Volume 22, Issue 9 e2021GC009749
Research Article
Free Access

Carbonatite Versus Silicate Melt Metasomatism Impacts Grain Scale 87Sr/86Sr and 143Nd/144Nd Heterogeneity in Polynesian Mantle Peridotite Xenoliths

Benjamin L. Byerly

Corresponding Author

Benjamin L. Byerly

Isotope Geochemistry Facility—Center for Mantle Zoology, Department of Earth Science, University of California, Santa Barbara, CA, USA

Thermo Fisher Scientific, Waltham, MA, USA

Correspondence to:

B. L. Byerly,

[email protected]

Search for more papers by this author
M. G. Jackson

M. G. Jackson

Isotope Geochemistry Facility—Center for Mantle Zoology, Department of Earth Science, University of California, Santa Barbara, CA, USA

Search for more papers by this author
M. Bizimis

M. Bizimis

School of the Earth, Ocean and Environment, University of South Carolina, Columbia, SC, USA

Search for more papers by this author
First published: 21 August 2021
Citations: 1

Abstract

The Earth's upper mantle is isotopically heterogeneous over large lengthscales, but the lower limit of these heterogeneities is not well quantified. Grain scale trace elemental variability has been observed in mantle peridotites, which suggests that isotopic heterogeneity may be preserved as well. Recent advances in isotope ratio mass spectrometry enable isotopic analysis of very small samples (e.g., nanograms or less of analyte) while maintaining the precision necessary for meaningful interpretation. Here we examine four peridotite xenoliths—hosted in lavas from Savai'i (Samoa hotspot) and Tahiti (Societies hotspot) islands—that exhibit grain scale trace element heterogeneity likely related to trapped fluid and/or melt inclusions. To evaluate whether this heterogeneity is also reflected in grain scale isotopic heterogeneity, we separated clinopyroxene, orthopyroxene, and (in the most geochemically enriched xenolith) olivine for single-grain 87Sr/86Sr and 143Nd/144Nd analyses. We find, in some xenoliths, extreme intra-xenolith isotopic heterogeneity. For example, in one xenolith, different mineral grains range in 87Sr/86Sr from 0.70987 to 0.71321, with corresponding variability in 143Nd/144Nd from 0.512331 to 0.512462. However, not all peridotite xenoliths which display trace elemental heterogeneity exhibit isotopic heterogeneity. Based on coupled isotopic and trace element data (i.e., a negatively-sloping trend in 87Sr/86Sr vs. Ti/Eu), we suggest that carbonatitic metasomatism is responsible for creating the intra-xenolith isotopic heterogeneities which we observe. This carbonatitic component falls off the array defined in 87Sr/86Sr-143Nd/144Nd space by Samoa hotspot basalts, which suggests a second, distinct EM2 (enriched mantle II) component is present in the Samoa hotspot that is not readily recognized in erupted products, but is instead seen only in mantle peridotite xenoliths.

Key Points

  • Peridotite xenoliths from ocean islands show variable within-sample (grain-scale) Sr and Nd isotopic heterogeneity

  • The extent of within-sample isotopic heterogeneity is controlled by whether the dominant metasomatic component is carbonatitic or silicic

  • The enriched mantle II (EM2) component sampled in the Samoan xenoliths has a different isotope composition than the EM2 component sampled by Samoan lavas

1 Introduction

Samples of the mantle (e.g., xenoliths, abyssal peridotites, and ophiolites) and products of mantle melting (e.g., mid-ocean ridge basalts and ocean island basalts) (Hofmann, 2007; O'Driscoll et al., 2018; Warren, 2016; Warren et al., 2009) provide strong evidence for compositional heterogeneity in the Earth's upper mantle. After decades of work, many questions remain, however, relating to the origins of mantle heterogeneities, the timescales over which they are preserved, and the lengthscales over which they exist (Hofmann, 2007; Hofmann & Hart, 1978; Zindler & Hart, 1986). In this study, we take advantage of a unique and extensively studied (Ashley et al., 2020; Farley, 1995; Hauri et al., 1993; Hauri & Hart, 1994; Poreda & Farley, 1992) suite of oceanic mantle-derived xenoliths from Savai'i (part of the Samoa hotspot track) and Tahiti (Dieu, 1996; from the Societies hotspot track) to better understand the length scales of radiogenic isotopic heterogeneity in the upper mantle.

First described by Hauri et al. (1993), the Savai'i xenoliths exhibit extreme compositional variability. Clinopyroxene (cpx) grains from these samples often display a wide range of incompatible trace element concentrations within a single xenolith, spanning several orders of magnitude for some incompatible elements, with extreme enrichments in many incompatible elements (e.g., heavy REE [rare earth elements], Sr, etc.) in cpx. While large scale (e.g., within a whole suite of xenoliths) trace element heterogeneity is not uncommon in studies of mantle peridotites (e.g., Bizimis et al., 2007; Song & Frey, 1989; Walker et al., 1989), short lengthscale variability, such as that observed in Savai'i xenolith (Hauri et al., 1993; Hauri & Hart, 1994), merit exploration of processes that operate to generate the fine-scale heterogeneities. Hauri et al. (1993) and Hauri and Hart (1994) conclude that, following melt extraction, carbonatitic and silicic melt metasomatism contributed to the extreme trace element variability observed in the Savai'i xenoliths. The observations prompt the question: are multiple, genetically distinct components responsible for the short length scale heterogeneity, or is this the result of complex distribution of a single component?

In the study of radiogenic isotope systems of lithophile elements (i.e., Sr, Nd, and Hf) in spinel peridotites, cpx is generally the phase of interest because it commonly hosts the bulk of these elements in a given sample. Owing to the low abundance of incompatible trace elements in mantle peridotites, these isotopic analyses are typically performed on bulk cpx mineral separates and thus provide insight on the variability within an entire suite of xenoliths, or submarine dredge of abyssal peridotites. Limited studies have also performed bulk mineral analysis of orthopyroxene (opx) and olivine in spinel peridotites (Byerly & Lassiter, 20142015; Comin-Chiaramonti et al., 2001; Deng & Macdougall, 1992; Dodson & Brandon, 1999; Stosch & Lugmair, 1986). While such studies can help to evaluate radiogenic isotopic heterogeneity amongst different mineral phases within a sample, they provide only limited information on the lengthscales of isotopic heterogeneity.

In this study, we utilize new advances in thermal ionization mass spectrometry (TIMS) to explore grain-scale and phase heterogeneity in oceanic mantle xenoliths by performing isotopic analysis on single mineral grains of cpx, opx, and olivine. The introduction of 1013 Ω amplifiers greatly improves signal/noise for small signals associated with isotopic analysis of small quantities of sample analyte. This enables accurate and precise measurement 87Sr/86Sr and 143Nd/144Nd in sample sizes on the order of ∼1 ng for Sr and ∼100 pg for Nd (Koornneef et al., 20142015). We present single mineral 87Sr/86Sr and 143Nd/144Nd analyses performed on cpx, opx, and olivine from four xenoliths from two Polynesian volcanic hotspot localities that each erupt lavas with EM2 (enriched mantle II) compositions: Savai'i (Samoan hotspot) and Tahiti (Societies hotspot). At present this is the first study that provides paired 87Sr/86Sr and 143Nd/144Nd, as well as trace element analyses, in single mineral grains from the oceanic mantle. From this data we are able to hypothesize about the nature of short lengthscale, intra-xenolith mantle heterogeneities and the processes which are responsible for their generation.

