Chlorine isotope constraints on fluid-rock interactions during subduction and exhumation of the Zermatt-Saas ophiolite

Authors


  • The copyright line for this article was changed on 30 April 2015 after original online publication.

Abstract

[1] Chlorine isotope compositions of high-pressure (∼2.3 GPa) serpentinite, rodingite, and hydrothermally altered oceanic crust (AOC) differ significantly from high- and ultrahigh-pressure (> 3.2 GPa) metasedimentary rocks in the Aosta region, Italy. Texturally early serpentinites, rodingites, and AOC have bulk δ37Cl values indistinguishable from those of modern seafloor analogues (δ37Cl = −1.0 to +1.0‰). In contrast, serpentinites and AOC samples that recrystallized during exhumation have low δ37Cl values (−2.7 to −0.5‰); 37Cl depletion correlates with progressive changes in bulk chemistry. HP/UHP metasediments have low δ37Cl values (median = −2.5‰) that differ statistically from modern marine sediments (median = −0.6‰). Cl in metasedimentary rocks is concentrated in texturally early minerals, indicating modification of seafloor compositions early in the subduction history. The data constrain fluid sources during both subduction and exhumation-related phases of fluid-rock interaction: (1) marine sediments at the top of the downgoing plate likely interacted with isotopically light pore fluids from the accretionary wedge in the early stages of subduction. (2) No pervasive interaction with externally derived fluid occurred during subsequent subduction to the maximum depths of burial. (3) Localized mixing between serpentinites and fluids released by previously isotopically modified metasediments occurred during exhumation in the subduction channel. Most samples, however, preserved protolith signatures during subduction to near-arc depths.

1. Introduction

[2] Arc magma chemistry requires large-scale transfer of fluid-mobile elements from the slab to the mantle wedge, and we currently know much about the time-integrated magnitude of the various chemical fluxes [e.g., Bebout et al., 1993; Brenan et al., 1998; Philippot et al., 1998; Jarrard, 2003; Straub and Layne, 2003; Lassiter, 2004; Scambelluri et al., 2004]. Our understanding of the specific mechanisms that lead to mass transfer at different depths during subduction is less advanced. Channelized fluid flow in veins and shear zones is well documented in subduction complexes [e.g., John et al., 2008; Spandler et al., 2011; Herms et al., 2012; John et al., 2012] and likely plays the major role in slab-to-mantle mass transfer. However, chemical changes adjacent to high-P veins show that infiltration of externally derived fluid is limited to the centimeter to meter scale [e.g., John et al., 2012]. Indeed, low rock permeability at high pressure should prohibit widespread porous flow, leaving most rocks unaffected by interaction with external fluids [e.g., Zack and John, 2007]. Evidence for limited mass transfer of fluid-mobile elements at pressures approaching subarc depths [e.g., Philippot and Selverstone, 1991; Busigny et al., 2003; Bebout et al., 2013] supports this view of fluid flow in channelized conduits that are separated by unaffected rock parcels. However, fluid that feeds the conduits must itself be derived from devolatilization reactions that occur on the grain scale in hydrous and carbonate-bearing parts of the slab. At the scale of a single dehydrating grain, the released fluid will be in equilibrium with surrounding rock, but modification of both fluid and rock occurs once fluid migration begins. The magnitude of the modifications will depend on fluid volume, rock heterogeneity, mechanism of fluid flow, and length scales of fluid transport.

[3] Here we use chlorine stable isotopes, in conjunction with other petrologic and chemical data, to examine the record of grain-scale fluid-rock interactions during subduction. We focus on exhumed high and ultrahigh-pressure (HP/UHP) oceanic rocks, spanning a wide range of slab lithologies, from the Aosta region in Italy. Because our focus is on grain-scale processes, we largely avoided collecting samples from obvious fluid conduits. A few vein and shear zone samples were included, however, for comparison.

[4] Chlorine has two important properties that make its isotopic composition a potentially useful tracer of subduction zone processes: (1) chlorine partitions strongly into aqueous fluids [e.g., Sharp and Draper, 2013] and (2) little to no isotopic fractionation between minerals and/or fluids is thought to occur at temperatures above the realm of diagenesis/low-grade metamorphism [Schauble et al., 2003]. As a result, rocks that are unaffected by interaction with externally derived Cl-bearing fluid during subduction should preserve their premetamorphic chlorine isotope compositions. In contrast, fluid-modified rocks will provide constraints on the source(s) of the infiltrating fluid(s). Heterogeneities in δ37Cl can constrain the spatial scale of metamorphic fluid-rock interaction and help to distinguish between percolative and channelized fluid flow where transient channels may not have left visible traces [e.g., Selverstone and Sharp, 2011].

[5] Known modern inputs into subduction zones are characterized by δ37Cl values ranging from −8 to +2‰ versus SMOC (standard mean ocean chloride). Sedimentary and basement pore fluids range from 0‰ to −8‰ [Ransom et al., 1995; Godon et al., 2004a, 2004b; Bonifacie et al., 2007], whereas solid seafloor reservoirs lie between −2 and +2‰ [Barnes and Sharp, 2006; Sharp et al., 2007; Barnes et al., 2009b; John et al., 2010; Barnes and Cisneros, 2012]. As the incoming plate enters the subduction zone, bending stresses cause normal faulting and downward penetration of seawater ± pore fluids [Ranero et al., 2003; Faccenda et al., 2009]. Infiltration of low-δ37Cl pore fluids derived from the accretionary wedge during faulting may impart an isotopically light signature to portions of the downgoing plate.

[6] Within this framework, we seek to address the following questions. (1) What are the chlorine isotope compositions of the major subduction reservoirs once the downgoing plate reaches HP/UHP conditions? (2) What is the spatial scale of chlorine isotope equilibration? (3) Do chlorine isotopes record evidence of pervasive fluid-rock interaction between different reservoirs at depth? (4) Are chlorine isotopes correlated with other indicators of fluid-assisted mass transfer during subduction? Together, answers to these questions will add to our understanding of the detailed processes by which mass transfer occurs at convergent plate boundaries, as well as the degree to which nominally fluid-mobile elements can be recycled into the postarc mantle.

2. Geologic Setting

2.1. Tectonic Setting

[7] Samples come from the Zermatt-Saas (Z-S) and Combin Zone ophiolites in Aosta province, Italy (Figure S1).1 The Z-S ophiolite represents a large slice of relatively intact Jurassic oceanic lithosphere that was subducted to HP/UHP conditions [Barnicoat et al., 1986; Rubatto et al., 1998; Compagnoni, 2003; Bucher et al., 2005; Angiboust et al., 2009]. It comprises serpentinized mantle peridotites; gabbros, basalts, and rodingites; zones of seafloor hydrothermal activity; manganese nodules and crusts; and other siliceous and calcareous marine metasediments. Eclogite-facies metamorphism occurred during Eocene subduction [e.g., Rubatto et al., 1998; Dal Piaz et al., 2001; Mahlen et al., 2005; Beltrando et al., 2010], with most rocks in the area of this study reaching pressures of 2.1–2.7 GPa [e.g., Barnicoat et al., 1986; Martin et al., 2008; Angiboust et al., 2009; Groppo et al., 2009]. A small sliver of mafic eclogites, Mn-rich metasediments, and calcschists at Lago di Cignana (LdC) preserves coesite [Reinecke, 1991] and diamond [Frezzotti et al., 2011] and provides clear evidence for subduction to even greater depths (P ≥ 3.2 GPa). The Z-S and LdC units provide an ideal target for investigation of fluid-rock interactions in major rock types—serpentinites, altered oceanic crust, and marine sedimentary rocks—characteristic of subduction zones.

