- © 2009 E. Schweizerbart’sche Verlagsbuchhandlung, D-70176 Stuttgart
Water contents of clinopyroxene and orthopyroxene in mantle peridotites from various xenolith occurrences in intraplate settings (both oceanic and continental) were determined by Fourier Transform Infrared Spectroscopy (FTIR). Samples from the following localities were studied: Sal Island (Cape Verde Archipelago); Baker Rocks and Greene Point (Northern Victoria Land, Antarctica); Panshishan and Lianshan (Subei Basin, Eastern China). They represent well-known localities where detailed petrographical and geochemical studies have already been carried out or areas which are currently under investigation. The water incorporated in these pyroxenes is low (cpx, 37–399 ppm; opx, 9–166 ppm) or very low (as in Greene Point, Antarctica; cpx, 5–16 ppm; opx, 9–16 ppm). Within each population there is no clear correlation with melting parameters such as MgO contents in any single mineral. Results are compared with the available literature data on water contents in mantle pyroxenes, which include peridotites from on-craton (hosted by kimberlitic-type magmas) and off-craton (hosted by alkaline basic magmas), as well as subarc mantle settings. The “relatively dry” (cpx, 140–528 ppm; opx, 38–280 ppm) sub-arc mantle xenoliths (Peslier et al., 2002) are shown to be wetter than the intraplate (off-craton) xenoliths. Cratonic mantle pyroxenes are only represented by a few determinations on garnet peridotites and eclogite from Kaapvaal and Colorado Plateau. They record the highest water contents (cpx, 342–1012 ppm; opx, 180–491 ppm) so far measured in mantle pyroxenes from various tectonic settings. Despite the limited data set, the indication that the cratonic mantle is strongly hydrated is compelling. Rehydration of cratonic mantle may be related to plate subduction (i.e., Colorado Plateau) or alternatively to metasomatic enrichment (i.e., Kaapvaal Craton). However, assuming that the water content was initially very low, metasomatic events in intraplate settings, similar to those presented here, would not be able to produce a significant increase in pyroxenes water content. According to our own data and those reported in the literature, it appears that substantial rehydration may instead occur at convergent plate margins.
Hydrogen is one of the most pervasive elements in the lithosphere. It is a mobile, reactive component which can cause major changes in the chemical, and physical properties of the Earth’s crust and mantle, thus strongly affecting their rheological behaviour (van der Lee et al., 2008; Yang et al., 2008a). The behaviour of this element along with its distribution with respect to tectonic settings is far from being fully understood. Most of the mantle petrologists’ interest in hydrous components has, so far, been focused on the volatile reservoirs in the upper mantle and on the role these components may play in understanding and characterizing mantle processes, such as melting and metasomatism.
Hydrogen, in the chemical form of OH and H2O, enters the structure of major minerals which are considered as being nominally anhydrous. Initial studies were focused on water in quartz because of its weakening effect upon the mechanical strength of this mineral (e.g., Paterson, 1982). Hydroxide groups have subsequently been found in other nominally anhydrous minerals, such as feldspars, nephe-line, garnets, sillimanite (and other aluminosilicates), rutile and zircon (Yesinowski et al., 1988; Rossman & Smyth, 1990; Xia et al., 2000). These molecules and ions are structurally bound in definite sites with distinct orientations and often persist in the minerals even at temperatures as high as 1200 °C (Aubaud et al., 2008).
After the early speculations of Fyfe (1970) – later reviewed by Ackermann et al. (1983) – who stated that OH-bearing anhydrous minerals might be a significant reservoir for water in the Earth’s mantle, numerous studies were undertaken on the H2O storage capacity of peridotitic minerals, namely olivine, orthopyroxene, clinopyroxene and garnet (Bell et al., 1995, 2003; Peslier et al., 2002). Measurements of OH concentrations in these nominally anhydrous minerals (NAMs) indicate that a large quantity of “water”, amounting to about one-third of the water in the oceans, is stored within the upper mantle (Ingrin & Skogby 2000). In this respect the upper mantle may represent the most important “water” reservoir on Earth. These mantle minerals can be brought up to the Earth’s surface as xenoliths or xenocrysts in volcanics, probably not in chemical equilibrium with the surrounding melt (Nazzareni et al., 2008; Wade et al., 2008). Geochemical evidence supports the retention of primary hydrogen concentrations in mantle minerals (Bell & Rossmann, 1992), and these concentrations may also be modified during eruption, especially by reduction-oxidation reactions in iron-rich mantle minerals (Skogby & Rossman, 1989; Ingrin & Skogby, 2000).
