Skip Navigation


Journal of Petrology Advance Access originally published online on May 3, 2008
Journal of Petrology 2008 49(6):1223-1251; doi:10.1093/petrology/egn023
This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow Supplementary Data
Right arrowOA All Versions of this Article:
49/6/1223    most recent
egn023v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Google Scholar
Right arrow Articles by Simon, N. S. C.
Right arrow Articles by Widom, E.
GeoRef
Right arrow GeoRef Citation
Social Bookmarking
 Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?

© 2008 The Author(s)
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/2.0/uk/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

Ultra-refractory Domains in the Oceanic Mantle Lithosphere Sampled as Mantle Xenoliths at Ocean Islands

Nina S. C. Simon1,*, Else-Ragnhild Neumann1, Costanza Bonadiman2, Massimo Coltorti2, Guillaume Delpech3, Michel Grégoire4 and Elisabeth Widom5

1Physics of Geological Processes, University of Oslo, Po Box 1048, Blindern, N-0316 Oslo, Norway
2Earth Science Department, University of Ferrara, Via Saragat, 1, 44100 Ferrara, Italy
3Umr Cnrs 8148 Ides, ‘Interactions Et Dynamique Des Environnements De Surface’, Faculté Des Sciences, Université Orsay-Paris Sud, Bât. 504, 91405 Orsay Cedex, France
4Observatoire Midi-Pyrénées, Umr 5562-DTP, Omp, Université Toulouse III, Avenue E. Belin, 31400 Toulouse, France
5Department of Geology, Miami University, 114 Shideler Hall, Oxford, Oh 45056, Usa

RECEIVED JUNE 29, 2007; ACCEPTED APRIL 11, 2008


    ABSTRACT
 TOP
 ABSTRACT
 INTRODUCTION
 COMPOSITIONAL VARIATIONS
 DISCUSSION
 CONCLUSIONS AND IMPLICATIONS FOR...
 SUPPLEMENTARY DATA
 REFERENCES
 
Many peridotite xenoliths sampled at ocean islands appear to have strongly refractory major element and modal compositions. To better constrain the chemistry, abundance and origin of these ultra-refractory rocks we compiled a large number of data for xenoliths from nine groups of ocean islands. The xenoliths were filtered petrographically for signs of melt infiltration and modal metasomatism, and the samples affected by these processes were excluded. The xenolith suites from most ocean islands are dominated by ultra-refractory harzburgites. Exceptions are the Hawaii and Tahiti peridotites, which are more fertile and contain primary clinopyroxene, and the Cape Verde suite, which contains both ultra-refractory and more fertile xenoliths. Ultra-refractory harzburgites are characterized by the absence of primary clinopyroxene, low whole-rock Al2O3, CaO, FeO/MgO and heavy rare earth element (HREE) concentrations, low Al2O3 in orthopyroxene (generally < 3 wt %), high Cr-number in spinel (0·3–0·8) and high forsterite contents in olivine (averages > 91·5). They are therefore on average significantly more refractory than peridotites dredged and drilled from mid-ocean ridges and fracture zones. Moreover, their compositions resemble those of oceanic forearc peridotites. The formation of ultra-refractory ocean island harzburgites requires potential temperatures above those normally observed at modern mid-ocean ridges, and/or fluid fluxed conditions. Some ultra-refractory ocean island harzburgites give high Os model ages (up to 3300 Ma), showing that their formation significantly pre-dates the oceanic crust in the area. A genetic relationship with the host plume is considered unlikely based on textural observations, equilibration temperatures and pressures, inferred physical properties, and the long-term depleted Os and Sr isotope compositions of some of the harzburgites. Although we do not exclude the possibility that some ultra-refractory ocean island harzburgites have formed at mid-ocean ridges, we favor a model in which they formed in a process spatially and temporally unrelated to the formation of the oceanic plate and the host plume. As a result of their whole-rock compositions, ultra-refractory harzburgites have a very high solidus temperature at a given pressure, low densities and very high viscosities, and will tend to accumulate at the top of the convecting mantle. They may be preserved as fragments in the convecting mantle over long periods of time and be preferentially incorporated into newly formed lithosphere.

KEY WORDS: mantle xenoliths; ocean islands; recycled oceanic mantle lithosphere; ultra-depleted mantle


    INTRODUCTION
 TOP
 ABSTRACT
 INTRODUCTION
 COMPOSITIONAL VARIATIONS
 DISCUSSION
 CONCLUSIONS AND IMPLICATIONS FOR...
 SUPPLEMENTARY DATA
 REFERENCES
 
Detailed information about chemical composition is essential in understanding the formation and evolution of different parts of the Earth's mantle. Small variations in the major element composition of the lithospheric mantle may strongly influence its physical properties and its behavior in response to stress (e.g. Karato, 1986Go; Hirth & Kohlstedt, 1996Go; Afonso et al., 2007Go; Simon & Podladchikov, 2008Go). It is therefore necessary to obtain reliable chemical compositions of different mantle domains to develop realistic geodynamic models and correctly interpret geophysical data.

Mantle xenoliths sampled in ocean islands (OI) cover a wide range in compositions. Some islands are dominated by spinel harzburgites that are more refractory than most peridotites sampled along mid-ocean ridges and fractures zones (MORP: mid-ocean ridge peridotites). The xenoliths in a second group of islands are relatively fertile, whereas in a third group of ocean islands the xenoliths range from strongly refractory to fertile. In the third group the fertile compositions are to a large extent the results of metasomatism associated with ocean island formation and/or melt–rock interaction during ascent, and many of the more fertile peridotites have strongly refractory pre-OI protoliths (e.g. Ryabchikov et al., 1995Go; Wulff-Pedersen et al., 1996Go; Grégoire et al., 1997Go, 2000aGo, 2000bGo; Moine et al., 2000Go; Neumann et al., 2002Go, 2004Go; Delpech, 2004Go; Bonadiman et al., 2005Go; Shaw et al., 2006Go).

The presence and abundance of ultra-refractory mantle not only in continental keels, but also in oceanic domains, is an important observation. Most of this evidence for ‘depleted compositions’ is based on isotope and trace element data on single minerals, samples and/or locations. However, to use compositional variations in geodynamic modeling, whole-rock major element compositions and modal mineralogy averaged over some larger area are actually the most important variables. This study represents a first step in providing such constraints for more petrologically consistent quantitative geodynamic modeling. Moreover, the mechanism for the generation of these samples is far from clear and this review will serve as a solid basis for further work.

Studies of mantle xenoliths in ocean islands have so far mainly been focused on mantle metasomatism and melt–rock interaction during ascent. These are recent processes in the evolution of the peridotites. We are interested in the pre-ocean island composition of ocean island peridotites and particularly in the most refractory xenoliths. The aim of this study is (1) to establish the distribution of strongly refractory peridotites in ocean islands, (2) to constrain their origin and mode of formation, and (3) to discuss the implications of strongly refractory mantle material for geodynamics with respect to physical and chemical properties.

