Origin of Water in Mantle Eclogites from the V. Grib Kimberlite Pipe, NW Russia

The water content in the garnet and clinopyroxene in the mantle eclogites from the V. Grib kimberlite pipe (Arkhangelsk Diamondiferous Province, NW Russia) was analysed using Fourier transform infrared spectrometry. The results show that all clinopyroxene grains contained structural water at concentrations of 39 to 247ppm, whereas two garnet samples contained detectable water at concentrations of 211 and 337 ppm. The low-MgO eclogites with oceanic gabbro precursors contained signi ﬁ cantly higher water concentrations in the omphacites (70 – 247ppm) and whole rock (35 – 224 ppm) compared to those with oceanic basalt protoliths (49 – 73ppm and 20 – 36ppm, respectively). The incorporation of water into the clinopyroxene may be associated with vacancies at the M2 site, Al in the tetrahedral position, and the elements that ﬁ lled the M2 site (mostly Na and Ca). The highest water content in the omphacite was detected in a nonmetasomatised sample and was assumed to represent residual water that survived during subduction. Other eclogite samples showed signs of modal and/or cryptic metasomatism and contained less water in the omphacites compared to the nonmetasomatised sample. The water content was heterogeneous within the eclogite section of the sampled lithospheric mantle. The lack of distinct and uniform correlations between the indices of eclogite modi ﬁ cation and their water content indicated that the saturation with water was disturbed during their residence within the lithospheric mantle.


Introduction
The study of the water content in the rock-forming minerals of mantle xenoliths that are entrained in kimberlites provides information concerning water storage in the lithospheric mantle of ancient cratons. In mantle xenoliths, the water can be identified as several percentages by weight in hydrous minerals (e.g., phlogopite and amphibole) and up to 2000 ppm in nominally anhydrous minerals (NAMs; e.g., olivine, pyroxene, and garnet). Since the hydrous phases occur sporadically in mantle xenoliths, their NAMs reserve the main water content in the lithospheric mantle. It is generally accepted that water in NAMs affects the mantle's physical and chemical properties (e.g., melting temperature, viscosity, rheology, and electrical and thermal conductivity) and can be the main factor controlling the long-term stability of ancient cratons (e.g., [1,2]). The concentration of water in the peridotite NAMs can provide information about the bulk water content in the lithospheric mantle of ancient cratons, as the peridotites are dominant in their composition. An evaluation of the origin of water in peridotite NAMs also provides insight into the reconstruction of the melting and metasomatic processes in the cratonic lithosphere [1,3,4].
The eclogites are minor components of the lithospheric mantle and their origin can be linked to several distinctively different processes: (1) metamorphism and the partial melting of various stratigraphic levels of the ancient oceanic crust during subduction (e.g., [5][6][7]), (2) crystallisation of the primary mantle melts at high pressure (e.g., [8,9]), and (3) the reaction between the peridotite and siliceous partial melts that are derived from the crustal rocks of the subducting plate (e.g., [10][11][12][13]). To date, the available data on the water content in the cratonic eclogite NAMs are sparse [14][15][16][17][18][19][20]. However, the juxtaposition and interpretation of the origin of eclogites with water content in their NAMs can provide additional information about water cycles in the lithospheric mantle and the factors that control their saturation with water. Data are presented on the water content in ten eclogite xenoliths from the V. Grib kimberlite pipe, which is the largest diamond deposit within the East European Platform. These data were used to evaluate the origin of the water in the eclogites and obtain information about the water cycle during the subduction process.

Background and Analytical Methods
2.1. Sample Overview. The 372 Ma V. Grib kimberlite pipe is in the southern part of the Kola Craton [21] within the Arkhangelsk Diamondiferous Province (ADP, NW Russia). The V. Grib kimberlite pipe contains significant varieties of the mantle and crustal xenoliths (e.g., [22][23][24]). The eclogite xenoliths comprise approximately 10% to 12% of the mantle constituents and are represented by two main groups: coarsegrained eclogites and fine-to-medium-grained equigranular zircon-bearing eclogites. The coarse-grained eclogites are the most abundant, whereas the zircon-bearing eclogites are rare.