2 Samples

We chose four xenoliths from two Polynesian ocean island basalt settings, Savai'i and Tahiti, to evaluate short lengthscale Sr an Nd isotope heterogeneity. Savai'i island is part of the Samoan islands and seamounts, which result from a mantle plume upwelling and melting beneath the Pacific plate (Hart et al., 2004; Workman et al., 2004). Samoan hotspot lavas exhibit an extreme range of 87Sr/86Sr compared to other OIB locations (Adams et al., 2021; Jackson et al., 2007; Jackson & Hart, 2006; Workman et al., 20042006). Tahiti is created by the Society Islands hotspot (Cheng et al., 1993; Duncan et al., 1994), an age-progressive hotspot (White & Duncan, 1996) fed by an upwelling mantle plume (French & Romanowicz, 2015). The xenoliths were selected, in part, based on previously demonstrated short lengthscale trace element heterogeneity (i.e., different cpx within a single sample have variable concentrations of incompatible trace elements). Each of the four xenoliths have been described elsewhere and are confirmed to be harzburgitic (i.e., <5 vol.% cpx) mantle spinel peridotites: SAV-1-1 and SAV-1-28 are discussed in (Hauri et al., 1993; Hauri & Hart, 1994); SAV-09-10 is presented in (Ashley et al., 2020); 85T95-5C is presented in (Dieu, 1996). The mineral modes are reported in Table S1. Two of the Savai'i xenoliths (SAV-1-1 and SAV-1-28) are from a young cinder cone on the eastern side of the island, which is mapped as part of the Mulifanua Volcanics, suggested to have erupted during the last glaciation (Kear & Wood, 1959). The third Savai'i xenolilth (SAV-09-10) was taken from a quarry on the outskirts of the town Salelolonga. The Tahiti xenolith was isolated from a boulder in the Fatauua River on the north side of Tahiti within the Papeete city limits (Dieu, 1996).

3 Methods

Xenoliths were crushed and sieved and individual mineral phases—olivine, opx, and cpx—were picked under a binocular microscope. Both inclusion-poor and inclusion-rich olivine and opx grains were selected in SAV-1-1 and SAV-1-28. Previous studies have shown that the inclusions in the Savai'i peridotites are dominantly CO2-bearing, often with carbonatite like trace element systematics, that appear to have been emplaced during sub-solidus annealing and growth of grain boundaries (Ashley et al., 2020). Samples were mounted in epoxy, and surfaces were exposed by grinding followed by polishing. Major and minor element concentrations in cpx, opx, and olivine were measured on a Cameca SX-100 electron microprobe housed at the University of California Santa Barbara (UCSB) using a 20 keV accelerating voltage, 60 nA beam current, and 10 μm spot size. New and previously published major and minor element concentration data for olivine, opx, cpx, and spinel associated with the four xenoliths are presented in Table S1.

Following electron microprobe analysis, individual mineral grains were removed from grain mounts and extensively acid leached to remove any surficial contamination prior to acid digestion. All leaching and subsequent wet chemistry was carried out at UCSB using double-distilled acids. Mineral grains are first leached in 15N HNO3 for 5 min at 150°C, sonicated for 10 min, then rinsed with 18.2 MΩ H2O. This step is repeated three times. Subsequently, mineral grains are leached with 6N HCl at 150°C for 5 min, sonicated for 10 min, then rinsed with 18.2 MΩ H2O. Lastly, 1N HNO3 is added, and the sample is sonicated for 10 min followed by repeated rinsing with 18.2 MΩ H2O.

Leaching, and subsequent dissolution and Sr and Nd purification by column chemistry, were optimized to reduce total procedural blanks while maintaining high yields for Sr and Nd. Sample mineral grains, unleached BCR-2 powder run as unknown together with mineral grains, and blanks are then dissolved in a 6:4 mixture of concentrated HF:concentrated HNO3 at 130°C. Most grains were <1,000 μm in diameter and were fully dissolved after 24 h using 225 and 150 μl of HF and HNO3, respectively. For larger grains (up to 3,000 μm in diameter) the amount of acid is doubled and grains typically require up to 48 h to fully dissolve. After drying down, 150 μl concentrated HNO3 is added to samples, heated at 130°C, and then dried, with the goal of removing any fluorides present.

Upon complete dissolution, approximately 10% (by mass) of the sample (or blank) solution was removed for major and trace element abundances by ICP-MS. The remaining 90% of sample dissolution was set aside for spiking and Sr and Nd isotopic work, see below. Precise and accurate masses of both aliquots for each sample were obtained on an analytical balance. The 10% solution aliquots, including total procedural blanks, were diluted by a factor of 30 in 2% HNO3, and elemental concentrations were analyzed by Inductively Coupled Plasma Mass Spectrometry (LA-ICPMS) using an Element 2 high-resolution ICPMS at the Center for Mass Spectrometry (CEMS), University of South Carolina following methods outlined in (Ashley et al., 2020; Frisby et al., 2016). Samples were introduced to the instrument with a self-aspirating 147 μl/min teflon nebulizer and a quartz cyclonic spray chamber. All reported elements (except Ti and Mg) where run at low resolution (m/Δm = 300). Following correction of the signals for the full procedural blank, trace element concentrations were calculated by using AGV-1 as external standard (concentrations from Jochum et al., 2016). We used the Mg concentration and signals of each sample for internal standardization, as Mg is known for each sample from electron probe analysis (this approach is inspired by trace element analyses by LA-ICP-MS, where the concentration of the element used for internal normalization is known independently). This approach is advantageous because (a) most of the grains were too small to obtain reliable masses on an analytical balance, (b) the 25Mg signals in medium resolution (m/Δm = 3,000) are interference-free and quite large (>106 cps) as required for an ideal internal standard, and (c) the calculated concentration is independent of any dilutions. In order to evaluate the accuracy of this approach, four small (3–5 mg) aliquots of BCR-2 were dissolved in the same batches of chemistry as the peridotite mineral grains, and from these dissolved aliquots small portions of the BCR-2 solutions (hosting approximately 600–800 picograms of Sr, and approximately 50–70 picograms Nd) were separated and analyzed along with the mineral grains. BCR-2 trace element concentrations were calculated by internal normalization to the MgO concentration of BCR2 of 3.599 wt.% (Jochum et al., 2016). The average trace element concentrations of the four separate BCR-2 aliquot digestions agree to better than 5% with the recommended values for BCR-2 from Jochum et al. (2016) for all reported elements (except Gd and Ta, at ∼13 and ∼15%, respectively).