[9] The Z-S ophiolite is tectonically overlain by the Combin Zone, which comprises Permo-Mesozoic rocks of continental affinity overlain by turbidites and an ophiolitic sequence that includes basalts, gabbro, serpentinites, and manganiferous cherts [Dal Piaz and Ernst, 1978; Pleuger et al., 2007]. The mafic rocks of the ophiolitic package are mainly epidote blueschists [Ring, 1995; Cartwright and Barnicoat, 2002], but garnet glaucophanites are locally present.

2.2. Rock Types

2.2.1. Serpentinites and Rodingites

[10] Serpentinites were collected in both the Z-S and Combin ophiolites (Figure S1). Within the Z-S zone, most are antigorite serpentinites [Li et al., 2004b] that formed from harzburgites. The protolith bodies were likely exposed on the seafloor in a slow-spreading ridge environment [Fontana et al., 2008], and initial serpentinitization thus may have occurred via interaction both with seawater and with sedimentary pore fluids [e.g., Barnes et al., 2006; Barnes and Sharp, 2006]. Z-S serpentinites and rodingites analyzed by Cartwright and Barnicoat [1999] have δ18O and δD values that are consistent with seafloor hydration, with little evidence for modification during subsequent metamorphism. The serpentinites experienced varying degrees of metamorphic recrystallization, leading to growth of metamorphic olivine ± Ti-clinohumite ± diopside ± chlorite ± amphibole [Li et al., 2004b; Bucher and Li, 2007; Fontana et al., 2008; Rossetti et al., 2009]. Samples from the Combin Zone include serpentinite and tremolite-bearing ophicarbonate breccia.

[11] Rodingitized gabbroic dikes and their selvages (Figures 1a and 1b) were collected at several localities. The rodingites contain grossular-diopside-amphibole ± chlorite ± epidote ± vesuvianite (Table 1); amphibole is tremolite in samples from the Valtournenche and actinolite/barroisite at Servette. Detailed descriptions of rodingites from the region are given by Li et al. [2004a], Ferrando et al. [2010], and Panseri et al. [2008].

Figure 1.

Field photographs of representative rock types and sample sites. (a) Rodingite dike with chloritic selvage in serpentinite. (b) Boudinaged rodingite dike in serpentinite. (c) HP garnet glaucophanite at Servette. Note lawsonite pseudomorph by pen tip. (d) Sulfide-rich talc-chloritoid-garnet schist in HP altered oceanic crust. (e) Phengitic calcmica schist with Mn nodules in UHP Lago di Cignana unit; hiking pole tip for scale. Garnet in nodules contains diamond inclusions. (f) Interlayered calcmica schist, mica schist, and marble in UHP LdC unit.

[12] We characterized serpentinites collected for this study according to the different types defined by Li et al. [2004b] for Z-S serpentinites (see Table 1):

[13] A. Pseudomorphic and mesh-textured serpentinites produced by hydration on the ocean floor. These serpentinites escaped subsequent overprinting, other than replacement of chrysotile and lizardite by antigorite and partial replacement of augite by diopside.

[14] B. Antigorite-magnetite mylonites, locally containing boudins of secondary olivine. Development of a pervasive, texturally early, mylonitic fabric throughout large volumes of the Z-S ophiolite is consistent with prograde metamorphism and shearing during subduction of the ophiolite.

[15] C. Folded and partially recrystallized mylonitic serpentinites related to the early stages of exhumation during Alpine orogenesis.

[16] D. Recrystallized serpentinites containing coarse, interpenetrating antigorite grains ± pseudomorphs after metamorphic olivine, associated with localized zones of late-stage deformation and/or hydrothermal veins. These rocks reflect the latest stages of metamorphism during exhumation.

2.2.2. Altered Oceanic Crust: Servette Mine

[17] Glaucophanites (Figure 1c), talc schists (Figure 1d), and chlorite schists at the Servette Cu-Fe sulfide mine record increasing degrees of seafloor hydrothermal alteration prior to subduction [Martin et al., 2008; Tumiati et al., 2010; Rebay and Powell, 2012]. The mineralized zone is overlain by Mn-rich quartzites and mica schists and is in tectonic contact with serpentinites and rodingites. The petrologic evolution of these rocks is described by Martin et al. [2008] and Rebay and Holland [2012].

[18] Glaucophanites contain glaucophane-garnet-chlorite-talc-rutile ± chloritoid ± paragonite± clinozoisite ± sulfides and locally preserve pseudomorphs after lawsonite (Figure 1c). Talc schists and chlorite schists contain centimeter-sized garnet with two distinct growth zones. Chloritoid porphyroblasts are present in most samples and show variable degrees of replacement by chlorite and talc. Rare garnetites contain >90% garnet surrounded by apatite-talc-chlorite.

2.2.3. Metasedimentary Rocks: HP

[19] High-pressure metasedimentary rocks were collected above the ore body at the Servette mine and from extensive exposures in the upper Valtournenche. Servette samples contain garnet-chloritoid-phengite-paragonite ± chlorite ± glaucophane. Pseudomorphs after lawsonite are prevalent, and partial replacement of glaucophane by chlorite + paragonite is evident in some samples. Valtournenche samples include garnet-bearing (calc)mica schists and piemontite + braunite-bearing manganiferous metacherts. Omphacite, glaucophane, and chloritoid occur in some samples, but the abundance of albite porphyroblasts indicates an extensive reequilibration history following HP metamorphism [e.g., Dal Piaz and Ernst, 1978; Dal Piaz et al., 1979].

2.2.4. Metasedimentary Rocks: UHP

[20] Ultrahigh-pressure schists, siliceous marbles, and Mn nodules and crusts were collected at Lago di Cignana (Figures 1e and 1f). Microdiamond inclusions in garnet were positively identified in sample LdC-1B2 [Frezzotti et al., 2011] and tentatively identified in all other samples from the Mn nodules and crusts.

[21] In contrast to the observations of Reinecke [1998], several of the UHP metasedimentary samples—in particular, the diamondiferous Mn crusts and nodules—show little evidence of retrogression other than the prevalence of quartz rather than coesite. In other samples, albite partially replaces paragonite, chlorite rims garnet, and pseudomorphs after lawsonite ± glaucophane ± omphacite are present. The effect of retrogression on chlorine systematics is discussed in section 4.4.