Olivine (ol) is the most abundant phase within the upper mantle. Therefore, measurements of water contents in mantle olivine are essential in order to model the Earth’s mantle H2O budget. However, diffusion profiles of hydrogen across olivine grains in mantle xenoliths suggest that olivine can lose significant proportions of its water during transportation to the surface (Peslier & Luhr, 2006). This fact, together with the crystallographic inability to incorporate key trace elements (i.e., alkalis, High Field Strength Elements, Rare Earth Elements), makes olivine of limited utility for interpreting mantle melting and magma differentiation processes. By contrast, the similarity of observed water content in pyroxenes with estimates of that expected for pyroxenes in equilibrium with mantle melts (i.e., Bell et al., 2004), together with estimates of the total amount of hydrous species stored in the mantle (200–550 ppm H2O; Bell & Rossman, 1992) suggests that the OH contents of pyroxenes in mantle xenoliths are preserved during the journey to the surface (Peslier et al., 2002). Moreover clinopyroxene, particularly in spinel facies conditions, is the most suitable mantle phase that can incorporate trace elements, allowing to investigate the relationships between water content and the main petrological processes, namely melting and metasomatism.
Besides T-P-X, many factors, such as water and oxygen fugacity, crystallographic charge-coupled substitutions and diffusive properties of hydrogen within the mineral structures, play an important role in determining the concentration of hydrous species preserved in a mineral (Johnson et al., 2002; Johnson & Rossmann, 2003). Petrologists are gathering an increasing number of data from pyroxenes in intraplate (cratonic and off-cratonic) and suprasubduction settings in order to evaluate whether the water content in NAMs may be related to the geological evolution, thus somehow representing a fingerprint of the various processes which have affected the lithosphere through time.
To reach this goal we present water measurements carried out on mantle orthopyroxene (opx) and clinopyroxene (cpx) from various xenolith occurrences in intra-plate settings (both oceanic and continental), thus providing a large data-set on which to base more robust models.
Samples from the following localities were studied: Sal Island (Cape Verde Archipelago; Bonadiman et al., 2005); Baker Rocks and Greene Point (Northern Victoria Land, Antarctica; Coltorti et al., 2004; Perinelli et al., 2006); Panshishan and Lianshan (Subei Basin, Eastern China; Hao et al., 2008). They represent very well-known localities where detailed petrographical and geochemical studies have already been carried out, or areas which are currently under investigation.
Results are then compared with the available literature data on water content in mantle pyroxene from cratonic (hosted by kimberlitic-type magmas), off-cratonic (hosted by alkaline basic magmas) and sub-arc mantle settings.
2. Sample description
The investigated samples are spinel-facies peridotites from three different intraplate mantle sections. The xenoliths are fresh and range from fertile lherzolite to cpx-bearing harzburgite (<5 % modal cpx). A brief description of each locality is provided below.
2.1. Cape Verde (AC)
The selected samples are anhydrous spinel-bearing fertile lherzolites [bulk mg# (MgO/(MgO+FeO) mol%), 89.4–90.2; Al2O3 bulk rock >2.1 wt%; modal cpx between 12 % and 18 %] included in alkaline basalts from Sal Island. They show protogranular textures and are devoid of any metasomatic texture, i.e., spongy pyroxenes, glassy patches, etc. The absence of garnet lherzolites restricts the P-T mantle section to between 1.3 and 2.1 GPa and 990–1100 °C, with T-P upper limit just below the spinel-garnet reaction boundary (Nickel et al., 1985; Robinson & Wood, 1998; Bonadiman et al., 2005). A detailed study of the pyroxene mineral chemistry (major and trace elements) and in situ 187Os–188Os isotopic determination on sulphide, reveal that these lherzolites represent a portion of sub-continental lithospheric mantle (SCLM) partially re-equilibrated from garnet to spinel conditions and left stranded by the drifting of the African plate during the opening of the Central Atlantic ocean (Bonadiman et al., 2005; Coltorti et al., 2008).
