Introduction

Xenolithic peridotites carried in rift-related alkali basalts from Tanzania offer a unique opportunity to investigate the effects of rifting and intrusion of rift-related melts on cratonic subcontinental lithospheric mantle (SCLM) that has experienced multiple previous enrichment episodes (Dawson 1984, 1999, 2002; Rudnick et al. 1992, 1993; Lee and Rudnick 1999; Chesley et al. 1999; Vauchez et al. 2005). The East African Rift (EAR) is widely attributed to the impingement of a plume or plumes (sensu lato) that delivered heat and fluids to the lithospheric mantle, leading to its thinning and enrichment in incompatible elements. Evidence for plume involvement includes the similarity of some EAR magmas to ocean island basalts (OIBs) (e.g., Marty et al. 1996; Bell and Simonetti 1996; George et al. 1998; Rogers et al. 2000), radiogenic Os isotopes in mantle xenoliths precipitated by rift magmas (Chesley et al. 1999), the volume of extruded and underplated melts (Latin et al. 1993) and geophysical data (e.g., Ebinger et al. 1989; Weeraratne et al. 2003).

Although it is postulated that plumes should exhibit a range of Li isotopic compositions, since they may contain deeply subducted slabs consisting of variably altered and dehydrated oceanic crust and mantle with distinct δ7Li (Li isotope composition relative to L-SVEC standard: {[(7Li/6Li)sample/(7Li/6Li)L-SVEC − 1] × 1,000} Chan et al. 1992, 2002; Zack et al. 2003; Nishio et al. 2004, 2005; Wunder et al. 2006), a direct connection between the lithium isotope composition of intraplate basalts and inferred slab components is yet to be established. Most OIBs have δ7Li overlapping with MORB (δ7Li = +2 to +6, see Ryan and Kyle (2004) and references therein), but a few OIBs and associated glasses have δ7Li trending toward both heavier and lighter δ7Li values (Gurenko and Schmincke 2002; Kobayashi et al. 2004; Ryan and Kyle 2004; Nishio et al. 2005). In particular, the δ7Li of the HIMU source has been suggested to be heavier than MORB (+6 to +7), either due to incorporation of slab-modified isotopically heavy mantle wedge materials (Jeffcoate and Elliott 2003) or incorporation of less altered (and hence, less dehydrated) oceanic crust (Nishio et al. 2005). In contrast, highly altered and dehydrated basaltic crust and pelagic sediment have been suggested to give rise to low δ7Li in the EMI source [inferred from unusually light Li in clinopyroxenes from mantle xenoliths from far-east Russia and SW Japan (Nishio et al. 2004)], although the basis of this interpretation has been recently questioned (Jeffcoate et al. 2007; Rudnick and Ionov 2007).

Volcanic rocks in the EAR have trace-element and isotopic compositions comparable to those of OIBs, but occupy a unique field in isotope compositional space compared to OIBs (Paslick et al. 1995; Rogers et al. 2000; Macdonald et al. 2001). Part of this uniqueness may stem from their interaction with the ancient, incompatible trace element-enriched lithospheric mantle. Rogers et al. (2000) suggest that the present-day plume has Sr and Nd isotope compositions of 0.7035 and 0.51275, respectively, which is the isotopic composition of the common end-member of Neogene to Recent Kenya rift basalts.

We report here the Li, Sr and Nd isotope and trace-element compositions of minerals from peridotite xenoliths from the Labait volcano, which erupted on the margin of the Archean Tanzanian craton. These data are used to determine the effects of the rift-related magmas on the Li, Sr and Nd isotope composition of the cratonic lithospheric mantle and to gain further insights into the composition of the plume that is feeding magmas of the EAR. These xenoliths have been the subject of previous detailed petrographic, major- and trace-element and Re-Os isotope investigations that documented heating and intrusion of the ancient lithosphere by rift-related melts (Dawson 1999; Lee and Rudnick 1999; Chesley et al. 1999).

Geology

The Gregory rift, the eastern branch of the EAR system in Tanzania, cuts along the border between the Archean Tanzanian craton and the Mozambique fold belt to the east. The latter is composed largely of Archean crust that was reworked in at least two Proterozoic orogenies, the ∼2.1 Ga Usagaran orogeny and the ∼600 Ma pan-African orogeny (Möller et al. 1998). Earliest manifestations of volcanic activity in the EAR system date back to 50–30 Ma in Kenya and the magmatic activity has moved southward since the mid Tertiary, leading to magmatism in Tanzania between 1 and 5 Ma (Dawson 1992). Voluminous basaltic to trachytic magmatism, forming large shield volcanoes in the mobile belt, was followed by faulting at about 1.2 Ma and eventually small-volume explosive, highly alkaline nephelinitic to carbonatitic volcanism, which is restricted to the craton and reworked craton margin (Dawson 1992; Smith and Mosley 1993) where the Labait volcano erupted.

The olivine–melilitite that forms the Labait volcano marks the southernmost point of the propagating Gregory Rift, where rifting is incipient (Dawson 1999). It sampled the SCLM beneath the eastern edge of the Tanzanian craton in the form of relatively large mantle xenoliths (Lee and Rudnick 1999). The volcano was indirectly dated to ca <0.4 Ma on the basis of U-Pb ages of metasomatic zircons in one of the peridotite xenoliths (Rudnick et al. 1999).