The samples analysed in this study were coarse-grained (grain sizes from 0.5 to 1.3 cm) bimineralic (garnet and clinopyroxene) eclogites with accessories of ilmenite and rutile. Nine of ten samples contained phlogopite with modes of <1 to 6 vol. %. The samples included high-MgO (three samples) and low-MgO (seven samples) groups ( Figure 1). The eclogites were interpreted as metamorphosed fragments of oceanic crustal rocks (basalt and gabbro for low-MgO eclogites and picritic/MgO basalt and troctolite for high-MgO eclogites) emplaced into the lithospheric mantle via a subduction event at 2.8 Ga [25]. Based on pressure-temperature estimates (44-78 kbar; 940°C-1275°C), eclogites were transported by kimberlite from depths of approximately 160 km to >200 km. The results of a detailed study of eclogites can be found in the work by Shchukina et al. [25]. The main characteristics of the eclogites examined in the present study are presented in Table 1. Quantitative electron microprobe data 2 Lithosphere The thickness of the transparent fragments varied from 80 to 410 μm. The polished grains were cleaned in acetone to remove the crystal-bond thermoplastics. The orientation of the pyroxenes was determined by observing the interference figures on a petrographic microscope. Pyroxene grains were analysed twice in two perpendicular orientations with the polariser parallel to the optical indicatrix α, β, or γ, which ensured the precise quantification of water in the anisotropic minerals [26]. Due to the isotropic nature of the garnets, they were analysed in nonpolarised light. Analyses were performed using a Bruker Vertex 70 spectrometer equipped with a Hyperion 2000 IR microscope (with a resolution of 4 cm -1 , 30 to 70 scans, aperture 50 × 50 μm). Polarised absorption measurements were conducted in a spectral range from 600 to 7000 cm -1 . The water content of the garnets and clinopyroxenes was calculated using the modified expression [27] of the Beer-Lambert law: A = c · t · ɛ, where ɛ is the (integrated) molar absorption coefficient, t is the thickness of the section, and c is the concentration of the absorbing species in the sample, which describes the linear relationship between the absorbance and the concentration. Therefore, the concentration of hydrogen can be calculated from the integral intensity of cðH2OÞ = A/t · ɛ. For the water content calculations, the coefficient ɛ was equal to 8:34 ± 1:46 · 10 4 for omphacite and 8:77 ± 0:51 · 10 3 (L/mol·cm 2 ) for garnet [28]. Note that the estimations with the coefficient ɛ from Bell et al. [29] revealed 54% and 15% uncertainties for clinopyroxene and garnet, respectively, compared to the water calculation results obtained with the coefficient ɛ from Katayama et al. [28]. To quantify the water content of the samples, the areas beneath the OH stretching bands were integrated for all three principal directions of the optical indicatrix (A α , A β , and A γ ) using the OriginPro software (OriginLab Corporation). Where possible, a minimum of five grains of garnet and clinopyroxene from each sample were analysed, and the average result was used. To avoid extra uncertainties in the calculations, the baseline corrections were conducted by hand a minimum of three times for each spectrum.

Main Element
Analysis by EPMA. The composition of the main elements in the minerals was measured using a JEOL JXA-8100 electron probe microanalyser (EPMA) at the Analytical Centre for Multi-element and Isotope Research at the Institute of Geology and Mineralogy Siberian Branch Russian Academy of Science (IGM SB RAS), Novosibirsk. The analytical conditions were a 20 kV accelerating voltage and a 50 nA beam current with a beam size of 1 μm. The instrument was equipped with five wavelengthdispersive spectrometers with various types of LIF, PET, and TAP crystals which allowed for a reliable analysis of the elements F-U. In-house natural mineral IGM SB RAS standards were used for calibration. Relative standard deviations were within 1.5%. Data were acquired for 10 s on-peak as well as 10 s on either side of the background. Raw data were corrected using a ZAF algorithm, and element abundances were transformed to oxides assuming stoichiometry, where FeO T represented total iron. Detection limits were <0.05 wt.% for all elements analysed, including 0.01 wt.% for Cr, Mn, Ni, Cl; 0.02 wt.% for Ti and Na; 0.05 wt.% for K; 0.04 wt.% for F; and 0.03 wt.% for Ba. The main element oxide contents in wt.% were recast into atoms per formula unit (apfu) using a standard protocol. Additional details have been reported elsewhere [30][31][32].

