Earth and Planetary Science Letters 404 (2014) 396–407

Contents lists available at ScienceDirect

Earth and Planetary Science Letters

www.elsevier.com/locate/epsl

Iron isotope tracing of mantle heterogeneity within the source regions

of oceanic basalts
Helen M. Williams a,∗ , Michael Bizimis b
a
Department of Earth Sciences, Durham University, Durham, DH1 3LE, UK
b
Department of Earth and Ocean Sciences, University of South Carolina, Columbia, SC 29208, USA

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

Article history: Mineralogical variations in the Earth’s mantle and the relative proportions of peridotitic versus enriched
Received 19 June 2013 and potentially crustally-derived pyroxenitic domains within the mantle have important implications for
Received in revised form 23 July 2014 mantle dynamics, magma generation, and the recycling of surface material back into the mantle. Here we
Accepted 29 July 2014
present iron (Fe) stable isotope data (δ 57 Fe, deviation in 57 Fe/54 Fe from the IRMM-014 standard in parts
Editor: T. Elliott
per thousand) for peridotite and garnet–pyroxenite xenoliths from Oahu, Hawaii and explore Fe isotopes
Keywords: as tracer of both peridotitic and pyroxenitic components in the source regions of oceanic basalts. The
Hawaii pyroxenites have δ 57 Fe values that are heavy (0.10 to 0.27h) relative to values for mid-ocean ridge
pyroxenite and ocean island basalts (MORB; OIB; δ 57 Fe ∼ 0.16h) and the primitive mantle (PM; δ 57 Fe ∼ 0.04h).
peridotite Pyroxenite δ 57 Fe values are positively correlated with bulk pyroxenite titanium and heavy rare earth
iron isotope element (REE) abundances, which can be interpreted in terms of stable isotope fractionation during
primitive mantle magmatic differentiation and pyroxene cumulate formation. In contrast, the peridotites have light δ 57 Fe
values (−0.34 to 0.14h) that correlate negatively with degree of melt depletion and radiogenic hafnium
isotopes, with the most depleted samples possessing the most radiogenic Hf isotope compositions and
lightest δ 57 Fe values. While these correlations are broadly consistent with a scenario of Fe isotope
fractionation during partial melting, where isotopically heavy Fe is extracted into the melt phase, leaving
behind low-δ 57 Fe peridotite residues, the extent of isotopic variation is far greater than predicted
by partial melting models. One possibility is derivation of the samples from a heterogeneous source
containing both light-δ 57 Fe (relative to PM) and heavy-δ 57 Fe components. While pyroxenite is a viable
explanation for the heavy-δ 57 Fe component, the origin of the depleted light-δ 57 Fe component is more
difficult to explain, as melting models predict that even large (>30%) degrees of melt extraction do not
generate strongly fractionated residues. Multiple phases of melt extraction or other processes, such as
metasomatism, melt percolation or the assimilation of xenocrystic olivine with light δ 57 Fe values may
need to be invoked to explain these light δ 57 Fe values; a caveat to this is that these processes must
either preserve, or generate correlations between δ 57 Fe and Hf isotopes. Published variations in δ 57 Fe in
mantle melting products, such as MORB and OIB, are also greater than predicted by melting models
assuming derivation from δ 57 Fe-homogeneous mantle. For example, OIB from the Society and Cook-
Austral islands, which have radiogenic Pb and Sr isotope compositions indicative of recycled components
such as subduction modified, low-Pb oceanic crust and terrigenous sediments have heavy mean δ 57 Fe
values (∼0.21h) significantly distinct to those of other OIB and MORB, which could explained by the
presence of heavy-δ 57 Fe pyroxenite cumulate or pyroxenitic melt components, whereas large degree
partial melts, such as komatiites and boninites, display light Fe-isotopic compositions which may reflect
sampling of refractory, light-δ 57 Fe mantle components. Iron stable isotopes may therefore provide a
powerful new means of fingerprinting mineralogical variations within the Earth’s mantle and identifying
the mineralogy of depleted and enriched components within the source regions of volcanic rocks.
© 2014 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/3.0/).

1. Introduction

Mineralogical variation in the Earth’s upper mantle and the

* Corresponding author. Tel.: +44 (0)191 334 2546; fax +44 (0)191 334 2546. potential existence of enriched (pyroxenitic or eclogitic, pyroxene-
E-mail address: [email protected] (H.M. Williams). dominated) and depleted (peridotitic, olivine-dominated) mantle

http://dx.doi.org/10.1016/j.epsl.2014.07.033
0012-821X/© 2014 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/3.0/).
 H.M. Williams, M. Bizimis / Earth and Planetary Science Letters 404 (2014) 396–407 397

components has been the subject of debate for several decades with which the incompatible trace element and radiogenic iso-
(Allègre and Turcotte, 1986; Hauri, 1996; Hofmann and White, tope signature of mantle rocks can be overprinted by metasomatic
1982; Workman et al., 2004). Numerous studies on erupted melts processes with little or no change in mineralogy (Niu and O’Hara,
have used major and trace elements and short- or long-lived ra- 2003).
diogenic nuclides to constrain mantle compositional and/or min- New tracers of both depleted and enriched mantle mineralogi-
eralogical variability (Elliott et al., 2007; Hauri, 1996; Humayun et cal components are thus required to compliment the extensive and
al., 2004; Jackson and Dasgupta, 2008; Prytulak and Elliott, 2007; rapidly growing evidence for mantle heterogeneity based on trace
Sigmarsson et al., 1998; Sobolev et al., 2005; Stracke et al., 1999; elements and radiogenic isotopes. Given the major advances that
Vlastelic et al., 1999). However, while pyroxenitic or eclogitic com- have been made in “heavy” (high atomic weight) metal stable iso-
ponents are often invoked to account for the enriched isotopic tope analyses over the last ten years, it is now timely to explore
signatures of oceanic basalts (Allègre and Turcotte, 1986; Hauri et the use of these stable isotope systems as tracers of mantle hetero-
al., 1996; Hirschmann and Stolper, 1996; Lassiter and Hauri, 1998; geneity. In this study, we explore the use of Fe stable isotopes as a
Lundstrom et al., 1999; Niu et al., 1999; Prinzhofer et al., 1989; tracer of mantle source mineralogy using peridotite and pyroxenite
Zindler et al., 1979; Zindler et al., 1984), their roles in generat- xenoliths from Hawaii as a case study.
ing mantle chemical heterogeneity remain controversial, as few
independent tracers of source mineralogy exist, such that resolv- 1.1. Iron isotopes as a tracer of mantle mineralogical variations
ing compositional (trace element or radiogenic isotope) enrich-
ment from mineralogical enrichment is challenging. For exam- Iron is a major cation in the Earth’s mantle, with a bulk
ple, radiogenic isotope systems such as Sr, Nd, Hf and Pb can partition coefficient close to 1 (Herzberg, 2004; Keshav et al.,
fingerprint crustally-derived components, but cannot distinguish 2004; Kogiso and Hirschmann, 2006; Pearce and Parkinson, 1993;
whether these components remain present as distinct lithological Pertermann and Hirschmann, 2003; Sobolev et al., 2005; Weyer
units (e.g. as pyroxenite or eclogite), or whether they are com- and Ionov, 2007). Consequently, Fe concentrations vary little with
pletely homogenized into the mantle by means of convective stir- degree of melting (at constant pressure) in primary MORB melts
ring (Gurenko et al., 2009; Jackson and Dasgupta, 2008) such that (Klein and Langmuir, 1987) and peridotites (Ionov and Hofmann,
only their geochemical signals remain. 2007), at least for degrees of melting appropriate for present day
Pyroxenite source components can be generated from both MORB and OIB (generally <15–20%). Theoretical (Polyakov and
crustal and mantle-sourced protoliths. ‘Crustal’ pyroxenite compo- Mineev, 2000) and empirical observations from equilibrated peri-
nents may be directly derived from recycled oceanic crust as eclog- dotites and pyroxenites (Weyer and Ionov, 2007; Williams et al.,
ite (Pertermann and Hirschmann, 2003) or may form by i) par- 2005) indicate that isotopically heavy Fe (high δ 57 Fe; parts per
tial melting of subducted eclogite and the reaction of these melts thousand deviation in 57 Fe/54 Fe from the IRMM-14 iron standard)
with mantle peridotite to form garnet pyroxenite (Hauri, 1996; will be concentrated in both low- and high-Ca pyroxenes rela-
Huang and Frey, 2005; Sobolev et al., 2005); this scenario has tive to olivine (by ca. 0.15 to 0.20h; Weyer and Ionov, 2007;
been invoked to explain the high SiO2 and Ni contents and high Williams et al., 2005) due to differences in bonding environment.
Fe/Mn ratios of some Hawaiian basalts (Sobolev et al., 2007; As melting preferentially consumes pyroxene over olivine, peri-
Sobolev et al., 2005), or ii) the extraction of silica-rich fluids or dotitic residues, melts and cumulates derived from melts of peri-
melts from oceanic crust during subduction (Kogiso et al., 2003). In dotitic and pyroxenitic source regions should inherit distinct Fe
contrast, ‘mantle’ pyroxenite components are considered to form as isotope signatures reflecting both the degree of melt extraction
high-pressure cumulates of low-degree mantle melts that infiltrate and the nature of the source mineralogy. Furthermore, as the Fe
and crystallize near the base of the oceanic mantle lithosphere contents of melts derived from peridotitic and pyroxenitic miner-
(Niu and O’Hara, 2003; Pilet et al., 2008) or through interaction alogies (e.g., Sobolev et al., 2005 their Table 1) are approximately
and melt–rock reaction between magmas and surrounding peri- similar, neither lithology should disproportionately contribute to
dotite wall-rock (Downes, 2007). For clarity, we use the term “py- the Fe budget of erupted melts and the Fe isotope compositions of
roxenite” to refer to source mineralogy (olivine-free, pyroxene and primitive lavas should therefore primarily reflect the mineralogies
garnet-bearing) rather than source origin unless this is referred to of their respective source regions.
specifically. While mineral-specific Fe-isotope partitioning effects have been
Depleted peridotitic components have also been invoked within recognized in a number of earlier studies (Beard and Johnson,
the source regions of OIB and MORB. Several studies have sug- 2004; Teng et al., 2008; Weyer et al., 2005; Weyer and Ionov,
gested that the Hawaiian plume mantle source contains long- 2007; Williams et al., 2004, 2009, 2005) the effects of variable
term depleted and compositionally variable peridotitic compo- source mineralogy have not yet been extensively explored in ei-
nents (Bizimis et al., 2013, 2005; Pietruszka and Garcia, 1999; ther models attempting to simulate Fe isotope fractionation dur-
Ren et al., 2006; Stracke et al., 1999) and depleted components ing partial melting or in calculations of the Fe isotope compo-
have also been invoked in the source of Reunion OIB (Vlastelic sition of the Earth’s mantle. Existing estimates of the Fe iso-
et al., 2006) and MORB. Evidence for the latter is provided by tope composition of the Earth’s mantle or the bulk silicate Earth
correlated Hf–Nd isotopes in MORB (Salters et al., 2011) and the (BSE) are constrained by sampling to the upper mantle and are
radiogenic Hf isotope compositions of Gakkel Ridge abyssal peri- generally based on suites of comparatively primitive basalts or
dotites (Stracke et al., 2011), while mantle peridotites found at their melting residues. In an early study (Weyer et al., 2005), it
mid-oceanic spreading centers and as xenoliths in OIB provide was observed that unmetasomatized peridotites from a variety of
some of the strongest evidence for the presence of ancient de- tectonic settings displayed near-chondritic δ 57 Fe values whereas
pleted peridotites in the convecting mantle (Bizimis et al., 2007; oceanic basalts displayed heavier δ 57 Fe values (mean ∼0.16h),
Burton et al., 2012; Liu et al., 2008; Stracke et al., 2011). How- similar to those of high-Mg lunar basalts but heavier than the
ever, identifying the presence of depleted components in man- chondritic δ 57 Fe values displayed by SNC meteorites and eucrites
tle source regions is not without difficulty. A major challenge is (Poitrasson et al., 2004; Schoenberg and von Blanckenburg, 2006;
the low incompatible element concentrations of refractory mantle Weyer et al., 2005). Weyer et al. (2005) concluded that the mean
peridotites, which means that they have little influence on the in- peridotite δ 57 Fe provided the best estimate of the BSE, noting that
compatible trace element and radiogenic isotope budget of erupted relative differences in planetary mantle δ 57 Fe could only be esti-
melts (Burton et al., 2012). Another challenge relates to the ease mated through comparisons of erupted basalts and their meteoritic
398 H.M. Williams, M. Bizimis / Earth and Planetary Science Letters 404 (2014) 396–407