2.3. Pressure-Temperature Histories

[22] Previously published P-T estimates for the Z-S ophiolite, LdC unit, and Combin Zone are summarized in Figure 2. The P-T path for the Z-S ophiolite constructed by Angiboust et al. [2009] is based on Thermocalc calculations from 27 localities throughout the unit. HP equilibration conditions determined from pseudosection analyses of different rock types at the Servette mine [Martin et al., 2008] are in good agreement with the overall Z-S path, although estimates from Rebay and Holland [2012] are offset to higher temperatures. Li et al. [2004b] used phase relations and textures in serpentinites to construct a P-T path for the Swiss portion of the Z-S ophiolite and to correlate different textural stages with different parts of the subduction/exhumation history. Their path reaches pressures slightly lower than those determined by Angiboust et al. [2009] (2.3 ± 0.1 GPa at 540 ± 20°C) at temperatures in excess of 600°C.

Figure 2.

Pressure-temperature plot showing previously published conditions of equilibration for units and localities in this study. Green letters denote conditions associated with different stages of serpentinite recrystallization delineated by Li et al. [2004b]: A, seafloor to early subduction; B, pressure peak; C, early exhumation; D, greenschist overprint. Shaded fields show calculated conditions of equilibration for representative serpentinite samples (Figure A2). “Peak” serpentinite and Servette conditions are consistent with P-T path of Angiboust et al. [2009]. Lago di Cignana unit records higher Pmax conditions, consistent with the presence of diamond [Frezzotti et al., 2011]. Quartz-coesite boundary from Mirwald and Massonne [1980]; graphite-diamond boundary is “best calorimetric” fit from Day [2012].

[23] We used bulk compositions of several serpentinites to calculate P-T pseudosections (Figure S2) and constrain the conditions at which different serpentinite mineral assemblages developed. We chose samples from the Z-S and Combin units that covered a range of compositions and that exhibited Type B, C, and D equilibration and/or overprinting. Data for three samples (CZ-1D, MA-1, and USS-1) that span the range of calculated conditions are shown in Figure 2. In all cases, fields that match the mineral assemblages, modes, and mineral compositions of the analyzed samples overlap the Angiboust et al. [2009] P-T path rather than the Li et al. [2004b] path. We thus suggest that Li et al. [2004b] overestimated Tmax. The relative conditions at which their textural stages developed remain the same, but Type B and C fabrics likely developed at lower temperatures than in their model. Stage D may reflect reheating at greenschist conditions, as claimed by Li et al. [2004b] but could also reflect partial reequilibration during nearly isothermal decompression of the Z-S ophiolite.

3. Methods

[24] All analytical work was carried out in the Department of Earth & Planetary Sciences at the University of New Mexico. Details of the procedures for XRF, microprobe, and Cl and H isotope analyses and P-T pseudosection construction can be found in Appendix A (supporting information). We emphasize that low chlorine concentrations determined by electron microprobe are approximate (though reproducible) and are used only in a relative sense in discussions of mineral data.

4. Results

4.1. Summary of Chlorine and Hydrogen Isotopic Compositions

[25] All of the chlorine isotope data are shown as a function of rock type in Figure 3. The data are also compared to the ranges of known subduction inputs. The most striking feature of Figure 3 is that serpentinites and AOC subducted to pressures >2 GPa show no significant change in chlorine isotope composition relative to their modern, unmetamorphosed equivalents. Metarodingite samples are indistinguishable from host serpentinites. HP and UHP metasedimentary rocks overlap the range documented for modern marine sediments [Barnes et al., 2008, 2009b] but extend to lower δ37Cl values. No isotopically positive metasedimentary rocks, such as those measured by John et al. [2010], were observed in the Aosta suite.

Figure 3.

Box-and-whisker plot showing δ37Cl values as a function of rock type. Colored boxes represent middle 50% of data for each rock type; horizontal lines within boxes show median values. AOC, altered oceanic crust represented by samples from Servette. Metasedimentary rocks from Z-S ophiolite (HP) and Lago di Cignana slice (UHP). Silicate-bound (SBC) δ37Cl data (bulk values where SBC not given) from modern seafloor samples shown for reference. B06, Barnes and Sharp [2006]; B08, Barnes et al. [2008]; B09a, Barnes et al. [2009a]; B09b, Barnes et al. [2009b]; B12, Barnes and Cisneros [2012]; Bon08, Bonifacie et al. [2008]. Sw, serpentinization via interaction with seawater; pf, serpentinization via interaction with sedimentary pore fluids [Barnes and Sharp, 2006].

[26] Hydrogen isotope compositions of mineral separates from a subset of serpentinites and metasediments are plotted versus δ37Cl in Figure 4. With the exception of two serpentinites, all of the δD values are in the range −62 ± 17‰, in agreement with previous studies from Alpine serpentinites [Cartwright and Barnicoat, 1999; Früh-Green et al., 2001] and similar to phengites from the UHP Dora Maira complex [Sharp et al., 1993]. There is no correlation with δ37Cl values. Two serpentinites have significantly lower δD values (−105 and −150‰), but their chlorine isotope compositions are the same as modern seafloor serpentinites. One of these samples overlaps serpentinites from Elba that experienced postserpentinitization interaction with meteoric water [Barnes et al., 2006], with no effect on δ37Cl values.

Figure 4.

Plot of δD versus δ37Cl for serpentinites and metasedimentary rocks. Data from obducted Elba serpentinites and modern seafloor serpentinites shown for comparison [Barnes et al., 2006, 2009a]. Shaded field gives ranges in δD and δ37Cl from HP Erro-Tobbio serpentinites [Früh-Green et al., 2001; John et al., 2011]. Ranges in δD values for phengite from Alpine metasedimentary rocks [S93, Sharp et al., 1993] and serpentinites from elsewhere in the Z-S ophiolite [CB99, Cartwright and Barnicoat, 1999] shown on right.

[27] Despite the overall similarities between the chlorine isotope compositions of modern and metamorphosed slab materials (Figures 3 and 4), it is useful to further parse the data in order to evaluate fluid-rock interactions within and between rock types. Here we discuss the detailed characteristics of each rock type and relationships between δ37Cl values and other petrologic and chemical data.

4.2. Serpentinites and Rodingites

[28] In most cases, serpentine is the main Cl host in the serpentinites (Table S1). Chlorite and Ti-clinohumite are less abundant than serpentine and have lower Cl concentrations. As a result, serpentine should dominate the δ37Cl signature of the serpentinites. As shown in Figure 5, up to three generations of compositionally distinct serpentine are present in single samples. Serpentine shows considerable variability in Al2O3 and Cl contents on the thin section scale in partially overprinted serpentinites (Table S1 and Figure 6). Individual Type B serpentinites are relatively homogeneous, but as a group they display a large range in Mg/Mg + Fe. Overall, with increasing degree of textural overprinting, serpentine evolves toward a magnesian endmember (Figure 6) that has higher Cl concentrations than older serpentine. Based on these compositional trends, we expect the chlorine isotope systematics of Type B serpentinites to be little affected by fluid-rock interactions following syn-subduction mylonitization. In contrast, Type C and D serpentinites will likely have isotopic compositions that reflect interaction with fluids that added chlorine to the rocks during exhumation of the subduction complex.

