A Garnet–Zircon Oxygen Isotope Record of Subduction and Exhumation Fluids from the Franciscan Complex, California

Abstract

Tectonic blocks of high-grade metamorphic rock hosted by the Franciscan mélange commonly preserve textural and geochemical evidence for metasomatism and blueschist-facies overprinting within a subduction zone. These exhumation events can obscure information on the nature of subduction fluids at depth that might otherwise be preserved by these rocks in this tectonic setting. Cation and oxygen isotope zoning in garnets from eclogite and garnet–hornblende rocks with minimal textural evidence of retrogression reveals separate prograde and retrograde fluid histories. Garnet cores preserve a homogeneous prograde oxygen isotope history with both high (δ¹⁸O ∼11‰) and low (∼4‰) values in different samples probably recording altered ocean-crust protolith compositions. Multiple episodes of garnet resorption and regrowth record a dramatically changing bulk oxygen isotope ratio resulting in 6–7‰ neoformed garnet rims with sharp core–rim boundaries in both samples. Matrix omphacite and hornblende are in profound oxygen isotope disequilibrium with garnet cores but are in equilibrium with garnet rims despite the appearance of overall textural equilibrium in thin section. This cryptic metasomatism found in samples from different locations within the Franciscan Complex brings into question existing thermobarometry on non-inclusion phases for this complex, as well as interpretations based on whole-rock geochemical data. Zircon ages and oxygen isotope equilibrium with late-forming garnet are consistent with post-metasomatic formation of zircon. The ability of garnet and zircon to record multiple fluid events and tie them to P–T–t–fluid history is a powerful tool in gaining an improved understanding of the complex fluid environments within subduction zones.

INTRODUCTION

The relationships between the record of fluids preserved in high-pressure metamorphic rocks exhumed from subduction zones, the fluids that are presumed to be responsible for the generation of arc magmas, and the characteristic subduction-zone trace element signature, are areas of great interest in arc petrology (e.g. Bebout, 2007). Dehydration reactions within the down-going slab generate substantial fluxes of aqueous fluid that migrate into the mantle hanging wall of the subduction zone, dictating the extent to which mantle melting will occur (e.g. Peacock, 1990; Kerrick & Connolly, 2001). The record of these fluids found in exhumed subduction-zone metamorphic rocks is mixed, however. Whole-rock stable isotope and trace element data have been used to establish extensive fluid–rock interaction and metasomatism within the subduction environment (Bebout & Barton, 1989, 1993; Sorensen & Grossman, 1989), as well as mechanical mixing along the slab–mantle interface (King et al., 2006, 2007; Marschall & Schumacher, 2012). Metasomatic rocks such as jadeitites and related lithologies are commonly found in serpentinite mélanges associated with subduction zones, and preserve evidence of complicated fluid histories (Johnson & Harlow, 1999; Harlow et al., 2003; Sorensen et al., 2006; Fu et al., 2010).

Other studies, principally on eclogite and related rocks, that report textural evidence of metamorphic fluids in the form of fluid inclusions (Scambelluri & Philippot, 2001) and veins (Spandler & Hermann, 2005) suggest that these features are the products of internal fluid flow and are unlikely to represent fluid entering or leaving the subducting slab. A large number of geochemical studies are similarly suggestive of limited fluid flow in samples of mafic metamorphic rocks formed in subduction zones (Philippot & Selverstone, 1991; Selverstone et al., 1992; Getty & Selverstone, 1994; Barnicoat & Cartwright, 1995; Ganor et al., 1996; Putlitz et al., 2000, 2001; Früh-Green et al., 2001).

A commonly cited reason for these disparate results is that fluid flow within subducting slabs may be highly channelized, and the discrepancy between studies reflects the low permeability of subducted mafic rocks and the complicated mélange structure of the slab–mantle interface (Bebout & Barton, 2002; Breeding et al., 2004; King et al., 2006, 2007; Ague, 2007).

Additionally, the prograde and peak metamorphic geochemical record of some samples can be obscured by variable metasomatism during exhumation by channelized flow immediately above the subducting slab. Geochemical studies of metamorphic rocks that preserve evidence of substantial fluid flow and metasomatism often employ whole-rock or bulk mineral separates, an approach that homogenizes zoned minerals and makes discerning between prograde and retrograde fluids challenging. Recent detailed work on high-pressure veins from alpine eclogites suggests that these rocks can preserve a chemically distinct record of both internally and externally sourced fluids, resolvable only by preserving the spatial context of geochemical analyses (Spandler et al., 2011). In addition, garnets from orogenic eclogites preserve subtle zoning in oxygen isotope ratios that records a previously unrecognized prograde fluid infiltration (Russell et al., 2013).

In this study we interrogate the stable isotopic record of subduction zone metamorphism preserved in garnet and zircon at a fine (µm) scale. In addition to its wide use as a geochronometer, zircon has become highly valued for its ability to preserve a robust record of a variety of isotopic and trace-element geochemical systems through rather extreme conditions (e.g. Hanchar & Hoskin, 2003; Bowman et al., 2011). Similarly, beyond its utility in thermobarometry, garnet is characterized by slow (although not as slow as zircon) intracrystalline diffusion rates, and the ability to retain geochemical information in the form of growth zoning up to granulite-facies conditions (Coghlan, 1990; Vielzeuf et al., 2005; Ganguly, 2010; Page et al., 2010). Zircon and garnet are both commonly found in eclogite and related subduction zone metamorphic rocks, and can record information from both the subduction and exhumation phases of metamorphism (Krogh et al., 1974; Krogh, 1982; Rubatto et al., 1999; Schmitz & Bowring, 2001; Page et al., 2003). By analyzing oxygen isotope ratios by ion microprobe in zircon and garnet, we can use the robust nature of these minerals to see through the multiple episodes of metasomatism that might be expected in the subduction zone environment. Furthermore, by identifying the oxygen isotope record of fluids in minerals that also record pressure–temperature and geochronological information, the fluid–rock interaction history may be tied directly to the P–T–t–fluid(s) history for a more complete understanding of subduction dynamics. In doing so, for example, it may be possible to resolve the complications of widespread fluid overprinting during exhumation, and gain more insight into the nature and timing of fluid fluxes at depth in subduction zones.

Two zircon- and garnet-bearing eclogite and hornblendite samples were chosen for this study from the Franciscan Complex of central California. The Franciscan is among several paleo-subduction zones in which tectonic blocks of high-pressure metamorphic rocks are hosted by a sedimentary–ultramafic mélange. These rocks are renowned for their textural, chemical, and isotopic evidence of large fluid fluxes and metasomatism. High-grade metamorphic blocks preserve textural evidence of metasomatism in the form of glaucophane- and actinolite-rich rinds (Coleman, 1980; Horodyskyj et al., 2009) as well as the presence of euhedral minerals (zoisite, aragonite, titanite, omphacite), which grew into fluid-filled cavities (Essene & Fyfe, 1967; Cloos, 1986; Page et al., 2007a). Geochemical evidence of metasomatism in the form of zoned minerals exists both for elements typically viewed as highly mobile, such as the large ion lithophile elements (LILE; e.g. Sorensen et al., 1997; Catlos & Sorensen, 2003), and those that are generally thought to be less soluble in aqueous solution, such as high field strength elements (HFSE) and the light rare earth elements (LREE; e.g. Sorensen & Grossman, 1989). To maximize the potential for recording both an early and late history of subduction fluids, a sample was chosen that contains garnets that record a prograde subduction history, and a second sample with garnet that records the oldest ages in the region.

REGIONAL GEOLOGY

The high-grade metamorphic rocks hosted by the Franciscan Complex in the Central and Northern California Coast Ranges have been the subject of intensive petrological and geochronological study for the past century. High-grade blocks (usually on the scale of 1–10 m) of eclogite, blueschist and garnet hornblendite (sometimes referred to as amphibolite despite the common lack of plagioclase) are hosted by a lower-grade metagreywacke mélange (e.g. Bailey et al., 1964; Cloos, 1986). Many high-grade blocks preserve a polymetamorphic history with evidence of significant rock–fluid interaction. A ubiquitous blueschist-facies overprint on all lithologies was recognized from the onset of petrological investigation in the area (e.g. Holway, 1904; Borg, 1956). Evidence of high-P and high-P/T overprinting (Essene & Fyfe, 1967; Moore, 1984; Moore & Blake, 1989) ultimately culminated in the proposal of counter-clockwise P–T paths for several eclogite blocks (Oh & Liou, 1990; Wakabayashi, 1990; Krogh et al., 1994). More recently, a series of studies have applied quantitative thermobarometry to inclusion phases in garnet from eclogite, generating a robust set of counter-clockwise P–T paths, with near-peak conditions of ∼600°C and ∼24 kbar, and blueschist-facies prograde paths (Ravna & Terry, 2004; Tsujimori et al., 2006a; Page et al., 2007a).

