Introduction

Water in Earth’s interior is held dominantly as trace hydroxyl in nominally anhydrous minerals (NAMs). Relative to its minor abundance, this “deep water” has a disproportionate effect on mantle physical properties, such as its solidus temperature and viscosity, thereby strongly influencing its phase relations and rheological behaviour as well as its geophysical expression (e.g., Kohlstedt 2006; Garber et al. 2018). During metasomatism, depending also on H2O fugacity, water may be added via crystallisation of hydrous minerals (e.g., amphibole, phlogopite), but can also be incorporated in NAMs as trace amounts of OH associated with point defects following diffusion into NAMs from fluids or melts (Demouchy and Bolfan-Casanova 2016). Water, sensu lato, in NAMs is reported as equivalent H2O concentration by weight, and may include intrinsic or structural protons or hydroxyl from lamellae of hydrous minerals (e.g., Stalder 2004; Chen et al. 2011; Moine et al. 2020). Subsequently, water in NAMs from magma-borne xenoliths may be lost upon transport to the surface depending on H diffusivities, which are mechanism-specific (Demouchy and Bolfan-Casanova 2016; Yang et al. 2007; Sundvall and Stalder 2011).

Based on the water contents of the grains with the least detectable IR signal for the typical sample thicknesses reported in the Online Resource, the detection limit is conservatively estimated to be ∼ 2 wt.ppm. Random errors reflecting those related to thickness measurement, baseline correction, possibly undetected contamination by optically invisible sheet silicates and contribution from non-ideal orientation (clinopyroxene), combined with uncertainties in experimentally-determined extinction coefficients are estimated to result in a total uncertainty in mineral c(H2O) of ∼ 40% (50% when uncertainties in bulk rock reconstruction are added). For comparison, the average relative uncertainty for c(H2O) based on one standard deviation for multiple measurements per sample is 11–19%, while the intra-grain heterogeneity, gauged by multiple measurements per mineral fragment is 5–7% (1σ) (Table S4 in Online Resource 3), both well within the estimated total uncertainty.

FTIR imaging

In order to further test sample homogeneity, transmission FTIR imaging was carried out at the Institute of Geological Sciences of the University of Bern using a Bruker Tensor II spectrometer with a globar MIRE light source and a KBr beam-splitter, coupled to a Bruker Hyperion 3000 microscope with a dry air-purged sample chamber. The instrument is equipped with a focal plane array (FPA) detector with 64 × 64 = 4096 liquid-nitrogen-cooled MCT elements on a square array with a pixel size of 2.7 μm × 2.7 μm. A 2 × 2 binning was used resulting in a 5.4 × 5.4 μm pixel resolution, higher signal quality and an improved signal to noise ratio. Spectra were acquired with 8 cm− 1 wavenumber resolution and 64 scans between 900 and 3850 cm− 1. For all acquired FTIR-FPA images, the atmospheric correction and concave rubber band correction with 64 points and four iterations was performed in OPUS® version 8.5. The spectra were then exported and further processed with the in-house developed MATLAB software SpecXY (Gies et al. 2024). The spectra were normalized to 1 cm thickness and integrated absorbance for different images were generated by extracting the chosen part of the signal, performing a linear baseline correction between the integration endpoints, and integrating the corrected spectra. In total, 23 clinopyroxene fragments from K1-2, K1-4, K1-6, K1-10, K1-13, K1-14, K1-16, K5-5 covering clear areas, as well as domains with cracks, inclusions and/or alteration features, were investigated. Grains were selected for imaging based on two criteria: firstly, the largest possible grains with clear areas and, secondly, grains with identifiable grain boundaries, for example, by direct contact with garnet. Two representative examples are shown in Fig. 4, and all imaging in Online Resource 2.