The mantle lithosphere beneath Labait is highly refractory to depths of ∼140 km, and shows a stepwise increase in fertility between 140 and 150 km (Lee and Rudnick 1999). Re-Os ages up to 2.9 Ga in peridotitic chromites from the upper 140 km of mantle lithosphere indicate stabilization by the Late Archean and testify to its longevity, despite the propagation of the rift into the craton; Proterozoic to future Os model ages for peridotites from the deeper, garnet-bearing, more fertile lithosphere may indicate later formation (i.e., in the Proterozoic) or modification of Archean lithosphere by interaction with the EAR magmas (Chesley et al. 1999). Multiple lines of evidence attest to intrusion and heating of the ancient lithosphere by rift-related magmas. These include: (a) breakdown of garnets into coronas (Lee and Rudnick 1999), (b) the presence of metasomatic veins containing more Fe-rich olivine, phlogopite, orthopyroxene, chromite, sulfide, 400 Ka zircons (Rudnick et al. 1999) and rutile with enhanced Zr concentrations on the rims, indicative of heating (Watson et al. 2006) and (c) recent Re addition (Chesley et al. 1999; Burton et al. 2000). Seismic signatures from the craton margin are transitional between lithospheric (in terms of anisotropy) and asthenospheric (in terms of seismic velocities) mantle, suggesting that the base of the lithosphere has been modified by a plume, which is likely responsible for the rifting (Vauchez et al. 2005).

Samples

Xenoliths from Labait can be divided into garnet-bearing peridotites, chromite-bearing garnet-free peridotites (containing chrome-rich spinel and Al2O3-poor opx typical of garnet-facies peridotites, but do not contain garnet) and spinel-facies peridotites (containing aluminous spinel and Al2O3-rich opx; Lee and Rudnick 1999; see also key parameters in Supplementary Appendix 1). Most also contain glassy pockets or veins (Dawson 1999). The mantle xenolith population is made up of >75% Fe-rich peridotites (of which ca 25% contain chrome diopside) and ca. 25% harzburgites or cpx-poor lherzolites, with rare glimmerites and wehrlites (Lee and Rudnick 1999). Olivine is fresh, with mosaic-porphyroclastic textures in garnet peridotites. Orthopyroxene forms small to large groundmass grains. Clinopyroxene displays well-developed morphology in wehrlites and garnet-bearing peridotites, and also occurs in garnet reaction rims and glassy patches, where it appears to be secondary. Garnet has mostly decomposed to kelyphite and an outer corona assemblage, with remnants occurring in the center of kelyphite. Olivine compositions may be heterogeneous, depending on habit and whether it occurs as inclusions or within the matrix. Olivine inclusions in garnet can have higher forsterite content than matrix olivines, reflecting Fe-enrichment post-dating garnet growth (Lee and Rudnick 1999).

Spinel-facies and garnet-free chromite-bearing peridotites are highly refractory (FeO = 6.5–7.4 wt%), whereas most garnet-bearing peridotites are more fertile, though still depleted relative to estimates for the primitive upper mantle composition (FeO = 7.6 to 10.6 wt%; Lee and Rudnick 1999). In contrast, Fe-rich peridotites (FeO = 11.2 to 14.1 wt%) are olivine-rich and have been interpreted as products of pervasive melt-rock reaction, rather than cumulates, on the basis of their significant Os concentrations (e.g., 0.9–2.9 ppb, Chesley et al. 1999). Garnet lherzolite LB-45 is unique in that it is derived from near the lithosphere–asthenosphere boundary at a temperature of ca 1,400°C, has a major element composition similar to primitive upper mantle estimates (Lee and Rudnick 1999) and radiogenic Os isotopic composition (Chesley et al. 1999). Based on these features, Chesley et al. (1999) suggested that this sample may derive from the plume beneath the EAR. More recent PGE analyses of this sample, using a high-temperature Carius tube digestion technique (Becker et al. 2006), show that the whole rock 187Os/188Os is somewhat less radiogenic than measured by Chesley et al. (1999), who used lower-temperature Carius tube digestion. These differences show that there are at least two populations of Os in the sample: an ancient, unradiogenic component, presumably armored within chromites, and a radiogenic component that is presumably contained within more acid-accessible, late-stage sulfides. Moreover, its platinum group element concentrations are much lower than primitive mantle (PM) and are accompanied by anomalous Re-enrichment. These features suggest that this sample is probably a piece of the original, Archean cratonic mantle that was strongly overprinted by rift-related magmas (Becker et al. 2006) near the base of the lithosphere.

Sample preparation, chemistry and analytical methods

Mineral separates, hand-picked to optical purity and cleaned in Milli-Q (18 MΩ) H2O in an ultrasonic bath for 15 min, were dissolved in a 3:1 mixture of concentrated HF–HNO3 in Savillex screw-top beakers and dried down. The dried residua were replenished with HCl–HNO3 mixture and HCl until clear solutions were obtained. The three-column chemistry for olivine is based on the procedure described by Moriguti and Nakamura (1998) and closely follows that described in Rudnick et al. (2004). Each column is loaded with Bio-Rad AG 50W-X12 (200–400 mesh) resin. Column calibrations using BHVO-1 as the low-Cr-Mg end of spectrum and a powdered peridotite as the high-Cr-Mg end were performed to ensure 100% yield for both types of matrix. In addition, the yield was monitored by taking aftercuts and measuring their Li contents on the Finnigan ELEMENT 2 magnetic sector Inductively Coupled Plasma Mass Spectrometer (ICPMS) in the Department of Geology at the University of Maryland. Analyses were rejected if more than 3% of Li was lost during any single column pass, which occurred during the processing of <10% of sample aliquots. Li loss was sporadic and could not be linked to differences in sample type or composition. Repeat analyses of samples in our dataset with up to 3% loss agreed within the analytical uncertainty.