Results
3.1. Clinopyroxene. The polarised infrared spectra of the clinopyroxenes with E//γ and E//α are shown in Figure 2. The spectra with E//β were similar to those with E//α and are not presented here. All three groups of absorption bands described by Koch-Muller et al. [16] were distinguished in our spectra: 3460 cm -1 (Group 1), 3530 to 3540 cm -1 (Group 2), and 3630 to 3640 cm -1 (Group 3). The bands of Groups 1 and 2 had the highest intensity with E//γ and those of Group 3 had the highest intensity with E//α.
All samples of the high-MgO eclogites had OH bands in Groups 1 and 3 with the absence of Group 2 bands. The intensity of the bands of Groups 1 and 3 was almost the same in the samples of the high-MgO eclogites and accounted for 42% to 70% and 30% to 58% of the calculated total OH content, respectively. The bands of Groups 1 and 3 were also identified in all samples of the low-MgO eclogites (except for the G3-2 sample, which only had the OH band of Groups 1 and 2). Five samples of the low-MgO eclogites (G3-1, G3-2, G3-4, G3-6, and G3-26) had the OH bands of Group 2. The intensities of the OH bands of Groups 1 and 3 were much higher compared to those of Group 2 and accounted for 22% to 89%, 8% to 58%, and 18% to 21% of the total calculated OH content, respectively. The measured water content in the clinopyroxenes varied from 39 to 84 ppm in the high-MgO eclogites and from 49 to 247 ppm in the low-MgO eclogites.
3.2. Garnet. Two samples of eclogites (G3-7 of the high-MgO group and G3-1 of the low-MgO group) had garnet that showed hydroxyl absorption bands. Three groups of OH absorption bands were distinguished in the garnet spectra: 3642 cm -1 , 3543 cm -1 , and 3448 cm -1 ( Figure 3). The first 4 Lithosphere 5 Lithosphere and second band groups were due to intrinsic water in the garnet structure, whereas the band near 3448 cm -1 was due to molecular water which may have occurred in submicroscopic fluid inclusions in the garnets (e.g., [33]) and, thus, was excluded from the total OH content calculation (e.g., [14,28,34,35]). The intensity of OH bands at 3543 cm -1 was significantly higher compared to 3642 cm -1 and accounted for 60% of the total OH content in the low-MgO eclogite, and approximately, the same in the high-MgO eclogite (47% and 53% of the total OH content, respectively). The calculated water content in the garnets amounted to 211 ppm for the G3-7 sample and 337 ppm for the G3-1 sample.
Based on the modal proportions of garnet and clinopyroxene in the eclogites, the estimated whole rock (WR) water content varied from 20 to 160 ppm in the high-MgO eclogites and from 20 to 224 ppm in the low-MgO eclogites.

Interpretation of Clinopyroxene FTIR Spectra.
Omphacite can be compositionally defined within the system Jd-Ac-Di-He-Tsch, with the chemical formula (M2)(M1)(Si,Al) 2 O 6 . The M2 mostly represents Ca and Na and the ratio of Na/(Na+Ca) is between 0.2 and 0.8 (i.e., jadeite component).
The clinopyroxene in the low-MgO eclogites of the V. Grib pipe is a typical omphacite with a Jd component of 23 to 56 mol.%. These omphacites are more Al 2 O 3 enriched (4.3-11.7 wt.%) compared to the clinopyroxene in the high-MgO eclogites. The clinopyroxene from the three samples of the high-MgO eclogites had a much lower Na 2 O content (1.7-2.9 wt.%), and the Jd component was equal to 13, 18, and 22 mol.%, respectively. According to classical terminology (e.g., [36,37]), two of the high-MgO clinopyroxenes were not omphacites, and they contained significant portions of diopside (G3-7) and enstatite (G3-8) components. Therefore, these rocks are not eclogites sensu stricto and can be classified as garnet clinopyroxenites. However, a detailed study of the mineralogy and geochemistry of the xenolith suite [25] concluded that all the rocks had an associated protolith linked to oceanic crustal rocks. Thus, we considered these "noneclogite" samples to be a part of the rock complex that originated from the Archean subduction; this fact that is important for the present study.
