Journal of Petrology Advance Access originally published online on September 3, 2004
Journal of Petrology 2004 45(10):2011-2044; doi:10.1093/petrology/egh046
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Journal of Petrology 45(10) © Oxford University Press 2004; all rights reserved
Layered Lithospheric Mantle Beneath the Ontong Java Plateau: Implications from Xenoliths in Alnöite, Malaita, Solomon Islands
DEPARTMENT OF EARTH AND PLANETARY SCIENCES, TOKYO INSTITUTE OF TECHNOLOGY, TOKYO 152-8551, JAPAN
RECEIVED APRIL 15, 2003; ACCEPTED MAY 3, 2004
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONSAMPLE DESCRIPTIONMAJOR ELEMENT MINERAL CHEMISTRY...DISCUSSIONSUPPLEMENTARY DATAREFERENCES |
|---|
A varied suite of mantle xenoliths from Malaita, Solomon Islands, was investigated to constrain the evolution of the mantle beneath the Ontong Java Plateau. Comprehensive petrological and thermobarometric studies make it possible to identify the dominant processes that produced the compositional diversity and to reconstruct the lithospheric stratigraphy in the context of a paleogeotherm. P–T estimates show that both peridotites and pyroxenites can be assigned to a shallower or deeper origin, separated by a garnet-poor zone of 10 km between 90 and 100 km. This zone is dominated by refractory spinel harzburgites (Fo91–92), indicating the occurrence of an intra-lithospheric depleted zone. Shallower mantle (
Moho to 95 km) is composed of variably metasomatized peridotite with subordinate pyroxenite derived from metacumulates. Deeper mantle (95–120 km) is represented by pyroxenite and variably depleted peridotites that are unevenly distributed; the least-depleted garnet lherzolite (Fo90–91) lies just below the garnet-poor depleted zone (100–110 km), whereas the presence of pyroxenite is restricted to the deepest region (110–120 km), together with relatively Fe-enriched garnet lherzolite (Fo87–88). This depth-related variation (including the depleted zone) can be explained by assuming that the degree of melting for a basalt–peridotite hybrid source was systematically different at each level of arrival depth within a single adiabatically ascending mantle plume: (1) the depleted zone at the top of the mantle plume, where garnet was totally consumed in the residual solid; (2) an intermediate part of the plume dominated by the least-depleted garnet lherzolite just above the depth of the peridotite solidus; (3) the deepest pyroxenite-rich zone, whose petrochemical variation is best explained by the interaction between peridotite and normative quartz-rich basaltic melt, below the solidus of peridotite and liquidus of basalt. We explain the obvious lack of pyroxenites at shallower depths as the effective extraction of hybrid melt from completely molten basalt through the partially molten ambient peridotite, which caused the voluminous eruption of the Ontong Java Plateau basalts. From these interpretations, we conclude that the lithosphere forms a genetically unrelated two-layered structure, comprising shallower oceanic lithosphere and deeper impinged plume material, which involved a recycled basaltic component, now present as a pyroxenitic heterogeneity. This interpretation for the present lithospheric structure may explain the seismically anomalous root beneath the Ontong Java Plateau. KEY WORDS: mantle xenolith; Ontong Java Plateau; peridotite; pyroxenite
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLE DESCRIPTIONMAJOR ELEMENT MINERAL CHEMISTRY...DISCUSSIONSUPPLEMENTARY DATAREFERENCES |
|---|
The island of Malaita, Solomon Islands, has been recognized as a part of the obducted southwestern margin of the Ontong Java Plateau (OJP), which is the world's largest oceanic plateau or large igneous province (LIP), occupying an area of at least 1600 km x 800 km (e.g. Coleman & Kroenke, 1981
; Tejada et al., 1996; Petterson et al., 1997) (Fig. 1a). The OJP was derived from episodic yet voluminous volcanic activity during the Cretaceous, which has been referred to as the ‘Pacific Superplume’ event (Larson, 1991). The average crustal thickness of the OJP is 33 km (Richardson et al., 2000) and the estimated total crustal volume is about 5·0 x 107 km3 (Schubert & Sandwell, 1989)—nearly twice as much as the modern total volumetric output of the Earth's sea-floor spreading for one million years (1·8 x 107 km3/Myr). Because such massive volcanism should have been caused by major dynamics of the Earth's mantle and should have had a significant influence on the surface environment at the time, numerous investigations utilizing geological, geochemical and geophysical methods have endeavored to understand the origin of the OJP (e.g. Larson, 1991; Coffin & Eldholm, 1994; Neal et al., 1997; Mahoney et al., 2001).
|
Malaita also represents a rare and famous locality for deep-seated garnet peridotite xenoliths from an oceanic environment (e.g. Nixon & Boyd, 1979
; Neal, 1985). Since the discovery of boulders of ultramafic rocks in streams from Malaita (Rickwood, 1957; Allen & Deans, 1965), various mantle xenoliths derived from the Malaitan alnöite have been investigated by many workers, with the aim of constraining the composition and evolution of the mantle beneath the OJP (e.g. Nixon & Coleman, 1978; Nixon & Boyd, 1979; Bielski-Zyskind et al., 1984; Neal, 1985, 1988; Neal & Nixon, 1985; Nixon & Neal, 1987; Neal & Davidson, 1989; Neal & Taylor, 1989; Collerson et al., 2000; Neal et al., 2001). Neal's (1985) and subsequent detailed studies have made several important statements concerning the xenoliths. However, the relationship between the diverse nature of the xenoliths and the unusual generation of the OJP is still ambiguous, because of insufficient petrochemical and geophysical knowledge of the xenoliths and the OJP itself. Therefore, Malaitan xenoliths are an important resource for added constraints on not only the origin of the OJP, but also the evolution of plume-related mantle in general. This study presents: (1) a discussion of the origin of the lithological diversity of the xenoliths from Malaita based on petrographic observations of newly collected samples; (2) a reconstruction of the lithospheric mantle stratigraphy on the basis of the equilibrated P–T estimates for the individual xenoliths; (3) a model for the formation and evolution model of the sub-OJP mantle and the OJP itself.
|
SAMPLE DESCRIPTION |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLE DESCRIPTIONMAJOR ELEMENT MINERAL CHEMISTRY...DISCUSSIONSUPPLEMENTARY DATAREFERENCES |
|---|
Sample location
The basement of Malaita mainly comprises Ontong Java Plateau basalts, erupted in the Early Cretaceous (Malaita Volcanic Group), and overlying sedimentary rock units (Hughes & Turner, 1977
; Petterson, 1995; Tejada et al., 1996; Petterson et al., 1997) (Fig. 1b). These units are overlain by Eocene-age alkaline basalts, which were recently named the Maramasike Formation (Petterson et al., 1997). Several pipe-like bodies and sills of alnöite are located in north–central Malaita (Nixon & Coleman, 1978; Nixon & Boyd, 1979; Neal & Nixon, 1985). The alnöite has an age of 34 Ma, given by U–Pb dating of megacrystalline zircon extracted from sand samples, and contains a varied suite of mantle-derived xenoliths (Davis, 1977).
