Isotopic evidence for iron mobility during subduction

Our study focuses on the Piemont zone of the south-western Alps where ophiolitic units were metamorphosed and juxtaposed during alpine subduction and collision in Late Cretaceous to Tertiary (e.g. Tricart, 1984). We have sampled a serpentinite suite from 5 meta-ophiolites that record metamorphic conditions ranging from greenschist to eclogite facies (see Debret et al., 2014 for more detailed about this sample suite). The Chenaillet ophiolite is located in the external Piemont zone, 6km west of Briançon (Fig. DR1). It is a thin tectonic nappe resting upon the Queyras Schistes Lustrés complex (Caby, 1995). It is composed of metagabbroic pods and serpentinites recording greenschist facies conditions (Mevel et al., 1978). The Queyras Schistes Lustrés complex is composed of metamorphosed Mesozoic oceanic sediments, strongly deformed during alpine subduction (Tricart and Schwartz, 2006). The complex encloses boudinaged ophiolite units composed of serpentinites, metagabbros and metabasalts from meter to kilometer size, and records blueschist facies P-T conditions (Schwartz et al., 2013). The Monte Maggiore ophiolite is located in the northern end of the Cap Corse, Northern Corsica (Fig. DR1). It is an ultramafic body of ∼4 km surrounded by eclogitized continental units. The massif is composed of slightly serpentinized peridotites, serpentinites and metagabbros recording blueschist facies conditions (e.g. Vitale-Brovarone et al., 2013). The Monviso meta-ophiolite is located in the extreme East of the Queyras Schistes Lustrés complex (Fig. DR1). This ultramafic body is composed of kilometric lenses of metabasites embedded in highly deformed serpentinites recording eclogite facies P-T conditions (Schwartz et al., 2001). The Lanzo massif is an eclogitized ultramafic body of 150 km2 bounded by the sediments of the Po plain to the East and South, meta-ophiolites and the Schistes Lustrés unit in the West, and by the HP metamorphic continental unit of Sesia in the North (Pelletier and Müntener, 2006).


INTRODUCTION
Magmas erupted at subduction zones display high ratios of oxidized to total iron (Fe 3+ /ƩFe) and are enriched in volatiles relative to other mantle-derived magmas (e.g., Kelley and Cottrell, 2009).It has therefore been suggested that the source region of arc magmas has been modified by oxidized and volatile-rich fluids released from the subducting slab.Possible oxidizing agents include sulfate-and/or carbonate-bearing fluids derived from subducted sediments or serpentinites (Kelley and Cottrell, 2009;Frezzotti et al., 2011;Evans, 2012;Debret et al., 2014aDebret et al., , 2015)), as well as Fe 3+ transferred directly via supercritical fluids and brines (Mungall, 2002;Kelley and Cottrell, 2009), which have been shown to mobilize other trivalent, normally fluid-immobile elements (Kessel et al., 2005).Evidence in support of direct Fe 3+ and/or SO 4 2transfer includes positive correlations between Fe 3+ /ƩFe and trace element ratios indicative of slab-derived fluids (e.g., Ba/La) in arc magma melt inclusions (Kelley and Cottrell, 2009).
The devolatilization of serpentinite within the subducting slab and the fluids released play a fundamental role in mantle wedge redox evolution (Evans, 2012) and arc magma genesis (Spandler and Pirard, 2013).At mid-oceanic ridges or oceanic-continent transition zones, serpentinites are formed by the hydration and oxidation of the oceanic lithosphere, which leads to the replacement of olivine (Ol) and orthopyroxene by Fe 3+ -rich lizardite, which contains as much as 13% water, and magnetite, increasing bulk-rock Fe 3+ /ƩFe (e.g., Andreani et al., 2013;Klein et al., 2014).During subduction, the lizardite (Liz, low-temperature form of serpentine) to antigorite (Atg, high-temperature form of serpentine) transition is accompanied by magnetite dissolution and a decrease in serpentine mineral Fe 3+ /ƩFe (Schwartz et al., 2013;Debret et al., 2014a).At higher pressures and temperatures, antigorite breakdown to secondary olivine results in the loss of H 2 O (e.g., Ulmer and Trommsdorff, 1995) and the growth of other Fe 2+ -rich minerals (Debret et al., 2015).The major consequence of serpentinite prograde metamorphism during subduction is thus a net decrease in bulkrock Fe 3+ /ƩFe.It is, however, unknown whether this reflects (1) the direct loss of Fe 3+ during prograde metamorphism and/or (2) the oxidation and loss of other redox-sensitive species (e.g., SO 4 2-) to the mantle wedge.