We have compiled available (published and unpublished) major element whole-rock and mineral composition data for 294 harzburgite and lherzolite xenoliths and their minerals from the Canary Islands (Neumann, 1991Go; Neumann et al., 1995Go, 2002Go, 2004Go; Wulff-Pedersen et al., 1996Go; Abu El-Rus et al., 2006Go), Madeira (Munha et al., 1990Go), the Azores (Johansen, 1996Go), Cape Verde (Ryabchikov et al., 1995Go; Bonadiman et al., 2005Go; Shaw et al., 2006Go), Kerguelen (Grégoire et al., 1997Go, 2000aGo, 2000bGo; Delpech, 2004Go), Grande Comore (Coltorti et al., 1999Go), Samoa (Hauri et al., 1993Go; Hauri & Hart, 1994Go), Hawaii (e.g. Jackson & Wright, 1970Go; Sen, 1987Go; Goto & Yokoyama, 1988Go; Sen & Leeman, 1991Go; Bizimis et al., 2003Go) and Tahiti (Tracy, 1980Go; Qi et al., 1994Go). We refer to samples from the same island as a xenolith suite. The data are presented in an Electronic Appendix (available for downloading at http://www.petrology.oxfordjournals.org). The database also gives the major element compositions of 450 MORP used for comparison.

Our study is based mainly on the most abundant major elements (SiO2, MgO, FeO, Al2O3 and CaO) because these are less easily reset by fertilization processes than incompatible trace elements. In addition to spinel harzburgites and lherzolites, mantle xenoliths from OI include dunites, wehrlites and pyroxenites. Dunites, wehrlites, pyroxenites, etc. are not discussed here as these rock-types are closely associated with fertilization processes (e.g. Frey, 1980Go; Sen, 1988Go; Neumann, 1991Go; Sen & Leeman, 1991Go; Kelemen et al., 1992Go, 1998Go; Grégoire et al., 1997Go, 2000bGo; Neumann et al., 2004Go). However, data on some of those samples are included in the database.


    COMPOSITIONAL VARIATIONS
 TOP
 ABSTRACT
 INTRODUCTION
 COMPOSITIONAL VARIATIONS
 DISCUSSION
 CONCLUSIONS AND IMPLICATIONS FOR...
 SUPPLEMENTARY DATA
 REFERENCES
 
Classification of samples
Because the focus of this study is on pre-ocean-island (pre-OI) compositions we apply petrographic criteria to recognize OI peridotites that have been least affected by melt–rock interaction processes. The effects of such processes range from very minor, reflected only in mild enrichment in strongly incompatible trace elements and resetting of radiogenic isotopes, to extensive petrographic changes such as those involving the formation of hydrous minerals, reactions involving the growth of ± olivine ± clinopyroxene ± spinel (± ol ± cpx ± sp) at the expense of orthopyroxene (opx), formation of poikilitic orthopyroxene and clinopyroxene, glass-rich reaction zones on primary minerals (e.g. orthopyroxene and clinopyroxene) and sieve-textured spinel and clinopyroxene (Fig. 1), in addition to significant enrichment in incompatible trace elements. Peridotites showing these textures are classified as OI2, indicating the presence of a second generation of minerals related to melt–rock reaction (Table 1).


Figure 1
View larger version (158K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 1. Photomicrographs of typical OI2 textures. (a) Melt–rock reaction in spinel harzburgite (Canary Islands) leading to growth of clinopyroxene (cpx) and olivine (ol) at the expense of orthopyroxene (opx; g, glass; crossed polars). (b) Poikilitic opx in spinel lherzolite (Canary Islands) enclosing corroded remnants of larger olivine grains and small oval olivine grains (crossed polars). (c) Poikilitic harzburgite (Kerguelen) with poikilitic cpx with spongy rims produced by interaction between protogranular harzburgites and alkaline, silica-undersaturated melt (crossed polars). (d) Poikilitic lherzolite (Canary Islands) showing cpx, locally with spinel exsolution lamellae, enclosing oval opx grains and spinel (crossed polars). Spinel occurs as single grains and as clusters or chains of very small grains. (e) Symplectitic intergrowth between cpx and spinel along grain boundaries in spinel harzburgite (Canary Islands). Small olivine domains (orange and bluish green) within the symplectite are in optical continuity with the neighboring olivine porphyroclasts (crossed polars). The texture suggests formation of the symplectite and cpx by reaction between melt and peridotite. (f) Spinel lherzolite from Cape Verde showing sieve-textured opx.

 

View this table:
[in this window]
[in a new window]

 
Table 1: Summary of textural and geochemical characteristics of the peridotite groups

 
Those xenoliths that show no petrographic evidence of metasomatism (as listed above and summarized in Table 1) are classified as OI1 (Fig. 2). These xenoliths typically show porphyroclastic to protogranular textures with deformed porphyroclasts of olivine and exsolved orthopyroxene; in some xenolith series the porphyroclast assemblage comprises exsolved clinopyroxene (e.g. from Hawaii and Tahiti) and/or spinel. The matrix consists of less deformed or undeformed neoblasts of olivine + orthopyroxene ± spinel ± clinopyroxene. In some cases the published petrographic descriptions do not give enough information to classify a given sample; such samples are assigned to the OI2 group. Because this study is aimed at investigating the composition of the upper mantle before the onset of ocean island magmatism the processes that have led to the petrographic changes recorded in the OI2 samples are of no interest here, and are not discussed. We therefore make no effort to distinguish between in situ metasomatism in the mantle and interaction with the host magma during ascent. Data points in the figures are restricted to OI1 xenoliths. However, as it is of interest to the discussion to know the extent to which melt–rock interaction has changed pre-OI peridotite compositions, we show the ranges covered by OI2 xenoliths in some diagrams.


Figure 2
View larger version (127K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 2. Photomicrographs of typical OI1 textures. (a) Protogranular texture in OI1-u xenolith (Kerguelen) showing olivine with undulatory extinction and exsolved opx (crossed polars). (b) Details of protogranular texture in spinel harzburgite (Cape Verde; crossed polars). (c) Opx porphyroclasts with cpx exsolution lamellae (exs opx) in spinel harzburgite (Canary Islands; crossed polars). Locally cpx exsolution lamellae have developed into larger domains (cpx2). Small equant cpx2 are scattered along the rims of the opx porphyroclasts. (d) Spinel lherzolite from Cape Verde with primary cpx. (plane-polarized light.). Abbreviations as in Fig. 1.

 
In addition to the petrographic distinction into OI1 and OI2, the xenoliths from different islands fall in two categories with respect to major element compositions (Table 1). One category comprises ultra-refractory harzburgites (OI-u) with low contents of clinopyroxene (< 3 vol. %) and other fusible components; the other category consists of more fertile xenolith series (OI-f), in which lherzolites with primary clinopyroxene are common. Essential petrographic and chemical differences between the two groups are outlined in Table 1. The focus of this study is on the OI-u xenoliths.