According to the recent tectonic reconstruction proposed by Petterson et al. (1999), the collision between the OJP and Solomon Islands arc commenced during the Oligocene (25–20 Ma), as shown by ceasing arc volcanism as the OJP blocked the subduction zone (cf. Coleman & Kroenke, 1981). Subsequently, a subduction flip from the SW to NE occurred at about 12 Ma and the emergence of Malaita above sea level was achieved by the formation of an extensive fold belt at 5 Ma. This geologic scenario clearly indicates that the eruption of the host alnöite occurred in an intra-plate setting and that the xenoliths were never affected by the presently subducting slab beneath Malaita. Unlike previous studies, detailed information on individual sample locations is not available because the xenoliths were usually found in the dense rain forests and river deposits, away from known localities of host alnöite.
Petrography
The xenoliths have rounded and weathered surfaces, and are generally smaller than 10 cm in diameter. However, specimens with a long axis of 30 cm are occasionally present. Previous studies classified Malaitan xenoliths into ultrabasic and discrete nodule (megacryst) suites (Nixon & Boyd, 1979). Based on the observation of a large and comprehensive suite of more than 200 newly collected samples, we describe a wide variety of rock types that were unreported in previous studies of the xenoliths from Malaita. In particular, there is a diverse suite of garnet-bearing pyroxenites, which have been rarely described previously. Therefore, we prefer to subdivide the samples into three rock suites: peridotite, pyroxenite and megacryst. Below, we will further distinguish these suites into two types: high-temperature type (HT-type) and low-temperature type (LT-type), based on their equilibrium conditions.
Peridotite suites
The highest proportion of the xenolith population of samples is classified as peridotites. Generally, the peridotite samples display severe alteration, represented by development of serpentine and carbonate replacing olivine, but fresh olivine is preserved in most specimens. Mineral assemblages and modal proportions were carefully determined using both thin sections and mineral separates. However, because of the alteration, small sample size and coarse-grained constituent minerals, some samples cannot preserve true and reasonable modal compositions.
Many of the peridotite xenoliths are garnet-free spinel peridotite. We collected more than 100 spinel peridotite xenoliths and 57 of them are included in this study. All specimens contain olivine, orthopyroxene, clinopyroxene and spinel, and are subdivided into spinel lherzolite (SL) and spinel harzburgite (SH). For convenience, SH, as defined here, has modal clinopyroxene <4% and/or Cr-number of spinel [= Cr/(Cr + Al)] >0·3 because the two values are strongly correlated (Table 1). Both spinel peridotites display a coarse granular texture and constituent silicate minerals (1–5 mm) have slightly curved or straight grain boundaries (Fig. 2a). Only one harzburgite (SAS1) shows porphyroclastic texture, with small olivine neoblasts as small as 0·1 mm (Fig. 2d). Abundance, size, shape and color of spinel are variable and are different between those in SL and SH. Transparent brown spinels in SL are larger and more abundant than those in SH. The spinel usually occurs as anhedral grains, and some examples contain a vermicular texture of spinel embedded in both pyroxenes (Fig. 2b). Rarely, spinel occurs as platelets, together with orthopyroxene exsolution lamellae within clinopyroxene. In contrast to the SL samples, spinel in SH samples is present in only trace amounts of small grains that are dark brown to black in color and are comparatively much less abundant. Amphibole is a common mineral in both SL (<5%) and SH (<17%), occurring as a replacement after pyroxenes and/or forming mantles around spinel similar to those described by Neal (1988) (Fig. 2c).
|
|
Garnet-bearing peridotite is relatively rare and comprises only 37 samples, including small chips (<2 cm). Garnet-bearing peridotite can be subdivided into two groups: garnet–spinel lherzolite (GSL) and garnet lherzolite (GL). Because of the low abundance of spinel (<3%), this classification may not be appropriate for samples of small size. However, the presence of spinel is closely related to the morphology of garnet. Rounded, small grains or an irregular shape of spinel usually appear within anhedral garnet, which is associated with olivine and both pyroxenes (Fig. 2e). In some cases, large amphibole perfectly surrounds the garnet including spinel (Fig. 2f). This textural evidence suggests that garnet was formed by a reaction between spinel and pyroxene and/or involving metasomatic melt–fluid from precursor spinel peridotite (Neal & Nixon, 1985
; Neal, 1988). The GSL showing the above texture comprises the 26 spinel-bearing garnet lherzolites and one of 11 spinel-free garnet lherzolites (sample SAG26), which may result from an advanced stage of the garnet-forming reaction. The modal proportions of GSL are highly variable (Table 1), and some low-clinopyroxene samples tend to be rich in amphibole (18%), indicating that replacement of clinopyroxene is also significant. On the other hand, GL is characterized by the presence of well-rounded garnet, without evidence of subsolidus reaction (Fig. 2g). Furthermore, texturally equilibrated amphibole is absent. GL samples show uniform modal proportions of primary minerals relative to those of GSL (Table 1). Most of the GL have coarse-grained textures, but two of them (SAG21 and SAG27) display porphyroclastic textures with large porphyroclasts surrounded by small olivine neoblasts (Fig. 2h).