Prior to Li isotope analysis, Na/Li was measured semi-quantitatively using voltage ratios obtained on the Nu Plasma Multi-Collector (MC) ICPMS at the University of Maryland, as ratios >5 are known to interfere with accurate Li isotope determination (Tomascak et al. 1999). Samples with Na/Li > 5 were dried down and passed a second time through column 3 to achieve greater Na–Li separation.

Purified Li solutions (∼50 ppb Li in 2% (v/v) HNO3) were carried by a mixture of Ar and N from an auto-sampler (ASX-100 Cetac Technologies) through a desolvating nebulizer (Aridus Cetac Technologies) fitted with a PFA spray chamber and micro-nebulizer (Elemental Scientific) to the Nu plasma MC-ICPMS at the University of Maryland. The standard bracketing method described in Teng et al. (2004) was used. Lithium concentrations were determined by comparison of sample signal intensities with those obtained from L-SVEC, with a two sigma uncertainty of ∼10%, as determined using isotope dilution methods (Teng et al. 2004).

The accuracy and precision of the data are assessed by repeated analysis of rock standards and the pure Li in-house standards Li-UMD-1 and of IRMM-016, with yielded average δ7Li of +54.6 ± 0.8 (2σ, n = 27) and +0.2 ± 1.0 (2σ, n = 17), respectively, during the course of the analyses presented here. In general, external precision (2σ on repeat runs) is ≤1‰ based on long-term analyses of these standards (Teng et al. 2004), where isotopic compositions are expressed as per mil (‰) deviations from the NIST L-SVEC Li2CO3 standard. The δ7Li for international rock standards agree within uncertainty with previously published values [note that serpentinite UB-N is the international standard with the closest matrix to our ultramafic samples (Supplementary Appendix 2)]. Total procedural blanks were always ≤400 pg Li (average 250 pg, n = 5) and blank contributions were, on average, ≤0.07% of the measured δ7Li, which is insignificant compared to the external precision.

Clinopyroxene separates were spiked with a 150Nd spike (99% abundance) prior to dissolution, which follows the procedure described above. Strontium and Nd cuts were obtained in two stages. In the first column, filled with Bio-Rad AG 50W-X12 (200–400 mesh) resin, Sr and Nd cuts were collected by eluting with 2.5 N HCl and 6.0 N HCl, respectively. Rb and Sr are separated quite efficiently during this first stage (total Rb in Sr cut = 0.2%). Nevertheless, a second-stage purification of the Sr cut was performed, using Eichrom Sr-SPEC resin. Separation of Nd from other REE was achieved with an Eichrom Ln-resin.

Strontium cuts were loaded on Re filaments with a Ta activator and isotope ratios measured in dynamic collection mode on the VG Sector 54 TIMS at the University of Maryland, using a 86Sr/88Sr of 0.1194 for mass fractionation correction. Multiple 87Sr/86Sr analyses of standards are reported in Supplementary Appendix 3.

Like lithium, purified Nd solutions were introduced into the Nu Plasma MC-ICPMS with an Aridus desolvating nebulizer. Neodymium isotope ratios were measured in static collection mode, where the collector configuration for Nd isotope analysis is such that all Nd isotopes plus 147Sm are analyzed (to monitor interference of 144Sm on 144Nd), with mass 144 on the axial channel. Mass fractionation is corrected by normalizing to a 146Nd/144Nd of 0.7219. Multiple 143Nd/144Nd analyses of AMES Nd standard averaged 0.512191 ± 0.000027 (2σ, n = 44). Rock standards and samples were normalized to a 143Nd/144Nd for AMES of 0.512138 (in-house TIMS measurements). Using this method, average 143Nd/144Nd for Shin-Etsu is 0.512121 ± 0.000032 (n = 8) and for BHVO-1 it is 0.512989 ± 0.000039 (n = 4). A single analysis of BCR-1 gave 0.512601 ± 0.000024 (2s.e.) (Supplementary Appendix 3). The blank concentration was 9.3 ng and thus did not require a correction.

Trace-element abundances in cpx were determined using either a frequency-quintupled Nd:YAG laser operating at 213 nm or an ArF Excimer operating at 193 nm (both from New Wave Research), coupled to the ELEMENT 2. Helium was used as the sample gas. Laser spot sizes ranged from ∼80 to 100 μm, at an 8–10 Hz laser repetition rate. A typical time-resolved analysis involved ∼20 s of background acquisition on the sample gas, followed by laser ablation for 40 s. Each batch of ca 20 analyses was preceded and followed by analysis of standards NIST 610 and BCR-2g, the latter analyzed as an unknown (Supplementary Appendix 4). Absolute trace-element abundances were obtained after processing the time-resolved spectra off-line (using a modified version of LAMTRACE by Simon E. Jackson) including background subtraction and normalization to 49Ti, based on analyses of Ti by electron microprobe (Lee and Rudnick 1999).