Fig. 4
figure 4

Typical FTIR spectra (vertically offset for graphical reasons) showing a.-c. OH absorption bands for clinopyroxene in clinopyroxene in representative eclogite xenoliths from African kimberlites, for polarisation ||α (red), ||β (green) and ||γ (blue), d.-f. corresponding absorption bands related to lattice overtones in representative eclogite xenoliths from African kimberlites, used to verify crystal orientation, and g. OH absorption bands for garnet in three samples. In a.-c., the wavenumber ranges of three types of absorbance bands (I, II, III) described in the text are shown as grey bars; note the similarity of α and β OH absorption bands. Strong absorbance ||γ relative to ||α and ||β in the Type III region is accompanied by little absorbance ||γ in the other regions (a, c), whereas absorbance ||γ similar to ||α and ||β is accompanied by bands in the Type I and II regions (b)

Full size image

EBSD

In order to estimate whether a crystallographic-preferred orientation (CPO) formed and to visualise internal misorientation indicative of grain boundaries and subgrain formation, EBSD data were acquired – as a preliminary investigation – for two metasomatised samples from Orapa (pyroxenite OE72 and eclogite OE83). The samples were polished to 0.25 micron with diamond and then for 10–15 min with colloidal silica. We employed a JEOL JSM-6490 scanning electron microscope, using an acceleration voltage of 15 kV, beam current of ∼ 8 nA, and a working distance of 20 mm. A Nordlys camera (Oxford Instruments) and the Channel 5 software (Flamenco, Oxford Instruments) were used to acquire the data. Maps were made of some garnet and omphacite grains with a step size of 5–7 μm. Due to the large grain size, the orientations were measured manually in Flamenco, with one measurement per grain. The orientations were plotted in equal area, upper hemisphere pole figures. Because of the typical medium to coarse grain size of xenolithic eclogite (Fig. S4 in Online Resource 1), only < 50 grains could be analysed per mineral. Consequently, the pole figures were not contoured, as more than 100–150 grains need to be measured to get a good representation of the orientations and to confirm the omphacite CPO, but in practice a limited number of grains are exposed in a given section of medium- to coarse-grained eclogite. The orientation is illustrated by three pole figures of [100], (010) and [001] for omphacite, and of {100}, {110} and {111} for garnet.

Results

For measurements along the refractive indices α and β in clinopyroxene in this study, bands with varying absorbance ratios are always recognised at wavenumbers ∼ 3630 − 3600 (Type I) and 3470 –3430 cm− 1 (Type III), whereas bands of Type II (3540 –3510 cm− 1) are weak or not observed. Three types of bands have been commonly described for clinopyroxene (Koch-Müller et al. 2004; Libowitzky and Beran 2006; Huang et al. 2014). The dominant absorption band ||γ is typically linked to the Type III region (Fig. 4). Averaging the results per sample as listed in Table S4 in Online Resource 3, OH absorption contributions in clinopyroxene (recalculated as water contents) from eclogite xenoliths range from 22 to 160 wt.ppm (average 81 wt.ppm) for the α-polarised component, from 24 to 160 wt.ppm (average 79 wt.ppm) for the β-polarised component, and from 34 to 130 wt.ppm for the γ-polarised component (average 67 wt.ppm), while the calculated total clinopyroxene c(H2O) ranges from 92 to 420 wt.ppm (average 230 wt.ppm; Table 1). Contributions for measurement along the refractive indices α and β are identical to within 20%, consistent with previous findings (Koch-Müller et al. 2004; Kolesnichenko et al. 2018; Agasheva et al. 2021; Radu et al. 2022; Aulbach et al. 2023). The OH band positions for garnet are between ∼ 3700 and 3300 cm− 1 (Fig. 4) (see also Schmädicke and Gose 2017; Gose and Schmädicke 2018). As in prior work, c(H2O) in garnet is on average lower, ranging from 2 to 48 wt.ppm (average 14 wt.ppm) when quantifiable, and H2O was entirely undetectable in garnet in eclogite xenoliths from Orapa.

Bulk rock c(H2O) was reconstructed assuming 55 wt% garnet and 45 wt% clinopyroxene in the bulk (see Aulbach et al. 2020 for rationale) and neglecting the presence of accessory rutile. Due to the low abundance of rutile in eclogite (median 0.6 wt; see Appendix 1 in Aulbach 2020), its contribution is neglected here. While rutile is indeed a very hydrogen-rich NAM, and can contain up to 380 wt.ppm H2O (median reported for rutile from mafic to felsic samples; see Appendix 3 in Lueder et al. 2024), rutile would only contribute 2.3 wt.ppm to the bulk eclogite. Bulk rock c(H2O) reconstructed from only garnet and clinopyroxene ranges from 43 to 190 wt.ppm, with an estimated total uncertainty of 50%, which reflects the uncertainties on mineral c(H2O) and on the mineral modes used in the bulk rock reconstruction. When no absorption related to OH in garnet was detected, its c(H2O) was calculated 2 wt.ppm (i.e. the detection limit) × sqrt(2)/2 = 1.4 wt.ppm, following Antweiler (2015) who suggests this procedure to safeguard against biasing datasets toward high values.

