Chemical Geology 314-317 (2012) 9–22

Contents lists available at SciVerse ScienceDirect

Chemical Geology

j ourna l homepage: www.e lsev ie r .com/ locate /chemgeo

Research paper

Using garnet to constrain the duration and rate of water-releasing metamorphicreactions during subduction: An example from Sifnos, Greece

Besim Dragovic a, Leah M. Samanta a, Ethan F. Baxter a,⁎, Jane Selverstone b

a Department of Earth Sciences, Boston University, 675 Commonwealth Ave., Boston, MA 02215, USAb Department of Earth & Planetary Sciences, University of New Mexico, MSC03 2040, Albuquerque, NM 87131‐0001, USA

⁎ Corresponding author. Tel.: +1 617 358 2844; fax:E-mail addresses: dragovic@bu.edu (B. Dragovic), lsa

(L.M. Samanta), efb@bu.edu (E.F. Baxter), selver@unm.e

0009-2541/$ – see front matter © 2012 Elsevier B.V. Alldoi:10.1016/j.chemgeo.2012.04.016

a b s t r a c t

a r t i c l e i n f o

Article history:Received 10 December 2011Received in revised form 14 April 2012Accepted 16 April 2012Available online 24 April 2012

Editor: K. Mezger

Keywords:GarnetDehydrationSubductionSifnosCycladesSm–Nd

We present a method to reconstruct the dehydration flux associated with garnet-forming reactions duringsubduction. Garnet-bearing blueschists from the island of Sifnos, Greece, in the Attic–Cycladic BlueschistBelt are used as a test case to extract information on the timescales of dehydration during subduction. Weuse garnet growth as a proxy for the net dehydration reaction. Thermodynamic (pseudosection) analysis ofa mafic blueschist in this unit (representative of an altered basalt protolith) indicates that garnet grew vianet reaction(s) of the form chlorite+chloritoid+glaucophane+phengite=garnet+pyroxene+lawsonite+paragonite+quartz+H2O. Garnet core and rim chemistry indicates that growth began at 2.0 GPa and 460 °Cand ended at 2.2 GPa and 560 °C. The stabilization of matrix lawsonite (promoted by the bulk chemical shift ofthematrix due to fractionation of the garnet and its inclusions from the system) throughout garnet growth limitsthe amount of water liberated from the rock over this P–T span. The average stoichiometry of the garnet-formingreaction(s) indicates an average molar production ratio of garnet to water of ~1.0:0.7. Given the 11 vol.% garnetobserved in the rock, this analysis indicates a loss of 0.3 to 0.4 wt.% H2O from the bulk rock during garnet growth.Zoned garnet geochronology from these rocks provides a constraint on the dehydration rate from this lithologyduring the time span of garnet growth. Two garnet grains, roughly 1.5 cm each in diameter, were microdrilledbased on major element zoning contours. Three concentric growth zones were extracted from each garnet forSm–Nd geochronology. Very low Nd concentrations in acid-cleansed garnet (0.03 to 0.09 ppm) yielded verysmall samples (~1 ng Nd) that were analyzed with TIMS using a NdO+ with Ta2O5 activator method. Untreated“garnet” powders rich inmineral inclusions fromeach garnet zonewere also analyzed. The powders fall off of thegarnet-matrix isochron, likely indicating age inheritance in the inclusions. Combining acid-cleansed garnet datafrom each garnet, multi-point garnet-matrix isochron ages of 46.50±0.80 Ma for the core, 46.49±0.53 Ma forthe intermediate zone, and 46.46±0.59 Ma for the rim were determined, indicating a brief growth duration of0.04 Ma with an upper bound (2 SD) on growth duration of 1.0 Myr. This equates to a release of 0.3–0.4 wt.%of water due to this reaction in this lithology, and heating of 100 °C, in less than 1.0 Myr. The short time intervalrepresents a focused, rapid pulse of dehydration and heatingwithin the context of the overall subduction descenttimescale of ~10–20Myr.

© 2012 Elsevier B.V. All rights reserved.

1 . Introduction and background

This study presents a new method to reconstruct both the durationand rate of particular dehydration reactions within subduction zones.Water bound structurally within subducted rocks, or contained withinpore spaces of the overlying sediments, is cycled through subductionzones. Global estimates indicate up to 7 wt.% structurally bound waterin sediments (Plank and Langmuir, 1998). Estimates of the amount ofwater entrained in the altered oceanic crust average around 1–2 wt.%but may be as high as 6 wt.% (Peacock, 1993). Staudigel et al. (1989) re-port 2.68 wt.% structurally bound water in Cretaceous age oceanic crust

+1 617 353 3290.manta@HaleyAldrich.comdu (J. Selverstone).

rights reserved.