These xenoliths were entrained in alkaline basalts of the McMurdo Volcanic Group (Northern Victoria Land) and were collected from two localities: Greene Points (GP) and Baker Rocks (BR) (Coltorti et al., 2004, 2006; Perinelli et al., 2006). The selected samples are protogranular in texture without evidence of superimposed secondary metasomatic assemblages. GP is characterized by the exclusive presence of anhydrous spinel-bearing lherzolites (bulk mg#, 90.5–91.5; Al2O3 bulk rock <1.3 wt%; modal cpx <9 %), whereas in the BR locality amphibole-bearing spinel lherzolites also occur. Amphibole (modal contents up to 4 %) can be present as disseminated grains in the peridotite matrix or in veins (bulk mg# 89.5–90.5; Al2O3 bulk rock 2.1–1.5 wt%; modal cpx ≃9–12 %; Coltorti et al., 2004). From these geochemical features, a slightly more refractory character of the Greene Point mantle section may be inferred. Pressure estimates for both groups indicate a relatively shallow lithospheric mantle (P: 0.9–1.6 GPa). The BR group presents lower temperatures (T: 800–1020 °C), fitting the McMurdo geotherm (Berg et al., 1989); whereas GP xenoliths show a higher thermal state (920–1080 °C; Coltorti et al., 2006; Perinelli et al., 2006).
2.3. Eastern China
Fourteen mantle xenoliths from Panshishan (PSS) and 18 from Lianshan (LS) were selected for H2O measurements in opx and cpx. These xenoliths occur in Neogene alkali basalts of the Subei basin, located east of the Tanlu fault zone within the Yangtze block (Chen & Peng, 1988; Chen et al., 1994; Reisberg et al., 2005; Xu et al., 2008).
Nearly all xenoliths are spinel-bearing lherzolites with rare harzburgites, as are most peridotites from eastern China (Fan et al., 2000; Xu et al., 2000). No hydrous phases were observed. The LS lherzolite compositions range from highly fertile (bulk mg# 90.0–91.3; Al2O3 bulk rock 5.6–7.1 wt%; 15–17 modal cpx) to rather depleted (mg# 91.3–92.8; Al2O3 2.1–5.0 wt%; ~5 vol% cpx) and show cpx/opx ratios of 0.5–0.6. The PSS lherzolites have a larger range of fertile compositions (mg#, 89.2–91.5; Al2O3, 4.3–7.1 wt% and 10–17 vol% cpx) and, consequently, higher cpx/opx ratios (up to 0.7). Xenoliths from both localities are coarse-grained and mostly protogranular or protogranular/porphyroclastic in texture, with the average grain size being larger for the Panshishan xenoliths (~6 mm for ol and opx). Metasomatic textures, mainly spongy clinopyroxene, are rarely observed. The two groups of xenoliths present similar equilibration temperatures (PSS = 760–1025 °C; LS =726–997 °C). If one assumes a geotherm similar to that of the nearby Nushan locality (Reisberg et al., 2005; Xu et al., 2008; Yang et al., 2008b), these temperatures suggest a derivation from the shallow lithosphere, mostly below 40 km depth.
3. Analytical methods
Double-polished thin sections with a thickness of about 0.2 mm were prepared for the FTIR investigation (resin was used during the polishing treatment). Infrared spectra were obtained at wavelengths from 650 to 6000 cm−1 using a Nicolet® 5700 FTIR spectrometer coupled to a Continμum microscope at the School of Earth and Space Sciences, University of Science and Technology of China (USTC) in Hefei. The samples were measured by unpolarized radiation with an IR light source, KBr beam-splitter and liquid-nitrogen cooled MCT-A detector. A total of 128 or 256 scans were counted for each spectrum at a 4 or 8 cm−1 resolution. Optically clean, inclusion- and crack-free areas, usually centred in the core region of each grain, were selected for the measurements with apertures of 30 × 30 or 50 × 50 μm. Large and fresh grains were selected for the H-profile analysis.