We emphasize that we distinguish between refractory/fertile, which refers to major element compositions and ultimately solidus temperatures, and the terms depleted/enriched, which refer to trace element and isotope characteristics. These chemical characteristics may be decoupled. A refractory harzburgite may, for example, be enriched in certain incompatible elements, but still retains its high solidus temperature and related physical properties. Most samples in the OI1-u group show some evidence of cryptic metasomatism [i.e. U-shaped rare earth element (REE) patterns]. We have, however, found a clear positive correlation between the extent of melt–rock interaction processes, as indicated by petrographic observations, and enrichment in strongly incompatible trace elements. On the basis of numerical modeling, Herzberg et al. (2007Go) concluded that a small fraction of trapped melt in a residual rock does not significantly affect the most abundant major elements. For the rocks discussed here this conclusion is supported by an excellent correlation between whole-rock major element chemistry and parameters such as cr-number [cation proportion Cr/(Cr + Al)] and mg-number [cation proportion Mg/(Mg + Fe2+)] in spinel, and the concentration of Al2O3 and Fo content in cores of orthopyroxene and olivine porphyroclasts, respectively.

In addition to chemical compositions we compare equilibration temperatures (Table 2) for the various peridotite suites. OI1-u harzburgites lack primary clinopyroxenes, and secondary clinopyroxenes in these rocks are commonly too small to analyze. OI2 xenoliths show two or more stages of mineral growth, and it is difficult to ascertain which mineral compositions represent equilibrium. We therefore chose to use the single-mineral CaO-in-opx geothermometer of Sachtleben & Seck (1981Go), assuming that the presence of clinopyroxene exsolved from orthopyroxene satisfies the four-phase requirements, also for harzburgites without primary clinopyroxene. This is in agreement with the conclusions of Neumann (1991Go) and Neumann et al. (1995Go) based on the application of different geothermometers for OI1-u harzburgites from the Canary Islands. The most important characteristics of the xenolith suites from each ocean island are summarized below.


View this table:
[in this window]
[in a new window]

 
Table 2: Equilibration temperatures estimated for ocean island xenoliths using the Sachtleben & Seck (1981Go) geothermometer based on CaO in orthopyroxene

 
Canary Islands
Xenoliths from the Canary Islands belong to the OI-u group (Table 1). They are dominated by spinel harzburgites with protogranular, porphyroclastic or poikilitic textures (Neumann, 1991Go; Siena et al., 1991Go; Neumann et al., 1995Go, 2002Go, 2004Go; Wulff-Pedersen et al., 1996Go; Abu El-Rus et al., 2006Go; E.-R. Neumann, unpublished data, 2007). The OI1-u group is restricted to porphyroclastic to protogranular harzburgites with porphyroclasts of olivine (ol1) and orthopyroxene (opx1); in some samples large equidimensional spinels may also represent porphyroclasts (sp1). The groundmass consists of equant to irregular grains of olivine, orthopyroxene and spinel, and minor amounts (generally <3 vol. %) of clinopyroxene. Clinopyroxene is present as exsolution lamellae in orthopyroxene and as small neoblasts surrounding orthopyroxene porphyroclasts (opx1), and is interpreted as a younger generation (cpx2) than the porphyroclasts (Fig. 2a). Rare, local occurrences of small, interstitial, vermicular cpx–sp intergrowths may represent melt infiltration.

OI2-u xenoliths (mainly from La Palma and Tenerife) comprise both spinel harzburgites and lherzolites. Clinopyroxene in OI2-u is found as small interstitial grains, as reaction rims on orthopyroxene (Fig. 1a and e), in cpx + sp ± opx symplectites along grain boundaries and/or as part of reaction textures (formation of clinopyroxene ± olivine at the expense of orthopyroxene; Fig. 1a and e), or as poikilitic grains enclosing round grains of olivine, embayed grains of orthopyroxene (Fig. 1d) and small spinel grains. Many xenoliths from Fuerteventura exhibit fibrous orthopyroxene. Some OI2 samples, particularly from La Palma and Tenerife, carry trace amounts of phlogopite.

No primary clinopyroxene porphyroblasts (cpx1) have been observed in the Canary Islands xenoliths; the clinopyroxene present (cpx2) is believed to have formed partly by exsolution from orthopyroxene porphyroclasts (opx1) and subsequent recrystallization (Fig. 2a), partly from interaction with infiltrating melts (Fig. 1a, d and e; Neumann et al., 1995Go, 2004Go). The samples with the highest modal proportions of clinopyroxene also show the strongest textural evidence of metasomatism.

Whole-rock major element relationships for Canary Islands OI1-u are plotted in Fig. 3. We have also plotted orthopyroxene–olivine (opx–ol) tie-lines, as defined by OI1-u xenoliths from Lanzarote. Major elements other than Na2O plot close to the opx–ol tie-line for most OI1-u xenoliths, consistent with the absence of primary clinopyroxene and confirming their strongly refractory composition. A few samples scatter at a moderate distance away from the opx–ol tie-line, for example in the SiO2–CaO diagram. We interpret this scatter as the result of minor metasomatism and/or melt–rock interaction. The wide range in Na2O is probably due to enrichment processes or contamination (Neumann, 1991Go; Siena et al., 1991Go; Neumann et al., 1995Go, 2004Go). The strongly melt-depleted nature of the OI1-u harzburgites is also reflected in low Al2O3 in orthopyroxene (0·3–3·4 wt %), Cr-rich spinels (cr-number = 0·29–0·95, with the majority of the samples in the range 0·53–0·71), and relatively high Fo-content [100Mg/(Mg + Fe)] in olivine (90·7–92·5; Fig. 4).


Figure 3
View larger version (28K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 3. OI1-u whole-rock major element variations. Data sources are as follows. Canary Islands: Neumann (1991Go), Wulff-Pedersen et al. (1996Go), Neumann et al. (2002Go, 2004Go) and Abu El-Rus et al. (2006Go); Kerguelen: Grégoire et al. (1997Go, 2000aGo, 2000bGo) and Delpech (2004Go); Cape Verde: Bonadiman et al. (2005Go) and C. Bonadiman (unpublished data, 2008). Dotted lines enclose the field of OI2-u xenoliths. The olivine (ol) and orthopyroxene (opx) fields and ol–opx tie-line (dashed grey line) are based on mineral data from Lanzarote. Blue and red symbols and arrows show the results of mantle melting experiments (Green et al., 2004Go). The blue arrows connect source (H, fertile Hawaiian pyrolite; T, Tinaquillo depleted lherzolite; MM3, lherzolite) and residual compositions (per cent melting given by blue numbers) in ‘first-stage melting’ experiments, aimed at modeling residues after extraction of MORB melts. The red arrows connect to residues of ‘second-stage’ experiments (additional 10%, H2O–CO2-fluxed partial melting of a residual chromite-bearing harzburgite at 1· 5 GPa and 1320–1350°C). (See text for discussion.)

