and HREE have abundances around 5–10 times those of chondrites. Clinopyroxene crystals further show a strong negative Sr anomaly of <0.1 times the Sr concentrations in chondrite. In addition, there is a significant negative Zr anomaly. Some crystals have a weak negative Eu anomaly. We have neither found significant core-rim variations in trace element concentrations, nor variations in trace element concentrations between clinopyroxene crystals that look “residual” (porphyroclasts) or “magmatic” texturally. REE concentrations in plagioclase crystals are generally low, between 0.5 and 0.01 times the chondritic values, with a strong positive Eu anomaly and a weak positive Sr anomaly (Figure 8b). Orthopyroxene crystals have very low LREE and MREE abundances, with HREE concentrations close to those in chondrites (Figure 8c). Comparison with analytical data from other peridotites from Othris [Barth et al., 2003] shows that clinopyroxenes in the studied sample are relatively enriched in REE compared those in other plagioclase-peridotites (Figure 8a, inset). Moreover, clinopyroxenes in GOF1 do not show “spoon-shaped” patterns with relatively high La and Ce concentrations as seen in some of the other plagioclase-peridotites from Othris [Barth et al., 2003].

Figure 8.

Measured Rare Earth Element (REE), Zr and Hf concentrations in cumulate and corroded clinopyroxene, cumulate plagioclase, and (cumulate?) orthopyroxene in contact with plagioclase, in plagioclase-bearing sample GOF1, determined using laser-ablation ICP-MS. Concentrations are normalized to those in chondrite. Inset in Figure 8a shows range of clinopyroxene compositions in plagioclase-peridotites and plagioclase-free harzburgites from Othris taken from Barth et al. [2003]. Clinopyroxenes in sample GOF1 have the highest REE concentrations (except La) of the available data set.

3. Discussion

3.1. A History of Melt-Rock Reaction and Fractional Crystallization

[22] The peridotites in Othris are predominantly clinopyroxene-poor harzburgites. They are, therefore, depleted peridotites which must have undergone partial melting at some time during their history, probably during mantle upwelling underneath the Othris paleo-ridge. However, no microstructural or petrographic evidence for this partial melting history remains in the peridotites, other than the overall depleted compositions. Only using spinel compositions and trace element modeling clinopyroxene compositions could Barth et al. [2003] shed some light on the early stage of melt extraction in Othris. The plagioclase-free harzburgites in Othris have probably undergone ∼15% dry partial melting; equally refractory spinel compositions preserved in plagioclase-bearing peridotites suggest similar amounts of melt extraction before melt impregnation [Barth et al., 2003].

[23] The melt-related petrographic features observed, linked with the olivine microstructures of the host peridotites, allow us to reconstruct a time sequence of magmatic and deformation events in the Othris peridotites following partial melting:

[24] 1. In peridotites which have relics of “asthenospheric” microstructures with a weak to moderate imprint of “lithospheric” deformation, petrographic evidence is preserved showing corrosion of orthopyroxene and clinopyroxene clasts by a pyroxene-undersaturated melt. Because of the ambiguity in the interpretation of “asthenospheric” microstructures discussed above, we are unable to determine whether these rocks have preserved structures produced in the thermal asthenosphere, TBL, or lithosphere.

[25] 2. Peridotites which have a strong “lithospheric” imprint contain evidence for extensive reaction with a melt which was at or close to orthopyroxene-saturation, leading to incomplete corrosion of orthopyroxene clasts and, at the same time, crystallization of orthopyroxene between olivine crystals, and within recrystallizing aggregates of olivine. Clinopyroxene crystals in these rocks still show signs of corrosion, while there is no evidence for clinopyroxene crystallization. This suggests that the melts present were still undersaturated in clinopyroxene. These reactions were coeval with deformation and dynamic recrystallization at stresses of 10–40 MPa. The deformation microstructures and high estimated stresses show that, by this time, the peridotites must have become part of the TBL or of basal part of the thermal lithosphere (Figure 2b).