An accurate measurement of hydrogen-species in anisotropic minerals requires orientation of single crystals and use of polarized IR radiation (Libowitzky & Rossman, 1996). This is a very demanding technique and so an alternative technique using unpolarized determinations on a statistically significant number of individual grains (e.g., 8–25) for each mineral in the same sample was chosen. Assuming that crystal orientation is randomly distributed within each sample (as evidenced by the variability in our FTIR results), an average value was used (Kovacs et al., 2008). Water contents were calculated by the modified form of Beer–Lambert law:
where Δ is the integral absorption area (cm−1) of absorption bands, I is the integral specific absorption coefficient (ppm−1·cm−2), c is the contents of hydrogen species (ppm H2O), t is the thickness of the section (cm), and γ is the orientation factor discussed by Paterson (1982). In this study, the integral region was 3000–3800 cm−1, and the integral specific coefficients for cpx and opx were taken from Bell et al. (1995, 2003). Thickness was measured with a digital micrometer and reported as an average of 30~40 measurements covering the whole section. Baseline corrections were carried out with a spline-fit method by points outside the OH-stretching region. Uncertainties in the obtained results derive from: (1) Unpolarized light. This is the main uncertainty during the analysis but it is estimated to be mostly less than 10 % considering the applied procedures and the recent approach of Kovacs et al.(2008). (2) Baseline correction. Some strongly rising, non-linear baselines may be an intrinsic part of the spectrum in the OH region. These baselines commonly arise from Fe2+ and may arise from silicate overtones in thick samples. A major, subjective source of uncertainty in IR measurements of OH in minerals remains the choice of baseline. Error introduced by different baseline corrections, e.g. spline-fit, polynomial-fit or slightly changing the points during the fitting, is usually <5 %. (3) Variation of thin sections thickness. This was less then than 6 % centred on the average value for each sample. (4) Absorption coefficients. There are slight differences between absorption coefficients in cpx and opx in our samples and those used to determine the mineral specific absorption coefficients (Bell et al., 1995), due to their different compositions and densities: these variations are estimated to be generally <10 %. On the whole, the total uncertainty, summing each single error, is estimated <30 %.
4. Hydrogen species and water content
The H2O measurements in the analyzed pyroxene grains exhibit several absorption bands in the typical OH-stretching vibration region (3000–3800 cm−1). The representative infrared spectra are shown for cpx and opx in Fig. 1a, b⇓, respectively. The absorption bands are 3440, 3520 and 3620 cm−1 for cpx and 3310, 3410, 3510 and 3580 cm−1 for opx. These absorption bands are in good agreement with those previously estimated by Skogby & Rossman (1989), Skogby et al. (1990) and Peslier et al. (2002) using polarized light on oriented grains. Analyses of a single pyroxene grain do not show consistent variations between core and rim, strongly suggesting that diffusion and loss of H during the xenolith’s ascent is not significant in cpx and opx. Analogous results are reported by Peslier et al. (2002) and Demouchy et al. (2006) for multiple measurements on polarized samples.
Coexisting olivine generally does not display absorption bands, but in two samples (PSS15 and PSS17) where very weak absorption bands were observed. Generally, olivine water content in peridotite xenoliths is very low (mostly <10 ppm; Bell & Rossman, 1992; Peslier & Luhr, 2006; Grant et al., 2007), due to (i) high partition coefficients between pyroxene and olivine (KdH2O Opx/Ol 10–40; KdH2O Cpx/Ol 10–80 for experimental and natural peridotite systems, Koga et al., 2003; Hauri et al., 2004; Hirschmann et al., 2005; Peslier & Luhr, 2006; Grant et al., 2007), (ii) rapid diffusion of H (H diffuses 1–3 orders of magnitude faster in olivine than in pyroxene, Ingrin & Blanchard, 2006; and references therein) and loss during ascent. However, a large difference between Kd obtained from experimental and natural samples exists, most probably due to H loss in olivine during the ascent (Peslier & Luhr, 2006). Taking all these facts into account, one may conclude that the contribution of olivine to the total water budget is negligible and this budget is therefore almost entirely controlled by the modal abundance of opx and cpx.
From the obtained spectra the water concentrations of pyroxenes were calculated and are summarized in Table 1⇓, together with their Al2O3 and MgO contents.
