Journal of Petrology Advance Access originally published online on August 5, 2004
Journal of Petrology 2004 45(12):2507-2530; doi:10.1093/petrology/egh039
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Journal of Petrology 45(12) © Oxford University Press 2004; all rights reserved
Geodynamic Information in Peridotite Petrology
DEPARTMENT OF GEOLOGICAL SCIENCES, RUTGERS UNIVERSITY, NEW BRUNSWICK, NJ 08903, USA
RECEIVED SEPTEMBER 10, 2003; ACCEPTED APRIL 21, 2004
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONRESIDUAL MANTLE PERIDOTITEARCHEAN KOMATIITES AND THEIR...INFERRING PRIMARY MAGMA...ARCHEAN PRIMARY MAGMASCONCLUSIONSSUPPLEMENTARY DATAREFERENCES |
|---|
Systematic differences are observed in the petrology and major element geochemistry of natural peridotite samples from the sea floor near oceanic ridges and subduction zones, the mantle section of ophiolites, massif peridotites, and xenoliths of cratonic mantle in kimberlite. Some of these differences reflect variable temperature and pressure conditions of melt extraction, and these have been calibrated by a parameterization of experimental data on fertile mantle peridotite. Abyssal peridotites are examples of cold residues produced at oceanic ridges. High-MgO peridotites from the Ronda massif are examples of hot residues produced in a plume. Most peridotites from subduction zones and ophiolites are too enriched in SiO2 and too depleted in Al2O3 to be residues, and were produced by melt–rock reaction of a precursor protolith. Peridotite xenoliths from the Japan, Cascades and Chile–Patagonian back-arcs are possible examples of arc precursors, and they have the characteristics of hot residues. Opx-rich cratonic mantle is similar to subduction zone peridotites, but there are important differences in FeOT. Opx-poor xenoliths of cratonic mantle were hot residues of primary magmas with 16–20% MgO, and they may have formed in either ancient plumes or hot ridges. Cratonic mantle was not produced as a residue of Archean komatiites.
KEY WORDS: peridotite; residues; fractional melting; abyssal; cratonic mantle; subduction zone; ophiolite; potential temperature; plumes; hot ridges
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONRESIDUAL MANTLE PERIDOTITEARCHEAN KOMATIITES AND THEIR...INFERRING PRIMARY MAGMA...ARCHEAN PRIMARY MAGMASCONCLUSIONSSUPPLEMENTARY DATAREFERENCES |
|---|
Samples of peridotite from the Earth's mantle occur as xenoliths in kimberlite and alkali basalt, tectonic crustal emplacements (i.e. Alpine, orogenic, massif and seamount peridotite), and abyssal peridotite dredged from oceanic spreading centers. Observed mineralogical and geochemical variations have been interpreted to reflect magma removal followed by metasomatism (e.g. Maaloe & Aoki, 1977
; Frey & Prinz, 1978; Herzberg, 1993; Parkinson & Pearce, 1998; Downes, 2001). Metasomatism can affect both trace and major element abundances (Parkinson & Pearce, 1998; Downes, 2001), but only the latter will be considered in this paper. The major element composition of residual peridotite depends on both the amount and the composition of the melt extracted, relative to more fertile peridotite, and these depend on the geodynamic environment in which they form. For example, hot mantle peridotite at high potential temperatures appropriate to hotspots will begin to melt deeper and sometimes more extensively than colder mantle, which melts below oceanic ridges. Relative to basalts that characterize oceanic ridges, primary magmas that form at hotspots tend to be enriched in MgO and FeO and depleted in Al2O3; they solidify to rocks called picrites and komatiites. Mass balance requires that extraction of primary magmas with a wide range of compositions must leave behind residues that reflect this diversity. Partial melting processes must, therefore, account for at least some of the variability in the major element geochemistry and petrology of mantle peridotites. It would clearly be desirable if this process could be inverted to allow extraction of geodynamic information from the petrology of mantle peridotite samples.
Liquid compositions produced in controlled melting experiments on fertile mantle peridotite (KR-4003; Walter, 1998) at a wide range of pressures offer new possibilities for constraining the geochemistry of its melting residues. KR-4003 is a fertile peridotite (Walter, 1998) that can be modeled by about 1% mid-ocean ridge basalt (MORB) extraction from the primitive McDonough & Sun (1995) composition. Liquid compositions produced in these experiments were modeled with mass balance equations that are appropriate for both equilibrium and fractional melting at mass fractions of melting that span the 0·0–1·0 range (Herzberg & O'Hara, 2002). Results previously reported for FeO and MgO have been extended to Al2O3 and SiO2 in this paper. The compositions of complementary residues produced by both equilibrium and fractional melting have been modeled by mass balance. These are then compared with natural peridotite samples from the sea floor near oceanic ridges and subduction zones, the mantle section of ophiolites, massif peridotites, and xenoliths of cratonic mantle in kimberlite. It will be shown that peridotite residues produced in geodynamic environments that differ in potential temperature can often be distinguished, but complexities can occur owing to subsequent stages of melt–rock reaction and addition of cumulus minerals.