In general, the omphacites from eclogites have three groups of OH FTIR: 3445 to 3465 cm -1 , 3500 to 3540 cm -1 , and 3600 to 3624 cm -1 [16]. It is well documented (e.g., [20,   7 Lithosphere as well as the total OH content, were assessed, and the results were compared with those previously obtained for natural [15,16,20,40] and experimental [41] samples. Several studies have shown that the intensity of the first band group (approximately 3460 cm -1 ) was positively correlated with the number of cation vacancies [16,20,38,40,41], indicating that the level of the vacancies controls the incorporation of the water. It was also shown that an increase in Ca+Na content reduces the number of vacancies in the M2 site and should, therefore, restrict the incorporation of OH into the omphacite structure (e.g., [16,38]). In the case of the samples analysed in the present study, the number of cation vacancies could not be accurately determined due to the lack of a reliable independent estimate of the Fe 3+ content (e.g., [42]). Cation quantities of >4.00 apfu in seven of ten samples, and the presence of Fe 3+ in these clinopyroxenes was expected (Table 2). Three samples with oceanic gabbro protoliths (G3-3, G3-10, G3-26) had cation quantities of <4.00 apfu, the lowest amount of Ca+Na that resulted in a cation deficit in the M2 site ( Table 2). These samples had the highest intensity of the Group 1 band (Figure 4(a)). The remaining samples had almost equal values of the Group 1 band intensity and showed no correlations.
The intensity of the second band group (approximately 3500 to 3540 cm -1 ) has been shown to be positively correlated with aluminium in the tetrahedral position, indicating that the charge-deficiency of Al for Si is responsible for the bands of Group 2 [16]. The OH band of Group 2 (3530-3540 cm -1 ) was identified in five samples of the low-MgO eclogites. The intensity of this band showed a distinct positive correlation with aluminium in the tetrahedral position (Figure 4(b)), demonstrating that Al IV was also involved in the incorporation of hydrogen in these samples.
The interpretation of the third group bands (approximately 3600 to 3624 cm -1 ) is controversial. Koch-Müller et al. [16] demonstrated that the strong absorption bands observed at 3600 to 3624 cm -1 in eclogite omphacites from Siberian kimberlite pipes were caused by nanoscale inclusions of clinochlore, amesite, and biotite and, therefore, should be omitted from the total OH content estimation. However, in most cases, these bands were included in the total OH calculation in omphacites from cratonic [15,20] and orogenic eclogites [35,40,43]. Smyth et al. [20] concluded that an α-polarised band at 3620 cm -1 is much stronger in diopside-rich eclogite pyroxenes than in jadeite-rich pyroxene, which has the strongest band at 3470 cm -1 . According to experimental data [41], the appearance of the hydroxyl-stretching band at 3600 cm -1 is due to hydrogen occupying the M2 site vacancy.
In the present study, the third group bands were identified in nine samples and had a significant total water content (up to 58%). However, the interpretation that the third group bands result from nanoscale inclusions in omphacites [16] was unlikely for our samples because no lamellae or inclusions were identified. The intensity of the third group bands was negatively correlated with the intensity of the first group bands, and the highest values were observed in the samples with the large quantities of Ca+Na (Figures 4(c) and 4(d)).