(their “super composite” metabasalt). The fate of this subducted wateris of paramount importance to intermediate depth earthquake nucle-ation (Kerrick and Connolly, 2001; Hacker et al., 2003; Husen et al.,2003), arc magmatism (Peacock, 1990; Bebout, 1991; Schmidt andPoli, 1998; Straub and Layne, 2003; Wallace, 2005), mantle rheology(Ulmer and Trommsdorff, 1995; Mibe et al., 1999; van Keken et al.,2002), and the global water cycle (Wallmann, 2001; Rüpke et al., 2004).

During subduction, the slab is subjected to changing P and T condi-tionswhichdrive chemical reactions and change thephysical propertiesof the slab. Metamorphic reactions release water that may either betaken up again in the structure of new minerals, or may be expelled asa free fluid into adjacent parts of the slab and/or the overlying mantlewedge. Even small amounts of slab dehydration and resultant water re-lease can decrease the melting temperature of the overlying mantle by100–150 °C, causing roughly 5–20% partialmelting of themantlewedge

10 B. Dragovic et al. / Chemical Geology 314-317 (2012) 9–22

(Bebout, 1991; Gaetani and Grove, 1998; Ulmer, 2001). Equilibriumthermodynamic–geodynamic models make predictions of the amountof dehydration in subduction zones (Peacock, 1990; Kerrick andConnolly, 2001; Hacker et al., 2003; Hacker, 2008; van Keken et al.,2011). The predictions of these models include two key assumptions:1) that natural rocks closely follow the modeled P–T-t paths, and2) that the reaction kinetics are fast enough that equilibrium-based pre-dictions prevail. If either of these assumptions is invalid, then theamount, rate, and location of water released could differ significantlyfrom the predictions. In this study, we present a method to constrainthe amount and rate of water released during subduction of a naturalblueschist, and compare our results to model predictions.

2 . Deriving a dehydration rate from natural samples

In order to place constraints on natural rates and timescales of dehy-dration reactions in subduction zones, we need rocks inwhich such reac-tions have occurred. The predicted dehydration histories of blueschistsand eclogites have been well documented (Schmidt and Poli, 1998;Kerrick and Connolly, 2001; Hacker et al., 2003; Rüpke et al., 2004),with up to 6 wt.%water originally held bymafic protoliths releasedwith-in the sub-arc setting. In natural samples, the challenge of accessing thetemporal record of this progressive dehydration is overcome by linkingthe release of water (which has left the rock system by the timewe sam-ple it) to the production of garnet (which is well preserved in manyblueschists and eclogites). Most garnet-forming reactions in blueschist/eclogite facies conditions involve dehydration of micas, chloritoid,epidote-group minerals, and amphiboles (e.g. El-Shazly and Liou, 1991;Spear, 1993; Okay, 2002; Hacker et al., 2003). Several studies haveshown that cores and rims of large garnet porphyroblasts that preserveprograde chemical zoning can be dated, constraining the duration (andhence rate) of garnet growth (e.g., Christensen et al., 1989; Vance andO'Nions, 1990; Ducea et al., 2003; Pollington and Baxter, 2010). Thus, ifgarnet growth can be linked to a particular garnet-forming reaction(or set of reactions), a dehydration flux over that P–T-t interval can beextracted.

3 . Garnet geochronology

We use the Sm/Nd method to date cores and rims of garnetporphyroblasts. Isochrons between each garnet zone and the surround-ingmatrix provide primary growth ages. Sm/Nd has been used success-fully to date garnet by many workers with significant improvements inrecent years. REE-rich inclusions within garnet still pose an importantchallenge, but this has been dealt with by means of partial dissolutioncleansing procedures (e.g., Amato et al., 1999; Scherer et al., 2000;Baxter et al., 2002; Thöni, 2002; Anczkiewicz and Thirlwall, 2003;Pollington and Baxter, 2011). Relatively uniform distribution of Smand Nd throughout garnet in most cases (e.g., Lapen et al., 2003; Skoraet al., 2006; Kohn, 2009; Pollington and Baxter, 2010) makes Sm/Nd agood choice for dating both the garnet core and rim. In some cases,Sm/Nd zonation in garnets does exist (e.g., Skora et al., 2009) so thisshould still be evaluated on a case by case basis. In the Lu/Hf system, incontrast, the parent Lu is strongly fractionated into the core (e.g., Lapenet al., 2003; Kohn, 2009); as a result, the Lu/Hf system is very sensitiveto the early stages of garnet growth, but less useful for dating later stagesof garnet growth.