To the remaining ∼90% of the sample solution, enriched-isotope 84Sr and 150Nd spikes are added, samples are given time to equilibrate, then dried down for column loading. Extraction chromatography resins from Eichrom Technologies are used to isolate Sr and Nd for isotopic analysis, modified after Price et al. (2014). Sr-spec, TRU-spec, and LN-spec resins are first precleaned with successive rinses in ultra-clean acid. 20 μL of precleaned Sr resin is loaded on an acid-cleaned pipette tip column (Koornneef et al., 2015). After the resin is loaded, it is washed with HF, HNO3, and then 18.2 MΩ H2O. The sample is loaded in 400 μL 3N HNO3, rinsed with additional 300 μL 3N HNO3—which elutes the matrix and rare earth elements (REEs) fraction—and then Sr is eluted with 18.2 MΩ H2O. We found that a double pass through the Sr spec column improved Rb separation from Sr. A separate acid-cleaned column is loaded with 100 μL of precleaned TRU resin. The matrix + REE fraction from the Sr column is then loaded in 500 μL 1N HNO3, rinsed with 750 μL 1N HNO3, and a REE fraction is collected in 750 μL 1N HCl. Lastly, 800 μL pre-cleaned LN spec is loaded in a 5 mL, acid-cleaned Pasteur pipette and washed with 6N HCl, 2N HF, and 18.2 MΩ H2O. The Nd-bearing fraction from the TRU-spec column is loaded in 1,000 μL 0.165 N HCl, rinsed with 8,000 μL 0.165 M HCl, and a purified Nd fraction in collected in 1,500 μL 0.165 N HCl and 2,000 μL 0.3N HCl. After drying down the purified Sr and Nd fractions, 20 μL of concentrated HNO3 and 10 μL of 0.025 M phosphoric acid (ULTREX II Ultra Pure, J.T. Baker) were added to the samples and dried down to a ∼1 μL drop.

The purified Sr and Nd fractions are loaded on to 99.999% zone-refined rhenium filaments (H-Cross), using outgassed single filaments for Sr and double filaments for Nd. A parafilm dam is melted at ∼1.2 A on both sides of the center of the filament ribbon in order to minimize sample spreading. For Sr isotopic analysis, the filament is turned down to 0.7 A and a TaF5 activator is added to the filament. Without letting the TaF5 activator dry, the sample is brought up in ∼1 μl 3N HNO3 and added to the TaF5 activator on the filament and the mixture is allowed to dry. The filament current is slowly increased to glowing (∼2.2–2.4 A). For Nd isotopic analysis, the sample is brought up in ∼1 μl 3N HNO3 and added directly to the filament.

Samples were measured for isotopic compositions using a Thermo Fisher Scientific Triton Plus TIMS housed at UCSB. The TIMS is equipped with five 1011 Ω amplifiers and five 1013 Ω amplifiers that are calibrated with a 3.3 pA gain board. 87Sr/86Sr analysis is performed statically using the 1011 Ω amplifiers. Amplifier rotation is not used, and a gain was run once at the start of each new turret. A mass fractionation correction assumes the exponential law and 86Sr/88Sr = 0.1194. For 87Sr/86Sr isotopic analysis, three to five filaments loaded with 1 ng NBS 987 standard are analyzed per turret, together with 10–12 unknowns (i.e., samples, blanks, and BCR-2). The isotopic compositions of unknowns (sample and BCR-2 analyses) are corrected for the offset between the preferred 87Sr/86Sr of NBS987 (0.710240) and the average of the measured values in each turret. Throughout the duration of this study (six months) 1 ng loads of NBS 987 run in the same turrets as unknowns presented in this study yielded an average 87Sr/86Sr of 0.710245 ± 0.000055 (2SD, n = 18) (Figure S2). Following normalization to the NBS987 value of 0.710240, a total procedural blank correction is applied (see below). To evaluate reproducibility, aliquots of BCR-2 reference material (hosting 6–8 ng Sr, i.e., from the same dissolutions processed for trace elements) were spiked with 84Sr and processed through column chemistry with mineral grains and analyzed by TIMS; following correction for the offset between measured and preferred standard values, and correction for blank, the average 87Sr/86Sr of the BCR-2 is 0.705030 ± 0.000053 (2SD, N = 4), which overlaps the BCR-2 value measured on larger sample aliquots by TIMS in Weis et al. (2006) of 0.705005 ± 0.000010 (2SD) (following renormalization of their data to an NBS987 value of 0.710240).

Neodymium isotopic analysis is performed using 1013 Ω amplifiers (Koornneef et al., 20142015). Mass fractionation correction assumes the exponential law and a 144Nd/146Nd = 0.7219. Again, amplifier rotation was not used, and a gain calibration was run once at the beginning of each new turret. As with Sr, an isotopic standard (JNdi) is run three to five times per turret and the isotope compositions of unknowns (mineral grain and BCR-2 analyses) are corrected for the offset between the preferred 143Nd/144Nd of JNdi-1 (0.512099; Garçon et al., 2018) and the average of the measured JNdi-1's in the same turret. Following normalization to the preferred JNdi-1 standard value, a blank correction is applied (see below). Analysis of 1 ng loads of JNdi-1, run together with samples in this study, yielded an average 143Nd/144Nd of 0.512111 ± 0.000022 (2SD, N = 11). Together with peridotite mineral grains examined here, aliquots of BCR-2 (containing approximately 0.5–0.7 ng of Nd) were dissolved, spiked with 150Nd, processed through columns, and analyzed for 143Nd/144Nd by TIMS in the same analytical sessions as the mineral grains. After correcting for the offset between measured and preferred JNdi-1 143Nd/144Nd, and a correction of blank, repeat measurements of BCR-2 yielded an average 143Nd/144Nd of 0.512612 ± 0.000009 (2SD, N = 3). This overlaps with the BCR-2 143Nd/144Nd of 0.512621 ± 0.000012 (2SD) reported by TIMS on larger sample loads in Weis et al. (2006), following correction of their BCR-2 results from the La Jolla to JNdi-1 reference frame from Tanaka et al. (2000) and renormalizing to a JNdi-1 143Nd/144Nd value of 0.512099.

Owing to the small quantity of analyte available (down to 1.1 ng Sr and 0.2 ng for Nd) when targeting single mineral grains (olivine, opx, or cpx) in peridotites for Sr and Nd isotopic analysis, it is necessary to ensure that the sample-to-blank ratio remains high in order to minimize the contribution of the total procedural blank to the measured isotope composition. Total procedural blanks (including all steps from sample dissolution to loading on Re filaments) varied from 3 to 26 pg for Sr, and 0.3 to 1.8 pg for Nd. Excluding two cpx grains that each contained >100 ng Sr (which had [Sr]sample/[Sr]blank > 60,000), sample-to-blank ratios range from 90 to 7,000. The sample-to-blank ratios for mineral grains analyzed for 143Nd/144Nd range from 300 to 10,000. Figure S1 illustrates the magnitude of the blank correction on each of the 87Sr/86Sr (upper panel) and 143Nd/144Nd (lower panel) analysis of xenolith phases as a function of sample size. In most cases, the blank correction resulted in less than a 30 ppm change in 87Sr/86Sr isotopic composition, and is small relative to the intra-sample range observed in xenolith samples SAV-1-28 (5,508 ppm total range in 87Sr/86Sr across different mineral grains) and SAV-1-1 (4,681 ppm). However, blank corrections are larger in two SAV-09-10 cpx (60–82 ppm), which approaches the range in 87Sr/86Sr observed in this relatively homogenous xenolith samples SAV-09-10 (151 ppm). For sample 85T95-5C, the range in 87Sr/86Sr (96 ppm) is greater than the largest blank correction (30 ppm). Additionally, he impact of the Sr blank on the isotopic composition of SAV-1-1 and SAV-1-28 is reduced because the blank composition (0.711) is similar to the isotopic composition of the constituent phases of the xenolith. The blank correction results in <6 ppm change in 143Nd/144Nd for all samples, which is small compared to the total intra-sample range in 143Nd/144Nd observed in xenolith samples SAV-1-28 (118 ppm total 143Nd/144Nd variability), SAV-1-1 (256 ppm), and 85T95-5C (87 ppm).