Isotopic dating of Franciscan metamorphic rocks has been continuing for almost 50 years, and many of these studies have been summarized by Wakabayashi & Dumitru (2007). Garnet hornblendites yield the oldest ages (hornblende Ar–Ar, 156–160 Ma; Ross & Sharp, 1986, 1988; Wakabayashi & Dumitru, 2007; garnet Lu–Hf, 153–168 Ma; Anczkiewicz et al., 2004). Garnet Lu–Hf mineral isochrons for eclogite (∼158 Ma) and blueschist (∼147 Ma) blocks yield generally younger ages than those for hornblendite, suggesting that the three high-grade lithologies in the Franciscan record the slow cooling of the subduction zone (Anczkiewicz et al., 2004). ⁴⁰Ar/³⁹Ar ages of phengite in Franciscan high-grade blocks generally cluster around 135–140 Ma, but range up to 156 Ma; phengite grains with younger ages are generally texturally late, whereas older phengite (e.g. 156 Ma from the Tiburon eclogite; Catlos & Sorenson, 2003) is apparently texturally in equilibrium with the eclogitic assemblage. Most phengite grains are variably enriched in the LILE, principally Ba (Sorensen et al., 1997; Catlos & Sorensen, 2003; Page et al., 2007a). The wide range in their ages has been proposed as evidence for long-lived fluid metasomatism within the Franciscan subduction zone (Catlos & Sorenson, 2003). A lawsonite Lu–Hf isochron yielded an age of 146 Ma for a blueschist block from Tiburon, consistent with other ages from blueschist blocks (Mulcahy et al., 2009). Nelson (1991) found similarly young ages from blueschist minerals in late veins and vugs in eclogite from Tiburon and the Diablo Range, with an Rb–Sr isochron of ∼150 Ma. In summary, ⁴⁰Ar/³⁹Ar and Lu–Hf ages from prograde and peak minerals in eclogite and hornblendite range from 153 to 168 Ma. Rb–Sr ages on late vug minerals, the ⁴⁰Ar/³⁹Ar chronology of texturally late phengite, and Lu–Hf ages for garnet and lawsonite in blueschists range from ∼135 to 150 Ma (Nelson, 1991). These younger blueschist-facies ages overlap ⁴⁰Ar/³⁹Ar phengite ages from coherent regions of Franciscan metamorphic rocks (120–154 Ma; Wakabayashi & Dumitru, 2007).

SAMPLE DESCRIPTIONS

Two high-grade metamorphic blocks were chosen for this study. The first block is an eclogite located ∼5 km SW of Healdsburg, California (38°35′42″N, 122°53′50″W), known as the Junction School (JS) eclogite (Sorensen et al., 1997; Page et al., 2007a). The sample in this study is from an ∼15 m diameter tectonic block, the largest and least visibly overprinted block at this locality. This block was chosen for study because it appears to be a relatively pristine sample of Franciscan eclogite with minimal evidence of overprinting in the form of significant veining or abundant late glaucophane; it does not have a metasomatic rind. The JS sample is a fine- to medium-grained eclogite consisting primarily of <1 mm omphacite and 0·5–1 mm garnet grains (Fig. 1a). A detailed textural and mineralogical description of this rock has been given by Page et al. (2007a) and is therefore presented here only in summarized form. Both omphacite and garnet are compositionally zoned, and this zoning is visible in backscattered electron (BSE) images (Fig. 1b and c). Large matrix rutile crystals (500–1000 µm) are partially to completely replaced by titanite (Fig. 1a). In more overprinted samples, partially resorbed garnets are rimmed by chlorite, phengite, and minor glaucophane. The outcrop has a mottled appearance on the centimeter scale, with darker patches containing a higher percentage of chlorite and phengite; joints in the rock contain abundant phengite (Sorensen et al., 1997). In thin section, more heavily altered patches contain euhedral epidote and omphacite crystals that may have originally grown into fluid-filled cavities (Page et al., 2007a), and are now surrounded by quartz (Fig. 1c) or titanite. Garnets in these areas are strongly zoned and show evidence for multiple episodes of growth and resorption. Abundant small zircons are found in garnet and omphacite rims (Fig. 1b and c). Mineral inclusions in garnet cores record the prograde metamorphism of this sample from ∼10 kbar and 400°C to a peak of ∼22 kbar and 550°C preserved in garnet mantles within a clear resorption boundary (Page et al., 2007a). Garnet rims record blueschist-facies conditions of 10–16 kbar and 350–400°C; although the prograde and retrograde P–T path of this rock is known in some detail, it has not yet been the subject of any geochronological studies.

The second sample is plagioclase-free garnet hornblendite from Panoche Pass (PNP) in the Diablo Ranges ∼40 km SE of Hollister, California (36°39′29″N, 121°06′48″W); it is composed of brown hornblende (0·5–2 mm long), garnet (1–3 mm) and ilmenite–rutile compound grains (∼300 µm, Fig. 1d and e). Hornblendite from the Panoche Pass locality was described by Hermes (1973), and the region has recently been mapped in detail by Wakabayashi & Dumitru (2007). Brown or green amphibole (hornblende) cores are surrounded by thin blue amphibole rims (Fig. 1d), and contain abundant inclusions of apatite (Fig. 1e and f). Cation zoning in garnets is not as dramatic as in the JS eclogite. Garnet has resorbed rims and abundant inclusions of hornblende, apatite, rutile, ilmenite and quartz. Ilmenite is often found in compound grains with rutile that has thin rims of titanite (Fig. 1e and f). Secondary compositionally zoned allanitic epidote, titanite, apatite, phengite, glaucophane and chlorite decorate grain boundaries, particularly around partially resorbed garnet (Fig. 1f). This block was chosen for study because it has one of the best-constrained chronologies of any Franciscan block. Garnet in this block crystallized at 168·7 ± 0·8 Ma (Lu–Hf; Anczkiewicz et al., 2004) and hornblende at 162 ± 2 Ma, with retrograde muscovite forming at ∼145 Ma (Ar–Ar; Ross & Sharp, 1988). Existing thermobarometric estimates are restricted to garnet–hornblende thermometry, yielding temperatures of 650–750°C (Anczkiewicz et al., 2004).

ANALYTICAL METHODS

Most chemical and isotopic analyses in this paper were performed on polished thin sections to preserve textural information. However, given the scarcity of relatively large zircon grains in these lithologies, zircons were separated and concentrated using standard gravimetric and magnetic techniques, followed by handpicking and mounting in epoxy. BSE and polychromatic cathodoluminescence (CL) images of minerals were made using the CAMECA SX100 electron microprobe at the University of Michigan and scanning electron microscopes at the University of Wisconsin–Madison and Oberlin College. Mineral chemistry was determined on a CAMECA SX100 electron microprobe at the University of Michigan using a focused 15 kV, 10 nA electron beam. Natural and synthetic silicate and oxide standards were used, and data were reduced using a CAMECA PAP-type correction calculating oxygen by stoichiometry.

Laser fluorination

Small-diameter (6 mm × ∼10 mm) plugs were drilled from hand samples of eclogite and hornblendite and were crushed to pieces of ∼0·5 mm in diameter. Garnet, clinopyroxene, hornblende and titanite separates were handpicked, and chips for analysis were selected for optical clarity and lack of visible inclusions or attached contamination. Aliquots of mineral chips weighing 1–2 mg were pre-treated overnight in the sample chamber with BrF₅, then individually heated with a CO₂ laser in the presence of BrF₅. Samples of whole-rock powder were not pre-treated in BrF₅ to avoid pre-reaction, but instead were evacuated overnight in an airlock chamber and individually moved under vacuum to the fluorination chamber for analysis (Spicuzza et al., 1998). All analyses were standardized against four analyses of oxygen isotope standard UWG-2 per day (δ¹⁸O = 5·80‰ VSMOW; Valley et al., 1995) and reported in standard per mil (‰) notation relative to VSMOW. The eight analyses of UWG-2 averaged 5·79 ± 0·14‰ (2 SD; 2 days of analyses), yielding an average correction of 0·01‰.

Ion microprobe

Separated zircons were mounted within 5 mm of the center of a 25 mm epoxy disk with standards AS3 (Paces & Miller, 1993; Schmitz et al., 2003) and 91500 (Wiedenbeck et al., 1995, 2004) and polished to the approximate centers of the sample grains. After initial CL and BSE characterization, zircon was analyzed for U–Th–Pb elemental ratios and Pb isotopic composition (details below). Correlation of U–Pb, trace element and oxygen isotope analyses followed a similar approach to that described by Cavosie et al. (2006). After U–Pb analysis, zircons were imaged with secondary electrons (SE) and CL to determine exact pit locations. Trace element analyses were made on the same surface as U–Pb analyses, typically within the same CL domain. After trace-element analysis, oxygen isotope standard KIM-5 zircon (Valley, 2003) was face-mounted into the existing disk and flattened and polished by hand. Care was taken to remove the least amount of material possible and to minimize any differential polishing relief between zircon and epoxy matrix as this can affect precision in oxygen isotopic analysis (Kita et al., 2009). After polishing to remove existing analysis pits, zircons were then re-imaged in CL, and δ¹⁸O analyses were targeted to the same CL zones as previous analyses, and in many cases directly underlie previous analysis spots.

Garnet and quartz were analyzed in polished thin section by grinding the section into a 25 mm circle, with the target grain or grains within 5 mm of the center. Oxygen isotope standards UWG-2 (garnet; Valley et al., 1995) and UWQ-1 (quartz; Kelly et al., 2007) were epoxied into small holes drilled near the center of the round thin sections, and polished to be co-planar with the sample surface. One highly zoned garnet from the JSE was located at the extreme edge of a rectangular thin section, and so an ∼7 mm × 7 mm square portion of the thin section containing the garnet was cut out and remounted at the center of a new epoxy mount containing UWG-2 and UWQ-1; sample and standard were also polished to be co-planar in this case.

U–Pb analyses were made using the CAMECA ims-1270 ion microprobe at the W. M. Keck Foundation Center for Isotope Geochemistry National Ion Microprobe Facility, University of California, Los Angeles. Analyses were guided by CL images of zircons and were made using an ∼30 µm diameter beam, and instrument conditions as described by Quidelleur et al. (1997). Rare earth elements (La–Lu) and other selected trace elements (Ti, P, Y, Hf, Th and U) were measured using a CAMECA ims-1280 ion microprobe at the WiscSIMS Laboratory, Department of Geoscience, University of Wisconsin–Madison, using an ∼4 nA O^– primary ion beam (23 kV total accelerating voltage) shaped to a diameter of 20–25 µm on the sample surface. Analyses were conducted in monocollection mode using the axial ETP electron multiplier, applying a combination of energy filtering (40 eV) and high mass resolution (mass resolving power of 5000) as described by Page et al. (2007b).