It is crucial to our study that the dated garnet retains its prograde agezonation. This may be evaluated by considering the diffusion lengthscalefor Nd within garnet. If the lengthscale for diffusion exceeds the domainsizewewish to sample (i.e. the radial dimension of the core or rimmicro-sample) then there is a chance that prograde garnet growth age informa-tionwill be partially re-equilibrated and compromised (e.g., Ganguly andTirone, 1998). This is broadly analogous to the “closure temperature”concept in that higher temperatures (or small domain sizes) will leadto loss of age information. Several studies have shown that garnet

with >1 mm domain size will experience negligible Nd diffusive ex-change as long as the garnet is not heated above ~700 °C for a significantperiod of time (i.e. millions of years) (Ganguly and Tirone, 1998; Baxteret al., 2002; Tirone et al., 2005; Pollington and Baxter, 2011). Here, wehave limited our sampling to rocks that never experienced temperaturesexceeding 600 °C.

4 . Geologic setting

The island of Sifnos is part of the Cycladic Islands in the AegeanSea of the southeast coast of mainland Greece. The Cyclades are partof the Attic–Cycladic crystalline complex, or ACCC, and they representpart of a subduction-related accretionary complex (Okrusch et al.,1978; Avigad, 1993). The ACCC consists of two major units, separatedby low angle faults (Durr et al., 1978; Matthews and Schliestedt, 1984;Schliestedt and Okrusch, 1988; Okrusch and Brocker, 1990). The upperunit comprises a sequence of unmetamorphosed sediments of Permianto Mesozoic age, ophiolites, Cretaceous to Tertiary greenschist rocks,and granitoids (Bröcker and Franz, 2006). The lower unit, termed theCycladic Blueschist Unit (CBU), consists of a crystalline basement over-lain by thrust sheets of ametamorphosed continentalmargin sequence.

On the island of Sifnos, four distinct lithologic units have been char-acterized (Fig. 1 and, Okrusch et al., 1978; Avigad, 1993; Trotet et al.,2001b; Ring and Layer, 2003; Bröcker and Pidgeon, 2007). From top tobottom, exposed in the northwestern part of the island are an uppermarble complex and a well-preserved eclogite–blueschist unit (EBU).The EBU contains significant lithologic heterogeneity, but it is not amélange. Rather it appears to be a structurally coherent continentalmargin sequence of rock including interbedded quartzites, acidic gneiss,metabasites, and metasediments that all experienced essentially thesame metamorphic history. Stratigraphically lower, in the central andsoutheastern parts of the island are the main marble complex and agreenschist unit, interposed with both marbles and blueschists. The topthree units (upper marble, eclogite–blueschist unit, main marble) arepart of a tectonically coherent unit called the Cherronissos Unit, whereasthe greenschist unit (part of different tectonic unit called the Faros Unit)is separated from the CherronissosUnit by a lowangle fault (e.g. Trotet etal., 2001b).

Many studies have discussed the metamorphic evolution of Sifnos,each involving different approaches, including structural (e.g. Avigad,1993; Lister and Raouzaios, 1996; Trotet et al., 2001b; Rosenbaum et al.,2002; Ring and Layer, 2003), petrologic (e.g. Matthews and Schliestedt,1984; Schliestedt, 1986; Schliestedt and Matthews, 1987; Trotet et al.,2001a; Groppo et al., 2009), and geochronologic (e.g. Altherr et al.,1979; Wijbrans et al., 1990; Bröcker and Pidgeon, 2007).

Earlier work delineated two stages of metamorphism on Sifnos.The first stage involved Eocene HP metamorphism under blueschistto eclogite facies conditions, dated at 48–41 Ma using K–Ar and Rb–Srin white micas (Altherr et al., 1979; Wijbrans et al., 1990). In addition,Forster and Lister (2005) reanalyzed 40Ar/39Ar apparent age spectrawork, determining a blueschist facies metamorphic age of ~41 Ma forwhite micas on Sifnos. Garnets from Sifnos have not been dated before,although garnet Lu/Hf ages of 50–52 Ma from nearby Syros eclogites(Lagos et al., 2007) presumably record a similar Eocene HPmetamorphicevent. This Eocene HPmetamorphic history occurred during the collisionof the Apulian microplate and the Eurasian continent (Avigad, 1993).Ring and Layer (2003) suggest that the entire Cycladic Blueschist Unitwas formed as one of a series of successive underplatings, this one initiat-ed at 60–55Ma, resulting in EoceneHPmetamorphism. Based onmodernsubduction rates and microplate reconstructions, Ring and Layer (2003)estimate the paleo-subduction convergence rate at ~2–3 cm/year witha dip angle of ~10–15°. This indicates a fairly slow prolonged descent ofsubducted lithologies to ~70 km maximum depths lasting ~10–20 Ma.Other work in the Cycladic Blueschist Unit suggests that subductioncould have occurred as early as 80 Ma (see Bröcker and Keasling,2006 for a review). Thesewell-preserved Eocene age HP rocks comprise

