Earth and Planetary Science Letters 450 (2016) 221–232

Contents lists available at ScienceDirect

Earth and Planetary Science Letters

www.elsevier.com/locate/epsl

Unusual δ56Fe values in Samoan rejuvenated lavas generated in the

mantle

Jasper G. Konter a,b,∗, Aaron J. Pietruszka b,c, Barry B. Hanan b, Valerie A. Finlayson a, Paul R. Craddock d,e, Matthew G. Jackson f, Nicolas Dauphas d

a Department of Geology and Geophysics, School of Ocean and Earth Science and Technology, University of Hawaii, Manoa, Honolulu, HI 96822, USAb Department of Geological Sciences, San Diego State University, San Diego, CA 92182, USAc United States Geological Survey, Denver Federal Center, Denver, CO 80225, USAd Origins Laboratory, Department of the Geophysical Sciences and Enrico Fermi Institute, The University of Chicago, Chicago, IL 60637, USAe Schlumberger-Doll Research, Reservoir Geosciences, Cambridge, MA 02139, USAf Department of Earth Science, University of California Santa Barbara, Santa Barbara, CA 93106, USA

a r t i c l e i n f o a b s t r a c t

Article history:Received 14 March 2016Received in revised form 9 June 2016Accepted 18 June 2016Available online xxxxEditor: B. Marty

Keywords:Samoarejuvenated lavashigh δ56Feisotope fractionationmantle source

Several magmatic processes contribute to the Fe isotope composition of igneous rocks. Most basalts fall within a limited range of δ56Fe (+0.10 ± 0.05�), although more differentiated lavas trend towards slightly elevated values (up to +0.3�). New data for basalts and olivine crystals from the Samoan Islands show higher δ56Fe values than have previously been reported for basalts worldwide. Common magmatic processes – from partial melting of average mantle to subsequent differentiation of melts – cannot sufficiently fractionate the Fe isotopes to explain the elevated δ56Fe values (∼+0.3�) in rejuvenated Samoan lavas. Instead, a mantle source with an elevated δ56Fe value – in conjunction with effects due to common magmatic processes – is required. The Samoan mantle source is known to be unique in its radiogenic isotope composition and indications that melting of the Samoan mantle source can generate elevated δ56Fe values in lavas comes from: (1) High f O2 values of Samoan lavas and their likely sources affecting Fe isotope fractionation during melting; (2) Metasomatism that caused elevated δ56Fe in the Samoan mantle, as observed in xenoliths; and (3) Involvement of a pyroxenite source lithology, based on the Zn/Fe ratios and TiO2 (and other high field-strength element) abundances of the lavas, that can generate melts with elevated δ56Fe values. Two models are presented to explain the elevated δ56Fe values in Samoan lavas: a metasomatized source (∼+0.07�) or the presence of a pyroxenite source component (∼+0.12�). Both models subsequently elevate δ56Fe values with both partial melting (∼+0.14�) and fractional crystallization (∼+0.1�). These processes may be related to an upwelling mantle plume with a pyroxenite component, or melting of previously metasomatized upper mantle.

© 2016 Elsevier B.V. All rights reserved.

1. Introduction

Magmatic systems have traditionally been characterized with major and trace element compositions, however stable Fe iso-tope ratios have recently been found to vary measurably in ig-neous rocks, potentially providing additional constraints on mag-matic processes. For example, stable Fe isotope fractionation has been suggested during partial melting (Williams et al., 2004;Weyer and Ionov, 2007; Dauphas et al., 2009a), magma differ-

* Corresponding author at: Department of Geology and Geophysics, School of Ocean and Earth Science and Technology, University of Hawaii, Manoa, Honolulu, HI 96822, USA.

E-mail address: jkonter@hawaii.edu (J.G. Konter).

http://dx.doi.org/10.1016/j.epsl.2016.06.0290012-821X/© 2016 Elsevier B.V. All rights reserved.

entiation (Schoenberg and von Blanckenburg, 2006; Teng et al., 2008, 2013; Schuessler et al., 2009; Sossi et al., 2012; Zambardi et al., 2014), fluid-related processes during differentiation (Heimann et al., 2008), and during thermal diffusion in melt subjected to temperature gradients (Huang et al., 2009; Richter et al., 2009). In many studies, the fractionation is assumed to be an equilib-rium process, where different phases have preferences for dif-ferent Fe isotopes partly as a function of the Fe oxidation state (e.g., Polyakov and Mineev, 2000; Shahar et al., 2008; Dauphas et al., 2014). Thus, when a mineral such as olivine excludes Fe3+ in contrast to an enrichment in associated melt, Fe isotope fractionation is possible (e.g., Sossi et al., 2012), potentially am-plified by kinetic fractionation effects (e.g., Huang et al., 2009;Richter et al., 2009).

222 J.G. Konter et al. / Earth and Planetary Science Letters 450 (2016) 221–232

Fig. 1. Trends in Fe isotope composition versus SiO2 in igneous rocks. Arrows indicate approximate Fe isotope evolution during magmatic processes. A) Two groups of trends in literature data: increasing in δ56Fe at low SiO2 and at high SiO2. Sources provided in legend, major element data for Societies: B. Marty, unpublished data. B) Data from this study shows Samoan values are elevated above the literature data and our own baseline data. Two altered samples are labeled, and two duplicated samples are circled.

Modeling partial melting or magma differentiation (and crys-tallization) with Fe isotope fractionation involves calculating the thermodynamically stable equilibrium phases and their isotopic compositions based on estimated equilibrium fractionation fac-tors for mineral pairs and mineral–melt pairs. However, these are only known to first-order because their magnitudes are close to the level of precision currently attainable. Currently, the best mineral–melt fractionation estimates (relevant to the terrestrial mantle) suggest that olivine and pyroxene differ from melt in δ56Fe value by −0.1� to 0�, while plagioclase and magnetite have higher values than melt (e.g., Polyakov and Mineev, 2000;Williams et al., 2005; Shahar et al., 2008; Dauphas et al., 2009a, 2014; Weyer and Seitz, 2012). These values appear to account for some but not all trends in terrestrial igneous Fe isotope composi-tions with melting and differentiation (Schoenberg and von Blanck-enburg, 2006; Weyer and Ionov, 2007; Teng et al., 2008, 2013; Dauphas et al., 2009a; Schuessler et al., 2009; Sossi et al., 2012;Williams and Bizimis, 2014).

In addition to magmatic processes, the Fe isotope composi-tions of igneous samples also depend on the original composition of the mantle that melted, which is not only a function of prior episodes of melt removal, but also mantle metasomatism and (re-lated) differences in oxidation state (e.g., Williams et al., 2004;Weyer and Ionov, 2007; Johnson et al., 2010). The worldwide Fe isotope dataset for igneous samples shows outliers with elevated δ56Fe values in particular (Fig. 1), and in ocean island basalts (hotspots) these values may reflect a unique source composition and/or mineralogy (Teng et al., 2013).

The ocean island basalts that are currently known to have dis-tinct δ56Fe values are samples from Polynesia. In particular, sam-ples from mainly the Society Islands have elevated δ56Fe values compared to other oceanic hotspots, including Hawai‘i (Teng et al., 2013). Teng et al. (2013) ascribe the elevated values to a different degree of melting, a more significant effect from fractional crys-tallization, and potentially source heterogeneity. Instead, Williams and Bizimis (2014) suggest that a pyroxenite source mineralogy is an important alternative.

The Society Islands are representative of an extreme radiogenic isotopic composition, of which the Samoan Islands are the most extreme example (the so-called EM2 mantle end-member; Jackson

et al., 2007), and are thus of particular interest for Fe isotopes. In this study, we examine the compositions of lavas and mineral separates from Samoan volcanoes that are significantly more frac-tionated in Fe isotopes than known lavas with similar MgO con-tents. The samples show distinctly high δ56Fe values for basalts, compared to the well-defined trend of the Hawaiian Kilauea Iki lava lake (Teng et al., 2008) and compared to the evolution of con-tinental igneous rocks (Heimann et al., 2008; Fig. 1). We discuss the cause for these compositions in terms of petrologic processes, mantle sources, and alteration processes.

2. Geological background

The Samoan islands form an age-progressive chain of volcanoes near the northern corner of the Tonga Trench (Fig. 2). The onset of volcanism in the main shields of these volcanoes shows an age progression from early construction in the east at Vailulu‘u, to on-set of volcanism dating back ∼5 Ma at the western Samoan island of Savai‘i, and continuing with submarine volcanoes to possibly ∼24 Ma (Hart et al., 2004). This age progression is in agreement with the expected absolute plate motion over that time period (e.g., Koppers et al., 2008), and as such Samoa is thought to be a hotspot volcanic chain in a tectonically complex area. Volcanic eruptions in individual Samoan volcanoes span at least 5 Ma, as demonstrated by the westernmost island of Savai‘i, which was ac-tive at ∼5 Ma (Koppers et al., 2008) and last erupted in 1905. Over this time span, eruption was not continuous, as exemplified by ero-sional contacts in the lava pile, and this has led to the recognition of multiple stages of activity.