Results

Lithium concentrations for olivine separates from Labait range from 1.8 to 4.8 ppm and δ7Li from +2.5 to +6.6 (Table 1). Lithium isotopic compositions of olivines correlate positively with Li concentration (Fig. 1a) and negatively with Fo and NiO contents (Fig. 2). Olivines in three Fe-rich peridotites have the heaviest Li (+5.2 to +6.6) and some of the highest Li concentrations (Fig. 1a), while at the other end of the spectrum, Li contents of olivines from the spinel and garnet-free chromite-bearing harzburgites are the lowest (∼2 ppm) and they have the lightest δ7Li (∼+3). Olivines from all three garnet-bearing peridotites, including the fertile garnet lherzolite (LB-45), have identical Li isotopic compositions (δ7Li = +4.7), which are intermediate between the Fe-rich peridotites and refractory harzburgites, and similarly intermediate Li concentrations (2.3–3.4 ppm).

Table 1 Lithium isotope compositions (‰), concentrations (ppm), apparent olivine–cpx distribution coefficients and inter-mineral fractionations (Δ7Li), where applicable, for peridotitic mineral separates and the host melilitite
Full size table
Fig. 1
figure 1

δ7Li (‰) against Li concentration (ppm) in a olivine (correlated with an r 2 of 0.59) and b cpx from Labait. Error bars are 1‰ and 10%, respectively. Estimate for range of Li contents in olivine from fertile to moderately depleted mantle from Eggins et al. (1998) and Seitz and Woodland (2000)

Full size image
Fig. 2
figure 2

a Mg-number [Mg number: 100Mg/(Mg + Fe)] against δ7Li (‰) (correlated with an r2 of 0.89) and b NiO (wt%) against δ7Li (‰) (correlated with an r2 of 0.80) for peridotitic olivine separates from Labait. Qualitative mixing trends for the addition of Fe and Li and dilution of NiO are also shown. Mg-numbers and NiO contents from Lee and Rudnick (1999)

Full size image

In all but one sample (gt-free harzburgite LB-21), pyroxenes have lithium concentrations that are equal to or lower than those in coexisting olivines, with apparent ol/cpxD (the concentration of Li in olivine divided by that in cpx) of 1.1–2.0, consistent with equilibrium partitioning (Table 1; Fig. 3, Brenan et al. 1998; Eggins et al. 1998; Seitz and Woodland 2000). The δ7Li of the orthopyroxenes are equal to or lighter than that of coexisting olivines by up to 7‰ and those of clinopyroxenes are consistently lighter than olivines by 2–12‰ (Fig. 3). These inter-mineral fractionations do not correlate with equilibration temperature and there is no correlation between δ7Li, Li content and Mg number in clinopyroxenes (Fig. 1b).

Fig. 3
figure 3

Apparent olivine–cpx partitioning coefficients (ol/cpx D Li) plotted against inter-mineral Li isotope fractionation (Δ7Liol–cpx). Gray bar shows range of equilibrium distribution coefficients (adapted from Eggins et al. 1998; Brenan et al. 1998; Seitz and Woodland 2000). Peridotites from far-east Russia (adapted from Rudnick and Ionov 2007) shown for comparison

Full size image

Strontium isotope compositions for seven cpx separates range from 0.702945 in the fertile garnet lherzolite (LB-45) to 0.705584 in an Fe-rich peridotite (Table 2). Unfortunately, Nd was lost for some of these samples and there is not enough material left to attempt repeat measurements. Nevertheless, ε Nd values for four cpx separates range from a low of −8.6 in a spinel peridotite to a high of +4.8 in an Fe-rich peridotite and an analytically indistinguishable value of +4.4 in the fertile garnet lherzolite, LB-45 (Table 2). δ7Liol shows a rough negative correlation with 87Sr/86Sr (Fig. 4a, with one outlier), and a positive correlation with ε Nd values in cpx (Fig. 4b), though the limited data make these correlations less than robust.

Table 2 Sr and Nd isotope compositions of peridotitic cpx mineral separates and of the Labait host melilitite, as in Supplementary Table 1; averaged values (avg) shown in italics; ε Nd is per 10,000 deviation from present-day chondrite (value of Wasserburg et al. 1981)
Full size table
Fig. 4
figure 4

a87Sr/86Sr and b143Nd/144Nd in cpx against δ7Li in olivine from Labait

Full size image

With the exception of the most incompatible elements, PM-normalized trace-element patterns of clinopyroxenes in the peridotites from Labait are mostly similar, with moderate, negative slopes in the REE (LREEN ∼10–50, HREEN ∼1–3) that flatten in the LREE (Supplementary Appendix 4, Fig. 5). The cpx in a garnet harzburgite (sample LB-4) has markedly lower incompatible trace-element concentrations, whereas that in a garnet-free chromite-bearing harzburgite (sample LB-21) has markedly higher incompatible trace-element abundances than the remainder of the samples. The latter sample shows extreme enrichment in Zr and Hf and is the only sample with Zr/Hf >PM and with ol/cpxDLi <1; the rest have PM-normalized Zr/Hf <1 and ol/cpxDLi >1. The similarity between the cpx patterns in garnet-bearing and garnet-free peridotites is surprising, considering that cpx is the only mineral into which these elements partition in the garnet-free samples, yet garnet should take in significant HREE when it is present. However, cpx in garnet-free samples do have significantly higher Sc concentrations than in garnet-bearing ones, suggesting that they are indeed garnet-free, as Sc is highly compatible in garnet.