Oxygen isotope ratios were measured at the WiscSIMS Laboratory, using the method described by Kita et al. (2009) and Valley & Kita (2009). Analyses were conducted using a focused 2·8–3·9 nA primary Cs⁺ beam with a spot size of ∼10 µm. Data were acquired during three separate analytical sessions. Two to four analyses of zircon standard KIM-5 (δ¹⁸O = 5·09% VSMOW; Valley, 2003), garnet standard UWG-2 (δ¹⁸O = 5·80% VSMOW; Valley et al., 1995), or quartz standard UWQ-1 (δ¹⁸O = 12·33% VSMOW; Kelly et al., 2007) were performed before and after every 10–12 sample analyses. The average value of the standards bracketing each block of unknowns was used to correct for instrumental bias using the method outlined by Kita et al. (2009). The external precision, defined as the spot-to-spot reproducibility of the standards before and after a given set of samples, is assigned to the bracketed sample analyses. Precision for zircon analyses ranges from 0·17 to 0·32‰ (2 SD), and for quartz analyses ranges from 0·16 to 0·35‰ (2 SD). The full ion microprobe datasets for oxygen isotope analyses, including both sample and standard data, can be found in Supplementary Data Electronic Appendices 1–3 (available for downloading at http://www.petrology.oxfordjournals.org).

Oxygen isotope analysis of garnet, with its multiple solid solutions, must also be corrected for compositional bias. This bias in Fe³⁺-poor garnets, such as those in this study, can be modeled successfully given current analytical uncertainty as a function of grossular content (Page et al., 2010). Garnet samples were corrected following this method using both UWG-2 and a set of secondary garnet standards that compositionally bracket the sample. Secondary standards were analyzed during the same analytical sessions as sample garnets and calibration curves were developed for each session (Electronic Appendix 2). To apply the calibration curves to correct for compositional bias, the cation composition of the sample garnets was determined adjacent to each ion microprobe pit using the CAMECA SX51 electron microprobe at University of Wisconsin–Madison. Garnets were analyzed in point beam mode with an accelerating potential of 15 kV and 20 nA beam current. The counting time was 10 s on peak and 5 s on both sides, off-peak. Natural and synthetic silicate and oxide standards were used. Data were reduced using the Probe for Windows software (Donovan et al., 2009), and oxygen was calculated by stoichiometry. The content (mol %) of garnet end-members for each ion microprobe analysis pit is reported alongside the isotopic data in Electronic Appendix 1. The matrix bias for garnets from the JS eclogite ranges from 0·9 to 1·4‰ relative to UWG-2, and the bias for garnets from the PNP hornblendite ranges from 0·8 to 1·2‰ relative to UWG-2. These corrections are substantially smaller than those required for zoned Ca- and Fe³⁺-rich skarn garnets, and are much smaller than the isotopic zoning observed within the sample garnets. Analytical precisions from repeated analysis of the UWG-2 standard range from 0·17 to 0·42‰ (2 SD), and are added in quadrature with uncertainties in the correction scheme of ∼0·3‰ (Page et al., 2010), yielding a combined measure of precision and accuracy of ion microprobe analysis of oxygen isotopes in garnet ranging from 0·34 to 0·51‰ (2 SD).

RESULTS

Zoned garnet

Junction School

Garnets from the JS eclogite are characterized by prograde zoning and, in some cases, dramatic evidence of multiple episodes of resorption and regrowth. The degree of retrograde zonation varies on the thin-section scale. Figure 2 shows compositional images and a cation and oxygen isotope traverse of a retrogressed garnet (JS-1) from the JS eclogite. In high-contrast BSE images (Fig. 2a) the average atomic number (Z) of the garnet can be seen to decrease slowly from the core of the garnet until there is a sharp decrease in Z ∼70 µm from the rim. An X-ray map (Fig. 2b) and compositional traverses (Fig. 2c) show that the garnet has a Mn-rich center (Alm₅₇Grs₂₇Pyp₆Sps₁₀) with a characteristic bell-shaped Mn prograde zoning profile coupled with slowly increasing Fe and Mg contents and relatively constant Ca concentrations (Alm₆₀Grs₂₇Pyp₁₂Sps₁) in the center region of the garnet. Representative cation analyses from this traverse are presented in Table 1. This core region contains inclusions of phengite and omphacite that together with the zoned garnet record prograde pressures and temperatures (Page et al., 2007a). The discontinuity between the prograde core region and the adjacent low-Z annulus is undulatory and contains deep embayments into the garnet core, indicating a period of garnet dissolution (Fig. 2a). The low-Z annulus itself is characterized by a sharp increase in Mn and decrease in Fe (Alm₅₄Grs₂₉Pyp₁₃Sps₄), and fills in a portion of the resorbed garnet. The outer margin of the low-Z annulus is another sharp compositional discontinuity leading to a high-Z annulus that records a spike in Fe, a decrease in Ca and Mg, and a drastic reduction in Mn (Alm₆₅Grs₂₇Pyp₈Sps₀). This high-Fe region appears to have filled much of the remaining resorbed garnet, and renewed growth along crystal faces (Fig. 2a). The inner portion of the high-Z annulus contains inclusions of phengite and omphacite that when combined with the host garnet yield blueschist-facies pressures and temperatures (Page et al., 2007a). Outside the high-Z annulus, a final intermediate-Z stage is visible (Fig. 2a). This outermost rim is characterized by a decrease in Fe and rise in Ca (Alm₅₉Grs₂₉Pyp₁₀Sps₂), and the reappearance of some Mn (Fig. 2b and c).

BSE images of the garnet rims record an even more complex story at a spatial resolution smaller than the 5 µm steps of the outer 100 µm of the traverse in Fig. 2c. Figure 3 shows a 1 µm step traverse performed on a rim section (Fig. 2a, B–B′) showing high variability. The inner margin of the low-Z annulus is defined by a peak of Mn and Ca that matches the sudden decrease in Fe content. The decrease of Mn that follows the peak is mirrored by an increase in Fe. The high-Z annulus is marked by an initial increase in Fe and decrease in Mg; Mg thereafter increases steadily towards the rim. BSE variation within both the high-Z annulus and the rim appear to be controlled by complementary fluctuations in Ca and Fe content, and the reappearance of Mn at the outermost (intermediate-Z) rim.

The Junction School eclogite garnets are also zoned in their oxygen isotope composition. Oxygen isotope data from both garnets analyzed are summarized in Table 2, with representative analyses of the different zones in Table 1. All analyses can be found in Electronic Appendix 1. Although the locations of the isotopic analyses were guided by BSE images and X-ray maps, analysis locations were chosen on a reflected light image of an Au-coated sample, leading to occasional analyses of inclusion-bearing regions (recognized by SEM after analysis, data excluded) and mixtures between garnet zones (included). The core region of garnet JS-1 (Fig. 2c) is homogeneous in δ¹⁸O within analytical error (4·3 ± 0·5‰, 2 SD, n = 5). Analyses made in the low-Z annulus (Traverse A–A′ in Fig. 2c at 50 µm, 4·0‰) and the high-Z annulus (930 µm, 3·9‰; 975 µm, 4·3‰) yield identical (within error) results to each other and the core analyses. The outermost two points on this traverse yielded substantially higher δ¹⁸O of 6·7‰ (at the end of the traverse) and a mixed analysis (a combination of the outermost rim and high-Z annulus) yielding 6·0‰ at the beginning of the traverse.

To better constrain the nature of isotopic zoning in the garnet rims, a number of analyses were made on a second garnet (JS-2) from the same sample and are shown superimposed on a high-contrast BSE image in Fig. 4. This garnet records a very similar cation zoning pattern to JS-1, with low- and high-Z annuli filling embayments in the partially resorbed core region. Similar 2–3 µm scale oscillations in BSE are visible within the high-Z annulus. Unlike JS-1, a small embayment into the high-Z annulus can be seen (Fig. 4) that suggests another period of resorption within the high-Z annulus. Multiple small ion microprobe traverses were made in this region to better constrain the rim δ¹⁸O. The δ¹⁸O zonation of this garnet is broadly similar to that observed in garnet from sample JS-1. The δ¹⁸O value of the core, low-Z annulus, and inner portion of the high-Z annulus are identical within analytical error (3·7–4·2‰, n = 10) with the exception of one point (5·5‰). Analyses made in the intermediate-Z rim and immediately outside the resorption line (where visible) in the high-Z annulus have substantially higher and more variable δ¹⁸O (5·6–7·2, n = 12), with one mixed analysis yielding intermediate results (5·1‰). It is clear that the rim is higher in δ¹⁸O value, but it is difficult to ascertain the fine details of the spatial pattern. Higher values appear to cluster in the middle of the rim with lower values towards the inner and outer margins, suggesting either that δ¹⁸O varies across single growth zones or at a scale smaller than the 10 µm spot diameter. Taken together, these garnets preserve in their cation chemistry a history of at least two periods of dissolution and regrowth. Garnet cores and annuli record homogeneous oxygen isotope ratios throughout their early history, with a significant increase in δ¹⁸O recorded in the outermost rims that crystallized after the second resorption episode.