To date there has been no direct means of tracing the loss of oxidizing components or Fe from the subducting slab and their transfer to the mantle wedge (Evans, 2012).New tools that can be used to trace and quantify the mobility of Fe during subduction-related devolatilization and metamorphism are thus required.Iron stable isotopes can serve as such a tracer, as significant isotopic variations can only be achieved by either the addition or loss of Fe-bearing components.Theory predicts that stable isotope variations will be driven by changes in oxidation state, coordination, and bonding environment (Polyakov and Mineev, 2000;Fujii et al., 2011), in particular for Fe the presence of Cl -and SO 4 2anions (Hill et al., 2010).Here we explore the mobility of Fe during slab devolatilization with a Fe stable isotope study of subducted serpentinites from Western Alps metaophiolites.We show a striking negative correlation between the Fe isotope compositions (d 56 Fe) of serpentinites and bulk-rock Fe 3+ /ƩFe, providing the first direct evidence for the open-system behavior of iron in serpentinites during subduction-related prograde metamorphism.

SELECTED SAMPLES AND RESULTS
Alpine metaophiolites are interpreted as remnants of oceanic lithosphere formed and serpentinized in magma-poor settings before being metamorphosed and devolatilized at various pressure-temperature (P-T ) conditions during subduction (Fig. 1A; e.g., Hattori and Guillot, 2007;Debret et al., 2013).We selected a set of 29 representative serpentinite samples encompassing the full range of prograde metamorphic grades for this study (Table DR1 and Fig. DR1 in the GSA Data Repository 1 ).These samples have been characterized in previous petrological and geochemical studies (Schwartz et al., 2013;Debret et al., 2013;Lafay et al., 2013).They display the progressive replacement of lizardite by antigorite (Atg/Liz-to Atg-serpentinites; Fig. 1A), and the first stages of serpentinite dehydration at eclogite facies (Ol/Atg-serpentinites).The Atgserpentinites are characterized by low amounts of magnetite relative to Atg/Liz-serpentinites (Debret et al., 2014a) and can display equilibrated hematite-magnetite assemblages (Fig. DR2).The Ol/Atg-serpentinites are generally present as metamorphic veins and shear zones composed of olivine and antigorite, and are interpreted as high-permeability reaction zones where the fluids released during serpentinite devolatilization have been localized (Debret et al., 2013).For reference we also analyzed a suite of Alpine slightly serpentinized peridotites (SSP), unmetamorphosed lizardite (Liz-) serpentinites, and abyssal serpentinites, which are considered to be representative of the presubduction lithospheric hydrated mantle.No retrograde phases (e.g., talc, chrysotile, amphibole) are observed in the studied sample suite, suggesting that the samples are poorly affected by retrograde metamorphism.