 

Figure 4
View larger version (36K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 4. Relationships between Fo contents and Al2O3 in the cores of coexisting olivine and orthopyroxene porphyroclasts, respectively, and cr-number–mg-number in the cores of equidimensional spinel grains in ultra-refractory OI1-u (upper panels) and more fertile OI1-f (lower panels). The ranges covered by OI2 from the various islands are indicated by colored dotted lines. Data sources are as follows. Canary Islands: Neumann (1991Go), Neumann et al. (1995Go, 2002Go), Wulff-Pedersen et al. (1996Go) and Abu El-Rus et al. (2006Go); Madeira: Munha et al. (1990Go); Azores: Johansen (1996Go); Cape Verde: Ryabchikov et al. (1995Go), Bonadiman et al. (2005Go) and Shaw et al. (2006Go); Kerguelen: Grégoire et al. (1997Go) and Delpech (2004Go); Samoa: Hauri & Hart (1994Go); Hawaii: Goto & Yokoyama (1988Go), Sen (1988Go) and Bizimis et al. (2007Go); Loihi seamount: Clague (1988Go); Tahiti: Tracy (1980Go) and Qi et al. (1994Go). Grey fields: compositional ranges of minerals in MORP (data sources: Hamlyn & Bonatti, 1980Go; Komor et al., 1990Go; Bonatti et al., 1992Go; Johnson & Dick, 1992Go; Cannat et al., 1995Go; Snow & Dick, 1995Go; Arai & Matsukage, 1996Go; Dick & Natland, 1996Go; Niida, 1997Go; Ross & Elthon, 1997Go; Stephens, 1997Go; Hellebrand et al., 2002Go; Brunelli et al., 2003Go, 2006Go; Hellebrand & Snow, 2003Go; Seyler et al., 2003Go). Because the ol–opx partition coefficient for Mg/(Mg + Fe) is close to unity for MORP (Seyler et al., 2003Go) we have used [Mg/(Mg + Fe)]opx instead of Fool where olivine data are not available for the MORP. Stars marked ‘exp’ indicate the composition of minerals in the experimental ‘second-stage’ residue of Green et al. (2004Go). (See text for discussion.)

 
Many OI2-u xenoliths also plot close to the opx–ol tie-line (Fig. 3), although the majority have lower SiO2 and MgO, and higher CaO–FeO* contents than the OI1-u xenoliths and show wide ranges in TiO2 and Na2O. Compared with the OI1-u rocks, the OI2-u rocks show a somewhat wider range in (Al2O3)opx, a similar range in cr-numbersp, but tend towards lower mg-number (Fig. 4). Estimated equilibration temperatures for OI1-u xenoliths are 873–1209°C with a mean of 1024°C (Table 2). The OI2-u xenoliths cover a similar range (862–1215°C) but have a significantly higher average, 1068°C.

Cores of OI1-u orthopyroxene porphyroclasts (opx1) are strongly depleted in REE (YbN = 0·09–0·41, LaN = 0·002–1· 2) and in light relative to heavy REE (LREE and HREE, respectively; Neumann et al., 1995Go, 2004Go; Fig. 5), which is typical for mantle rocks that represent residues after extensive partial melting (e.g. Johnson et al., 1990Go; Neumann et al., 2004Go). A tendency for enrichment in LREE relative to the middle REE (MREE) is most probably due to contamination (Neumann et al., 2002Go).


Figure 5
View larger version (30K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 5. Rare earth element patterns (normalized to Primordial Mantle; McDonough & Sun, 1995Go) for orthopyroxene porphyroclasts in (a) ultra-refractory OI1-u. Data sources: Canary Islands: Neumann et al. (2002Go, 2004Go); Cape Verde: Bonadiman et al. (2005Go) and C. Bonadiman (unpublished data, 2008); Kerguelen: Grégoire et al. (2000bGo) and Delpech (2004Go); (b) fertile OI1-f from Cape Verde (Bonadiman et al., 2005Go). Shaded field: orthopyroxene porphyroclasts in MORP (data sources: Tartarotti et al., 2002Go; Brunelli et al., 2003Go; Hellebrand et al., 2005Go).

 
In contrast to the LREE-depleted opx1 in the OI1-u xenoliths, cpx2 in the same rocks shows a wide range in REE patterns and concentrations, ranging from LREE-depleted in the clinopyroxenes with the lowest HREE, through U-shaped and flat patterns, to LREE-enriched patterns in clinopyroxenes with the highest HREE (YbN = 0·54–1· 1, LaN = 0·22–128; Fig. 6).


Figure 6
View larger version (30K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 6. REE patterns for clinopyroxenes in OI1. (a) OI1-u from the Canary Islands (Neumann et al., 2002Go, 2004Go), Cape Verde (Bonadiman et al., 2005Go; C. Bonadiman, unpublished data, 2008), Kerguelen (Mattielli et al., 1999Go; Grégoire et al., 2000aGo, 2000bGo; Moine et al., 2000Go; Delpech, 2004Go), and Samoa (Hauri et al., 1993Go). (b) OI1-f from Cape Verde (Bonadiman et al., 2005Go), Hawaii (Sen et al., 1993Go; Yang et al., 1998Go) and Tahiti (Qi et al., 1994Go).

 
Madeira
Mantle xenoliths from Madeira comprise refractory deformed harzburgites and lherzolites (OI-u; Fo90–91) and less deformed dunites, websterites, wehrlites and clinopyroxenites (Fo77–88; Munha et al., 1990Go). Harzburgites and lherzolites are granular (most common) to porphyroclastic with deformed olivine and exsolved orthopyroxene porphyroclasts in a fine-grained matrix of ol + opx + sp ± cpx; minor amounts of phlogopite may be present interstitially or as inclusions in clinopyroxene (in lherzolite).

OI1-u xenoliths from Madeira (Munha et al., 1990Go) have lower (Al2O3)opx (1· 5–3· 6 wt %) than orthopyroxene in MORP, but fall within the depleted range of MORP with respect to spinel compositions (cr-number: 0·26–0·55) and Fool (90·4–91·2; Fig. 4). Temperature estimates for the OI1-u xenoliths are 891–1017°C with a mean of 936°C (Table 2); OI2-u xenolith give the wider range of 892–1182°C, with a mean of 1037°C.