Komatiites of Archean age left behind residues whose lithological identity is important for a comprehensive understanding of their origin. An examination is made of the mass balance requirements of computed residues of Archean komatiites with the goal of testing the commonly accepted theory that they are similar to kimberlite-hosted xenoliths of cratonic mantle. This paper concludes with an identification of the FeOT and MgO contents of primary magmas that are complementary to residues of mantle peridotite, and these are compared with primary magmas that have been estimated for MORB and various hotspot occurrences. It is shown that many samples of cratonic mantle of Archean and Proterozoic ages are likely to be hot residues of picritic primary magmas that are similar in composition to those forming below Hawaii today.
|
RESIDUAL MANTLE PERIDOTITE |
|---|
|
TOPABSTRACTINTRODUCTIONRESIDUAL MANTLE PERIDOTITEARCHEAN KOMATIITES AND THEIR...INFERRING PRIMARY MAGMA...ARCHEAN PRIMARY MAGMASCONCLUSIONSSUPPLEMENTARY DATAREFERENCES |
|---|
Model residue compositions of fertile mantle peridotite
For the limiting case of equilibrium melting, residue compositions were computed by mass balance from equilibrium liquid compositions of fertile peridotite KR-4003 (Herzberg & O'Hara, 2002
); further details are given in Electronic Appendix 1 (available at http:www.petrology.oupjournals.org). For the case of accumulated fractional melting, the primary magma has a unique residue, but the accumulated liquid is not in equilibrium with that residue; only the final drop of liquid extracted is in equilibrium with the residue. Computed FeO and MgO contents of residues that develop during decompression fractional melting were not reported by Herzberg & O'Hara (2002), but are discussed in Electronic Appendix 1 and illustrated in Fig. 1a. Computed Al2O3 and SiO2 contents of liquids and residues were also not given by Herzberg & O'Hara (2002); they are, however, given in Electronic Appendix 1 and Fig. 1b and c.
|
Peridotite residue compositions are important functions of the pressures of initial and final decompression melting. At constant MgO, increasing pressure of initial melting will produce residues that are lower in FeO (Fig. 1a), higher in Al2O3 (Fig. 1b) and higher in SiO2 (Fig. 1c). When the pressure of initial melting is 2–3 GPa, residues can be strongly depleted in Al2O3 but remain unchanged in FeO. At very high pressures of initial melting, in the 7–10 GPa range, residues can be depleted or enriched in Al2O3, depending on the amount of melt extracted. Melt fraction increases when the pressure of final melting decreases, and this is universally reflected in residues with elevated MgO. Residues almost always have lower SiO2 than the source.
Composition fields for harzburgite and dunite residues at magmatic conditions are also shown in Fig. 1. Additionally, there are numerous residue lithologies with compositions that plot between those for the peridotite source and harzburgite. These are typically spinel and garnet lherzolites, but include lherzolites, garnet harzburgites, and an assemblage with low-Ca pigeonitic clinopyroxene (L + Ol + low-Ca Cpx + Gt) (Herzberg & O'Hara, 2002). Natural peridotite samples typically have petrographic characteristics that correspond to those defined by the compositional fields in Fig. 1. However, small amounts of clinopyroxene, spinel, or garnet observed in peridotites that plot within the field defined by harzburgite can be of exsolution or metasomatic origin (Cox et al., 1987; Boyd et al., 1997).
The pressure at which melting begins increases with rising mantle potential temperature (e.g. McKenzie & Bickle, 1988), and two hypothetical cases are illustrated in Fig. 2a. One is for cold mantle that has a potential temperature of 1320°C and intersects the anhydrous solidus at 2 GPa. The other is an example of hot mantle that has a potential temperature of 1600°C and a pressure of initial melting at 5 GPa. For each case, fractional melting terminates at three pressures indicated by the closed circles, chosen to illustrate residue lithology and composition. Each closed circle in Fig. 2b shows the MgO and Al2O3 contents of the simulated residue produced at the conditions of melting shown in Fig. 2a. The three residues produced at each potential temperature define a simple array that is coincident with a line of initial melting pressure. Although the simulation in Fig. 2a is shown only for MgO–Al2O3, similar arrays must be observed in MgO–FeO and MgO–SiO2 space. In the following sections, the simulated residues in Fig. 2 are replaced with the contents of MgO, FeO, Al2O3, and SiO2 in naturally occurring peridotite samples to infer pressures of melting and melt fractions, information that is used to constrain mantle potential temperature. Arrays of peridotite compositions that do not plot along common initial melt pressure lines when viewed in combined MgO–SiO2, MgO–Al2O3, and MgO–FeO space cannot be simple residues of fertile peridotite KR-4003. Examples will be discussed in the following sections.
|
Observed mantle peridotite samples: treatment of data, assumptions, limitations
A brief survey is made of relatively young peridotites as a prelude to an evaluation of cratonic peridotites of Archean age. ‘Young’ refers to peridotites from modern ocean basins (abyssal peridotites), Tethyan ophiolites, and the Ronda orogenic massif. Although obduction and emplacement ages are usually young, Proterozoic Re–Os ages have been reported, considered to reflect a magmatic depletion event (e.g. Reisberg & Lorand, 1995
).