4 Results

Major element compositions of olivine, opx, and cpx are given in Table S1. Olivine Mg# (molar Mg/(Mg + Fe) × 100) values range from 90.8 to 91.1. The low CaO concentrations of the xenolith olivines are much lower than the CaO concentrations of phenocrystic olivine typical of Samoan lavas (Figure 1), precluding an accidental magmatic cumulate origin for the minerals examined here. The xenoliths from both locations have low modal abundance of cpx (<6%) and high spinel Cr# [molar Cr/(Cr + Al) × 100] that place them in the range of highly melt-depleted abyssal peridotites (Figure 2).

Details are in the caption following the image

Olivine in xenoliths from this study (colored symbols) as well as other xenoliths from Savai'i (gray squares; Hauri & Hart, 1994) and Tahiti (gray triangles; Dieu, 1996) have low CaO compared to olivine phenocrysts hosted in Samoan lavas (gray field; Jackson & Shirey, 2011).

Details are in the caption following the image

Xenoliths in this study have low modal concentrations of clinopyroxene and high spinel Cr# which are both indicative of a high degree of melt extraction. Data are from this study, Hauri and Hart (1994), Dieu (1996), and Ashley et al. (2020). Also shown are Tahiti xenoliths (gray triangles; Dieu, 1996), Savai'i xenoliths (gray squares; Hauri & Hart, 1994), and variably melt depleted abyssal peridotites (black diamonds; Warren, 2016).

Clinopyroxene grains from each of the four xenoliths show within-sample variability in incompatible trace element compositions (Table S2, Figure 3). With the exception of SAV 09-10, the overall trace element pattern is generally consistent amongst cpx grains from the same sample. SAV 1-1 and SAV 1-28 have cpx grains that are the most enriched in incompatible trace elements, with high LREE/HREE concentration ratios (light rare earth element/heavy rare earth element). Cpx grains from these two xenoliths also have large depletions in high field strength elements (HFSEs; e.g., Zr, Hf, and Ti) relative to similarly compatible rare earth elements (Figure 3), which was noted in prior work by Hauri et al. (1993) and Hauri and Hart (1994). SAV 09-10 and 85T95-5C have cpx grains which display broadly similar patterns (e.g., LREE-enriched and HFSE-depleted), but to less-extreme degrees compared to SAV 1-1 and SAV 1-28, and SAV 09-10 cpx do not exhibit the same magnitude Zr and Hf depletions as the other cpx grains in this study. Like cpx grains form SAV 1-1, opx grains from SAV 1-1 are incompatible trace element enriched; the two opx grains from SAV 1-28 are less enriched in incompatible elements than SAV-1-1 opx, and exhibit a concave up REE pattern (Figure 4). Most opx do not display a positive anomaly in HFSE relative to similarly compatible REE that is often seen in opx trace element patterns (Bedini & Bodinier, 1999) (Figure 4a). Olivine grains from SAV 1-1 also display extreme LREE enrichment and HFSE depletions, and like opx have middle-REE that are depleted relative to HREE (Figure 4b).

Details are in the caption following the image

Trace elements in single clinopyroxenes normalized to primitive mantle (McDonough & Sun, 1995). Trace elements were determined by solution ICP-MS from aliquots of dissolutions made for isotopic analysis.

Details are in the caption following the image

Trace elements in single orthopyroxene grains (upper panel) and single olivine grains (lower panel) normalized to primitive mantle. As with clinopyroxene, orthopyroxene, and olivine were determined by solution ICP-MS from aliquots of dissolutions made for isotopic analysis. Note in SAV 1-1 the extreme depletions in high field strength elements (Zr, Hf, and Ti) relative to similarly compatible rare Earth elements. Lanthanum (La) was not measured for all of these samples, so La is not shown.

In addition to the variability in cpx incompatible trace element concentrations in SAV 1-1 and SAV 1-28 (e.g., cpx Sr concentrations ranging from 132 to 443 ppm and 100 to 403 ppm, and cpx Nd ranging from 21 to 71 ppm and 9 to 68 ppm, for SAV 1-1 and SAV 1-28, respectively), minerals from these xenoliths also display extreme variability in their Sr and Nd isotopic compositions (Figure 5). A large intra-sample range in 87Sr/86Sr is present in minerals from SAV 1-28 (0.70863–0.71256; Figure 5b) and SAV 1-1 (0.70988–0.71321; Figure 5a), even when a single outlying opx analysis from each sample is excluded. This contrasts greatly with the very narrow range of 87Sr/86Sr observed in SAV 09-10 and 85T95-5C (87Sr/86Sr range from 0.70561 to 0.70572 and 0.70475 to 0.70482, respectively; Figures 5c and 5d). There are fewer available Nd isotopic measurements for these samples. However, intra-sample heterogeneity for 143Nd/144Nd is identified in the three xenoliths for which 143Nd/144Nd was analyzed.

Details are in the caption following the image

Sr isotope analyses of minerals from SAV 1-1 and SAV 1-28 (upper left and upper right, respectively) illustrate an extreme range in 87Sr/86Sr, whereas clinopyroxene from 85T95-5C and SAV 09-10 display a limited range of 87Sr/86Sr.

Details are in the caption following the image

Olivine, orthopyroxene, and clinopyroxene (cpx) from SAV 1-1 and cpx from SAV 1-28 plot near, but below, the array formed by Samoa shield-stage lavas in Sr-Nd isotope space; a SAV-1-1 olivine grain approaches the lowest 143Nd/144Nd measured in Samoan basalts. Cpx from Tahiti xenolith sample 85T95-5C overlap with the range in isotopes for Tahiti lavas. Sample SAV09-10 was not analyzed for Nd isotopes. SAV 1-1 bulk cpx and glass and SAV 1-28 bulk cpx data are from Hauri et al. (1993). We note that the bulk cpx measurement for SAV-1-1 from Hauri et al. (1993) falls on the Samoan shield basalt array, while the single grain analyses we report fall off the array and plot with Hauri et al. (1993) SAV-1-1 glass and SAV-1-28 bulk cpx measurements: this may be due to the possibility that bulk cpx measurement (which was made by pooling many different cpx grains) included a population of cpx that was not represented in the relatively small number of mineral grains from this sample that were included in our study. Samoan shield and rejuvenated lavas are shown as separate fields; due to the highly altered nature of the samples, the Samoan shield basalt array does not include lavas from dredge 118 of the Samoan ALIA cruise (Jackson et al., 2007).