Figure 5.

BSE images of HP serpentinites and rodingites. (a) Relic augite core surrounded by diopside in Type A partially serpentinized lherzolite. (b) Type C Srp2 replacing Type A serpentine (MA-1A). Srp1 core has lower Mg# and Cl content than Srp2. (c) Serpentine pseudomorph after spinifex olivine, set in serpentine matrix. Srp1 has lower Mg# and higher Al content than texturally younger Srp2. Chlorine is concentrated in Srp2 in pseudomorph. Sample CZ-1D, Type B with C overprint. (d) Type B/C serpentinite (USS-1) showing transposition of diopside + Ti-clinohumite + serpentine foliation into younger serpentine ± chlorite fabric (left side). (e) Type C serpentinite (CZ-1C) with three distinct compositions of serpentine. Srp1 has highest Al2O3, Srp2 the highest Mg#, and srp3 the highest Cl content. (f) Type D serpentinite (CZ-1B) with coarse serpentine intergrown with dolomite.

Figure 6.

Serpentine composition as a function of textural type. (a) Mg/Mg + Fe versus wt % Al2O3. Dashed arrows show textural younging within and between samples. MA-1A is partially serpentinized lherzolite; other samples have harzburgitic protoliths. (b) Mg/Mg + Fe versus Cl concentration. Samples with Type C overprint evolve toward higher Cl content. Type D sample has low Cl concentration.

[29] Chlorine isotope compositions are shown versus bulk chemical data (normalized to anhydrous basis; raw data in Table S2) from 17 serpentinites in Figure 7 and compared to average compositions of mid-ocean ridge basalt (MORB)-type depleted mantle (compiled from http://earthref.org/GERMRD/). Type A (partially overprinted) and Type B serpentinites have δ37Cl values that lie within the range of depleted mantle (−0.5 to +0.5‰) [Sharp et al., 2007, 2013], and bulk chemistry that deviates little from that of the peridotitic source. Type C and Type D serpentinites, however, record a progression to lower δ37Cl values. A vein sample from a Type D serpentinite records the lowest δ37Cl value (−2.7‰). There is a statistically significant correlation between δ37Cl values and loss-on-ignition data for all serpentinite samples (r2 = 0.724; r2 = 0.524 without Type D vein sample MA-2A): samples with the lowest δ37Cl have the highest loss on ignition (LOI) values. (Fully serpentinized peridotite will have ∼12 wt % H2O; the presence of carbonate results in higher LOI values in Type C and D serpentinites). This correlation suggests that δ37Cl values decrease with increasing degree of interaction with externally derived fluid(s).

Figure 7.

Whole-rock composition versus δ37Cl for different serpentinite types. Bulk chemical data are normalized to anhydrous basis to facilitate comparison between samples. Gray boxes show expected range in compositions for depleted mantle protoliths, based on elemental data from Earthref.org and isotopic data from Sharp et al. [2007, 2013].

[30] Type C and D serpentinites show positive trends on plots of SiO2 versus δ37Cl and MgO versus δ37Cl, consistent with the hypothesis that these rocks represent different degrees of interaction with a single low-δ37Cl fluid. However, the Cr data in Figure 7 complicate this interpretation: Cr increases with decreasing δ37Cl in Type C serpentinites, but Type D rocks are characterized by uniformly low Cr contents. Differences between Type C and D serpentinites are also apparent in FeO*, CaO, and Sr. These data are most readily explained by variable interaction of the HP serpentinites with at least two distinct low-δ37Cl fluids at different stages in their exhumation history.

[31] In addition to measuring chlorine isotope ratios in the silicate-bound fraction, we also analyzed the water-soluble chloride fraction (WSC) 7from a Type C serpentinite, a vein in a Type D serpentinite, one rodingite, and the tremolite-bearing selvage around that rodingite. In three of the four cases, the WSC δ37Cl value is 0.5–1.6‰ lower than the silicate value for the same sample (Table 1). In the serpentinite vein, δ37ClWSC = δ37ClSBC = −2.7‰, indicating vein formation from an isotopically light fluid. δ37ClWSC from the rodingite selvage is 1‰ lower than δ37ClWSC from the adjacent rodingitized dike, but δ37Clbulk values are essentially identical for the two samples.

Table 1. Mineral Assemblages, Cl Concentrations, and Isotopic Data in Analyzed Samplesa
SampleDescriptionGrtCldChlWMCarbAmpSrpTlcOlCpxEp/CzTypeOtherδ37Cl (‰)Cl (ppm)cCl (ppm)dδD (‰)δD Phase%H2O
  1. a

    Mineral abbreviations from Whitney and Evans [2010], except WM, white mica; Carb, carbonate; Type, serpentinite metamorphic stage according to textural classification scheme of Li et al. [2004b]: A, ocean-floor serpentinization; B, mylonitic serpentinite developed during subduction; C, early exhumation stage; D, greenschist-facies overprint; X, phase is present; −, phase is absent; ±, minor phase; psd, pseudomorph; alter, alteration; FIs, fluid inclusions; phyllo conc., phyllosilicate concentrate used for δ37Cl analysis.

  2. b

    Two distinct textural types.

  3. c

    Cl content calculated from peak area referenced to seawater standards on day of analysis.

  4. d

    Cl content determined by XRF analysis of washed rock powder. SBC, silicate-bound chloride; bulk, total chloride; WSC, water-soluble chloride; where not indicated, chlorine isotope value represents SBC.