Panoche Pass

Cation zoning in the Panoche Pass hornblendite garnet is less variable than the Junction School eclogite. BSE imaging (Fig. 5a) records somewhat patchy zoning in the garnet core, with a brighter center and darker outer region (less distinct than the low-Z annulus of the JS eclogite sample), coupled with a 50–100 µm thick high-Z rim. A Mn X-ray map (Fig. 5b) shows slight enrichment in the core of the garnet with a relatively high-Mn rim coupled with faint radial zoning. A compositional traverse (C–C′, Fig. 5a and c) shows a core composition of (Alm₆₀Grs₂₆Pyp₁₀Sps₄) with gradually decreasing Fe and Mn and increasing Mg, coupled with constant Ca, moving rimward (Alm₅₈Grs₂₅Pyp₁₄Sps₃). Closer to the rim, Mn increases gradually whereas Ca remains stable, and Fe continues to decrease and Mg continues to increase (Alm₅₅Grs₂₄Pyp₁₆Sps₆). This outermost region of the garnet core has less well-defined margins than the zoning in the JS eclogite garnet, but is visible as the low-Z portions of Fig. 5a that have elevated Mn in Fig. 5b. Finally, the outermost garnet rim is characterized by an abrupt jump in Fe coupled with decrease in Ca (Alm₅₉Grs₁₉Pyp₁₆Sps₆), followed by a gradual increase in Ca and decrease in Fe towards the edge of the crystal (Alm₅₆Grs₂₃Pyp₁₆Sps₆), whereas Mn and Mg remain relatively constant. The thickness of the outer rim is highly variable given the resorbed nature of these garnets.

Three garnets (PNP-1, PNP-2 and PNP-3) from the Panoche Pass hornblendite were analyzed for their oxygen isotope composition and were also found to be zoned (Tables 1 and 2; Electronic Appendix 1). In traverse C–C′ (Fig. 5c), the garnet core from PNP-1 (including Mn-enriched edges of the core region) is homogeneous in δ¹⁸O (11·1 ± 0·3‰, n = 7). However, the isotopic composition of the outer rim is strikingly different, although similarly homogeneous (6·5 ± 0·4‰, n = 5). The other garnets measured were found to have a similar pattern, with the addition of distinct patches of intermediate δ¹⁸O in the low-Z outer portions of the cores close to the rims. Figure 6 shows a high-contrast BSE image near the rim of garnet PNP-2. The core δ¹⁸O in this garnet is similar to or slightly higher (11·3–11·7‰) than that of PNP-1 in Fig. 5c. The thin, high-Z outer rim of the garnet has a similarly low δ¹⁸O (<6·8‰, constrained by a single analysis that overlaps with the core). In this traverse near the edge of the ∼1 mm garnet, the gradient in δ¹⁸O at the rim–core margin is constrained to 4·5‰ over 12 µm. Unlike garnet PNP-1 (Fig. 5), PNP-2 was found to contain faint ∼100–200 µm diameter patches of intermediate δ¹⁸O (8·9–9·5‰, Fig. 6). These intermediate δ¹⁸O zones appear to be small domains (perhaps connected in the third dimension) and not additional concentric zoning. These patches (Alm₆₂Grs₂₅Pyp₁₂Sps₁) are compositionally similar to the surrounding garnet (Alm₆₂Grs₂₆Pyp₁₁Sps₁), and are only faintly visible in high-contrast BSE owing to a difference in average Z of ∼0·1% (Fig. 6).

δ¹⁸O of matrix minerals

The δ¹⁸O of the whole-rock and selected matrix phases from the JS eclogite and PNP hornblendite were analyzed by laser fluorination on ∼2 mg samples (Table 3; Fig. 7). Bulk samples of garnet from eclogite (δ¹⁸O = 4·8 ± 0·1‰, 2 SD, n = 4) and hornblendite (10·0 ± 0·1‰, 2 SD, n = 2) do not reflect the significant isotopic heterogeneity apparent in the ion microprobe analyses, but have intermediate δ¹⁸O between rim and core values as determined by ion microprobe. Bulk garnet δ¹⁸O is closer to that of the cores, as would be expected for mixed analyses of small volumes of rim material with larger volumes of core material. Omphacite from the JS eclogite (5·8 ± 0·1‰, 2 SD, n = 2) and hornblende from the PNP hornblendite (6·1 ± 0·1‰, 2 SD, n = 2) are substantially different from the coexisting garnet, the other major phase in these rocks. Given the small degree of fractionation between pyroxene and garnet, and hornblende and garnet, even at low temperatures (e.g. Kohn & Valley, 1998a, 1998b), these differences imply disequilibrium. Retrograde titanite has δ¹⁸O = 5·3‰ in the eclogite sample and 4·7‰ in the hornblendite sample. The δ¹⁸O values of whole-rock powders are 6·3‰ (eclogite) and 7·2‰ (hornblendite). The combined mode of garnet and omphacite, or hornblende, is greater than 90% in these samples. Although the hornblendite whole-rock δ¹⁸O value is a plausible combination of garnet and hornblende δ¹⁸O values, the whole-rock δ¹⁸O value of the eclogite sample is higher than those of garnet and omphacite, indicating the presence of some volumetrically minor, but significantly ¹⁸O-enriched reservoir.

Texturally late quartz from the JS eclogite was analyzed for δ¹⁸O by ion microprobe (Table 2; Electronic Appendix 1). Matrix quartz in the JS eclogite occurs along grain boundaries between garnet and omphacite (Figs 1c and 8). Polygonal crystals grade into micrometer-scale veins that appear to wet grain boundaries between zoned omphacite crystals. Texturally late quartz has uniformly high δ¹⁸O (25·4 ± 0·2‰, 2 SD, n = 4, Fig. 8), providing an explanation for the high whole-rock values described above.

Zoned zircon

Junction School

Zircon in the Junction School eclogite occurs as inclusions in garnet, omphacite, epidote and titanite, as well as between grains of matrix phases. Inclusions within zircon grains are generally small (∼1 µm) and are not identifiable in thin section, either because they were plucked during polishing, or because they are not mineral but fluid inclusions. Zircon found along the margins of matrix minerals, as inclusions in texturally late epidote, chlorite and titanite, or in the rims of omphacite or garnet (Fig. 1b and c) is typically euhedral. In Fig. 1b euhedral zircon is found in the outer acmite-rich rims of matrix omphacite, and in Fig. 1c an aggregate of small euhedral zircon grains is found within the blueschist-facies high-Z annulus in garnet. These zircon grains or aggregates are nearly featureless in BSE images and have a diffuse oscillatory zoning in CL. Rare, larger zircons separated from the Junction School eclogite are clear and subhedral, 150–200 µm in length, with poorly developed terminal pyramids. They contain inclusions of phengite, omphacite, quartz, epidote, and chlorite. Zircons separated from the JS eclogite are similar in morphology, BSE, and CL characteristics to texturally late zircon observed in thin section, but are much larger. Separated zircons are relatively featureless in BSE images, but in CL images have distinct cores surrounded by fine-scale, but somewhat diffuse, oscillatory zoned rims (Fig. 9a–f). Many separated zircons appear to be fragments of longer crystals with chevron-shaped oscillatory CL zones appearing to grow in one direction (Fig. 9a–c). Zircon cores (when apparent) are featureless and dark (Fig. 9d), have faint, diffuse zoning (Fig. 9e) or are patchy (Fig. 9f).

The δ¹⁸O values of zircons from the JS eclogite are bimodal and correlate with the zoning patterns identified in CL (Figs 9 and 10a). Ion microprobe δ¹⁸O data are summarized in Table 2 with complete results presented in Electronic Appendix 3. Patchy CL cores are uniformly low in δ¹⁸O (4·6 ± 0·6‰, n = 5). Oscillatory-zoned rims are substantially higher in δ¹⁸O (6·4 ± 0·5‰, n = 8).

Selected zircons were also analyzed from the JS sample for trace elements by ion microprobe (Fig. 10a and b; Table 4). Zircon cores have higher REE (454–2590 ppm) and Y (615–4145 ppm) concentrations than zircon rims (68–110 ppm REE; 141–182 ppm Y). Zircon cores also have higher Th/U ratios (0·27–0·60) than oscillatory-zoned rims (0·01–0·03; Fig. 10a). On chondrite-normalized REE diagrams, zircon cores display a positive Ce anomaly (Ce/Ce* = 13–33) and a negative Eu anomaly (Eu/Eu* = 0·3–0·5, Fig. 10b). Zircon rims are more strongly depleted in LREE and middle REE (MREE), with virtually no Ce anomaly (Ce/Ce* = 0·7–2·4) and a small negative Eu anomaly (Eu/Eu* = 0·5–0·6). Both cores and rims have steep heavy REE (HREE) patterns.

JS zircons were also analyzed for U–Pb isotopes by ion microprobe (Table 5); ages were calculated and Tera–Wasserburg concordia diagrams (Fig. 10c) were constructed with Isoplot 3.0 (Ludwig, 2003). Analyses were not corrected for common lead, and lie on mixing lines between common lead (Mattinson, 1986) and the reported concordia intercept age. U–Pb isotopic analysis of zircon from the JS eclogite was made difficult by the low concentrations of U (rims 5–24 ppm; cores 11–68 ppm). Uranium and thorium concentrations measured during the geochronology session are similar to those measured during the trace element session (Tables 4 and 5), with the exception that one rim geochronology analysis yielded 55 ppm U and a Th/U ratio of 0·5. Unfortunately, this grain was plucked during sample preparation after U–Pb analysis and could not be analyzed further. With the exception of this grain, zircon rims analyzed for geochronology have an average U concentration of 9 ppm with Th/U = 0·01–0·03. Previous attempts to date Franciscan eclogite metamorphism by U–Pb on titanite (Mattinson, 1986) were also problematic because of the extremely low levels of U in that phase. Although chemically and texturally distinct, zircon cores and rims from the JS eclogite do not have a resolvable difference in age, with cores yielding a low-precision concordia intercept age of 141 ± 22 Ma (n = 8, MSWD = 2·4) and rims yielding 144 ± 10 Ma (n = 9, MSWD = 1·3).

