Nature and distribution of dykes and related melt migration structures in the mantle section of the Oman ophiolite



[1] We conducted a comprehensive field, petrographic, and microprobe study of the dykes and porous flow channels cropping out in the Oman harzburgites. The 36 rock types we recognized among of about 1000 samples can be grouped in two main magma suites contrasted in terms of structural and textural characteristics, modal composition, order of crystallization, and phase chemistry. One suite (troctolites, olivine gabbros, opx-poor gabbronorites, and rare oxyde gabbros) derives from MORB-like melts. The other suite (pyroxenites, opx-rich gabbronorites, diorites, and tonalite-trondhjemites) derives from melts richer in silica and water than MORBs and ultradepleted in incompatible elements. Dykes and porous flow channels from the MORB suite are restricted to a few areas, covering only 25% of the mantle section. This is an unexpected result as the deep Oman crust is made essentially of cumulates from MORB-like melts. Their composition, texture, and relations with the host harzburgites point to high mantle temperatures at the time of crystallization (likely above 1100°C, up to 1200°C for part of them), i.e., conditions close to the “asthenosphere/lithosphere” boundary. The largest outcrop of mantle harzburgites enclosing MORB like dykes is a 80 km long and 10 km wide corridor, parallel to the strike of the sheeted dyke complex and centered on an area where a former mantle upwelling has been unambiguously defined (the Maqsad “diapir”). A few other occurrences of mantle cumulates from the MORB suite are smaller than the Maqsad area and have a lesser abundance of troctolites (i.e., of high-temperature cumulates). We interpret the troctolite zones of Oman as the witnesses of former diapirs frozen at various stages of their development. Dykes belonging to the depleted suite are the most common in Oman harzburgites. Their structural and textural characteristics show that they crystallized in a mantle colder than the melt (likely in the range 600°C to 1100°C). A possible origin for the parent melts of this suite is in situ partial melting of the shallow and partly hydrated lithosphere residual after MORB extraction. Our data support the view that feeding magma chambers with MORBs is a focused (and likely episodic) process involving the rise of hot mantle to the base of the crust through a lithospheric lid accreted during a previous diapiric event. They suggest also that the shallow mantle beneath spreading centers is a place of important petrologic processes, some of them predicted on the basis of MORB composition (e.g., fractionation inside melt conduits) and other ones unexpected (e.g., remelting of the depleted lithosphere).

1. Introduction

[2] The Oman ophiolite offers a unique opportunity to observe and sample the deep horizons of the oceanic crust poorly accessible in the present-day oceans. Data collected in Oman have greatly contributed to the debate about the structure and petrologic evolution of magma chambers beneath oceanic spreading centers [e.g., Pallister and Hopson, 1981; Smewing, 1981; Lippard et al., 1986; Nicolas et al., 1988; Juteau et al., 1988; Reuber et al., 1991; Boudier et al., 1996; Korenaga and Kelemen, 1997; Chenevez et al., 1998; MacLeod and Yaouancq, 2000]. The way magma chambers are fed is another puzzling aspect of oceanic crust genesis. Here also, field observations in Oman have been widely used to infer possible melt migration mechanisms in the shallow mantle [e.g., Nicolas, 1986; Ceuleneer and Rabinowicz, 1992; Ildefonse et al., 1993; Kelemen et al., 1995, 1997a; Ceuleneer et al., 1996]. The mantle section of the Oman ophiolite is essentially made of harzburgites with a mean composition of about 80% olivine 20% orthopyroxene and minor disseminated chromite, interpreted as residual after partial melting and melt extraction. Igneous features whose lithological nature contrasts with these monotonous harzburgites are interpreted as “melt migration structures.” They constitute a minor but ubiquitous component of the Oman mantle section. They present various structural relations with the host harzburgites (interstitial minerals, irregular bodies, planar features with diffuse or clear cut boundaries, etc.) that have been attributed to different melt migration processes: diffuse melt flow in a porous matrix, reactive flow, melt flow confined in cracks, etc.

[3] Field evidence is, however, of limited use when not associated to a meticulous characterization of the lithologies defining the melt migration structures. This step is necessary to determine likely physical conditions that prevailed during melt migration and to infer the composition of these melts. In spite of the fact that the Oman ophiolite is one of the most studied igneous bodies in the world, no comprehensive petrographic and microprobe study of these melt migration features has been published up to now. Previous works either focused on structural aspects alone or, when including petrology, concerned a quite limited amount of samples and/or a specific area of Oman (see next section for a short review). The aim of the present paper is to fill this gap in our knowledge of the Oman ophiolite record. We described in thin section and used the microprobe to analyze about 1,000 samples of melt migration structures we collected all along the Oman ophiolite. Our goal was to answer a few basic questions concerning melt migration in the shallow mantle on a statistically valid basis:

[4] • What is the lithological diversity of these melt migration structures?

[5] • What are their relative abundance, their mutual relations and their distribution?

[6] • What can we infer concerning the nature of their parent melts?

[7] • Are these melts the ones that fed the overlying crustal section?

[8] • What are the relations between melt migration and crustal building processes?

[9] • Are these data helpful in the frame of the long standing debate concerning the tectonic setting of the Oman ophiolite?

[10] In this study, we focused mainly on dykes sensu stricto (i.e., slabs of rocks limited by two clear cut and parallel contacts with the host harzburgites (see Figure 6a) because they are by far the most abundant melt migration structures and are thus the witnesses of one (or more) important stage in the igneous history of the ophiolite. We included also in our survey structures that share many characteristics with dykes (e.g., planar aspect, lithology that contrasts with the mantle harzburgites) but that have diffuse boundaries with their host (Figure 4a). We will use the term “porous flow channels” [Kelemen et al., 1995] when referring to these structures. We have also sampled in some places patches of interstitial minerals but, as these features are extremely tenuous, it was impossible to map their distribution at the scale of the ophiolite. We have not included in the present study the abundant melt migration structures cropping out in the dunitic mantle/crust transition zone [e.g., Ceuleneer and Rabinowicz, 1992; Korenaga and Kelemen, 1997; Kelemen et al., 1997b] because the origin of this horizon (residual versus cumulative) remains controversial. As our purpose was to study mantle processes, to group these structures with the ones observed in the underlying harzburgites, of undisputable residual nature, could be misleading. We have also ignored, during our systematic sampling, the dunites that crop out inside the harzburgites and present a wide variety of modes of occurrence (from wispy layers to kilometer sized bodies). The petrological characteristics and distribution of these dunites have been and are currently studied by Peter Kelemen and coworkers [Kelemen et al., 1995, 2000; Braun and Kelemen, 2002]. Finally, chromite ore bodies were not considered here because their origin raises several specific (still unanswered) petrogenetic questions [e.g., Lorand and Ceuleneer, 1989; Leblanc and Ceuleneer, 1992; Schiano et al., 1997; Arai, 1998], not necessarily related to the general problem of ocean crust genesis.

[11] In the context of the present study, we have surveyed the various massifs comprised between wadi Hatta in the North and wadi Khabbah in the Ibra area at the southeastern limit of the ophiolite (Figure 1), i.e., a distance of about 350 km roughly parallel to the average strike of the sheeted dyke complex (inferred azimuth of the paleo-ridge axis). The present paper is based on the thin section description and microprobe study of about 800 samples we collected during a few recent field seasons (1997, 1999, 2000 and 2001) and integrates microprobe data we previously collected on about 200 samples from the Maqsad area that are partly integrated in previous publications [Ceuleneer et al., 1996; Benoit et al., 1999] but that were never fully published before.

Figure 1.

Geographical names used in the text.

2. Previous Work

[12] Mafic dykes associated with Oman “serpentinites” were described by early surveyors [Carter, 1850; Pilgrim, 1908; Lees, 1928]. They interpreted these features as resulting from a magmatic event unrelated to the formation of the ultramafic rocks. In his pioneer petrological study of the Oman ophiolite, Reinhardt [1969], noticed that “eucritic gabbro” dykes cross cutting the peridotites are particularly abundant in a ∼500 m thick zone on the “ultramafic side of the peridotite-gabbro contact,” while ultramafic rocks never occur as dykes in the layered gabbros. He used this and other arguments to support his view of a recent (mid-cretaceous) intrusion of gabbroic magmas into a much older (pre-Permian) peridotitic igneous body. This picture may look somewhat naive today but Reinhardt demonstrated that the peridotitic and gabbroic units cannot be interpreted in the frame of a single cumulate suite like in the volcano-pluton model, popular at that time, and so largely contributed to the development of modern concepts about the Oman ophiolite. T. Peters [Allemann and Peters, 1972] completed the picture drawn by Reinhardt: he showed that peridotites were equilibrated at rather great depths (“mantle pressures”) and described in the peridotite unit a wide diversity of “intrusive” lithologies; in addition to different kinds of gabbros, he reported the occurrence of pyroxenites and of more “evolved” intrusions (diorites-trondhjemites). He showed, in the northern part of the ophiolite, that the mantle dykes are not restricted to the neighborhood of the mantle/crust boundary as in Reinhardt's picture and interpreted these structures as channels used by mantle partial melts to reach the crust.