DISCUSSION
Although the progressive increase in d 56 Fe during prograde metamorphism is suggestive of open-system behavior of iron during slab devolatilization, several processes must be first taken into account.These are (1) the mobility of Fe and associated isotopic fractionation during the initial ocean-floor serpentinization of lithosphere, and (2) preexisting protolith Fe isotope heterogeneity.The Liz-serpentinites display a broad range in d 56 Fe values, which range from -0.098‰ to +0.047‰ and are similar to those of SSP, which range from -0.069‰ to +0.063‰.Both sets of values overlap with the d 56 Fe values of abyssal serpentinites (-0.094‰ to +0.108‰; Craddock et al., 2013), and may be considered representative of presubduction oceanic lithosphere.It is important that, while ridge-axis serpentinization processes are known to increase Fe 3+ /ƩFe (Evans, 2008), the d 56 Fe values of oxidized and hydrous Liz-serpentinites are similar to those of the comparatively reduced and anhydrous SSP and are uncorrelated with Fe 3+ /ƩFe (Fig. DR4).This suggests that Fe behaves conservatively during oceanfloor serpentinization and that no appreciable Fe stable isotope fractionation takes place during this process.Previous studies have demonstrated that a relationship between bulk d 56 Fe values and rock fertility exists, with pyroxene-rich and less depleted peridotites displaying isotopically heavier compositions (Williams et al., 2005).In Alpine ophiolites, serpentinite protolith fertility can be assessed using fluid-immobile elements poorly affected by serpentinization (e.g., Al 2 O 3 / SiO 2 , Zr/Nb; Bodinier and Godard, 2013).As illustrated in Figure 3, the d 56 Fe values of Alpine Liz-serpentinites and SSP form broad positive arrays with Al 2 O 3 /SiO 2 and Zr/Nb, which provide evidence for some level of source fertility control on protolith d 56 Fe.Similar correlations between the d 56 Fe values of Atg/Liz-or Atg-serpentinites and indices of peridotite fertility are present, but are demonstrably offset to overall heavier d 56 Fe values (Fig. 3).
The comparatively heavy d 56 Fe values of the Atg/Liz-and Atg-serpentinites relative to the SSP and Liz-serpentinites and the correlation between bulk serpentinite d 56 Fe and Fe 3+ / ƩFe (Fig. 2) must therefore relate to the loss or addition of Fe-bearing fluids during prograde metamorphism.Sediment dehydration releases a significant amount of fluids and volatiles that can potentially interact with serpentinites and modify their isotopic signatures during subduction (Deschamps et al., 2010).However, no correlations between d 56 Fe and elemental tracers of sediment-serpentinite interaction (e.g., As, Sb, Cs) are present (Fig. DR5), suggesting that interactions between sediment-derived melts or fluids and serpentinites cannot explain the heavy d 56 Fe values of the Atg/Liz-and Atgserpentinites.Furthermore, sediment melting requires an elevated subduction geotherm (e.g., Bouilhol et al., 2015), which is inconsistent with the inferred P-T conditions of Alpine ophiolites (Fig. 1).The most straightforward means of interpreting the d 56 Fe and Fe 3+ /ƩFe correlation is thus the release of isotopically light Fe-bearing fluid during prograde serpentinite devolatilization.In agreement with this interpretation, the Ol/Atg-serpentinites, which are interpreted as high-permeability reaction zones where fluids released during prograde metamorphism have been concentrated, display isotopically light d 56 Fe values relative to the Atg-serpentinites (Fig. 1) and high concentrations of fluid-mobile elements (Debret et al., 2013).
Iron stable isotope fractionation is usually attributed to redox effects and experimental and theoretical studies that predict that, at  equilibrium, Fe 2+ -bearing phases should be isotopically lighter than Fe 3+ -bearing phases (Polyakov and Mineev, 2000;Hill et al., 2010).This is in apparent contradiction with the inferred loss of a low-d 56 Fe fluid and the inverse correlation between bulk serpentinite d 56 Fe and Fe 3+ /ƩFe.One possible explanation is that the isotopically light composition of the fluids reflects kinetic (disequilibrium) isotope fractionation and the preferential mobility of the lighter isotope.In this case, the inverse correlation between d 56 Fe and Fe 3+ /ƩFe could reflect the direct loss of Fe 3+ during serpentinite prograde metamorphism.However, the mobility of Fe 3+ relative to Fe 2+ in aqueous fluids and brines is known to be low (e.g., Ding and Seyfried, 1992) and unrealistic amounts of Fe 3+ (~3 wt% Fe 2 O 3 ) would need to be lost to explain the observed decrease in Fe 3+ /ƩFe from 0.7 to 0.5 (Fig. DR6).An alternative explanation is that isotopically light Fe is lost in the form of Fe(II)-S or Fe(II)-Cl complexes, as theory predicts that these complexes should show a preference for isotopically light Fe (Hill et al., 2010).In agreement with this scenario, subducted Alpine serpentinites have low S and Cl contents relative to abyssal serpentinites, providing evidence in support of the mobility of these elements in serpentinitederived fluids (e.g., Alt et al., 2012;Debret et al., 2014b).This is confirmed by the presence of Cl-and S-rich polyphase fluid inclusions preserved within metamorphic olivine and garnet from Alpine metaophiolites (e.g., Scambelluri et al., 2015).The release of such fluids during serpentinite devolatilization can have a large impact on element mobility, since many nominally immobile elements are mobile in Cl-rich fluids (e.g., Rapp et al., 2010).