Pyroxenite suites
The Malaitan alnöite also contains xenoliths that are composed mainly of pyroxenes and garnet. The xenoliths have no omphacitic clinopyroxene and are different from eclogite in kimberlite or metamorphic terrains. Therefore, they should be referred to as garnet pyroxenite and not as eclogite. Although Nixon & Boyd (1979) reported the presence of garnet-free pyroxenite, such as spinel wehrlite and spinel pyroxenite, all of the pyroxenites in this study contain garnet. The garnet pyroxenites show a broad spectrum of mineral assemblages, modal compositions and textures, which were used for the classification.
Garnet websterites (GW) are mainly composed of coarse-grained clinopyroxene (up to 3 cm) and smaller grains of orthopyroxene and garnet. The GW are dominated by exsolution textures; most of the garnet and orthopyroxene occur as small blebs and plates within large clinopyroxene and/or as anhedral crystals along the grain boundary of clinopyroxene, as described in many previous studies of mantle eclogites and pyroxenites (e.g. Jerde et al., 1993) (Fig. 3a). This texture indicates that most of the garnet and orthopyroxene exsolved from primary grains of clinopyroxene. Some specimens also contain large discrete grains of orthopyroxene (up to 4 mm), which are probably primary grains in the rocks (Fig. 3b). Spinel and olivine occur as minor phases in some samples. Spinel is always mantled by garnet, which is associated with anhedral olivine. This texture shows that the garnet originated not only by exsolution from aluminous clinopyroxene and orthopyroxene, but also by reaction between both pyroxenes and spinel.
|
Spinel–garnet clinopyroxenites (SGC) are mainly composed of clinopyroxene, garnet and spinel, whereas orthopyroxene is generally absent. Only sample SAE101 contains a small, rounded grain of orthopyroxene (
0·1 mm) within clinopyroxene. In many samples, brown spinel is mantled by garnet and isolated from clinopyroxene similar to the spinel occurring in GSL and GW. Anhedral olivine is rarely present around the grains of garnet (Fig. 3c). In other specimens, spinel is not isolated from clinopyroxene. Many of these samples display a granoblastic texture, and are composed of polygonal, equant garnet and clinopyroxene grains, as well as gray–green spinel with euhedral and straight-edged crystals (Fig. 3d).
Garnet clinopyroxenites (GC) are principally composed of large, subhedral or rounded grains of garnet (1–6 mm) in an interstitial matrix of clinopyroxene (<1 cm), and are easily distinguished from the GW and SGC by no textural indication of subsolidus exsolution and reaction (Fig. 3e). The proportion of garnet to clinopyroxene varies from 1:9 to 1:1 (Table 1), and has relevance to the size and morphology of garnet. Subhedral and smaller garnet generally occurs in clinopyroxene-rich specimens, although well-rounded and larger garnet tends to occur in garnet-rich specimens. Although most members of the GC are bimineralic, some samples contain other accessory phases. Sample SAE116 is a quartz–garnet clinopyroxenite, mainly composed of clinopyroxene (75%), with minor anhedral to subhedral garnet (21%) and small-quartz aggregates (4%) occurring along grain boundaries (Fig. 3f). Small grains of quartz also appear as discrete inclusions in the clinopyroxene and garnet. SAE132 and SAE141 contain orthopyroxene as an accessory phase, but the textures in the two samples differ distinctly. Clinopyroxene in sample SAE132 is extremely large (3 cm) and entirely surrounds well-rounded garnet (<1 cm) associated with small orthopyroxene (<1 mm). Presumably, this specimen represents a transitional example between the GC and isolated megacrysts because the constituent minerals have sizes equivalent to those in the megacryst suites. Sample SAE141 shows distinct porphyroclastic textures. Porphyroclasts of clinopyroxene, euhedral orthopyroxene and subhedral to anhedral garnet with thick kelyphitized margins are enclosed by a matrix of clinopyroxene neoblasts (0·2 mm). Rarely, orthopyroxene occurs as thin exsolution lamellae in the core of clinopyroxene porphyroclasts.
Garnet orthopyroxenites (GO) are mainly composed of coarse-grained (up to 1·5 cm) orthopyroxene and interstitial garnets (<2 mm) with irregular and curvilinear grain outlines (Fig. 3g). The proportion of garnet to orthopyroxene varies from 1:9 to 5:4 (Table 1). Similar to the GC, textural evidence of exsolution is apparently absent. Many specimens contain small, well-rounded inclusions (<1 mm) of olivine and clinopyroxene within orthopyroxene, and, rarely, within garnet. Some specimens also contain large clinopyroxene (<1 cm) with anhedral interfaces between the orthopyroxene and interstitial garnet. Three samples (SAE136, SAE137 and SAE139) have megacrystalline garnets (up to 2·5 cm) scattered through the orthopyroxene-dominated matrix (Fig. 3h). The irregular rims of megacrystalline garnets are texturally continuous to anhedral garnets in the matrix, indicating that second-generation garnet grew on pre-existing rounded grains of garnet. This argument is supported by the fact that marked chemical zonation is retained in the megacrystalline garnet.
Megacryst suites
Megacryst suites are defined by large, rounded, single crystals, and mainly consist of garnet, subcalcic diopside, augite, bronzite and ilmenite (e.g. Nixon & Boyd, 1979). Phlogopite and zircon are also regarded as members of megacryst suites. Moreover, the presence of clinopyroxene–ilmenite graphic intergrowths, which are common in kimberlitic inclusions, have been described in previous studies. The most common mineral is garnet and some specimens contain smaller grains of subcalcic diopside. Compared with the garnet megacrysts, the abundance of clinopyroxene megacrysts is small, owing to their rare occurrence in the field. We investigated 18 clinopyroxene megacrysts, including nine subcalcic diopsides, six augites, two clinopyroxene–ilmenite intergrowths and one clinopyroxene–rutile intergrowth. The clinopyroxene–rutile intergrowth displays a texture that is similar to the clinopyroxene–ilmenite intergrowths and also contains ilmenite along thin surfaces and inner cleavage planes of the rutile. One of the miscellaneous samples comprising mica, amphibole and ilmenite with minor apatite and clinopyroxene is very similar to the MARID suite in kimberlite (e.g. Cawson & Smith, 1977) and is probably related to the augite megacrysts, based on compositional similarity.