HP Serpentinites, by Locality (All Contain Magnetite)
CZ-1ASerpentinite/ophicarbonate×Tr×±×D/E −1.48<10   
CZ-1BSerpentinite with Carb vein±Dol×±D/E −0.52729   
CZ-1CSerpentinite±±×±CTi-Chu−1.1, −0.747, 3532   
CZ-1DSerpentinite, bastite and mesh; spinifex Ol××A + DSpl−0.2160161−150Srp12.4
CZ-1ESerpentinite breccia/ophicarbonate××Tr×±D/ESpl, Ti-Chu−1.57<10   
LdC-10ASerpentinite 35 cm below rodingite×B  bdna−59, −57Srp12.4
LdC-10GSerpentinite 30 cm above rodingite×B/C  bdna   
LdC-12BSerpentinite mylonite±×B −0.11215−67Srp12.3
LDC-36Serpentinite××BTi-Chu−0.45573−105Srp/mix6.8
MA-1ASerpentinite with relic px; bastite, mesh××××A ± C 0.1, 0.641, 5785   
MA-1BSerpentiniteNo thin section        A/B 0.935na−60Srp10.1
MA-2ACoarse vein in serpentinite: SBC, WSC××D −2.7, −2.71110−78Srp12.0
MA-2BSerpentiniteNo thin section        C −0.67<10   
PM-12Serpentinite mylonite with relic bastiteMgs → Dol×B −0.416<10−71Srp11.3
SER-12AMag-rich serpentinite, folded myloniteNo thin section        C −1.36<10−67, −61Srp11.6
SER-12BMag-rich serpentinite, folded myloniteNo thin section        C −1.016na   
SER-30HMag-rich serpentinite, folded myloniteNo thin section        C −1.410<10−66Srp11.7
USS-1Folded serpentinite, abund Ol, Ti-Chu××××BTi-Chu0.36<10   
VT-1DSerpentinite 3 m below rodingite×××BTi-Chu0.413<10−58Srp11.7
VT-1ESerpentinite, folded mylonite±×CTi-Chu−1.1, −1.0, −0.77, 7, 16<10   
VT-2ESerpentinite, folded mylonite: SBC, WSC×Tr××C −0.9, −1.410na   
HP Rodingites and Selvages, by Locality
LdC-10DRodingite             −1.629na   
LdC-13ARodingite: SBC, bulk, WSC××Tr×× FIs in Cz0.2, 0.3, −0.6, −1.317, 16, 3018   
LdC-13CRodingite selvage: SBC, bulk, WSC×Tr× FIs in Amp−0.7, −0.8, −2.33, 8na   
LdC-14CSelvage between serp and gabbro×Act±× Mag0.513na−56Amp/mix8.1
SER-11ARodingite selvage, pyroxenite× Ap, Ttn, FIs in Cpx0.816na   
SER-11BRodingite×Act, Brs×× Vsv, Ap−1.27<10   
SER-11CRodingite××Act, Brs±× Vsv, Ap−0.42724   
SER-11ERodingite selvage×Act ×× ±Vsv0.73245   
VT-1ARodingiteNo thin section          −0.136176   
VT-1BRodingite selvageNo thin section          −0.39<10   
HP Altered Oceanic Crust, by Rock Type; All Samples Contain Rutile
SER-1JGlaucophanite, lower mine××PgGln×× Lws psd0.622na−55Tlc5.7
SER-10BGlaucophanite, lower mine×××PgGln×  0.01647−95Cld7.0
SER-10CGlaucophanite, lower mine×××PgGln×  0.91618   
SER-10DGlaucophanite, shear zone, lower mine×××Gln×  −1.51318   
SER-10EGlaucophanite, lower mine×××PgGln×  0.02825   
SER-30CGlaucophanite, lower mine×××±Gln×  0.116na   
SER-30LGlaucophanite and Chl schist, upper mine×××Gln×  −0.72057   
SER-10FTlc-Cld schist, lower mine×××in Grt×  0.510095−82Cld8.8
SER-30ITlc-Cld schist, upper mine××in Grt× FIs in Grt−0.32520   
SER-10CChl-Cld schist, lower mine×××PgGln×× Lws psd0.811na   
SER-30JChl-Cld schist, upper mine×××in Grt×  −1.24035   
SER-1IChl schist, lower mine××±Gln×  0.922na   
SER-30FGarnetite, upper mine×±± abund Ap0.8, 1.07496   
HP Metasedimentary Rocks, by Rock Type (All Contain Quartz and Rutile)
PM-10AiiCalcmica schist×Ph×±  −1.77na−58Ph3.7
PM-14FCalcmica schist××Pg××Omp× Ttn−2.67na−66Ph3.4
PM-15ACalcmica schist××Ph±××  −2.577   
PM-15BCalcmica schist×±Ph×××× Tur, Ttn−2.31210   
PM-4Mn metachert/schist±PhPmt Ap, Ab−0.5910−51Ph3.9
PM-5Mn metachert/schist×PhPmt Braunite, Ap, Ab0.02240−55, −45Ph2.5
PM-6BMn metachert/schist×PhPmt Ap−3.4, −2.914na−50Ph2.3
SER-11GMica schist×××Ph, Pg± Gln psd: Chl + Pg−3.3, −3.28, 57−62, −59Ph, Pg4.2
SER-30DMica schist××±psdPh, Pgin psd× Tur, Lws psd−1.9, −1.813, 9<10−55Ph, Pg2.4
UHP Metasedimentary Rocks, by Rock Type (All Contain Quartz ± Coesite, Rutile, ±Mn Oxides; Diamond-Bearing Fuid Inclusions Present in Mn Nodules)
LdC-30BVolcaniclastic? Retrogressed Omp quartzitebin GrtHbl±Ep, Zo Ap, retro Ab−3.447   
LdC-30CVolcaniclastic? Retrogressed Omp quartzitebPgin GrtHbl±Ep, Zo Ap, retro Ab−3.2814   
LdH-16HCalcmica schist×Pgin Grt± Tur, retro Ab, Bio bdna−54Pg4.3
LdC-19BCalcmica schistPh×  −2.814na−54Ph5.0
LdC-31C-mtxCalcmica schist: matrixPg±×  −2.71716−51Ph3.5
LdC-31C-veinCalcmica schist: epidote veinPg×  −2.74na   
LdC-31DCalcmica schist: phyllo conc.No thin section          −2.9820   
LdC-32FCalcmica schistPh×  −1.710<5   
LdC-32EMica schist: phyllo conc.×Ph  −2.5712   
LdC-34Mica schist: phyllo conc.×retroPh, Pg Ap, Al-phosphate−1.8610−53Ph, Pg4.6
LdC-32BMn schist×retroPh× Ap, Gln psd: Chl + Ab−2.2610−63Ph1.3
LdC-1B2Mn nodule and schist: phyllo conc.×PhMgs, Dol Ap, Dia−3.1, −3.015, 6na−55Ph4.2
LdC-31BMn nodule and schist: phyllo conc.××PgMgs, Dol×× Ap, Mn-Fe phosphate−1.7, −1.23, 25−55Ph3.7
LdC-31FMn nodule and schist: phyllo conc.×PhMgs, Dol  −2.31119   
LdC-37Mn nodule and schist: phyllo conc.×PhMgs, Dol Ap−2.969   
LdC-35ASiliceous marble: phyllo conc.×PhCal×  −3.7, −3.6, −3.54, 4, 4na−60Ph0.6
LdC-35BSiliceous marble: phyllo conc.××Ph, PgCalZo Ap, Ttn, Gph, Dia?−1.24043−72Ph, Pg3.8

4.3. Altered Oceanic Crust

[32] Bulk chemical data from the Servette mine (Table S2) [Martin et al., 2008] are compared to modern oceanic crust from black smoker systems [Humphris et al., 1998; Teagle and Alt, 2004; McCaig et al., 2007] and MORB in Figure 8. With the exception of the CaO data, glaucophanites overlap compositions of basalts altered via low-T interaction with seawater [e.g., Seyfried and Bischoff, 1981; Seyfried and Mottl, 1982]. Talc and chlorite schists are similar to hydrothermally altered oceanic crust from black smoker fields. The linear trends in Figure 8 are consistent with formation of the Servette schists by progressive loss of SiO2, MgO, CaO, and Na2O and residual enrichment of Al2O3 and FeO* during interaction between mafic oceanic crust and a single hydrothermal fluid. Only two of the Servette samples have Cl concentrations that differ significantly from the range reported for fresh MORB (Figure 8), and all values are within the range reported by Barnes and Cisneros [2012] for altered oceanic crust.