[26] 3. During the last stages of lithospheric deformation recorded in the fine-grained tectonites, melts started to crystallize plagioclase, orthopyroxene, and clinopyroxene in lenses, veins, and dykes in the uppermost exposed levels of the mantle section in Othris (now the plagioclase-peridotites). The plagioclase-peridotites are surrounded by an area in which there is evidence for crystallization of only clinopyroxene. Crystallization outlasted the tectonite deformation; the only deformation post-dating the crystallization of plagioclase-bearing lenses, veins, and dykes was localized into peridotite mylonite shear zones. The high anorthite content of the plagioclase and the high Mg# of the clinopyroxene show that these melt products are not simply frozen-in pockets of melt, but rather cumulates derived from fractional crystallization of melt. The crystallization of orthopyroxene together with plagioclase and clinopyroxene is again consistent with the melt being at orthopyroxene-saturation. By this time, the Othris peridotites must have been fully part of the thermal and rheological lithosphere. The temperature of the peridotites had probably decreased below 1100°C (as suggested by the crystallization of gabbro-norites, companre Ceuleneer et al. [1996]). The widespread occurrence of gabbroic dykes coeval with plagioclase-impregnation suggests that the plagioclase-bearing domains represent the transition between diffuse porous flow and melt transport through dykes produced by hydrofracturing.

[27] Plagioclase-peridotites are found in a relatively large area, but the layer in which plagioclase crystallized may be only a few hundreds of meters thick (Figure 3). However, evidence for orthopyroxene crystallization and corrosion is widespread in almost all harzburgites, suggesting that melt-rock reaction occurred at a scale of at least 1–2 km.

3.2. Melt Composition

[28] Our petrological observation suggest that the melts which interacted with peridotites in Othris evolved from orthopyroxene-undersaturation, via orthopyroxene-saturation, to saturation in orthopyroxene, clinopyroxene, and plagioclase. The fact that the melts in Othris became saturated in orthopyroxene before plagioclase suggests that the melt was more silica-rich than typical MORB. The high Mg# of the cumulate clinopyroxene and the high anorthite content of the plagioclase suggest that the last melts which traveled through the peridotites had a high Mg/Fe and Ca/Na ratio. Ti enrichment in clinopyroxene suggests aTi-rich melt. Such compositions are typical for depleted to ultra-depleted melts (UDM) occasionally found in melt inclusions [Sobolev and Shimizu, 1993]. High Mg# in clinopyroxene and high An% in plagioclase have also been found in cumulate minerals in oceanic gabbros [Ross and Elthon, 1993; Coogan et al., 2000], in gabbroic dykes in peridotites from the Oman Ophiolite [Kelemen et al., 1997; Benoit et al., 1999], and in lenses and grain interstices in mantle rocks from the Ligurian and Corsican ophiolites [Rampone et al., 1996, 1997]. In all these examples, plagioclase and clinopyroxene, sometimes together with orthopyroxene [Rampone et al., 1997], are thought to be the products of fractional crystallization of depleted melts.

[29] As noted above, the major element chemistry of the clinopyroxene and plagioclase in the relic melt lenses, suggest an origin by fractional crystallization and subsequent melt extraction rather than by complete crystallization of an interstitial melt. However, it is possible that the cumulate aggregates contained some trapped melt. There is no phase with an intercumulus texture or a composition that could represent a trapped melt fraction. However, the trapped melt could have crystallised as clinopyroxene ± plagioclase ± orthopyroxene after it became sealed off from the main body of melt, followed by redistribution of REE's over all the phases in the rock. To investigate this effect of a “cryptic trapped melt fraction,” we have carried out a simple “backstripping” procedure using a method very similar to that discussed in Bédard [1994]. We have assumed trace element concentrations and crystal/melt distribution coefficients to be negligible in olivine. In our calculations, we have used clinopyroxene/melt distribution coefficients derived using the model of Wood and Blundy [1997] and plagioclase/melt and orthopyroxene/melt distribution coefficients obtained from the measured partitioning of trace elements between plagioclase and clinopyroxene, and between orthopyroxene and clinopyroxene in the studied sample (Table 2). We have calculated the compositions of clinopyroxene and plagioclase after removing 0–40% (of the total rock volume) coexisting trapped melt (Figures 8a and 8b, see also figure captions for more information about method). With the given mineral proportions and melt modes, plagioclase and clinopyroxene disappeared from the rock at 40% trapped melt removal. The calculations show that neglecting a possible trapped melt fraction leads to overestimation of the REE contents of clinopyroxene and plagioclase before crystallization of the trapped melt. In Figure 8c we have shown the calculated compositions of the liquid in equilibrium with clinopyroxene. All the calculated melt compositions are depleted with respect to typical MORB melt, in particular in the LREE, consistent with a depleted or ultra-depleted melt composition. On the basis of our calculations we suggest that the studied plagioclase-peridotite has incorporated 5–10% trapped melt, as melt compositions calculated for this range of trapped melt fractions have no significant Eu anomaly and only a minor negative Sr anomaly. A 5–10% trapped melt fraction can also explain the relatively enriched character of cumulate phases in the studied sample with respect to those in other plagioclase-bearing peridotites from Othris (Figure 9a, see also Barth et al. [2003]). In particular, variations in HREE concentrations in cumulate phases in Othris plagioclase-peridotites are difficult to explain otherwise. The measured negative Zr anomaly in clinopyroxenes cannot be removed by backstripping a cryptic trapped melt fraction, and seems to be a characteristic of the melt that percolated through the Othris peridotites.