The Azores
Mantle xenoliths from the Azores comprise Mg-Cr-rich, porphyroclastic spinel harzburgites (<2 vol. % clinopyroxene) and lherzolites (≤15 vol. % clinopyroxene) as well as granular Fe–Ti-rich spinel wehrlites, clinopyroxenites and dunites (Johansen, 1996Go; E. R. Neumann, unpublished data, 2007). Harzburgites predominate over lherzolites. The oldest generation of minerals in the harzburgites and lherzolites is irregular, deformed porphyroclasts of olivine and orthopyroxene with clinopyroxene exsolution lamellae, and occasional large spinels. The matrix appears to consist of two generations of neoblasts. The older neoblast generation consists of equigranular domains of polygonal neoblasts of olivine (occasionally with undulatory extinction) and rare unexsolved orthopyroxene; the younger generation is made up by fine-grained, glass-rich domains of irregular neoblasts of olivine, orthopyroxene, clinopyroxene, and spinel, without undulatory extinction (Johansen, 1996Go). Clinopyroxene occurs partly as small interstitial grains, partly as poikilitic grains. All Azores xenoliths are classified as OI2.

The Azores xenoliths are dominated by fertile whole-rock major element compositions, but are depleted with respect to (Al2O3)opx (2·1–2·3 wt %) and cr-numbersp (0·51–0·72; Fig. 4). Fool falls within the range 90·0–91·2 in all samples, with the exception of one with Fool of 86·5 (Fig. 4). Equilibration temperatures are high (1003–1232°C, mean: 1135°C; Table 2). The textures of these rocks suggest that the contrast between whole-rock and mineral chemistry is the result of recent melt infiltration into a refractory peridotite protolith that in general has not had time to affect the cores of porphyroclasts and large spinels.

Cape Verde
Xenoliths from Sal in the Cape Verde archipelago have been described by Ryabchikov et al. (1995Go), Bonadiman et al. (2005Go) and Shaw et al. (2006Go). The xenoliths have protogranular to porphyroclastic and mylonitic textures and comprise refractory harzburgites without primary clinopyroxene (≤3 % cpx) as well as mildly depleted spinel lherzolites with primary clinopyroxene (8–18 vol. % cpx). Many samples show extensive evidence of reaction (e.g. glass-bearing sieve-textured rims on clinopyroxenes and spinels) and rims of ol + sp + cpx + glass on orthopyroxene. Ryabchikov et al. (1995Go) also reported phlogopite in some samples. Glass is common, both interstitially and as small inclusions in olivine and orthopyroxene. Bonadiman et al. (2005Go) proposed that the strongly refractory spinel harzburgites and the mildly depleted spinel lherzolites represent two unrelated suites. Both suites include OI1 and OI2 samples (Fig. 2b, d and Fig. 1f, respectively). The presence of two distinct xenolith suites is supported by the relationship between modal proportions of clinopyroxene and olivine (Fig. 7). Within overlapping ranges of olivine the lherzolites (OI-f) define a trend of strongly decreasing clinopyroxene with increasing olivine, whereas the harzburgites (OI-u) show roughly constant, low clinopyroxene contents. Equilibration temperatures are given in Table 2.


Figure 7
View larger version (24K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 7. Relationships between modal proportions of clinopyroxene and olivine in mantle xenoliths from Cape Verde (data sources: Bonadiman et al., 2005Go; C. Bonadiman, unpublished data, 2008). The different positions and slopes of linear regression lines R1 and R2 calculated for the harzburgites and lherzolites, respectively, strongly support suggestions by Doucelance et al. (2003Go) and Bonadiman et al. (2005Go) that the harzburgites and lherzolites represent two unrelated rock suites.

 
The OI-u harzburgites are strongly refractory with whole-rock major element contents falling close to the opx–ol tie-line (e.g. 44·9–47·9 wt % MgO, 0·56–1· 1 wt % CaO, <0·8 wt % Al2O3; Fig. 3). Somewhat higher TiO2 in a few samples is probably due to minor contamination.

Mineral major element data are available only for one OI1-u harzburgite, which has Cr-rich spinel (cr-number = 0·55), Fo-rich (91·3) olivine and Al2O3-poor orthopyroxene (2·6 wt %; Fig. 4). Cores of orthopyroxene porphyroclasts in OI1-u have low HREE contents and show strongly LREE-depleted to U-shaped REE patterns (YbN = 0·04–0·29, LaN = 0·02–0·35; Fig. 5). Clinopyroxene in OI1-u shows LREE-depleted to LREE-enriched patterns (YbN = 0·7–4·5, LaN = 0·3–12·2; Fig. 6).

The Cape Verde OI1-f lherzolites are relatively fertile with 39·7–45·5 wt % MgO, 1· 2–4·6 wt % CaO and 0·5–2·0 wt % Al2O3 (Fig. 8); cr-numbersp is ~0·3, Fool = 87–89 and (Al2O3)opx = 4·1–6·0 wt % (Fig. 4). Orthopyroxene porphyroclasts in OI1-f are depleted in HREE and have moderately LREE-depleted patterns (YbN = 0·23–0·36, LaN = 0·03–0·5; Fig. 5), whereas OI1-f clinopyroxene have upward-convex REE patterns (YbN = 0·9–1· 5, LaN = 1· 7–2· 0; Fig. 6).


Figure 8
View larger version (26K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 8. Major element variations in OI1-f from Cape Verde (Bonadiman et al., 2005Go; C. Bonadiman, unpublished data, 2008), Hawaii (Jackson & Wright, 1970Go; Sen, 1987Go; Goto & Yokoyama, 1988Go; Sen & Leeman, 1991Go; Bizimis et al., 2003Go) and Tahiti (Tracy, 1980Go; Qi et al., 1994Go). Opx and ol fields and opx–ol tie-lines (grey dashed lines) as in Fig. 3. Blue and red symbols and arrows show the results of mantle melting experiments by Green et al. (2004Go). Light grey field, OI1-u.

 
Kerguelen
Mantle xenoliths from Kerguelen belong to the OI-u group and the suite is dominated by protogranular (OI1-u, Fig. 2a) and poikilitic harzburgites (OI2-u, Fig. 1c). Some OI2 samples are cut by thin mafic veinlets (of, e.g. pyroxenite, glimmerite, hornblendite), which may show reactions against spinel or orthopyroxene (e.g. Mattielli et al., 1996Go; Grégoire et al., 1997Go, 2000aGo, 2000bGo; Moine et al., 2000Go; Delpech, 2004Go). In OI1-u, clinopyroxene is generally present as small dispersed, interstitial neoblasts (cpx2), sometimes associated with opx and vermicular spinels. In OI2, clinopyroxene also occurs in symplectites, in or along veinlets, and in reaction zones around orthopyroxene and spinel. A few poikilitic OI2 samples contain phlogopite and amphibole. Trace amounts of apatite, feldspar, rutile, chromite, ilmenite and armalcolite have been observed in glass-bearing domains or veinlets.