With the exception of peridotites from subduction zones, reported values of FeOT are recalculated to Fe2O3 and FeO using the relation
 | (1) |
for a wide range of peridotite samples. In most cases, recalculated FeO contents, rather than FeOT, are plotted in Fig. 1a even though the computed Fe2O3 content is usually very small. The reason for making this distinction is because the FeO contents displayed in Fig. 1a were computed strictly for iron that is exchangeable between olivine and liquid, i.e. Fe2+ (Herzberg & O'Hara, 2002). Whole-rock peridotite analyses were normalized to 100% anhydrous, and FeO and MgO are plotted in Fig. 1a. Subduction zone peridotites might have considerably more Fe2O3 than that calculated from equation (1), and this is discussed below. All interpretations concerning the pressure of initial and final melting extracted from Fig. 1 assume that the initial source had a composition similar to that of fertile peridotite (i.e. KR-4003; 44·90% SiO2, 4·26% Al2O3, 8·02% FeO, 38·12% MgO). This assumption will be repeatedly tested by combining plots of residue compositions in MgO–SiO2, MgO–Al2O3, and MgO–FeO space.
Unless stated otherwise, a comparison is made of the compositions of observed mantle peridotites with computed residues expected from fractional melting of fertile peridotite rather than equilibrium melting (e.g. McKenzie, 1984; Johnson et al., 1990). However, abyssal peridotite residues are compared with residues expected from both equilibrium and fractional melting, and this is discussed in the following section.
Abyssal peridotites
Abyssal peridotites are fragments of mantle that have been dredged from modern ocean basins (e.g. Dick & Fisher, 1984; Dick, 1989; Johnson & Dick, 1992). Serpentinization is pervasive and whole-rock chemical analysis is compromised (Snow & Dick, 1995). Original whole-rock compositions are reconstructed using primary mineral modes and either analyzed (Dick & Fisher, 1984; Dick, 1989) or calculated (Niu, 1997; Baker & Beckett, 1999) phase compositions. The common database presented by Dick and coworkers has resulted in two very different model whole-rock reconstructions (Niu, 1997; Baker & Beckett, 1999) owing to different assumptions in calculated phase compositions.
The Baker & Beckett (1999) model abyssal peridotite compositions define trends that are mostly coincident with residues of fertile peridotite produced by fractional melting in combined MgO–SiO2, MgO–Al2O3, and MgO–FeO space (Fig. 3). Inferred initial melting pressures are 2–3 GPa, final melting pressures are 0·5–2·0 GPa, and melt fractions ranged from 0·09 to 0·25. They are also similar to residues produced by equilibrium melting as discussed in Electronic Appendix 2 on the Journal of Petrology website (http:www.petrology.oupjournals.org). These similarities do not permit a conclusion to be drawn about the relative importance of equilibrium vs fractional melting (Electronic Appendix 2). The Niu (1997) model defines a trend that is positively correlated in FeO–MgO space (Fig. 3a), inconsistent with a trend expected of simple residues (see discussion of Fig. 2). Niu (1997) and Niu et al. (1997) interpreted the model abyssal peridotites as residues into which cumulus olivine was later added by partial crystallization of MORB. Both models are nearly coincident with residues at initial melting pressures of 2–3 GPa when viewed in MgO–Al2O3 space, but possible olivine addition is revealed by low SiO2 and Al2O3 contents in some cases (Fig. 3b and c). More recent data by Seyler et al. (2003) are consistent with both interpretations (Fig. 3). Fractional melting cannot account for elevated TiO2 and Na2O in abyssal peridotites, a matter that has received considerable attention elsewhere (Elthon, 1992; Asimow, 1999; Baker & Beckett, 1999; Niu & O'Hara, 2003).
|
Pressures of initial melting of 2–3 GPa obtained here are similar to other estimates for abyssal peridotite (Asimow, 1999
), and for MORB generation (McKenzie & Bickle, 1988; Langmuir et al., 1992; Asimow et al., 2001; Herzberg & O'Hara, 2002). Using the T–P diagram in Fig. 2a and adiabatic gradients given by Iwamori et al. (1995) and Herzberg & O'Hara (2002), the potential temperatures can be inferred to be 1300–1450°C, in good agreement with 1280–1400°C for most MORB estimates (Herzberg et al., submitted). These potential temperatures are relatively cold compared with plume occurrences (Herzberg & O'Hara, 2002). Residues of abyssal peridotite are therefore considered ‘cold’.