The FTIR spectra interpretations demonstrated that the OH bonds that generated three different groups of absorption bands may be associated with features of crystal chemistry, at least concerning vacancies at the M2 site, the Al in the tetrahedral position, and the elements that fill the M2 site (mostly Na and Ca). The water content was positively correlated with the jadeite component in most of the samples (Figure 4(e)), and the highest values were observed in the eclogites with oceanic gabbro protoliths. Additionally, the set of samples with the largest quantities of Ca+Na showed a positive correlation between the water content and the ratio of calculated Fe 3+ to Fe total (Figure 4(f)). In summary, uniform correlations among the samples were not observed, indicating that a complex of local conditions controls the incorporation of water in these minerals.   Two groups of OH absorption bands observed in the studied eclogite garnets were associated with intrinsic water: 3642 cm -1 and 3543 cm -1 . Similar band groups were previously identified in garnets from xenoliths of mantle peridotites [1,3,34] and eclogites [19,34] of kimberlites, as well as garnets from orogenic eclogites [28,35,43,44]. It has been well-documented that the appearance of the first group bands (approximately 3610 to 3660 cm -1 ) can be explained by the incorporation of water via tetrahedral [SiO 3 (OH)] groups that are charge-balanced by cation vacancies on dodecahedral [45] or dodecahedral and octahedral sites [46,47]. The second band group (approximately 3540 to 3600 cm -1 ) is associated with structural OH incorporation in the form of the SiO 4 4-↔(O 4 H 4 ) 4substitution [33,47]. We have not provided results on correlations between the intensity of the OH band groups and the chemical composition of the garnet as such correlations would not be informative due to the critically low number of samples. However, we can report a single observation: the OH band at 3543 cm -1 was significantly prominent in the low-MgO eclogite garnet, which contained elevated TiO 2 (0.76 wt.%). This observation corroborates the results of a previous study on the dependence of the increasing intensity of the OH band near 3520 to 3575 cm -1 on the Ti content in garnets from mantle peridotites of the Kaapvaal Craton [1] and Udachnaya [3] kimberlites and the Udachnaya eclogites [34].

Factor 1. The Initial Water Content in Eclogite
Protoliths and Dehydration during Subduction. During subduction, the downgoing slab releases water at various depths and varying amounts which depend on numerous factors such as the thermal regime and the geometry of the subduction zone, the rate of plate convergence, and the age of the subducting slab [48][49][50]. Additionally, the magnitude and depth of water loss can vary dramatically from layer to layer of the oceanic crust [49] due to differences in the mineralogy, WR composition, and initial water content. Intensive dehydration during subduction occurs at shallow depths (where most of the hydrous minerals are broken down [51]) and decreases with the subduction depth [50]. Less than 20% of the slab "water" survives as OH groups in the NAMs at the mantle depths of diamond stability field conditions [18,28]. The upper oceanic crust is more water-enriched (35% of the total H 2 O content in the slab) compared to the lower oceanic crust (28% of the total H 2 O content in the slab; [49]). Theoretically, at equal conditions of water loss after dehydration and partial melting of the upper and lower oceanic crustal rocks during subduction, the high pressure metamorphosed/residue analogues (as mantle eclogites) should contain the same proportion of water as their protoliths. However, the V. Grib pipe eclogites show the reverse ratio, i.e., the eclogites of the lower oceanic crustal precursor are more water-enriched compared with those of the upper oceanic crustal protoliths. This observation can be explained in several ways: (1) the initial content of the water was higher in the lower oceanic crustal protolith, (2) the lower oceanic crustal rocks were less dehydrated than the upper crustal rocks, (3) the eclogites of the upper oceanic crustal protoliths underwent more intensive partial melting, and (4) the eclogites of the lower oceanic crustal protoliths became more water reenriched during their residence within the lithospheric mantle. To obtain more clarity regarding the origin of the water in the V. Grib eclogite suite, the indices of eclogite modification during subduction and their residence in the lithospheric mantle must be identified, i.e., partial melting and mantle metasomatism, which are the main processes that can alter the water content.

Factor 2. Partial Melting.