To the best of the authors’ knowledge, data reported for Cape Verde represent the first measurements of water contents in pyroxenes from intraplate oceanic mantle lithosphere. Water contents range from 37 to 109 ppm in cpx and 9 to 43 ppm in opx. Water contents in clinopyroxenes and orthopyroxenes from GP (Antarctica) are from 5 to 22 ppm and from 9 to 16 ppm respectively, whereas BR amphibole-bearing xenoliths have H2O concentrations varying from 82 to 399 ppm and from 39 to 166 ppm in cpx and opx, respectively (Table 1⇑, Fig. 2⇓). The BR pyroxenes appear to retain a much higher water content than those of GP, which have the lowest content so far analyzed. In the two occurrences from China, H2O contents vary between 64 and 183 ppm in cpx and between 16 and 61 ppm in opx from PSS, while cpx and opx from LS contain 37 to 102 ppm and 13 to 45 ppm, respectively (Table 1⇑, Fig. 2⇓). Both PSS and LS samples show a restricted range of water contents in pyroxenes, largely overlapping those of Nunshan samples (Yang et al., 2008b) which lie at the lower end of the range (Fig. 2a, b⇓).
Taking into account the modal abundances of opx and cpx of the investigated localities, a whole-rock water content of 8–26 ppm for Cape Verde, 2–8 ppm for GP, 14–50 ppm for PSS and 7–25 ppm for LS can be estimated. This calculation cannot be performed for BR samples due to the presence of amphibole, whose OH content is under investigation. It has to be noted that irrespective of the variability in the cpx and opx modal abundances, the calculations reproduce the relative variations of water contents shown by pyroxenes (Fig. 2a, b⇑) in relation to different tectonic settings. For example GP maintain the lowest whole-rock water content.
The average partitioning coefficient for H2O between cpx and opx for all opx/cpx pairs for the studied samples and from literature is (2.6 ± 0.5) slightly but not significantly higher than the values reported in literature for experimentally determined partition coefficients (2.2 ± 0.5) (Table 1⇑, Fig. 3⇓; Hirschmann et al., 2005).
5. OH content and pyroxene geochemical features
In this section we discuss and compare pyroxene water contents with the geochemical parameters that are commonly used to characterize mantle processes such as melting and enrichment. For this purpose all of the xenoliths selected for this study are texturally free of superimposed pyrometamorphic textures (glassy patches, spongy pyroxenes, etc.) related to recent metasomatism. In particular we focussed our attention on the MgO and Al2O3 contents in pyroxene and REE abundances of clinopyroxenes.
We preferred to use the MgO content of pyroxenes instead of the most popular geochemical parameter mg# to represent the mantle depletion degree, because MgO is a better indicator of the degree of melt extraction with respect to chemical-physical components (e.g., Palme & Nickel, 1985; Lee et al., 2003; Niu, 2004). Because water behaves incompatibly during mantle processes (e.g., Dixon et al., 1988; Michael, 1988; Michael, 1995), we can expect the water content to decrease with increasing MgO. In this respect we could explore the processes which enrich or deplete the lithospheric mantle in hydrous species with time. The water dissolved in intraplate mantle pyroxenes is low (or very low as in GP, Antarctica) and, among each population, no clear correlation with the MgO content in a single mineral apparently exists (Fig. 4a, b⇓). Only a weak negative correlation emerges if we take into account the entire off-craton intraplate pyroxene population (in particular clinopyroxene): the most depleted pyroxenes (highest MgO) tend to have the lowest water content (Fig. 4a, b⇓). Peslier et al. (2002) described the correlations between whole rock compositions and water contents in anhydrous minerals of the spinel facies in the mantle-wedge (positive correlations for Al2O3, TiO2 and Na2O and negative for MgO). Despite this general relationship, we did not observe a direct correlation between MgO and H2O contents in individual grains of mantle wedge pyroxenes (Fig. 4a, b⇓). On the other hand, a broad positive correlation between pyroxene water contents and Al2O3 manifested itself in sub-arc mantle wedge samples (Fig. 5a, b⇓; Peslier et al., 2002; Grant et al., 2007), and in the off-craton intraplate mantle (Fig. 5a, b⇓; Grant et al., 2007; Yang et al., 2008b), with different slopes due to the higher values of water contents in the mantle wedge with respect to the intraplate setting (Fig. 5a, b⇓). This correlation is also supported by experimental results of Rauch & Keppler (2002) which show that water solubility increases with Al2O3 content (up to 1 wt%) in synthetic enstatite. For anomalously high Al2O3 contents however, the water budget of the mantle may dominate the water absorption in opx (Mierdel et al., 2007).