The most significant erosional contact separates what is known as the main shield stage from a later, volumetrically less sig-nificant, rejuvenated stage (e.g., Natland, 1980; Konter and Jack-son, 2012). The rejuvenated lavas make up ∼2% of the erupted volume on the island of Savai‘i, and they cover the entire is-land. The areal coverage decreases to about half on Upolu and to a small fraction on Tutuila (Natland, 1980; Konter and Jack-son, 2012). The rejuvenated lavas are most easily distinguished from the shield-stage lavas by their radiogenic isotope compo-sitions. In particular, the rejuvenated lavas can be distinguished from earlier shield-stage Samoan volcanism by their relatively low

J.G. Konter et al. / Earth and Planetary Science Letters 450 (2016) 221–232 223

Fig. 2. Map of the Samoan Islands and seamounts. Sample locations are shown with the same symbols as Fig. 1. On the islands of Savai‘i, Upolu, and Tutuila, the blue color shows coverage of the rejuvenated lavas. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

206Pb/204Pb ratios (approximately <18.9; e.g., Koppers et al., 2008;Konter and Jackson, 2012). The shield lavas are influenced by multiple mantle components, displaying two globally-unique end-member compositions (Jackson et al., 2007, 2014): 1) the high 87Sr/86Sr submarine lavas around Savai‘i, and 2) the high 3He/4He lavas that were found on Ofu island. The shield compositions are thought to relate to two distinct mantle plume components, while the rejuvenated composition has a more complex origin.

The high 87Sr/86Sr shield component and the rejuvenated com-ponent form separate trends in Sr–Nd–Pb isotope space that con-verge on a common component that features high 3He/4He ratios. This pattern has been interpreted as a mantle plume source, con-sisting of a matrix of high 3He/4He material that hosts the other components (Jackson et al., 2014). The unique high 87Sr/86Sr com-ponent most likely derives from continental sediment that was recycled into the mantle at a subduction zone (White and Hof-mann, 1982; Jackson et al., 2007). The rejuvenated lavas may de-rive their unusual areal coverage and composition from mantle source changes and differences in volcanic plumbing related to plate bending (Konter and Jackson, 2012), a process previously pro-posed for all Samoan volcanism (e.g., Natland, 1980). Given these unusual source compositions and processes, Fe isotope variations in Samoa should not only be evaluated in light of magmatic pro-cesses, but also source characteristics.

3. Sample descriptions

The sample suite studied here consists of a set of igneous baseline samples (for Fe isotopes), and a set of Samoan samples. The set of baseline samples include continental basalts (Columbia River, Snake River Plain), basalts from a mid-oceanic ridge (Aus-tralian Antarctic Discordance) and several hotspots (Réunion, Gala-pagos, Hawai‘i), as well as reference materials (multiple analyses of BCR-1, BHVO-1, BHVO-2, BIR-1, W-2, IF-G). We also included mineral separates from a Kilbourne Hole peridotite, and a dacite glass from Kilauea’s East Rift Zone, recovered from a drill core when melt flowed into the well (Teplow et al., 2009). The Samoan suite contains both shield and rejuvenated lavas, including shield lavas with the two extreme Sr and He isotope compositions. The shield sample set (Fig. 2) consists of dredged lavas and olivine sep-arates (2006 ALIA cruise, R/V Kilo Moana, KM0506) from Savai‘i,

including the most extreme EM2 end-member samples ever mea-sured (e.g., Jackson et al., 2007; Koppers et al., 2008). The high 3He/4He shield component is represented by samples from Ofu Is-land (Jackson et al., 2014). The shield samples are supplemented by subaerial samples from rejuvenated flows on Savai‘i and Upolu. The samples in the dredges and from subaerial collection range from nearly picritic basalts to trachytes, spanning a range in mag-matic evolution (e.g., fractional crystallization). The dredge sam-ples analyzed for Fe isotopes range from basalt to hawaiites and basaltic trachy-andesite (e.g., Koppers et al., 2008), the Ofu sam-ples straddle the basalt–basanite boundary (Jackson et al., 2014), and the rejuvenated lavas range from basalt to basanite (Table 1). The rejuvenated samples are all taken from fresh historical flows on the islands, the Ofu shield samples originate from a fresh boul-der and dike, while the submarine samples have been variably al-tered. Major and trace element abundances and radiogenic isotope compositions of the shield samples have been presented elsewhere (Jackson et al., 2007).

4. Methods

Major and trace element abundances and radiogenic isotope compositions are already available for the samples from Savai‘i (submarine) and Ofu (Jackson et al., 2007). New Fe isotope data are presented here for these samples, accompanied by baseline δ56Fe values for several other igneous samples from a variety of tectonic settings. In addition, new major-element and Rb and Ba abundances, and isotope ratios are presented for the rejuvenated lavas from Savai‘i (subaerial) and Upolu.

For isotope analyses, the samples were wrapped in heavy paper or plastic and broken by hammer, crushed in a ceramic or WC jaw crusher, rinsed in 18 M� water, picked under a microscope, and then dissolved in a HF–HNO3 mixture. For 26 of the 28 samples, Fe isotope compositions were obtained using a separation technique employing AG1-X4 resin in HCl medium, and subsequent analy-sis with a 58Fe–57Fe double spike technique, using a Nu Plasma 1700 multi-collector inductively-coupled plasma mass spectrom-eter (MC-ICP-MS) at San Diego State University (SDSU). Briefly, Fe isotopes were analyzed avoiding polyatomic interferences in (pseudo)high-resolution mode, and using a desolvating nebulizer (DSN-100). Isobaric interferences by Ni and Cr were corrected by

224 J.G. Konter et al. / Earth and Planetary Science Letters 450 (2016) 221–232

Table 1Sample major element compositions.

Location (source) Sample ID SiO2 Al2O3 TiO2 FeO MnO CaO MgO K2O Na2O P2O5 Total

Samoan samplesShield, Savai‘i D115-03 49.16 12.59 3.50 9.74 0.13 8.17 7.97 3.09 2.85 0.51 97.71Shield, Savai‘i D115-18 52.28 14.20 2.45 7.09 0.07 6.08 5.09 4.01 3.64 0.36 95.27Shield, Savai‘i D115-28 50.00 13.44 3.30 7.86 0.12 9.62 6.37 2.73 3.15 0.52 97.11Shield, Savai‘i D128-01 48.94 12.68 2.90 9.73 0.17 10.68 8.22 1.58 2.41 0.37 97.68Shield, Savai‘i D128-21 48.95 12.11 2.28 9.67 0.17 9.81 11.65 1.23 2.12 0.25 98.24Shield, Ofu Ofu-04-03 44.97 9.75 3.83 13.51 0.19 9.79 14.74 1.08 1.72 0.42 100.00Shield, Ofu Ofu-04-15 45.01 11.78 4.88 12.31 0.17 10.48 10.89 1.14 2.82 0.52 100.00Rejuvenated, Savai‘i SAM6-7 47.73 13.20 2.90 11.48 0.168 10.18 10.51 1.01 2.74 0.34 100.24Rejuvenated, Savai‘i SAM6-8 45.11 11.97 3.70 11.99 0.170 10.56 12.18 1.77 2.63 0.36 100.45Rejuvenated, Upolu SAM6-5 44.77 14.12 4.01 11.94 0.177 11.69 7.42 1.52 3.47 0.83 99.95Rejuvenated, Upolu SAM6-9 43.83 12.10 4.14 12.45 0.173 9.95 11.23 2.02 3.08 0.65 99.63Rejuvenated, Upolu SAM6-10 42.30 11.63 4.37 13.07 0.182 10.02 12.19 1.97 3.80 1.01 100.55

Baseline samplesGalapagos DR 53-01 50.47 14.05 2.06 11.91 0.21 10.32 6.36 0.45 2.76 0.25 98.84Australian-Antarctic MW8801-16-12 48.85 16.48 1.57 9.35∗ 0.16 11.03 8.24 0.20 3.03 0.23 99.14Discordance MW8801-29-5 51.08 16.09 1.2 8.26∗ 0.14 10.67 8.45 0.17 3.28 0.15 99.49Reunion 983-18w1 48.04 13.60 2.94 13.48∗ 0.17 9.89 8.57 0.84 2.77 0.52 100.82Reunion ED1964 48.96 14.68 2.69 11.97∗ 0.17 11.86 6.74 0.81 2.64 0.34 100.9Reunion SR1977 44.61 8.46 1.48 14.06∗ 0.19 6.7 22.64 0.47 1.42 0.16 100.2Columbia River Basalt CR02-362 48.39 16.25 2.53 11.46 0.15 8.42 5.33 1.01 3.25 0.46 97.25Snake River Plain W02-3645 1.95 4.96 1.2 0.07Snake River Plain W02-3740 2.00 8.71 0.58 0.23Mauna Loa ML400-1843 51.65 13.81 2.16 11.95∗ 0.17 10.45 6.64 0.47 2.12 0.27 99.69Mauna Loa ML177-1984 51.29 13.63 2.08 12.09∗ 0.17 10.53 6.66 0.38 2.65 0.24 99.72Puna dacite Puna 74.72 0.06

Kilbourne Hole Olivine KH7-Ol 40.75 0.05 0.01 10.09 0.07 50.25 101.22Kilbourne Hole Cpx KH7-Cpx 51.64 7.37 0.49 2.93 20.19 15.45 1.73 99.80Kilbourne Hole 1 Ola KH7-1-Ol 40.75 0.05 0.01 10.09 0.07 50.25 101.22

Previously published data in italics; see Supplementary Material for sources. CR02-362 unpublished data from V. Camp and Puna unpublished data from B. Marsh. ML400-1843 values given are the average of the 2 other samples from the same 1843 flow (Rhodes and Hart, 1995).

a Single olivine used.

monitoring non-interfering 60Ni and 52Cr isotopes. Instrumental mass-dependent isotopic fractionation is corrected using the dou-ble spike (57Fe–58Fe) technique. The external reproducibility of δ56Fe measurements with this double spike technique as imple-mented at SDSU is ±0.06�(2σ). Further details of the Fe double spike technique are provided in the online supplementary mate-rial. One of the submarine Savai‘i samples was duplicated at SDSU, and five Savai‘i and Ofu samples were analyzed at the University of Chicago (UC) using chromatographic separation on AG1-X8 resin in HCl medium and measurement by standard-sample-bracketing (SSB) on a Thermo Scientific Neptune MC-ICP-MS in medium- or high-resolution wet plasma mode (Craddock and Dauphas, 2011). Dauphas et al. (2009b) have shown that the SSB method uncer-tainty is ±0.03� on δ56Fe (external 2σ reproducibility, averaging 5–9 runs per sample; averaging ∼3 runs per sample with the SDSU double spike yields the same precision; Finlayson et al., 2015). Blank levels are insignificant (6 ng Fe). All Fe isotopic composi-tions are reported in δ notation relative to the reference standard IRMM-014 (or its equivalent IRMM-524a; Craddock and Dauphas, 2011).