equivalents, as meteorites representing the residues of mantle par- unknown, we recast αmelt-residue as a function of the fractiona-
tial melting on the Moon, Mars and Vesta are not available. In a tion factor between melt and clinopyroxene (αmelt-cpx ; R melt / R cpx ),
latter study (Dauphas et al., 2009) the δ 57 Fe value of the BSE was which we treat as a free variable, and subsolidus clinopyroxene-
estimated i) by intersecting a correlation between δ 57 Fe and Fe/Si mineral fractionation factors (αmineral-cpx ; R min / R cpx ), which we
in chondrites with the current Fe/Si estimated for the BSE, giving estimate from observed differences in δ 57 Fe between equilibrated
a BSE δ 57 Fe value of 0.02 ± 0.07h, and ii) from the mean δ 57 Fe phases in mantle peridotites, pyroxenites and eclogite xenoliths
of boninites (0.04 ± 0.01h), which as high-degree melts should (Weyer and Ionov, 2007; Williams et al., 2009, 2005).
approximate the isotopic composition of their source region. More At each melting increment, the modal abundance of each phase
recent estimates of the BSE have been obtained from abyssal peri- (n, Eq. (2)) and its Fe concentration is calculated by mass balance
dotites (0.04 ± 0.04h; Craddock et al., 2013) and the ∼90 Ma from the original source mineralogy and mineral Fe contents cou-
Gorgona komatiite suite (−0.13 ± 0.05h; Hibbert et al., 2012). pled with literature partition coefficients for Fe. Melting reactions
While all these estimates of the BSE δ 57 Fe value are useful, they are defined using liquid modes (p values) where positive values
do not provide much information on the relative Fe isotope com- indicate consumption during melting, negative values crystalliza-
positions of planetary mantles, as samples of comparable petrology tion:
have not yet been obtained from the other terrestrial planets and
insufficient information is available regarding the degree to which nmineral residue = nmineral initial − ( F × p mineral ) (2)
Fe isotopes fractionate during mantle melting and the extent to where F = the degree of partial melting, and p describes the
which this may depend on variations in planetary mantle oxida- relative proportion of a given mineral in the melting reaction.
tion state and mineralogy. More critically, current estimates of BSE Melting reactions for the lherzolite lithologies were taken from
δ 57 Fe do not constrain the potential variability in δ 57 Fe that may Robinson et al. (1998). Clinopyroxene is exhausted at 25% melt-
exist in the Earth’s mantle, let alone that of the other terrestrial ing after which harzburgite melting modes (given in the Ap-
planets. For example, do the mantle source δ 57 Fe values estimated pendix A and based on the study of Parman and Grove, 2004) are
from either boninites (Dauphas et al., 2009) or the Gorgona ko- used to take into account this change in lithology. The Fe con-
matiites (Hibbert et al., 2012) provide a true measure of PM or tent of the bulk residue is then calculated as the weighted sum
BSE δ 57 Fe, or do they represent preferential sampling of more de- of the residual mineral modal abundances (normalized to 100%)
pleted, low-δ 57 Fe domains within the mantle? and their Fe contents. The fractionation factor between the melt
and bulk residue, αmelt-residue is calculated at each melting in-
1.2. Models of Fe isotope fractionation during partial melting crement using the Fe content of the bulk residue and the value
assigned to αmelt-cpx , where n = normalized mineral modal abun-
Although there are, to our knowledge, no experimental stud- dance:
ies of Fe isotope fractionation between silicate melts and minerals, 

n
it is widely observed that the products of partial melting (e.g. αmelt-residue = αmelt-cpx ∗ [ni · FeOmineral ]
MORB and OIB) are displaced to higher δ 57 Fe values (typically by
i =1
0.1–0.2h) than unmetasomatized mantle peridotites (Dauphas et 
al., 2009; Weyer and Ionov, 2007; Williams et al., 2004, 2005). This 
n
/ [ni · αmin-cpx · FeOmineral ] (3)
effect could relate to the more incompatible nature of Fe3+ dur-
i =1
ing melting (Dauphas et al., 2009; Williams et al., 2004), and/or
to the non-modal nature of partial melting, where (isotopically The isotopic compositions of the instantaneous melt and residue
heavier) pyroxene contributes disproportionately to the melting as- at each step can then be calculated using α melt-residue and Eq. (1).
semblage. The effects of non-modal melting can be explored with The value of R melt calculated at the first melting increment uses
partial melting models where observed differences in δ 57 Fe be- the initial isotopic composition assumed for the source (0.36259
tween equilibrated olivine and clinopyroxene in mantle peridotites or 0.04h relative to IRMM-14). The value of R residue at this
and clinopyroxene and olivine in mantle eclogites (Weyer and and subsequent melting increments is calculated by mass bal-
Ionov, 2007; Williams et al., 2009, 2005) are used to approximate ance between the original source and value of R melt calculated
relative differences in mineral-melt Fe isotope partitioning. These for the extracted melt. The concentration and isotope composi-
differences reflect subsolidus mineral equilibration at temperatures tion of the aggregate melt for any given melt fraction is calcu-
ranging from ca. 900 to 1200 ◦ C (Weyer and Ionov, 2007; Williams lated by the sum of the weighted melt increments. The model
et al., 2009, 2005) rather than mantle potential temperatures and assumes that each melt increment is completely extracted and
it may be expected that they will decrease with increasing tem- each subsequent melting step uses the concentration, modal
perature (as 1/ T 2 ). In the models discussed below, we have not abundances and Fe-isotope composition of the previous residue.
applied a temperature correction due to the absence of experimen- The model is checked so that mass balance is satisfied, i.e. the
tal data describing the temperature dependence of inter-mineral Fe combined sum of the concentration and isotope composition of
isotope fractionation and because no correlations between mineral the residue and cumulate melt equates to that of the original
fractionation factors and apparent equilibration temperature have source.
been observed in the existing studies. This model approach ensures that at each step each mineral re-
We use an incremental batch non-modal melting model (see mains in isotope equilibrium with the melt and the other minerals
Appendix A for details) where the modal mineralogy of the residue in the residue. Note that αmelt-residue is not simply a weighted av-
is recalculated at each melting increment. At each melting step the erage of αmelt-min values; rather it is dictated by the equilibrium
Fe-isotope fractionation between melt and residue is dictated by with the mineral assemblage, and buffered by the presence of each
the melt and residue fractionation factor, α : mineral. Although this calculation does not implicitly model the
greater incompatibility of Fe3+ during partial melting relative to
R melt = αmelt-residue ∗ R residue (1) Fe2+ (Woodland and Koch, 2003) or the preferential concentration
of isotopically heavy Fe into Fe3+ -bonding environments (Polyakov
57 54
Where R melt and R residue are the absolute Fe/ Fe isotopic com- et al., 2007; Polyakov and Mineev, 2000), these effects are allowed
positions of the melt and the residue at any given melting in- for via the dominance of clinopyroxene in the melting assemblage
crement. As melt-mineral α (and hence melt-residue) values are relative to olivine.
 H.M. Williams, M. Bizimis / Earth and Planetary Science Letters 404 (2014) 396–407 399

Fig. 1. Melt and residue δ 57 Fe values generated by fractional melting of a primitive mantle source of δ 57 Fe = 0.04h plotted against degree of partial melting, or mole fraction
( F ) of FeO removed for αmelt-cpx = 1. Mineral modes, liquid modes, iron contents and isotopic compositions used in the models are given in Table A.1. Both lherzolite and
pyroxenite sources were constrained to have initial δ 57 Fe values of 0.04h in line with the current PM estimate and both sources also have the same starting FeO contents
of 8.3 wt%.