Figure 9.

Evalution of melt composition and the effect of a cryptic trapped melt fraction. (a) Calculated trace element compositions of clinopyroxene before crystallization of a coexisting trapped melt fraction of 0–25% (percentages as part of total rock volume). Modeled using a backstripping method similar to that of Bédard [1994], which is conceptionally equivalent to remelting the cumulate aggregates. We used distribution coefficients given in Table 2, rock mineral modes cpx:plag:opx:ol = 0.1:0.1:0.2:0.6 and melting modes cpx:plag:opx:ol = 0.25:0.25:0.25:0.25. We find that the model is very insensitive to the melting modes, as the trace element budget is dominated by clinopyroxene. Shaded field represents the range of compositions of clinopyroxenes in Othris plagioclase-peridotites, taken from Barth et al. [2003]; (b) Trace element compositions of plagioclase before crystallization of a coexisting trapped melt fraction of 0–25%, calculated using the same backstripping method. Some element concentrations (Zr, Hf, HREE) were not calculated because of unreliable plag/melt distribution coefficients (Table 1); (c) Calculated compositions of melts in equilibrium with clinopyroxene compositions in Figure 9a normalized to N-MORB [Sun and McDonough, 1989]. At F = 0.40 (40% trapped melt removal) plagioclase and clinopyroxene disappear from the ‘residue’, which corresponds to complete “remelting” of the cumulate aggregates. Calculated melt compositions differ strongly from those of boninites. Boninite compositions (no values for Pr, Ho and Tm) taken from Hickey and Frey [1982] and Taylor et al. [1994].

Table 2. Solid/Melt Distribution Coefficients
 LaCeSr 2+PrNdZrHfSmEuGdTbDyYHoErTmYbLu
  • a

    Calculated for T = 1150°C and P = 4 Kbar, for a melt composition in equilibrium with olivine with a Mg# of 90, and for the average clinopyroxene composition in the studied sample determined by electron microprobe, using the predictive model of Wood and Blundy [1997] for trivalent cations. Sr was calculated assuming E(2+) = 2/3 · E(3+) and D0(2+) = 0.14 · D0(3+) [see Blundy and Wood, 1994]. E is the apparent Young's Modulus of the pyroxene M2-site, where D0 is the distribution coefficient of an element with the same ionic radius as the M2-site [Wood and Blundy, 1997]. Solid/melt distribution coefficients for high valence ions Zr and Hf were taken from Green et al. [2000], multiplied by the ratio between DSm from table above and DSm from Green et al. [2000] in order to preserve internal consistency.

  • b

    Calculated from measured partitioning of trace elements between plagioclase and clinopyroxene, and orthopyroxene and clinopyroxene. Also shown are one standard deviation errors. Elements for which no distribution coefficients are shown had calculated coefficients smaller than uncertainty (1 standard deviation) and were not used in backstripping calculations (Figure 9).

Cpx/melta0.1560.2330.0400.3270.4310.1900.3260.6110.6720.7160.7400.7410.7310.7230.6910.6520.6100.569
Plag/meltb0.160 ± 0.1110.11 ± 0.0320.59 ± 0.210.072 ± 0.0350.048 ± 0.020--0.022 ± 0.0190.39 ± 0.120.014 ± 0.0120.013 ± 0.0080.008 ± 0.0050.004 ± 0.003-----
Opx/meltb-----0.015 ± 0.005----0.030 ± 0.0150.043 ± 0.0170.064 ± 0.0260.062 ± 0.0140.072 ± 0.0230.106 ± 0.0320.154 ± 0.0360.185 ± 0.066