5 Discussion

5.1 Mineral Grain Trace Element Budgets

The incompatible trace element compositions that we measured in cpx, opx, and olivine are incongruous with the mineral major element compositions. For example spinel Cr# indicate the xenoliths have experienced high degrees of melt extraction which is typically accompanied by extreme depletion in incompatible trace elements—not the extreme enrichments which we observe. Additionally, the measured trace element abundances in the different phases relative to one another are not consistent with those expected from equilibrium partitioning (Agranier & Lee, 2007; Witt-Eickschen & O'Neill, 2005). This is also abundantly clear from the observation that minerals of the same mineral phase within a given sample have different isotopic compositions.

These xenoliths are known to host fluid inclusions, especially within opx and olivine (Ashley et al., 2020; Burnard et al., 1998; Farley, 1995). It is likely that in olivine and opx (phases that typically have very low abundances of incompatible trace elements) the incompatible trace element budgets are in fact dominated by the inclusions and not the mineral phase itself. This is also likely true for cpx but to a lesser extent than opx and olivine, owing to greater partitioning of incompatible trace elements into the cpx structure relative to those phases. It is for this reason that we chose inclusion-rich minerals so that we can infer the isotope composition of the inclusions as well. For example, the olivine lattice will host very little of the incompatible elements of interest in this study, so the Sr and Nd isotopic compositions and incompatible trace element ratios measured on the inclusion-rich olivine separates provides an estimate of the composition of the inclusions hosted in the olivines. A detailed study of the inclusions hosted in the phases is beyond the scope of this study, but studies of inclusions in the Savai'i mantle xenolith suite (including SAV 09-10) are available elsewhere (Ashley et al., 2020; Burnard et al., 1998). For simplicity, we use the names of the analyzed phases (cpx, opx, olivine) knowing that the incompatible trace element composition (and therefore Sr and Nd isotope composition) of these minerals represents an admixture of a mineral and its hosted inclusions.

5.2 Generation of Sr and Nd Isotopic Heterogeneity in Peridotite Xenoliths

While lengthscales of isotopic heterogeneity in the mantle are well documented at the degree-2 (hemispheric) scale (e.g., Castillo, 1988; Hart, 1984), and at the scale of meters down to decimeter scale in exposed peridotite bodies (Borghini et al., 2013; Rampone et al., 2018; Rampone & Hofmann, 2012; Warren et al., 2009), the lengthscale of isotopic heterogeneity at the scale of a single peridotite xenolith is not well known, particularly in oceanic settings. Osmium and lead isotopes have been shown to be heterogenous over the lengthscale of a peridotite hand sample (Blusztajn et al., 2014; Burton et al., 2012; Fitzpayne et al., 2020; Harvey et al., 20062011; Lassiter, 2018; Warren & Shirey, 2012), and Sr and Nd isotopes have been shown to be heterogeneous at the scale of individual abyssal peridotite samples but this effort requires pooling multiple grains to obtain sufficient Sr and Nd for isotopic analysis (e.g., Warren et al., 2009). Thus, there is opportunity to provide new insights into the shortest lengthscales of oceanic mantle peridotite heterogeneity through characterization of individual mineral phases (olivine, opx, and cpx) for paired 87Sr/86Sr and 143Nd/144Nd analysis, which has not been carried out in samples of the oceanic mantle.

We identify clear intrasample Sr and Nd isotopic heterogeneity in two of the four peridotite samples examined here, but two related questions emerge: (a) how was the isotopic heterogeneity generated in the two samples (Savai'i xenolith samples SAV-1-1 and SAV-1-28) that show clear intrasample radiogenic isotopic disequilibrium and, (b) why do two of the peridotite xenoliths (Savai'i xenolith sample SAV 09-10 and Tahiti xenolith sample 85T95-5C) lack clear radiogenic isotopic heterogeneity?

Intrasample isotope heterogeneities in mantle xenoliths are hypothesized to result from a number of processes including interaction with the host magma, radiogenic ingrowth, and metasomatic overprinting (Alard et al., 2005; Giuliani et al., 2018; Jagoutz et al., 1980; Schmidberger et al., 2003; Shu et al., 2014). Furthermore, we identify no relationship between 87Sr/86Sr and Rb/Sr in either SAV 1-1 or SAV 1-28 (Figure S3), both of which exhibit significant intra-sample 87Sr/86Sr heterogeneity, indicating that radiogenic ingrowth is not the mechanism responsible for the intra-sample 87Sr/86Sr variability.

The two xenoliths that exhibit extreme intrasample isotopic heterogeneity, SAV 1-1 and SAV 1-28, show clear evidence for carbonatite metasomatism, and we suggest carbonatite metasomatism is responsible for the extreme intrasample isotopic heterogeneity. Samples SAV 1-1 and SAV 1-28 have high LREE/HREE coupled with clear depletions in Ti, Zr and Hf relative to REE, which are features that have been attributed to carbonatite metasomatism in these two xenoliths (Hauri et al., 1993), and in peridotites at other localities (Coltorti et al., 1999; Klemme et al., 1995; Neumann et al., 2002; Rudnick et al., 1993; Yaxley et al., 1998). Ashley et al. (2020) also find evidence for both carbonatite and silicate metasomatism in Savai'i xenoliths. Critically for the current study, Ti depletions relative to adjacent REE (i.e., low Ti/Eu)—which signal interaction with carbonatite melts (Mattielli et al., 1999; Rudnick et al., 1993)—show an inverse correlation with 87Sr/86Sr in cpx (Figure 7, upper panel) and olivine (Figure 7, lower panel). This provides key evidence that the intrasample 87Sr/86Sr heterogeneity is driven by interaction with carbonatite, where stronger carbonatite signatures (lower Ti/Eu) are associated with higher 87Sr/86Sr. Multiple pulses of metasomatism (as suggested by Burnard et al., 1998 for a harzburgite from the Savai'i xenolith suite), possibly each having different 87Sr/86Sr compositions, could explain the clear relationship between an indicator of carbonatite metasomatism (i.e., low Ti/Eu) and 87Sr/86Sr in single mineral grains from SAV 1-1 and SAV 1-28. For example, Giuliani et al. (2018) argue that multiple coeval yet isotopically distinct fluids variably overprinted the mantle beneath Kimberley, South Africa resulting in peridotites with clear intrasample Sr isotope heterogeneity. We favor a similar mechanism for the origin of isotope variability in the two carbonatite metasomatized Savai'i peridotite xenolith samples examined here.

Details are in the caption following the image

Clinopyroxene (upper panel) and olivine (lower panel) have Ti/Eu that negatively correlate with 87Sr/86Sr. Low Ti/Eu is commonly associated with carbonatite metasomatism, while high Ti/Eu is associated with silicates. The figure shows that carbonatite metasomatism in the SAV-1-1 and SAV-1-28 xenoliths is linked to more radiogenic 87Sr/86Sr.