Peslier et al. (2002) considered their samples to be “relatively dry” and, in our opinion correctly, interpreted the “relatively dry” sub-arc mantle wedge to be a possible effective medium through which subducted water is transported from slabs toward the surface. However, if we compare these results with the large data set presented in this study, it is evident that the mantle wedge pyroxenes are definitely wetter than the intraplate (off-craton) mantle pyroxenes (Fig. 3a, b⇑).
Based on the available data, the MgO and Al2O3 contents in both pyroxenes are not correlated to the water contents of the Colorado Plateau (associated with spinel or garnet) nor of the kimberlitic mantle pyroxenes (Fig. 4⇑, 5⇑). Bell & Rossmann (1992) found that OH concentrations in garnet from the Monastery kimberlite (South Africa) were correlated with their mg# but this seems not to be replicated in the pyroxenes from the same samples nor from other South African kimberlite sites.
Orthopyroxene crystals synthesized in a water-saturated environment are compositionally similar to those of natural mantle orthopyroxene (Al2O3 contents in the range of 0.67–5.79 wt%) but the total integrated OH absorbance is about 10 times higher than in natural samples and they show no correlation between Al and water contents as recorded in recent experiments (Mierdel et al., 2007).
To further explore possible relationships between water contents and different geochemical parameters, we looked at pyroxene trace element contents. Unfortunately this comparison was only possible for our own data, and is restricted to off-craton mantle. We limited the comparison to the REE (in clinopyroxenes), since they are commonly used to investigate melting (HREE) and/or enrichment processes (LREE). No correlations were found, since even those clinopyroxenes homogeneous in major elements very often have a large variation in REE. We are currently acquiring more water and trace element measurements for the same pyroxene grain in order to investigate a possible relationship between these two parameters in relation to metasomatic events.
6. Concluding remarks
In this section, pyroxene water contents from the five intraplate (off-craton) localities are compared with the available data from literature for mantle pyroxenes from different tectonic settings.
Clino- and orthopyroxenes from intraplate mantle domains are on average relatively poor in water and, among each population, quite homogeneous. In contrast, pyroxenes from the mantle wedge, expected to have low water contents in response to oxidizing environments (Peslier et al., 2002), produce a larger variation in water contents (cpx, 140–528 ppm; opx, 38–280 ppm) and significantly higher values than those recorded for oceanic and continental intraplate lithospheric mantle (Fig. 3⇑).
The few H2O measurements of Antarctic pyroxenes do not show conclusive evidence of mantle hydration. However, significant distinctive characteristics emerge between the two Antarctic mantle sectors and when these are compared to other localities. Water contents of GP pyroxenes are very homogeneous and are the lowest among the entire data set (Table 1⇑; Fig. 3⇑). The GP population has water contents at least 2 times lower than those from other intraplate off-craton xenolith populations, comparable only with the water contents measured in the abyssal peridotite minerals (Peslier, 2007). The low water contents of abyssal peridotites were explained as the consequence of the slow adiabatic decompression of oceanic mantle beneath ridges. In fact both melting (H is incompatible in the basaltic system) and decreasing H solubility in mantle minerals with decreasing pressure, combine to deprive abyssal peridotite minerals of water (Aubaud et al. 2004, 2008; Keppler & Bolfan-Casanova 2006). In this respect, the water contents of GP pyroxenes could reflect a slow uplift of this mantle section accompanied by melting.
The FTIR study on the Colorado Plateau mantle xenoliths (Li et al., 2008) revealed that the lithospheric mantle is wetter than typical continental lithosphere, with water contents higher than those recorded in the mantle-wedge environment (Fig. 3a⇑). Li et al. (2008) speculated that hydration inducing lithospheric thinning might have further modified the western North American lithosphere and thus, together with slab truncation (Bird, 1984, 1988) accounted for its present-day structure. Because the effect of water on lithospheric thickness is magnified at greater depths and at higher water contents, a subduction-induced hydration could lead to the weakening, thinning and eventual recycling of craton keels on a global scale (van der Lee et al., 2008).