[30] On the basis of the REE geochemistry, we can rule out that the melt had the composition of a boninite or island arc basalt (Figures 9d and 10a). Such melts are generally LREE enriched with respect to the middle and heavy REE. There is, therefore, no geochemical evidence that suggests that the Othris Peridotite Massif is part of a supra-subduction ophiolite, as has been proposed by Bizimis et al. [2000]. Its relation with boninitic lavas locally found in the Othris Mountains [Cameron et al., 1979], and to peridotite bodies in Othris which show evidence for extraction of boninitic melts [Bizimis et al., 2000; Rassios and Smith, 2000], therefore remains enigmatic. It is possible that there is not simply one Othris Ophiolite, but rather that oceanic fragments with different tectono-magmatic affinities were assembled in the accretionary complex of the Othris Mountains.

Figure 10.

Diagram showing fractionation of MREE (Nd) with respect to HREE (Yb) in different melts, modified from Figure 7 of Koga et al. [2001]. Boninite field (BON) based on data from Hickey and Frey [1982] and Taylor et al. [1994]. Data from ultra-depleted melt inclusions (UDM) from Sobolev and Shimizu [1993] and Shimizu [1998]. MORB field from references given in Koga et al. [2001]. Composition of Othris melts (open diamonds) in equilibrium with measured clinopyroxene trace element data overlap with those of UDM. Dark squares show average melt compositions in equilibrium with clinopyroxene after backstripping the effect of a hypothetical trapped melt fraction of 0–40% (see text and caption of Figure 9 for details about method).

3.3. Origin of Depleted Melts

[31] Depleted melts for which evidence is found at mid-ocean ridges are thought to be derived from low-pressure partial melting of already depleted peridotite [Ross and Elthon, 1993; Sobolev and Shimizu, 1993], and therefore represent the last melt fractions produced by melting underneath ocean ridges. Alternatively, melts can acquire a depleted, high SiO2 character by extensive reaction with depleted peridotites during porous melt flow. For instance, REE melt compositions very similar to those described here were found to be in equilibrium with cumulate plagioclase and corroded clinopyroxene in peridotites from the Internal Ligurides and from Corsica; these melts also crystallized orthopyroxene [Rampone et al., 1997]. Piccardo et al. [2002] suggested that these depleted melts were produced by pyroxene dissolution during porous flow. However, such a process probably requires a high rock/melt ratio during melt-rock reaction, and would probably lead to spoon-shaped REE patterns as the result of chromatographic effects [e.g., Navon and Stolper, 1987]. Although such spoon-shaped patterns may be present in clinopyroxenes in some of the plagioclase-peridotites from Othris [Barth et al., 2003], they are notably absent in sample GOF1 discussed here. We therefore favor an interpretation in which melts for which we see evidence in the Othris peridotites were produced by shallow (but probably still in the spinel lherzolite field) melting of already depleted mantle rocks. For comparison, Seyler et al. [2001] report evidence for a impregnation event in abyssal peridotites by depleted melts which were undersaturated in orthopyroxene, and which were probably produced at relatively deep levels, close to the base of the spinel-lherzolite field. This suggests that production of depleted melts is not necessarily restricted to shallow levels underneath mid-ocean ridges.

3.4. Implications for Melt Transport Mechanisms

[32] The earliest phase of melt-rock reaction for which evidence is locally preserved in the Othris peridotites produced strongly corroded orthopyroxene clasts. The spatial association of the corroded orthopyroxene clasts and large replacive dunite bodies in the western block of the Katáchloron area suggest that the melts responsible for this stage of melt-rock reaction were at least partly channeled. This stage could represent the interaction of rising mantle melts with peridotites at near-adiabatic temperatures (Figure 11a). As discussed by Aharonov et al. [1995] and Kelemen et al. [1995a, 1995b], mantle melts traveling upward at adiabatic temperatures will tend to dissolve enclosing peridotite, which leads to the melts becoming saturated in olivine (but not orthopyroxene). Such melts thus tend to dissolve pyroxenes and crystallize olivine, and will thus create their own, highly permeable, dunite melt channels [Aharonov et al., 1995; Kelemen et al., 1995a, 1995b; Spiegelman et al., 2001]. As soon as such melts are completely surrounded by olivine (dunite), in which they are saturated, they effectively cease to interact with their wall rocks and travel upward in a nearly closed-system. If near-adiabatic temperatures persist all the way to the base of the crust, as expected at fast spreading ridges, melts can be delivered almost directly from dunite channels to the crustal magma-chamber.