|
MAJOR ELEMENT MINERAL CHEMISTRY AND EQUILIBRIUM CONDITIONS |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLE DESCRIPTIONMAJOR ELEMENT MINERAL CHEMISTRY...DISCUSSIONSUPPLEMENTARY DATAREFERENCES |
|---|
Mineral compositions were measured by electron microprobe (JEOL-JXA8800) with a wavelength-dispersive analyzer system, housed in the Tokyo Institute of Technology. Natural silicate minerals and synthetic oxides were used as standards. All analyses were performed with an accelerating voltage of 15 kV, and a focused electron beam with a 12 nA current. An oxide ZAF correction scheme was employed. For checking the compositional homogeneity of individual phases in a single sample, automatic X-ray concentration maps were produced. Generally, constituent minerals are quite homogeneous in individual samples, although some of the samples contain zoned minerals (details below). Compositional ranges of major minerals in each rock type are summarized in Table 1. Constituent mineral compositions of representative samples from each rock type are listed in Table 2. Stoichiometric calculation for Fe3+ was applied only for spinel. Total Fe-content was assumed as Fe2+ for silicates because low apparent concentrations of Fe3+ inhibited precise evaluation from microprobe analyses. The entire set of mineral compositions for the studied specimens is available for downloading from the Journal of Petrology website at http://www.petrology.oupjournals.org.
|
Pyroxenes
Clinopyroxene and orthopyroxene are common minerals among the three rock suites and their compositions are distinct in terms of a Ca–Mg–Fe* (Fe* indicates total Fe assumed as Fe2+) diagram (Fig. 4) and minor element concentrations such as Na, Ti and Cr (Fig. 5). Clinopyroxene and orthopyroxene in most of the spinel lherzolite and harzburgite samples (SL–SH) are characterized by high Mg-number [= Mg/(Mg + Fe*)] and are classified as diopside and enstatite, respectively. Both pyroxenes in the garnet–spinel lherzolite (GSL) and the garnet websterite (GW) have almost identical compositions to SL–SH pyroxenes, except for lower Cr2O3 in the GW clinopyroxene (Fig. 5a). Spinel–garnet clinopyroxenites (SGC) also contain Mg-rich diopsides, but they tend to have slightly higher XCa [= Ca/(Ca + Mg + Fe* + Mn)] and lower Cr2O3 than those in SL–SH, GSL and GW. High Ca content in diopside from SGC is accompanied by elevated Al2O3 (up to
11 wt %; Table 1), indicating a high Ca-Tschermak (CaTs) component. Al contents of both pyroxenes in most GW and SGC are significantly zoned and systematically decrease from core to rim. Textural observations indicate that this zoning is closely associated with secondary generation of garnet. Although similar zoning is recognized in both pyroxenes in some of the SL–SH and GSL, the observed heterogeneity in individual samples is smaller than those in the GW and SGC.
|
|
Compositions of both pyroxenes in garnet lherzolite (GL), garnet clinopyroxenite (GC) and garnet orthopyroxenite (GO) are remarkably homogeneous in individual samples (except for SAE141 in GC) and they are characterized by Ca-poor clinopyroxene and Ca-rich orthopyroxene that plot outside of the compositional fields of SL–SH, GSL and GW in the Ca–Mg–Fe* diagram (Fig. 4). This deviation indicates that the equilibrium temperatures of these rocks are higher than those of other types (see below). Similar deviations of pyroxene compositions are found in six samples of spinel harzburgite. Although harzburgite with high-Ca diopside and low-Ca enstatite is common, there is a significant discrepancy for CaO content in orthopyroxene between the two types of harzburgites (high-temperature spinel harzburgites or HTSH 1·1–1·6 wt %, SH <0·7 wt %). For this reason, the HTSH is distinguished from the other spinel peridotite (SL–SH). Clinopyroxenes in HTSH and GL are characterized by large Ca variations with limited variation of Mg-number, but pyroxenes in the porphyroclastic GL (SAG21 and SAG27) are substantially less magnesian (clinopyroxene 0·868 and 0·861, orthopyroxene 0·883 and 0·877, respectively). GC and GO are mainly composed of pyroxenes with much lower Mg-numbers and these clinopyroxene compositions display the lowest XCa among all the xenolith suites. Clinopyroxene and orthopyroxene in Fe-rich GL and all of GC–GO compositionally resemble the megacrystalline subcalcic diopside and bronzite, as shown in the Ca–Mg–Fe* diagrams. It has been recognized that clinopyroxenes of the Malaitan megacryst suites, ranging from subcalcic diopsides to clinopyroxene–ilmenite intergrowths and to augites, exhibit regular chemical trends; with decreasing Mg-number, there is a decrease in Cr and an increase in Na and Ti contents (Nixon & Boyd, 1979
; Neal & Davidson, 1989). In terms of these minor element concentrations, most of the GC–GO pyroxenes and megacrystalline diopside–bronzite are compositionally identical, indicating that these may be genetically linked. However, clinopyroxene in quartz-bearing GC (SAE116), GC with lowest Mg-number (SAE152) and GC with porphyroclastic texture (SAE141), significantly deviates from the regular trends toward lower Na and Ti contents at a given Mg-number. Also, clinopyroxene in Fe-rich GL has a distinctively higher Cr content than the megacryst cpx.
Garnet
Garnets in GSL and GW have similar compositions, characterized by high Mg-number, low XCa and low TiO2 (Figs 4 and 6). The garnet in GSL tends to be more Cr-rich than the GW garnet (GSL 1·0–1·9 wt %, GW 0·35–1·4 wt %). Sample SAG26, which is the only sample without spinel among the GSL, contains distinctively high-Cr garnets (2·9 wt %), indicating that perfect spinel consumption resulted in elevation of Cr contents in garnet. On the other hand, garnet in SGC demonstrates highly variable XCa, with relatively constant Mg-number on Ca–Mg–Fe* diagram, and the Mg-number is slightly higher and Cr2O3 is lower than those in GW. Similarly to pyroxene compositions, garnets in GW and SGC display compositional zoning; XCa systematically increases from core to rim. The variation is closely related to Al zoning in coexisting pyroxenes, suggesting that the exsolved garnet tends to be more Ca-rich than pre-existing garnet and that homogenization occurred imperfectly.