Figure 8.

Bulk chemical data for AOC samples from Servette compared to low-T interaction with seawater (black squares) [Seyfried and Bischoff, 1981], hydrothermally altered basalts from black smoker fields [Humphris et al., 1998; Teagle and Alt, 2004; McCaig et al., 2007], and unaltered MORB (Earthref.org database).

[33] The mineralogy of major rock types present at the Servette locality is described in detail by Martin et al. [2004, 2008]. Our data (Tables S3 and S4) are in good agreement with previously published mineral analyses but also include chlorine contents. In glaucophanites, amphibole is the dominant host of chlorine. Chlorite and talc host small amounts (up to 0.02 wt %), whereas chloritoid generally contains no detectable Cl. In chlorite and talc schists, Cl is distributed subequally between chloritoid and chlorite at concentrations up to ∼0.04 wt %. Talc consistently has the lowest Cl contents. In SER-30J, Cl is entirely hosted in a saponitic retrograde phase, indicating resetting of the Cl distribution by interaction with a late-stage fluid.

[34] δ37Cl values for all unretrogressed glaucophanites and AOC schists are between −0.7 and +0.9‰. Two samples with extensive retrogression have values of −1.5 and −1.2‰. Spatial relationships between the samples are shown in Figure S3. Passage of an isotopically light fluid through a shear zone had no effect on adjacent glaucophanites, suggesting that fluid infiltration after attainment of the pressure peak was highly channelized and did not affect most rocks.

4.4. Metasedimentary Rocks

[35] HP and UHP metasedimentary rocks have the lowest chlorine contents of all rock types investigated in this study (Table 1), and show no obvious correlation between Cl concentration and any other element (Table S5).

[36] In HP schists, chlorite is the major Cl host (Table S6). Chlorine is hosted in minor amounts by phengite and apatite but is below detection limits in piemontite in Mn metacherts (Figure 9a). Mixed analyses of fine-grained phases in pseudomorphs after glaucophane (Figure 9b) in schists are lower than concentrations in matrix phengite and chlorite. High-Si primary phengite (3.45–3.52 Si p.f.u.) contains measurable Cl, whereas phengite associated with breakdown of chloritoid (3.14–3.36 Si p.f.u.) contains no detectable chlorine (Table S6) and secondary chlorite has lower Cl contents than matrix chlorite. Although chloritoid breakdown requires fluid infiltration (e.g., 3 Cld + 3 Celad + 2 H2O = Chl + 3 Mu + 4 Qz), mineral compositions show that no significant chlorine was added during retrogression. Micas in pseudomorphs after lawsonite have higher Cl contents than matrix micas. Lawsonite breakdown likely occurred via a prograde reaction (e.g., 24 Lws + 5 Celad = 12 Cz + 5 Mu + Chl + 14 Qz + 38 H2O). Because this reaction releases rather than consumes H2O, pseudomorph phases likely acquired their chlorine from precursor minerals rather than from interaction with an externally derived fluid.

Figure 9.

BSE images of HP/UHP metasedimentary rocks. (a) Piemontite + braunite-bearing HP schist PM-5. (b) Chlorite + phengite pseudomorph after amphibole in HP sample SER-11G. (c) Phengite, paragonite, and calcite matrix, UHP sample LdC-35B. (d) Inclusions of intergrown carbonate + chlorite (circles) in garnet; Cl is higher in included chlorite than in matrix chlorite (LdC-35B, UHP). (e) Prograde phengite with Cl-free biotite rims (LdC-31B, UHP). (f) Pseudomorph after glaucophane (UHP, LdC-32B). Pseudomorph has lower Cl content than matrix phases.

[37] Different minerals host the bulk of the Cl in UHP metasediments of different bulk compositions (Table S7). Where phengite and paragonite coexist (Figure 9c), Cl contents are higher in paragonite. Texturally early phases, such as chlorite in pseudomorphs enclosed in garnet (Figure 9d), have higher Cl contents than the same phases in the matrix (Table S7). Obvious secondary minerals, such as biotite rims on phengite (Figure 9e) and chlorite pseudomorphs after glaucophane (Figure 9f), contain little to no chlorine. Likewise, lawsonite pseudomorphs are nearly devoid of chlorine (Tables S4 and S7). All of these observations suggest that the chlorine isotope composition of the UHP metasedimentary rocks is biased toward the prograde or highest pressure stage of their history, with little to no effect from fluid infiltration during exhumation.

[38] The δ37Cl values for the metasediments range from −3.6 to 0‰ (n = 25), with a median of −2.5‰ (Table 1 and Figure 3). An unpaired Student's t test shows that the mean chlorine isotope compositions of HP versus UHP metasedimentary rocks in the Z-S ophiolite package are not statistically different from one another (−2.3‰ versus −2.7‰; P = 0.1099).

[39] One striking feature of the metasedimentary data is the heterogeneity in δ37Cl values that is preserved in closely spaced samples. Two UHP siliceous marble samples collected from different layers within centimeters of one another (Figure 1f) differ by >2‰ from each another. The same is true for two HP mica schists collected a meter apart at Servette (Table 1 and Figure S3). Two UHP Mn nodules at outcrop LdC-31 differ by ∼1‰ from each other (Table 1) and also differ from the enclosing calcmica schists and an early epidote vein.

[40] The only previously published chlorine isotope data from eclogite-facies schists have a median value of 0‰ and a range of −2.2 to +2.2‰ (Raspas complex, Ecuador, n = 8) [John et al., 2010]. The median and the positive values differ significantly from the data presented here. Our data also differ statistically from modern marine sediments, which have median δ37Cl = −0.6 and a range of −2.5 to +0.5‰ (n = 21) [Barnes et al., 2008, 2009b]. In particular, there is no overlap in δ37Cl values between pelitic and marly samples analyzed in this study and published protolith data (Figure 10). These metasedimentary rocks are the only samples in this study that show a significant change in chlorine isotope composition relative to plausible presubduction compositions, indicating some degree of isotopic resetting. As argued below, it is unlikely that dehydration drove them toward low δ37Cl values. Instead, interaction with an externally derived fluid appears to be required. In contrast, the Mn crusts and nodules show nearly complete overlap with data from modern radiolarites and siliceous oozes and hence preserve no conclusive evidence of fluid-rock interaction.

Figure 10.