The calculated δ 57 Fe values of the evolving residues and asso- 1.3. The Fe isotope composition of the primitive mantle
ciated melts are shown in Fig. 1, for fertile lherzolite and pyrox-
enite mineralogies with the same initial Fe isotope compositions As discussed above, MORB and OIB have heavy δ 57 Fe (Beard
and bulk FeO contents (detailed in the Appendix A). The results et al., 2003; Dauphas et al., 2009; Teng et al., 2008, 2013; Weyer
show that pyroxenite melts are heavier than lherzolitic partial et al., 2005) relative to mantle peridotites (Craddock et al., 2013;
melts at the same degree of partial melting (e.g. for F = 0.10 and Weyer et al., 2005; Williams et al., 2005) and chondrites (Dauphas
αmelt-cpx = 1, lherzolite and pyroxenite melts have δ 57 Fe values of et al., 2009). While this difference is qualitatively consistent with
0.15 and 0.25h, respectively) although the difference decreases melting models, the overall range in δ 57 Fe displayed by MORB and
OIB (Fig. 2) is over 0.2h and 0.4h, respectively, for a restricted
with increasing degree of melting. At a wide range of melting
range of MgO contents (see Figure caption for details), and is sub-
degrees (0.01 < F < 0.25) for αmelt-cpx = 1, instantaneous and ag-
stantially greater than that predicted in our melting models and
gregate melts calculated for the lherzolite source show limited
those of Dauphas et al. (2009) and is much larger than can be ex-
variation in δ 57 Fe, with values ranging from 0.141 to 0.149 and
plained by analytical uncertainty or processes such as fractional
0.131 to 0.149h. Melting residues also display a limited range in crystallization (Schuessler et al., 2009), olivine accumulation (Teng
δ 57 Fe, from 0.008 to 0.040h. Increasing the value of αmelt-cpx in- et al., 2008) and fluid exsolution (Heimann et al., 2008). A similar
creases the δ 57 Fe values of the calculated melts and the difference discrepancy also exists for melting residues as the δ 57 Fe variation
in δ 57 Fe between melt and residue, although it should be noted displayed by fresh abyssal peridotites (Craddock et al., 2013), com-
that αmelt-cpx values >1 are not required to explain the observed monly thought to represent the residues of MORB-melt extraction,
offset in δ 57 Fe between mean MORB or OIB and mantle peridotites. is 0.24h, far greater than that predicted by any model.
The models therefore demonstrate that non-modal partial melting One explanation for this observed variability in melt and
scenarios, where the melting phases is dominated by clinopyrox- residue δ 57 Fe is derivation of these samples from a δ 57 Fe-hetero-
ene, can readily explain the observed offset in δ 57 Fe between mean geneous mantle, as previously suggested for Ko’olau OIB (Teng et
MORB or OIB and mantle peridotites without requirement for addi- al., 2013). In order to evaluate the potential for mantle Fe-isotope
heterogeneity, the underlying causes of that heterogeneity and the
tional mineral-melt fractionation. Critically, the models also predict
suitability of Fe isotopes as a tracer of mantle mineralogy the
that there will be minimal variation in the δ 57 Fe values of erupted
following questions need to be addressed: i) the extent of Fe-
melts and residues across a wide range in melting degree.
isotope variation that exists between different mantle lithologies;
Models of Fe isotope fractionation during partial melting have
ii) the nature of the processes generating Fe isotope variations be-
been presented in other studies (Weyer and Ionov, 2007; Williams tween and within different mantle lithologies and iii) the extent to
et al., 2009, 2005), most recently by Dauphas et al. (2009). The which mantle region source Fe-isotope heterogeneity can explain
latter study differs from our own in that it does not use data the variations in δ 57 Fe observed in oceanic basalts.
from natural samples and it does not take into account relative To address these questions we have determined the Fe isotope
Fe isotope partitioning between minerals; rather Fe isotope frac- compositions of silicate minerals (Table 1) from well-characterized
tionation is driven by differences in Fe3+ –Fe2+ partitioning, where peridotite and pyroxenite xenoliths from Oahu, Hawaii. The peri-
+
it is assumed that the magnitude of fractionation between Fe3melt dotites are considered to be fragments of 90–100 Ma Pacific
+ oceanic lithosphere or even ancient recycled mantle material
and Fe3source is ca 0.30h (for δ 56 Fe, to convert to δ 57 Fe the value
would be 0.45h) and there is no fractionation between Fe2melt +
and within the Hawaiian mantle plume (Bizimis et al., 2007, 2004).
+ The garnet pyroxenites are considered to be high-pressure cumu-
Fe2source . Both these models and our models can account for the
lates from OIB-like melts that formed close to the lithosphere–
observed difference in MORB and OIB from mantle peridotites. The
asthenosphere boundary (60–90 km) (Bizimis et al., 2013, 2005;
slight difference between our models and those of Dauphas et al. Sen et al., 2005, 2011), while some samples with majorite pseu-
(2009) may stem from the fact we have not incorporated an addi- domorphs (Keshav and Sen, 2001) and nanodiamonds (Wirth and
tional α term to allow for the more incompatible behavior of Fe3+ , Rocholl, 2003) potentially originated at depths >150 km. In this
although, as emphasized above, our models do take the behavior study, we use peridotite Fe isotope compositions to infer the Fe-
of Fe3+ into account indirectly, though the greater contribution of isotope systematics of the peridotitic upper mantle and pyroxenites
clinopyroxene to the melting assemblage. to explore the Fe-isotope systematics of mineralogically enriched
400 H.M. Williams, M. Bizimis / Earth and Planetary Science Letters 404 (2014) 396–407

Fig. 2. Literature iron isotope data for MORB (a) (EPR, East Pacific Rise; MAR, Mid Atlantic Ridge) and OIB (b). Data sources and key to legend: W-2007 (Weyer and Ionov,
2007); W-2005 (Weyer et al., 2005); T-2013 (Teng et al., 2013); B & J – 2003 (Beard et al., 2003); T-2008 (Teng et al., 2008). The Fe isotope composition inferred for the PM
(0.04h) is shown as a horizontal line on both plots. Errors on isotopic composition are typically 0.02 to 0.09h and smaller than the overall variation in δ 57 Fe observed
(our 2 S.D. long term reproducibility is shown for reference as an error bar on this plot). Icelandic basalts (Schuessler et al., 2009) were not plotted as i) they sample both
MORB-source and enriched mantle source regions ascribed to the Icelandic plume, which could make identifying differences in Fe-isotope systematics of MORB and OIB more
difficult ii) they are comparatively fractionated, as detailed by (Schuessler et al., 2009). Samples <5.5 wt% and >14 wt% MgO were excluded to avoid magmatic fractionation
(Schuessler et al., 2009) and olivine accumulation effects (Teng et al., 2008) respectively.

lithologies in the mantle, and we evaluate these results in the con- 2001), ranges from ∼1–12% (Table 1). Clinopyroxene and spinel
text of published Fe isotope data for MORB and OIB. Cr# and clinopyroxene HREE contents are all highly correlated
(Bizimis et al., 2007), consistent with major and trace element
2. Materials and methods equilibration between the primary mineral phases. All the clinopy-
roxenes from the samples reported here exhibit different degrees
2.1. Samples of light rare earth element (LREE) enrichment, with concave down
to concave up chondrite-normalized REE patterns, consistent with
The samples studied here are mantle xenoliths from the Salt variable refertilization by incompatible element enriched melts
Lake Crater (SLC) Pali and Kaau vents that belong to the Honolulu (Bizimis et al., 2007, 2004). Based on their combined major, trace
Volcanics series in Oahu, Hawaii, which are part of the rejuvenated element and Sr–Nd–Hf–Os isotope systematics, the Pali and Kaau
or post-erosional stage of the Hawaiian volcanism (Clague and Frey, peridotites are generally thought to represent comparatively un-
1982; Ozawa et al., 2005; Sen et al., 2005). A location map is pro- modified parts of the Pacific lithosphere beneath Oahu (Bizimis et
vided in Appendix A. In this study we focus on previously studied, al., 2007, 2004; Sen et al., 1993). In contrast, the SLC peridotites
well-characterized spinel lherzolite and garnet pyroxenite xeno- have experienced, on average, a greater extent of melt depletion
liths (Bizimis et al., 2013, 2004, 2005; Sen et al., 2011) as well (e.g. higher pyroxene Mg# and Cr#), and re-enrichment (higher
as some newly reported samples. The samples are from the Dale Na and LREE) relative to the Pali and Kaau peridotites. They have
Jackson and Dean Presnall collections at the Smithsonian Institu- highly radiogenic Hf isotope (Bizimis et al., 2007, 2004; Salters and
tion. The peridotites are all spinel lherzolites with ∼5–12 modal Zindler, 1995) and unradiogenic Os isotope compositions, with Re-
% clinopyroxene, and are fresh, free of serpentinization and visi- depletion ages up to 2 Ga (Bizimis et al., 2007), which have been
ble melt infiltration (e.g. veins). The bulk rock Mg# (reconstructed explained by a scenario in which these peridotites represent frag-
from mineral abundance and compositions) vary from 0.89 to 0.90, ments of ancient (>1 Ga old) recycled lithosphere entrained as
ranging from fertile (McDonough and Sun, 1995) or Depleted Man- part of the upwelling Hawaiian plume.
tle (Salters and Stracke, 2004) compositions to more Mg-rich, and The pyroxenites are classified as garnet–clinopyroxenites with
therefore more depleted values. The degree of melting experienced 10–30 volume % garnet, >50% clinopyroxene (augite) and sub-
by the peridotites, as calculated from spinel Cr# (Hellebrand et al., ordinate amounts of olivine, orthopyroxene, spinel and traces of
 H.M. Williams, M. Bizimis / Earth and Planetary Science Letters 404 (2014) 396–407 401

Table 1
Iron isotope data for Hawaiian peridotite and pyroxenite xenoliths.