Peridotites from the Ronda massif
Whole-rock data for peridotites from the Ronda orogenic massif (Frey et al., 1985) define trends that are coincident with model residues formed by initial melting at 2–5 GPa and final melting at 1–2 GPa (Fig. 4). Misfits occur for samples that have been contaminated by mafic layers (Frey et al., 1985), and unpublished data indicate that modification of bulk-rock mg-number is also possible by infiltration metasomatism or melt–rock reaction (Bedini et al., 2003; J. L. Bodinier, personal communication, 2004). Inferred melt fractions from Fig. 4 are 0–0·30, in excellent agreement with previous estimates (Frey et al., 1985). Samples from Ronda with the highest MgO contents exhibit the highest pressures of initial melting (i.e. Po = 3–5 GPa; Fig. 4a–c). Inferred potential temperatures range from about 1450 to 1550°C (Fig. 2a), hotter than abyssal peridotites. These residues are therefore considered ‘hot’. Samples with the lowest MgO contents exhibit initial melting pressures at about 2 GPa, similar to abyssal peridotites.
|
Hot Ronda residues are similar to a model residue composition for the Ontong Java Plateau, computed by mass balance from the calculated primary fractional melt composition presented by Herzberg (2004a)
. The Ontong Java Plateau is a region of thickened crust that might have formed by melting in a Cretaceous plume (Neal et al., 1997; Fitton & Godard, 2004). The initial source is very similar in composition to fertile peridotite KR-4003 (Herzberg, 2004a). A model residue for Hawaii would be equally desirable, but there is some indication that the initial source is not similar to KR-4003 (Herzberg & O'Hara, 2002; Feigenson et al., 2003). Even more desirable would be xenoliths of residues from the active Hawaiian plume. However, xenoliths from Hawaii (Sen, 1987) are more likely to be magmatic cumulates, and lherzolite xenoliths from Oahu are probably pieces of the lithosphere (Yang et al., 1998). The Ontong Java Plateau example is used throughout this paper as a reference model residue formed by fractional melting of a fertile peridotite in a modern plume situation.
The results presented here are in good agreement with the suggestion of Frey et al. (1985) that the Ronda peridotite massif is a fossil plume or mantle diapir. Residues with the highest MgO contents may have been located in the hot core of the plume where melting initiated at the highest pressures. Conversely, residues with the lowest MgO contents may have been located at the periphery or diapir–wall rock interface (Frey et al., 1985) where the initial melting pressure and extent of melting were both at a minimum.
Peridotite xenoliths from the Kaapvaal, Siberian and Tanzanian cratons
Kimberlite-hosted peridotite xenoliths from the Kaapvaal, Siberian and Tanzanian cratons have Re–Os model ages that range from Proterozoic to Archean, with the majority being Archean (Pearson et al., 1995; Carlson et al., 1999; Chesley et al., 1999). Of the low- and high-temperature equilibrated xenoliths, only the low-temperature types are considered here because metasomatism is extensive in the high-temperature types (e.g. Smith & Boyd, 1987; Griffin et al., 1989). Xenolith compositions plotted in Fig. 5 (Boyd & Mertzman, 1987; Boyd et al., 1993, 1997, 1999; Lee & Rudnick, 1999) have been divided into two populations. The first consists of peridotites that display the properties of residues produced by initial melting at 3–5 GPa in combined MgO–SiO2, MgO–Al2O3, and MgO–FeO space (Fig. 5a–c; open cross symbols). All others are too enriched in orthopyroxene to be residues (Fig. 5a–c; filled cross symbols), a conclusion reached previously (Kelemen et al., 1992, 1998; Herzberg, 1993, 1999; Smith et al., 1999). Addition of orthopyroxene will create a whole rock with high apparent initial melting pressures when viewed in MgO–FeO (Fig. 5a) and MgO–SiO2 (Fig. 5c) space, but impossibly low pressures when viewed in Al2O3–MgO space (Fig. 5b). Xenoliths that exhibit contradictory pressures cannot be simple residues. These results are consistent with low vanadium contents, which point to Opx enrichment by either melt–rock reaction or cumulus addition (Lee et al., 2003). Peridotite xenoliths from the Slave Craton (Kopylova & Russell, 2000) also display variable orthopyroxene enrichment, and have not been plotted to preserve clarity. Peridotites from Somerset Island (Schmidberger & Francis, 2001) are discussed as a separate issue in Electronic Appendix 3 (http:www.petrology.oupjournals.org).
|
Orthopyroxene-poor xenoliths from cratonic mantle could have been produced at initial melting pressures of 3–5 GPa and potential temperatures of about 1450–1600°C. Most cratonic peridotites plot in the compositional space defined by harzburgite residues in Fig. 5, in agreement with petrographic observations; the small amounts of garnet and clinopyroxene observed in these samples have been interpreted to be of exsolution or metasomatic origin (Cox et al., 1987
; Boyd et al., 1997). These are hot residues, similar to those for Ronda and the Ontong Java Plateau. Similar pressures of initial melting indicate similar potential temperatures for melt extraction (Fig. 2) in geodynamic settings that range from the Archean to the Cretaceous. Some harzburgite xenoliths differ in displaying higher MgO and higher melt fractions compared with Ronda and the Ontong Java Plateau, but dunite is a rare lithology. Reference to Fig. 5 shows that residues with elevated MgO contents can be produced by lower pressures of final melting, indicating the possible involvement of thinner lithosphere in the Archean at the time of melt extraction.