Fig. 5
figure 5

Trace-element patterns of cpx in peridotites from Labait. Normalized to primitive mantle of McDonough and Sun (1995)

Full size image

Discussion

Li isotopic disequilibria

Lithium concentrations in both olivines and pyroxenes from the Labait peridotites range to significantly higher values than “normal” mantle, as defined by Seitz and Woodland (2000). These enrichments are evidence for Li addition to the samples some time after the melt depletion that they experienced in the Archean. With the exception of the very refractory harzburgite (LB-21), which has a low ol/cpxDLi of 0.6, the ol/cpxDLi values of the Labait peridotites fall within the range considered to reflect equilibrium partitioning (ol/cpxDLi = 1.1–2.0, Brenan et al. 1998; Eggins et al. 1998; Seitz and Woodland 2000). This result is noteworthy, given that in many peridotites metasomatized by a silicate melt, cpx is preferentially, sometimes strongly, enriched in Li relative to olivine (Seitz and Woodland 2000; Paquin and Altherr 2002; Scambelluri et al. 2006; Rudnick and Ionov 2007), resulting in disequilibrium ol/cpxDLi <1 (Fig. 3). In contrast, the rather large differences in Li isotopic compositions between coexisting olivine and pyroxene (Table 1 and Fig. 3), discussed below, show that they are not equilibrated with respect to Li isotope composition. Thus, isotopic disequilibrium exists despite apparent elemental equilibrium.

The systematic offset of δ7Li between olivine and cpx in the Labait peridotites is in the same direction as that documented by some previous studies of mantle xenoliths (Seitz et al. 2004; Rudnick and Ionov 2007), with cpx systematically lighter than coexisting olivine. Whereas Seitz et al. (2004) found a maximum Δ7Liol–cpx of ∼3.5‰ and attributed this to equilibrium fractionation, Δ7Liol–cpx of our samples ranges between 2 and 12. Similar and even larger fractionations (Δ7Liol–cpx = 3–23) are reported by Rudnick and Ionov (2007) for olivine–clinopyroxene pairs from far-east Russian xenoliths, from some of the same localities investigated by Nishio et al. (2004), who documented unusually low δ7Licpx (down to −17), but did not report any data for coexisting olivines. In contrast, other studies of peridotite xenoliths have found that Δ7Liol–cpx ranges to both positive and negative values (−2.4 to +1.2‰, Magna et al. 2006; −3.6 to +13.5‰, Jeffcoate et al. 2007), and does not correlate with equilibration temperature. It is thus not clear whether measurable equilibrium fractionation exists between olivine and cpx at mantle temperatures, but it is apparent that minerals in the Labait peridotites are not in Li isotope equilibrium.

Interestingly, whereas Li concentrations appear to be equilibrated between olivine and cpx, Mg–Fe are not. The Mg number of cpx in the only two Fe-rich peridotites that contain this phase (LB-46 and LB-51) is much too high relative to that of coexisting olivine to be in equilibrium (Mg number cpx = 89.4 and 89.5 vs. Mg number olivine = 86.9 and 87.5, respectively, Fig. 6a, Supplementary Appendix 1). In contrast, these two samples fall on a correlation between Li in cpx and Fo content of coexisting olivines (Fig. 6b), but do not fall on a correlation between Li and Mg number in cpx (Fig. 6c). It thus appears that diffusion of Li in cpx is faster than Fe–Mg diffusion in cpx and at least as fast as Fe–Mg diffusion in olivine, such that the olivine Mg number reflects reaction with a metasomatic melt, whereas cpx does not (yet). This also means that the Li abundances of cpx are linked to the Fe and Li ingress seen in olivine. Furthermore, it suggests that the cpx Mg number in the Fe-rich peridotites records an older signature (higher Mg number), whereas olivine equilibrated with an infiltrating melt leading to lower Mg number, consistent with faster Fe–Mg diffusion in olivine relative to cpx (Klügel 2001). If so, the isotopic disequilibrium is not due to primary olivine being combined with secondary cpx crystallized from a melt. In fact, from textural evidence, only cpx in garnet-free peridotite LB-6 was considered to be secondary (Lee and Rudnick 1999).

Fig. 6
figure 6

a Mg number olivine against Mg number cpx, b Mg number olivine against Li concentration in cpx (ppm) and c Mg number cpx against Li concentration in cpx (ppm)