FTIR images (Fig. 3; Online Resource 2) show the total integrated absorbance of H2O (3050–3720 cm− 1), the C-H bonds (2750–3000 cm− 1) as a measure of epoxy contamination, which often also shows bands in the OH range, and the Si-O overtones of as a measure of clinopyroxene thickness changes (1450–2350 cm− 1). Of the 23 fragments analysed, 21 show flat hydrogen profiles with no clear diffusive loss or gain (Fig. 3a, b). In these grains the total H2O absorption is only affected by the contamination of epoxy or hydrous inclusions, which can be detected by changes in the absorption of the C-H bonds or a change in the absorption of the Si-O overtones (Fig. S2 in Online Resource 1). Only two grain fragments of sample K1-6, which exhibit constant Si-O overtones and no contamination by epoxy, have profiles with lower total integrated H2O absorbance towards the edge, but still show a plateau in the core of the fragment overtones (Online Resource 2). However, the differences in absorption only decrease for grain fragment K1-6_1 and K1-6_2 from 670 cm− 2 to 600 cm− 2 and from 750 cm− 2 to 650 cm− 2, respectively. The within-grain H2O concentration variability is therefore estimated < 15%, which is well within the total estimated uncertainty of 40% relative. The images also validate the measurement strategy for the water quantification and clearly show that only measurements of optically clear areas with no sign of epoxy contamination in the spectrum are representative for the sample.

Sample OE72 is a medium-grained pyroxenite, with correspondingly jadeite-poor, apple-green clinopyroxene (Jadeite mole fraction 0.07) and grossular-poor garnet (grossular mole fraction 0.12), which shows a mild foliation defined by some grains of both garnet and omphacite (Fig. S4 in Online Resource 1). The preliminary data suggest the formation of a CPO where [100] and (010) form girdles in the YZ plane and [001] forms a point maximum parallel to X (Fig. S5 in Online Resource 1). Orientation was determined for 46 omphacite grains. The garnet grains show some minor internal misorientation (Fig. S5 in Online Resource 1). Sample OE83 is an eclogite with jadeite mole fraction of 0.28 and grossular mole fraction of 0.20, and also shows a foliation (Fig. S4 in Online Resource 1). Again, subgrain boundaries (i.e. lines with a misorientation < 10°) or variability in misorientation are observed in garnet, but not in omphacite. CPO is detected in both omphacite and garnet (based on only 42 grains each). The omphacite CPO is similar to the CPO of sample OE72; the [001] axes are parallel to X, and [100] and poles to (010) form a girdle in the YZ plane (Fig. S5 in Online Resource 1). The foliation and lineation of the samples are not known, and the reference frame for the pole figures was based on Fig. S4 in Online Resource 1, where the horizontal of the image is X, and vertical is Y. Therefore, X was parallel to the long axes of the elongated grains in sample OE72, but perpendicular to the long grain axes in sample OE83 (Fig. S4 in Online Resource 1). Modelling and previous measurement of omphacite CPO shows that the [001] axes are expected to align with the lineation, and that [100] and (010) form girdles (e.g., Bascou et al. 2001; Ulrich and Mainprice 2005).

Discussion

Nature of mantle metasomatism in cratonic eclogite and its timing

Making use of observational data, thermodynamic models and experimental constraints, mantle eclogite was shown to have an origin as spreading ridge-derived and variably differentiated oceanic crust (comprising melts and cumulates) that was subsequently exposed to varying degrees of seawater alteration and metamorphic processes (dehydration, partial melting) during recycling, probably in subduction zones, followed by mantle metasomatism (e.g., Aulbach and Smart 2023). The geochemical effects of mantle metasomatism in eclogites include – patently – addition of hydrous minerals or carbonates and – cryptically – enrichment in incompatible elements as well as enriched radiogenic isotope compositions, with melts from the kimberlite-carbonatite spectrum frequently invoked as metasomatic agents (e.g., Ireland et al. 1994; Barth et al. 2002; Heaman et al. 2006; Jacob et al. 2009; Smart et al. 2009; Czas et al. 2018). The geochemical hallmarks of this metasomatism are recognised in eclogite and pyroxenite suites globally, and include a decrease of jadeite component in clinopyroxene and of grossular component in garnet, a decrease in FeO and Li, Cu ± Zn abundances, and an increase in MgO, Cr2O3 and moderately to highly incompatible elements (Sr, Pb, Th, U ± Zr, Nb) (Aulbach et al. 2020), as illustrated in Fig. 2b and c. These signatures, which are often prevalent at mid-lithospheric depths (60–150 km; see Fig. 2d), occur in some 20–40% of individual xenolith populations. As a result of the varied protoliths and multi-stage evolution of xenolithic eclogite, bivariate plots typically exhibit significant scatter and weak correlations among various parameters.