Peridotites Pyroxenites
Pa 27 KAPS-36 77SL-405 77SL-466 77SL-341 77SL-402 77SL-594 77SL-601 77SL-582 NMNH-114954-20A
% cpx 10 10 5 12 10 10 84.7 70 70 82
% opx 25 28 25 24 20 28 5 0 0 0
% garnet – – – – – – 10 30 30 8
% olivine 63 60 68 63 68 60 – – – –
% spinel 2 2 2 2 1 2 0.3 – – –
δ 57 Fe cpx 0.16 0.12 −0.07 0.13 −0.21 −0.04 0.17 0.33 0.14 0.25
2 S.D. 0.03 0.04 0.09 0.06 0.07 0.05 0.06 0.03 0.02 0.01
δ 56 Fe cpx 0.08 0.08 −0.04 0.07 −0.14 −0.03 0.12 0.23 0.09 0.16
2 S.D. 0.04 0.06 0.06 0.03 0.03 0.04 0.01 0.02 0.01 0.02
δ 57 Fe opx −0.09 0.12 −0.20 0.03 −0.08 −0.04
2 S.D. 0.06 0.01 0.01 0.04 0.09 0.04
δ 56 Fe opx −0.07 0.08 −0.13 0.04 −0.07 −0.02
2 S.D. 0.02 0.04 0.01 0.02 0.06 0.02
δ 57 Fe ol −0.07 0.15 −0.38 0.10 −0.21 0.00
2 S.D. 0.03 0.06 0.06 0.04 0.04 0.02
δ 56 Fe ol −0.06 0.09 −0.25 0.05 −0.15 −0.02
2 S.D. 0.02 0.07 0.04 0.06 0.04 0.01
δ 57 Fe gt 0.15 0.21 0.04 0.03
2 S.D. 0.01 0.03 0.05 0.07
δ 56 Fe gt 0.08 0.10 0.01 0.02
2 S.D. 0.01 0.06 0.02 0.03
δ 57 Fe bulk −0.07 0.14 −0.34 0.09 −0.19 −0.01 0.17 0.27 0.10 0.21
2 S.D. 0.08 0.08 0.11 0.08 0.12 0.07 0.06 0.05 0.05 0.07
δ 56 Fe bulk −0.06 0.09 −0.23 0.05 −0.14 −0.02 0.11 0.16 0.05 0.14
2 S.D. 0.05 0.10 0.07 0.06 0.08 0.05 0.01 0.06 0.02 0.04
57 Fe cpx-ol 0.23 −0.02 0.31 0.03 0.01 −0.03
2 S.D. prop 0.04 0.08 0.11 0.07 0.08 0.05
57 Fe cpx-opx 0.25 0.01 0.13 0.10 −0.13 0.00
2 S.D. prop 0.07 0.05 0.09 0.07 0.11 0.06
57 Fe cpx-gt 0.02 0.13 0.11 0.22
2 S.D. prop 0.06 0.05 0.05 0.07
FeOt bulk 8.69 8.49 8.52 8.22 8.22 8.66 7.44 8.00 7.74 7.29
Cr# bulk 0.08 0.06 0.17 0.07 0.13 0.11 0.00 0.00 0.01 0.00
Mg# bulk 0.89 0.89 0.90 0.90 0.90 0.89 0.75 0.76 0.79 0.73
% melting 1 .5 1 .5 12.3 2 .4 9.3 5.8
T (C) 1008 1012 1079 1041 1039 1046 1006 1236 1275 1283

ε Hf(cpx) ± 2 18.9 ± 0.8 11.8 ± 0.7 26.3 ± 1 19.3 ± 0.5 24.3 ± 0.7 28.3 ± 0.1 12.9 ± 0.2 15.4 ± 0.5 15.1 ± 0.8 14.3 ± 0.4

Mineral abbreviations: cpx, clinopyroxene; opx, orthopyroxene; ol, olivine; gt, garnet. Errors on Fe isotope measurements are 2 S.D. calculated for replicate analyses. Errors
on bulk values are propagated using mineral errors and standard error propagation techniques. Iron isotope measurements followed protocols detailed in Methods. Bulk rock
molar Mg# = Mg/(Fe + Mg) and Cr# = Al/(Cr + Al) values were calculated based on mineral modes and compositions from Bizimis et al. (2004, 2005) and new data
generated as in those studies. Equilibration temperatures are taken from Bizimis et al. (2004, 2005) and peridotite and garnet pyroxenite pressures are assumed to be 18 and
25 kbar, respectively. Temperature data for samples 77SL-402, 77SL-594 and NMNH-114954-20A are calculated as in the previous studies. Hafnium isotope compositions are
from Bizimis et al. (2004, 2005) with new data for samples 77SL-402, 77SL-594 and NMNH-114954-20A reported here (see Methods section). The epsilon notation refers to
the deviation from chondritic Hf (176 Hf/177 Hf = 0.282785) in parts per 10,000. The degree of partial melting was calculated following Hellebrand et al. (2001).

phlogopite. Based on major and trace element modeling and their eral grains between 120–200 μm in size. The samples were picked
Sr, Nd, Hf, Pb, and Os isotope compositions they have been largely under ethanol using a binocular microscope avoiding crystals with
interpreted as high pressure (>2 GPa, >60 km) cumulates near the obvious cracks, external alteration and mineral or fluid inclusions
base of the Pacific lithosphere from melts similar to the Hawai- and were subsequently cleaned in ultrapure Millipore® 18.2  wa-
ian alkali rejuvenated lavas (Bizimis et al., 2013, 2005; Keshav et ter. Dissolution, iron purification and isotopic analyses were under-
al., 2007; Sen et al., 2010). Some samples display ilmenite exso- taken at Durham University using established procedures (Hibbert
lutions within garnet (including samples NMNH-114959-20A and et al., 2012; Williams et al., 2012). Isotopic analyses were car-
77SL-582 analyzed here), majorite pseudomorphs and nanodia- ried out on a multiple-collector inductively coupled plasma mass
monds and probably initially crystallized at >5 GPa (>150 km) spectrometer (MC-ICPMS; Thermo Neptune) (Williams et al., 2012).
(Keshav and Sen, 2001; Keshav et al., 2007), deeper than the Sample solutions consisted of 0.9 to 1.5 ppm Fe (different con-
80–90 km seismically defined base of the lithosphere beneath centrations were chosen on different days according to instrument
Oahu (Li et al., 2004). The radiogenic Hf and Nd isotopic compo- sensitivity) in 0.1M HNO3 , and instrumental mass bias was cor-
sitions of the pyroxenites analyzed thus far (Bizimis et al., 2013, rected for by sample–standard bracketing where the sample and
2005), including the high pressure pyroxenite NMNH-114959-20A, standard Fe beam intensities (typically 35–40 V 56 Fe for a stan-
place them at the depleted end of the OIB array, suggesting an ori- dard 1011  resistor) were matched to 5%. Mass dependence,
gin from a depleted mantle source, distinct from the isotopically long-term reproducibility and accuracy were evaluated by anal-
enriched “eclogitic” component inferred for the Hawaiian plume ysis of an in-house FeCl salt standard (δ 57 Fe = −1.06 ± 0.07h;
(Hauri, 1996; Huang and Frey, 2005). δ 56 Fe = −0.71 ± 0.06h 2 S.D., n = 35) previously analyzed in other
studies (Hibbert et al., 2012; Williams et al., 2012). The interna-
2.2. Iron isotope analyses tional rock standards BIR-1 (Icelandic basalt) and Nod-PI (Pacific
ferromanganese nodule) were also analyzed over the course of this
Iron isotope analyses were carried out on hand-picked mineral study (Table A.2). The mean Fe isotope compositions of these stan-
grains, where individual sample aliquots consisted of 40–60 min- dards are: BIR-1, δ 57 Fe = 0.082 ± 0.01h; δ 56 Fe = 0.062 ± 0.01h
402 H.M. Williams, M. Bizimis / Earth and Planetary Science Letters 404 (2014) 396–407

(2 S.D., n = 6), Nod-P1 δ 57 Fe = −0.837 ± 0.02h; δ 56 Fe = −0.569 ± 4. Discussion