Fig. 1.Map of Sifnos, Greece showing sample location (star). Stratigraphic column shows relationship between major units on Sifnos. The Cherronissos and Faros units are separatedby a low angle fault.Map is modified after Matthews and Schliestedt (1984) and Trotet et al. (2001b).

11B. Dragovic et al. / Chemical Geology 314-317 (2012) 9–22

the eclogite–blueschist unit in the northern part of Sifnos, and includeblueschist and eclogites from mafic compositions, jadeite-gneisses fromfelsic compositions, and glaucophane-bearing schists and marbles fromthe sedimentary layers (Schliestedt and Okrusch, 1988; Mocek, 2001).

The second stage of metamorphism involved Miocene greenschistto amphibolite facies overprinting of the precursor blueschist to eclo-gite facies metamorphism. This metamorphism ranges in age from 24to 18 Ma using K–Ar, Rb–Sr (Altherr et al., 1979) and 40Ar/39Ar(Wijbrans et al., 1990; Forster and Lister, 2005) in white micas fromthe greenschists in the southern part of the island. This is related to ex-humation caused either by retreat of the Hellenic subduction zone(Avigad, 1993), or by extrusion of the Anatolian microplate causedby the collision of the Arabian plate with Eurasia (Gautier and Brun,1994; Ring and Layer, 2003). Overprinting of the southern part ofthe islandwas determined to be enhanced by infiltration of an ascend-ing fluid phase (Bröcker, 1990; Breeding et al., 2003). Due to the factthat the northern part of the island is sandwiched between relatively im-permeable main and upper marble complexes, the eclogite–blueschistunit remained protected from fluids and thus escaped retrogressivemetamorphism during exhumation (Matthews and Schliestedt, 1984;Schliestedt and Matthews, 1987). Alternately, Avigad (1993) proposedthat the higher degree of greenschist overprinting in the rest of the islandjust implies a different cooling history of the northern blueschist belt.

Pressure–temperature paths and conditions for peak metamor-phism of the eclogite–blueschist unit have been estimated using a vari-ety of techniques, including conventional thermobarometry and multi-phase equilibria (Matthews and Schliestedt, 1984; Schliestedt, 1986;Schliestedt and Matthews, 1987; Schliestedt and Okrusch, 1988;Avigad, 1993; Trotet et al., 2001a; Schmadicke and Will, 2003; Groppoet al., 2009), oxygen isotope thermometry (Matthews and Schliestedt,1984), and Zr-in-rutile thermometry (Spear et al., 2006). These previousestimates for the pressure–temperature conditions of Eocene HP meta-morphism range widely from 440 to 600 °C at 1.2–2.1 GPa, but with themore recent thermodynamic calculations setting the higher P–T, upperbounds, of 550–600 °C at 2.0 GPa (Trotet et al., 2001a; Schmadicke andWill, 2003) and 525–565 °C at >2.1 GPa (Groppo et al., 2009). Groppoet al. (2009), whose analysis suggests the highest pressures, were thefirst to fully incorporate ferric iron in their geothermobarometric analy-sis of these often highly oxidized rocks.

A decompression path from 480–520 °C at 1.2–1.5 GPa to 450 °C at0.8–1.0 GPa, followed by cooling and decompression to 450 °C at0.5 GPa, was proposed by Avigad et al. (1992). Trotet et al. (2001a)suggested a cooling and decompression path from 600 °C at 2.0 GPa to350 °C at 0.9 GPa for the northern eclogite–blueschist unit, with furthercooling and decompression for the greenschist unit. Schmadicke andWill (2003) assign greenschist facies conditions of b450 °C at 0.9–1.0 GPa.

12 B. Dragovic et al. / Chemical Geology 314-317 (2012) 9–22

5 . Sample description

We collected blueschist samples from just south of Cheronissosand from cliffs north of Vroulidia Bay in the northern part of Sifnos(Fig. 1). Preliminary analysis of several samples revealed that garnetin sample 06MSF-6Cwas easily cleansed of inclusions yielding relative-ly high garnet 147Sm/144Nd ratios, thus providing a good opportunityfor acquiring precise age information. As a result, sample 06MSF-6C isthe focus of the present study.