|
Garnet compositions of GL can be discriminated from GSL by higher Mg-number and Cr2O3, except for Fe-rich garnets coexisting with low Mg-number pyroxenes. Previous studies reported that some garnet-bearing lherzolites have garnets with much higher Cr content—up to 5·5 wt % (Nixon & Boyd, 1979
; Neal, 1988, referred to the rocks as Cr-pyrope lherzolite). These garnets coexist with Ca-rich orthopyroxenes and Ca-poor clinopyroxenes, indicating that the Cr-pyrope lherzolites have a deep connection with GL rather than GSL, despite the fact that two of them contain Cr-rich spinel. Garnet in GC–GO is Fe-rich and overlaps with the compositional field of the megacrysts. In general, most garnets in GC, GO and GL are quite homogeneous (as are their coexisting pyroxenes), but coarse-grained garnets in three specimens of GO and one specimen of Fe-rich GL exhibit compositional zoning, even though other coexisting phases are homogeneous (Figs 4 and 6). Garnet in SAG21 (Fe-rich GL) and the largest garnet in SAE136 (GO) retain a high Mg-number core (SAG21 0·84, SAE136 0·865), which is equivalent to those in other GL garnet. These garnets show appreciable increase in Ti content (both SAG21 and SAE136 0·3–0·5 wt %) associated with decreasing Mg-number (SAG21 0·80, SAE136 0·795) towards the rims, indicating that the Fe-rich compositions of these rocks were attained by some kind of enrichment process (e.g. Smith & Boyd, 1987; Burgess & Harte, 1999). The covariation relationship between Mg-number and Ti content is well defined in garnet megacrysts similar to the case of the pyroxene megacrysts, implying that they track the crystallization trends of a magma that crystallized the megacrysts (Fig. 6b). Most of GO and three GC garnets plot near the Mg-rich side of the megacryst field with equivalent Ti content. However, GC garnets with low Mg-number tend to deviate from the megacryst trend. In particular, garnets in SAE152 with lowest Mg-number and quartz-bearing SAE116 are clearly depleted in Ti and rich in Ca relative to the megacryst at a given Mg-number. In terms of Cr content, there is no systematic difference between megacryst and GC–GO garnets. However, zoned garnets in GO have a distinctively high Cr content in their cores (up to 2·85 wt %, Fig. 6a). The zoned garnet in SAE136 shows complex variation of Cr content; highest Cr content appears in the inner rim of the grain (up to 2·4 wt %) and the Cr content in the core and outermost rim decreases to equivalent values of other small anhedral garnets.
Olivine and spinel
Cr-number of spinel [(= Cr/(Cr + Al)] and forsterite (Fo = 100 Mg number) content in olivine are generally used as sensitive indicators for the degree of depletion in mantle peridotite. It has been recognized with experiments (e.g. Jaques & Green, 1980) and natural samples (e.g. Arai, 1994) that the Cr-number of spinel and Fo content in olivine progressively increase with enlargement of the extent of melt extraction. Cr-number in spinel of Malaitan spinel peridotites ranges from 0·1 in clinopyroxene-rich SL to 0·6 in clinopyroxene-poor SH, and shows a negative correlation with modal abundance of clinopyroxene (Fig. 7a) and a positive correlation with Fo in olivine (Fig. 7b). In the latter diagram, most samples plot along the high-Fo and low-Cr-number side of the olivine–spinel mantle array (OSMA) representing the compositional field of mantle-derived residual peridotites (Arai, 1994). This evidence is consistent with the above partial-melting relationship. However, some of the amphibole-bearing SL and SH plot outside of the OSMA trend toward the lower-Fo and higher-Cr-number side, implying that extensive crystallization of amphibole can modify the coexisting spinel and/or olivine compositions. In the same way, the high-Cr nature of spinel in GSL (0·15–0·43) at a given Fo content (88·5–91) can be explained by the secondary formation of garnet, which resulted in the selective Al consumption from the spinel (Neal & Nixon, 1985). This explanation is largely supported by the fact that spinel Cr-number in GSL is roughly correlated with the modal abundance of garnet (Fig. 7c).
|
Spinel with the highest Cr-number is found in HTSH and generally coexists with high-Fo olivine (91–92). Although one sample (SAS63) contains considerably lower-Fo olivine than those in the other HTSH (
90·5), all of the HTSH plot within the OSMA trend (Fig. 7b). It can be difficult to discriminate the HTSH from the entire population of spinel peridotites on the OSMA trend, but HTSH can be discriminated from other peridotites by their high Cr-number in spinel at a given modal abundance of clinopyroxene (Fig. 7a). Additionally, Cr-pyrope lherzolites (PHN3539 and CRN209) also contain spinel with significantly high Cr-numbers (0·54 and 0·50, respectively), indicative of the strong similarity to HTSH rather than GSL. In contrast to high-Cr-number spinel in harzburgites, highly aluminous spinel commonly occurs in SGC (Cr-number <0·105). Most of olivines in GL have equivalent Fo (90–91) to those in other lherzolites, except for SAG21 and SAG27, which contain low-Fo olivine (87·0 and 87·6, respectively). Olivine with lower Fo contents is present in GO (85–87), consistent with the low Mg-number of other coexisting phases. Thus, olivine in the xenoliths exhibits a relatively large variation in Fo content. Additionally, there is also significant variation of CaO, despite the low concentration (<0·13 wt %), which is apparently irrespective of the variations in Fo content; olivine in HTSH, GL and GO contains 0·08 wt % of CaO or above, whereas olivine in GSL, SL–SH and GW–SGC contains 0·07 wt % of CaO or below.
Equilibrium conditions
Equilibrium P–T estimates for the samples that have garnet, clinopyroxene and orthopyroxene were calculated with a combination of the Al-in-orthopyroxene barometer and the two-pyroxene thermometer of Brey & Köhler (1990), which are believed to yield the most accurate P–T estimates for natural peridotitic rocks. Those workers noted that the above combination reproduced their natural experiments to ±20°C and ±3 kbar (1
). For garnet-free peridotite, only temperature was calculated by the two-pyroxene thermometer, using various pressure values. Also, the Ca-in-orthopyroxene thermometer of Brey & Köhler (1990) was applied for peridotite samples, to test overall reliability. For SGC and most of GC that do not include orthopyroxene, the garnet–clinopyroxene Fe–Mg exchange thermometers of Ellis & Green (1979) and Krogh (1988) were applied. For most of GO that do not include clinopyroxene, the garnet–orthopyroxene Fe–Mg exchange thermometer of Harley (1984) was used. Consistency between the two-pyroxene thermometer and other methods was tested for samples where different thermometers could be applied.