Care should be taken when interpreting cpx trace element patterns alone as HFSE partition preferentially into opx relative to cpx, which can result in low HFSE/REE in cpx and high HFSE/REE in opx (Bedini & Bodinier, 1999; Byerly & Lassiter, 2015; Condie et al., 2004; Garrido et al., 2000; Harvey et al., 2012; Kalfoun et al., 2002; Witt-Eickschen & O'Neill, 2005). We rule out the cpx-opx partitioning explanation for the carbonatitic signature of SAV 1-1 on the basis of extreme HFSE depletions observed in both cpx and opx (Figures 3 and 4). Furthermore, widespread depletions in HFSE relative to REE are observed in several other Savai'i peridotite opx (Ashley et al., 2020) pointing to carbonatitic metasomatism in the Savaii lithosphere, not to equilibrium cpx-opx partitioning. Additionally, the extreme high Ce/Yb observed in SAV 1-28 and SAV 1-1 (Figure 8) xenoliths is a signature consistent with carbonatitic metasomatism. Equilibrium conditions clearly were not operating in these xenoliths during the carbonatite metasomatic event(s), as evidenced by the extreme Sr and Nd isotopic heterogeneity.

Details are in the caption following the image

Clinopyroxene (cpx) from SAV 1-1 and SAV 1-28 have high Ce/Yb and low Ti/Eu, which are indicators of carbonatite metasomatism. In contrast, SAV 09-10 and 85T95-5C have high Ti/Eu and low Ce/Yb, which is more typical of silicate metasomatism. Also shown are abyssal peridotite cpx for reference (data from Warren, 2016).

In contrast to SAV 1-1 and SAV 1-28, cpx from SAV 09-10 and 85T95-5C exhibit a limited range of 87Sr/86Sr. Unlike SAV-1-1 and SAV-1-28, which exhibit both Ti depletion (low Ti/Eu) and high Ce/Yb associated with carbonatite metasomatism, cpx from SAV-09-10 and 85T95-5C have high Ti/Eu and low Ce/Yb, consistent with silicate metasomatism or a combination of silicate and carbonatite metasomatism (Coltorti et al., 1999) (Figure 8). This is supported by Ashley et al. (2020), who observe opx trace element systematics in SAV 09-10 (and other Savai'i xenoliths) that are consistent with multiple generations of metasomatism, one carbonatite-like and another consistent with silicate metasomatism. This is also consistent with the multiple fluid pulses in a Samoan harzburgite mantle xenolith from Savai'i identified by Burnard et al. (1998). The lack of significant intra-sample 87Sr/86Sr heterogeneity in SAV 09-10 and 85T95-5C may be associated with higher degrees of silicate metasomatism relative to SAV 1-1 and SAV 1-28, a hypothesis that we explore below.

In light of the extreme 87Sr/86Sr heterogeneity in the two xenoliths that exhibit strong evidence for carbonatite metasomatism (Figures 7 and 8), and the far smaller 87Sr/86Sr heterogeneity in the two xenoliths that have stronger silicate metasomatism signatures (Figures 7 and 8), we explore a mechanism for the generation of intrasample 87Sr/86Sr heterogeneity that relates to the dominant metasomatic agent, carbonatite versus silicate melt. Silicate melts are produced in great abundance at hotspot volcanoes, and the islands of Savai'i, Samoa (∼1,700 km2 subaerially exposed) and Tahiti, Society Islands (∼1,000 km2)—where the xenoliths in this study were recovered—are large volcanic edifices composed primarily of mafic silicate material. The peridotite xenoliths from Samoa are interpreted to sample the oceanic mantle lithosphere (Hauri & Hart, 1994), and the Tahiti xenolith is inferred to have a similar origin (Dieu, 1996). Due to the large abundance of silicate melt that passed through the oceanic lithosphere during hotspot volcano construction, the mass of melt passing from the melting asthenosphere through the oceanic mantle lithosphere is expected to be large, yielding large melt/rock ratios in the oceanic mantle lithosphere beneath a hotspot volcano. One hypothesis is that xenoliths sampling the portion of the oceanic lithosphere dominated by this type of silicate melt metasomatism would have reduced isotopic variability that reflects the dominant radiogenic isotopic compositions of the silicate melts that constructed the volcano. Indeed, the xenolith 85T95-5C from Tahiti has Sr-Nd isotopic compositions that are similar to the host volcano, consistent with overprinting by silicate metasomatism (Figure 6). Similarly, the SAV-09-10 xenolith (with cpx varying from 0.7056 to 0.7057), hosted in a rejuvenated Savai'i lava, has a Sr isotopic composition that overlaps with the range found in rejuvenated lavas from Savai'i (0.7053–0.7069; Hauri & Hart, 1994; Konter & Jackson, 2012; Workman et al., 2004; Wright & White, 1987).

Carbonatites are rarely observed erupting in oceanic islands (Hoernle et al., 2002), but their presence has been inferred (Dixon et al., 2008). The time-integrated volume of carbonatite passing through the lithosphere beneath a hotspot volcano is expected to be low compared to the volume of silicate melt, resulting in low melt/rock ratios. Petrographic observation of relatively low modal abundances of metasomatic phases (in particular, low abundances of cpx), even in the most metasomatized xenoliths (including SAV 1-1 and SAV 1-28), supports the hypothesis of a low mass of carbonatitic melt passing through the mantle peridotite (Hauri & Hart, 1994). This carbonatite melt component is sampled by the two xenoliths (SAV 1-1 and SAV 1-28) in this study with strong carbonatite signatures and highly variable isotope compositions.

Hauri et al. (1993) suggested the following reaction for the generation of metasomatic cpx in the Savai'i xenoliths:
urn:x-wiley:15252027:media:ggge22611:ggge22611-math-0001(1)

The absence of surviving carbonate in the xenoliths, and the presence of low modal abundances of metasomatically produced cpx (2.1 and 3.1 modal percent cpx in samples SAV-1-1 and SAV-1-28, respectively; Hauri & Hart, 1993), are consistent with very small volumes of carbonatite that reacted to completion to generate low modal abundances of the metasomatic cpx (which dominates the cpx present in the xenoliths). The presence of abundant CO2 trapped in fluid-vapor inclusions is further evidence of conversion of carbonatite melt to CO2 (Ashley et al., 2020). If the melt/rock ratio for carbonatite is low in the oceanic lithosphere where metasomatism occurred (Hauri & Hart, 1994), then the influence of the carbonatite on the geochemistry of the peridotite may be highly localized and therefore highly heterogeneous at the scale of the carbonatite metasomatized xenoliths. Again, if the xenoliths examined here (SAV-1-1 and SAV-1-28) also experienced multiple pulses of carbonatite melt, and if the different pulses of carbonatite were isotopically heterogeneous, then the carbonatite would have the potential to enhance isotopic variability in the xenoliths down to the grain scale.