Full size image

Recent studies have suggested that Li diffusion in cpx is faster than in olivine (Jeffcoate et al. 2007; Rudnick and Ionov 2007; Parkinson et al. 2007). If so, the low δ7Licpx relative to δ7Liol in the Labait peridotites, coupled with the lack of correlations between δ7Licpx and other parameters, may reflect very recent diffusion of Li into cpx (but not olivine), and attendant kinetic isotope fractionation, possibly during transport in the melilitite host. Even small amounts of Li addition via diffusion from a grain-boundary source, such as the host melilitite, may produce large isotopic effects, while having little effect on overall concentration. The degree of isotopic fractionation is governed by the relative mobility of 7Li and 6Li, which relates to the exponent β, and the concentration contrast between the mineral and the source of Li (Richter et al. 2003). Modelling shows that for simple boundary conditions (one-dimensional diffusion in a semi-infinite medium of a 2 mm diameter sphere), using a diffusivity of Li in cpx of 10−11 m2 s−1 (Coogan et al. 2005), β values of 0.19–0.27 (Parkinson et al. 2007) and a concentration contrast of 9 ppm (similar to that between the host melilitite and a 1 ppm Li cpx), total δ7Li in the cpx decreases by ∼5.7–8.2‰ in only ∼5 min. At the same time, total Li concentration of the hypothetical cpx increases from 1 ppm to only 1.7 ppm. Thus, very recent Li addition to the cpx may have produced the observed isotopic disequilibria without strongly affecting the elemental partitioning between olivine and cpx.

The apparent Li elemental equilibrium between cpx and olivine for our sample suite suggests that this process was incipient, and had not lead to the significant Li enrichment and extreme isotope fractionation such as that seen in cpx in peridotites from far-east Russia (Rudnick and Ionov 2007). In this scenario, the δ7Licpx is unrelated to the Li and Fe enrichment event evidenced by many Labait peridotites. We thus propose that, for the Labait samples, olivine best records the δ7Li of the metasomatic agent responsible for the Fe-enrichment (giving rise to the correlations we see) and, due to its modal abundance, the δ7Li of the whole rock. We thus focus the rest of our discussion on Li in olivine.

The Fe-enrichment event

The Li budget in mantle peridotite is controlled by olivine, which is the most abundant mineral and typically has Li concentrations (∼1.6–1.9 ppm) that are slightly higher than those of coexisting pyroxenes (Ryan and Langmuir 1987; Brenan et al. 1998; Eggins et al. 1998; Seitz and Woodland 2000). Melt depletion leads to lower concentrations in all minerals, due to the moderate incompatibility of Li, whereas metasomatism leads to higher concentrations for the same reasons (e.g., Seitz and Woodland 2000). As mentioned above, most samples in our study have higher Li concentrations than olivine from primitive peridotites and thus probably experienced Li addition (Fig. 1a).

These elevated Li contents, combined with the correlations between δ7Li and Li content (Fig. 1a), forsterite content and NiO in olivine (Fig. 2) suggest that the peridotites are the products of mixing between two components. One component is the ancient mantle lithosphere, which is refractory (Fe-poor) and isotopically light. The second component is an isotopically heavy, Li- and Fe-rich melt. The Re-enrichment seen these peridotites is likely also due to the same event (Chesley et al. 1999). If so, this event must represent an earlier phase of rift magmatism, since there is no correlation between Re-enrichment and Os isotope composition. Similar reaction between Li- and 7Li-depleted lithospheric mantle with a Li-rich and isotopically heavy melt has been suggested for cratonic peridotites from southern Africa, but it is unclear whether the relatively light Li in ancient refractory mantle is due to earlier metasomatism or whether it reflects a secular evolution toward more isotopically heavy mantle due to addition of high-δ7Li crustal components (Bell et al. 2005).

The Fe- and Li-rich and high-δ7Li character of olivines in Fe-rich peridotites and the deep-seated garnet-bearing peridotites (including fertile garnet lherzolite, LB-45) are interpreted to be related to a prolonged metasomatic event involving plume-derived silicate melts associated with the EAR. It is likely that these peridotites, including cpx, have experienced metasomatic overprinting during earlier phases of rift magmatism. Indeed, the precipitation of Pleistocene zircons (Rudnick et al. 1999), coupled with enhanced Zr abundances on the rims of metasomatic rutiles in a vein from one of the ancient Labait harzburgites (Watson et al. 2006) suggests that this section of lithosphere has probably seen repeated injection of rift-related magmas over the last million years or so. A positive correlation between δ7Liol and 143Nd/144Nd in cpx and a rough negative correlation between δ7Liol and 87Sr/86Sr in cpx (Fig. 4a, with the exception of a single Fe-rich peridotite) would be consistent with mixing between a metasomatic component associated with the rift that has high 143Nd/144Nd, low 87Sr/86Sr and heavy Li relative to the ancient, refractory mantle (Fig. 4b). In this context, the similarity of the cpx trace-element patterns (with the exception of highly incompatible elements Th, U, Nb and Ta, and Li abundances), in Fe-rich peridotites, garnet peridotites and spinel peridotite reflect interaction with similarly fractionated melts at different times (and depths) that could explain why the patterns are so similar, while at the same time the cpx have variable radiogenic isotope compositions.

In addition to identifying an ancient lithospheric mantle component and a rift component, another component is apparent in the lithospheric mantle beneath Tanzania. The single Fe-rich peridotite that plots off the 87Sr/86Sr in cpx vs. δ7Liol trend (LB-51) has unusually radiogenic Sr relative to its mildly unradiogenic Nd, thus plotting in an unusual space on a 143Nd/144Nd vs. 87Sr/86Sr diagram (Fig. 7). Several lavas from the EAR (Macdonald et al. 2001), including the host melilitite measured here, plot in this area of the diagram, as do cpx from mantle xenoliths from Lashaine. These samples may reflect an exotic component present within the lithospheric mantle beneath Tanzania. Despite the complexities apparent in the radiogenic isotope data for Labait peridotites, we can conclude that the least radiogenic (with regard to Sr), plume-like isotope composition is observed for the most fertile, PM-like sample (LB-45), whose significance is discussed in the following section.