All the OI1-u and most of the OI2-u xenoliths fall close to the opx–ol tie-line in most major element diagrams (Fig. 3). OI1-u have orthopyroxenes with low Al2O3 contents (1· 8–2· 4 wt %), high cr-numbersp (0·42–0·61) and high, uniform Fool of 91· 0–91· 8 (Fig. 4). Minerals in OI2-u tend towards more fertile compositions. OI1-u xenoliths give lower equilibration temperatures (933–1104°C, average 1014°C; Table 2) than OI2-u (960–1193°C, average 1062°C).

Pyroxene REE patterns indicate minor interaction with REE-enriched melts or fluids (e.g. Mattielli et al., 1996Go, 1999Go; Grégoire et al., 1997Go, 2000aGo, 2000bGo; Delpech, 2004Go). Most OI1-u orthopyroxenes show low concentrations in HREE and MREE/HREE <<1, but some have irregular REE patterns (e.g. YbN = 0·03–0·23, LaN = 0·011–0·93; Fig. 5). Clinopyroxenes in OI1-u xenoliths show a range in REE patterns from LREE-depleted, through U-shaped, to LREE-enriched, with YbN = 0·13–0·97 and LaN = 0·013–21 (Fig. 6).

Grande Comore
Mantle xenoliths from Grande Comore comprise spinel lherzolites, dunites and wehrlites; no harzburgites have been reported (Coltorti et al., 1999Go). The lherzolites have protogranular to porphyroclastic textures, and consist of large, deformed porphyroclasts of olivine and orthopyroxene and small, strain-free grains of clinopyroxene and spinel, which occur interstitially or as inclusions in orthopyroxene (Ludden, 1977Go; Coltorti et al., 1999Go). The primary textures have been overprinted by pyrometamorphic textures (e.g. melt infiltration), resulting in reactions leading to formation of significant amounts of clinopyroxene, olivine and spinel at the expense of orthopyroxene. Melt–rock interaction processes in the Grande Comore lherzolites are also recorded in grain-to-grain compositional differences within single samples, and zoning (Hauri et al., 1993Go; Hauri & Hart, 1994Go). All Grande Comore xenoliths are classified as OI2.

Despite the reaction textures, the Grande Comore OI2 xenoliths have low (Al2O3)opx (0·8–3·4 wt %), high cr-numbersp (0·4–0·7; Fig. 4) and moderate Fool (90·0–91· 3), classifying them as OI2-u. Estimated equilibration temperatures fall in the range 965–1139°C, with a high mean temperature of 1109°C (Table 2).

Samoa
Mantle xenoliths from Savai'i (Samoa) are dominated by strongly refractory spinel harzburgites with protogranular to porphyroclastic (most common) textures (Hauri et al., 1993Go; Hauri & Hart, 1994Go) and belong to the OI-u group. Symplectites are common in the porphyroclastic rocks. Clinopyroxene occurs mostly along the edges of orthopyroxene + spinel symplectites, but is also found as small interstitial grains rimming orthopyroxene and olivine. In some rocks, melt–rock interaction led to interstitial glass + clinopyroxene ± apatite-bearing patches in various stages of reaction with adjacent porphyroclasts (Hauri et al., 1993Go). Three harzburgites with porphyroclastic textures (≤0·5% cpx) have been classified as OI1-u.

Whole-rock major element data are available only for OI2-u harzburgites; these fall close to the opx–ol tie-line (not shown). Mineral compositions are typical of OI1-u (1· 3–2· 6 wt % (Al2O3)opx; cr-numbersp 0·49–0·70; Fool 90·5–91· 4; Fig. 4). OI2-u xenoliths tend towards more fertile compositions, particularly less Mg-rich spinel and olivine. OI1- and OI2-u rocks give roughly similar equilibration temperatures (983–1142°C with average 1063°C and 1045–1097°C with average 1072°C, respectively; Table 2).

Samoan OI1-u clinopyroxenes show a wide range in REE contents (YbN = 0·045–8·6, LaN = 0·011–2·5; Hauri et al., 1993Go; Hauri & Hart, 1994Go; Fig. 6) both between and within samples. Hauri & Hart (1994Go) found a strong correlation between REEcpx and textures.

Hawaii
The Hawaiian mantle xenoliths belong to the mildly depleted OI-f group. They range in rock types from spinel- and spinel–plagioclase-bearing harzburgites and lherzolites, to dunites, wehrlites, websterites and pyroxenites, some of which contain garnet (e.g. Jackson & Wright, 1970Go; Sen, 1987Go; Goto & Yokoyama, 1988Go; Sen & Leeman, 1991Go; Bizimis et al., 2003Go). Textures range from porphyroclastic to coarse-grained and fine-grained granular. Clinopyroxene-rich spinel lherzolites are common (up to 24 vol. % cpx; e.g. Jackson & Wright, 1970Go; Sen, 1987Go, 1988Go; Goto & Yokoyama, 1988Go; Sen & Leeman, 1991Go; Bizimis et al., 2003Go). Primary clinopyroxene porphyroclasts have spinel or orthopyroxene exsolution lamellae (e.g. Sen, 1988Go).

The whole-rock major element compositions of all the Hawaiian harzburgites and lherzolites plot far away from the opx–ol tie-lines in Fig. 8; Hawaiian OI1 xenoliths include some of the lowest MgO (37–43 wt %), and highest CaO (1·1–3·9 wt %), TiO2 (≤0·43 wt %), Al2O3 (≤4 wt %) and Na2O (≤0·7 wt %) contents reported from ocean islands. OI1-f mineral compositions are also relatively fertile [1·7–4·9 wt % (Al2O3)opx, cr-numbersp 0·09–0·55, Fool 88·5–92·0; Fig. 4]. There is a strong tendency for lower Fool in the OI2-f compared with the OI1-f. Equilibration temperatures are similar for OI1-f (914–1205°C with a mean of 1004°C) and OI2-f (881–1161°C, mean 1005°C; Table 2).

A large amount of clinopyroxene REE data from Hawaiian peridotites has been published, but no orthopyroxene REE data. Clinopyroxene in one OI1-f plagioclase–spinel harzburgite and four OI1-f spinel lherzolites from Oahu exhibit smooth REE patterns with strong LREE depletion (YbN = 3·4–4·7, LaN = 0·02–1· 0; Yang et al., 1998Go; Sen et al., 2005Go; Fig. 6). Minor metasomatism is indicated by increasing La and Ce from cores to rims in a couple of samples.

In contrast to the OI-f xenoliths from the Hawaiian islands, mantle xenoliths from the Loihi Seamount comprise dunites and spinel harzburgites, but no lherzolites (Clague, 1988Go). The harzburgites have cataclastic textures. OI-u compositional characteristics are reflected in low (Al2O3)opx (1· 5–2·0 wt %), high cr-numbersp (0·53–0·75; Fig. 4) and high Fool (90·6–92·3; Fig. 4). Estimated temperatures for the OI1-u are even higher than in Hawaiian OI1-f, 1139–1182°C, with a mean of 1163°C (Table 2).