[13] This way of thinking was followed by Boudier and Coleman [1981], who documented the same field relations in the southern part of the ophiolite (Wadi Tayin massif). They proposed that pyroxenite dykes emplaced while the host harzburgites were still hot, before or just after their incorporation to the lithosphere, an episode followed by the injection of gabbroic dykes at shallower depth. On the basis of bulk rock major element data and of petrographic studies, they suggested that mantle dykes are cumulates from an olivine-poor tholeiite somewhat different from the parent liquid of the crustal section, assumed to be picritic. Gregory [1984] argued that the composition of the melts traveling through the mantle section was dramatically modified by reaction with the host peridotites so that it was hopeless to characterize their origin simply.

[14] Ph.D. students from the Open University (mostly Browning [1982]), whose results are summarized by Lippard et al. [1986], described mantle dykes in the central part of the Oman ophiolite. They concluded that olivine gabbro dykes are the most abundant at shallow depth in the mantle section and are equivalent to the layered crustal gabbros in terms of modal composition, order of crystallization and phase composition. As pyroxenite dykes appear more abundant at greater depth, they proposed that pyroxenites are higher pressure cumulates issued from the same kind of melts as the gabbro dykes and the crustal cumulates. This interpretation was corroborated by other field surveys [e.g., Boudier et al., 1983; Nicolas et al., 2000b]. Lippard et al. [1986] reported also on the occurrence of rare gabbronoritic dykes issued from a source depleted in incompatible elements they attributed to late subduction zone magmatism (i.e., cogenetic with the Alley volcanics).

[15] More recently, melt migration structures of Oman have been essentially studied by three teams adopting contrasted but complementary approaches: a team of structural geologists from Montpellier University, a team of petrologists from the Woods Hole Oceanographic Institution, and our team from the Observatoire Midi-Pyrénées (Toulouse). The Montpellier team systematically mapped the dyke orientations in order to determine the paleo-stress field at the time of melt injection whatever the nature of these melts. Their main result (reviewed by Nicolas et al. [2000b]) was to show that the dyke injection pattern is broadly consistent with the mantle flow structures and with the spreading directions deduced from the attitude of the sheeted dyke complex. The Woods Hole team performed detailed geochemical studies of samples of gabbros and pyroxenites collected in a few outcrops of Oman in order to constrain the composition of the melts migrating through the mantle section and the effects of assimilation of the depleted harzburgites on the composition of these melts [Kelemen et al., 1995, 1997a]. They developed specific techniques to map the distribution of dunitic channels in mantle harzburgites [Kelemen et al., 2000; Braun and Kelemen, 2002]. The main concern of our team is the accurate determination of the gabbroic and pyroxenitic dykes composition and of the variability of this parameter in Oman, our objective being to better constrain the physical and petrological processes involved in the ocean crust generation. Our first results [Ceuleneer et al., 1996; Amri et al., 1996; Benoit et al., 1996, 1999] concerned the Maqsad area where a paleo-mantle upwelling has been recognized [Rabinowicz et al., 1987; Ceuleneer et al., 1988; Ceuleneer, 1991]; they will be integrated in the discussion of the present paper.

3. Field, Petrographic, and Geochemical Characteristics of the Oman Mantle Dykes and Related Melt Migration Structures

[16] Results of our field, petrographic and microprobe survey of Oman mantle dykes are presented in Figure 2 and auxiliary material Tables A1 and A2. Figure 2 is a list of the various petrographic types we recognized in our sample population, with indication of abundances, grouping in lithological families, and caption of the symbols used throughout this paper. In Tables A1 and A2, we present the main characteristics of each of the 1,001 samples we analyzed at the microprobe: the coordinates of the outcrop, some field relationships (orientation, size, relation with the country rocks, etc.), a digest of the petrographic description (estimated modal proportions, textures) and the average phase compositions (major and minor elements determined at the microprobe). It is the database we used to construct our diagrams and maps.

Figure 2.

Summary of main characteristics for each lithological type determined in the sample population. Column “Symb” shows the symbols used in all figures: N = total number of samples; S = estimated surface of the mantle section where each lithological family is dominant; Mcum.: mesocumulate; Acum.: adcumulate; Mos.: mosaic; Diff.: diffuse segregation; PFC: porous flow channels; Int.: intrusion.

3.1. Lithological Diversity

[17] On the basis of modal and textural criteria, we have identified, among our sample population, 36 rock types (Figures 2 and 3). Rock types are named according to Streckeisen [1976] except that we did not follow the “10% limit” because it is generally meaningless in terms of petrogenesis while the absence or presence of a given mineral, even in concentration lower than 10%, can be diagnostic of an important boundary, in terms of parent melt composition for example. In a few specific cases, we have used textural characteristics to complete the classification based on modal abundances. In order to present our results in a synthetic way, we have grouped the so defined rock types in a few major lithological families. Rock types in each family share common compositional and textural characteristics. The name of a family refers to the main rock type in this family. Accordingly, the “olivine gabbro” family, for example, includes rocks that are not, sensu stricto, olivine gabbros, like olivine gabbronorites. This grouping is interpretative and thus a subjective stage of our study.

Figure 3.

Diagrams showing mineralogical modes for the four main phases: olivine (Olv), plagioclases (Plg), clinopyroxene (Cpx) and orthopyroxene (Opx).

3.1.1. Troctolite Family

[18] The rocks from the troctolite family are essentially made of cumulate plagioclase and olivine, with minor (<1%) chromian spinel. They are characterized by the absence of clinopyroxene that coprecipitated with plagioclase and olivine (Figures 4b and 5). Clinopyroxene may be present, however, as irregular veins of poikilitic crystals (Figures 4c and 4d) and may appear abundant at the scale of a thin section (in which case the rock is called a cpx-troctolite) (Figure 4d) but it remains a minor component at the scale of a dyke or of a porous flow channel. The modal proportions, although variable, present a well-defined peak at values of 25% olivine and 75% plagioclase (Figure 5). Although the crystallization of olivine and plagioclase appears largely contemporaneous, when an order of crystallization can be determined, plagioclase is generally the earliest phase: euhedral to subhedral plagioclase crystals, ranging in size from about 0.1 mm to a few mm, exceptionally 1 cm, are included in more irregular olivine crystals (Figure 4b). Troctolites frequently show medium grained mosaic textures with triple junctions at 120° diagnostic of sub-solidus grain boundaries readjustment and of textural equilibrium. When this is the case, the contact between olivine and plagioclase is generally underlined by a thin reaction rim made of pyroxene (both clinopyroxenitic and orthopyroxenitic rims can occur within a single thin section; see Figure 4b). This recrystallization is not related to shear deformation, as far as we can determine from field and thin section observations.

Figure 4.

(a) Field view of a porous flow channel in harzburgites. (b) Thin section (crossed polars) of a troctolite showing plagioclase partially included in olivine, mosaic texture and pyroxene reaction rims between olivine and plagioclase. (c) Sample of troctolite containing a one_centimeter clinopyroxene vein, Cpx appears poikilitic in the vein and is almost totally absent elsewhere in the sample. (d) Thin section (crossed polars) of a sample of Cpx-troctolite showing small subhedral to euhedral plagioclase included in olivine and a background of poikilitic Cpx.

Figure 5.

Histograms showing the modal proportions of plagioclase, olivine, clinopyroxene and orthopyroxene for rocks from the troctolite family.

3.1.2. Olivine Gabbro Family

[19] The olivine gabbro family constitutes another homogeneous group of rocks. It is characterized by the coprecipitation of olivine, plagioclase and clinopyroxene. Coprecipitation is deduced from the adcumulate texture and from the subhedral to anhedral character of these three minerals, although clinopyroxene in many samples shows a more interstitial character than plagioclase and olivine (Figures 6b6d), some samples being transitional, in terms of texture, with Cpx-troctolites. Modal proportions are quite variable from one dyke to the other but, here also, a more frequent modal composition can be defined unambiguously (Figure 7): 50% plagioclase, 15% olivine and 35% clinopyroxene. About one fifth of the rocks from the olivine gabbro family are actually olivine gabbronorites, containing interstitial orthopyroxene in minor amounts (typically less than 10%, rarely reaching 20%); olivine gabbros or gabbronorites with significant amounts (>5%, Figures 6c and 6d) of amphibole and/or oxides (ilmenite, hematite and magnetite) have been observed but are quite uncommon. The texture and grain size of olivine gabbros are generally homogeneous within a single dyke, fine to medium grained (a few hundred microns to one or two millimeters) adcumulates being the most frequent (Figure 6). Coarser grained olivine gabbros do exist but are uncommon. Olivine gabbronorites are among the finest grained rocks from the olivine gabbro family, some samples having a texture approaching that of dolerites.

Figure 6.

(a) Field view of several parallel dykes of fine grained olivine gabbro. (b) Thin section (crossed Polars) of a sample of olivine gabbro showing adcumulate texture and the interstitial character of clinopyroxene. (c) and (d) Thin section of an amphibole and oxide rich olivine gabbro (plane polarized (c); crossed Polars (d)) showing mesocumulate texture with interstitial hornblende (Hb) and hematite (He).

Figure 7.