Mantle metasomatism in xenolithic eclogite was suggested to involve addition of a diopside-rich pyroxene from a kimberlite-like melt, followed by diffusive homogenisation and/or recrystallisation to account for the general compositional homogeneity of garnet and clinopyroxene in metasomatised eclogite and pyroxenite xenoliths (Aulbach et al. 2020), although exceptions exist (e.g., Korolev et al. 2021). In peridotite xenoliths, secondary clinopyroxene introduction was inferred for samples having a coherent olivine ± orthopyroxene fabric, but clinopyroxene of different orientation based on EBSD (Puziewicz et al. 2023). Similar observations could not be made for the two metasomatised eclogite xenoliths studied here. Several samples amongst the Orapa suite do show weak shape-preferred orientation of some but not all grains of both clinopyroxene and garnet (Fig. S4 in Online Resource 1), and clinopyroxene orientation data additionally suggest a CPO (Fig. S5 in Online Resource 1). Based on profiles for mineral pairs in metasomatised vs. pristine Koidu eclogite xenoliths (Aulbach et al. 2019), trace element abundances show no consistent zoning across garnet-clinopyroxene interfaces of strongly metasomatised eclogites, except possibly at the outermost 50–100 μm. This includes Li, which is the fastest-diffusing element amongst those quantifiable by laser ablation inductively-coupled plasma mass spectrometry (Fig. S6 in Online Resource 1). This homogeneity suggests that any chemical gradients associated with metasomatism were removed by diffusion by the time of entrainment, perhaps aided by melt-advected heat, or obliterated by melt-assisted recrystallisation.

Subgrain boundaries (i.e. lines with a misorientation < 10°) in garnet from eclogite OE83 (Fig. S3 in Online Resource 1) may reflect partial recovery after deformation. Conversely, no subgrain boundaries are observed in clinopyroxene. This is taken to indicate that, in contrast to garnet, subgrain boundary migration in clinopyroxene led to formation of new high-angle grain boundaries (e.g., Buatier et al. 1991). Combined with the compositional equilibration, this qualitatively indicates that some time had elapsed between metasomatism and exhumation in the host kimberlite. More data are needed to assess whether this interpretation is robust.

Effects of mantle metasomatism on H2O distribution

Values of clinopyroxene−garnetDH2O for Orapa vary widely because none of the garnets contained measurable H2O, as also reported for other localities on the Kaapvaal craton (Huang et al. 2014) (a value of 1.4 wt.ppm was assumed for calculation, see Results). Prior work showed that clinopyroxene−garnetDH2O decreases with increasing grossular content of garnet and with increasing temperature, both of which favour partitioning of incompatible components, including H2O, in eclogitic garnet (Aulbach et al. 2023). However, such systematics are not (or only weakly) observed in the present dataset (Fig. 5a-b), nor is there a relationship with jadeite component in clinopyroxene (Fig. 5c). There is a weak relationship of clinopyroxene−garnetDH2O with indicators of metasomatism, such as elevated MgO, Ce or Th abundances (Ce shown as an example in Fig. 5d). While some scatter in clinopyroxene−garnetDH2O could reflect diffusive loss or gain of hydrogen related to entrainment, we note that that pristine eclogite xenoliths, which should likewise be affected by hydrogen diffusion, do show systematic variations. This is confirmed by comprehensive FTIR imaging of clinopyroxene grain fragments from some of the samples analysed in this study, which, with one exception (K1-6) show homogeneous H2O absorption (Fig. 3; Online Resource 2). We therefore conclude that the scattered clinopyroxene−garnetDH2O in metasomatised eclogites reflects changes in the crystal chemistry (mostly lowering the jadeite and grossular components in clinopyroxene and garnet, respectively; Fig. 2b-c). As these changes are also prevalent over specific temperature intervals (Fig. 2d), there are superposed effects on how much hydrogen is incorporated in these two minerals in equilibrium with a hydrous melt.