As mentioned above, individual phases in a sample are generally homogeneous and averaged compositions can be used for thermobarometric calculations for most samples. However, most GW and SGC, and some GSL, are composed of the minerals showing systematic compositional zoning within individual grains; Al contents of both pyroxenes decrease and Ca content in garnet increases towards grain margin, owing to exsolution caused by partial reequilibration in a cooling situation. Therefore, for GW and GSL, both pyroxenes with the lowest Al content and garnet with highest XCa were selected as the pair of last equilibrated compositions, which are commonly present in the outermost rim of individual minerals. Although the Al content in clinopyroxene depends on the amount of jadeite component, the observed variation of Na2O in a single sample is limited and small, so that the variation does not affect the temperature calculation. In the case of applying the garnet–clinopyroxene thermometer for SGC, there is no systematic variation in the Mg-numbers of constituent minerals, but XCa heterogeneity in garnet significantly influences the results. Therefore, the combination of garnet with maximum XCa and clinopyroxene with minimum Al content was used. On the other hand, P–T estimates for GO and GL with heterogeneous garnet were made by combination of the composition of the rim of zoned garnet and the averaged composition of other homogeneous phases. Although SAE141 (GC) and SAE136 (GO) contain garnet, clinopyroxene and orthopyroxene, P–T calculations were not performed because their complex inter-mineral variations make it difficult to define the equilibrium mineral assemblages.
All of the P–T points obtained by the combination of Al-in-orthopyroxene barometer and the two-pyroxene thermometer yield a linear trend (r2 of a linear least-squares regression = 0·975) in the P–T field, ranging from 1·8 GPa and 850°C to 3·6 GPa and 1350°C (Table 3; Fig. 8a). This method places all of the P–T points in the garnet lherzolite stability field determined by experiments with both natural (Green & Ringwood, 1970) and CMAS (O'Neil, 1981; Klemme & O'Neil, 2000) systems, implying that the estimation is reasonable. However, the result of temperature estimations for spinel peridotite (SL, SH and HTSH, shown by the lines with pressure dependence) demonstrates that the variation is very large and almost covers the entire T range of garnet-bearing peridotite. It is known that the spinel-to-garnet peridotite transition depends on the degree of depletion; the spinel–garnet reaction boundary continuously shifts to higher pressures with increasing Cr content and Mg-number in the system (O'Neil, 1981; Robinson & Wood, 1998). This evidence indicates that most of the spinel peridotite was equilibrated within the garnet lherzolite field of the pyrolite III and CMAS systems, which represent very fertile compositions. Therefore, the appearance of garnet reflects differences in not only equilibrate condition but also bulk composition.
|
|
Compared with the above results, the combinations of other thermometers and the Al-in-orthopyroxene barometer give significantly different P–T arrays as a result of systematic discrepancies in estimated temperatures (Fig. 9a–d). It is noted that the calculated temperatures with Fe–Mg exchange thermometers of garnet–clinopyroxene and garnet–orthopyroxene are directly affected by the presence of Fe3+ in the requisite minerals. Because total Fe was assumed as Fe2+ and no correction for Fe3+ was attempted in the temperature estimation, the deviations may be partly because of the presence of Fe3+. However, the magnitude of scattering and tendency of deviations are equivalent to those expected from the ability of each thermometer to reproduce Fe3+-free experimental conditions (Brey & Köhler, 1990
) (Fig. 9a–c). This implies that the observed temperature discrepancies can be mainly attributed to errors of individual calibrations.
|
T = calculated T – TBK90(Two–Px) against TBK90(Two–Px). Shaded fields in (a), (b) and (c) indicate the reproductions of experimental temperatures with each thermometer, taken from Brey & Köhler (1990)The Ca-in-orthopyroxene thermometer also yields significantly different results from those of the two-pyroxene thermometer. The temperature gap between HT-type and LT-type becomes larger, as shown by the two arrays individually oblique to the zero-line (Fig. 9d). Although both equations were calibrated using the same experimental results, this may also be as a result of inadequate calibration of the Ca-in-orthopyroxene thermometer, which is strongly influenced by Na content in orthopyroxene (Brey & Köhler, 1990
). It is evident that all samples plotted above zero contain low-Na orthopyroxene (Na <0·005 in structural formulae based on six oxygens), similar to the pattern seen in application to natural garnet peridotites from Kaapvaal craton (Brey & Köhler, 1990). As noted by Brey & Köhler, this evidence supports the argument that the occurrence of significant Na variations necessitates caution in the use of Ca-in-orthopyroxene thermometer.
Thus, all other methods give broad agreement with the two-pyroxene thermometer, as seen in their somewhat correlated patterns (Fig. 9a–d), but the presence of systematic deviations does not allow us to reconcile temperature values calculated from different methods. For example, it is clear that the estimated temperature for SGC and GC must be compared with those of other samples using the same method. Using pressure values estimated by the methods of Brey & Köhler (1990) in the equations of the garnet–clinopyroxene thermometer of Ellis & Green (1979) reveals that isopleths for GC and SGC lie in a comparable range to those of GW and GO, respectively (Fig. 8b). As shown by the compositional resemblance, we can safely state that the lithological pair SGC–GW equilibrated at equivalent conditions; similar arguments apply to the pair GC–GO. The origin of higher T estimates for two SGC samples (SAE115 and SAE125) is unclear. It is noted that both samples are composed of garnet and clinopyroxene with the highest XCa and CaTs, respectively. Therefore, it is possible that errors in calculated temperatures for the two samples are not comparable with those observed in GW and GSL.