Histograms showing the modal proportions of plagioclase, olivine, clinopyroxene and orthopyroxene for rocks from the olivine gabbro family. Orthopyroxene rich samples are olivine gabbros containing interstitial orthopyroxene.

3.1.3. Gabbronorite Family

[20] The rocks from the gabbronorite family show contrasted characteristics with the ones from the two previous families. Orthopyroxene is here a major (up to 40%) cumulus phase presenting a subhedral shape, suggesting that it appeared early in the crystallization sequence (Figures 8b and 8d) while clinopyroxene and plagioclase are more frequently anhedral and interstitial. Clinopyroxene is totally absent in about 10% of the samples from this family (these samples are thus true norites; see Figure 9). Olivine is uncommon and, when present, never abundant (<15%), but amphibole (hornblende to pargasite) is frequently present as a late interstitial mineral (Figure 8d), generally epitaxial on pyroxenes. When the entire family is considered, modal proportions are quite variable and the definition of an average or of a more frequent composition appears to be a meaningless exercise. When the gabbronorite type only is considered (excluding olivine gabbronorites and norites), the more frequent phase proportions are: 50% plagioclase, 25% clinopyroxene and 25% orthopyroxene (Figure 9). Textures are also quite variable in this family, even at the scale of a single dyke: they range from fine grained adcumlates to pegmatites, “giant” crystals (up to several tens of cm, Figure 8c) being quite common. Harrisitic growth on the inner dyke walls frequently results in dykes that are coarse grained on their margins and fine grained in their center. Grain size layering parallel to the dyke walls is frequently observed and can be interpreted in terms of multiple intrusions within a single crack or in terms of a complex nucleation and growth history. It is rather common to observe porphyroclastic to mylonitic textures in gabbronorites, especially in the amphibole and/or oxide-rich gabbronorites (Figures 8e and 8f), while this type of (relatively) low-temperature, high-stress deformation texture is virtually absent in samples of troctolites and olivine gabbros.

Figure 8.

(a) View of a 10 meter thick sub-horizontal dyke of gabbronorite. (b) Thin section (crossed Polars) of a fine grained gabbronorite; texture is adcumulate but orthopyroxenes appear subhedral to euhedral and are included in more interstitial plagioclases and clinopyroxenes. (c) Field view of a pegmatitic dyke of gabbronorite showing crystals larger than ten centimeters, the compass is about 6 cm wide. (d) Thin section (crossed Polars) of an amphibole_rich gabbronorite showing sub-doleritic texture; note the interstitial character of the plagioclase including small subhedral to euhedral crystals of pyroxene; amphibole (Am, hornblende) appears poikilitic. (e) Field view of a mylonitic gabbronorite dyke, deformation is heavier near the contact with harzburgite than in the core of this dyke. (f) Thin section (crossed Polars) of a mylonitic amphibole-rich gabbronorite showing relics of plagioclase and orthopyroxene in a fine grained matrix.

Figure 9.

Histograms showing the modal proportions for samples from the gabbronorite family for plagioclase, olivine, clinopyroxene and orthopyroxene. Very few samples contain olivine or are Cpx-free (norites), histograms show no preferred phase proportions for those facies.

3.1.4. Pyroxenite Family

[21] Rocks from the pyroxenite family range in composition from clinopyroxenites (rocks made of 100% clinopyroxene are not uncommon) to orthopyroxenites (up to 98% orthopyroxene, orthopyroxenites totally devoid of clinopyroxene or olivine have not been observed). In terms of modal composition, there is a perfect continuum between clinopyroxenite and websterites made of 50% clinopyroxene and of 50% orthopyroxene, while compositions intermediate between websterites and orthopyroxenites are uncommon (Figure 10). Clinopyroxenites are four times as abundant as orthopyroxenites (Figures 2 and 10). Most of the pyroxenites are devoid of olivine; in other words wehrlitic lithologies are uncommon among mantle dykes of Oman. There is no compositional gap between the clinopyroxenites-websterites and the gabbronorites: some samples can be described as “pyroxene-rich gabbronorites” or “plagioclase-bearing websterites,” although such transitional samples are not abundant (Figure 3). Textures of pyroxenites are similar to those described in gabbronorites, ranging from coarse grained to pegmatitic with a frequent overprint of high-stress deformation (undulose extinctions, development of sutured grain boundaries and of a fine grained matrix (Figures 11c11f). However, in contrast to the gabbronorites, textures are rarely mylonitic. There is no marked contrast, in terms of modal composition, between pyroxenite layering and discordant pyroxenitic dykes, although the modal composition of layering with diffuse contacts with the host harzburgites cannot be defined unambiguously.

Figure 10.

Histograms showing estimated modal proportion for samples from the pyroxenite family for plagioclase, olivine, clinopyroxene and orthopyroxene. Most of the samples are plagioclase and olivine free, orthopyroxene proportions show a continuum between clinopyroxenite (less than 5% Opx) and websterite to 50% Opx.

Figure 11.

(a) Field view of a large dyke of clinopyroxenite, dyke width for pyroxenite range usually from a few centimeters to a few tens of centimeters. (b) Sawn surface of a sample of clinopyroxenite showing the pegmatitic character of that rock. (c) Thin section (crossed Polars) of a coarse grained clinopyroxenite containing less than 5% orthopyroxene which appears as small subhedral to euhedral crystals included in clinopyroxene. (d) Thin section (crossed Polars) of a sample of orthopyroxenite showing recrystallized texture and small interstitial clinopyroxene at grain boundaries. (e) and (f) thin sections (crossed Polars) of websterites showing adcumulate texture (e) and partly recrystallized and deformed texture (f).

3.1.5. Diorite-Granite Families

[22] Other lithologies observed in mantle dykes of Oman belong to a suite ranging in composition from diorites (frequently transitional with amphibole bearing gabbronorites) to more felsic rocks like tonalites and trondhjemites. True granites do occur but are uncommon. Although frequently reported in previous field descriptions, diabase dykes (i.e., frozen melts) are extremely uncommon. Such lithologies do exist in rare places but numerous reported occurrences are actually black chloritites resulting from complete alteration of troctolites or gabbros.

3.1.6. Diopsidite Family

[23] We have also observed a type of mantle dykes never reported in previous studies, made of pure diopside (locally associated to pure anorthite or pure forsterite) with mineral compositions departing from magmatic trends (black triangles in Figures 1418) and with texture reminiscent of skarns. We attribute these lithologies to high-temperature hydrothermal processes; this particular point will be further developed in another publication.

3.2. Mineral Composition

[24] In order to further characterize the mantle dykes of Oman, we have systematically analyzed the main mineral phases for major and minor elements with the electron microprobe. Analytical conditions are given in the caption of Table A2. We had to restrict the total number of analyses to a maximum of 40 points per thin section, i.e., 5 to 10 points for each mineral phase. Apart from felsic lithologies where plagioclase presents marked compositional zoning, the variations of concentrations in a single thin section and, a fortiori, in a single mineral are within the analytical errors: accordingly, the average values calculated on a sample by sample basis, reported in Table A2 and plotted in figures are generally significant. This is consistent with the conclusions of Coogan et al. [2000], who have shown that major elements in oceanic gabbros are largely insensitive to complex processes associated to the circulation of late or post-cumulus liquids. When present, minor phases like ore minerals (sulfides, chromites and other oxides, etc.) have been analyzed (see Table A1), but variations in their compositions will not be discussed in the present paper.

[25] Mineral compositions show marked variations: the compositional fields and trends appear consistent with the classification into rock types and families we have proposed on the basis of petrographic observations alone (Table 1). Moreover petrographic determinations of the modal composition, order of crystallization and texture have shown that there is a closer affinity between rocks of the gabbronorite and pyroxenite families than between gabbronorites and other gabbroic lithologies. On the other hand, some rocks from the troctolite family show clearly transitional characteristics with rocks from the olivine gabbro family, allowing us to infer some genetic links between these families. Accordingly, we can schematically group our lithological families into two “super groups.” Rocks from the diorite-tonalite family appear, in that respect, closer to the pyroxenite-gabbronorite kindred than to the troctolite-olivine gabbros. These preliminary conclusions justify the grouping we made in some of the following chemical diagrams.