0.03h (2 S.D., n = 7). The data for BIR-1 are in excellent agree-
ment with earlier studies (Hibbert et al., 2012; Millet et al., 2012; 4.1. Peridotite Fe isotope compositions and the role of melt extraction
Weyer et al., 2005) and the results for Nod-PI fall within the
median of literature values which range from δ 56 Fe = −0.67 to Strong negative correlations exist between bulk peridotite δ 57 Fe
δ 56 Fe = −0.49h (Asael et al., 2013; Gagnevin et al., 2012). Iron and indicators of melt extraction where the most depleted peri-
yields were quantitative and chemistry blanks were <0.5 ng Fe, dotites (i.e. highest spinel and bulk rock Cr#, lowest clinopyroxene
negligible compared to the quantities of sample Fe (>300 μg) pro- Al-contents; Fig. 3a, b, c) display the lightest δ 57 Fe values. A neg-
cessed. ative array is also present between δ 57 Fe and hafnium isotopes
(ε Hf; Fig. 3d), a radiogenic isotopic system that records the time-
integrated fractionation of Hf from Lu during partial melting and
2.3. Hafnium isotope analyses
which has been shown to be generally resistant to overprint by
mantle metasomatism (Bizimis et al., 2004; Lazarov et al., 2012;
New hafnium isotope data on clinopyroxene from samples Stracke et al., 2011). While Fe stable isotopes do not record tempo-
77SL-402 and 77SL-594 were obtained at the Center for Elemen- ral information, the observed correlation between Fe isotopes and
tal Mass Spectrometry, University of South Carolina, on a Thermo Hf isotopes does suggest that the samples with the greatest radio-
Neptune MC-ICPMS using established chemical procedures (Bizimis genic isotope signatures of time-integrated melt extraction are also
et al., 2007, 2013), and are reported here for data completeness. the most isotopically fractionated in terms of Fe. One possibility
The JMC Hf standard yielded values of 176 Hf/177 Hf = 0.282142 ± is that Fe and Hf isotopes record the same partial melting events.
5 (2 S.D., n = 10; 20 ng runs), and all ratios are reported relative However, such a direct link is not supported by the melting models
to 176 Hf/177 Hf = 0.282160 for this standard. Hafnium blanks were in Section 1.2 coupled with the estimated degree of melt extraction
<50 pg. Hafnium isotope data reported in Table 1 in the epsilon (Hellebrand et al., 2001) recorded by the peridotites, which are not
notation relative to a chondritic 176 Hf/177 Hf = 0.282785. sufficient to generate the observed range in peridotite δ 57 Fe. Other
explanations for the range in peridotite δ 57 Fe must therefore be
3. Results considered.
While hydrothermal alteration and seawater contamination are
processes that can influence the trace element and radiogenic iso-
Bulk peridotite δ 57 Fe values (Table 1 and Fig. 3) were calcu-
tope budgets of oceanic basalts and abyssal peridotites (Burton et
lated from mineral separate data, mineral modal abundances and
al., 2012), the samples studied here are pristine with little evidence
Fe concentrations (Table 1) and range from −0.34 ± 0.11h to
of post-magmatic alteration (Bizimis et al., 2004, 2005; Sen et al.,
0.14 ± 0.08h; bulk pyroxenite δ 57 Fe varies from 0.10 ± 0.05h
2011). Moreover, composite oceanic crust (Rouxel et al., 2003) and
to 0.27 ± 0.05h (all errors are 2 S.D., propagated using sum-of-
sheeted dyke samples (Williams et al., 2009) display δ 57 Fe values
squares and the 2 S.D. errors in Table 1). These differences are
within error of mean MORB, suggesting that these processes have
also evident at the mineral scale as peridotite clinopyroxene δ 57 Fe
little impact on Fe isotopes. Incorporation of xenocrystic olivine,
values range from −0.21 ± 0.07h to 0.16 ± 0.03h, whereas py-
discussed in the previous section, could in principle contribute to
roxenite clinopyroxenes have heavier δ 57 Fe values of 0.14 ± 0.02h
the low δ 57 Fe value of 77SL-405, but it is not clear how this pro-
to 0.33 ± 0.03h. Apparent inter-mineral fractionation factors are
cess would generate correlations between δ 57 Fe and Hf isotopes.
also presented in Table 1 and Fig. 4. In the peridotites, clinopy-
Metasomatic refertilisation, and melt-peridotite reaction are pro-
roxene is either isotopically heavier, or within error identical to
cesses that can explain the Hf-decoupling from Sr and Nd isotopes
coexisting olivine. A similar pattern is observed for clinopyroxene– and the elevated light rare earth element (LREE) abundances of the
orthopyroxene pairs with the exception of sample 77SL-341, where Oahu peridotites (Bizimis et al., 2004). However these mechanisms,
orthopyroxene displays a heavier δ 57 Fe value relative to clinopy- without any additional stable isotope fractionation process, can-
roxene. In the pyroxenites, clinopyroxenes display δ 57 Fe values that not explain the range in peridotite δ 57 Fe as relationships between
are heavier or, in one case (sample 77SL-594) within error of co- δ 57 Fe and Nd and Sr isotopes or Na and LREE contents are absent.
existing garnet. While the overall tendency of clinopyroxene to Furthermore, mass balance calculations show that unfeasibly large
display heavy δ 57 Fe values with respect to olivine, orthopyrox- ratios of melt to peridotite (91:9, for a melt with a δ 57 Fe value
ene and garnet is consistent with theory and studies of natural of ∼0.15h and the peridotite 77SL-405 with the lowest bulk rock
peridotites, pyroxenites and eclogites (Beard and Johnson, 2004; δ 57 Fe) are required, which are inconsistent with the high Mg#’s
Weyer and Ionov, 2007; Williams et al., 2009, 2005), the variability and Cr#’s of the peridotites (Table 1).
observed in calculated intra-mineral fractionation factors (Table 1) One means of reconciling the relatively large variation in peri-
is not consistent with isotopic equilibrium over the narrow range dotite δ 57 Fe and the correlations between peridotite δ 57 Fe and
of equilibration temperatures inferred for these samples (Table 1). indicators of partial melt extraction could be derivation from a
This variation could reflect late-stage Fe-isotope redistribution be- δ 57 Fe-heterogeneous mantle. In this scenario, the correlations be-
tween minerals, or potentially melt–rock reaction and diffusion tween peridotite δ 57 Fe, ε Hf and proxies for melt extraction in-
processes. However, no systematic correlations between mineral dicate that the processes responsible for Fe-isotope heterogene-
fractionation factors or other mineral or whole-rock chemical pa- ity must be magmatic in nature. Distinct mantle source regions
rameters are evident, suggesting that the processes which induced have previously been invoked to explain the elemental (Dasgupta
mineral isotopic disequilibria did not modify the Fe isotope com- et al., 2010; Hirschmann and Stolper, 1996; Prytulak and Elliott,
positions of the bulk samples. An alternative scenario could be 2007) and radiogenic isotope signatures (Chauvel et al., 1992;
the incorporation of isotopically light xenocrystic olivine, which Hofmann, 1997; Salters et al., 2011; Stracke et al., 2005, 1999;
is known to be present in the Hawaiian system as a function of White and Hofmann, 1982; Workman et al., 2004) of many oceanic
fractional crystallization processes (Teng et al., 2008). This process basalt suites. At least two modified mantle source components in
would increase measured clinopyroxene–olivine fractionation fac- addition to primitive mantle (PM; 0.04h; Dauphas et al., 2009)
tors and reduce bulk δ 57 Fe, and could potentially account for the must be invoked in order to explain the peridotite δ 57 Fe arrays. In
light olivine δ 57 Fe values observed in 77SL-405, which also has an this scenario, a long-lived depleted source with radiogenic ε Hf and
elevated bulk MgO content relative to the other SLC peridotites. light δ 57 Fe would be required to account for the light δ 57 Fe val-
 H.M. Williams, M. Bizimis / Earth and Planetary Science Letters 404 (2014) 396–407 403

Fig. 3. a) Bulk sample δ 57 Fe versus Cr# (molar Cr/Cr+Al) SLC: Salt Lake Crater. R2 values for peridotite correlations: 0.90 (all); 0.98 (SLC). Bulk sample δ 57 Fe and elemental
abundances were calculated using mineral modes, Fe contents and isotopic compositions in Table 1. Bulk δ 57 Fe errors were calculated using standard propagation techniques
and given 2 S.D. errors. As detailed in the main text, the long-term reproducibility is 0.07h for the in-house standard and is shown as an error bar in all panels. b) Bulk
sample δ 57 Fe versus clinopyroxene Al2 O3 (wt%). Peridotite array R2 values: 0.84 (all); 0.84 (SLC). c) Bulk sample δ 57 Fe versus % partial melt extraction (Hellebrand et al.,
2001). R2 values: 0.81 (all); 0.98 (SLC). d) Bulk sample δ 57 Fe versus clinopyroxene hafnium isotope compositions. e) Bulk sample δ 57 Fe versus TiO2 (wt%). Model curve shows
the evolution of cumulate residues in equilibrium with fractional melts generated from an initial melt with δ 57 Fe = 0.15h and TiO2 = 1.96 wt%. Tick marks show % melt
removed. The model uses a bulk garnet pyroxenite Ti partition coefficient of 0.31 (Johnson, 1998) and was fitted to the data by adjusting the δ 57 Fe of the initial melt and
α melt-residue . The optimal value of α melt-residue was 1.00007 (0.07h). The calculated TiO2 content of the initial melt is 1.96 wt%, assuming equilibrium with the most primitive
pyroxenite (Bizimis et al., 2005; Sen et al., 2011), and is similar to TiO2 contents displayed by Hawaiian OIB, which may reflect the small amounts of altered oceanic crust
in their source regions (Prytulak and Elliott, 2007). The δ 57 Fe value of this primitive melt is estimated to be 0.15h, within error of mean OIB, MORB and oceanic crust. The
large degree of fractional crystallization required to generate the full range of pyroxenite TiO2 contents could also be indicative of source TiO2 heterogeneity, but this does
not affect the overall conclusions drawn here.
404 H.M. Williams, M. Bizimis / Earth and Planetary Science Letters 404 (2014) 396–407

Fig. 4. Mineral–mineral Fe isotope plots for clinopyroxene vs olivine (a), orthopyroxene vs olivine (b), clinopyroxene vs orthopyroxene (c) and clinopyroxene vs garnet (d).
Symbols are as in Fig. 3. The long-term reproducibility for δ 57 Fe is shown as an error cross on all plots. Dotted curves show a reference line with a slope of 1. The linear
regression shown in (b) was calculated using standard least-squares fitting procedures; errors are 95% confidence.

ues and radiogenic ε Hf of the most depleted samples. If 77SL-405 (FeO = 8.10 wt%; δ 57 Fe ∼ 0.15h, assumed to be similar to MORB
is excluded from the data array and the sample 77SL-341 (bulk or OIB) and PM (FeO = 8.07 wt%; δ 57 Fe ∼ 0.04h), as unrealistic
δ 57 Fe = −0.19 ± 0.12h; ε Hf = 24.2) taken as the most depleted melt–rock ratios (92:8) are needed. A similar argument applies to
endmember, the δ 57 Fe value of this source is required to be ∼ mixtures of mantle peridotite and altered oceanic crust or oceanic
−0.15h (from the models in Section 1.2, allowing for a source- sediments as these have δ 57 Fe values within error of mean MORB
residue fractionation of 0.04h at 30% melt extraction), extremely (Rouxel et al., 2003), or, in the case of ferromanganese sediments,
similar to that inferred for the Gorgona komatiites (Hibbert et al., extremely light δ 57 Fe values (<−0.6h) (Levasseur et al., 2004).
2012). An additional means of generating mantle source components with
Because the models in Section 1.2 demonstrate that single-stage heavy δ 57 Fe values is therefore required.
melt extraction from a BSE with an assumed δ 57 Fe value of 0.04h
cannot create strongly fractionated residues, this source region
4.2. Iron isotopes in the Oahu pyroxenite xenoliths and the formation of
must have acquired this low δ 57 Fe value either through multiple
enriched mantle lithologies
phases of melt extraction or a combination of processes including
melt–rock reaction and melt percolation. Experimental studies of
Fe isotope fractionation during partial melting, in both spinel and The pyroxenites display δ 57 Fe values up to 0.27 ± 0.05h and
garnet facies, coupled with experiments simulating the effects of may represent an analogue for the heavy δ 57 Fe mantle source
percolation and melt–rock reaction processes are required to test components discussed above. Positive correlations exist between
and refine these hypotheses. pyroxenite δ 57 Fe and TiO2 at the bulk rock and mineral level
At the high-δ 57 Fe end of the peridotite array, the peridotite (Fig. 3), which likely reflect isotopic fractionation and the concen-
KAPS-36 (δ 57 Fe 0.14 ± 0.08h; Table 1) has a relatively low spinel tration of heavy δ 57 Fe in evolved melts (Schuessler et al., 2009;
Cr#, low bulk Mg# and slightly LREE enriched clinopyroxene Teng et al., 2008) and hence in the cumulates derived from these
chondrite-normalized REE patterns (Bizimis et al., 2004), all of melts. A model for fractional crystallization and the evolution of
which are indicative of derivation, via minimal levels of melt ex- cumulate residues is shown in Fig. 3; the δ 57 Fe value of the initial
traction, from relatively fertile mantle. The heavy δ 57 Fe value of melt is estimated to be 0.15h, within error of mean OIB, MORB
the peridotite KAPS-36, which is similar to mean MORB or OIB, and altered oceanic crust. The calculated TiO2 content of the initial
indicates derivation from a source with a similar δ 57 Fe value, as melt is 1.96 wt%, assuming equilibrium with the most primitive
the models discussed in Section 1.2 show that the δ 57 Fe values pyroxenite (Bizimis et al., 2005; Sen et al., 2011), similar to the
of melting residues at low degrees of melting are almost indistin- TiO2 contents displayed by alkali rejuvenated stage Hawaiian lavas
guishable from the initial source. However, a source region with a (Clague and Frey, 1982; Garcia et al., 2010), the presumed parental
δ 57 Fe value of ∼0.14h cannot be created by mixing mantle melts melt of these pyroxenites (Bizimis et al., 2013).
 H.M. Williams, M. Bizimis / Earth and Planetary Science Letters 404 (2014) 396–407 405