Fig. 7
figure 7

87Sr/86Sr-143Nd/144Nd (εNd on right axis) correlation diagram for cpx mineral separates and melilitite from Labait. Olmani peridotites from Rudnick et al. (1993); Lashaine peridotites (authors’ unpublished data); field for East African carbonatites, DM, HIMU and EMI mantle end-members from Bell and Tilton (2001); Kenya rift volcanics (divided into those penetrating the Tanzanian craton, the rifted craton and the mobile belt) from Macdonald et al. (2001) and estimated plume composition from Rogers et al. (2000); Tanzanian ultramafics from Cohen et al. (1984), inset shows extended diagram including samples with extremely radiogenic Sr

Full size image

Sr-Nd-Li isotope composition of the East African plume

Estimating the isotopic composition of a plume from derivative lavas erupted through ancient lithosphere can be difficult because the lavas may interact with and become contaminated by isotopically evolved lithospheric crust and mantle during ascent, as demonstrated previously for lavas of the EAR (e.g., Paslick et al. 1995; Bell and Simonetti 1996; Macdonald et al. 2001). Samples from the lithospheric mantle that have been metasomatized by plume-derived melts, as postulated for Labait peridotites (Lee and Rudnick 1999), may better reflect the isotopic affinity of the plume, provided their trace-element inventory has been sufficiently overprinted. This is obviously not the case for peridotites with evolved Sr and Nd and unradiogenic Os isotope compositions (Chesley et al. 1999), such as the shallow and ancient peridotites from Labait (spinel-facies and garnet-free peridotites, as well as the one Fe-rich peridotite that sits at high Sr relative to Nd (LB-51), discussed above). The host melilitite (LB-2) also appears to have acquired an anomalously high 87Sr/86Sr signature during passage through the lithospheric mantle (Table 2, Fig. 7).

In contrast, the isotopic composition of the plume feeding the EAR may be deduced from the samples that show the greatest overprinting by rift-related magmas [Fe-rich peridotites and the high-temperature fertile lherzolite (LB-45)]. The latter sample is derived from the lithosphere–asthenosphere boundary (ca 1,400°C equilibration temperature), where it was heavily overprinted by plume-derived asthenospheric melts, and it has a highly unusual major-element composition for cratonic peridotite that is similar to primitive upper mantle estimates (Lee and Rudnick 1999). It contains radiogenic 187Os/188Os that is similar to that of the host melilitite and a glimmerite (LB-49) that are believed to represent those of the plume-like asthenospheric mantle (Chesley et al. 1999), and its non-mantle-like PGE concentrations and evidence for the presence of multiple Os isotope components are consistent with such overprinting (Becker et al. 2006).

Even though at first glance Fe-rich peridotites (FeO 11.2–14.1 wt%) seem to have experienced a higher degree of melt infiltration than fertile peridotite LB-45 (FeO 8.3 wt%), they likely formed by melt-rock reaction between asthenosphere (plume)-derived melts and lithospheric mantle. This would lead to decreasing MgO and NiO contents in the residual melt without affecting the relative abundances of most other elements, in accord with the lower MgO and NiO contents in the Fe-rich peridotites. In contrast, LB-45 is suggested to have reacted with a “primitive” asthenospheric melt close to the lithosphere–asthenosphere boundary.

Although we did not measure 206Pb/204Pb, which defines the HIMU mantle reservoir, the Sr-Nd isotope composition of LB-45 (0.7029 and 0.5129, respectively) is similar to HIMU (Fig. 7), and we suggest that this value is likely to best reflect the Sr and Nd isotopic composition of the plume feeding this portion of the EAR. This is supported by the HIMU-like isotopic compositions of some carbonatites and mafic lavas from Tanzania and Kenya (Paslick et al. 1995; Bell and Simonetti 1996; Macdonald et al. 2001; Bell and Tilton 2001; Furman et al. 2006). Because LB-45 is derived from the base of the lithosphere, and its bulk rock composition has been very strongly overprinted, its HIMU signature is proposed to have a sublithospheric origin (like the lavas from Turkana, Kenya, Furman et al. 2006), rather than originating in ancient enriched continental lithospheric mantle, as proposed by Paslick et al. (1995) for a number of the Tanzanian rift-related lavas.

The Fe-rich peridotites from Labait form one end-member of the lithium mixing line (Figs. 1a, 2), with δ7Li up to +6.6. These rocks are suggested to have formed by rift-related melt-rock reaction (Lee and Rudnick 1999; Chesley et al. 1999), thus their Li isotope composition likely represents that of the source of the rift basalts. The fertile garnet peridotite that was used to define the Sr and Nd composition of the plume beneath Labait, LB-45, has a slightly lower δ7Li (+4.7), although within uncertainty (±1‰) of the Fe-rich peridotites. Collectively, these data suggest that the plume source of the rift basalts in Tanzania has a heavy lithium isotope composition similar, though not exclusive to OIBs with HIMU affinity (δ7Li up to +7; Jeffcoate and Elliott 2003; Ryan and Kyle 2004; Nishio et al. 2005).