Table 1. Summary of Main Chemical Characteristics for Each Lithological Type Determined in the Sample Population
FamilyLithologyPlag. An%OlivinesOrthopyroxenesClinopyroxenesMagmatic Amphiboles
Fo%MnONiOXmgMnOCr2O3Al2O3XmgWo%Cr2O3Al2O3TiO2Na2OXmgXmg*8-Si(Na + K)A
TrotctolitesPl dunites80–9088–900.10–0.200.20–0.4085–910.15–0.250.60–0.701.50–2.50≃92≃480.60–0.802.50–2.90≃0.70≃0.35
Pl and Cpx dunites80–9085–900.10–0.200.20–0.4085–910.15–0.25≃0.50≃1.5087–9134–480.80–1.402.50–2.500.10–0.250.20–0.4080–900.90–1.001.60–1.750.40–0.65
Pl harzburgites80–9086–910.10–0.200.20–0.4085–910.15–0.250.50–0.701.50–2.00≃9036–381.10–1.302.00–3.000.10–0.400.20–0.45
Pl and Cpx harzburgites80–9586–910.10–0.200.20–0.4085–910.15–0.250.50–0.901.50–2.0088–9237–470.40–1.402.00–3.000.40–0.700.15–0.25
Cpx troctolites70–9575–910.15–0.300.05–0.3579–890.10–0.300.10–0.901.00–4.5080–9032–500.60–1.402.00–4.000.05–0.400.15–0.4080–900.90–0.981.40–2.000.45–0.75
Olivine GabbrosOl gabbros50–9570–900.10–0.40<0.4070–900.10–0.300.10–0.401.00–2.0080–9038–52<1.101.80–3.50<0.800.20–0.5075–920.80–1.000.40–0.801.25–2.00
Amph-rich Ol gabbros60–9069–820.25–0.500.10–0.3080–850.20–0.30≃0.101.00–1.5080–8745–480.20–0.501.80–3.500.30–1.100.25–0.5070–850.80–0.900.45–0.701.40–1.80
Opaques-rich Ol gabbros70–8083–870.10–0.20<0.30≃850.25–0.300.10–0.301.00–1.2580–8542–480.20–1.002.00–3.000.05–1.00≃0.50≃850.85–0.900.45–0.601.50–1.80
Opx-rich Ol gabbros60–9570–910.10–0.45<0.5075–900.10–0.45<0.701.00–2.5077–9034–49<1.102.00–3.00<1.200.10–0.4080–900.85–0.950.30–0.601.15–1.75
Amph and Opx-rich Ol gabbros60–9568–820.25–0.45<0.2075–800.25–0.45<0.201.00–1.5077–8746–480.05–0.802.00–3.000.15–0.750.30–0.4575–850.75–0.900.50–0.701.50–1.80
Amph-rich gabbros00–65≃68≃0.550.00≃1.0075–8544–48<0.801.50–2.500.20–0.700.20–0.3060–700.70–0.801.50–1.900.55–0.70
GabbronoritesOl gabbronorites85–9580–900.15–0.35<0.3080–900.10–0.300.20–0.701.50–2.0080–9034–480.40–0.802.00–3.000.15–0.400.10–0.20
Amph-rich gabbronorites78–9865–850.20–0.40<0.200.50–1.5075–8546–48<0.400.50–2.50<0.400.10–0.3050–850.70–0.980.50–1.750.10–0.60
Opaques-rich gabbronorites88–9855–650.40–0.80≃0.000.50–1.25≃9042–46≃0.400.50–2.000.10–0.200.05–0.2030–650.50–0.701.00–1.100.05–0.30
Amph-rich Norites50–9555–750.20–0.60≃0.001.00–2.0060–700.75–0.951.00–1.600.20–0.60
Amph and opaques-rich Norites30–9550–600.40–0.80≃0.000.75–1.5060–700.80–0.850.75–1.00<0.10
PyroxenitesPl websterites90–9575–800.20–0.250.10–0.20≃1.5083–8545–470.05–0.252.00–2.50<0.100.20–0.30
Ol websterites85–900.10–0.300.10–0.3085–91<0.150.30–0.501.00–2.0087–9042–500.20–1.402.00–2.500.25–0.35<0.10
Ol clinopyroxenites85–900.10–0.200.20–0.3088–90≃0.200.50–0.60≃1.5090–9346–490.60–1.401.75–2.75<0.300.10–0.25
Ol orthopyroxenites≃900.15–0.250.40–0.50≃90≃0.100.60–0.90≃2.0090–9349–511.00–1.403.50–4.00<0.10<0.05
Pl wehrlites85–8780–850.25–0.300.30–0.4089–9149–50≃1.202.50–3.00≃0.250.10–0.20
Pyroxenitic layerings87–910.05–0.200.20–0.7085–930.05–0.200.30–0.900.75–2.7589–9333–490.60–1.401.00–3.50<0.20<0.20
Hydrothermal diopsidites and gabbros>9591–95<0.20.30–0.50>92<0.100.70–0.80≃1.50>9349–52<0.10<0.75<0.15<0.10≃1.00>0.95<1.50<0.70
GranitesDiorites and Granodiorites02–95600.400.101.25≃87≃50≃0.00<0.25≃0.00≃0.1025–800.55–1.000.30–1.75<0.70
Trondhjemites and Tonalites00–9905–700.10–0.801.00–1.750.10–0.50

[26] Forsterite content of olivine ranges from about 92 to 70 in both super groups while the Mg/(Mg+Fe) (mgn) of pyroxenes shows a more pronounced variation range in the gabbronorite family (50 to 90%). As expected in the case of coprecipitation, Fo and mgn of pyroxenes are well correlated (correlation coefficient close to 0.95). The less magnesian gabbronorites are devoid of olivine (Figure 12). The NiO content of olivine and the Cr2O3 content of clinopyroxene are largely correlated with the Fo content and with the Mg number. These minor elements show an abrupt decrease as the Mg/(Mg + Fe) ratio departs from “primitive” values (i.e., Fo or mgn close to 90, see Figures 13 and 14). In both super groups, a contrast exists between a primitive family (troctolite and pyroxenite, respectively) and a more evolved family (olivine gabbro and gabbronorite, respectively), consistent with the crystallization order deduced from petrographic observations. In terms of Fo, troctolites are extremely homogeneous (variations are in the 88–92 range), the troctolites with poikilitic clinopyroxene being, on average, slightly more ferric. Olivine gabbros show a well-defined preferred Fo around values of 82–86 Fo, the number of samples decreasing rapidly as Fo decreases. Uncommon very evolved olivine gabbros sometimes contain amphibole. However, evolved and primitive olivine gabbronorites are equally abundant. Pyroxenites are always rather primitive in terms of mgn (that is never less than 80), but there is no more frequent value as well defined as in the troctolite and olivine gabbro families. Similarly, gabbronorites are very scattered in terms of mgn, a poorly defined more frequent value being observed for values comprised between 70 and 80.

Figure 12.

Histograms showing mgn variation in olivines for troctolite (in red) and olivine gabbro (in blue) families and in orthopyroxene for gabbronorite (in orange) and pyroxenite (in green) families. Very few gabbronorite and pyroxenite samples contain olivine so that such a diagram for olivine is significant only for troctolite and olivine gabbro families; on the other hand, few samples of troctolite or olivine gabbro contain orthopyroxene.

Figure 13.

Evolution of NiO with forsterite content in olivines; (a) each point represents the compositional average for each sample, see Figure 2 for legend of symbols; (b) density clouds for troctolite and olivine gabbro families. There are too few samples of olivine bearing pyroxenite or gabbronorite so that density contours can only be represented for troctolite and olivine gabbro families. NiO decreases with decreasing forsterite suggesting a differentiation trend from troctolites to olivine gabbros.

Figure 14.

Evolution of Cr2O3 with mgn in clinopyroxenes. (a) each point represents the average for each sample, see Figure 2 for legend of symbols; (b) and (c) density clouds show domains of composition where each lithological family is present. Cr2O3 decrease from troctolite family to olivine gabbro and gabbronorite families; values are very dispersed for the pyroxenite family but the majority of the samples from that group is less chromian than the average of troctolites.

[27] The Al2O3 content of pyroxenes is different from one family to the other (Figure 15). Pyroxenes from the pyroxenite family have low Al2O3 compared to the pyroxenes from the cpx-troctolites and olivine gabbros with the same mgn. Most of the clinopyroxenes from the pyroxenite and gabbronorite families have Al2O3 contents lower than 2%, with some higher values in pyroxenite layers. Al2O3 content is generally higher than 2% in cpx-troctolites and olivine gabbros. There is no marked evolution of the Al2O3 content of pyroxenes with mgn.

Figure 15.

Evolution of Al2O3 with mgn in clinopyroxenes (a) each point represents the compositional average for each sample, see Figure 2 for legend of symbols; (a) the gray-green zone shows the chemical domain of pyroxenitic layering; (b) and (c) density clouds showing compositional domains for each lithological family. Samples from troctolite and olivine gabbro families present about the same values for Al2O3, samples from pyroxenite family show lower concentrations.

[28] TiO2 content in clinopyroxenes show well-defined variations from one family to the other (Figure 16): in pyroxenites, this value is generally below the detection limit (about 0.12%). In the gabbronorite family, a slight increase with decreasing mgn from 0.15% to 0.25% is observed. The behavior of Ti in the troctolites and olivine gabbros families is different. Primitive cpx-troctolites present a wide scatter in TiO2 content, from values below the detection limit to about 0.3%, while the TiO2 content of clinopyroxenes in olivine gabbros increases abruptly to values of 0.8% for a slight decrease of the mgn. The behavior of Na in clinopyroxenes largely mimics that of Ti (see Figure 17).

Figure 16.

Evolution of TiO2 with mgn in clinopyroxenes. (a) each point represents the compositional average for each sample, see Figure 2 for legend of symbols; (b) and (c) density clouds showing compositional domains for each lithological family. Compositional domain for olivine gabbro family includes that for troctolite family (represented by Cpx-_troctolite) and reaches the highest values of TiO2. Most samples from the pyroxenite family are below analytical detection limit.