Perhaps most important, the proposed loss of isotopically light Fe associated with SO X species during serpentinite devolatilization may also have a large effect on mantle wedge redox evolution (e.g., in the case of SO 4 2-, 1 mol of S 6+ can oxidize 8 mol of Fe 2+ as SO 4 2-is reduced to S 2-in the mantle wedge).The S budget of abyssal serpentinites is dominantly controlled by sulfides (e.g., Alt and Shanks, 2003).The decrease in serpentinite S concentration with prograde metamorphism therefore suggests that serpentinite dehydration is associated with sulfide breakdown and the release of sulfur (Hattori and Guillot, 2007).The presence of equilibrated hematite-magnetite assemblages (Fig. DR2) in the most depleted, low Fe 3+ /ƩFe and heavy d 56 Fe serpentinites shows that S transport in serpentinite-derived fluids must have taken place at high oxygen fugacity (f O 2 ), close to the hematite-magnetite buffer (Debret et al., 2015;Tumiati et al., 2015).At such P-T-f O 2 conditions, sulfur will be mobilized as SO 4 2-(aq) (aq-aqueous) rather than reduced species such as HS - (aq) (Fig. DR7).This is consistent with recent observations of SO 4 2-in diamond-bearing fluid inclusions from Western Alps metaophiolites (Frezzotti et al., 2011) (1) Given that typical ocean-floor serpentinites and eclogite facies Atg serpentinites have Fe 3+ /ƩFe ratios of ~0.7 and ~0.5, respectively, and both contain ~8 wt% of Fe 2 O 3total (Fig. DR6), Equation 1 predicts that as much as ~800 ppm sulfur and a minor quantity of Fe (~0.1 wt%) will be released in serpentinite-derived fluids.This value, although speculative, is broadly consistent with the observed decrease in sulfur content (Hattori and Guillot, 2007;Alt et al., 2012;Debret et al., 2014b) and the relatively constant Fe 2 O 3total (Fig. DR6) of serpentinites during subduction.This is, however, a minimum estimate.For example, if S was originally present as S -and released as SO -, S 2 O 3 2-, or SO 3 2-, then the amount of S released would be significantly larger for the same amount of Fe loss.Our results therefore suggest that slab serpentinite devolatilization is associated with Fe reduction and the oxidation and mobilization of sulfides, and potentially other reduced species.The migration of these fluids from the slab to the slab-mantle interface or mantle wedge and the reduction of slab fluid SO X to (Fe)S 2-and concomitant oxidation of mantle wedge Fe(II) to Fe(III) may therefore provide an oxidized mantle source region for primary arc magmas (e.g., Kelley and Cottrell, 2009;Fig. 4).

Figure 2 .
Figure 2. Increase in d 56 Fe and Fe 3+ /ΣFe ratio in serpentinites during subduction.Bulk sample iron isotope composition versus Fe 3+ /ΣFe in subducted Alpine Liz-, Atg/Liz-, and Atg-serpentinites are compared to the one of abyssal serpentinites (small white circles).The error crosses are 2 standard deviations; the long-term reproducibility is shown for reference as an error bar in the lower left corner of this plot.The slope of the correlation is -0.92 ± 0.28 and the 95% confidence interval is represented by two lines (mean square of weighted deviates = 3.6).