Pb isotope compositions for the rejuvenated lavas were deter-mined on separate sample splits, using the SDSU Nu Plasma 1700 MC-ICP-MS. Chemical separation and purification used the same method as the Samoan samples reported by Konter and Jackson(2012). MC-ICP-MS data were corrected for instrumental mass frac-tionation and machine bias, by both standard-sample-bracketing and Tl doping. Within-run fractionation was monitored using NIST SRM 997, and final ratios were corrected to bracketing standards using NIST SRM 981 of Todt et al. (1996). Blank levels were found to be <50 pg Pb.

Major element and some trace element analyses of the reju-venated lavas were analyzed in the same laboratory (Washington State University GeoAnalytical Lab) and using the same technique

(XRF) as Samoan samples published in Jackson et al. (2007) and Konter and Jackson (2012).

5. Results

Samoan shield samples studied here range from 5.1–14.7 wt.% MgO, overlapping with the Samoan rejuvenated lavas (7.4–12.2 wt.%; Table 1), where rejuvenated lavas are distinguished from shield lavas by their 206Pb/204Pb ratios below 18.9 (Table 2; Fig. 3). The baseline samples bracket this entire range (5.0–22.6 wt.%), however the majority of these samples fall within 6–8 wt.% MgO. One highly evolved sample was also studied (the Puna dacite with <0.1 wt.% MgO). Samoan samples range in LOI from −0.72 to 4.40%, where all the rejuvenated samples have negative values, and the large positive values are associated with the submarine samples. Ba and Rb abundances are quite variable, with submarine samples straying from the canonical Ba/Rb value ∼12 (Hofmann and White, 1983), indicating element mobility during surface al-teration (Ba/Rb: 6.3–12.1; Table 2).

The baseline samples define a narrow range of δ56Fe values (0.06–0.12�; Fig. 1 and 4) that overlaps with a variety of igneous rocks (Fig. 1, grey background symbols). There are two exceptions to this narrow range: First, the Puna dacite (+0.41�), which has a δ56Fe value in line with its high Si content (e.g., Fig. 1), and which plots along an extrapolated trend of Kilauea Iki samples (Teng et al., 2008; Fig. 4). Second, the single crystal and batch olivine separates and the clinopyroxene separates of the Kilbourne Hole peridotite (+0.02 to +0.07�; all within error of each other) have lower values than average igneous rocks, similar to other peridotites that have previously been studied (e.g., Williams et al., 2004; Craddock et al., 2013; Weyer and Ionov, 2007; Poitrasson et al., 2013).

The Fe isotope compositions for the Samoan samples range from +0.08 to +0.34� in δ56Fe. All rejuvenated Samoan sam-

J.G. Konter et al. / Earth and Planetary Science Letters 450 (2016) 221–232 225

Table 2Key sample parameters for assessment of alteration and Pb isotope compositions.

Location (source) Sample ID LOI Rb Ba 206Pb/204Pb 2se 207Pb/204Pb 2se 208Pb/204Pb 2se

Shield, Savai‘i D115-03 1.9 83 569 19.011 15.633 39.347Shield, Savai‘i D115-18 4.4 97 646 18.957 15.643 39.399Shield, Savai‘i D115-28 2.51 81 510 19.034 15.634 39.419Shield, Savai‘i D128-01 2 44 345 19.150 15.639 39.465Shield, Savai‘i D128-21 1.43 28 240 18.963 15.628 39.273Shield, Ofu Ofu-04-03 27 196 19.257 15.591 39.418Shield, Ofu Ofu-04-15 25 230 19.141 15.580 39.169Rejuvenated, Savai‘i SAM6-7 −0.68 25 265 18.817 1.0E−03 15.624 2.2E−03 39.096 3.5E−03Rejuvenated, Savai‘i SAM6-8 −0.42 44 389 18.826 5.0E−04 15.575 5.0E−04 38.936 1.1E−03Rejuvenated, Upolu SAM6-5 0.13 45 545 18.809 5.0E−05 15.605 6.0E−04 39.006 1.2E−03Rejuvenated, Upolu SAM6-9 −0.34 58 596 18.609 5.0E−06 15.593 5.0E−04 38.759 1.2E−04Rejuvenated, Upolu SAM6-10 −0.72 56 547 18.881 1.3E−03 15.597 1.4E−03 39.073 4.7E−03

Data in italics from literature; see Supplementary Material for sources.

Fig. 3. Pb isotope compositions of Samoan lavas. The rejuvenated lavas can easily be distinguished from shield lavas in Pb isotopes. Samoan isotopic compositions define 4 trends (Jackson et al., 2014; shown as 99% confidence regions around best-fit trends; colored streaks under grey data points) that all converge on a central high 3He/4He component (central ellipse); the rejuvenated lavas form one of these trends (yellow confidence interval). Submarine Savai‘i lavas overlap with other shield lavas (grey triangles), but not with the rejuvenated lavas (grey hour glasses). Symbols as in Fig. 1.

ples have values above +0.2�. Two submarine shield samples have δ56Fe >+0.2�, although the majority have lower δ56Fe val-ues (Table 3; Fig. 4). The values found for the rejuvenated sam-ples are significantly higher than lavas with similar SiO2 or MgO from a global compilation (Fig. 1 and 4). For example, δ56Fe val-ues are significantly higher at a given MgO content than basalts (and their derivatives) of the Kilauea Iki lava lake in Hawai‘i, prob-ably the closest analog to the compositional range of Samoan lavas for which δ56Fe values are available. In order to verify that these high values were not an analytical artifact, an olivine phenocryst was analyzed and its δ56Fe value was similarly high (+0.34 vs. +0.30� for the whole rock). One of the Samoan samples with a higher δ56Fe value (115-28) was replicated at SDSU (0.25 vs. 0.28�), and further duplicates of two samples were measured in-dependently (115-28: 0.30(9)� at UC vs. 0.25–0.28� at SDSU; 115-03: 0.18(8)� at UC vs. 0.11� at SDSU). Data quality can also be assessed with the results for the reference materials; differ-ences between our results and recommended reference values (e.g. Craddock and Dauphas, 2011), and differences between duplicated sample analyses are all within external errors (0.06 and 0.03�;

Table 3 and Supplementary Table S3). A clinopyroxene separate from one of these samples (115-28) gave a lower value (+0.15vs. 0.25–0.31� for the whole rock), potentially reflecting the state of alteration of the sample, as described below.

6. Discussion

There are several possible explanations to consider for the ele-vated δ56Fe values in Samoan lavas, including an anomalous man-tle source composition, isotopic fractionation associated with mag-matic processes, or overprinting of the original igneous composi-tion by surface alteration. Below, the potential for alteration of the primary isotopic signatures is first assessed. Thereafter, magmatic processes, specifically the processes of partial mantle melting and fractional crystallization, are examined as an explanation for the measured Fe isotopic compositions. Finally, we assess if unusual mantle source compositions can explain the high Fe isotopic val-ues in Samoan lavas.

6.1. Has alteration affected the Fe isotopic compositions of Samoan volcanic samples?

Altered oceanic crust can range from δ56Fe = −0.37 to +0.55�(Rouxel et al., 2003), where elevated values are ascribed to leach-ing of Fe2+ with low δ56Fe values. All Samoan samples in this study fall within this range, so it is essential to filter out altered samples. This is particularly important for the rejuvenated lavas that all have unusually high δ56Fe (>+0.2�). The Fe isotopic compositions of the rejuvenated lavas plot above the trends de-fined by the majority of basaltic rocks (Fig. 4). However, their low LOI and their fresh appearance due to their recent eruption rule out significant alteration. The δ56Fe values of these lavas therefore faithfully preserve their original igneous composition.