Panoche Pass

Zircon in the Panoche Pass hornblendite occurs in thin section along grain boundaries at the rims of matrix minerals or along fractures (Fig. 1e and f). Zircon separated from the PNP hornblendite is clear, generally euhedral, and 150–200 µm in length, but with a slightly rounded appearance. It is substantially more abundant than that separated from the JS eclogite. CL images show oscillatory and sector zoning (Fig. 9g–j), but do not have the distinct core–rim textural differences visible in the JS zircons. In one case (Fig. 9i) a diffuse oscillatory-zoned rim appears to truncate more distinct zoning in the core region, implying a multi-stage growth history for this grain. Panoche Pass zircon has very few inclusions, although one large (∼20 µm) titanite inclusion is present in one zircon. No zircons similar in size to the separated fraction have been observed in thin section.

The δ¹⁸O values of zircons (Figs 9 and 10d) from the PNP hornblendite are summarized in Table 2 with the complete dataset presented in Electronic Appendix 3. In general, zircon δ¹⁸O is higher (8·2 ± 1·1‰, n = 22) in the PNP sample than in the core or rim populations of the JS eclogite, although no systematic zoning pattern correlated with CL patterns was observed.

The trace element chemistry of selected zircons is reported in Table 4, with analysis locations indicated in Fig. 9. As with oxygen isotopes, trace element analyses reveal a relatively homogeneous composition, with Y (136–404 ppm) and REE (107–323 ppm) intermediate between the core and rim analyses of the JS eclogite. Th/U ratios range from 0·02 to 0·13 (Fig. 10d). On a chondrite-normalized REE diagram, zircons have a small positive Ce anomaly (Ce/Ce* = 6·5–17) and a negative Eu anomaly (Eu/Eu* = 0·7–0·9), and generally steep HREE (Fig. 10e).

Zircons from the Panoche Pass hornblendite were also analyzed for U–Pb isotopes by ion microprobe (Table 5, Fig. 10f). Th and U concentrations and Th/U ratios were consistent with those measured along with the other trace elements. The PNP hornblendite zircons have higher U concentration than those from the JS eclogite, and provided more precise results and concordant analyses with lower counting times, yielding an intercept age of 147 ± 4 Ma (MSWD = 0·94, n = 16).

DISCUSSION

Compositional and isotopic zoning in garnet and zircon have long been used to reconstruct pressure, temperature, fluid and time histories of metamorphic rocks. Zircon is a highly refractory mineral with slow intracrystalline diffusion of both cations (e.g. Hanchar & Hoskin, 2003; Cherniak, 2010) and oxygen (Valley et al., 1994; Watson & Cherniak, 1997; Valley, 2003; Page et al., 2007c; Bowman et al. 2011). Garnet is also an extremely robust mineral with sluggish diffusion rates for both oxygen and cations, particularly at the low temperatures (<600°C) recorded in these rocks (e.g. Coghlan, 1990; Vielzeuf et al., 2005; Ganguly, 2010). The snapshot of geochemical conditions provided by these minerals during their crystallization is unlikely to be substantially modified, barring wholesale dissolution, particularly at the low temperatures found in subduction regimes. Indeed, the very heterogeneities observed in both cation and isotopic compositions are testimony to the ability of these minerals to preserve geochemical information. Correlation of δ¹⁸O data and REE patterns allows a synthesis of records preserved in the zircon and garnet. U–Pb data from zircons allow this relative history to be placed in the context of absolute chronology and compared with other Franciscan tectonic blocks. Thermobarometry on inclusions in zoned garnets (where available) can be used to tie this record to the P–T path of these rocks, generating whole or partial P–T–t–fluid paths in a subduction tectonic setting.

Cation zoning in garnets

Junction School

Garnets from the JS eclogite record a rather complete record of prograde growth during subduction from early blueschist-facies conditions in the Mn-rich cores (10 kbar, 400°C) to the highest-grade eclogite-facies conditions (22 kbar, 550°C) recorded in the outer-core region just inside the high-Mn, low-Z annulus (Page et al., 2007a). This prograde growth zonation is truncated by an irregular boundary followed by a sharp compositional shift of elevated Mn and depleted Fe (Figs 2 and 3). This increase in Mn is both outside the resorption boundary and very sharp, unlike the commonly observed (in higher grade rocks) upward turn seen inside the resorption boundary characteristic of Mn back-diffusion into the garnet during dissolution.

There are several possibilities for the origin of Mn enrichment in this annulus. The source of Mn could be locally derived from previously resorbed garnet or another Mn-bearing phase that was liberated during the lacuna in garnet growth and scavenged by the neoformed garnet. Alternatively, the Mn could have been introduced into the rock by an external fluid during garnet resorption or at the beginning of new garnet growth. The sharp rise at the resorption boundary and prolonged elevation in Mn content of this annulus may be evidence for an external source of Mn. Annuli of compatible elements in garnet that have been interpreted as having formed in a closed system typically have a sharp peak in Mn followed by a smooth decrease in Mn consistent with a Rayleigh model of compatible element fractionation (e.g. Yang & Rivers, 2003). The plateau in spessartine content visible in the low-Z annulus (Fig. 3) and the oscillating behavior in the outermost rim are not consistent with a simple dissolution–regrowth history, despite the presence of resorption boundaries inside the Mn annuli. The Mn source may have been in part the resorbed garnet, but some substantial, fluid-mediated redistribution of Mn is the most likely cause of these Mn annuli. The low-Mn, high-Z annulus located between both Mn-enriched annuli contains a blueschist-facies mineral assemblage (Page et al., 2007a), requiring that rim development occurred after substantial exhumation.

Oscillatory-zoned garnets were first reported from Franciscan eclogites by Dudley (1969), from the Tiburon peninsula. Their presence there was recently reconfirmed by the optical observations of Tsujimori et al. (2006a). Although zoning patterns in eclogite garnet have been described as a kinetic phenomenon (Ghent, 1988), more recently, oscillatory zonation in garnets from mélange-hosted eclogite has been interpreted to record inflections in the exhumation P–T path (García-Casco et al., 2002; Blanco-Quintero et al., 2011). The zoning described by García-Casco et al. (2002) is similar to that seen here. However, that reported by Blanco-Quintero et al. (2011) is much broader than that observed in the other studies described above, with zones as wide as 500 µm in a 5 mm garnet. One possible origin for the >10 µm-scale zoning in the JS eclogite is a similar series of inflections in the P–T path. However, extremely fine-scale zoning found exclusively in the rims of eclogite garnets (Dudley, 1969; García-Casco et al., 2002; Tsujimori et al., 2006a) is more likely to preserve a record of changing fluid conditions during exhumation, similar to the record preserved in hydrothermal skarn garnets (e.g. Jamtveit et al., 1993; Jamtveit & Hervig, 1994; Clechenko & Valley, 2003; Page et al., 2010; D’Errico et al., 2012).

Panoche Pass

Cation zoning in the PNP hornblendite garnet is much more diffuse than that found in the JS eclogite. The core regions of garnet preserve relatively subdued prograde zonation with increasing Mg and decreasing Fe, suggestive of growth under increasing T. There is no existing detailed P–T work on inclusions in this garnet, which are limited to hornblende, quartz, ilmenite and apatite. The faint zones of Mn enrichment inward from the rim (Fig. 5c) suggest recrystallization along cracks or perhaps dissolution–recrystallization phenomena, leaving a somewhat ‘ghost-like’ high-Z core with the diffuse margins visible in Fig. 5a. The broad and relatively homogeneous Mn-rich rim is not consistent with diffusion of Mn into the garnet during resorption because of its flat-topped morphology in compositional traverse (Fig. 5c). If it were the product of regrowth from a fixed reservoir after resorption, an asymmetrical peak would be expected, followed by depletion of Mn owing to its removal from the matrix. The homogeneity of the Mn contents in the rim suggests growth with a continual supply of Mn, either at the expense of a dissolving Mn-rich phase (perhaps ilmenite) or from an external fluid. The sudden increase in Fe with concomitant decrease in Ca at the outer rim may be due to a fluid change—for example, by entrainment in a reducing environment such as a serpentinite, or owing to renewed crystallization at a different P and T.

Oxygen isotopes

The oxygen isotope results for all mineral and whole-rock analyses (with the exception of quartz in the JS eclogite; Table 2) are within the range commonly observed for eclogites and related rocks (e.g. Putlitz et al. 2000; Russell, 2011; Russell et al. 2013, and references therein). The JS eclogite and PNP hornblendite are basaltic in composition, and are generally believed to have a mid-ocean ridge basalt (MORB) protolith (Hermes, 1973; Sorensen et al., 1997). Unaltered gabbros from ophiolites and unaltered MORB record a consistent whole-rock δ¹⁸O of 5·6‰, whereas basalts that experienced low-T alteration with seawater are enriched in ¹⁸O (δ¹⁸O_WR = 5·6–14‰) and gabbros that interacted with water at high T take on lower values of δ¹⁸O (2–5·6‰; see Eiler, 2001, for a review). Previous rock- and mineral-scale studies have demonstrated that many eclogites and related rocks preserve this seafloor alteration signal through high-P metamorphism and exhumation (e.g. Putlitz et al., 2000, 2001; Früh-Green et al., 2001; Fu et al., 2010; Halama et al., 2011; Russell et al., 2013).