Comparison of Cl isotope compositions of HP/UHP metasedimentary rocks (this study) with marine sedimentary protoliths [Barnes et al., 2008, 2009b]. Not all lithologies can be correlated between the different studies; the data set here is thus smaller than in Figure 3.

5. Discussion

5.1. Protolith Versus Modified Chlorine Isotope Compositions

[41] Chlorine isotope compositions of the HP serpentinites and rodingites overlap serpentinite values from the modern seafloor (δ37Cl = −1.5 to +0.7‰) [Barnes and Sharp, 2006; Bonifacie et al., 2008; Barnes et al., 2009a]. HP samples from the Erro-Tobbio peridotite body (Italy) and the Almiraz peridotite (Spain) record a similar range (δ37Cl = −1.6 to +0.4‰) [Bonifacie et al., 2008; John et al., 2011] and provide evidence that subduction-zone metamorphism does not, per se, result in chlorine isotope fractionation during HP deserpentinization [John et al., 2011]. Most AOC samples also record chlorine isotope compositions that show little modification from expected seafloor values. Modern hydrothermal vent fluids at temperatures of 236–340°C have δ37Cl values that are indistinguishable from those of seawater (0.0‰) [Bonifacie et al., 2005], and experimental work confirms that minimal Cl isotope fractionation occurs during liquid-vapor phase separation of aqueous fluids [Liebscher et al., 2006].

[42] Although the metasedimentary rocks have δ37Cl values that are the least like their modern analogues, they also preserve isotopic variations between different horizons. Because the highest-Cl concentrations in these rocks are associated with the texturally earliest mineral phases and compositions, it is likely that the isotopic heterogeneity between samples is inherited from the protoliths and does not reflect HP fluid interaction.

[43] Oxygen and hydrogen isotope data from rocks throughout the Z-S ophiolite also indicate widespread preservation of seafloor values [Barnicoat and Cartwright, 1995; Cartwright and Barnicoat, 1999], and HP vein minerals have isotopic compositions consistent with derivation from locally derived fluid. Cartwright and Barnicoat [1999] suggest that in situ breakdown of lawsonite and/or glaucophane liberated the H2O necessary for vein formation. This conclusion is in accord with our observations of pseudomorphs after lawsonite and glaucophane in rocks that show no evidence for introduction of chlorine during reaction.

[44] In contrast to the above, isotopic and bulk-rock chemical data (Figures 4, 7, and 11) show that Type C and D serpentinites were locally modified via interaction with a 37Cl-depleted fluid. Water-soluble chloride data from a Type C serpentinite, a Type D serpentine vein, and fluid-inclusion-rich rodingites range from δ37Cl = −2.7 to −1.3‰ (Figure 11). The silicate-bound δ37Cl values of Type C and D serpentinites can be explained by binary mixing between Type A/B serpentinite and fluid(s) with the composition(s) of the WSC fractions (arrows marked 3 in Figure 11). Based on correlations with the data of Li et al. [2004b], this fluid interaction did not occur until after the onset of exhumation. Similarly, at Servette, infiltration of a low-δ37Cl fluid through a shear zone caused late-stage retrogression and locally reset the isotopic composition of glaucophanite.

Figure 11.

Box-and-whisker plot showing δ37Cl values as a function of serpentinite textural type. Data from pore waters in oceanic basement and accretionary complexes shown for comparison. Water-soluble chloride data are from Type C and D serpentinites and rodingites. Blue arrows show proposed model for modification of Cl isotope compositions of metasedimentary rocks and Type C/D serpentinites: 1, interaction of slab-top sediments with accretionary-wedge pore fluids produces low-δ37Cl sedimentary signature; 2, HP/UHP devolatilization of metasedimentary rocks releases fluids with negative δ37Cl values, represented by WSC fraction in serpentinites; 3, interaction of Type A and B serpentinites with low-δ37Cl fluid derived from metasediments produces Type C and D serpentinites.

[45] Where rocks were affected by interaction with late-stage fluids, that interaction was patchy rather than pervasive. Overall, there was little to no modification of the chlorine isotope composition of the Z-S ophiolite during subduction to depths around 80 km. Had these rocks continued to greater depths in the subduction zone, most would likely have retained their seafloor signatures to subarc depths, in agreement with the conclusions of John et al. [2011].

5.2. Origin of Low-δ37Cl Metasedimentary Rocks and Fluids

[46] Although most of the chlorine isotope data are consistent with inherited seafloor values, the strikingly low δ37Cl values of some metasedimentary rocks—in addition to reset Type C and D serpentinites and sheared glaucophanite at Servette—require further explanation. At least three hypotheses could account for the compositions of these rocks. (1) Existing data do not span the full range of Cl isotope compositions of modern marine sediments. (2) Dehydration reactions during subduction released fluid that was enriched in 37Cl relative to remaining minerals, leaving the solid residue enriched in the lighter isotope. (3) The sedimentary protoliths interacted with a low-δ37Cl fluid early in their subduction history and preserved isotopically light compositions throughout their metamorphic history.

[47] Testing the first hypothesis must await analysis of a larger suite of seafloor sediments. The second hypothesis requires (a) that devolatilization occurred at low temperature, to allow for fractionation, and (b) that Δ37Clmineral-fluid was negative, in opposition to theoretical predictions [Schauble et al., 2003]. Speciation and polymerization in high-P aqueous fluids can differ significantly from low-P fluids [e.g., Manning, 2004a, 2004b], and it is possible that different bonding behaviors in such fluids affect Δ37Clmineral-fluid. In the absence of specific evidence that this is the case, however, the second hypothesis remains highly speculative.

[48] Of all of the plausible subduction inputs that have been analyzed to date, the only confirmed low-δ37Cl source is pore water from the accretionary wedge (Figure 11). Pore fluids from oceanic basement rocks are also isotopically negative [Bonifacie et al., 2007] but show a more restricted range in values. Compaction within the wedge coupled with bending stresses should act to drive these fluids downward, toward, and into the top of the subducting plate (Figures 12a and 12b). In the case of the Z-S ophiolite, the top of the plate comprised a relatively thick sequence of marine sediments. Outboard of the trench, bending stresses should channel seawater into the downgoing plate, producing rocks with near 0‰ values. As the plate entered the trench, however, low-T interaction between subducting sediments, their pore fluids, and isotopically light accretionary-wedge pore fluid could produce a low-δ37Cl horizon at the top of the downgoing plate. If brine infiltrating from the wedge had δ37Cl ≤ −3‰, diagenetic hydration reactions could produce minerals with low δ37Cl values (arrow 1, Figure 11). The degree of equilibration with low-δ37Cl fluids would depend on proximity to bending stress-related faults and the initial mineralogy—hence reactivity—of the sediments. The net result would be a sequence of isotopically heterogeneous metasedimentary rocks capping relatively homogeneous oceanic lithosphere, in accord with the predictions of the third hypothesis above. Similar downward fluid flow across the subduction interface, induced by plate-bending stresses, has been proposed by Halama et al. [2010, 2012] and John et al. [2011].