The two submarine shield lavas (115-18 and 115-28) with high δ56Fe values have the highest LOI (>2.5%) and the lowest Ba/Rb ra-tios (<6.7). This is likely due to an increase in Rb during seafloor alteration (e.g., Koppers et al., 2003) and elevated δ56Fe values from leaching of isotopically light Fe2+ (Rouxel et al., 2003). The clinopyroxene separate from sample 115-28 also yields a much lower δ56Fe value (+0.15�; Table 3) than the whole rock, which is consistent with the fact that pyroxene is resistant to alteration (e.g., Koppers et al., 2003) and its Fe isotopic composition may re-flect the original igneous Fe isotope composition (the rest of the matrix being altered). Given the altered state indicated by LOI and Ba/Rb, the elevated δ56Fe values of the whole rock likely do not reflect pristine igneous signatures, and these two values are not considered further. The remaining Samoan shield stage lavas do not have elevated δ56Fe values for their SiO2 compositions (Fig. 1) and follow the trend defined by the majority of published val-ues for basaltic rocks. Thus, the common Fe isotope compositions

226 J.G. Konter et al. / Earth and Planetary Science Letters 450 (2016) 221–232

Fig. 4. δ56Fe versus MgO of various igneous suites. A) The majority of igneous samples cluster around 0.09� (black horizontal line), although two types of trends are indicated with arrows; one group quickly increasing in δ56Fe with decreasing MgO, another group that increases in δ56Fe at very low MgO. B) Anomalously high values in Samoa compared to literature data and our baseline samples. In black lines, fractional crystallization models show that even at the most extreme fractionation (�δ56Feolivine-meltof −0.3), common primitive mantle compositions do not reach rejuvenated Samoan compositions. Altered samples and duplicated samples indicated.

of the other shield lavas are likely pristine, whereas the elevated δ56Fe values are most prominent in the rejuvenated lavas.

6.2. Extent of Fe isotopic fractionation due to magmatic processes

6.2.1. Fractional crystallizationThe δ56Fe values of co-genetic suites of volcanic rocks are

thought to be partially controlled by fractionation of Fe-bearing minerals from an evolving magma and should therefore correlate with indices of fractional crystallization. For example, the two ma-jor crystallizing phenocryst phases containing Fe are olivine and clinopyroxene and correlations between δ56Fe and MgO (±CaO) content could indicate Fe isotope fractionation by these crystal-lizing phases (e.g., Teng et al., 2013). The specifics of the trend in δ56Fe during crystallization will depend on the nature and relative redox state of the melt and crystallizing phases (e.g. Schuessler et al., 2009; Dauphas et al., 2014) as well as whether the system is open or closed condition to oxygen exchange with surrounding rocks (Sossi et al., 2012). Multiple natural systems show varia-tions in δ56Fe with SiO2 and MgO. For example, trends for Kilauea Iki (Hawai‘i) and Red Hill (Australia) show increasing δ56Fe val-ues, starting from basalts with ∼47% SiO2 and ∼ 0.1�δ56Fe val-ues to andesitic compositions with ∼57% SiO2 and ∼ 0.2�δ56Fe(Sossi et al., 2012; Teng et al., 2013). In contrast, suites from Hekla (Iceland) and Cedar Butte (USA) remain below ∼0.10� in δ56Fe until ∼70% SiO2 values are reached (Schuessler et al., 2009;Zambardi et al., 2014). The elevated δ56Fe values in Kilauea Iki and Red Hill samples have been explained with fractional crystal-lization of Fe3+-poor phases (olivine, pyroxene under oxygen-poor conditions) in a system closed to oxygen exchange (Sossi et al., 2012), and this explanation provides a causal relationship between Fe isotope and major element trends.

The fractionation of Fe isotopes in the Samoan rejuvenated lavas can be approximated with a Rayleigh fractionation model (Fig. 4, black curves), similar to the model applied to other ocean island basalts by Teng et al. (2008; 2013) in order to simulate crys-tallization under closed (for O2) conditions. For the model, a prim-itive melt major element composition was constructed by equilib-rium olivine addition to a representative sample (SAM6-8) to reach 18 wt.% MgO (assumed to represent a primitive melt, since this

corresponds to equilibrium olivine with Fo = 91, close to esti-mates for many hotspots; Putirka, 2008). Equilibrium olivine was removed stepwise from this parental melt until the residual melt had ∼0.2 wt.% MgO, using MgO/FeO partitioning between olivine and melt from Roeder and Emslie (1970). In order to model melt evolution due to fractional crystallization, a relationship between the calculated melt evolution from stepwise olivine removal and melt fraction during fractional crystallization is needed. This was accomplished by tracking the calculated K2O content to determine the related remaining melt fraction during fractional crystalliza-tion, assuming DK2O = 0.04 (following Teng et al., 2008, 2013). These melt fractions are then linked to the Rayleigh distillation model (similar to Teng et al., 2008, 2013), using a range of frac-tionation factors (�δ56Feolivine-melt: 0, −0.1, −0.2, −0.3). Assuming a δ56Fe of 0.09� in the primary melt envelops most existing Fe isotope data between fractionation factors of 0 and −0.2 (Fig. 4, black curves, −0.2 is dashed). However, even the most extreme fractionation factor (−0.3) cannot explain the elevated rejuvenated Samoan values (Fig. 4). In fact, a primary melt with a δ56Fe value of 0.20–0.25� at 18 wt.% MgO is needed to explain the mea-sured δ56Fe up to +0.34�. The discrepancy of at least 0.11�between this primary melt and the common δ56Fe melt composi-tion of 0.09� relies on the simplistic isotope fractionation model above, and warrants closer examination.

Recent work has suggested that isotopic fractionation of Fe2+between olivine and melt is minimal, and instead that the change in melt redox conditions due to Fe2+ removal (concentrating Fe3+with elevated δ56Fe in the melt) is most important (Sossi et al., 2012; Dauphas et al., 2014). Furthermore, Mg and Fe in-situ iso-tope profiles across olivine crystals are consistent with diffusion and not crystal growth (Sio et al., 2013), suggesting kinetic Fe iso-tope fractionation may be more significant than equilibrium Fe isotope fractionation during olivine removal. Thus, multiple fac-tors may affect Fe isotope fractionation during crystallization, yet a simple equilibrium fractionation model that assumes crystalliza-tion under closed conditions for well-studied systems such as Ki-lauea Iki and Red Hill provides a reasonable fit to the data (Fig. 4). Such a first order “effective fractionation” model can also explain

J.G. Konter et al. / Earth and Planetary Science Letters 450 (2016) 221–232 227

Table 3Sample Fe isotope compositions.

Location (source) Sample ID δ56Fe 2sea δ57Fe 2se Laboratory

Samoan volcanic samplesShield, Savai‘i D115-03 0.11 0.03 SDSUShield, Savai‘i duplicate 0.18(8) 0.028 0.28(3) 0.050 UCShield, Savai‘i D115-18 0.24 0.04 SDSUShield, Savai‘i D115-28 0.28 0.03 SDSUShield, Savai‘i replicate 0.25 0.03 SDSUShield, Savai‘i duplicate 0.30(9) 0.028 0.45(0) 0.050 UCShield, Savai‘i D115-28 (Cpx) 0.15(0) 0.029 0.22(6) 0.052 UCShield, Savai‘i D128-01 0.08 0.03 SDSUShield, Savai‘i D128-21 0.11 0.03 SDSUShield, Ofu ofu-04-03 0.10(5) 0.029 0.18(0) 0.052 UCShield, Ofu ofu-04-15 0.16(7) 0.029 0.23(8) 0.052 UCRejuvenated, Savai‘i SAM6-7 0.31 0.04 SDSURejuvenated, Savai‘i SAM6-8 0.30 0.04 SDSURejuvenated, Savai‘i SAM6-8 (Ol) 0.34 0.05 SDSURejuvenated, Upolu SAM6-5 0.33 0.04 SDSURejuvenated, Upolu SAM6-9 0.21 0.04 SDSURejuvenated, Upolu SAM6-10 0.23 0.05 SDSU

Baseline samplesGalapagos DR 53-01 0.07 0.04 SDSUAustralian-Antarctic MW8801-16-12 0.12 0.03 SDSUDiscordance MW8801-29-5 0.09 0.02 SDSUReunion 983-18w1 0.08 0.04 SDSUReunion ED1964 0.12 0.03 SDSUReunion SR1977 0.09 0.03 SDSUColumbia River Basalt CR02-362 0.10 0.03 SDSUSnake River Plain W02-3645 0.06 0.03 SDSUSnake River Plain W02-3740 0.06 0.03 SDSUMauna Loa ML400-1843 0.06 0.03 SDSUMauna Loa ML177-1984 0.06 0.03 SDSUPuna dacite Puna 0.41 0.06 SDSU

Kilbourne Hole olivine KH7 (Ol) 0.06 0.04 SDSUKilbourne Hole Cpx KH7 (Cpx) 0.07 0.03 SDSUKilbourne Hole 1 Olb KH7-1 (Ol) 0.02 0.04 SDSU

Reference Materials δ56Fe 2sea δ57Fe 2se Recommended δ56Fec

SDSU resultsBasalt, Br. Columb. (N = 4) BCR-1 0.11 0.06 0.091 ± 0.011 (BCR2)Basalt, Iceland (N = 4) BIR-1 0.09 0.04 0.053 ± 0.015 (BIR1a)Diabase, Virginia (N = 11) W-2 0.05 0.08 0.056 ± 0.013Basalt, Hawai‘i (N = 7) BHVO-1 0.07 0.04 0.105 ± 0.008

UC resultsBasalt, Hawai‘i BHVO-2 0.10(8) 0.029 0.17(0) 0.058 0.114 ± 0.011Iron Formation, Isua IF-G 0.63(8) 0.029 0.96(8) 0.058 0.639 ± 0.013

Replicate is reanalysis of the same purified Fe fraction, while duplicate is analysis on a separately dissolved and purified Fe fraction.a Uncertainties for SDSU analyses are 2se internal errors from single analysis, except the reference materials, which show 2σ of multiple (N) runs. Uncertainties for UC

analysis are 2se external errors from average of nine non-consecutive analyses.b Single olivine crystal analysis.c Reference material recommended values from Craddock and Dauphas (2011).

the δ56Fe values of Samoan shield lavas, but additional processes are needed for the elevated values in the rejuvenated lavas.