The systematic P–T trend obtained by the methods of Brey & Köhler (1990) is consistent with the theory that the xenolith suites recorded the P–T conditions controlled by the geotherm when the host alnöite erupted (Nixon & Boyd, 1979; Kawasaki, 1987). As pointed out in previous studies, the P–T array has a much higher gradient than those calculated by kimberlite xenoliths (e.g. Rudnick & Nyblade, 1999), suggesting that the geotherm reflects the higher heat flow of the oceanic environment (oceanic geotherm) than that of continental areas (continental geotherm). In Fig. 8, the geotherm constructed using the xenolith data is compared with the results of simple one-dimensional heat-flow calculations using the following equation (Turcotte & Schubert, 1982):
 | (1) |
is the thermal diffusivity, taken to be 1 mm2/s. As the OJP plume would have thermally affected the xenoliths or the lithospheric mantle, it is difficult to evaluate the thermal evolution of the lithospheric mantle quantitatively. However, in giving several reasonable parameters, such as the age of the lithosphere after a thermal perturbation (t = 60–100 Ma) and adiabatic mantle temperature (Tm = (1300–1400) + 0·4°C), the xenolith geotherm can be broadly reproduced by this simple model, implying that the thermal state within the lithospheric mantle was in a cooling situation at the time of the xenolith host eruption, similar to normal oceanic lithosphere. Given that the concept of the xenolith geotherm is acceptable, equilibrium conditions obtained from individual samples represent each of their derived depths, suggesting that the lithological structure of sub-OJP mantle can be reconstructed. From this point of view, it is clear that both peridotite and pyroxenite suites can be assigned to a low-temperature (LT-type) or a high-temperature type (HT-type)—differences that seem to be derived from the distinct depth intervals. The LT-type of pyroxenite and peridotite suites consists of GW–SGC and SL–SH–GSL, respectively; whereas GC–GO and GL–HTSH belong to the HT-type of pyroxenite and peridotite suites, respectively.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLE DESCRIPTIONMAJOR ELEMENT MINERAL CHEMISTRY...DISCUSSIONSUPPLEMENTARY DATAREFERENCES |
|---|
Origin of LT-type peridotite and pyroxenite
The above thermobarometric results indicate that the shallower mantle beneath the OJP was composed of LT-type peridotite (SL–SH and GSL) and pyroxenite (GW–SGC). The higher abundance of peridotite relative to pyroxenite within our collection suggests that peridotitic mantle seems to be widely distributed beneath the OJP. Although temperature estimates obtained for SL–SH vary significantly, samples equilibrated at lower temperatures (<950°C) are commonly SL–SH rather than GSL, suggesting that the uppermost mantle consists of spinel peridotite and that garnet–spinel peridotite is dominant in slightly deeper regions.
In terms of chemical composition, SL–SH and GSL generally contain Mg-rich minerals and are considered as a series of melting residues. However, the compositional variations of constituent minerals are also influenced by post-melting modification. It is evident that all rocks reequilibrated at subsolidus conditions as indicated by abundant exsolution textures, either as lamellae or granular forms. Furthermore, many samples display evidence for metasomatic enrichment, summarized as follows: (1) amphibole texturally equilibrated with other primary phases is ubiquitous; (2) depleted SH usually contains clinopyroxene, with elevated Na and Ti contents as high as those in fertile SL and GSL (Fig. 10a and b); (3) enriched light rare earth elements (LREE), or convex-upward rare earth element (REE) patterns in clinopyroxene are common in SL–SH and GSL (Neal, 1988; Ishikawa, 2002). These lines of evidence suggest that the signature of the former partial melting event was substantially masked by subsequent recrystallization and metasomatism, which has been recognized in the studies of other mantle xenoliths and orogenic peridotites (e.g. McDonough & Frey, 1989; Bodinier et al., 1990).
|
Residual signatures are largely preserved in depletion indicators such as Cr-number in spinel, Fo content in olivine and heavy rare earth element (HREE) concentrations in clinopyroxenes, which were analyzed using a Cameca IMS-3f ion mass spectrometer (SIMS) at Tokyo Institute of Technology [detailed analytical procedure is the same as that described by Yurimoto et al. (1989)
], because they are less sensitive to the effects of subsequent metamorphism and metasomatism. Figure 10c illustrates that YbN [subscript N indicates normalization to chondritic values taken from Anders & Grevesse (1989)] in clinopyroxene in the majority of the samples is negatively correlated with the Cr-number in spinel (unpublished data; Ishikawa, 2002). Presumably, the deviations observed in some SH can be ascribed to metasomatism that modified the Yb abundance or both values. The correlation is expected in residual mantle that has undergone progressive depletion through melt extraction from a fertile source, and is very similar to the trend defined by the compilation of global abyssal peridotite data (Hellebrand et al., 2001), which are regarded as the main constituent of oceanic lithosphere. This evidence indicates that the shallower mantle represents normal oceanic lithosphere and that a former partial melting event occurred in a mid-oceanic ridge setting.
According to theoretical considerations for partial melting of mantle in a mid-oceanic ridge setting, melting occurs in response to adiabatic decompression of ascending asthenospheric mantle (e.g. Klein & Langmuir, 1987). From this viewpoint, we can expect that the above depletion indicators correlate with estimated temperature because the oceanic lithosphere probably forms a vertically stratified column of melt-depleted residues, in which the extent of depletion increases from bottom to top (e.g. Plank & Langmuir, 1992). It is possible that the uppermost mantle beneath the OJP (<850°C) displays depleted characteristics, as represented by higher Cr-number in spinel and Fo content of olivine, and lower Na content in clinopyroxene (Fig. 11a–c), consistent with the above hypothesis. However, SL–SH samples showing a similar degree of depletion also occur at significantly greater depth. This could be partly explained by local heterogeneity in the depletion of peridotite formed during adiabatic ascent beneath a spreading ridge axis. It has been proposed that focused extraction of mid-ocean ridge basalt (MORB) through the formation of high-permeability channels within upwelling peridotite causes melt–rock reactions involving precipitation of olivine and dissolution of pyroxene (e.g. Kelemen et al., 1995). Though essentially proposed for the formation of dunites, this mechanism may increase the apparent degree of depletion of the host peridotite and amplify the lateral variation at particular depths. However, the evidence that the majority of SH display higher temperature than those of SL and GSL (>1000°C) implies that another process is responsible for the creation of SH at great depth; this will be discussed later.