Pyroxenite cumulate residues (as well as evolved melts) gen- lithology), OIB with heavier δ 57 Fe values of ∼0.22h can be ac-
erated by high-pressure basaltic magma fractionation therefore counted for by a 30:70 mixture of 10% lherzolite melt and 20%
appear to possess heavy δ 57 Fe signatures and high incompatible garnet pyroxenite melt, or, if it assumed that such OIB represent
element concentrations. Therefore, pyroxenite cumulate compo- smaller overall melting degrees, by a 40:60 mixture of 5% lher-
nents, irrespective of their origin, could contribute to the level zolite melt and 10% garnet pyroxenite melt. In these calculations,
of Fe-isotope heterogeneity observed in MORB and OIB source re- the FeO contents of both endmembers are assumed to be equal;
gions. For example, a source region with a δ 57 Fe value of ∼0.14h, given the tendency of pyroxenites to have generally higher FeO
such as that required to generate the KAPS-36 peridotite, could contents relative to depleted peridotitic components this means
be formed through a 55:45 mixture of cumulate pyroxenite melt that if anything, the amount of pyroxenite melt required is slightly
to unmelted mantle peridotite (60% cumulate melt as modeled overestimated. Hence, the presence of pyroxenite in mantle source
in Fig. 3, δ 57 Fe = 0.21h, peridotite δ 57 Fe = 0.04h, for simplicity regions is a viable mechanism for explaining the heavy δ 57 Fe val-
both peridotite and pyroxenite melt are assumed to have the same ues of the Society and Cook-Austral OIB, and potentially places
FeO content of 8 wt%). This scenario is broadly consistent with the direct constraints on the mineralogy of the HIMU and EM2 mantle
relatively fertile nature of the KAPS-36 peridotite, where the strong components themselves.
nature of LREE enrichment that such metasomatism would gen-
erate is then moderated by subsequent episode(s) of minor melt
5. Conclusions
extraction.

4.3. Evidence for Fe-isotope heterogeneity in the mantle source regions Iron stable isotopes are a powerful new tracer for detect-
of MORB and OIB ing mineralogically distinct components in the source regions of
oceanic basalts. Depleted peridotite and pyroxenite lithologies dis-
Multiple mantle source regions with distinct δ 57 Fe are also play contrasting Fe isotope signatures that correlate with elemental
required to explain the variation in MORB and OIB δ 57 Fe. Con- and radiogenic isotope tracers of magmatic processes. The light
siderable variability exists in the Fe isotope composition of fresh Fe isotope compositions of the peridotites are considered to arise
MORB and OIB (Fig. 2), which is not readily explained by magmatic through melt extraction, whereas the heavy δ 57 Fe signatures of the
differentiation or post-magmatic processes. For example, at a com- pyroxenites reflect fractional crystallization processes near the base
paratively narrow range of MgO (6.4–11.2 wt%) MORB δ 57 Fe ranges of the oceanic lithosphere. The absence of correlations between Fe
from 0.09 to 0.29h, a range far greater than that predicted by any isotopes and indices of metasomatism provides evidence that Fe
melting model. At a similar range of MgO contents (5.5–13.6 wt%), isotopes are highly resistant to metasomatic overprinting.
OIB δ 57 Fe varies from −0.05 to 0.36h. If Hawaiian OIB are con- Melting models cannot explain the level of Fe-isotope variation
sidered alone, the extent of δ 57 Fe variation is marginally reduced observed in MORB and OIB. We conclude that the source regions
(δ 57 Fe ranges from 0.07 to 0.31h) but is still greater than the of both MORB and OIB are heterogeneous in terms of mineralogy
extent of isotopic variation predicted by the melting models in Sec- and Fe isotopes. The heavy δ 57 Fe values of OIB suites such as So-
tion 1.2 (∼0.03h for 2–20% partial melting). ciety and Cook-Austral are indicative of pyroxenite components in
Interestingly, OIB from the Cook-Austral and Society islands, their mantle source, while the low δ 57 Fe values of the most de-
which represent type examples of HIMU (high-μ, high 238 U/204 Pb; pleted peridotites provide strong new evidence for the presence of
206
Pb/204 Pb > 20.5, 87 Sr/86 Sr < 0.703) (Stracke et al., 2005) and highly depleted domains in the mantle. Because the latter material
EM2 OIB (enriched mantle; high 87 Sr/86 Sr) (Chauvel et al., 1992), is so refractory, it is not widely sampled in most basalts (Salters
respectively, have particularly heavy mean δ 57 Fe values (0.21 ± et al., 2011), with the exception of those generated by exception-
0.05h 2 S.D., n = 10, MgO 6.3–10.1 wt% and 0.19 ± 0.02h 2 ally large degrees of partial melting, such as komatiites (Hibbert
S.D., n = 2, respectively, MgO 5.5–7.3 wt%) relative to other OIB. et al., 2012) and boninites (Dauphas et al., 2009), which display
Student’s t-tests demonstrate that this difference in mean δ 57 Fe light δ 57 Fe values relative to MORB and OIB. Iron isotope studies
between Society and Hawaiian OIB (0.15 ± 0.09h 2 S.D., n = 34, of large-degree partial melts and depleted peridotites may thus re-
MgO 5.8–13.6 wt%) is resolvable at the 95% confidence level, even veal the melt depletion and enrichment history of the convecting
when samples from Ko’olau, which define a subtlety lighter mean upper mantle.
δ 57 Fe (0.14 ± 0.07h 2 S.D., n = 13, MgO 6.3–10.1 wt%) than other
Hawaiian OIB (Teng et al., 2013) are omitted from the Hawaiian Acknowledgements
dataset (Hawaii without Ko’olau: 0.17 ± 0.09h 2 S.D., n = 18, re-
spectively, MgO 5.8–13.6 wt%). While the radiogenic Pb and Sr
The authors would like to thank Joel Baker, Kevin Burton,
isotope compositions of EM2 and HIMU have been linked with re-
Karsten Haase, Ian Parkinson, Frank Poitrasson and Pete Tollan
cycled components such as subduction modified, low-Pb oceanic
for constructive discussions. Fang-Zhen Teng is acknowledged for
crust (HIMU) and terrigenous sediments (EM2) (Hofmann, 1997),
helpful discussions and providing access to new OIB Fe isotope
these components do not have Fe isotope compositions signifi-
data prior to its publication. Geoff Nowell is thanked for his in-
cantly different to that of mean MORB (Rouxel et al., 2003) and
valuable and dedicated technical support. Stefan Weyer and an
therefore cannot explain the heavy δ 57 Fe values of Society and
anonymous reviewer are thanked for their constructive and help-
Cook-Austral OIB. However, our data suggest that the process of
ful reviews and Tim Elliott, as editor, for his numerous sugges-
pyroxenite cumulate formation from fractionated mafic melts (ir-
tions, which truly helped us improve this paper. Helen Williams
respective of their origin) can generate enriched mantle mineralo-
is funded by NERC (Advanced Fellowship NE/F014295/1) and the
gies and associated melts with heavy δ 57 Fe values and our models
European Research Council (ERC Starting Grant 306655 “Habitable-
demonstrate that pyroxenite source lithologies will generate melts
Planet”). Michael Bizimis is funded by NSF (NSF-OCE 1129280).
with heavier δ 57 Fe values than those from olivine-bearing sources
under similar conditions. For example, oceanic basalts with δ 57 Fe
values of ∼0.18h can be accounted for by a 70:30 mixture of Appendix A. Supplementary material
10% lherzolite melt and 20% garnet pyroxenite melt (both mod-
eled in Section 1.2; a higher melting degree is assumed for the Supplementary material related to this article can be found on-
pyroxenite component given the greater melt productivity of that line at http://dx.doi.org/10.1016/j.epsl.2014.07.033.
406 H.M. Williams, M. Bizimis / Earth and Planetary Science Letters 404 (2014) 396–407