Fig. 5
figure 5

Clinopyroxene-garnet H2O partition coefficient D versus (a) temperature (as in Fig. 1), (b) grossular component in garnet, (c) jadeite component in clinopyroxene and (d) whole-rock (WR) Ce abundance, showing weak relationships, if any (data for partly metasomatised eclogite and pyroxenite xenoliths in this study and for mostly pristine eclogite and pyroxenite xenoliths in Aulbach et al. 2023)

Full size image

Effects of mantle metasomatism on mean OH dipole orientation in omphacite: further evidence for H2O retention in xenolithic eclogite minerals

Hydroxyl in clinopyroxene occurs in different local environments with correspondingly different orientations. The mean orientation of the OH dipole in clinopyroxene may be gauged as the ratio of the absorption contribution of the γ and the α component (hereafter γ/α), given that contributions from α and β are indistinguishable. Prior work on pristine eclogite xenoliths finds that mean dipole orientation shows weak anticorrelations with temperature, diopside component and with some incompatible elements, such as Nb (Aulbach et al. 2023). In this study comprising also strongly metasomatised eclogite and pyroxenite xenoliths, the weak anticorrelation with temperature and diopside persists (Fig. 6a-b), and a marked relationship with a wider range of incompatible elements is highlighted (Sr, Nb, Ce, Th) that was not evident from the pristine eclogite dataset. That is, γ/α for samples with strong enrichment in these elements is restricted to values < ∼ 1.2, whereas nearly the full range of γ/α (0.4–3.1) is observed for unenriched samples (shown for MgO, Ce, Th and Li in Fig. 6c-f). The relationship with REE disappears for Gd and heavier elements (not shown). This is suggested to reflect that mantle metasomatism affects both the concentrations of incompatible elements and mean OH band polarisation via its effect on crystal chemistry. More importantly, we take this regularity – where hydrogen and trace element concentration estimates are obtained by completely different methods (FTIR and laser ablation microprobe inductively-coupled plasma mass spectrometry, respectively) – as circumstantial evidence that our estimates of the contributions from the three refractive indices are largely accurate, with insignificant effects of diffusive H2O loss or contributions from contamination.

Fig. 6
figure 6

Mean dipole orientation, gauged as the ratio of the absorption contribution along the γ and α refractive indices to bulk H2O content in clinopyroxene against a. temperature and b.-f. various indicators of mantle metasomatism, illustrating that ratios remain low in metasomatised eclogite and pyroxenite xenoliths, whereas they vary widely in pristine samples (data for partly metasomatised eclogites in this study and for mostly pristine eclogites in Aulbach et al. 2023)

Full size image

Effects of mantle metasomatism on H2O in cratonic eclogite – expectations and observations

Omphacite, with its high vacancy contents related to jadeite or Ca-Tschermaks component, has the highest H2O contents among rock-forming minerals in the uppermost mantle (Demouchy and Bolfan-Casanova 2016), and it additionally constitutes some 40–60% of xenolithic eclogite (e.g., Aulbach et al. 2020). Combined with the knowledge that H2O is incompatible, therefore enriched in small-volume kimberlite-like mantle melts, metasomatised eclogite might be expected to be a particularly H2O-rich reservoir in cratonic lithospheric mantle. Aubaud et al. (2008) derived the dependence of clinopyroxene-melt and garnet-melt H partition coefficients (clinopyroxene−garnetDH2O) in the system metabasalt + 6 wt% H2O on Al2O3 and TiO2 content, respectively. This allows estimating the hypothetical bulk c(H2O) in eclogite minerals for equilibrium with a melt containing, as an example, 3 wt% H2O. The uncertainties on the estimates so obtained are large, due to possible pressure-temperature effects, differences in the crystal chemistry of experiments and natural samples (Fig. 2a), additional crystal-chemical effects not considered by Aubaud et al. (2008) and the possibility that CO2 in the metasomatic kimberlite melt decreases the activity of H2O (Novella et al. 2015). Bearing this in mind, bulk rock c(H2O) reconstructed based on these hypothetical values are then compared to bulk c(H2O) reconstructed from measured mineral c(H2O), hereafter referred to as “hypothetical” and “observed” for brevity (Fig. 7a).