Figure 12.

Cartoons illustrating possible physical scenario consistent with chlorine isotope data. (a) Unbending stresses at depth in the subduction system result in cross-slab and updip, within-slab fluid flow [after Faccenda et al., 2012]. (b) Enlarged view showing downward fluid flow (along normal faults) induced by bending stresses. (c) Modified from Figure 12a to show subduction of thinned continental margin and upward migration of tectonic slices within the subduction channel, consistent with the Alpine setting [e.g., Agard et al., 2009]. (d) Enlargement showing cross-slab fluid flow and modification of Cl isotope signature induced by unbending stresses and/or exhumation-driven devolatilization in tectonic slivers in subduction channel. Yellow stars, positions of representative serpentinites; circles, sediment/metasedimentary rocks.

[49] During subsequent subduction, Δ37Clmineral-fluid decreases with increasing T such that localized devolatilization reactions will have little effect on chlorine isotope compositions except where exotic fluids migrate across lithologic boundaries. During subduction to the peak conditions of the Z-S ophiolite, only metapelitic bulk compositions are predicted to lose a significant amount of H2O via dehydration reactions [e.g., Rüpke et al., 2004; Hacker, 2008]. The fluid released by these rocks will have a Cl isotope composition identical to that of the parent rock (arrow 2, Figure 11). Serpentinites and hydrated oceanic crust, in contrast, are likely to carry most of their H2O to greater depths before widespread dehydration of antigorite occurs [Ulmer and Trommsdorff, 1995; van Keken et al., 2011].

5.3. Fluid-Rock Interaction During Exhumation in the Subduction Channel

[50] Chlorine isotope data from Type C and D serpentinites, fluid-inclusion-rich rodingites, sheared glaucophanite, and retrogressed AOC schist record interaction with a low-δ37Cl fluid during exhumation of the Z-S ophiolite (3 in Figure 11). Figure 12a shows regions of both updip and cross-slab fluid flow associated with unbending stresses at depth in an oceanic subduction zone, based on modeling by Faccenda et al. [2012]. In Figure 12c, we superimpose Faccenda et al.'s model results on a schematic cross section showing subduction of thinned continental crust that is more compatible with the Alpine tectonic setting. Preservation of relatively coherent slices that record different Pmax conditions (UHP LdC slice, HP Z-S ophiolite, Combin ophiolite) indicate that large slices of the downgoing plate detached at different depths and were transported upward in a subduction channel. Juxtaposition of different rock types against one another during exhumation in the channel could allow for cross-lithology fluid flow. During exhumation, rock P-T paths can cross-reaction boundaries at high enough angles to enhance dehydration, which will produce fluids with chlorine isotope compositions identical to those of the parent rock. In the case of the low-δ37Cl metasediments, aqueous fluids with δ37Cl around −3.5 to −1‰ will be liberated (arrow 2, Figure 11 and 12d). Cross-slab faulting and fluid flow predicted by Faccenda et al. [2012] will lead to localized infiltration of these fluids into structurally higher slices in the subduction channel. Interaction with overlying serpentinites and glaucophanites results in binary mixing of seafloor and metasediment-derived chlorine isotope signatures (3 in Figures 11 and 12d), yielding δ37Cl values that are 1–2‰ lower than adjacent unaltered rocks. Deep regions of updip fluid flow are likely to transport fluids derived from continuously subducting serpentinites beneath the subduction channel; these fluids will have δ37Cl values around 0‰ (Figure 12). At depths beyond the subduction channel, serpentinite breakdown should produce widespread aqueous fluid with δ37Cl ≈ 0‰.

5.4. Implications for Grain-Scale Mass Transfer During Subduction

[51] The chlorine isotope data presented here are consistent with other studies of exhumed subduction zone rocks showing that many such rocks retain stable isotope compositions indicative of their original seafloor histories [e.g., Getty and Selverstone, 1994; Barnicoat and Bowtell, 1995; Cartwright and Barnicoat, 1999; Putlitz et al., 2000; John et al., 2011]. They also support studies showing that little fluid-assisted mass transfer occurs on an intergranular scale during subduction to 2–3 GPa, despite considerable devolatilization [Selverstone et al., 1992; Spandler and Hermann, 2006; Spandler et al., 2007; Bebout et al., 2013; Spandler and Pirard, 2013]. Large chlorine isotope heterogeneities preserved between adjacent samples require that there was little fluid communication between layers during their metamorphic history. Localized cross-lithology fluid flow capable of modifying chlorine isotope compositions only occurred during exhumation-driven dehydration. The general absence of evidence for percolative cross-slab flow leads us to conclude that pervasive mass transfer sufficient to reset chlorine isotope compositions may not occur within an intact slab until pressures greater than ∼3.2 GPa are reached (constrained by LdC samples). Our conclusions do not rule out channelized cross-slab mass transfer, however, because our sampling strategy intentionally avoided HP/UHP veins and similar structures. Updip transport of coherent slices within a subduction channel can lead to heating and dehydration that release fluids into the mantle wedge over a wide range of pressures. Corner flow within the mantle wedge may transport this metasomatized mantle material to greater depths, where serpentine breakdown can trigger hydrous melting of slab and mantle components [Spandler and Pirard, 2013].

6. Summary and Conclusions

[52] Subduction of serpentinites and variably altered oceanic crust resulted in no change in chlorine isotope compositions relative to samples from the modern seafloor during transport to ≥80 km. However, sediments at the top of the downgoing plate acquired anomalously low δ37Cl values via interaction with accretionary-wedge pore fluids as the plate entered the trench. Large differences in chlorine isotope composition between adjacent rock types were preserved throughout the subduction history, and fluid-enhanced mixing between lithologies only occurred during exhumation within the subduction channel. Cross-lithology fluid flow, induced by unbending stresses or differential exhumation of rock slices in the subduction channel, allowed for localized fluid infiltration from the metasedimentary rocks into ultramafic and mafic rocks. It is likely, however, that most rocks beneath the subduction channel carried unmodified chlorine isotope signatures to greater depths. Widespread homogenization of chlorine isotopes throughout the slab may occur in association with antigorite breakdown, but if fluids are channelized along unbending-related faults, chlorine isotope heterogeneities could be preserved between faults to postarc depths. This scenario may explain the negative δ37Cl values documented in some ocean island basalts [John et al., 2010], which appear to require a stored crustal component in the mantle.

Acknowledgments

[53] G. Bebout and an anonymous reviewer provided excellent journal reviews. We thank A.-M. Ali and K. McCeague for XRF analyses, V. Atudorei for the hydrogen isotope data, J. D. Barnes for discussions, M. Halick and A. T. H. Whittaker for assistance in the field, and then-Congressman Martin Heinrich's staff for their efforts in securing the release of our samples from U.S. Customs and Border Patrol. Financial support was provided by NSF grant EAR-0911669.

Ancillary