References Hofmann, A.W., White, W.M., 1982. Mantle plumes from ancient oceanic-crust. Earth
Planet. Sci. Lett. 57, 421–436.
Allègre, C.J., Turcotte, D.L., 1986. Implications of a two component marble-cake man- Huang, S., Frey, F.A., 2005. Recycled oceanic crust in the Hawaiian Plume: evidence
tle. Nature 323, 123–127. from temporal geochemical variations within the Koolau Shield. Contrib. Min-
Asael, D., Tissot, F.L.H., Reinhard, C.T., Rouxel, O., Dauphas, N., Lyons, T.W., Ponzevera, eral. Petrol. 149, 556–575.
E., Liorzou, C., Chéron, S., 2013. Coupled molybdenum, iron and uranium sta- Humayun, M., Qin, L., Norman, M.D., 2004. Geochemical evidence for excess iron in
ble isotopes as oceanic paleoredox proxies during the Paleoproterozoic Shunga the mantle beneath Hawaii. Science 306, 91–94.
Event. Chem. Geol. 362, 193–210. Ionov, D.A., Hofmann, A.W., 2007. Depth of formation of subcontinental off-craton
Beard, B.L., Johnson, C.M., 2004. Inter-mineral Fe isotope variations in mantle- peridotites. Earth Planet. Sci. Lett. 261, 620–634.
derived rocks and implications for the Fe geochemical cycle. Geochim. Cos- Jackson, M.G., Dasgupta, R., 2008. Compositions of HIMU, EM1, and EM2 from global
mochim. Acta 68, 4727–4743. trends between radiogenic isotopes and major elements in ocean island basalts.
Beard, B.L., Johnson, C.M., Skulan, J.L., Nealson, K.H., Cox, L., Sun, H., 2003. Applica- Earth Planet. Sci. Lett. 276, 175–186.
tion of Fe isotopes to tracing the geochemical and biological cycling of Fe. Chem. Johnson, K.T., 1998. Experimental determination of partition coefficients for rare
Geol. 195, 87–117. earth and high-field-strength elements between clinopyroxene, garnet, and
basaltic melt at high pressures. Contrib. Mineral. Petrol. 133, 60–68.
Bizimis, M., Sen, G., Salters, V.J.M., 2004. Hf–Nd isotope decoupling in the oceanic
Keshav, S., Sen, G., 2001. Majoritic garnets in Hawaiian xenoliths: preliminary re-
lithosphere: constraints from spinel peridotites from Oahu, Hawaii. Earth Planet.
sults. Geophys. Res. Lett. 28, 3509–3512.
Sci. Lett. 217, 43–58.
Keshav, S., Gudfinnsson, G.H., Sen, G., Fei, Y., 2004. High-pressure melting experi-
Bizimis, M., Sen, G., Salters, V.J.M., Keshav, S., 2005. Hf–Nd–Sr isotope systematics of
ments on garnet clinopyroxenite and the alkalic to tholeiitic transition in ocean-
garnet pyroxenites from Salt Lake Crater, Oahu, Hawaii: evidence for a depleted
island basalts. Earth Planet. Sci. Lett. 223, 365–379.
component in Hawaiian volcanism. Geochim. Cosmochim. Acta 69, 2629–2646.
Keshav, S., Sen, G., Presnall, D.C., 2007. Garnet-bearing xenoliths from Salt Lake
Bizimis, M., Griselin, M., Lassiter, J.C., Salters, V.J.M., Sen, G., 2007. Ancient recycled
crater, Oahu, Hawaii; high-pressure fractional crystallization in the oceanic man-
mantle lithosphere in the Hawaiian plume: osmium–hafnium isotopic evidence
tle. J. Petrol. 48, 1681–1724.
from peridotite mantle xenoliths. Earth Planet. Sci. Lett. 257, 259–273.
Klein, E.M., Langmuir, C.H., 1987. Global correlations of ocean ridge basalt chem-
Bizimis, M., Salters, V.J., Garcia, M.O., Norman, M.D., 2013. The composition and
istry with axial depth and crustal thickness. J. Geophys. Res., Solid Earth 92,
distribution of the rejuvenated component across the Hawaiian plume: Hf–Nd–
8089–8115.
Sr–Pb isotope systematics of Kaula lavas and pyroxenite xenoliths. Geochem.
Kogiso, T., Hirschmann, M.M., 2006. Partial melting experiments of bimineralic
Geophys. Geosyst. 14, 4458–4478.
eclogite and the role of recycled mafic oceanic crust in the genesis of ocean
Burton, K.W., Cenki-Tok, B., Mokadem, F., Harvey, J., Gannoun, A., Alard, O., Parkin-
island basalts. Earth Planet. Sci. Lett. 249, 188–199.
son, I.J., 2012. Unradiogenic lead in Earth’s upper mantle. Nat. Geosci. 5,
Kogiso, T., Hirschmann, M.M., Frost, D.J., 2003. High-pressure partial melting of gar-
570–573.
net pyroxenite: possible mafic lithologies in the source of ocean island basalts.
Chauvel, C., Hofmann, A.W., Vidal, P., 1992. HIMU EM – the French–Polynesian Con-
Earth Planet. Sci. Lett. 216, 603–617.
nection. Earth Planet. Sci. Lett. 110, 99–119.
Lassiter, J.C., Hauri, E.H., 1998. Osmium-isotope variations in Hawaiian lavas: evi-
Clague, D.A., Frey, F.A., 1982. Petrology and trace-element geochemistry of the Hon-
dence for recycled oceanic lithosphere in the Hawaiian plume. Earth Planet. Sci.
olulu Volcanics, Oahu – implications for the Oceanic Mantle Below Hawaii. J.
Lett. 164, 483–496.
Petrol. 23, 447–504.
Lazarov, M., Brey, G.P., Weyer, S., 2012. Evolution of the South African mantle—a case
Craddock, P.R., Warren, J.M., Dauphas, N., 2013. Abyssal peridotites reveal the near-
study of garnet peridotites from the Finsch diamond mine (Kaapvaal craton);
chondritic Fe isotopic composition of the Earth. Earth Planet. Sci. Lett. 365,
Part 2: multiple depletion and re-enrichment processes. Lithos 154, 210–223.
63–76.
Levasseur, S., Frank, M., Hein, J.R., Halliday, A.N., 2004. The global variation in the
Dasgupta, R., Jackson, M.G., Lee, C.-T.A., 2010. Major element chemistry of ocean
iron isotope composition of marine hydrogenetic ferromanganese deposits: im-
island basalts: conditions of mantle melting and heterogeneity of mantle source.
plications for seawater chemistry? Earth Planet. Sci. Lett. 224, 91–105.
Earth Planet. Sci. Lett. 289, 377–392. Li, X., Kind, R., Yuan, X., Wolbern, I., Hanka, W., 2004. Rejuvenation of the litho-
Dauphas, N., Craddock, P.R., Asimow, P.D., Bennett, V.C., Nutman, A.P., Ohnenstetter, sphere by the Hawaiian plume. Nature 427, 827–829.
D., 2009. Iron isotopes may reveal the redox conditions of mantle melting from Liu, C.-Z., Snow, J.E., Hellebrand, E., Bruegmann, G., von der Handt, A., Buechl, A.,
Archean to Present. Earth Planet. Sci. Lett. 288, 255–267. Hofmann, A.W., 2008. Ancient, highly heterogeneous mantle beneath Gakkel
Downes, H., 2007. Origin and significance of spinel and garnet pyroxenites in the ridge, Arctic Ocean. Nature 452, 311–316.
shallow lithospheric mantle: ultramafic massifs in orogenic belts in Western Eu- Lundstrom, C., Sampson, D., Perfit, M., Gill, J., Williams, Q., 1999. Insights into
rope and NW Africa. Lithos 99, 1–24. mid-ocean ridge basalt petrogenesis: U-series disequilibria from the Siqueiros
Elliott, T., Blichert-Toft, J., Heumann, A., Koetsier, G., Forjaz, V., 2007. The origin of Transform, Lamont Seamounts, and East Pacific Rise. J. Geophys. Res. 104,
enriched mantle beneath Sao Miguel, Azores. Geochim. Cosmochim. Acta 71, 13035–13048.
219–240. McDonough, W.F., Sun, S.-S., 1995. The composition of the Earth. Chem. Geol. 120,
Gagnevin, D., Boyce, A., Barrie, C., Menuge, J., Blakeman, R., 2012. Zn, Fe and S 223–253.
isotope fractionation in a large hydrothermal system. Geochim. Cosmochim. Millet, M.-A., Baker, J.A., Payne, C.E., 2012. Ultra-precise stable Fe isotope measure-
Acta 88, 183–198. ments by high resolution multiple-collector inductively coupled plasma mass
Garcia, M.O., Swinnard, L., Weis, D., Greene, A.R., Tagami, T., Sano, H., Gandy, C.E., spectrometry with a 57 Fe–58 Fe double spike. Chem. Geol. 304, 18–25.
2010. Petrology, Geochemistry and Geochronology of Kaua’i Lavas over 4.5 Myr: Niu, Y.L., O’Hara, M.J., 2003. Origin of ocean island basalts: a new perspective from
implications for the Origin of Rejuvenated Volcanism and the Evolution of the petrology, geochemistry, and mineral physics considerations. J. Geophys. Res.,
Hawaiian Plume. J. Petrol. 51, 1507–1540. Solid Earth 108.
Gurenko, A.A., Sobolev, A.V., Hoernle, K.A., Hauff, F., Schmincke, H.-U., 2009. En- Niu, Y., Collerson, K.D., Batiza, R., Wendt, J.I., Regelous, M., 1999. Origin of enriched,
riched, HIMU-type peridotite and depleted recycled pyroxenite in the Canary E-type mid-ocean ridge basalt at ridges far from mantle plumes: The East Pacific
plume: a mixed-up mantle. Earth Planet. Sci. Lett. 277, 514–524. Rise at 11◦ 20 N. J. Geophys. Res. 104, 7067–7087.
Hauri, E.H., 1996. Major-element variability in the Hawaiian mantle plume. Na- Ozawa, A., Tagami, T., Garcia, M.O., 2005. Unspiked K–Ar dating of the Honolulu
ture 382, 415–419. rejuvenated and Ko’olau shield volcanism on O’ahu, Hawai’i. Earth Planet. Sci.
Hauri, E.H., Lassiter, J.C., DePaolo, D.J., 1996. Osmium isotope systematics of drilled Lett. 232, 1–11.
lavas from Mauna Loa, Hawaii. J. Geophys. Res., Solid Earth 1978–2012 (101), Parman, S.W., Grove, T.L., 2004. Harzburgite melting with and without H2 O: experi-
11793–11806. mental data and predictive modeling. J. Geophys. Res. 109, B02201.
Heimann, A., Beard, B.L., Johnson, C.M., 2008. The role of volatile exsolution and sub- Pearce, J.A., Parkinson, I.J., 1993. Trace element models for mantle melting: appli-
solidus fluid/rock interactions in producing high Fe-56/Fe-54 ratios in siliceous cation to volcanic arc petrogenesis. In: Magmatic Processes and Plate Tectonics.
igneous rocks. Geochim. Cosmochim. Acta 72, 4379–4396. Geol. Soc. (Lond.) Spec. Publ. 76, 373–403.
Hellebrand, E., Snow, J.E., Dick, H.J.B., Hofmann, A.W., 2001. Coupled major and trace Pertermann, M., Hirschmann, M.M., 2003. Partial melting experiments on a MORB-
elements as indicators of the extent of melting in mid-ocean-ridge peridotites. like pyroxenite between 2 and 3 GPa: constraints on the presence of pyroxenite
Nature 410, 677–681. in basalt source regions from solidus location and melting rate. J. Geophys. Res.,
Herzberg, C., 2004. Partial crystallization of mid-ocean ridge basalts in the crust and Solid Earth 108.
mantle. J. Petrol. 45, 2389–2405. Pietruszka, A.J., Garcia, M.O., 1999. A rapid fluctuation in the mantle source and
Hibbert, K.E.J., Williams, H.M., Kerr, A.C., Puchtel, I.S., 2012. Iron isotopes in an- melting history of Kilauea Volcano inferred from the geochemistry of its histor-
cient and modern komatiites: evidence in support of an oxidised mantle from ical summit lavas. J. Petrol. 40, 1321–1342.
Archean to present. Earth Planet. Sci. Lett. 321, 198–207. Pilet, S., Baker, M.B., Stolper, E.M., 2008. Metasomatized lithosphere and the origin
Hirschmann, M.M., Stolper, E.M., 1996. A possible role for garnet pyroxenite in the of alkaline lavas. Science 320, 916–919.
origin of the “garnet signature” in MORB. Contrib. Mineral. Petrol. 124, 185–208. Poitrasson, F., Halliday, A.N., Lee, D.C., Levasseur, S., Teutsch, N., 2004. Iron isotope
Hofmann, A.W., 1997. Mantle geochemistry: the message from oceanic volcanism. differences between Earth, Moon, Mars and Vesta as possible records of con-
Nature 385, 219–229. trasted accretion mechanisms. Earth Planet. Sci. Lett. 223, 253–266.
 H.M. Williams, M. Bizimis / Earth and Planetary Science Letters 404 (2014) 396–407 407