Figure 17.

Evolution of Na2O with mgn in clinopyroxenes. (a) each point represents the compositional average for each sample, see Figure 2 for legend of symbols; (b) and (c) density clouds showing compositional domains for each lithological family.

[29] The An content of plagioclase is the chemical parameter that is the most consistent with our grouping in rock families: the variation diagram where An is plotted as a function of the Mg/(Mg + Fe) of coexisting ferromagnesian minerals (Figure 18) shows that troctolite and olivine gabbro families, on the one hand, and gabbronorites, on the other hand, define contrasted differentiation trends. Gabbronorites are remarkably An-rich, An always being higher than 85%, and higher than 90% in many samples. No overall decrease of the An content with the Mg/(Mg + Fe) is observed, in spite of the wide variation range of mgn in the gabbronorite family. In the troctolite family, An content reaching 85 is observed only for the more primitive samples. The An content in troctolites varies from 70 to 90%. An content shows a vague tendency to decrease with decreasing Mg/(Mg + Fe) in the cpx-troctolites and olivine gabbros.

Figure 18.

Evolution of An% in plagioclase with mgn in clinopyroxene (or from a “virtual” clinopyroxene calculated from the Fo of olivine in the case of troctolites (ss), and from the mgn of orthopyroxene in the case of norites (ss). This calculation is based on the excellent correlation between mgn (cpx), Fo (ol) and mgn (opx) in lithologies where cpx is associated to olivine and/or opx (a) each point represents the compositional average for each sample, see Figure 2 for legend of symbols; (b) and (c) density clouds showing compositional domains for each lithological family. Samples from pyroxenite family do not contain any plagioclase, so they are not present; gabbronorite family shows a large variation in clinopyroxene mgn with almost no variation in plagioclase An% while, in contrast, troctolite and olivine gabbro families show a larger variation in plagioclase An% than in clinopyroxene mgn.

3.3. Field Characteristics and Distribution

[30] The distribution of the various mantle dyke lithologies in the Oman ophiolite is reported on the map of Figure 19. Typical field characteristics of these dykes are illustrated in several photographs (Figures 4, 6, 8, and 11). Contacts between mafic cumulates and host mantle harzburgites are in most cases sharp, cumulates having crystallized between two parallel surfaces of variable spacing. Dyke tips [e.g., Rubin, 1995] are frequently observed and indicate that dykes were upward, propagating cracks. The generic name of “dyke” given to mantle cumulates refers to these field relations and, implicitly, to a brittle emplacement mechanism. As mentioned above, diffuse contacts between melt transport features and host harzburgites are also observed from place to place; this texture is generally attributed to former intergranular percolation of the melt in the host peridotites, accompanied by partial dissolution of the residual orthopyroxene leading to “depleted” margins that can evolve to dunites [Kelemen et al., 1995]; the strong heterogeneity of cumulate crystal proportions in the case of diffuse structures evokes channelized rather than homogeneous porous flow [see Ceuleneer and Rabinowicz, 1992; Kelemen et al., 1995; Ceuleneer et al., 1996]. In the case of gabbroic cumulates, the distinction between the residual host rock and the cumulate crystals is relatively easy to draw as the residual mantle rocks of Oman are harzburgites devoid of plagioclase and virtually devoid of clinopyroxene. However, the distinction is much trickier, generally impossible on the base of field observation alone, in the case of olivine or orthopyroxene rich cumulates.

Figure 19.

Simplified geological map of the ophiolite of Oman showing the location of each dyke sampled and the distribution of the main lithological families inferred from these data.

[31] The present study has confirmed an observation we made in the Maqsad area: the troctolite family is the only lithological family where diffuse relationships (Figure 4a) between cumulates and the host mantle peridotites are commonly observed. We have attributed this to the fact that the ductile/brittle transition for melt migration occurs, in the shallow mantle, at a temperature of about 1200°C, i.e., within the crystallization field of troctolites [Ceuleneer et al., 1996]. The unexpected fact that the present study has demonstrated is that the central part of the Maqsad paleo-upwelling is the only area in Oman where troctolites (both in dykes and in porous flow channels) are the dominant lithology. Away from the Maqsad area, the occurrence of troctolites associated with olivine gabbro dykes can be followed northwestward in the Nakhl area and southeastward in the Samad area. It forms a 80 km long corridor centered on the Maqsad paleo-upwelling and parallel to the local strike of the sheeted dyke complex (i.e., N130°E); its width reaches a maximum of about 10 km in Maqsad, narrowing slightly to the NW and to the SE.

[32] While the troctolite dykes are generally rather thick (one to a few meters) and isolated, olivine gabbros occur more frequently in groups, or “swarms” of parallel dykes a few centimeters to a few decimeters thick (Figure 6a). Olivine gabbro dykes have a preferred orientation parallel (in terms of strike and of dip) to the local orientation of the sheeted dyke complex and thus probably were emplaced in a regime of pure extension (see Figure 20). While depleted margins are common in the case of troctolite dykes, such features are rarely observed in the case of olivine gabbro dykes; similarly, harrisitic growth on their inner wall is not frequent.

Figure 20.

Stereonets (pole of planes projected on the lower hemisphere) showing orientation of dykes for each family of rocks in Hilti and Fizh areas. Sheeted dyke complex orientations are represented by black and gray lines and are annotated as follows: KWD: Kahwad block [Misseri, 1982]; MSB: Musibit block [Misseri, 1982]; HAY: Haylayn block [Pallister and Hopson, 1981; Browning, 1982; Ceuleneer, 1986]; SAM: Sarami block [Ceuleneer, 1986]; NAE: East Nakhl block [MacLeod and Yaouancq, 2000]; NAW: West Nakhl block [Browning, 1982; Ceuleneer, 1986]; HIL: Hilti block [Ceuleneer, 1986]; NRS: North Suma'il block [Misseri, 1982; Ceuleneer, 1986]; SDS: South Suma'il block [Misseri, 1982]; MQD: Maqsad area [Ceuleneer, 1986; Jousselin et al., 1998]; IBR: Ibra block [Pallister and Hopson, 1981; Misseri, 1982]; WTN: Wadi-Tayin block [Misseri, 1982; Ceuleneer, 1986]; WUQ: Wuqbah block [Pallister and Hopson, 1981; Browning, 1982; Ceuleneer, 1986]; FZS: South Fizh block [Ceuleneer, 1986]; FZN: North Fizh block [Reuber, 1988]. See Figure 2 for legend of symbols.

[33] At the scale of our global map (Figure 19), the distribution of troctolite and olivine gabbro families can hardly be distinguished, confirming the deduction from petrographic and microprobe study that these two families are genetically related. Another unexpected conclusion of our survey is that occurrences of olivine gabbros are not common in the Oman ophiolite. Apart from the large corridor centered on Maqsad, olivine gabbros constitute the dominant dyke lithology (associated to minor troctolites) only in a zone parallel to the Maqsad corridor, about 20 km to the northeast going from the Kahwad massif to the south of the Wadi-Tayin massif, and a zone of smaller extent (about 15 km) centered on wadi Hilti in the northern part of the Oman ophiolite. In all other areas we surveyed, dykes described as “gabbros” by previous workers on the basis of field observations belong to the family of depleted opx-rich gabbronorites.

[34] Rocks from the pyroxenite and gabbronorite families are generally associated in the field, although either pyroxenites or gabbros are generally dominant in a given area. The frequently reported distribution of pyroxenites and gabbros according to paleo-depth is not supported by our observations: there are areas where pyroxenite dykes are observed up to Moho levels while gabbronorites may cross cut deep horizons of the mantle section. As in the troctolite-olivine gabbro kindred, this association supports a common origin for pyroxenites and opx-rich gabbronorites. Cumulates from these families always occur as dykes or as “intrusions” (i.e., irregular bodies reaching a few hundred meters in extent). Dyke thickness is extremely variable, ranging from a few centimeters to more than 10 meters (Figure 8a). Wall rock reaction, leading to the development of depleted to dunitic margins is not uncommon at the margins of pyroxenite dykes, and is commonly observed in the case of pyroxenite layering, but is hardly observed in the case of gabbronorites. Harrisitic growth on the inner dyke walls is the rule in these families, and dykes are frequently composite. They may contain xenoliths of harzburgite.

[35] Abundance and thickness of pyroxenite and gabbronorite dykes is quite variable. Interestingly, the pyroxenite-gabbronorite pegmatites reach their maximum abundance in the vicinity of domains occupied by troctolites and olivine gabbros. This is the case all along the Maqsad corridor, whose periphery is massively injected by pyroxenites and gabbronorites [see also Amri et al., 1996; Ceuleneer et al., 1996], but also around the olivine gabbro area of wadi Hilti. Conversely, in other massifs, like the Rustaq and Sarami area, the abundance of pyroxenites and gabbronorites is not exceptionally high, although these lithologies are always present. The orientation of pyroxenite and gabbronorite dykes (Figure 20) appears much more random than in the case of troctolites and olivine gabbros, and is generally not parallel to the strike of the sheeted dyke complex. The dip of pyroxenite and gabbronorite dykes can be particularly shallow. It is common to observe, at the scale of an outcrop or of an area, two preferred conjugate orientations, showing that these dykes were not emplaced in an extensional stress field related to spreading.