Disequilibrium among minerals

A first-order question in the interpretation of the oxygen isotope record of these rocks is an assessment of equilibrium and the scale of exchange between the minerals. Even at the low metamorphic temperatures (<600°C) typical of Franciscan high-grade blocks, equilibrium fractionation between the major phases in these feldspar-absent mafic lithologies is small. The fractionation between omphacite and almandine-rich garnet [Δ¹⁸O (omph–grt)] was found to be 0·02–0·46‰ for 600–650°C alpine eclogites (Kohn & Valley, 1998b). Larger fractionations [Δ¹⁸O (omph–grt) = 0·72‰ at 500°C] were found among the lower T eclogites of Syros, Greece (Putlitz et al., 2000). Fractionation between the core and rim values of the JS eclogite garnet and bulk analyzed omphacite yield Δ¹⁸O (omph–grt core) = 1·8‰, and Δ¹⁸O (omph–grt rim) = −0·2‰. The garnet cores are clearly out of equilibrium with the bulk omphacite. Although the fractionation between garnet rims and bulk omphacite could be consistent with high-T equilibrium, given the results of Kohn & Valley (1998b), a late high-T event is not recorded in Franciscan blocks and suggests oxygen isotope disequilibrium between garnet rims and matrix omphacite as well. Matrix omphacite in this rock is strongly zoned in cations, and may preserve a heterogeneous δ¹⁸O record to complement that preserved by the garnets. Texturally late titanite (δ¹⁸O = 5·3‰; Table 3) similarly must be in disequilibrium with all garnet core and rim values at the T recorded by the inclusions (Valley et al., 2003; Page et al., 2007a). Finally, the extremely elevated δ¹⁸O void-filling quartz (25‰) would require T < 100°C to be in equilibrium with any other phase measured in the rock, and is probably a late, fluid-precipitated phase (Valley et al., 2003). The high δ¹⁸O value of the quartz is consistent with low-temperature (∼200°C) equilibrium with vein calcite (δ¹⁸O ∼20‰) found within Franciscan host sediments (Sadofsky & Bebout, 2001).

Measured δ¹⁸O of the phases present in the PNP hornblendite similarly reveals profound disequilibrium. The fractionation of oxygen isotopes between hornblende and almandine-rich garnet is negligible at any reasonable temperature (Kohn & Valley, 1998a). Therefore, garnet cores are in extreme disequilibrium with bulk matrix hornblende [Δ¹⁸O (hbl–grt core) < –5‰], whereas the thin, low-δ¹⁸O rims could be in equilibrium with the matrix hornblende [Δ¹⁸O (hbl–grt rim) = −0·3‰]. Texturally late matrix titanite may also be in equilibrium with garnet rims, with Δ¹⁸O (sph–grt rim) = –1·7‰ yielding T ∼500 ± 80°C, or might be in disequilibrium if they were formed at lower temperatures (Valley et al., 2003).

The mineral-scale disequilibrium found in these two samples is consistent with recent results from high-grade blocks throughout the Franciscan (Pincus et al., 2010; Errico et al., 2013). The evidence for late fluid exchange, extreme heterogeneity and disequilibrium recorded in these Franciscan metamorphic rocks stands in contrast to the apparently pristine ocean-floor alteration record preserved in most eclogites (e.g. Putlitz et al., 2000, 2001; Früh-Green et al., 2001; Halama et al., 2011).

Garnet zoning

Although δ¹⁸O zoning in garnet has been recognized in skarn garnets (e.g. D’Errico et al., 2012), and in an ultrahigh-pressure metasediments (Sobolev et al., 2011), such zoning in almandine-rich garnets in pelitic and mafic lithologies is reported rarely in the literature, and then is generally <1‰ (Kohn et al., 1993; Skelton et al., 2002; Martin et al., 2006; Lancaster et al., 2009). One notable exception to this is the ∼4‰ zoning reported in garnets from migmatites in the Pyrenees (Vielzeuf et al., 2005). A recent study of eclogite garnets from nine European localities has revealed the presence of up to 2·5‰ zoning with lower δ¹⁸O cores and higher δ¹⁸O rims, suggesting a late prograde fluid interaction with a higher δ¹⁸O, more mantle-like, reservoir (Russell et al., 2013). The ∼3‰ zoning observed in the JS garnets and the ∼5‰ zoning observed in the PNP garnets are thus rather dramatic for metamafic rocks. Recently, Errico et al. (2013) have documented similar ∼3‰ zoning in other Franciscan eclogite garnets. Because of the low temperatures recorded in these samples (<600°C) and the slow intracrystalline diffusion of oxygen in garnet, the zoning represents a primary record of crystallization (or, locally, recrystallization; e.g. Martin et al., 2011a; Raimondo et al., 2012) conditions as opposed to diffusional resetting.

Ion microprobe oxygen isotope analyses of garnets from the JS eclogite yield a bimodal distribution with low-δ¹⁸O cores and high-δ¹⁸O rims (Fig. 11a). The low-δ¹⁸O (∼4‰) portions of the garnets include the garnet cores, as defined by cation zonation, the low-Z annulus, and the blueschist-facies high-Z annulus (Figs 2 and 4), which have broadly homogeneous δ¹⁸O. Only the outermost garnet rims, outside a second resorption boundary, have elevated δ¹⁸O (6–7‰, Fig. 4).

The JS garnet core δ¹⁸O of 4·3‰ is low for the range of garnet values recorded in eclogite, and is consistent with that recorded in metagabbros that have experienced higher-temperature hydrothermal alteration at a mid-ocean ridge. Feldspar-free metamorphic lithologies such as those in this study are characterized by relative homogeneity of δ¹⁸O between phases because of the small fractionation between the limited number of minerals that make up the rock (Kohn & Valley, 1998a, 1998b). In general, the δ¹⁸O of eclogitic garnet is typically less than 0·5‰ lower than the equilibrium whole-rock composition even at T < 600°C (e.g. Putlitz et al., 2000). Most JS garnet core analyses are more than 0·5‰ lower in δ¹⁸O than the range of unaltered MORB whole-rocks (Fig. 11a), suggesting that the MORB protolith of the JS eclogite was altered by seawater at medium to high temperatures, similar to ophiolitic gabbros (Eiler, 2001). Because of the absence of a significant high-δ¹⁸O mineral reservoir in this eclogite, the only plausible explanation for the extreme (2–3‰) shift in δ¹⁸O at the garnet rim is the influx of an externally derived, high-δ¹⁸O fluid.

Panoche Pass hornblendite garnets preserve three δ¹⁸O domains (Fig. 11b). The δ¹⁸O values of garnet cores are 10·9–11·7‰, recrystallized patches are 8–9‰, and the outermost high-Z rims yield uniformly low ratios (6·2–6·8‰). Negligible oxygen isotope fractionation between hornblende and almandine garnet allows direct comparison among these minerals and equilibrium whole-rock compositions. The elevated ¹⁸O of PNP garnet cores is well above the value of unaltered MORB (Fig. 11b), requiring that a MORB protolith be altered at low temperatures on the ocean floor, prior to metamorphism. The outermost low-δ¹⁸O rim is in equilibrium with matrix hornblende, and records the influx of a less ¹⁸O-enriched fluid. Given the slow rate of intracrystalline oxygen diffusion in garnet and textural evidence of resorption, both before and after rim formation, the intermediate δ¹⁸O patches are best explained as zones of recrystallization, perhaps along healed fractures. Although texturally less distinct than the evidence of zoning and resorption recorded in the JS garnets, these intermediate δ¹⁸O domains are likewise best explained by resorption and regrowth of garnet in a changing δ¹⁸O fluid environment. These patches may have formed in a dissolution–reprecipitation process similar to that documented by Martin et al. (2011a).

Zircon zoning

Zircon from the JS eclogite preserves a bimodal distribution of oxygen isotope ratios (Fig. 11c), similar to those found in the garnets of the same rock. The distinctive core–rim CL textures present in JS zircons are, in some ways, similar to magmatic zircons with metamorphic rims. However, a magmatic origin at a mid-ocean ridge is not consistent with the oxygen isotope ratios (4·6‰) found in the cores of JS zircons. Igneous zircons occurring in oceanic gabbros display a distinctively narrow range of δ¹⁸O (5·2 ± 0·5‰; Cavosie et al., 2009; Grimes et al., 2011), characteristic of their formation from a mantle reservoir. The oxygen isotope record of zircons from the ocean-floor environment is robust: igneous zircon from the ocean floor can even retain a mantle signature through metasomatic conversion of its host-rock to jadeitite in a subduction environment (Fu et al., 2010). Although some δ¹⁸O values from JS zircon cores overlap the mantle range, most are not consistent with a mantle origin for the JS zircon cores, as would be expected for igneous zircons from MORB.

The oxygen isotope fractionation between zircon and almandine-rich garnet is negligible, allowing the zircon and garnet record to be directly compared (Valley et al., 1994, 2003). All JS zircon core δ¹⁸O values overlap with the core compositions of coexisting garnets (Fig. 11a and c), suggesting that they formed from the same oxygen reservoir as garnets: altered oceanic crust. Zircon rims (δ¹⁸O = 6·4‰) from the JS eclogite are more enriched in ¹⁸O than the cores, similar to values found in garnet rims (6·4‰; Fig. 11a and c). Because they have the same δ¹⁸O, zircon rims are in equilibrium with garnet rims, but may also have formed any time after the influx of the high-δ¹⁸O fluid responsible for the isotopic zoning in garnet.