Water in NAMs is thought to be incompatible, and similar to light rare earth element (LREE) their content should decrease during partial melting [52]. Generally, there are several indices of partial melting recorded in eclogite WR composition: strong depletion of LREE towards heavy rare earth elements (HREE) (e.g. [53,54]), fractionation of the middle rare earth elements (MREE) to HREE (e.g., [54]), a simultaneous increase in the MgO content and decrease in the SiO 2 content relative to the protolith (e.g., [55,56]), and an increase in the Sm/Nd ratio to the Nd content [5]. The LREE depletion in WR REE spectra was observed only in the three eclogites with the oceanic gabbro protoliths (Figure 5(a)). The remaining samples showed almost flat MORB-like REE distribution or slight and extreme enrichment of LREE to HREE ( Figure 5(b)). Two of ten samples were plotted in the residual field obtained from experimental metabasalt melting in the SiO 2 /MgO diagram ( Figure 6(a); [55]) but did not show signs of partial melting in their WR RE patterns ( Figure 5(b)). Additionally, a single sample (G3-26) matched the field of "residues" after oceanic crust partial melting in the Sm/Nd versus Nd diagram ( Figure 6(b); [5]). If it was assumed that three samples with oceanic gabbro protoliths ( Figure 5(a)) had yet to experience partial melting, they should have been the most water depleted; however, the reverse was true: the water content in their cpx (70-247 ppm) was comparable or higher than those in the other eclogite samples (39-111 ppm). Based on the WR REE patterns, the G3-6 sample was the best candidate to suggest that it experienced partial melting: the HREE content was within the range of MORB values, and the fractionating of MREE to HREE was observed. However, their enrichment in LREE did not match the residual composition which suggests that they were modified via metasomatic processes.

Factor 3. Mantle Metasomatism.
It is well documented that mantle metasomatism enriches mantle rocks with incompatible elements and, therefore, should also increase their water content (e. g., [1, 3, 15]). The primary evidence 10 Lithosphere that demonstrates eclogite metasomatic enrichment is elevated LREE relative to HREE in the WR profile and the disturbance of the initial isotopic compositions (i.e., cryptic metasomatism; e.g., [7,11,54]), as well as the growth of secondary minerals (e.g., phlogopite) and its effect on the composition of the primary mineral assemblage (i.e., modal metasomatism; [57,58]). All the high-MgO samples and two of the low-MgO eclogites (G3-6 and G3-1) showed slight (La/Yb n up to 1.2) to extreme (La/Yb n =7 in G3-8 sample) enrichment of the WR LREE relative to the HREE ( Figure 5) and were observed to have experienced the influence of cryptic metasomatism and, therefore, should have been the most water-enriched. However, these samples had a similar or lower water content compared to the other samples (Figures 7(a) and 7(b)). Moreover, the water content in the clinopyroxenes was negatively rather than positively correlated with La/Yb ratio (Figure 7(c)). Additionally, there were no observations of correlations between the water content and the indices of cryptic metasomatism, such as La/Yb, La/Sm, or Ce (Figures 7(b)-7(f); [1,3,4]) in any of the samples. However, as previously demonstrated [22,23,59,60], the lithospheric mantle beneath the V. Grib pipe was metasomatised by silicate and carbonatite melts. Therefore, the interaction of the eclogites with the mantle melts/fluids during their residence within the lithospheric mantle was possible. To obtain an approximate estimate of the water content in a hypothetical melt in equilibrium with the eclogitic minerals, we used the mineral/melt partition coefficient described by O'Leary et al. [61] and Tenner et al. [62] for clinopyroxene and garnet, respectively. The water content of the melts in equilibrium with the eclogite clinopyroxene ranged from 0.4 to 1.4 wt.% (average of 0.6 wt.%). These values were not comparable with the water content in the kimberlite (~5 wt.%; [63]) and carbonatite (6-10 wt.%; [64]) and were lower than the calculated values for the metasomatic melts equilibrated with the clinopyroxenes from the metasomatised Roberts Victor Type I eclogites (average of 4-5.6 wt.%; [15]) and Kaapvaal   Diagrams were adapted from Tappe et al. [56] and Agashev et al. [5], respectively. Experimental data on metabasalt melting [55]. Oceanic crust section [70][71][72][73].   [69] 12 Lithosphere (2.95 wt.%; [1]) and Udachnaya (2 wt.%; [3]) peridotites, but were equal to those observed in the mid-ocean ridge basalt (MORB) glasses (0.04-0.5 wt.