Fig. 7
figure 7

(a) Hypothetical H2O abundance in reconstructed whole rock (WR) eclogite and pyroxenite xenoliths, calculated for clinopyroxene and garnet based on experimentally-derived mineral-melt partition coefficients (Aubaud et al. 2008) for equilibrium with a melt containing 3 wt% H2O, compared to estimated whole-rock c(H2O) based on measured mineral c(H2O). Difference between hypothetical (hypoth) and measured (meas) WR c(H2O) vs. (b) jadeite component in clinopyroxene, (c) reconstructed whole-rock (WR) MgO content and (d) Ce abundance, as indicators of metasomatism by kimberlite-like melt. Metasomatised samples tend to have H2O abundances that agree within 100 wt.ppm with those expected for equilibrium with a metasomatic melt containing ∼ 3 wt% H2O, whereas greater deficits are observed for pristine eclogite xenoliths with high-jadeite clinopyroxene and low MgO and Ce contents (data for partly metasomatised eclogites in this study and for mostly pristine eclogites in Aulbach et al. 2023; other parameters for partly metasomatised eclogites and for mostly pristine eclogites plotted here and in following figures are listed in Table S6 in Online Resource 3, and in the supplement to Aulbach et al. 2023)

Full size image

For samples from all suites investigated in this and a prior study (Aulbach et al. 2023), therefore representing an isogenous dataset, the difference between hypothetical and observed bulk c(H2O) increases with increasing clinopyroxene jadeite content (Fig. 7b). At the same time, the difference between hypothetical and observed c(H2O) is smaller for samples with high MgO, Cr2O3, Sr, Ce and Th concentrations (shown for MgO and Ce in Fig. 7c-d), which along with lower jadeite and grossular contents are identified as geochemical signatures of kimberlite melt metasomatism. If CO2 in kimberlite (∼ 5–10 wt%; e.g., Becker and Le Roex 2006) significantly reduces H2O activity, then the actual H2O content in the kimberlite would have had to be correspondingly higher, which is possible given the range of reported concentrations (3–7 wt%; Bell et al. 2004; Becker and Le Roex 2006). Despite the uncertainties, the observations suggest that metasomatised eclogites with jadeite-poor clinopyroxene did interact with an agent containing H2O, whereas samples with high-jadeite clinopyroxene, therefore little affected by metasomatism (including carbonatite), indeed contain less H2O than expected for equilibrium with a hydrous agent.

Inhibited hydrogen uptake in metasomatised cratonic eclogite: the role of crystal chemistry

As discussed in the previous section, most eclogite and pyroxenite xenoliths identified as metasomatised have reconstructed bulk rock H2O abundances commensurate with equilibrium with hydrous melt, whereas those identified as pristine have lower abundances than expected for such equilibrium. This finding translates into overall higher H2O abundances for metasomatised than for pristine samples, with broad correlations with various indicators of metasomatism (Fig. 8a-e). Although there is no correlation of bulk c(H2O) with Zr/Hf, high ratios of which have been linked to carbonatite metasomatism (Rudnick et al. 1993), we note that samples with the highest Zr/Hf values are restricted to bulk c(H2O) < 150 wt.ppm (Fig. 8f), which may reflect low H2O activity in carbonatite (Novella et al. 2015).

Fig. 8
figure 8

H2O contents in clinopyroxene, garnet and in whole rock (WR) eclogite and pyroxenite xenoliths reconstructed from measured mineral H2O contents plotted against various geochemical indicators of mantle metasomatism, such as (a) low jadeite component in clinopyroxene, (b) low grossular component in garnet, high whole-rock (c) MgO contents and (d) Ce abundances and (e) low Li abundances. (f) shows Zr/Hf ratios as an indicator of carbonatite metasomatism (Rudnick et al. 1993). This illustrates that H2O is moderately enriched in kimberlite-metasomatised eclogite and pyroxenite xenoliths (data for partly metasomatised xenoliths in this study and for mostly pristine xenoliths in Aulbach et al. 2023)