[36] On the map of Figure 19, it seems that regions occupied by the gabbronorite-pyroxenite suite and the troctolite-olivine gabbro suite overlap in many places, but in the field it is exceptional to observe cross-cutting relationships between dykes having different lithologies. When cross-cutting relationships are observed, they generally occur in a shear zone which does not allow us to determine a relative age among the two series.

4. Discussion

4.1. Parent Melts of Oman Mantle Cumulates

[37] A major result of our survey is that mantle dykes of Oman belong to two main groups or “suites” that are contrasted in terms of all characteristics: field relations and textures, modal composition and order of crystallization, and chemical composition. This implies that the parent melt(s) of these two cumulate suites had contrasted compositions but also that the physical conditions during emplacement were very different.

4.1.1. “MORB” Suite

[38] In terms of modal and chemical composition and of order of crystallization, the troctolite and olivine gabbro families present the characteristics predicted for cumulates formed by fractional crystallization of a primitive MORB [e.g., Grove et al., 1992]: troctolites correspond to cumulates which crystallized from the most primitive known mid-ocean ridge basalts (mgn around 70 [e.g., Presnall and Hoover, 1987]): the most frequent modal proportions (25% olivine and 75% plagioclase) are those determined for the olivine + plagioclase cotectic of common MORBs (olivine tholeiites); the fact that plagioclase seems to be the (slightly) earliest crystallizing phase implies low crystallization pressures (≪0.5 MPa) and a relatively Al-rich parent melt. Experiments on low-pressure fractional crystallization of MORBs predict that clinopyroxene joins plagioclase and olivine in the crystallization sequence after a moderate (<25%) degree of fractional crystallization, which is consistent with the gradational relationship between troctolites and olivine gabbros and with the fact that, on average, olivine gabbros have a lower Fo content than troctolites, and Cpx-troctolites have intermediate values; here also the 3-phase cotectic proportions determined for low-pressure crystallization of MORBs (50% Plg + 15% Olv + 35% Cpx [Grove et al., 1992]) are consistent with the more frequent modal content of olivine gabbro dykes.

[39] Troctolites show coarse grained mosaic and coronitic textures requiring efficient diffusion and implying that they were emplaced in a mantle that remained sufficiently hot (very close to magmatic temperatures, i.e., about 1200°C) during a “relatively long” period of time. Accordingly, we propose that the troctolites allow us to map former asthenospheric windows, i.e., places where hot mantle made its way through a lithospheric lid up to, or at least close to, the mantle/crust boundary. Olivine gabbro dykes were emplaced in still very hot mantle at the immediate vicinity of these asthenospheric windows. The parallelism of olivine gabbro dykes with the sheeted dyke complex, and the elongation of the zones of mantle troctolites and olivine gabbros in this same direction, suggest that asthenospheric upwelling occurred in an extensional, i.e., spreading, setting.

[40] Gabbroic cumulates sampled along present-day ocean ridges present generally well defined differentiation trends in a given location, although these trends can be quite variable from one location to the other. In Figure 21 we have compared the variations of An and TiO2 contents according to mgn of Oman mantle dykes to the same variations in several occurrences of oceanic gabbros. It appears that the variation fields of Oman mantle dykes from the MORB suite encompass the field of oceanic gabbros in terms of variations of An and TiO2 for a given mgn, although oceanic gabbros are, on average, more evolved in terms of mgn. That points to the extreme variability of Oman mantle dykes in terms of incompatible element contents. The broad variations of An can result from variations of the Na/Ca ratio of the parental melt, pressure, or variations of parameters like the water content of the melt.

Figure 21.

Comparison between Oman mantle dykes and the oceanic crust for some geochemical characteristics. Global compositional domains are shown for each family of Oman dykes while oceanic gabbros are plotted by geographic area. (a) anorthite content in plagioclases versus mgn in clinopyroxenes; (b) TiO2 versus mgn in clinopyroxenes.

[41] The similarity between melts in equilibrium with the troctolite-olivine gabbro mantle cumulates and actual MORBs is confirmed in Figure 22b for TiO2 and mgn. Of course, in spite of common characteristics, these melts still present wide geochemical variations, not necessarily related to a simple fractional crystallization process (i.e., not strictly correlated to mgn). Compatible major and minor element trends (Fo and Ni in olivine, mgn and Cr in pyroxenes) inside each family and from one family to the other are broadly consistent with fractional crystallization as a major process accounting for the geochemical variability of the Oman mantle dykes. However, it is worth noting that a significant variation in Ni and Cr contents is observed, even in the most primitive end-members (in terms of Fo or mgn), implying heterogeneity of parental melts; this variability might result from melt rock interaction.

Figure 22.

Calculated liquids in equilibrium with clinopyroxenes and plagioclases. Equilibrium mgn are calculated using a KDFe-Mg value of 0.29 determined from experiments on MORBs [Grove et al., 1992]; equilibrium. (a) histogram showing the mgn variations for each main family of dykes; (b) evolution of TiO2 of equilibrium liquid (Kd = 0.47 [Putirka, 1999]) with mgn, see Figure 2 for legend of symbols; the yellow area is the MORBs field.

[42] The fact that the correlation between mineral composition and mineral content is not perfect can be interpreted either as the result of heterogeneity in the major element composition of the parent melts or as an indication that cumulates do not result from pure fractional crystallization. For example, the fact that the crystallization of interstitial orthopyroxene is not restricted to the more evolved olivine gabbros may result from heterogeneity in Si/Mg ratio of the parent melt or from the fact that some rocks are not pure adcumulates (some may include trapped melt). The excellent correlation between Fo (Ol) and mgn (Opx and Cpx) combined with frequent adcumulate textures leads us to favor the hypothesis of heterogeneity of the parent melt composition. An answer to this question requires trace element studies. Similarly, although most of the olivine gabbros have more evolved phase compositions than the troctolites, some olivine gabbros are more primitive in terms of Mg/Fe ratio than some troctolites; these cumulates cannot, strictly speaking, be derivative from exactly the same melt. Accordingly, even if we can speak of a MORB-like parent melt, it remains apparent that in each mantle dyke or porous flow channel, a liquid has circulated that was not strictly equivalent to the one that circulated in its neighborhood.

4.1.2. “Depleted” Suite

[43] The parent melt of our second cumulate suite can be characterized as follows: it was richer in SiO2 than standard MORBs, closer to quartz tholeiites or andesites (SiO2 in the range of 52% to 55% is likely a good estimate) in order to account for early and abundant crystallization of orthopyroxene; this melt was depleted in incompatible elements like Na and Ti (and also in incompatible trace elements like the LREE [Kelemen et al., 1997a; Benoit et al., 1999]), depleted peridotites left after MORB extraction are the best candidate for the source of these melts. The very low Al2O3 content of both clino- and orthopyroxenes in pyroxenites from this suite allows us to rule out a popular previous interpretation of pyroxenites as high-pressure cumulates from a MORB-like melt, equivalent to the parent melt of the olivine gabbros [see Lippard et al., 1986]. The case of pyroxenite layering is more ambiguous. We did not sample systematically pyroxenitic layering but the 19 samples that we studied are, on average, richer in Al2O3 than pyroxenes in dykes (see Figure 15). Given their early crystallization (before the accretion of the mantle to the lithosphere, as they are transposed by solid state deformation, whether they are ancient melt migration features or not), an origin at higher pressure than the discordant dykes cannot be ruled out.

[44] Contrasting with the high-temperature context of emplacement of mantle cumulates from the MORB suite, it seems that the mantle was much cooler than the melt at the time of crystallization of the depleted dykes; above 400°C (frozen melts and/or dykes with chilled margins are extremely uncommon) but below solidus temperatures as shown by evidence for rapid crystal growth on the dyke walls (pegmatites, etc.). Fluids also played an important role in the development of these textures, as attested by the ubiquitous association of amphiboles with the last stage of crystallization of the pegmatites (amphibole rims epitactic on pyroxene crystals). When these dykes are affected by mylonitic deformation, which is rather common, this occurred along shear zones of variable thickness (a few millimeters to several decimeters in the case of shear zones that developed inside the dykes themselves, reaching hundreds of meters where the shear zones affect both the dykes and their host peridotites); grain size reduction occurred without significant change in mineral assemblage and chemistry, i.e., at high-grade conditions. Although there is no experimental or theoretical data to quantify precisely the temperature of the wall rock during crystallization and deformation of these pegmatites, temperatures between 600°C and 1100°C are probably good estimates. On the other hand, these structures are not transposed by high-temperature plastic flow and were emplaced not far from the Moho. Accordingly, we infer that the depleted suite was generated and emplaced in relatively cool mantle lithosphere, partly altered by hydrothermal fluids [cf. Benoit et al., 1999]. The diopsidite skarns we found from place to place in the harzburgites and developing at the expense of serpentinized harzburgites are the footprint of very high temperature fluid penetration in the shallow mantle and support this scenario.