Peridotites from active subduction zones
Peridotites from active subduction zones have been described from the Itinome-gata back-arc in Japan (Kuno & Aoki, 1970; Aoki & Shiba, 1973), the Cascade back-arc (Brandon & Draper, 1996), the Patagonian back-arc (Laurora et al., 2001), the Luzon arc (Maury et al., 1992), the South Sandwich forearc (Pearce et al., 2000), the Izu–Bonin–Mariana forearc (Parkinson & Pearce, 1998), and the Papua New Guinea forearc (McInnes et al., 2001). Whole-rock geochemical data are summarized in Fig. 6. Most arc peridotites have low Al2O3 and are more orthopyroxene-rich than residues of fertile mantle peridotite. Similarly, many arc peridotites are enriched in SiO2 (Fig. 6c), similar to many cratonic mantle samples (Fig. 5c). Some are enriched in FeOT compared with cratonic mantle, a difference that is discussed below.
|
Peridotites from active subduction zones contain pyroxenes and spinels with high Fe3+, indicative of elevated oxygen fugacities (Wood et al., 1990
; Brandon & Draper, 1996; Parkinson & Pearce, 1998; Parkinson & Arculus, 1999; Pearce et al., 2000). However, vanadium contents reveal a more reducing oxygen fugacity (Canil, 2002; Lee et al., 2003). It has been suggested that vanadium is recording the oxygen fugacity of the residue protolith, and Fe3+ is a record of subsequent melt–rock reaction (Lee et al., 2003). Elevated oxygen fugacity during melt–rock reaction will increase the content of Fe2O3 in arc peridotites relative to those computed with equation (1); accordingly, only total iron contents (i.e. FeOT) are given in Fig. 6a. Some arc peridotites are low in FeOT, comparable with hot residues. Others are unusually iron-rich, probably owing to addition of Fe3+ from melts or hydrous fluids from the subducting slab (Lécuyer & Ricard, 1999; Parkinson & Arculus, 1999). This oxidized metasomatic component might add SiO2 that reacts with peridotite in the mantle wedge to produce orthopyroxene.
Although oxygen fugacities are highly elevated in peridotites from the back-arc occurrences of Itinome-gata and the Cascades (Brandon & Draper, 1996; Parkinson & Arculus, 1999), Fe2O3 might be lower compared with arc locations owing to reduced slab fluid or melt fluxes behind the volcanic front. Indeed, a few peridotites from back-arc occurrences in Japan, the Cascades, and Patagonia are very similar to hot peridotite residues (Fig. 6b and c). These are the only reported peridotites from active subduction zones that display internally consistent melting pressures and melt fractions when viewed in combined MgO–SiO2, MgO–Al2O3, and MgO–FeO space, and they also have the lowest FeOT. Indeed, xenoliths from Patagonia (Laurora et al., 2001) are located about 400 km from the Chile trench and show the least enrichment in SiO2 and FeOT. Clearly, some peridotites from back-arc regions have survived the effects of melt–rock reaction.
Peridotites from subduction zones that have the lowest FeOT and highest mg-numbers might be characteristic of first-stage residues before they become affected by melt–rock interaction. As first-stage residues from subduction zones are similar in composition to hot residues expected in plumes, they might have formed in buoyant oceanic plateaux. Niu et al. (2003) examined this possibility, and they concluded that subduction might be initiated at locations where there is a change from ridge-type to plateau-type oceanic lithospheric mantle. Oceanic plateaux might become buoyant platforms on which new continental crust can accrete (e.g. Jordan, 1978, 1988; Ben-Avraham et al., 1981; Abbott & Mooney, 1995; Albarède, 1998; Herzberg, 1999; Niu et al., 2003). The petrology of arc peridotite residues reported here provides strong support for this class of model. However, buoyant residues might also have been produced in hot ridge type environments during the Archean, an alternate possibility that is examined below.