Tahiti
Tahiti mantle xenoliths are porphyroclastic to equigranular, mildly refractory OI-f spinel lherzolites (most abundant) and harzburgites (Tracy, 1980Go; Qi et al., 1994Go). Clinopyroxene occurs partly as large porphyroclasts, partly in vermicular intergrowths with spinel and olivine. Some clinopyroxenes have spongy glass-bearing rims, interpreted as being a result of metasomatism (Qi et al., 1994Go). No other evidence of metasomatism has been given by Tracy (1980Go) or Qi et al. (1994Go). Samples with spongy glassy rims have been classified as OI2-f, and the rest of the Tahiti samples as OI1-f.

Like the Hawaii xenoliths, most OI1-f from Tahiti have lower MgO and higher SiO2, CaO, FeO* and Al2O3 than OI1-u harzburgites, but they have moderate TiO2 and Na2O contents (≤0·12 and ≤0·05 wt %, respectively; Fig. 3). They also have relatively fertile mineral compositions with (Al2O3)opx 2·3–5·6 wt %, cr-numbersp 0·12–0·41 and Fool 84·6–90·7 (Fig. 4). OI1-f equilibration temperatures are 996–1108°C with a mean of 1058°C (1093°C for the single OI2-f sample; Table 2). Clinopyroxenes in OI1-f have uniform, mildly LREE-enriched REE patterns (YbN = 2·2–4·0, LaN = 1· 6–3· 4; Fig. 6).


    DISCUSSION
 TOP
 ABSTRACT
 INTRODUCTION
 COMPOSITIONAL VARIATIONS
 DISCUSSION
 CONCLUSIONS AND IMPLICATIONS FOR...
 SUPPLEMENTARY DATA
 REFERENCES
 
Different ocean islands show significant contrasts in their OI1 mantle xenolith populations. Xenoliths from the Canary Islands, Madeira, Kerguelen, Samoa and Loihi seamount are strongly refractory (OI1-u and OI2-u), whereas those from Hawaii and Tahiti are mildly refractory (OI1-f and OI2-f). The textural and compositional characteristics of each group are summarized in Table 1. Cape Verde xenoliths appear to comprise both categories. Although no data on OI1 xenoliths are available for the Azores and Grande Comore, the low (Al2O3)opx, high cr-numbersp and high Fool in many OI2 samples suggest that ultra-refractory OI1-u domains may also be present in the upper mantle beneath these islands. Below we summarize the characteristics of OI-u and OI-f xenoliths, compare the OI-u with MORP, and speculate on the origin and evolution of OI-u. Furthermore, we provide some estimates of the proportion of the uppermost mantle that might consist of ultra-refractory peridotites and discuss some implications for geodynamics.

Ultra-refractory OI peridotites (OI-u)
As shown in Fig. 3, the whole-rock major element compositions of OI1-u and OI2-u xenoliths (the Canary Islands, Madeira, Cape Verde, Kerguelen, Samoa and Loihi) essentially reflect different modal proportions of olivine and orthopyroxene; clinopyroxene generally makes up <2 vol. % and occurs as secondary neoblasts (Figs 1 and 2). OI1-u have high cr-numbersp, high Fool, low (Al2O3)opx and low HREE concentrations in pyroxenes (Table 1; Figs 4–6GoGo). Their ultra-refractory compositions strongly suggest that they are residues after partial melting beyond the stability of primary clinopyroxene. This is in agreement with the degrees of partial melting proposed, for example, for Samoa (33–42% melting; Hauri & Hart, 1994Go) and the Canary Islands (25–30% melting; Neumann et al., 2004Go). Furthermore, the OI1-u compositions are similar to residues formed in two-stage melting experiments (Green et al., 2004Go; Fig. 3) in which a residual, dry, chromite-bearing harzburgite with a composition close to MORP was melted by an additional 10% under H2O–CO2-fluxed conditions at 1· 5 GPa and 1320–1350°C.

The secondary origin of clinopyroxene in OI1-u and OI2-u xenoliths from the Canary Islands is reflected in their REE characteristics. Two OI1-u samples with mildly LREE-depleted cpx patterns have opx/cpx REE ratios close to the range of published equilibrium opx/cpx REE ratios in spinel peridotites (Fig. 9), supporting petrographic evidence (Fig. 2) that these clinopyroxenes (cpx2) formed through exsolution from orthopyroxene porphyroclasts (opx1), followed by recrystallization. Many OI2-u, in contrast, have opx/cpx REE ratios that fall significantly below published REE partition coefficients (Fig. 9), implying disequilibrium between the pyroxenes. We interpret these low ratios as the result of a much stronger influence of LREE-enriched fluids or melts on clinopyroxene than on orthopyroxene; some of the most LREE-enriched clinopyroxenes may have formed directly from LREE-rich infiltrating melts.


Figure 9
View larger version (38K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 9. Opx/cpx REE ratios for selected samples from the Canary Islands, compared with published partition coefficients for pyroxenes in equilibrium in spinel harzburgites and lherzolites (Eggins et al., 1998Go; Garrido et al., 2000Go; Green et al., 2000Go; Hellebrand et al., 2005Go). Continuous lines, OI1-u; dashed lines, OI2-u. The opx/cpx ratios for OI1 and some OI2 fall within the range of the experimental data, indicating equilibrium between the two pyroxenes. The ratios obtained for the most REE-rich OI2 clinopyroxenes fall far below the experimental range, indicating REE enrichment in clinopyroxene relative to orthopyroxene.

 
The less refractory nature of the OI2-u lherzolites does not imply that they represent pristine mantle or mantle domains that have undergone lower degrees of partial melting than the OI1-u harzburgites. Instead, we interpret the OI2-u as being formed by interaction between melts and OI1-u protoliths. Evidence that interaction between melts and refractory mantle is an important mode of formation of lherzolites comes from abyssal peridotites, peridotite massifs, ophiolites and mantle xenoliths (e.g. Hellebrand et al., 2002Go; Grégoire et al., 2003Go; Müntener & Piccardo, 2003Go; Simon et al., 2003Go; Le Roux et al., 2007Go; Godard et al., 2008Go), and from experimental work (Yaxley & Green, 1998Go; Yaxley, 2000Go; Bulatov et al., 2002Go).

Mildly refractory OI peridotites (OI-f)
The OI-f xenoliths (Hawaii, Tahiti, the lherzolite suite from Cape Verde) are significantly more fertile than the OI1-u series [lower whole-rock MgO and higher CaO–FeO*–TiO2–Al2O3 contents, higher (Al2O3)opx, lower cr-numbersp and lower Fool, common presence of primary clinopyroxene; Figs 4, 6 and 8; Table 1]. A primary origin of clinopyroxenes in many xenoliths from Hawaii is supported by their strongly LREE-depleted REE patterns (Yang et al., 1998Go; Fig. 6). The OI1-f peridotites appear to have undergone a lower degree of partial melting than OI1-u. Yang et al. (1998Go) estimated about 2–8% melting of a depleted mantle source for a group of spinel peridotites from Oahu. The OI1-f xenolith suites from Hawaii, Tahiti and Cape Verde indicate the presence of mildly refractory domains beneath these islands; the upper mantle beneath Hawaii–Loihi and Cape Verde includes both strongly refractory and more fertile domains.