[45] The comparison between the inferred parent melts of our depleted suite and magma types erupted at the Earth's surface is less straightforward than in the case of the MORB suite. At least can we demonstrate that the parent melt was very different from MORBs. The contrast between liquids in equilibrium with our depleted cumulates and present-day MORBs is clearly illustrated in Figure 22b. Similarly, Oman dykes from the depleted suite differ from oceanic gabbros (Figure 21) in terms of differentiation trends in mgn-An and mgn-TiO2 diagrams, apart from the gabbronoritic cumulates of ODP Site 334 that are closer to Oman depleted cumulates than to cumulates from the MORB suite, particularly in terms of TiO2 content. Most authors who have worked on Oman lavas distinguished 3 main types of lavas [e.g., Alabaster et al., 1982; Lippard et al., 1986; Pflumio, 1991]. The Lasail or V2 unit is derived from a particularly depleted magma, so it could be related to our depleted suite. But that hypothesis does not explain why corresponding cumulates are almost absent in the crustal section (see Figure 23). In Figure 24, we have compared the composition of Oman lavas to the two dyke suites and MORBs. There is a large dispersion in basalt compositions but we can see a global trend that is between tholeiitic and depleted mantle suites, roughly similar to the MORB trend apart from being generally more evolved (lower mgn). A boninitic kindred has been proposed for pyroxenite dykes from other ophiolitic massifs [e.g., Varfalvy et al., 1996]. This seems unlikely in the case of Oman because olivine is particularly rare in pyroxenite dykes while olivine is very common in boninites. We will return to the possible origin of these cumulates below.

Figure 23.

Comparison between mantle and crustal cumulates of Oman; representation for mantle dykes is the same as in Figure 21: (a) anorthite content in plagioclases versus mgn in clinopyroxenes; (b) TiO2 versus mgn in clinopyroxenes.

Figure 24.

Comparison between the chemistry of lavas from Oman and mantle dykes on the one hand and MORB on the other hand. Blue zone is the chemical domain of the troctolite-olivine gabbro suite; green zone is the chemical domain of the pyroxenite-gabbronorite suite.

4.2. Focused MORB Delivery in the Crustal Section

[46] Most of the crustal section of the Oman ophiolite is made of cumulates formed by fractional crystallization of MORB-like melts [e.g., Pallister and Hopson, 1981; Browning, 1982; Ernewein et al., 1988; Juteau et al., 1988; Korenaga and Kelemen, 1997; MacLeod and Yaouancq, 2000]. Exceptions concern (1) some rare outcrops where gabbronoritic and/or websteritic cumulates interlayered with olivine gabbros have been reported (Smewing [1981], in the northern part of the ophiolite; Lachize et al. [1996], in the central part of the ophiolite, wadi Haylayn area), (2) the gabbronorite-tonalite intrusions rooting in the mantle section and invading the crustal section at the periphery of the Maqsad diapir [Amri et al., 1996; Benoit et al., 1999] and (3) wehrlite - melano-troctolite intrusions, ubiquitous in Oman [e.g., Ernewein et al., 1988; Reuber, 1988; Koga et al., 2001]. The nature and origin of the magmas parental to the wehrlitic intrusions is still debated; it seems that they can derive from a depleted mantle source [Ernewein et al., 1988], although this conclusion has been questioned recently [Koga et al., 2001]; unlike our depleted suite of mantle cumulates, they did not crystallize from an SiO2-rich melt. We have confirmed in the present study that these intrusions have no counterpart in mantle dykes. The relation between the gabbronorite-tonalite intrusions of Maqsad and the depleted suite of mantle cumulates has, however, been clearly established; although abundant at the periphery of the Maqsad area, such intrusions seem rather uncommon at the scale of the ophiolite. The case of gabbronorites interlayered with olivine gabbros is more puzzling. We have compared the mineral compositions of Lachize et al. [1996] gabbronorites to those of our mantle dykes and it appears that they are closer to the MORB suite than to the depleted suite in terms of An and Ti (cpx) contents (Figure 23). An content of gabbronorites studied by Smewing [1981] appear closer to the ones of our depleted suite; no Ti contents of associated clinopyroxenes have been published, unfortunately.

[47] Although a more systematic survey of the Oman crustal section looks necessary to establish the real proportion between crustal cumulates issued from MORB-like melts and those issued from other kinds of melts, it remains that on mgn-An and mgn-TiO2 variation diagrams (Figure 23) there is no marked difference between our mantle cumulates from the MORB suite and crustal cumulates for which mineral analyses are available in the literature, the main difference being that the degree of differentiation is, on average, more pronounced in the case of crustal cumulates. Accordingly dykes of the MORB suite were expected to be the more abundant in the Oman mantle section but the present survey has shown that it was not the case. Mantle dykes of the MORB suite occur in a few specific areas; globally, we estimate that MORB dykes crop out in only a quarter of the mantle section presently exposed in Oman. This paradox may concern large (massif wide) scales. For example, in the well-studied Haylayn block (about 60 km in length), we have found that the mantle section is devoid of dykes from the MORB suite while the overlying crustal section is made essentially of MORB cumulates [Browning, 1982; Lachize et al., 1996].

[48] There are two contrasted ways to account for this puzzling observation:

[49] 1. MORB delivery into the Oman crust is highly focused, i.e., there is a ratio of about equation image between the surface of the “asthenospheric windows” and the surface of the crust fed by the melts that circulated through these windows, as far as we can estimate from the outcrops presently exposed in the ophiolite.

[50] 2. Mantle troctolites and gabbros do allow us to map zones where MORB-like melts did crystallize, i.e., where they migrated in a context of decreasing temperature, that can be quite different in size, shape and distribution from the zones where these melts did circulate.

[51] The first interpretation is implicitly based on the hypothesis that the circulation of melts at shallow mantle depths can be tracked through the distribution of their crystallization products. Strictly speaking, the areas where we presently observe dykes in the mantle section are zones where the mantle cooled down sufficiently to trigger fractional crystallization of the melts migrating to the crust. In this respect, dykes are, by definition, the witnesses of a “dying” system. Accordingly, the structures observed in Oman could reflect very peculiar processes happening during the latest stage of spreading and during the initiation of the oceanic detachment leading eventually to the obduction. Temperature decrease in the shallow mantle could also be a natural consequence of the episodic character of the magma supply processes at “healthy” oceanic spreading centers. As a matter of fact, gabbroic and pyroxenitic dykes are currently observed in abyssal peridotites exposed along present-day ocean ridges [e.g., Cannat et al., 1997]. As a conclusion, we have established that the zones of MORB-like melts crystallization in the Oman mantle section were rather restricted; this could reflect focusing of MORB circulation in the mantle but it cannot be excluded that this distribution is related to the peculiar tectonic setting during the genesis of the Oman crust even if this hypothesis could hardly be tested.

[52] The second interpretation supposes that the troctolite and gabbro dykes are late or off-axis features and do not reflect the distribution of former melt channels that fed the bulk of the Oman crust. Tracks of these high-temperature channels should rather be looked for in the distribution of the dunitic lithologies currently observed in mantle harzburgites [e.g., Kelemen et al., 1995]. It will be possible to test this hypothesis in the future, but in the present state of our knowledge it remains an open question, first because no global map showing the distribution of dunites in Oman harzburgite has been published up to now and second because it is not clear which proportion of the Oman dunites result from reaction of the harzburgites with MORB-like melts. As a matter of fact, where associated to crystallization products, dunites have been observed at the wall of both troctolites (i.e., reaction with former MORBs) and pyroxenites (i.e., reaction with depleted melts).

4.2.1. Mantle Diapirs

[53] Maps of high-temperature plastic deformation structures of mantle peridotites have revealed several areas where the fossil mantle flow lines are at a high angle to the mantle/crust boundary, i.e., to an inferred paleo-horizontal surface [e.g., Boudier and Coleman, 1981; Rabinowicz et al., 1987; Ceuleneer et al., 1988; Nicolas et al., 1988, 1989, 2000a; Ceuleneer, 1991; Jousselin et al., 1998]. These areas have been interpreted as fossil asthenospheric diapirs but are documented with variable accuracy. Moreover, the interpretation of the orientation of plastic deformation structures in terms of paleo-flow directions is uncertain. In the case of Maqsad, we have seen that the correlation between zones of near-vertical lineation and the occurrence of high-temperature cumulates in mantle harzburgites (troctolites) supports the view that a hot mantle upwelling reached Moho level there [Ceuleneer et al., 1996; this study]. In this context, it is interesting to determine whether this correlation holds also for the many other diapirs (10 of them, according to Nicolas et al. [2000a]) defined on the basis of structural data.