Silica-rich mantle peridotites from active subduction zones and cratonic mantle
Many peridotites from active subduction zones and cratonic mantle are very similar in displaying orthopyroxene enrichment relative to hot residues (Figs 5 and 6). This is displayed by similar systematics in MgO–SiO2 and MgO–Al2O3 space. There is, however, an important difference as shown in Fig. 7. Opx-rich cratonic mantle tends to be low in FeOT (Fig. 5a). In contrast, some Opx-rich subduction zone peridotites are rich in FeOT (i.e. Itinome-gata), whereas others are not (i.e. Luzon arc, Patagonia back-arc).
|
Orthopyroxene-rich cratonic mantle has been interpreted as a mixture of residues and cumulus orthopyroxene (Herzberg, 1993
, 1999), and as a reaction product of residues with a silica-rich melt (Kelemen et al., 1992, 1998; Simon et al., 2003). These two processes cannot be distinguished in plots of MgO–Al2O3 and MgO–SiO2, but they might produce very different MgO–FeOT systematics. Kelemen et al. (1998) proposed that SiO2 was added to cratonic mantle residues by partial melts of subducted eclogitic basalt and sediment. Recent progress in numerical simulations indicates that elevated mg-number in peridotite might be a possible consequence of the interaction of a basaltic melt with a first-stage residue (Bedini et al., 2003). A melt–rock reaction model is also a suitable explanation for many of forearc peridotites; however, silica might have been added as an oxidized Fe2O3 solute-rich hydrous fluid rather than a melt (Parkinson & Pearce, 1998; Parkinson & Arculus, 1999). Melt–rock reaction models for cratonic and subduction zone mantle must, therefore, differ in terms of the melt or fluid compositions that originate from the slab and mantle wedge; these differences have not yet been explored in forward numerical simulations. If progress cannot be made in simulating the formation of FeO-depleted and Opx-rich cratonic mantle by melt–rock reaction, then alternative models must be preferred (e.g. Herzberg, 1993; Francis, 2003).
Peridotites from Tethyan ophiolites
Plotted in Fig. 8 are peridotite compositions from ophiolites in Greece, Cyprus, and Oman (Menzies & Allen, 1974; Lippard et al., 1986; Godard et al., 2000; Takazawa et al., 2003). Many ophiolitic peridotites differ from abyssal peridotites in being depleted in Al2O3 and enriched in SiO2. They are similar to peridotites from subduction zones; however, SiO2 enrichments are not as extreme. Peridotites in ophiolite sections cannot, therefore, be single-stage residues of fertile peridotite. They also cannot be second-stage residues of a depleted mantle source as shown in Electronic Appendix 4 (http:www.petrology.oupjournals.org). These results favor models of Tethyan ophiolite formation in back-arc basins and subduction zones rather than open oceans (e.g. Robertson, 2002).
|
|
ARCHEAN KOMATIITES AND THEIR RESIDUES |
|---|
|
TOPABSTRACTINTRODUCTIONRESIDUAL MANTLE PERIDOTITEARCHEAN KOMATIITES AND THEIR...INFERRING PRIMARY MAGMA...ARCHEAN PRIMARY MAGMASCONCLUSIONSSUPPLEMENTARY DATAREFERENCES |
|---|
Aluminum-undepleted komatiites
CaO and Al2O3 can be fractionated from each other in komatiites that had augite and garnet in the residue. This is likely to be important for understanding the origin of high CaO/Al2O3 in aluminum-depleted komatiites, as discussed below. However, fractionation is not significant when melting occurs with residual harzburgite or dunite (e.g. Herzberg & O'Hara, 2002
). Therefore, CaO/Al2O3 values for liquids and coexisting harzburgite [L + Ol + Opx] or dunite [L + Ol] residues should be similar to that of the peridotite source, as indicated in Fig. 9. Aluminum-undepleted komatiites have CaO/Al2O3 
1·0 (e.g. Nesbitt et al., 1979) as also shown in Fig. 1, similar to fertile mantle peridotite and its depleted residues (Herzberg, 1993; McDonough & Sun, 1995), but considerable scatter is evident for altered samples with low Al2O3 owing to Ca mobility (e.g. Herzberg, 1992). Consequently, if we consider this parameter alone, the residue of aluminum-undepleted komatiites could be either harzburgite (Nisbet et al., 1977; Nesbitt et al., 1979; Herzberg & O'Hara, 1998) or dunite (Herzberg & O'Hara, 1998; Sproule et al., 2002). To distinguish between these possibilities, an examination is made of FeOT and MgO contents of komatiites.
|
Shown in Fig. 10 are FeOT and MgO contents of
2700 Ma aluminum-undepleted komatiites from Belingwe (Zimbabwe), Kambalda (Australia), Alexo and Pyke Hill (Canada) (Bickle et al., 1975, 1993; Arndt et al., 1977; Nisbet et al., 1977, 1987; Arndt, 1986; Lesher & Arndt, 1995; Lahaye & Arndt, 1996; M. Shore, personal communication, 2004). Olivine was incrementally added to and subtracted from representative sample Z5 from Belingwe (Bickle et al., 1993) using a procedure described in the caption to Fig. 10. Compositions of liquids are shown as the gray and black trajectories, and these are connected to coexisting olivine compositions. The coherent negative sloping trend displayed by aluminum-undepleted komatiites from three continents is coincident with the two trajectories, demonstrating that they can be successfully modeled by olivine fractionation. A parental magma having 28–30% MgO would precipitate olivine with an mg-number = 94, similar to the maximum observed mg-numbers of olivines in the lava flows (Arndt, 1986; Renner et al., 1994; Lesher & Arndt, 1995). These parental magma compositions are in excellent agreement with previous estimates (Arndt, 1986; Lesher & Arndt, 1995).