The oxygen isotope ratios of zircons from the PNP hornblendite record a different history. The 8·2‰ δ¹⁸O value of these zircons precludes an igneous origin in equilibrium with the mantle, despite the MORB-like composition of the whole-rock. Similarly, a metamorphic origin in equilibrium with high-δ¹⁸O garnet cores (11·3‰) or low-δ¹⁸O garnet rims (6·6‰) is also ruled out (Fig. 11b and d). The zircon oxygen isotope ratios overlap substantially with intermediate δ¹⁸O value recrystallized zones in garnet (Fig. 11b and d), suggesting that they formed in a similar fluid environment.

Zircon textures, geochronology and trace elements

Texturally, the Junction School eclogite and Panoche Pass hornblendite zircon grains show diffuse oscillatory zoning in CL, features that in other rocks have been attributed to an igneous origin. Despite a superficial similarity to igneous zircon, a number of lines of evidence argue against this conclusion. The formation of oscillatory-zoned zircon is not exclusive to igneous environments, but is also found in zircons formed in high-pressure hydrothermal veins. Liati & Gebauer (1999) and Rubatto et al. (1999) reported metamorphic ages for eclogites based on euhedral, oscillatory-zoned zircon from deformed quartz veins in the Rhodope zone of northern Greece and the Sesia–Lanzo zone of the western Alps, respectively. Rubatto & Hermann (2003) found similar textures in zircon from a vein in eclogites from Monviso in the western Alps. In the case of the JS and PNP zircons, oscillatory zoning is confined to high-δ¹⁸O zircon that crystallized during or after high-pressure fluid metasomatism, and not in equilibrium with primitive mantle-derived magmas at a spreading center.

High-pressure, zircon-bearing vein material contains significant amounts of both rutile and zircon, suggesting that metamorphic fluids at eclogite-facies P–T have the ability to transport Zr and Ti over at least short distances. The formation of vein zircon and the movement of Zr and Ti in aqueous fluids suggest a high fluid/rock ratio over a prolonged period (Nadeau et al., 1993; Philippot, 1993). Alternatively, NaAlSi₃O₈-bearing fluids dramatically increase rutile solubility, and probably also play a role in HFSE mobility in subduction systems (Antignano & Manning, 2008). Alkaline fluids such as those formed during serpentinization also increase Zr solubility and have been shown to result in neoformed zircon in serpentine-hosted eclogite (Rubatto et al., 2008; Martin et al., 2011b), rodingite (Li et al., 2010), and jadeitite (e.g. Tsujimori et al., 2006b; Mori et al., 2011). Fu et al. (2010) found evidence of both igneous zircons with mantle-like δ¹⁸O and low-δ¹⁸O hydrothermal zircons in jadeitites from Japan.

The ages of zircon from both Franciscan samples are further suggestive of a metamorphic origin. The low-precision ages for JS eclogite zircon cores (141 ± 22 Ma) and rims (144 ± 10 Ma) are indistinguishable from one another, and the more precise PNP hornblendite zircon age (147 ± 4 Ma). These ages match the youngest metamorphic ages found in Franciscan blocks, typically associated with garnet blueschist- or blueschist-facies overprinting (Wakabayashi & Dimitru, 2007, and references therein), and are incompatible with an igneous origin.

Although there is no published geochronology for the JS eclogite, the PNP hornblendite has been the subject of several chronological studies that place the new zircon ages in context. Ross & Sharp (1988) performed some of the first ⁴⁰Ar/³⁹Ar geochronology in the Franciscan on a garnet hornblendite from Panoche Pass. They interpreted a 162 ± 2 Ma plateau as the hornblende cooling age, and attributed the total gas age of ∼145 Ma (identical to the zircon age in this study) to late white mica intergrown with the hornblende. Anczkiewicz et al. (2004) produced a Lu–Hf age of 168·7 ± 0·8 Ma for this block, similar to, but slightly older than the hornblende age. The younger hornblende age is probably compromised by the later metasomatism, as any fluid pulse that shifts the bulk δ¹⁸O by ∼5‰ would probably at least partially reset the Ar geochronometer. However, it seems unlikely that the hornblende age dates the last stage of fluid metasomatism, as the higher δ¹⁸O zircons post-date the hornblende age.

In addition to CL textures, oxygen isotopes, and arguments based on geochronology, trace-element chemistry is commonly used as a discriminant between igneous and metamorphic zircon (e.g. Rubatto, 2002). A canonical interpretation of the Th/U systematics of these Franciscan zircons would be that the high Th/U values of the JS zircon cores (Fig. 10a) record an igneous origin. The low U and Th concentration and the generally low Th/U ratios (0·01–0·13) of JS zircon rims and PNP zircons (Fig. 10a and d) are generally interpreted as evidence for a metamorphic origin. However, although Th/U ratios are commonly higher in igneous zircons than in metamorphic zircon, this signature is by no means unequivocal. Metamorphic zircon can preserve elevated Th/U through growth in a closed system, recrystallization of magmatic zircon, or by growth at the expense of a high-Th/U phase (Möller & Kennedy, 2006). Because the preponderance of CL, oxygen isotope, and geochronological evidence points to a metamorphic origin for all Franciscan zircons in this study, our preferred interpretation is that the JS zircon cores are also metamorphic, despite the lack of universally low Th/U ratios.

REE patterns are perhaps the most commonly used tool to relate zircon ages to metamorphic events recorded in the matrix mineralogy, particularly in tracing zircon that has grown during or after the growth of garnet (e.g. Rubatto, 2002; Kelly & Harley, 2005; Lancaster et al., 2009). The compatibility of the HREE in garnet reduces the availability of these elements during zircon growth, and results in a characteristic flat or negatively sloped HREE pattern on chondrite-normalized diagrams, despite the increasing compatibility (with increasing atomic number) of these elements in zircon (Rubatto, 2002). However, despite the abundance of garnet in the rocks in this study, all zircons analyzed have the steep HREE pattern found in zircons from garnet-free lithologies (Fig. 10b and e). A conventional interpretation of these data in garnet-bearing rocks would be that zircons formed before the onset of garnet crystallization. However, the oxygen isotope and geochronological data presented above are similarly incompatible with such an interpretation for all zircons except the JS cores, which may have grown before JS garnet but from the same oxygen reservoir.

Zircons with steep REE patterns may have grown late in the history of both rocks, during periods of garnet resorption, and possibly with an external source of REE. Alternatively, zircon with a steep HREE pattern may have formed simultaneously with garnet containing even higher concentrations of HREE. REE data from both garnet and zircon are required to fully address these possibilities, but are beyond the scope of the current study.

The JS eclogite zircon cores are likely to have grown after ocean-floor metasomatism (judging from the preserved low δ¹⁸O values) during the early stages of metamorphism, before the onset of garnet growth (yielding a steep HREE pattern). Both the JS eclogite and PNP hornblendite garnets display textural evidence of resorption before a shift in δ¹⁸O. The oxygen isotope ratios of JS rim zircons overlap with those preserved in garnet rims, and PNP zircon δ¹⁸O values match the intermediate recrystallized garnet values (Fig. 11). Depending on the HREE composition of the garnet with similar δ¹⁸O values, these zircons may have grown during garnet resorption, or during or after the formation of neoformed garnet with similar oxygen isotope ratios.

Oscillatory-zoned high-δ¹⁸O zircon from Franciscan samples is also notable for the small size of positive Ce anomalies in chondrite-normalized REE patterns (Fig. 10b and e). Trail et al. (2011, 2012) recently developed an experimentally calibrated oxybarometer to allow estimates of f_O2 based on the magnitude of the Ce anomaly in zircon and temperature. The combination of a small Ce anomaly and low temperatures of metamorphism characteristic of the Franciscan yield extremely reducing estimates based on the current calibration of the oxybarometer. Even with a temperature estimate as high as 600°C (the highest recorded by the JS eclogite; Page et al., 2007a), the resulting calculated oxygen fugacities are below the iron–wüstite buffer for JS zircon rims and PNP zircons. Uncertainties in the low concentrations of LREE and extrapolation of experimental results to low temperatures are probably responsible for the unreasonable magnitude of this result. However, extremely reducing, alkaline fluids generated by serpentinization are a possible candidate for metasomatism and Zr mobility in subduction zones (e.g. Johnson & Harlow, 1999; Tsujimori et al., 2005; Evans, 2010). Hydrothermal zircon in metasomatic jadeitite (Tsujimori et al., 2006b; Fu et al., 2010; Yui et al., 2010, 2012; Mori et al., 2011) and metasomatized ultramafic rocks (Dubinska et al., 2004; Li et al., 2010) is perhaps more analogous to Franciscan zircons than zircon in orogenic eclogite. Hydrothermal zircons (discriminated by Th/U and δ¹⁸O) from jadeitite (Mori et al., 2011) and serpentinite (Tsujimori et al., 2006b; Fu et al., 2010) have similarly small Ce anomalies that would result in similarly low oxygen fugacity based on the Trail et al. (2012) calibration. Although these results suggest that the Trail et al. (2012) calibration does not extend to extremely low-T, low-f_O2 conditions, late zircon growth in the JS and PNP samples may have been formed by an influx of low-f_O2 serpentinite-derived, reducing fluids, perhaps the same fluids that caused garnet dissolution.

Fluid history and P–T–t–fluid paths

Taken together, the data described above present a series of metamorphic and fluid events recorded in two different high-grade tectonic blocks within the Franciscan Complex. Federico et al. (2007) have shown that adjacent eclogite and blueschist blocks in a serpentinite mélange in the western Alps can have diachronous P–T histories, and indeed different P–T paths, consistent with the tectonic models of Gerya et al. (2002). New chronological and geochemical data for these complex, polymetamorphic rocks are best evaluated in terms of P–T–t–fluid paths, as each block probably represents a different path through the subduction zone at differing times. The pooled paths can be used to discuss the subduction zone evolution as a whole, rather than directly comparing ages and geochemical traces of fluid between blocks with an assumed common history.