6.2.2. Partial meltingThe modeled crystal fractionation trends only pass through the

Samoan δ56Fe values if the starting δ56Fe composition for the frac-tionating primary melt is between +0.2 to 0.25� (Section 6.2.1). We modeled the maximum fractionation in Samoan samples that can be attributed to partial melting following Dauphas et al.(2009a), finding that unrealistic parameters are needed to ex-plain δ56Fe values >0.3�. Specifically, such elevated δ56Fe values can only be achieved with unreasonably small degrees of melting (F � 0.01), and high Fe3+/Fe2+ ratios (>0.2) in the source.

The melting model of Dauphas et al. (2009a) quantifies frac-tionation relative to a melt fraction of “0”, which is subsequently modified at greater melt fractions through either batch or frac-tional melting. The expected range in rejuvenated lavas is confined to only a very small melt fraction. For example, Hawaiian rejuve-nated lavas range from 0.0002 up to 0.026 melt fraction (Garcia et al., 2010). Thus, the maximum isotopic fractionation, produced by

either model at nearly “0” melt fraction is appropriate here. Us-ing the Fe isotope fractionation factor between Fe3+ versus Fe2+in melt from Dauphas et al. (2009a; �δ56FeFe3+–Fe2+ = 0.3�), the maximum fractionation between melt and source calculated here is approximately +0.16� (Fig. 5). If we assume a mantle source composition of 0 to 0.02� as a starting point (similar to Weyer and Ionov, 2007; Dauphas et al., 2009a), partial melting cannot generate the primitive melt composition that is required to ex-plain the Samoan rejuvenated lavas with a combination of partial melting, followed by the effective fractionation during crystalliza-tion (a maximum of 0.18� is an upper limit during melting, whereas 0.20–0.25� is needed). Moreover, a “0” melt fraction par-tial melt would not segregate from its mantle source, implying the achievable δ56Fe of a primitive melt is likely slightly lower, at ∼+0.14–0.15�. These values assume a ∼0.01 melt fraction, and since Fe3+/Fe2+ is not well-constrained in Samoa, we assume a mantle source that lies between the values for samples from Ki-lauea Iki, Hawaii (Teng et al., 2008) and arcs (Kelley and Cottrell, 2009); Fig. 5). Thus, starting with average mantle (0–0.02�), melt-

228 J.G. Konter et al. / Earth and Planetary Science Letters 450 (2016) 221–232

Fig. 5. Modeled Fe isotope compositions for various degrees of melting, at differ-ent Fe3+/Fe2+ ratios. The maximum isotope fractionation is achieved at near-zero degree melting, and at high Fe3+/Fe2+ ratios.

ing under these conditions (0.14–0.15�) produces primary melts that are still ∼0.03–0.09� lower in δ56Fe than the compositions required by the fractional crystallization model (+0.20–0.25�). Similarly, Williams and Bizimis (2014) found melting cannot ex-plain the δ56Fe difference (>0.2�) between melts and residues in Hawai‘i. If the recent force constants of Fe for olivine and melt studied by NRIXS (Dauphas et al., 2014) are considered, equilib-rium fractionation during melting may be less significant, increas-ing the discrepancy between the primitive melt and the starting point for crystal fractionation even further.

6.2.3. Other magmatic processes?Three other processes have been proposed to affect the Fe iso-

tope composition of melts, but two of them are limited to highly silicic systems. For example, fluid exsolution was proposed to ex-plain enriched δ56Fe values in granitic materials (e.g., Poitrasson and Freydier, 2005; Heimann et al., 2008). However, the Samoan lavas are substantially more mafic (i.e., 42 to 52 wt.% SiO2; Ta-ble 1). Secondly, isotope fractionation under a thermal gradient (Soret diffusion) may fractionate stable isotopes in igneous sys-tems (e.g., Richter et al., 2009; Huang et al., 2009; Zambardi et al., 2014), although this was proposed for systems much more silicic than relevant for Samoan magmatism (e.g. Zambardi et al., 2014). Lastly, xenocrysts might affect the melt composition. In this sce-nario, an increase in δ56Fe values would require assimilation of a mineral with elevated δ56Fe, like magnetite, which would be obvi-ous in MgO and Al2O3 characteristics (not observed). Alternatively, diffusional exchange of Fe with xenocrystic olivine would prefer-entially remove 54Fe from the melt, compared to 56Fe (Dauphas et al., 2010; Sio et al., 2013). This requires Fe diffusion into a signifi-cant mass fraction of primitive olivines with elevated δ56Fe values (and low Fe/Mg), while primitive olivines tend to have low δ56Fe values.

6.3. Unusual δ56Fe composition of the Samoan source

6.3.1. Negligible contribution from a recycled mantle componentNone of the predominant magmatic processes can explain the

elevated δ56Fe compositions of the Samoan rejuvenated lavas, ei-ther as a single process, or as a combination. Instead, the observed Fe isotope compositions may contain a contribution from an un-usual source composition, such as the recycled continental sedi-ment component in the Samoan mantle source (e.g., Jackson et al., 2007). Continental igneous rocks can have elevated δ56Fe values up

to ∼0.4� (e.g., Heimann et al., 2008). However, isotopic mass bal-ance argues against an explanation invoking a recycled continental component because δ56Fe values of continental igneous rocks are similar to Samoan rejuvenated lavas. More importantly, the Fe con-tent of subducted continental sediment is less than that of typical mantle peridotite. The fraction of recycled continental sediment re-quired in the mantle source to satisfy Fe isotopic data is an order of magnitude larger than allowed based upon radiogenic isotope compositions (e.g., Sr) of the Samoan lavas (e.g., Jackson et al., 2007). Moreover, compositional evidence for recycled continental sediments is strongest in the submarine shield lavas studied here (Jackson et al., 2007), yet these lavas show near “average” basaltic compositions (∼+0.09�). In contrast, elevated Fe isotopic com-positions are mainly observed in the rejuvenated lavas, which do not feature the extreme radiogenic isotope compositions thought to derive from recycled continental sediments. Thus, other aspects of the source, such as metasomatism (e.g., Hauri et al., 1993) or lithology (e.g. Prytulak and Elliot, 2007) should be assessed.

6.3.2. Potential contribution from mantle metasomatismMantle metasomatism is an alternative process that can elevate

the δ56Fe mantle source values (e.g. Weyer and Ionov, 2007) of the Samoan rejuvenated lavas. This process may explain trace and volatile element enrichments in rejuvenated lavas from Hawai‘i (Ni‘ihau; Dixon et al., 2008). Petrography and elemental abun-dances of peridotites from Savai‘i (Samoa) suggest metasomatic overprinting (e.g., Hauri et al., 1993), which may also explain ra-diogenic isotope compositions of rejuvenated Samoan lavas (Konter and Jackson, 2012). The mineralogical result of this metasomatism is the creation of secondary clinopyroxene in Samoan xenoliths (Hauri et al., 1993). The specific mineral reactions and the final products of metasomatism largely determine the resulting Fe iso-tope compositions (Poitrasson et al., 2013). For the specific case of clinopyroxene creation, it is thought that this results in ele-vated δ56Fe values (Weyer and Ionov, 2007; Poitrasson et al., 2013;Williams and Bizimis, 2014). The metasomatized peridotite can then function as the starting material for melting, but at an ele-vated δ56Fe value compared to average mantle, such that partial melting can then generate the expected primary melt δ56Fe com-positions.

An indication that elevated mantle values are indeed plausible comes from the slightly elevated δ56Fe values in peridotite xeno-liths from Savai‘i (+0.07�; Finlayson et al., 2015), compared to average peridotite mantle (0 to +0.02�; Weyer and Ionov, 2007;Dauphas et al., 2009a; Craddock et al., 2013). The difference in δ56Fe value between xenolith values from Savai‘i (+0.07�) com-pared to average peridotite (0 to +0.02�) is very similar to the component of Fe isotope fractionation in Samoan rejuvenated lavas that cannot be explained with partial melting and fractional crys-tallization (0.03–0.09�). Thus, such elevated values could serve as a starting point for partial melting and fractional crystallization to explain the Samoan rejuvenated compositions.

Metasomatism may also affect the Fe3+/Fe2+ ratio of the man-tle source, as suggested by the fact that metasomatized mantle xenoliths are generally oxidized (high f O2; e.g., Mattioli et al., 1989), and redox state affects the Fe isotope composition of partial melts (Section 6.2.2). Although no precise published measurements exist for Samoa, recent XANES measurements (described and pre-sented by Cottrell and Kelley, 2014) suggest more oxidized con-ditions in Samoan lavas (Fe3+/Fe2+ = 0.19–0.26) than Kilauea Iki (Fig. 5). This would increase the degree of Fe isotopic fraction-ation during partial mantle melting and close the gap between source composition and primary melt (e.g., Williams et al., 2004;Weyer and Ionov, 2007; Dauphas et al., 2009a; Craddock et al., 2013).