Plume vs. lithosphere signatures

The approach taken above, to define the EAR plume composition from mantle xenoliths that either formed directly from, or were strongly overprinted by rift-related magmas, can be extended to localities beyond Labait. Clinopyroxenes in dunites and wehrlites from the Olmani volcano, northern Tanzania, formed by rift-related carbonatite overprint of ancient refractory mantle lithosphere (Rudnick et al. 1993). These peridotites have highly variable parent–daughter ratios, yet remarkably homogeneous 87Sr/86Sr (0.7034–0.7039) and 143Nd/144Nd (0.51281–0.51283) that fall between the HIMU and EMI components. The isotopic homogeneity at this locality, despite highly variable parent–daughter ratios, indicates that the metasomatism was recent and that the carbonatitic melt effectively overprinted any original isotopic signature of these rocks (Rudnick et al. 1993). In contrast, the plume beneath Labait is inferred to have HIMU-affinity based on the Labait peridotites described above (see Sect. “Sr-Nd-Li isotope composition of the East African plume”).

The small but significant difference in Sr-Nd isotope composition of samples considered to best represent the plume component at Labait and at Olmani, respectively, which are located within ∼150 km of each other, may point to isotopic heterogeneity in the plume. Plume heterogeneity has been suggested for carbonatites that are located at a similar distance from each other beneath the Kenya rift and that show HIMU and EMI affinities, respectively (Bell and Tilton 2001).

It is also possible that the EMI component in the metasomatic carbonatite from Olmani was acquired during percolation of the melt through metasomatized mantle (e.g., Bell and Simonetti 1996), rather than indicating true plume heterogeneity. For example, rift-related volcanic rocks from northern Tanzania have OIB-like trace-element compositions, whereas their isotopic compositions, in particular unradiogenic Nd, were suggested to point to a lithospheric mantle source that has been enriched by small-volume melts (Paslick et al. 1995). In addition, the low δ7Li of +3.4 for a primitive olivine–melilitite from the nearby Oldoinyo Lengai carbonatite volcano (Halama et al. 2007), similar to refractory LREE-enriched spinel peridotite LB-31 (+3.3, see Sect. “The Fe-enrichment event”), may reflect assimilation of ancient refractory metasomatized mantle by the olivine–melilitite, or a lithospheric source.

We argue that the convergence of Sr-Nd isotope systematics for different materials and localities (carbonatite-metasomatized peridotites from Olmani; Rudnick et al. 1993; authors’ unpublished Data; ankaramite, peridotitic cpx and vein material from Pello Hill and Eledoi; Cohen et al. 1984) suggests that a 143Nd/144Nd of ∼0.51280–0.51284 and 87Sr/86Sr of ∼0.7034–0.7035 expresses the plume characteristics beneath some regions of Tanzania, while beneath Labait the plume may be more HIMU-like. The uniformity of Sr-Nd isotope ratios from Olmani are all the more striking as the mantle beneath Tanzania has been variably and unsystematically metasomatized (Macdonald et al. 2001; Dawson 2002).

The Sr-Nd isotope compositions of peridotites from Labait and Olmani differ slightly from the common end-member for Neogene to Recent basalts that has a more radiogenic 143Nd/144Nd of ∼0.51275 and 87Sr/86Sr of ∼0.7035 and that has been suggested to represent the isotopic composition of the plume beneath the Kenya Rift, which comprises the area between Turkana and northern Tanzania (Rogers et al. 2000). Recently, the plume generating volcanism in Turkana and Afar, north of Tanzania, has been proposed to be heterogeneous, containing lenses of isotopically distinct materials (Furman et al. 2006) and our data suggest that this may also be true for the plume beneath Tanzania.

Summary and conclusions

The δ7Li of peridotitic olivine, combined with 87Sr/86Sr and 143Nd/144Nd of coexisting cpx are used to unravel the rift-related and pre-rift history of the lithospheric mantle beneath the Tanzanian craton margin (Labait) and to place constraints on the composition of the plume feeding the Tanzanian section of the EAR:

  1. (1)

    Addition of Fe and isotopically heavy Li is inferred from positive correlations between Li, δ7Li and FeO of olivine in peridotites from Labait that are proposed to have interacted with plume-derived silicate melts (Lee and Rudnick 1999; Chesley et al. 1999).

  2. (2)

    A fertile, high-temperature garnet lherzolite with radiogenic Os, inferred to have been entrained from close to the lithosphere–asthenosphere boundary, may best represent the composition of the plume component beneath Labait (Lee and Rudnick 1999; Chesley et al. 1999). It has Sr-Nd isotope compositions (87Sr/86Sr = 0.7029 and 143Nd/144Nd = 0.51286) similar to the HIMU mantle reservoir, which is an end-member in EAR-related volcanism. This is slightly less evolved than a previous estimate for the plume beneath the rift (0.70354 and 0.51275, respectively; Rogers et al. 2000) and may reflect plume heterogeneity or, alternatively, point to assimilation of old metasomatised mantle for the latter estimate, which is based on rift-related lavas.

  3. (3)

    The Fe- and Li-rich peridotites have δ7Li of +5.2 to +6.6. These, along with the deep-seated garnet lherzolite, which has a δ7Li of +4.7, may best represent the Li isotope composition of the EAR plume beneath Labait, which is isotopically heavy, lying near the high end of estimates for MORB and overlapping with previous estimates of the δ7Li of HIMU oceanic basalts.