One future test for the hypothesis that multiple pulses of carbonatite metasomatism are responsible for the Sr and Nd isotopic heterogeneity in the xenoliths would be to examine within-grain isotopic variability in the metasomatically generated cpx. Different generations of fluid inclusions might be expected to have different isotope compositions. If the timescale between metasomatic events is sufficiently short (to avoid diffusive homogenization), and if the metasomatic melts have geochemically distinct Sr and Nd isotopes, intra-grain Sr isotopic variability may be present. The presence of extreme inter-grain Sr isotopic heterogeneity suggests that the metasomatic process is a disequilibrium, transient process that leaves significant Sr and Nd isotopic variability in its wake. While the current work targets single-grains for Sr and Nd isotopic analysis to evaluate intra-sample isotopic variability, future work targeting intra-grain variability will allow a test of the hypothesis of multiple pulses of heterogeneous carbonatitic melts.

5.3 Preservation of Sr and Nd Isotopic Heterogeneity in Peridotite Xenoliths

In this section, we focus on the mechanism responsible for preserving isotopic heterogeneity in the two Samoan xenoliths from Savai'i—SAV-1-1 and SAV-1-28—that exhibit extreme radiogenic isotopic disequilibrium at the scale of individual opx, cpx, and (for SAV-1-1) olivine grains. These three phases are out of isotopic equilibrium with each other within a particular xenolith, and different grains of the same phase also exhibit isotopic disequilibrium within a single xenolith (Figure 5). An important question is how long Sr and Nd isotopic heterogeneity can be preserved at the scale of a xenolith (4 × 2 cm for SAV-1-1, 2 × 2 cm for SAV-1-28; Hauri et al., 1993) such that individual olivine, opx, and cpx grains can preserve different radiogenic isotopic compositions. This is because, following the most recent metasomatic event that generated the isotopic heterogeneity, diffusion of Sr and Nd limits the amount of time that the isotopic heterogeneity can be preserved in a xenolith prior to the entrainment and eruption of the xenolith in a Samoan lava.

In order to calculate the time for diffusive relaxation of isotopic heterogeneities in the xenoliths, we note that olivine is the most abundant phase and forms an interconnected network of grains in the xenoliths, while opx and cpx grains tend to form isolated “islands.” Additionally, given available diffusivity of Sr in olivine, where a single result is available at 1,275°C for Sr (D = 2 × 10−15 cm2/s; Remmert et al., 2008), the diffusivity of Sr at 1,275°C is over 1 (Sneeringer et al., 1984) to over 2 (Cherniak & Dimanov, 2010) orders of magnitude higher in cpx than that of olivine. We are not aware of published data for Sr diffusion in opx. Nonetheless, because olivine forms an interconnected network in the xenoliths (and cpx and opx do not), and because Sr diffuses slower in olivine than in cpx (the primary host of Sr in the xenolith), diffusion in olivine will be the rate limiting step for diffusive relaxation of Sr isotopic heterogeneities in the xenolith.

We examine volume diffusion through olivine in our calculation for diffusive relaxation. While grain boundary diffusion is more rapid than volume diffusion, the amount of Sr (or Nd) transported along grain boundaries is likely to be small compared to volume diffusion. Furthermore, because Sr diffuses more quickly in olivine than REE in olivine and cpx at the same conditions (Cherniak & Dimanov, 2010; Remmert et al., 2008; Sneeringer et al., 1984), we consider Sr diffusion only, which places a tighter constraint on the amount of time that passed between metasomatism and xenolith eruption (i.e., faster Sr diffusion results in shorter diffusive relaxation times than slower Nd diffusion).

Assuming a xenolith with a spherical geometry and radius (r) of 1 cm, the time for 95% diffusive equilibrium is 0.4 × (r2)/D. Using a Sr diffusivity of 2 × 10−15 cm2/s (i.e., the diffusivity of Sr in olivine, which is the rate limiting step), the time for 95% diffusive equilibrium in the xenolith is ∼6.4 million years. This time for 95% diffusive relaxation of Sr heterogeneities would have been longer if temperatures in the lithosphere experienced by the xenoliths were lower than 1,275°C (i.e., the temperature for the Sr diffusivity measurement in olivine [Remmert et al., 2008]), as suggested by studies of these (Hauri & Hart, 1993) and other related (Ashley et al., 2020) Samoan xenoliths; therefore, we describe the 95% diffusive relaxation time of the Sr in the xenoliths as >6.4 Ma. More complex geometries could be explored in the diffusion model, but such a model would be limited by the diffusion data itself, which is highly uncertain due to differences relating to dislocation density, oxygen fugacity, composition effects, and grain boundary effects (i.e., working on this problem in more sophisticated ways might give the reader a false sense of “precision”). For example, we use a diffusivity of Sr in olivine calculated at 1,275°C, which is not to imply metasomatism happened at those temperatures, instead we are simply using the data that are available.

The key message from this simple model is that the diffusion timescales (a minimum of ∼6.4 Ma) permits preservation of Sr isotopic heterogeneities (at the scale observed in xenoliths) in the mantle beneath Samoa, particularly if the metasomatic event that generated the isotopic disequilibrium was generated by melting associated with the Samoan hotspot. The island has been volcanically active since 5.29 Ma (Koppers et al., 2008), and was last active from 1905 to 1911 (see discussion in Konter & Jackson, 2012). If carbonatite metasomatism responsible for generation of the Sr isotopic heterogeneity in the two Savai'i xenoliths is the result of low-degree melts from the Samoan plume under Savai'i, as suggested previously (Hauri et al., 1993; Hauri & Hart, 1994), then the metasomatism occurred over a time period (i.e., <5.29 Ma) that is less than the diffusive relaxation time of the Sr isotopic heterogeneity in the xenoliths (>6.4 Ma), which will allow some preservation of Sr isotopic heterogeneity (and, hence, Nd isotopic heterogeneity) in the xenoliths. However, if the metasomatic event occurred closer in time to the eruption of the xenolith at the cinder cone (<<1 Ma) than during incipient volcanism at Savai'i (5.29 Ma), the mechanism for preservation of Sr isotopic heterogeneity in the xenolith is even more straightforward.

These results also suggest that, on longer timescales, Sr isotopic variability will be reduced at the short lengthscales examined here. However, lithophile radiogenic isotopic systems with slower diffusivities, like Nd, will be better suited for longer-term preservation of isotopic heterogeneity. Assuming a Nd diffusivity in olivine of 2.9 × 10−16 cm2/s, the time for 95% diffusive equilibrium is ∼44 Ma (which is, again, a lower limit, as temperatures experienced by the xenoliths may be lower than the 1,275°C temperature for the Nd diffusivity measurement in olivine [Remmert et al., 2008]). More efficient preservation of Nd isotopic heterogeneity compared to Sr isotopic heterogeneity over long timescales may result in decoupling of these two isotopic systems in the mantle, which will be the subject of future work.

An additional key outcome of this study is the demonstration of disequilibrium among the phases of the xenolith, which precludes two-phase thermometry. This is because calculation of the temperature of mantle residence of xenoliths requires the assumption that the relevant phases are in equilibrium. However, the cpx, which have a metasomatic origin, are almost certainly not equilibrium phases. We know this because radiogenic isotopes (this study) and trace elements (Hauri & Hart, 1994) in the cpx are not in equilibrium with each other, or with the other phases in the xenoliths. Furthermore, new major element analyses reported on cpx from SAV-1-1 and SAV-1-28 reported here (Table S1) are not in equilibrium with cpx analyses on these two xenoliths reported by Hauri and Hart (1994). Nonetheless, we provide temperatures using the a cpx-opx thermometer in Table S1 for reference (Brey & Kohler, 1990), but the isotopic and major and trace element disequilibrium observed in SAV-1-1 and SAV-1-28 makes it difficult to assert that these calculated temperatures are representative of the mantle residence of the xenoliths.