J.G. Konter et al. / Earth and Planetary Science Letters 450 (2016) 221–232 229

6.3.3. A contribution from source lithologyPyroxenites represent a potential source lithology for hotspot

volcanoes that may impact the melting behavior (deeper and to a higher degree than a peridotite source) and the chemistry of the lavas. Moreover, recent Fe isotopic data show that mantle pyroxen-ites can have elevated δ56Fe values compared to those of man-tle peridotites (+0.18�; Williams and Bizimis, 2014; +0.16�; Finlayson et al., 2015). The first of three common types of pyrox-enites (Bodinier and Godard, 2014) – the vein-type pyroxenites – have elevated δ56Fe values ascribed to a cumulate origin by frac-tional crystallization in the mantle, based on the correlation of pyroxenite Fe isotopes with TiO2 and heavy rare earth element abundances (Williams and Bizimis, 2014). Although Fe isotope frac-tionation factors measured by Dauphas et al. (2014) suggest a minimal effect for crystal fractionation of (ortho)pyroxene, it is not clear if these factors can be extrapolated to cumulate–pyroxenite formation under mantle conditions. The second pyroxenite type – pyroxenites derived from subduction-recycled oceanic crust – may also carry elevated δ56Fe values due to devolatilization reactions removing fluids with low δ56Fe values from the subducting slab, as suggested by data from serpentinites (Debret et al., 2016). The third type of pyroxenite – replacive pyroxenites – effectively rep-resent melt–rock reactions that produce pyroxene, which should generate an elevated δ56Fe value like cumulate or metasomati-cally produced pyroxene. Regardless of the origin of the pyroxenite, the existence of pyroxenites with elevated δ56Fe values motivates evaluation of their potential role in generation of the Samoan re-juvenated lavas. Evidence that such a pyroxenite contribution is important comes from several major and minor element abun-dances and their ratios (discussed below). Elements such as Ca, Mg, Fe, Ti, and Zn partition differently between melt and olivine versus melt and clinopyroxene, and can thereby identify pyroxen-ite sources, although the exact pyroxenite origin is not specifically defined.

Suggestions for the presence of pyroxenite in the mantle source for Samoan lavas come from the CaO–MgO relationship, high TiO2content and Ti anomalies, and transition element ratios (particu-larly Zn/Fe). The CaO–MgO relationship for shield and rejuvenated volcanism in Samoa suggests mainly a crystal fractionation trend for the shield lavas that includes clinopyroxene and/or plagio-clase fractionation, reducing CaO to near-zero at low MgO values (Fig. 6). Although Samoan shield stage lavas plot in the field for peridotite melts that evolved under olivine fractionation and later Ca-bearing phases, the rejuvenated lavas plot at low CaO values that can indicate pyroxenite melting (based on experimental data; Herzberg and Asimow, 2008). Secondly, the high TiO2 content of hotspot lavas, after correction for mineral fractionation, has been interpreted as an indicator for a pyroxenite source lithology, and Samoan lavas are unusually enriched in TiO2 (Prytulak and El-liott, 2007; online supplementary Fig. S5) and have a positive TiO2anomaly compared to elements of similar compatibility (Jackson et al., 2008). In the rejuvenated lavas, this anomaly extends be-yond the shield lavas by ∼30%, to the highest values that are found in hotspots (Prytulak and Elliott, 2007; Konter and Jackson, 2012;Jackson et al., 2014; online supplementary Fig. S5). Thirdly, a high Zn/Fe (×104) ratio is thought to indicate pyroxenite in the mantle source, since peridotite partial melting generally cannot produce Zn/Fe (×104) ratios >13 (Le Roux et al., 2010; Davis et al., 2013). Consequently, there are three independent geochemical indicators for pyroxenite in the source for the rejuvenated Samoan lavas. We note that the three indicators used here (Prytulak and Elliott, 2007;Herzberg and Asimow, 2008; Le Roux et al., 2010) all assume a dif-ferent pyroxenite type and origin. None of these methods specify a clear way to distinguish between different types of pyroxenite. Thus, the presence of pyroxenite with elevated δ56Fe values in the

Fig. 6. MgO versus CaO of Samoan shield and rejuvenated lavas. Rejuvenated lavas plot below the pyroxenite-peridotite source line (Herzberg and Asimow, 2008), al-though it is possible to acquire these compositions with clinopyroxene fractionation at mantle depths.

source of Samoan rejuvenated lavas seems supported, even if its origin cannot be uniquely constrained.

An important unknown in this scenario is the relative con-tribution of pyroxenite versus peridotite in the source. Peridotite melts are thought to be produced alongside pyroxenite melts (e.g., Prytulak and Elliott, 2007), and thus, melting of the peridotite will likely dilute the heavy Fe isotopic signature from the pyroxen-ite (i.e., bring the melt composition closer to the peridotite aver-age Fe isotope composition). However, Bizimis and Peslier (2015)have suggested pyroxenites in Hawai‘i with elevated δ56Fe values (Williams and Bizimis, 2014) may melt to nearly 50% before peri-dotite melting would occur. Moreover, Prytulak and Elliott (2007)suggested that melting of carbonated garnet-pyroxenite could ex-plain high TiO2 lavas, because melting experiments by Dasgupta et al. (2006) using such material produced a carbonatitic liquid and a Ti-enriched alkalic liquid without causing surrounding peridotite to melt. Prytulak and Elliott (2007) acknowledged that the melts produced would likely react with peridotite, but they implied that carbonated pyroxenite melting should increase TiO2 in the lavas. By extension, it may also preserve an elevated Fe isotopic com-position while also providing a natural link to the metasomatism described above, and therefore the carbonatite metasomatism de-scribed for Samoa by Hauri et al. (1993).

Relative peridotite and pyroxenite proportions have been mod-eled in dynamic hotspot models, where pyroxenite may exist as an inherent component of an upwelling mantle plume, and it may contribute slightly more to the melt during the rejuvenated stage (>80% for Hawai‘i, e.g., Ballmer et al., 2013). However, estimates for pyroxenite contributions to rejuvenated lavas are strongly de-pendent on the exact solidus and melt productivity (Ballmer et al., 2013), and therefore may or may not be greater than during the shield stage. Further work is needed to establish whether the Fe isotope signatures are related to a pyroxenite plume compo-nent, and where in the plume this component might be located. A speculative alternative model could explain the pyroxenites as part of the upper mantle that had previously been metasoma-tized over the Rarotonga hotspot (Konter and Jackson, 2012). In this scenario, the pyroxenite represents veins that result from crys-tal segregation from a mafic melt as it travels toward the surface (e.g., Bodinier and Godard, 2014). This material could then prefer-entially melt due to its lower solidus compared to peridotite (e.g.,

230 J.G. Konter et al. / Earth and Planetary Science Letters 450 (2016) 221–232

Bizimis and Peslier, 2015), during the flexure-derived melting pro-posed by Konter and Jackson (2012). Thus, several feasible models exist that explain the observed Fe isotope data for Samoa, and these are in line with existing models for the origin of rejuvenated lavas in Samoa.

7. Summary

Samoan lavas exhibit a large range of Fe isotopic composi-tions compared to those observed in any other oceanic basalts. In particular, Samoan rejuvenated lavas have elevated δ56Fe (up to +0.34�) given their MgO contents. In contrast, shield stage lavas representing extreme mantle source compositions (high 87Sr/86Sr and 3He/4He) show similar δ56Fe values to other hotspot lavas (e.g. Hawai‘i, Societies; Teng et al., 2008, 2013; Fig. 4). Therefore, the range in δ56Fe values between the various mantle end-member components defined to date is small compared to the range in Samoan lavas, implying processes other than those that generate radiogenic isotope anomalies should be considered to explain δ56Fe variability in hotspot lavas. This is particularly necessary to explain the elevated values in Samoan rejuvenated lavas.

Modeling suggests that a combination of processes is required to explain the Fe isotope ratios in rejuvenated Samoan lavas. Frac-tional crystallization and partial melting can only explain part of the elevated δ56Fe values. An additional source contribution is needed, either from metasomatism of the source peridotite, or from melting a pyroxenite source (Fig. 7).

Samoan metasomatized peridotites have δ56Fe values up to +0.07� (Finlayson et al., 2015), and partial melting of metaso-matized mantle might elevate the δ56Fe value by a maximum ∼0.14�. Subsequent extreme fractionation (crystallization with �δ56Feolivine–melt = −0.3) is needed to produce melts with δ56Fe values similar to the highest values in Samoa. This sequence of fractionating processes starting with metasomatism agrees with the idea that Samoan rejuvenated lavas are produced by flexural-based melting of upper mantle that had previously been metaso-matized over the Rarotonga hotspot (Konter and Jackson, 2012).

Alternatively, the presence of pyroxenite source material (e.g., Hirschmann et al., 2003; Dasgupta et al., 2006) is consistent with the observed silica-undersaturated nature and the CaO–MgO, Zn/Fe and Ti characteristics of Samoan rejuvenated lavas. A pyroxenite contribution to the melt Fe isotope composition could be impor-tant because pyroxenites: 1) generally have elevated δ56Fe (up to +0.15�; e.g., Williams and Bizimis, 2014), and 2) have a lower solidus. Since Dauphas et al. (2009a, 2014) have argued that Fe isotope fractionation during melting is more sensitive to redox conditions than mineralogy during melting, we might expect the same isotope fractionation trends (Fig. 7) for both peridotite and pyroxenite melting, assuming the same redox conditions. As a re-sult, the primary melt from mantle melting involving pyroxenite, is slightly elevated in δ56Fe compared to the peridotite-only melt, and after crystal fractionation, may produce melts that very closely overlap with observed Samoan rejuvenated compositions (Fig. 7).