Although the widespread oxygen isotope disequilibrium between the major phases in these rocks might initially cause dismay, it records a wealth of now interpretable data. Integration of oxygen isotope and trace element data measured using an in situ technique on garnet and zircon allows for the development of a robust fluid and tectonic history for each sample that can be presented in the form of a P–T–t–fluid path (Fig. 12), or, in the case of the PNP hornblendite where there is no robust P–T information, a fluid–time path (Fig. 13).

The igneous protolith of the Junction School eclogite was probably a mid-ocean ridge gabbro that was altered by high-temperature, sub-solidus, hydrothermal interaction. High-temperature fluids shifted the protolith δ¹⁸O (whole-rock) to <5‰ before subduction (δ¹⁸O is preserved in zircon and garnet cores formed after the beginning of metamorphism). The low-precision age of the metamorphic zircon cores (141 ± 22 Ma) is consistent with other Franciscan eclogite ages of 150–160 Ma (Wakabayashi & Dumitru, 2007). The P–T record of the subduction path begins with blueschist-facies garnet cores with no evidence of subduction-related fluid metasomatism from an external source, through the peak of eclogite-facies metamorphism over 70 km deep beneath the Franciscan trench (22 kbar, Fig. 12). The garnet record was truncated by resorption at or after the peak of metamorphism, and then begins again after substantial exhumation to 30–50 km depth. The δ¹⁸O of the eclogite garnet remained constant through this first stage of exhumation, with no evidence of exchange with external fluids. If this initial resorption episode was the result of changing fluid conditions reducing the stability of garnet, these fluids were apparently equilibrated with the eclogite and left no lasting mark on the oxygen isotope record of this rock. After another, less-pronounced period of resorption, new garnet growth began, this time with elevated δ¹⁸O (Fig. 12). This metasomatism is approximately dated at 144 ± 10 Ma by high-δ¹⁸O zircon. The P–T conditions of the elevated-δ¹⁸O garnet and zircon rims are not constrained, but the temperature is unlikely to be substantially lower than for the blueschist-facies garnet, given the already low temperatures of ∼400°C. Finally, late hydrothermal fluids precipitated high-δ¹⁸O quartz along grain boundaries in a separate, and probably much later event, perhaps during the final stages of exhumation and emplacement in a sedimentary mélange at low temperatures.

It is unclear why garnet in the JS eclogite stopped growing after the highest P recorded by the inclusions in garnet, was resorbed, then began to grow again at blueschist-facies conditions. One possibility is that the rock moved out of the garnet stability field because of a change in bulk composition, such as metasomatism by an oxidizing fluid stabilizing epidote in favor of garnet (e.g. Donohue & Essene, 2000). Another possibility is that garnet was resorbed after exhumation of the eclogite block and re-equilibration below the pressures required to form garnet, then reburied to blueschist-facies conditions. This process has been described in the Alps by Rubatto et al. (2011) as ‘yo-yo subduction’, and has also been recorded in zoned garnet in mélange-hosted eclogite (Blanco-Quintero et al., 2011). The broad-scale cation zoning in the JS eclogite bounded by resorption boundaries could be a result of changing P conditions and convection in the subduction channel, and the finer-scale oscillations in outermost garnet rims may represent a changing fluid record, similar to a skarn garnet, superimposed on the tectonic zoning. Following this line of reasoning, and given the two resorption boundaries present in these garnets, the arrow in Fig. 12 might be redrawn as a series of loops with P–T conditions returning to somewhere near the origin during the two periods of garnet resorption, and representing three burial–exhumation events in the same subduction channel. Recent work on serpentinites in the Franciscan has also uncovered evidence of 2–3 cycles of subduction–exhumation (e.g. Wakabayashi, 2012). Yo-yo subduction provides a plausible mechanism for the ubiquitous blueschist-facies overprint found in Franciscan blocks (e.g. Essene & Fyfe, 1967; Moore & Blake, 1989), as well as an alternative model for the counter-clockwise subduction P–T path first proposed by Wakabayashi (1990).

Whether or not the eclogite block was on its way back to the surface following the exhumation path shown in Fig. 12, or if it was undergoing repeated reburial beneath the Franciscan trench, at the time of metasomatism, the JS block was no longer part of a coherent subducting slab, but was already a block within the mélange of the subduction channel (e.g. Cloos, 1986; Bebout, 2007). A fundamental difference in this oxygen isotope zonation and that described by Russell et al. (2013) is that the metasomatism in this case happened after the peak of metamorphism and after substantial exhumation. Although the high δ¹⁸O rules out a fluid in equilibrium with a mantle reservoir (e.g. Putlitz et al., 2000), the metasomatizing fluid in this channel could have been in equilibrium with serpentinized ultramafic rocks, sedimentary matrix, or both. The serpentinite body adjacent to the JS eclogite was recently found to have an elevated δ¹⁸O (whole-rock) value of 7·9‰ (Barnes et al., 2013). Other work on Franciscan serpentinite blocks has suggested that they, like the garnet-bearing blocks in this study, were similarly metasomatized by a high-δ¹⁸O fluid (King et al., 2003). Given the degree of metasomatism in the JS eclogite block, and the requisite high fluid/rock ratios needed to generate this shift in δ¹⁸O, the most likely cause is the high volume of fluids probably flowing up the subduction channel, originating from devolatilization of the subducting slab (e.g. Bebout, 2007). The zoning observed in the JS eclogite differs from that observed by Errico et al. (2013) in three other Franciscan eclogites, all of which had lower δ¹⁸O rims than cores, explained by residence in low-δ¹⁸O serpentinites. One possible explanation of this discrepancy is local differences in permeability and fluid–rock buffering within the subduction channel.

Like the Junction School eclogite, the Panoche Pass hornblendite has a basaltic composition probably formed at a mid-ocean ridge. Evidence of low-temperature hydrothermal alteration of the basalt is preserved in the elevated δ¹⁸O (∼11‰) of garnet cores (Fig. 13) that began to grow at ∼169 Ma (Anczkiewicz et al., 2004). Garnet rims (∼6‰) are in oxygen isotope equilibrium with matrix hornblende (∼6‰) that formed at 162 ± 2 Ma (Ar–Ar; Ross & Sharp, 1988). This suggests that garnet growth, resorption, and rim growth occurred in the first 7 Myr after the beginning of garnet growth, recording a 4–5‰ shift in δ¹⁸O (Fig. 13). This rapid metasomatism after the peak of metamorphism is followed by another fluid episode that increases rock δ¹⁸O between 162 and 147 Ma. Zircon growth at 147 ± 4 Ma probably dates this event, synchronous with limited garnet recrystallization, although the δ¹⁸O of the block may have already shifted prior to zircon growth. The decrease and subsequent increase in oxygen isotope ratios in this block records changing fluid environments in the subduction channel, either during exhumation or possibly during multiple exhumation–reburial cycles.

Taken together, the zoned garnets from these two distinct blocks in the Franciscan preserve a consistent story of hydrothermal events at an oceanic spreading center that are preserved through subduction metamorphism. Most of the garnet in these rocks records an unchanging δ¹⁸O, suggesting limited fluid interaction from external sources during prograde metamorphism and garnet growth while these rocks were part of the subducting slab. Garnet rims and texturally late zones of recrystallization record profound shifts in δ¹⁸O consistent with metasomatism at high fluid/rock ratios during exhumation or during exhumation–reburial cycles. However, if cation zoning and resorption in the JS eclogite garnets are a result of exhumation and reburial in the subduction channel, constant δ¹⁸O across the first resorption boundary is suggestive of limited fluid flow through the block even when hosted by mélange (e.g. Ague, 2007). Perhaps the JS eclogite block was larger (or less permeable) during its first reburial, and mechanical changes to the block allowed for high-δ¹⁸O fluid infiltration during a second reburial. If this form of cryptic metasomatism is widespread throughout the Franciscan (and beyond), then a vast unexploited record of subduction zone fluid history is available, provided that the metamorphic rocks that preserve it are closely scrutinized at the sub-grain mineral scale or smaller.

ACKNOWLEDGEMENTS

L. J. Sandler is thanked for assistance in field work. S. Passelaqua and J. Polidora generously allowed access to the Junction School locality. B. Hess, J. Kern, and P. Munk assisted in sample preparation. The authors are grateful to C. E. Henderson and J. Fournelle for assistance with the electron microprobe analyses; to M. Grove and C. W. Carrigan for help with U–Pb analyses; and to M. Spicuzza and N. T. Kita for assistance with oxygen isotope analyses. Comments by J. Barnes, T. Raimondo, an anonymous reviewer, and editor J. Hermann greatly improved this contribution.

FUNDING

This work was supported by the National Science Foundation (NSF) (grant EAR 00-87448) to S.B.M. and E.J.E. and (grant EAR 05-09639) to J.W.V., by a Sigma Xi Grant-in-aid and by the Turner Fund of the Department of Geological Sciences at the University of Michigan to F.Z.P. The electron microprobe and electron microscopes used in this work were funded by the NSF (grants EAR 96-28196, EAR 99-11352, CCLI 00-87895). WiscSIMS is partly supported by the NSF (grants EAR 03-19230, 07-44079, 10-53466) and the US Department of Energy (grant 93ER14389).

SUPPLEMENTARY DATA

Supplementary Data for this paper are available at Journal of Petrology online.

REFERENCES

Author notes