%; [65]). If we assume that the water enrichment of the V. Grib pipe eclogite clinopyroxene is associated the reaction with the mantle melt of the composition close to the MORB mantle source, a similarity in La, Ce, or (La/Sm) n , and H 2 O behaviour should have been observed [65]; however, no correlations were detected (Figures 7(d)-7(f)). The H 2 O/Ce ratio in the clinopyroxene was significantly higher in the eclogites with the oceanic gabbro precursors compared to those of the oceanic basalt protoliths and high-MgO suite (Figure 7(g)); however, this did not reflect the influence of the metasomatic processes (e.g., [4]). The hypothetical melt in equilibrium with the eclogite garnets should have contained at least 6 to 10 wt.% of H 2 O, which was significantly higher compared to the coexisting clinopyroxene. These values were equal to the water content in the carbonatite (6-10 wt.%; [64]); however, no signs of carbonatite metasomatism were identified in the garnet composition, and the coexisting clinopyroxene was not extremely LREE-enriched (La n = 38 and 1.5 chondritic units). Additionally, during metasomatic events, the H 2 O tends to be incorporated into the clinopyroxene rather than the garnet, as evidenced from the available clinopyroxene/garnet partition coefficient (e.g., [66]). The partition coefficients for water between the garnet and clinopyroxene (D grt/cpx ) in the studied samples were equal to 2.5 and 3.1, respectively, and were significantly higher compared to those from natural and experimental lithospheric mantle assemblages (e.g., [1,3,66]). These coefficients were also significantly higher than those that were previously derived for ultrahigh pressure (UHP) eclogites from the Kokchetav (D grt/omp~0 .1-0.2; [28]) and Erzgebirge (D grt/omp = 0.07-0.18; [43]) massifs for which it has been suggested that water in the deeper subducted crust is preferentially hosted in the omphacite rather than garnet. However, Sheng et al. [35] showed that the water D grt/omp for the UHP eclogites of the Dabie orogen varied greatly (from 0.3 to 5.8) and concluded that the garnet could be as important as the clinopyroxene in incorporating and transporting water into the mantle during subduction. In the case of the samples G3-7 and G3-1, the garnet was more favourable for water incorporation; however, the reasons for this and why the garnet from the other samples had no detectable water are unclear. Moreover, the water partitioning between the garnet and the clinopyroxene was in complete disequilibrium in these samples, indicating the heterogeneity of water saturation even within small samples.
Nine of ten eclogites examined contained modal phlogopites in the form of large (up to 0.5 cm) platy grains and small (<0.1 cm) irregular grains with modes that varied from <1 to 6 vol.%. As demonstrated by Shchukina et al. [25], the large phlogopite grains correspond to "primary" phlogopite [67] and could be an early generation of this mineral, whereas the small irregular phlogopite grains could be formed later during prekimberlite metasomatism or kimberlite emplacement. Therefore, at least two metasomatic events by H 2 O-K-enriched fluid (e.g., [58]) may have affected the eclogite suite and increased the water content in their NAMs. However, the major and trace element content in the garnet and clinopyroxene did not provide evidence for the disturbance of their compositions by phlogopite metasomatism (e.g., [57]). The exception was the G3-6 sample (Figure 1(f)) with 6 vol.% of phlogopite, in which the composition of garnet and clinopyroxene were in major-element disequilibrium, and the garnet had elevated strontium (Sr; 9 ppm) and barium (Ba; 16 ppm) concentrations. However, the water content in the clinopyroxene from the most phlogopite-rich sample was equivalent to that of the remaining samples, and the garnet did not contain water at all. Among all the samples analysed, the highest water content in the clinopyroxene and garnet were identified in phlogopite-rich (G3-7 eclogite with 6 vol.% of phlogopite and water content of 84 ppm in cpx and 211 ppm in garnet) and phlogopite-poor (G3-1 eclogite with <1 vol.% of phlogopite and water content of 111 ppm in cpx and 337 ppm in garnet) samples. Additionally, the maximum water content (247 ppm) in the clinopyroxene was detected in the phlogopite-free eclogite (G3-26). The lack of a correlation between the water content in the NAMs and the WR with a modal amount of phlogopite (Figure 8) 13 Lithosphere demonstrates that the water saturation of the eclogite NAMs was not strongly controlled by phlogopite metasomatism.