Overall, a combination of processes appears to be necessary to explain the unique Fe isotope signatures in Samoa, and it appears that unusual Fe isotope signatures may provide a gauge for unusual mantle source materials for igneous rocks.

Acknowledgements

We thank the following colleagues for sharing major element data and samples: Bruce Marsh for the Puna drill core melt (“dacite”), Vic Camp for the Columbia River Basalt, Francis Albarède for the Réunion basalt, Elizabeth Anthony for the Kilbourne Hole peridotite, and Bernard Marty for the Society Islands elemental data. We thank Michael Bizimis and Helen Williams for construc-tive reviews, and Bernard Marty for editorial handling.

Fig. 7. Potential generation of elevated Fe isotope values in Samoa through source and magmatic processes. (A) In red, peridotite mantle is metasomatized, after which small degree partial melting under high Fe3+/Fe2+ conditions generates primary melts with elevated δ56Fe. Fractional crystallization is then used to model the Fe isotope fractionation observed in Samoan rejuvenated lavas. In blue, pyroxen-ite involvement is shown: pyroxenite source material has elevated δ56Fe. Melt-ing of peridotite–pyroxenite mantle source therefore may produce primary melts that are elevated compared to peridotite-only melts. (B) Fractional crystallization is used to model a range of compositional trends, varying the primary melt com-position and the fractionation factor between melt and crystals. Peridotite-only sourced melts in red, peridotite–pyroxenite sourced melts in blue. Values refer to �δ56Feolivine–melt (difference in δ56Fe between melt and olivine), similar to Teng et al. (2008). The most extreme fractionation model for peridotite-sourced melts overlaps with the error bars of the samples, however all samples plot above the modeled curves. A source with pyroxenite relaxes the extreme fractional crystal-lization model needed. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Appendix A. Supplementary material

Supplementary material related to this article can be found on-line at http://dx.doi.org/10.1016/j.epsl.2016.06.029.

References

Ballmer, M.D., Ito, G., Wolfe, C.J., Solomon, S.C., 2013. Double layering of a thermo-chemical plume in the upper mantle beneath Hawaii. Earth Planet. Sci. Lett. 376, 155–164. http://dx.doi.org/10.1016/j.epsl.2013.06.022.

Bizimis, M., Peslier, A.H., 2015. Water in Hawaiian garnet pyroxenites: implica-tions for water heterogeneity in the mantle. Chem. Geol. 397, 61–75. http://dx.doi.org/10.1016/j.chemgeo.2015.01.008.

Bodinier, J.L., Godard, M., 2014. Orogenic, Ophiolitic, and Abyssal Peridotites, Treatise on Geochemistry. Elsevier.

J.G. Konter et al. / Earth and Planetary Science Letters 450 (2016) 221–232 231

Cottrell, E., Kelley, K.A., 2014. Redox heterogeneity of the mantle inferred from hotspots. In: AGU Fall Meeting. San Francisco. V23G-02.

Craddock, P.R., Dauphas, N., 2011. Iron isotopic compositions of geological refer-ence materials and chondrites. Geostand. Geoanal. Res. 35, 101–123. http://dx.doi.org/10.1111/j.1751-908X.2010.00085.x.

Craddock, P.R., Warren, J.M., Dauphas, N., 2013. Abyssal peridotites reveal the near-chondritic Fe isotopic composition of the Earth. Earth Planet. Sci. Lett. 365, 63–76. http://dx.doi.org/10.1016/j.epsl.2013.01.011.

Dasgupta, R., Hirschmann, M.M., Stalker, K., 2006. Immiscible transition from carbonate-rich to silicate-rich melts in the 3 GPa melting interval of eclog-ite + CO2 and genesis of silica-undersaturated ocean island lavas. J. Petrol. 47, 647–671. http://dx.doi.org/10.1093/petrology/egi088.

Dauphas, N., Craddock, P.R., Asimow, P.D., Bennett, V.C., Nutman, A.P., Ohnenstet-ter, D., 2009a. Iron isotopes may reveal the redox conditions of mantle melting from Archean to Present. Earth Planet. Sci. Lett. 288, 255–267. http://dx.doi.org/10.1016/j.epsl.2009.09.029.

Dauphas, N., Pourmand, A., Teng, F.-Z., 2009b. Routine isotopic analysis of iron by HR-MC-ICPMS: how precise and how accurate? Chem. Geol. 267, 175–184. http://dx.doi.org/10.1016/j.chemgeo.2008.12.011.

Dauphas, N., Teng, F.-Z., Arndt, N.T., 2010. Magnesium and iron isotopes in 2.7 Ga Alexo komatiites: mantle signatures, no evidence for Soret diffusion, and iden-tification of diffusive transport in zoned olivine. Geochim. Cosmochim. Acta 74, 3274–3291. http://dx.doi.org/10.1016/j.gca.2010.02.031.

Dauphas, N., Roskosz, M., Alp, E.E., Neuville, D.R., Hu, M.Y., Sio, C.K., Tissot, F.L.H., Zhao, J., Tissandier, L., Médard, E., Cordier, C., 2014. Magma redox and structural controls on iron isotope variations in Earth’s mantle and crust. Earth Planet. Sci. Lett. 398, 127–140. http://dx.doi.org/10.1016/j.epsl.2014.04.033.

Davis, F.A., Humayun, M., Hirschmann, M.M., 2013. Experimentally determined mineral/melt partitioning of first-row transition elements (FRTE) during par-tial melting of peridotite at 3 GPa. Geochim. Cosmochim. Acta 104, 232–260. http://dx.doi.org/10.1016/j.gca.2012.11.009.

Debret, B., Millet, M.A., Pons, M.L., Bouilhol, P., Inglis, E., Williams, H., 2016. Isotopic evidence for iron mobility during subduction. Geology 44, 215–218. http://dx.doi.org/10.1130/G37565.1.

Finlayson, V.A., Konter, J.G., Ma, L., 2015. The importance of a Ni correction with ion counter in the double spike analysis of Fe isotope composition using a 57Fe/58Fe double spike. Geochem. Geophys. Geosyst. 2015GC006012.

Garcia, M.O., Swinnard, L., Weis, D., Greene, A.R., Tagami, T., Sano, H., Gandy, C.E., 2010. Petrology, Geochemistry and Geochronology of Kaua’i Lavas over 4·5 Myr: implications for the Origin of Rejuvenated Volcanism and the Evolu-tion of the Hawaiian Plume. J. Petrol. 51, 1507–1540. http://dx.doi.org/10.1093/petrology/egq027.

Hart, S.R., Coetzee, M., Workman, R.K., Blusztajn, J., Johnson, K.T.M., Sinton, J.M., Steinberger, B., Hawkins, J.W., 2004. Genesis of the Western Samoa seamount province: age, geochemical fingerprint and tectonics. Earth Planet. Sci. Lett. 227, 37–56. http://dx.doi.org/10.1016/j.epsl.2004.08.005.

Hauri, E.H., Shimizu, N., Dieu, J.J., Hart, S.R., 1993. Evidence for hotspot-related carbonatite metasomatism in the oceanic upper mantle. Nature 365, 221–227. http://dx.doi.org/10.1038/365221a0.

Heimann, A., Beard, B.L., Johnson, C.M., 2008. The role of volatile exsolution and sub-solidus fluid/rock interactions in producing high 56Fe/54Fe ratios in siliceous igneous rocks. Geochim. Cosmochim. Acta 72, 4379–4396. http://dx.doi.org/10.1016/j.gca.2008.06.009.

Herzberg, C., Asimow, P.D., 2008. Petrology of some oceanic island basalts: PRIMELT2. XLS software for primary magma calculation. Geochem. Geophys. Geosyst. 9, 1–25. http://dx.doi.org/10.1029/2008GC002057.

Hirschmann, M.M., Kogiso, T., Baker, M.B., Stolper, E.M., 2003. Alkalic magmas generated by partial melting of garnet pyroxenite. Geology 31, 481–484. http://dx.doi.org/10.1130/0091-7613(2003)031<0481:AMGBPM>2.0.CO;2.

Hofmann, A.W., White, W.M., 1983. Ba, Rb, and Cs in the Earth’s mantle. Z. Natur-forsch., A 38, 256–266. http://dx.doi.org/10.1515/zna-1983-0225.

Huang, F., Lundstrom, C.C., Glessner, J., Ianno, A., Boudreau, A., Li, J., Ferré, E.C., Marshak, S., DeFrates, J., 2009. Chemical and isotopic fractionation of wet an-desite in a temperature gradient: experiments and models suggesting a new mechanism of magma differentiation. Geochim. Cosmochim. Acta 73, 729–749. http://dx.doi.org/10.1016/j.gca.2008.11.012.

Jackson, M.G., Hart, S.R., Koppers, A.A.P., Staudigel, H., Konter, J., Blusztajn, J., Kurz, M., Russell, J.A., 2007. The return of subducted continental crust in Samoan lavas. Nature 448, 684–687. http://dx.doi.org/10.1038/nature06048.