|
Fe = 0·05 (black trajectory) and 0·15 (gray trajectory). Tie-lines connect liquid compositions in each trajectory with coexisting olivine compositions. Crosses-in-circles, parental magmas that would precipitate olivine with mg-number = 94; these contain 28% MgO for Fe3+/To evaluate the residuum mineralogy, the assumption is made that the estimated parental magma compositions shown in Fig. 10 are identical to the primary magma. Primary magmas that exit the melting regime in the mantle must erupt directly to the surface without crystallization for this assumption to be valid. Primary komatiite magmas that enter sills and fractionate olivine will be transformed to derivative liquids with MgO
28–30%. A comparison can then be made of these primary magmas with the range of possible FeOT and MgO contents in primary magmas that form by equilibrium and accumulated fractional melting of an assumed fertile peridotite source, using the computational procedures described by Herzberg & O'Hara (2002). The fertile source composition is that of McDonough & Sun (1995) with FeOT adjusted to 0·27% Fe2O3 and 7·79% FeO using the method of Canil et al. (1994). Fe2O3 is treated as totally incompatible in olivine and highly incompatible in orthopyroxene (Herzberg & O'Hara, 2002). Results in Fig. 11 show the full range of possible liquid compositions that can be extracted with olivine and harzburgite as residual assemblages. All liquids produced by equilibrium melting are too low in FeOT compared with the primary komatiite magma. The match is improved for the case of accumulated fractional melting where olivine is the only residuum phase (Fig. 11). There are, however, several possible difficulties with this model. The melt fraction is about 0·7, which might be excessive. The residuum olivine would have an mg-number of 98 (Herzberg & O'Hara, 2002, fig. 7b), and dunites having this composition have never been reported. Dunites are a rare lithology in xenoliths of cratonic mantle and, where reported, the most forsteritic olivines have mg-numbers of 93 (Lee & Rudnick, 1999; see Fig. 5a). Komatiites are rare in comparison with basalts in Archean greenstone belts, and it is reasonable to expect that residuum dunites would be even more rare, given that they resided in the mantle; they may also have been subjected to modification by iron-rich magmas in the mantle. Alternatively, dunites having mg-numbers of 98 may not be observed because fractional melting of fertile peridotite may be an inappropriate model.
|
Another model is shown in Fig. 12 for a depleted peridotite composition. The abyssal peridotite source used in Fig. 9 was found to be too depleted to provide any liquid composition that could match the komatiite primary magma. However, successful solutions were found, by trial and error, using a slightly less depleted source having 41·0% MgO, 7·8% FeO, and 0·27% Fe2O3. Total iron contents of depleted and fertile source compositions can be very similar despite the large variations in Al2O3 as shown in Fig. 3 for abyssal peridotite. The major element geochemistry of such a source is similar to a residue produced by about 10% basaltic melt extraction from a McDonough & Sun (1995)
fertile peridotite. A depleted source is required from observed light rare earth element (LREE) and isotopic depletions (Arndt, 1986; Bickle et al., 1993; Chauvel et al., 1993; Lesher & Arndt, 1995; Lahaye & Arndt, 1996). Inspection of Fig. 12 shows that the primary komatiite magma is similar to a liquid that could have formed by 0·5 mass fractions of equilibrium melting. Accumulated fractional melting is not a possible model solution because it would form liquids with FeOT contents higher than that of the komatiite primary magma. The residue would have been pure dunite with an mg-number of 94; harzburgite is not an acceptible residue. Olivines approaching mg-numbers of 94 have been reported for cratonic mantle xenoliths (Boyd et al., 1997; Lee & Rudnick, 1999). However, these are rare and they occur in harzburgites rather than dunites. The lack of dunite residues of aluminum-undepleted komatiites with mg-numbers of either 94 or 98 does not permit a determination to be made about whether melting was equilibrium or fractional.
|
Aluminum-depleted komatiites
As noted above, aluminum-depleted komatiites are characterized by much higher CaO/Al2O3 ratios than aluminum-undepleted komatiites (Fig. 9). High ratios must reflect the source composition when olivine and harzburgite are residues because CaO and Al2O3 fractionation is not significant (e.g. Herzberg & O'Hara, 2002
). Parman et al. (1997) and Grove et al. (1999) have proposed that aluminum-depleted Barberton komatiites with CaO/Al2O3 
1·8 were generated by hydrous partial melting leaving a harzburgite residuum. This residue must also have CaO/Al2O3 1·8 and this value must be similar to that of the source, which is considerably higher than 0·8 for fertile peridotite (McDonough & Sun, 1995) and 0·9 for average depleted mantle peridotite (Herzberg, 1993). This model requires an unusual source composition that is not observed in fertile peridotites (Fig. 13). The few peridotites with CaO/Al2O3 that are comparable with Barberton komatiites are also exceptionally depleted in Al2O3, and cannot be a plausible initial source. Additionally, depletions in heavy rare earth elements (HREE) exhibited by Barberton komatiites are a feature of a garnet-bearing residue rather than harzburgite (Sun & Nesbitt, 1978; Nesbitt et al., 1979). Other difficulties with this model have been discussed elsewhere (Arndt et al., 1998; Arndt, 2003).