Other
Factors. An additional factor that could control the incorporation of water into the eclogite minerals could be the equilibration pressure. However, the influence of pressure on the water content in omphacites has not yet been studied experimentally [1]. In the case of the massif eclogites, the water content in the omphacites systematically increases with an increase in the metamorphic pressure (e.g., [28,35,40]). The reverse was observed in Siberian eclogite omphacites [16] (i.e., the water content decreased with increasing pressure). Huang et al. [15] demonstrated a positive correlation regarding the pressure versus water content in the omphacites in the lesser or nonmetasomatised eclogites of the Roberts Victor kimberlite pipe. In the case of the V. Grib eclogites in the present study, a distinct negative correlation was observed between the water content and the equilibrium pressure in the clinopyroxene from the high-MgO eclogites, whereas the same correlation was not observed in the low-MgO samples (Figure 7(h)). Moreover, the garnet did not provide any convincing results in the present study due to the low number of samples (G3-7 and G3-1) in which water was detected. Nevertheless, those samples were extreme representatives of the calculated pressures for the eclogite suite (44 and 78 kbar) and showed a positive correlation of the water content with the pressure in the garnet and the WR.
Marshall et al. [4] proposed the equilibration of the water activity between the water-rich and water-poor portions of the lithospheric mantle (e.g., the rocks from the recycled slab and metasomatised veins versus the depleted peridotites) as the main process that explains the rarity or lack of uniform correlations between the water content in NAMs and the indices of mantle metasomatism or melt extraction. Note that FTIR-determined water content for clinopyroxene from the Obnazhonnaya eclogite [18] are shown; the SIMS-determined water content in Obnazhonnaya clinopyroxene is much higher and equal to 4850 ppm [18].
14 Lithosphere Nevertheless, the V. Grib eclogites with oceanic gabbro precursors contained significantly higher water concentrations in the omphacites (70-247 ppm) and the WR (35-224 ppm) compared to those with oceanic basalt protoliths (49-73 ppm in the omphacites and 20-36 ppm in the WR). However, this set included the samples with signs of cryptic and/or modal metasomatism (G3-1, G3-3, G3-10) and those without (G3-26). The water saturation in the G3-26 clinopyroxene potentially had not been disturbed by the metasomatic processes and may reflect the "slab" water, whereas the remaining samples most likely contained the water disturbed during mantle residence time in varying degrees. The results of this study suggest that during residence in the lithospheric mantle, the water content in the omphacites does not increase but decreases compared with the residual water retained in the omphacites from the subducted slab. Moine et al. [18] demonstrated that the main factor controlling the water content and H isotope composition in omphacites during residence time in the lithospheric mantle is the exchange of water between fluids and minerals in the H 2 form. This process leads to an increase in D 2 and a decrease in H 2 and total water content in the solid.

Conclusions
The water in the V. Grib kimberlite pipe eclogites was mostly stored in the clinopyroxene at 39 to 247 ppm. The incorporation of water into the clinopyroxene may be associated with vacancies at the M2 site, aluminium in the tetrahedral position, and the elements that fill the M2 site (mostly Na and Ca). The garnet in two eclogite samples had a water content that was two to three times higher than that of the coexisting clinopyroxene, indicating the complete disequilibrium of water between the minerals and the heterogeneity of water saturation even within small samples. The clinopyroxene in the nonmetasomatised sample had the highest water content, and the eclogites with the oceanic gabbro precursors contained significantly higher concentrations of water than those with oceanic basalt protoliths or a high-MgO suite. However, the juxtaposition of the water content in the V. Grib eclogites with the indices of their modification during subduction and residence in the lithospheric mantle showed no distinct or uniform correlations, indicating that a complex of various local conditions controls the incorporation of water into these minerals. Moreover, our data indicate a loss of water by the eclogitic clinopyroxene rather than incorporation during residence in the lithospheric mantle.
Data Availability Lithosphere