Polyakov, V.B., Mineev, S.D., 2000. The use of Mössbauer spectroscopy in stable iso- Stracke, A., Salters, V.J.M., Sims, K.W.W., 1999. Assessing the presence of pyroxenite
tope geochemistry. Geochim. Cosmochim. Acta 64, 849–865. in the source of Hawaiian basalts: hafnium–neodymium–thorium isotope evi-
Polyakov, V.B., Clayton, R.N., Horita, J., Mineev, S.D., 2007. Equilibrium iron isotope dence. Geochem. Geophys. Geosyst. 1999GC000013.
fractionation factors of minerals: reevaluation from the data of nuclear inelastic Stracke, A., Hofmann, A.W., Hart, S.R., 2005. FOZO, HIMU, and the rest of the mantle
resonant X-ray scattering and Mossbauer spectroscopy. Geochim. Cosmochim. zoo. Geochem. Geophys. Geosyst. 6.
Acta 71, 3833–3846. Stracke, A., Snow, J.E., Hellebrand, E., von der Handt, A., Bourdon, B., Birbaum, K.,
Prinzhofer, A., Lewin, E., Allegre, C., 1989. Stochastic melting of the marble cake Guenther, D., 2011. Abyssal peridotite Hf isotopes identify extreme mantle de-
mantle: evidence from local study of the East Pacific Rise at 12◦ 50 N. Earth pletion. Earth Planet. Sci. Lett. 308, 359–368.
Planet. Sci. Lett. 92, 189–206. Teng, F.Z., Dauphas, N., Helz, R.T., 2008. Iron isotope fractionation during magmatic
Prytulak, J., Elliott, T., 2007. TiO2 enrichment in ocean island basalts. Earth Planet. differentiation in Kilauea Iki Lava Lake. Science 320, 1620–1622.
Sci. Lett. 263, 388–403. Teng, F.Z., Dauphas, N., Huang, S., Marty, B., 2013. Iron isotopic systematics of
Ren, Z.-Y., Shibata, T., Yoshikawa, M., Johnson, K.T., Takahashi, E., 2006. Isotope oceanic basalts. Geochim. Cosmochim. Acta 107, 12–26.
compositions of submarine Hana Ridge lavas, Haleakala volcano, Hawaii: im- Vlastelic, I., Aslanian, D., Dosso, L., Bougault, H., Olivet, J., Geli, L., 1999. Large-scale
plications for source compositions, melting process and the structure of the chemical and thermal division of the Pacific mantle. Nature 399, 345–350.
Hawaiian plume. J. Petrol. 47, 255–275. Vlastelic, I., Lewin, E., Staudacher, T., 2006. Th/U and other geochemical evidence for
Robinson, J., Wood, B., Blundy, J., 1998. The beginning of melting of fertile and de- the Reunion plume sampling a less differentiated mantle domain. Earth Planet.
pleted peridotite at 1.5 GPa. Earth Planet. Sci. Lett. 155, 97–111. Sci. Lett. 248, 379–393.
Rouxel, O., Dobbek, N., Ludden, J., Fouquet, Y., 2003. Iron isotope fractionation during Weyer, S., Ionov, D.A., 2007. Partial melting and melt percolation in the mantle: the
oceanic crust alteration. Chem. Geol. 202, 155–182. message from Fe isotopes. Earth Planet. Sci. Lett. 259, 119–133.
Salters, V.J., Stracke, A., 2004. Composition of the depleted mantle. Geochem. Geo-
Weyer, S., Anbar, A.D., Brey, G.P., Munker, C., Mezger, K., Woodland, A.B., 2005. Iron
phys. Geosyst. 5.
isotope fractionation during planetary differentiation. Earth Planet. Sci. Lett. 240,
Salters, V.J., Zindler, A., 1995. Extreme 176 Hf/177 Hf in the sub-oceanic mantle. Earth
251–264.
Planet. Sci. Lett. 129, 13–30.
White, W.M., Hofmann, A.W., 1982. Mantle heterogeneity and isotopes in oceanic
Salters, V.J.M., Mallick, S., Hart, S.R., Langmuir, C.E., Stracke, A., 2011. Domains of de-
basalts. Nature 295, 363–364.
pleted mantle: new evidence from hafnium and neodymium isotopes. Geochem.
Williams, H.M., McCammon, C.A., Peslier, A.H., Halliday, A.N., Teutsch, N., Levasseur,
Geophys. Geosyst. 12.
S., Burg, J.P., 2004. Iron isotope fractionation and the oxygen fugacity of the
Schoenberg, R., von Blanckenburg, F., 2006. Modes of planetary-scale Fe isotope frac-
mantle. Science 304, 1656–1659.
tionation. Earth Planet. Sci. Lett. 252, 342–359.
Williams, H.M., Peslier, A.H., McCammon, C., Halliday, A.N., Levasseur, S., Teutsch, N.,
Schuessler, J.A., Schoenberg, R., Sigmarsson, O., 2009. Iron and lithium isotope sys-
Burg, J.P., 2005. Systematic iron isotope variations in mantle rocks and minerals:
tematics of the Hekla volcano, Iceland – evidence for Fe isotope fractionation
the effects of partial melting and oxygen fugacity. Earth Planet. Sci. Lett. 235,
during magma differentiation. Chem. Geol. 258, 78–91.
435–452.
Sen, G., Frey, F.A., Shimizu, N., Leeman, W.P., 1993. Evolution of the lithosphere be-
neath Oahu, Hawaii – rare-earth element abundances in mantle xenoliths. Earth Williams, H.M., Nielsen, S.G., Renac, C., Griffin, W.L., O’Reilly, S.Y., McCammon, C.A.,
Planet. Sci. Lett. 119, 53–69. Pearson, N., Viljoen, F., Alt, J.C., Halliday, A.N., 2009. Fractionation of oxygen and
Sen, G., Keshav, S., Bizimis, M., 2005. Hawaiian mantle xenoliths and magmas; com- iron isotopes by partial melting processes: implications for the interpretation of
position and thermal character of the lithosphere. Am. Mineral. 90, 871–887. stable isotope signatures in mafic rocks. Earth Planet. Sci. Lett. 283, 156–166.
Sen, I.S., Bizimis, M., Sen, G., 2010. Geochemistry of sulfides in Hawaiian garnet Williams, H.M., Wood, B.J., Wade, J., Frost, D., Tuff, J., 2012. Isotopic evidence for
pyroxenite xenoliths: implications for highly siderophile elements in the oceanic internal oxidation of the Earth’s mantle. Earth Planet. Sci. Lett., 321–322.
mantle. Chem. Geol. 273, 180–192. Wirth, R., Rocholl, A., 2003. Nanocrystalline diamond from the Earth’s mantle un-
Sen, I.S., Bizimis, M., Sen, G., Huang, S., 2011. A radiogenic Os component in the derneath Hawaii. Earth Planet. Sci. Lett. 211, 357–369.
oceanic lithosphere? Constraints from Hawaiian pyroxenite xenoliths. Geochim. Woodland, A., Koch, M., 2003. Variation in oxygen fugacity with depth in the upper
Cosmochim. Acta 75, 4899–4916. mantle beneath the Kaapvaal craton, Southern Africa. Earth Planet. Sci. Lett. 214,
Sigmarsson, O., Carn, S., Carracedo, J.C., 1998. Systematics of U-series nuclides in 295–310.
primitive lavas from the 1730–36 eruption on Lanzarote, Canary Islands, and Workman, R.K., Hart, S.R., Jackson, M., Regelous, M., Farley, K.A., Blusztajn, J., Kurz,
implications for the role of garnet pyroxenites during oceanic basalt formations. M., Staudigel, H., 2004. Recycled metasomatized lithosphere as the origin of
Earth Planet. Sci. Lett. 162, 137–151. the enriched mantle II (EM2) end-member: evidence from the Samoan volcanic
Sobolev, A.V., Hofmann, A.W., Sobolev, S.V., Nikogosian, I.K., 2005. An olivine-free chain. Geochem. Geophys. Geosyst. 5.
mantle source of Hawaiian shield basalts. Nature 434, 590–597. Zindler, A., Hart, S., Frey, F., Jakobsson, S., 1979. Nd and Sr isotope ratios and rare
Sobolev, A.V., Hofmann, A.W., Kuzmin, D.V., Yaxley, G.M., Arndt, N.T., Chung, S.L., earth element abundances in Reykjanes Peninsula basalts evidence for mantle
Danyushevsky, L.V., Elliott, T., Frey, F.A., Garcia, M.O., Gurenko, A.A., Kamenetsky, heterogeneity beneath, Iceland. Earth Planet. Sci. Lett. 45, 249–262.
V.S., Kerr, A.C., Krivolutskaya, N.A., Matvienkov, V.V., Nikogosian, I.K., Rocholl, A., Zindler, A., Staudigel, H., Batiza, R., 1984. Isotope and trace element geochemistry
Sigurdsson, I.A., Sushchevskaya, N.M., Teklay, M., 2007. The amount of recycled of young Pacific seamounts: implications for the scale of upper mantle hetero-
crust in sources of mantle-derived melts. Science 316, 412–417. geneity. Earth Planet. Sci. Lett. 70, 175–195.