|
A substantial amount of pyroxenite (GW–SGC) is also present at varying depths in the shallower mantle, but these rocks are hypothesized to occur only locally, as inferred from their restricted occurrence. Compared with LT-type peridotite, exsolution and reaction textures attributed to subsolidus recrystallization are prominent in LT-type pyroxenite, as mentioned above. Although the compositional zoning caused by partial re-equilibration is commonly preserved in the constituent minerals, the present mineral compositions do not represent the ‘primary’ compositions established when the individual rocks were generated. Therefore, in order to discuss the protolith formation, it is necessary that the primary mineralogy and chemistry of each rock are deduced from the textural, modal and compositional variations. Although we do not describe the detailed results here, the series of the primary mineralogy systematically changes from GW to SGC, as follows: clinopyroxene, clinopyroxene + orthopyroxene, clinopyroxene + orthopyroxene + spinel, clinopyroxene + spinel + garnet (Table 4). Both the assemblage and sequence can be adequately accounted for through a single fractionation process, according to the liquidus phase equilibrium data below 3·0 GPa for the model basalt tetrahedron (Suen & Frey, 1987
; Milholland & Presnall, 1998). This evidence indicates that the LT-type pyroxenite may represent an early cumulate that crystallized from a mafic magma originating from the deeper mantle (Ishikawa et al., in preparation).
|
Origin of HT-type peridotite and pyroxenite
Presence of a depleted zone
The results of geothermobarometric calculations show the occurrence of an apparent gap, separating the LT- and HT-types of garnet-bearing peridotite (GSL and GL, respectively) in P–T space, ranging from 2·7 GPa and 1050°C to 3·0 GPa and 1220°C (Fig. 8a). Although the origin of this apparent gap may result from sampling bias, made by either host alnöite or ourselves, some of the SH and HTSH samples record temperatures corresponding to this gap (1050–1120°C for 10 SH samples, 1190°C for SAS21 and 1224°C for SAS46, Table 5), as shown in Fig. 11, which displays the various mineral compositions as a function of temperature. This implies that the apparent gap in the Malaitan geotherm is dominated by spinel harzburgite and may represent a ‘garnet-poor depleted zone’. Presumably, the low abundance of the appropriate samples is because of weak resistance of an olivine-rich petrography against surface alteration.
|
It is important to note that the possible boundary between LT- and HT-types lies within the ‘garnet-poor depleted zone’ and two different spinel harzburgites occur above and below the boundary (SH and HTSH, respectively), as marked by the abrupt jump in Na content of clinopyroxene; SH samples with high-Na clinopyroxene occur above the boundary, whereas HTSH samples with distinctively low-Na clinopyroxene occur below the boundary (Fig. 11c). Nixon & Boyd (1979)
reported the presence of samples that seem to be derived from just below the boundary on the basis of the two-pyroxene temperatures calculated with the mineral compositions in the literature (1176°C for PHN3538 Cr-pyrope lherzolite, 1192°C for PHN3539 spinel-bearing Cr-pyrope lherzolite and 1155°C for PHN3549B spinel lherzolite). Although the mineral assemblages and abundances may be variable, constituent minerals in these samples (including SAS21 and SAS46) consistently display greater depletion. For example, the constituent phases are characterized by high-Mg-number (100Mg-number = 92·0–92·3 in orthopyroxene, 92·6–93·2 in clinopyroxene, 85·8 in garnet, 91·9–92·3 in olivine; olivine compositions in PHN samples are not reported). The clinopyroxenes have Ti- and Na-poor composition (TiO2 <0·03 wt %, Na2O <0·3 wt %, except for 0·7 wt % in PHN3539; Fig. 11c), and the garnets in PHN3538 and PHN3539 have distinctively Cr-rich compositions (Cr2O3 5·5 wt %). These lines of evidence suggest that the lower part of the ‘garnet-poor depleted zone’ is appropriate for melt-depleted residual peridotite generated by a high degree of melting, whereas the upper part of the ‘garnet-poor depleted zone’ comprises SH with a strongly metasomatized signature, which is not truly depleted in melt components.
This interpretation implies the presence of intra-lithospheric boundary, subdividing the whole mantle section beneath the OJP into two layers, differing in origin because the most depleted residue is generated in an upper portion of the adiabatically ascending mantle (e.g. Klein & Langmuir, 1987; Watson & Mckenzie, 1991; Plank & Langmuir, 1992). Therefore, formation of this compositional structure requires at least a two-stage process for the partial melting of the peridotitic mantle: (1) formation of shallower LT-type peridotite by upwelling and depletion of mantle with relatively lower temperature; (2) formation of HT-type peridotite by melting of ascending mantle with higher temperature. This argument suggests that the depleted zone originated from a hot mantle plume, which impinged upon the pre-existing lithosphere. It is possible that plume impingement caused thermochemical erosion at the bottom of pre-existing lithosphere because of infiltration of a plume-derived melt at high temperature. This process could explain the production of highly metasomatized SH by a melt–rock reaction involving pyroxene dissolution and olivine precipitation at the bottom of the pre-existing lithosphere, which would receive a high flux of infiltrated melt (Kelemen et al., 1992; Bedini et al., 1997; Xu et al., 1998). Given these considerations, we envisage that the intra-lithospheric boundary indicates a boundary between two bodies of upwelling mantle, differing in time of emplacement and mantle potential temperature.
Recently, Griffin et al. (1999) provided an attractive interpretation of the genesis of lithospheric mantle beneath the Lac de Gras area of the Slave craton, on the basis of integrated data from a wide variety of kimberlite-borne xenolith materials. They concluded that underplating of a rising mantle plume beneath the pre-existing lithosphere resulted in the generation of a two-layered structure, defined by a compositional discrepancy observed in garnet xenocrysts. Although the model for the entire lithospheric section beneath the Slave craton provided by Griffin et al. (1999) is significantly different from our model for the sub-OJP lithospheric mantle, they concluded that the garnet-rare depleted zone occurs near the boundary between the two layers. The occurrence of intra-lithospheric depleted zones in other mantle sections beneath Archean cratons has been also suggested by previous studies utilizing either xenoliths or garnet concentrates (e.g. Finnerty & Boyd, 1987; Boyd et al., 1993; Griffin et al., 20