Pyroxenes of cratonic mantle xenoliths (brought up to the surface by kimbelite-like melts), represented by few determinations on Kaapvaal and Colorado Plateau garnet peridotites and eclogite (Bell & Rossman, 1992; Grant et al., 2007; Li et al., 2008), record the highest water content so far measured in mantle pyroxenes in various tectonic settings, including the mantle wedge (cpx, 342–1012 ppm; opx, 180–491 ppm; Fig. 3⇑). This (unexpected) result manifests itself only when the cratonic mantle is compared with worldwide off-cratonic setting. Cratons are the ancient (Precambrian) cores of continents that have, for the most part, remained tectonically quiescent over a timescale of billions of years. The lithospheric mantle roots beneath cratons are found to be thicker and more refractory than most Phanerozoic (off-craton) lithosphere (e.g., Hawkesworth & Norry, 1983; Menzies & Hawkesworth, 1987; Nixon, 1987). These depleted mantle roots, together with the overlying crust, are thought to be chemically buoyant and strong enough to resist convective disruption on a timescale of billions of years (e.g., Jordan, 1978; Pollack, 1986; Pearson et al., 1995; Griffin et al., 1999; Griffin et al., 2003b; Lee, 2006). Part of this lithospheric strength is attributed to the cooler thermal state of cratons, but low temperatures alone are insufficient to prevent disruption by mantle convection over prolonged periods of geological time, as thermal re-equilibration should also occur within a timescale of billions of years (Pollack, 1986; O’Reilly et al., 1996). Cratonic mantle roots may also be intrinsically strong because they were dehydrated during long-term melt depletion processes, resulting in a dry residue compared to the “damp” fertile mantle which makes up most of the off-craton mantle (Pollack, 1986; Hirth & Kohlstedt, 2004). However, the data presented in this contribution indicate that the cratonic mantle is strongly hydrated (at least Kaapvaal and the Colorado Plateau). If dehydration does enhance lithospheric strength and ensure longevity, the question that arises is how cratonic mantle can ever be rehydrated, and if it can, whether such rehydration can ultimately lead to weakening or even destabilization of the lithosphere.
As far as the Colorado Plateau cratonic mantle rehydration is concerned, it could be related to subduction of the Farallon plate as suggested by Li et al.(2008). The Kaapvaal sublithospheric mantle (including eclogite) could have acquired its water content from metasomatic events which affected the highly depleted cratonic mantle during Cretaceous (Griffin et al., 2003a; Kobussen et al., 2008). If this is the case, it is very likely that unmetasomatized harzburgitic or even dunitic (Beyer et al., 2006) Archean mantle have not been measured yet, and that further work on cratonic mantle samples has to be carried out. However, assuming that the pristine cratonic mantle was also highly depleted in water, the question of the water’s provenance is still debatable. It appears in fact that metasomatism affecting off-craton intraplate settings does not carry a large quantity of OH. So far, rehydration could only be seen in off-craton settings above a subducted slab, such as Colorado Plateau (Li et al., 2008).
The data from intraplate oceanic and continental mantle lithosphere presented in this paper and comparison with the literature data from various tectonic settings suggest that water is more easily retained in cratonic and supra-subduction environments. If this statement is confirmed by additional water measurements in pyroxenes from samples of lithospheric mantle, it could then be concluded that water content variations in the entire upper mantle could be related to geodynamic setting (horizontal variation) rather than depth (vertical variation), as recently suggested by Peslier (2007).
We thank Norman Pearson and Sabrina Nazzareni for providing critical and thoughtful reviews, and guest editor Paola Comodi for the challenging comments. This work was supported by the National Science Foundation of China (90714009). We also thank Ross Angel for the critical reading of the manuscript. YanTao Hao also acknowledges support from Bruker Company© to present this research at the International School of Mineralogy “HP-HT mineral physics: implication for geosciences” held in Bressanone (Italy), February 2008. IUSS (Istituto Universitario di Studi Superiori) of Ferrara University is also acknowledged for providing a three-years PhD fellowship to YanTao Hao.
- Received 1 December 2008.
- Modified version received 21 January 2009.
- Accepted 30 April 2009.