Full size image

Increasing bulk-rock Ce concentrations, as a gauge for mantle metasomatism in eclogite, are accompanied by a decrease in Al2O3 content (Fig. 9a), and this is consequently similarly true for Ce and jadeite component (hence Al2O3 content) in clinopyroxene (Fig. 2b). Due to interaction with an H2O-bearing melt, metasomatised clinopyroxene with correspondingly low Al2O3 content does show elevated H2O concentrations (Fig. 9b). However, the positive effect of Al2O3 concentration in clinopyroxene clinopyroxene−meltDH is well established (Aubaud et al. 2008 and references therein), and the lowering of Al2O3 during metasomatism is thus expected to counteract the uptake of hydrogen from the metasomatic melt. As a consequence, the decrease in Al2O3 content and jadeite component in clinopyroxene leads to a decoupling of H2O and Ce, which have similar compatibilities during mantle melting (Aubaud et al. 2004), such that the lowest H2O/Ce are observed at low jadeite component in clinopyroxene and vice versa (Fig. 9c). Because the H2O inventory in eclogite is dominated by clinopyroxene, reconstructed bulk rocks similarly show low H2O/Ce at low Al2O3 contents (Fig. 9d).

Fig. 9
figure 9

(a) Whole-rock (WR) Al2O3 vs. Ce content, (b) clinopyroxene Al2O3 vs. H2O content, (c) jadeite component vs. H2O/Ce in clinopyroxene and (d) Al2O3 content vs. H2O/Ce in reconstructed whole rocks (data for partly metasomatised eclogite and pyroxenite xenoliths in this study and for mostly pristine eclogite and pyroxenite xenoliths in Aulbach et al. 2023). The arrows in b. indicate the competing effects of equilibration with a hydrous melt on the one hand, and lower hydrogen incorporation resulting from lower Al2O3 content in clinopyroxene as a result of melt metasomatism on the other hand

Full size image

In sum, in contrast to other incompatible components, the uptake in particular by clinopyroxene of hydrogen from kimberlite-like metasomatic agents seems to be impeded by attendant compositional and crystal-chemical changes that diminish hydrogen partitioning into clinopyroxene, and possibly additionally by lower H2O activity in carbonated melts. With a median H2O abundance of 94±47 wt.ppm (this study and Aulbach et al. 2023; assuming a 50% total uncertainty), eclogite and pyroxenite do not disproportionately (relative to their subordinate volume of < 20%; Garber et al. 2018) contribute to the bulk cratonic H2O inventory, compared to median abundances of 80 wt.ppm in peridotites from the Slave, Siberian and Kaapvaal cratons (Peslier et al. 2010, 2012; Doucet et al. 2014; Kilgore et al. 2020; see also Jackson and Gibson 2023).

Summary

We estimated, by FTIR, hydroxyl contents in clinopyroxene and garnet from eclogite and pyroxenite xenoliths that show evidence for metasomatism by kimberlite-like melt evident in elevated contents of MgO and of highly incompatible elements (e.g., Sr, Ce, Th), and compare them to published data for xenoliths that do not show strong evidence for metasomatic overprint. The main findings are as follows:

  • In this study, the difference between H2O abundances derived from mineral-specific (Bell et al. 1995) vs. wavenumber-dependent calibrations (Libowitzky and Rossman 1997) is on average 33%.

  • FTIR imaging reveals that the rim of only one of 8 samples shows variability (< 15%) of H2O, but well within the estimated overall uncertainty on mineral H2O of 40%, mostly related to choice of absorption coefficient.

  • The calculated total clinopyroxene c(H2O), estimated for both pristine and metasomatised eclogite and pyroxenite xenoliths using the calibration of Libowitzky and Rossman (1997), ranges from 90 to 420 wt.ppm, while that in garnet is on average lower (< 2–50 wt.ppm).

  • Bulk rock c(H2O) ranges from 40 to 210 wt.ppm, and – to the extent that experimental hydrogen partition coefficients derived in the eclogite-H2O system can be applied to kimberlite-metasomatised eclogites and pyroxenites - for many metasomatised samples corresponds within ~100 wt.ppm with hypothetical contents calculated for equilibrium with a melt having a typical H2O content of 3 wt%. In contrast, most pristine samples have lower bulk H2O abundances than expected for equilibrium with such melt.

  • Although metasomatised eclogite and pyroxenite xenoliths have on average higher H2O abundances than pristine ones, metasomatism by kimberlite-like melt is shown to lower clinopyroxene Al2O3 contents, which in turn inhibits hydrogen incorporation, such that the abundances of highly incompatible elements like Ce become increasingly decoupled from those of hydrogen.

  • Eclogite does not disproportionately contribute to the H2O inventory of the bulk cratonic mantle.