5.4 A Different EM2 Component Sampled by Carbonatite in Samoa

The mineral-specific isotopic and trace element data on the Samoan xenoliths most strongly influenced by carbonatite metasomatism—SAV-1-1 and SAV-1-28—show a clear trend when considered together in 87Sr/86Sr versus Ti/Eu space (Figure 7) that suggest mixing between two components: a low Ti/Eu component with high 87Sr/86Sr, consistent with a carbonatite contribution, and a component with higher Ti/Eu and lower 87Sr/86Sr. Two component mixing is also supported by the 87Sr/86Sr versus 143Nd/144Nd data, where the samples with lower 87Sr/86Sr and higher 143Nd/144Nd trend in the direction of the less geochemically enriched lavas in Samoa (possibly a component similar to the geochemically depleted portion of the Samoan shield array). In contrast, the xenoliths with more geochemically enriched 87Sr/86Sr and 143Nd/144Nd, associated with low Ti/Eu in Figure 7, clearly trend off of the array formed by Samoan shield lavas. It is relatively straightforward to show that the geochemically enriched (carbonatite influenced) endmember sampled by xenoliths does not have substantially higher 87Sr/86Sr (and, by inference, substantially lower 143Nd/144Nd) because the trends formed by the xenolith cpx and olivine in 87Sr/86Sr versus Ti/Eu space place a firm upper limit on the 87Sr/86Sr of the low Ti/Eu carbonatite endmember. Carbonatites have low Ti/Eu, and extrapolating the trends in Figure 7 to a Ti/Eu value of zero shows that the maximum 87Sr/86Sr of the carbonatite endmember is ∼0.714 (i.e., on the 87Sr/86Sr vs. Ti/Eu array defined by the xenoliths, a Ti/Eu value of zero defines the maximum possible 87Sr/86Sr on the trends shown in Figure 7). Thus, the geochemically enriched carbonatite endmember has an 87Sr/86Sr value only marginally higher than the highest 87Sr/86Sr identified in the single-grain mineral analyses from SAV-1-1 and SAV-1-28.

What is notable is that this carbonatite endmember sampled in Samoa lies off the 143Nd/144Nd—87Sr/86Sr mantle array defined by Samoan shield lavas. This indicates that the carbonatite components recognized here in the metasomatized xenoliths sample a different EM2 composition than the EM2 composition sampled by Samoan lavas (which form a more shallowly sloping array in 143Nd/144Nd—87Sr/86Sr space than the xenoliths). The extreme EM2 carbonatite metasomatized xenoliths are from the same island, Savai'i, where extreme EM2 lavas (>0.720) were dredged. Thus, all known lavas and xenoliths with 87Sr/86Sr >0.712 are from the same island, a critical location for defining EM2 in the mantle. Notably, the xenoliths—which form an array in 143Nd/144Nd—87Sr/86Sr space that terminates at 87Sr/86Sr of ∼0.714—clearly reveal an enriched mantle composition, with lower 143Nd/144Nd at a given 87Sr/86Sr, that is different from enriched mantle compositions thus far sampled in the Samoan shield array. Lead isotopes in bulk cpx separates from both xenoliths—SAV-1-1 and SAV-1-28—were reported in Hauri et al. (1993) (206Pb/204Pb varies from 18.845 to 18.877, and 208Pb/204Pb varies from 39.720 to 39.845), and are quite unlike Pb isotopic compositions identified in Samoan lavas from Savai'i with the most extreme EM2 signatures from Samoa (where 206Pb/204Pb varies from 18.954 to 19.070, and 208Pb/204Pb varies from 39.347 to 39.452). Therefore, the Pb isotopic data are consistent with the suggestion that the Savai'i xenoliths sample a different enriched endmember than Samoan basalts. This heterogeneity in the EM2 endmember—with one EM2 component sampled by carbonatites in Savai'i xenoliths and the another sampled by Savai'i lavas—may result from heterogeneity in crustal materials that were subducted and recycled into the Samoan plume. Importantly, and irrespective of the origin of these endmembers, the carbonatite metasomatic signatures in these peridotites reveal an additional component in the plume not readily recognized in the erupted magmas. Therefore, the detailed study of isotope systematics of plume influenced lithosphere, at the mineral scale level, opens up new opportunities to understand the different species of the mantle zoo.

6 Conclusions

We show that some peridotite xenoliths that sample the oceanic lithosphere exhibit extreme grain-scale variability in Sr and Nd isotopic compositions, which mirror extreme heterogeneity in trace element abundances at the grain scale. This can have implications for the interpretation of bulk-mineral isotope analysis depending on the ubiquity of within-xenolith heterogeneities. In the case of the xenoliths studied here, the style of metasomatism (carbonatitic vs. silicate) appears to be the primary control on the formation of short scale isotopic heterogeneity, as the xenoliths dominated by carbonatitic metasomatism exhibit significant grain-to-grain Sr and Nd isotopic heterogeneity, while the two xenoliths dominated by silicate metasomatism do not. How long these heterogeneities can be preserved against diffusion at mantle conditions is a function of the temperature and the amount of time that passed since generation of the intra-xenolith Sr and Nd isotopic heterogeneities by carbonatite metasomatism. In the Samoa xenoliths that display short scale isotopic heterogeneity, Sr isotopic heterogeneity can be preserved against diffusive homogenization for at least ∼6.4 Ma. The timeframe of volcanic activity (and, hence, likely metasomatism associated with proximal mantle melting) at the host volcano, from 5.3 Ma to recent historical eruptions, make it likely that metasomatic events associated with Samoan hotspot volcanism at Savai'i occurred sufficiently recently for preservation of Sr (and, thus, Nd) isotopic heterogeneity at the lengthscale of the xenoliths. We show that the EM2 carbonatite endmember sampled in Samoa lies off the 143Nd/144Nd—87Sr/86Sr mantle array sampled by Samoan shield lavas, and indicates the carbonatites that metasomatized the xenoliths have a different EM2 composition than the EM2 composition sampled by Samoan lavas. This heterogeneity in the EM2 endmember may result from heterogeneity in continental crustal materials that were subducted and recycled into the Samoan plume.

Acknowledgments

The authors thank Felix Genske and Andrea Giuliani for reviews. M.G. Jackson acknowledges funding from NSF grants OCE-1736984, EAR-1429648, and EAR-1900652. The authors thank Frank Spera and Stan Hart for discussion. M.G. Jackson is thankful for mentorship from Erik H. Hauri, who initially suggested that we should target olivines for Sr isotopic analysis. M. Bizimis acknowledges support from NSF grant OCE-1624315.

    Data Availability Statement

    All data are currently available in the EarthChem data repository at https://doi.org/10.26022/IEDA/112050.