Figure 11.

Interaction between rising mantle melts and peridotite wall rock, illustrated using pseudo-ternary phase diagrams on the plane forsterite olivine (Fo)-diopside clinopyroxene (Di)-silica (Si), projected from spinel (after Figure 2 of Kelemen et al. [1995a]). A central element of the analysis is that the olivine phase field expands drastically with decreasing pressure [Andersen, 1915; Kelemen et al., 1995a, 1995b; Niu, 1997]. (a) Adiabatically ascending melt fractions produced at high pressures are above their liquidus and will tend to dissolve enclosing peridotite, creating a pyroxene-free, dunitic porous flow channel. The flow channel will consist of only olivine, in which the melts will be saturated. Upon cooling, such melts will first crystallize olivine, until they reach the Fo-Di co-tectic, after which they will also crystallize clinopyroxene (and plagioclase). Orthopyroxene will be the last phase to crystallize. As the bulk of the melts produced at depth migrate through the mantle in conduits, interaction between melts and wall rock mantle rocks will be minimal and shallow peridotite will not be in equilibrium with mid-ocean ridge basalts [Kelemen et al., 1992, 1995a, 1995b]. (b) Melt fractions produced by shallow mantle melting underneath mid-ocean ridges will have more SiO2-rich compositions than high-pressure mantle melts, and will be close to orthopyroxene-saturation. Such melts may initially dissolve some enclosing harzburgites. This will lead to saturation in olivine and olivine crystallization combined with pyroxene dissolution will rapidly drive the melt to the olivine-enstatite phase boundary. This will cause orthopyroxene crystallization (or reaction of olivine to orthopyroxene if the phase boundary is an incongruent reaction boundary, as expected at low pressures), which will immediately lead to choking of porous melt channels which may have formed. Therefore low-pressure melt fractions are inherently unable to create their own, stable dunite melt channels. They will mainly travel by diffuse porous flow and remain isolated from the bulk melt produced underneath mid-ocean ridges which flows through dunite conduits. Clinopyroxene and plagioclase are expected to be the last phases to crystallize from such low-pressure melts.

[33] At the Othris ridge-system, however, the base of the TBL and the thermal lithosphere penetrated well into the mantle, probably because of a slow spreading-rate and/or as the result of the cooling effect of a nearby transform fault [Dijkstra et al., 2001], and melts produced at depth would have had to travel through a cold lid of mantle rocks. Melts do not immediately crystallize completely as soon as they enter the thermal lithosphere; as long as they remain above their solidus temperature (<900–1000°C) they can traverse part of the cold mantle lid [e.g., Seyler and Bonatti, 1997]. The issue is whether they migrate by channel or diffuse porous flow through the base of the thermal lithosphere. The majority of the Othris peridotites contain widespread evidence for reaction between an orthopyroxene-saturated, depleted melt and harzburgitic peridotites, and subsequent fractional crystallization of cumulate phases, within the TBL and base of the thermal lithosphere. These reaction or crystallization products are not spatially restricted, suggesting that melt migrated through the rocks by diffuse porous flow. This behavior of depleted melts can be explained by their composition and their possible shallow origin. Low-pressure melts (i.e., melts produced at shallow levels in the melting column) are much closer to orthopyroxene saturation than melt fractions produced at high pressures (Figure 11b). Because low-pressure are formed just below or within the base of the TBL, they will start to cool shortly after their formation. They will be able to dissolve peridotite as they rise, as long as they remain at near-adiabatic temperatures, but the solubility of peridotite in low-pressure melt fraction is much less than in high-pressure melts, as shown by calculations by Kelemen et al. [1995a]. As a result of peridotite dissolution, the melts will become saturated in olivine; olivine crystallization drives the melts toward the olivine-orthopyroxene phase boundary, resulting in orthopyroxene crystallization and continuing peridotite dissolution (Figure 11b, see also [Kelemen et al., 1995a]). Since the olivine-orthopyroxene phase boundary is an incongruent reaction boundary at low pressure, melts following this boundary will cause transformation of olivine into orthopyroxene [Andersen, 1915; Niu, 1997], which is in good agreement with our petrographic observations. In this way, permeability will decrease and olivine-rich channels produced by low-pressure melts will choke and become unstable almost as soon as they are formed.