|
Barberton alumina-depleted komatiites are similar in composition to experimental anhydrous liquids in equilibrium with olivine, subcalcic clinopyroxene, and garnet [i.e. L + Ol + low-Ca Cpx + Gt] produced in excess of 7 GPa (Fig. 13; Herzberg, 1992
; Walter, 1998). For an assumed anhydrous fertile peridotite source (McDonough & Sun, 1995), aluminum-depleted komatiites with elevated CaO/Al2O3 shown in Fig. 13 must be mass-balanced with residua having low CaO/Al2O3 (i.e. olivine + subcalcic clinopyroxene + garnet; Herzberg, 1992; Walter, 1998). Residual garnet is consistent with the REE characteristics of Al-depleted komatiites that show depletions in the HREE. Residues of aluminum-depleted komatiites have not been reported, and this might again reflect the general rarity of komatiites in Archean greenstone belts.
Archean komatiite residues and cratonic lithospheric mantle: a misconnection
There is a rough complementarity between the major element geochemistry of Archean komatiites and cratonic lithospheric mantle xenoliths, an observation that has led many workers to propose a genetic link (O'Hara et al., 1975; Maaloe & Aoki, 1977; Hanson & Langmuir, 1978; Takahashi, 1990; Canil, 1992; Herzberg, 1993, 1999; Walter, 1998). In particular, cratonic lithospheric mantle is highly depleted in incompatible elements, consistent with it being a residue from high degrees of partial melting. However, with the exception of Herzberg (1999), none of the above workers provided quantitative mass balance calculations with respect to any particular mantle source. Herzberg (1999) noted that equilibrium melting of fertile peridotite produces melts with FeO contents lower than those of Al-undepleted komatiites, a misfit that is reproduced in this study (Fig. 11). The suggestion that this misfit might be explained by some form of fractional melting (Herzberg, 1999) appears to hold (Fig. 11). However, this model might be problematic because it predicts olivines in the residue with an mg-number of 98 (see Herzberg & O'Hara, 2002), and these have never been observed.
Samples of low temperature-equilibrated Archean cratonic mantle are mostly harzburgites with lesser lherzolites (Nixon & Boyd, 1973; Boyd & Mertzman, 1987; Boyd et al., 1993, 1997, 1999; Lee & Rudnick, 1999; Kopylova & Russell, 2000; Hanghøj et al.,2001; Schmidberger & Francis, 2001). Petrographic descriptions provided by these researchers are in good agreement with whole-rock data that plot in the compositional fields defined by harzburgite and lherzolite, not dunite (Fig. 5). The few cases of dunite that have been reported are shown in Fig. 5, and these have olivines with mg-numbers that are typically <93 (Lee & Rudnick, 1999). In contrast, the inferred residue of aluminum-undepleted komatiite is dunite with mg-numbers of 94 and 98 for equilibrium and fractional melting models, respectively. The inferred residue of aluminum-depleted komatiite is an unusual assemblage of Ol + low-Ca Px + Gt, which is expected to re-equilibrate in the subsolidus to garnet harzburgite with minor clinopyroxene. Residues of aluminum-depleted and -undepleted komatiites have never been observed as xenoliths of cratonic mantle peridotite. It is concluded, therefore, that cratonic lithospheric mantle did not form as a residue from Archean komatiite melt extraction. How then do we explain its geochemistry? One way to explore this problem is to seek alternatives to Al-depleted and -undepleted komatiites. This is examined in the next section.
|
INFERRING PRIMARY MAGMA COMPOSITIONS FROM MANTLE PERIDOTITES |
|---|
|
TOPABSTRACTINTRODUCTIONRESIDUAL MANTLE PERIDOTITEARCHEAN KOMATIITES AND THEIR...INFERRING PRIMARY MAGMA...ARCHEAN PRIMARY MAGMASCONCLUSIONSSUPPLEMENTARY DATAREFERENCES |
|---|
Estimates have been made of primary magma compositions for modern MORB and Phanerozoic plume lavas produced by accumulated fractional melting of fertile and depleted sources (Herzberg & O'Hara, 2002
; Herzberg, 2004a, 2004b). Results displayed in Fig. 14a show that there is a simple and linear relationship between MgO and FeOT, and this can be described by the equation
 | (2) |
|