Comparison between OI1-u and MOR peridotites
Peridotites that are dredged and drilled at ‘normal’ segments along mid-ocean ridges and at fracture zones (MORP) make up a large part of the available oceanic peridotite database and are often regarded as representative of all oceanic lithospheric mantle. We therefore explore here how MORP compare with OI1-u peridotites. We restricted the MORP samples (recalculated to 100% dry) presented in Fig. 10 to harzburgites and lherzolites (using modal compositions as criteria). However, as modal data are not available for all samples, it is likely that some of the rocks with the lowest SiO2 contents (<42 wt %) are in fact dunites, even though they are classified as harzburgites in the literature. The samples that Seyler & Bonatti (1997Go) and Tartarotti et al. (2002Go) described as strongly infiltrated by melts are excluded from Fig. 10.


Figure 10
View larger version (36K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 10. Variations in bulk-rock major element compositions of MORP from various ridge–transform systems (data sources: Miyashiro et al., 1969Go; Aumento & Loubat, 1971Go; Dick, 1989Go; Blum, 1990Go; Cannat et al., 1995Go; Snow & Dick, 1995Go; Casey, 1997Go; Niu & Hékinian, 1997aGo; Stephens, 1997Go; Brandon et al., 2000Go; Lee et al., 2003Go; Seyler et al., 2003Go; Niu, 2004Go; Harvey et al., 2006Go; Godard et al., 2008Go). Grey field, OI1-u xenoliths.

 
MORP show wide ranges in whole-rock and mineral major element compositions. Most samples have 35–49 wt % MgO, 40–49 wt % SiO2, <0·1–4·5 wt % CaO and 0·3–4·5 wt % Al2O3 (Fig. 10). (Al2O3)opx falls in the range 1·7–6·1 wt %, cr-numbersp is 0·12–0·57, and in most MORP Fool is 90–91 (Fig. 4). It is generally accepted that MORP do not represent pristine residues after extraction of mid-ocean ridge basalts, but have been chemically modified by open-system melt migration and crystallization, melt–rock interaction by reactive porous flow, metamorphism, serpentinization and marine weathering (e.g. Elthon, 1992Go; Hékinian et al., 1993Go; Cannat et al., 1997Go; Casey, 1997Go; Niu, 1997Go, 2004Go; Niu & Hékinian, 1997aGo; Asimow, 1999Go; Baker & Beckett, 1999Go; Tartarotti et al., 2002Go). The combination of these processes explains much of the strong compositional heterogeneity observed in MORP. The low CaO/Al2O3 ratios exhibited by many MORP (ranging from about 10 to <0·01; Fig. 11) are probably due to leaching of Ca during serpentinization (e.g. Bach et al., 2004Go; Palandri & Reed, 2004Go). OI1-u xenoliths have CaO/Al2O3 ratios close to 1· 0, which is typical of harzburgites and lherzolites that have not been subjected to serpentinization (Boyd, 1996Go).


Figure 11
View larger version (22K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 11. Variations in CaO/Al2O3 ratios of (a) MORP and (b) OI1-u. Data sources as in Figs 3 and 10. MORP show a strong tendency for lower CaO/Al2O3 ratios than OI1-u, probably as a result of serpentinization.

 
Because of post-melting processes, the origin of clinopyroxene in MORP has been a subject of debate. Elthon (1992Go) claimed that MORP experienced partial melting to the exhaustion of clinopyroxene. He proposed that ≤2% clinopyroxene may be explained by granular exsolution from orthopyroxene, whereas proportions >2% most probably formed from infiltrating melt. Other workers (e.g. Dick & Bullen, 1984Go; Edwards et al., 1996Go; Seyler & Bonatti, 1997Go; Tartarotti et al., 2002Go; Seyler et al., 2003Go; Niu, 2004Go; Brunelli et al., 2006Go) have presented extensive petrographic and geochemical evidence that a significant proportion of the clinopyroxene found in MORP is primary, implying melting within the stability field of ol + opx + cpx. However, these researchers have acknowledged that some clinopyroxene has formed through melt infiltration.

On average OI1-u harzburgites have higher MgO, lower (Al2O3)opx and higher cr-numbersp than MORP and are thus significantly more refractory (Fig. 4). Furthermore, OI1-u has no residual clinopyroxene, whereas residual clinopyroxene is present in many MORP. Orthopyroxenes in MORP (Tartarotti et al., 2002Go; Brunelli et al., 2003Go; Fig. 5) are, on average, less depleted in REE than those in OI1-u (e.g. YbN 0·17–1· 4 in MORP; YbN <0·05–0·41 in OI1-u). Olivine in MORP has very uniform compositions with average Fo contents between 89·6 (Central Indian Ridge) and 90·7 (Mid-Atlantic Ridge; Fig. 12), whereas most cores of olivine porphyroclasts in OI1-u have Fo contents between 91 and 92 (e.g. Hierro, Lanzarote, Kerguelen, Loihi). OI-u xenolith suites that include a significant proportion of samples with hydrous minerals or reaction textures (OI2-u) show a wider range in Fo contents and lower Fo averages (e.g. La Palma, Tenerife, Cape Verde) than OI1-u suites. The tails towards lower Fo can be explained by metasomatic overprinting with some addition of Fe in OI2-u. Clear peaks at Fo contents between 92 and 93 in some islands with heterogeneous olivine compositions (e.g. Gran Canaria and Samoa) reflect the ultra-refractory nature of the protolith (OI1-u). Olivines in OI-f from Hawaii and Tahiti show wider ranges in Fo, and lower averages than OI1-u.


Figure 12
View larger version (41K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 12. Fo contents of cores in olivine porphyroclasts in OI spinel peridotite xenoliths compared with those in MORP. Olivine porphyroclasts in OI1-u (dry) xenoliths show peaks in Fo contents between 91 and 92 (e.g. Hierro, Lanzarote, Cape Verde, Kerguelen). Islands with a high proportion of samples with hydrous minerals (OI2-u) show wider ranges in Fo contents and lower averages (e.g. La Palma, Tenerife). Olivine in the fertile OI-f from Hawaii and Tahiti shows lower ranges in Fo contents. Olivine in MORP has very uniform Fo contents with average Fo between 89·6 and 90·6. The dashed line shows the MORP average of 90·5. Data for OI peridotites are from Jackson & Wright (1970Go), Tracy (1980Go), Clague (1988Go), Goto & Yokoyama (1988