[34] The seemingly coeval dissolution and precipitation for which evidence is observed in the Othris peridotites can perhaps be explained by a stress-dependent incongruent reaction [Dijkstra et al., 2002]: in melts close to the reaction boundary, the reaction could be orthopyroxene-consuming in low-stress areas and orthopyroxene-producing in high-stress areas. In the fine grained tectonites there is indeed some evidence for such a stress-dependence of the reaction, as evidence for orthopyroxene dissolution is restricted to the pressure shadows of large orthopyroxene porphyroclasts (low-stress areas), whereas orthopyroxene precipitation seems to have mainly occurred along grain boundaries subparallel to the foliation in olivine-rich domains away from porphyroclasts (high-stress areas).

[35] The Othris peridotites also show that, upon further cooling, clinopyroxene and plagioclase started to crystallize. This would have further decreased the permeability of the rocks, arresting diffuse porous flow [Korenaga and Kelemen, 1997; Kelemen and Aharonov, 1998]. From that moment on only hydrofracturing produced new melt pathways to the surface, as evidenced by the large amount of gabbroic veins and dykes cutting the peridotites.

[36] In summary, high-pressure melts become orthopyroxene-undersaturated during adiabatic decompression and consequently dissolve orthopyroxene on their way up, creating high permeability dunite conduits [Kelemen et al., 1995a, 1995b; Aharonov et al., 1995; Spiegelman et al., 2001]. Because interaction between melt and wall rock is minimized, and because melt ascent rates are fast in high-permeability channels, such melts can retain their heat very effectively and stay at near-adiabatic temperatures up to very shallow levels, even if they have to travel through a cold mantle lid. In contrast, our study of the Othris peridotites leads us to suggest that low-pressure melts are inherently unable to create their own (large) dunite melt channels. They can actually start to crystallize orthopyroxene when they enter the TBL soon after their formation, reducing permeability and therefore forcing themselves to migrate by diffuse porous flow. In this way, high- and low-pressure melt fractions can remain separated during transport to the Earth surface, migrating by different mechanisms (Figure 1d). This provides an explanation why depleted melts are occasionally preserved in melt inclusions [Sobolev and Shimizu, 1993], and can crystallize separately in depleted cumulate gabbros at mid-ocean ridges [Ross and Elthon, 1993; Coogan et al., 2000] without being aggregated into MORB [Kelemen et al., 1997]. Furthermore, our results show that diffuse porous flow and melt-rock reaction can occur at the scale of at least 1–2 km.

4. Conclusions

[37] The majority of the Othris peridotites contain evidence for reactions between harzburgites and SiO2-rich, LREE-depleted melts on a km-scale, and for fractional crystallization of LREE-depleted plagioclase, orthopyroxene and clinopyroxene within the TBL and the basal part of the thermal lithosphere. The depleted melts were close to orthopyroxene-saturation and migrated by diffuse porous flow as they were unable to create their own high permeability channels. Such depleted melts, probably produced by melting of an already depleted mantle source at relatively shallow levels underneath a spreading center, can remain isolated from melt fractions produced at deeper levels, since they migrate by different mechanism. On the basis of REE geochemistry, we can rule out a boninitic or island arc origin for these melts. We find, therefore, no support for the hypothesis that the Othris Peridotite Massif represents the mantle part of a supra-subduction zone ophiolite. The existence of ophiolitic fragments with supra-subduction zone characteristics elsewhere in the region [Bizimis et al., 2000] suggests that there is not one Othris Ophiolite, but that different ophiolitic fragments were assembled during subduction-accretion in the Othris Mountains.

Acknowledgments

[38] Anna Rassios (IGME, Kozani) is thanked for support in Greece and for sharing some ideas and unpublished material on the Othris Peridotite Massif. Thanks are due to the helpful people of Makrirahi, Smokovou, and Loutra Smokovou, for providing accommodation in their villages during several field seasons in Othris. Rob Wilson is thanked for assistance with microprobe analyses at Leicester University. Further thanks to Wim van Westrenen for his help with calculating partition coefficients. Constructive critical reviews by Elisabetta Rampone, Georges Ceuleneer, and by volume editor Peter Kelemen improved the manuscript considerably. The research in this paper was mostly carried out at Utrecht University and was part of a doctorate project of A.H.D. funded by NWO Pionier subsidy #030-7534. A.H.D. was funded by a European Commission Marie Curie Postdoctoral Fellowship grant while writing this manuscript at Leicester University.

Ancillary