[54] Figure 25 shows the location of these inferred diapirs superimposed on our map of mantle dyke lithologies. It appears that 4 diapirs crop out in the Maqsad corridor. Apart from the Maqsad diapir itself, one has been located at the northwestern termination of this corridor (Nakhl area) and the two other ones are very close-spaced (<10 km) at its southeastern termination (Samad area). This observation can be accounted for in different ways. If the interpretation of the small Nakhl and Samad diapirs as former narrow, more or less cylindrical, upwelling zones is correct, it means that the mantle flow can be segmented in an irregular way. Another explanation is that vertical orientation of mantle flow lines in the Nakhl and Samad areas is unrelated to upwelling, but instead formed as complex deformation features related to the solid intrusion of asthenospheric material into a lithospheric lid. Our survey has shown that the only large area in harzburgites where troctolites constitute the only cumulate lithology is centered on Maqsad. Away from Maqsad, a continuous decrease in the abundance of troctolites and increase in the abundance of olivine gabbros is observed, together with a narrowing of the “MORB” corridor itself. Neither Nakhl nor Samad diapirs show an increased abundance in troctolites. This lends us to favor the hypothesis of a single elongated mantle upwelling parallel to the spreading axis, more vigorous in its central part (Maqsad) than in its distal parts.

Figure 25.

Simplified map of the Oman ophiolite showing the main zones for each dyke family and structural diapirs. Diapir locations and inferred ridge axes are from Nicolas et al. [2000a].

[55] The only other case where a diapir coincides with a zone of MORB cumulates is the one of the Musibit massif (Mansah area of Jousselin and Nicolas [2000]). There, we have found a few troctolites in a zone where olivine gabbros constitute the dominant lithology. Jousselin and Nicolas [2000] have mapped complex mantle flow patterns there that they interpret as the footprint of an “off-axis” diapir. As illustrated in Figure 26, our data on mantle dykes is consistent with the existence there of a former mantle upwelling that was less “successful” than the Maqsad diapir, i.e., that was frozen at greater depth, before reaching Moho level.

Figure 26.

Interpretative sketch based on a cross section trough Maqsad and Musibit areas: see text for comments.

[56] Other structurally defined diapirs, in the Miskin, Bahla and Haylayn and East Wadi Tayin massifs do not have any corresponding “hot” dyke region. Some of them crop out in massifs where all the mantle dykes belong to the depleted family. The case of the Wuqbah diapir is more ambiguous since we have not yet surveyed this area systematically.

[57] Finally, the small troctolite zone centered on the wadi Hilti area does not correspond to any mapped diapir. The wadi Hilti area may correspond to the tip of a troctolite zone that is presently hidden below crustal outcrops. It may also result from a situation similar to the one invoked in the case of the Musibit area, i.e., a mantle upwelling that did not reach Moho level.

4.2.2. Mantle Dyke Orientations

[58] Our observations on mantle dyke orientations can be summarized as follows (Figure 20 and auxiliary material Figure A1) [see also Ceuleneer et al. [1996]): troctolites show a highly variable orientation in both strike and dip. Olivine gabbro dykes are generally sub-vertical with a strike parallel to the sheeted dyke complex, supporting the view that they were injected in a stress field dominated by the regional extension, although many dykes are oriented at high angle to this azimuth.

[59] The case of gabbronorites and pyroxenites is even more complex. In some areas (Sarami and Haylayn massifs), these dykes show a rather constant azimuth but a variable dip. In other massifs, the orientation of pyroxenite and gabbronorite dykes is variable in both strike and dip. The orientation pattern of gabbronorite and pyroxenites dykes is close to random, evoking hydraulic fracture, i.e., a local stress field dominated by overpressure of melts and little influenced by the regional tectonic stress field.

[60] Given these observations, to map dyke trajectories [Nicolas et al., 2000b] is difficult and potentially misleading. Accordingly, the regular orientation of dyke trajectory strikes in the Haylayn and Sarami massif poorly reflects the variability of dyke dip in these massifs. To ignore the lithological nature of the dykes leads also to misinterpretations, for example when two different events are interpreted as a single event with a rotating stress field. This situation is well illustrated in the southern part of the Nakhl area, where gabbronorite dykes are almost orthogonal to olivine gabbro dykes.

4.3. Tectonic Setting of the Oman Ophiolite: The Mantle Dykes Perspective

[61] The tectonic setting of the Oman ophiolite (back-arc versus mid-ocean ridge) has been debated for more than 30 years [e.g., Welland and Mitchell, 1977; Coleman, 1981; Alabaster et al., 1982; Nicolas, 1989], but there is still no consensus. Mantle dykes alone will certainly not bring a definite answer to this question but at least can we add some new pieces of evidence.

[62] We have seen that two contrasted magma suites crystallized in the mantle harzburgites of Oman. One, with “andesitic” characteristics is almost ubiquitous and formed from a depleted source, likely by hydrous melting at shallow depth. The other one has a parental melt that has characteristics similar to primitive mid-ocean ridge basalts but occurs only in a few areas of Oman (less than 25% of the presently exposed surface of the mantle section).

[63] We might be tempted to use this evidence to support the hypothesis of a back-arc setting for the Oman ophiolite. In that context, the area where the MORB suite crops out could be interpreted as evidence for opening of small marginal basins where petrogenetic processes and the mantle source (partial melting of spinel lherzolites not or little contaminated by fluids) were similar to those beneath mid-ocean ridges.

[64] Things are not so simple, unfortunately. The main problem is that cumulates from the depleted suite are rarely present in the Oman crust itself, apart from a few occurrences reported by Smewing [1981] and Amri et al. [1996], in the plutonic section, and, in the extrusive section, the so-called Cpx-phyric and Alley volcanics [Alabaster et al., 1982], attributed to a late (V2) magmatic event [Ernewein et al., 1988]. It would be surprising that, in a back arc spreading context, the major magma type cross cutting the mantle section has no major counterpart in the crust. On the other hand, although they are uncommon, gabbronoritic cumulates similar to cumulates from our depleted suite have been described in an ocean ridge setting (ODP Site 334, off the Mid-Atlantic Ridge [e.g., Ross and Elthon, 1993]). Highly depleted, orthopyroxene-saturated liquids are also known as inclusions in phenocrysts of mid-ocean ridge basalts [e.g., Sobolev and Shimizu, 1993]. Thus liquids with highly depleted incompatible element compositions and higher SiO2 content than MORB do form in a mid ocean ridge context, even if they are never erupted, as far as we know.

[65] Another observation relevant to the possible origin of the depleted suite is that the occurrence of Oman depleted cumulate dykes may be related to the occurrence of mantle cumulates from the MORB suite: pyroxenite and gabbronorite pegmatites are more abundant at the immediate periphery of “asthenospheric windows” defined by the occurrence of troctolites and olivine gabbros. This is very spectacular around the Maqsad area where the peridotites are invaded by giant intrusions of depleted cumulates [see also Amri et al., 1996; Benoit et al., 1999].

[66] Accordingly we propose the following scenario for the genesis and emplacement of mantle cumulates in Oman. Hot mantle upwellings episodically made their way through an axial lithospheric lid, residual after MORB extraction at depth greater than presently exposed in Oman (a few kilometers below the Moho). Temperature increase in these hydrated and depleted peridotites triggered their remelting, causing the formation of SiO2-rich melts. The sudden overpressure related to the formation of melts of lower density than the country rocks induced fracturation of the peridotites along more or less randomly oriented cracks, independent of the extension direction related to spreading. Due to the steep temperature gradient in the shallow lithosphere, likely maintained by active hydrothermal circulation (suggested by the presence of lithologies like hydrothermal diopsidites) these melts crystallized as swarms of pegmatites before reaching the crust. The arrival of the asthenospheric upwelling at (or close to) Moho level delivered large quantities of MORB-like melt into the crust. This scenario could apply as well to mid-ocean ridge setting or to marginal basins related to subduction.

5. Conclusion

[67] Previous classification schemes of Oman mantle dykes mainly stressed the contrast between pyroxenites and gabbros, because these lithologies are easily distinguished in the field. It was proposed that the pyroxenites are primitive cumulates derived from the same primary melts as gabbros. We have shown that, in terms of petrogenesis, there is an important distinction between cumulates with olivine and plagioclase as early crystallizing phases (troctolites, olivine gabbros and olivine gabbronorites) and cumulates with pyroxenes as early crystallizing phases (pyroxenites, gabbronorites-norites and diorites-tonalites-trondhjemites). The contrast is due to distinctly different parental melts (including fluid contents) and mantle source compositions. These cumulates define two main magma suites, a first one with MORB characteristics, a second one with characteristics closer to depleted andesites. The fact that mantle dykes from the andesitic kindred are the more widespread in Oman can be used to favor a back-arc setting for the ophiolite, but this can also be explained in the context of episodic diapirism inducing hydrated remelting of the shallow lithosphere in a mid-ocean ridge setting. Delivery of MORB like melts into the Oman crust may have been focused in some well-defined areas.


[68] We are very indebted to Philippe de Parseval from the microprobe service of the Toulouse University for his help during data acquisition, to Anne-Marie Roquet, Fabienne de Parseval and Raphael Peyron who made the polished thin sections, to Camille Ceuleneer and Marc Monnereau for their precious help in the field, to Marc Monnereau, Michel Rabinowicz, and Henry Dick for stimulating discussions, and to Hilal Al Azri, from the Oman Geological Survey, who made possible field work in Oman. We thank two anonymous reviewers, the Associate Editor (Peter Kelemen) and Editor (Bill White) for their constructive comments. Financial support for this work was provided by the Centre National de la Recherche Scientifique in the frame of the programs “Dorsale” and “Dynamique et Bilan de la Terre.”