Jackson, M.G., Hart, S.R., Saal, A.E., Shimizu, N., Kurz, M.D., Blusztajn, J.S., Skov-gaard, A.C., 2008. Globally elevated titanium, tantalum, and niobium (TITAN) in ocean island basalts with high 3He/4He. Geochem. Geophys. Geosyst. 9, 21. http://dx.doi.org/10.1029/2007GC001876.

Jackson, M.G., Hart, S.R., Konter, J.G., Kurz, M.D., Blusztajn, J., Farley, K.A., 2014. He-lium and lead isotopes reveal the geochemical geometry of the Samoan plume. Nature 514, 355–358. http://dx.doi.org/10.1038/nature13794.

Johnson, C.M., Bell, K., Beard, B.L., Shultis, A.I., 2010. Iron isotope compositions of carbonatites record melt generation, crystallization, and late-stage volatile-transport processes. Mineral. Petrol. 98, 91–110. http://dx.doi.org/10.1007/s00710-009-0055-4.

Kelley, K.A., Cottrell, E., 2009. Water and the oxidation state of subduction zone magmas. Science 325, 605. http://dx.doi.org/10.1126/science.1174156.

Konter, J.G., Jackson, M.G., 2012. Large volumes of rejuvenated volcanism in Samoa: evidence supporting a tectonic influence on late-stage volcanism. Geochem. Geophys. Geosyst. 13, 1–23. http://dx.doi.org/10.1029/2011GC003974.

Koppers, A.A.P., Staudigel, H., Pringle, M.S., Wijbrans, J.R., 2003. Short-lived and discontinuous intraplate volcanism in the South Pacific: hot spots or ex-tensional volcanism? Geochem. Geophys. Geosyst. 4. http://dx.doi.org/10.1029/2003GC000533.

Koppers, A., Russell, J., Jackson, M., Konter, J., Staudigel, H., Hart, S., 2008. Samoa reinstated as a primary hotspot trail. Geology 36, 435.

Le Roux, V., Lee, C., Turner, S.J., 2010. Zn/Fe systematics in mafic and ultramafic sys-tems: implications for detecting major element heterogeneities in the Earth’s mantle. Geochim. Cosmochim. Acta 74, 2779–2796. http://dx.doi.org/10.1016/j.gca.2010.02.004.

Mattioli, G.S., Baker, M.B., Rutter, M.J., Stolper, E.M., 1989. Upper mantle oxygen fugacity and its relationship to metasomatism. J. Geol. 97, 521–536. http://dx.doi.org/10.1086/629332.

Natland, J.H., 1980. The progression of volcanism in the Samoan linear volcanic chain. Am. J. Sci. 280-A, 709–735.

Poitrasson, F., Freydier, R., 2005. Heavy iron isotope composition of granites de-termined by high resolution MC-ICP-MS. Chem. Geol. 222, 132–147. http://dx.doi.org/10.1016/j.chemgeo.2005.07.005.

Poitrasson, F., Delpech, G., Grégoire, M., 2013. On the iron isotope heterogeneity of lithospheric mantle xenoliths: implications for mantle metasomatism, the ori-gin of basalts and the iron isotope composition of the Earth. Contrib. Mineral. Petrol. 165, 1243–1258. http://dx.doi.org/10.1007/s00410-013-0856-7.

Polyakov, V.B., Mineev, S.D., 2000. The use of Mössbauer spectroscopy in stable iso-tope geochemistry. Geochim. Cosmochim. Acta 64, 849–865. http://dx.doi.org/10.1016/S0016-7037(99)00329-4.

Prytulak, J., Elliott, T., 2007. TiO2 enrichment in ocean island basalts. Earth Planet. Sci. Lett. 263, 388–403. http://dx.doi.org/10.1016/j.epsl.2007.09.015.

Putirka, K., 2008. Excess temperatures at ocean islands: implications for man-tle layering and convection. Geology 36, 283–286. http://dx.doi.org/10.1130/G24615A.1.

Richter, F., Dauphas, N., Teng, F., 2009. Non-traditional fractionation of non-traditional isotopes: evaporation, chemical diffusion and Soret diffusion. Chem. Geol. 258, 92–103. http://dx.doi.org/10.1016/j.chemgeo.2008.06.011.

Roeder, P.L., Emslie, R.F., 1970. Olivine–liquid equilibrium. Contrib. Mineral. Petrol. 29, 275–289. http://dx.doi.org/10.1007/BF00371276.

Rouxel, O., Dobbek, N., Ludden, J., Fouquet, Y., 2003. Iron isotope fractiona-tion during oceanic crust alteration. Chem. Geol. http://dx.doi.org/10.1016/j.chemgeo.2003.08.011.

Schoenberg, R., von Blanckenburg, F., 2006. Modes of planetary-scale Fe isotope fractionation. Earth Planet. Sci. Lett. 252, 342–359. http://dx.doi.org/10.1016/j.epsl.2006.09.045.

Schuessler, J., Schoenberg, R., Sigmarsson, O., 2009. Iron and lithium isotope sys-tematics of the Hekla volcano, Iceland — evidence for Fe isotope fractiona-tion during magma differentiation. Chem. Geol. 258, 78–91. http://dx.doi.org/10.1016/j.chemgeo.2008.06.021.

Shahar, A., Young, E., Manning, C., 2008. Equilibrium high-temperature Fe isotope fractionation between fayalite and magnetite: an experimental calibration. Earth Planet. Sci. Lett. 268, 330–338. http://dx.doi.org/10.1016/j.epsl.2008.01.026.

Sio, C.K.I., Dauphas, N., Teng, F.-Z., Chaussidon, M., Helz, R.T., Roskosz, M., 2013. Dis-cerning crystal growth from diffusion profiles in zoned olivine by in situ Mg–Fe isotopic analyses. Geochim. Cosmochim. Acta 123, 302–321. http://dx.doi.org/10.1016/j.gca.2013.06.008.

Sossi, P.A., Foden, J.D., Halverson, G.P., 2012. Redox-controlled iron isotope frac-tionation during magmatic differentiation: an example from the Red Hill in-trusion, S. Tasmania. Contrib. Mineral. Petrol. 164, 757–772. http://dx.doi.org/10.1007/s00410-012-0769-x.

Teng, F.-Z., Dauphas, N., Helz, R.T., 2008. Iron isotope fractionation during mag-matic differentiation in Kilauea Iki lava lake. Science 320, 1620–1622. http://dx.doi.org/10.1126/science.1157166.

Teng, F.-Z., Dauphas, N., Huang, S., Marty, B., 2013. Iron isotopic systematics of oceanic basalts. Geochim. Cosmochim. Acta 107, 12–26. http://dx.doi.org/10.1016/j.gca.2012.12.027.

Teplow, W.J., Marsh, B.D., Hulen, J., Spielman, P., Kaleikini, M., Fitch, D., Rickard, W., 2009. Dacite melt at the Puna geothermal venture wellfield, Big Island of Hawaii. Geotherm. Resour. Counc. Trans. 33, 989–994.

Todt, W., Cliff, R.A., Hanser, A., Hofmann, A.W., 1996. Evaluation of a 202Pb–205Pb double spike for high-precision lead isotope analysis. In: Basu, A., Hart, S.R. (Eds.), Earth Processes: Reading the Isotopic Code. In: Geophysical Monograph Series. American Geophysical Union, Washington, DC, pp. 429–437.

Weyer, S., Ionov, D., 2007. Partial melting and melt percolation in the mantle: the message from Fe isotopes. Earth Planet. Sci. Lett. 259, 119–133. http://dx.doi.org/10.1016/j.epsl.2007.04.033.

Weyer, S., Seitz, H.M., 2012. Coupled lithium- and iron isotope fractionation dur-ing magmatic differentiation. Chem. Geol. 294–295, 42–50. http://dx.doi.org/10.1016/j.chemgeo.2011.11.020.

232 J.G. Konter et al. / Earth and Planetary Science Letters 450 (2016) 221–232

White, W.M., Hofmann, A.W., 1982. Sr and Nd isotope geochemistry of oceanic basalts and mantle evolution. Nature 296, 821–825. http://dx.doi.org/10.1038/296821a0.

Williams, H.M., McCammon, C.A., Peslier, A., Halliday, A.N., Teutsch, N., Levasseur, S., Burg, J., 2004. Iron isotope fractionation and the oxygen fugacity of the mantle. Science 304, 1656–1659. http://dx.doi.org/10.1126/science.1095679.

Williams, H., Peslier, A., Mccammon, C., Halliday, A., Levasseur, S., Teutsch, N., Burg, J., 2005. Systematic iron isotope variations in mantle rocks and minerals:

the effects of partial melting and oxygen fugacity. Earth Planet. Sci. Lett. 235, 435–452. http://dx.doi.org/10.1016/j.epsl.2005.04.020.

Williams, H.M., Bizimis, M., 2014. Iron isotope tracing of mantle heterogene-ity within the source regions of oceanic basalts. Earth Planet. Sci. Lett. 404, 396–407. http://dx.doi.org/10.1016/j.epsl.2014.07.033.

Zambardi, T., Lundstrom, C.C., Li, X., McCurry, M., 2014. Fe and Si isotope variations at Cedar Butte volcano; insight into magmatic differentiation. Earth Planet. Sci. Lett. http://dx.doi.org/10.1016/j.epsl.2014.08.020.