Sample 06MSF-6C is a 1560 cm3 block of float from the Cheronissosarea, just south of Pentzoulas, containing porphyroblasts of garnet up to15 mm in diameter; the primary outcrop source of this hand samplewas later confirmed nearby (N37° 01′56.1″, E24° 39′45.2″). The assem-blage for 06MSF-6C includes sodic amphibole, epidote, garnet, par-agonite, rutile, and less abundant quartz, phengite, jadeite, hematite,chloritoid and lawsonite; the latter two minerals occur only as inclu-sions in other minerals. Large crystals of garnet and epidote are easilyvisible within the glaucophane–paragonite matrix.

6 . Petrography

Garnet (Fig. 2A) is riddled with ~0.2 mm inclusions of quartz thatare aligned subparallel to the matrix foliation. Quartz is by far themost abundant inclusion in garnet. Inclusions of glaucophane, rutile,chloritoid, jadeitic pyroxene, and epidote are also common. A fewsmall white mica inclusions (later confirmed to be both phengite andparagonite) in garnet were found. In addition, two rare lawsonite

Fig. 2. Thin section photos of sample 06MSF-6C in plane polarized light. (A) Interior ofa large garnet crystal showing abundant inclusions. Key inclusion phases are labeled.(B) Representative matrix showing abundant glaucophane and paragonite with a largeepidote crystal overprinting them. Scale bar is 1 mm.

inclusions were discovered: one in garnet and the other in a large ma-trix glaucophane.

In the matrix, Na-amphibole (glaucophane) occurs (Fig. 2B) bothas large porphyroblasts and as fine-grained, inclusion-free aggre-gates. Epidote is the largest matrix porphyroblast (5–10 mm) phase(beside garnet); the grains are typically euhedral and contain abundantinclusions of Na-amphibole and quartz, along with rare chloritoid.Paragonite and phengite occur both as inclusion-rich porphyroblastsand as smaller, inclusion-free matrix grains. In thin section the whitemicas are difficult to distinguish, but microprobe observation suggestsroughly 50/50 phengite to paragonite. White mica contains inclusionsof Na-amphibole, quartz, chloritoid, and epidote. Jadeitic pyroxene ispresent in thematrix as anhedralmasses. Rutile is an abundant accesso-ry phase. Hematite is sparsely distributed both as inclusions and in thematrix. Chloritoid and lawsonite are absent from the matrix.

7 . Analytical methods

7.1 . Mineral and whole rock chemistry

Thin sections of 06MSF-6C were prepared for analysis on a JEOLJXA-733 Superprobe (Massachusetts Institute of Technology) for garnettraverses, spot analyses on garnet inclusions, and matrix spot analyses.The microprobe was run at a 10 nA beam current with a 15 kV acceler-ating voltage. Spot size was 10 μm for hydrous phases and 1 μm foranhydrous phases. Representative major element data for each phaseis given in Table 1a and 1b.

A 45 cm3 slice of 06MSF-6C whole rock (including garnet) was cutfrom the hand sample for whole rock major element analysis. Allweathered edgeswere removed; the slicewas then powdered in an alu-mina ball mill and sieved through 100 μm-mesh sieve. Similarly, garnetcrystals were physically removed from a second portion of rock leaving31 g of garnet-free matrix for powdering and analysis. The powderedsamples were dried in an oven overnight at 150 °C, then placed in por-celain crucibles. The crucibles were heated in a furnace at 950 °C for45 min; upon removal the crucibles were weighed to determine losson ignition (LOI). A portion was then taken from each sample andmixed with lithium metaborate (LiBO2) in a graphite crucible. Alongwith four standards (SDC-1, QLO, IORC, SCo-1), the sampleswere placedin the furnace at 1050 °C for 15 min. After flux fusion, the samples/stan-dardswere dissolved in acid and filtered to remove any carbon contam-ination. The samples and standards were then diluted in a 4000:1 ratioand run on the ICP-ES at BostonUniversity.Whole rock andmatrix-onlymajor element chemistry is given in Table 2. A portion of thewhole rock(powdered as discussed above), along with whole rock standards QLO,W2, JA2, andMAS-1722, was prepared in a 2000:1 dilution and run on aICP-MS at Boston University to conduct whole rock trace element anal-ysis. Whole rock trace element chemistry is given in Table 3.

7.2. Modal analysis

Large slabs (56 cm2) of the whole rock were point counted to de-termine the garnet mode. Point counts over several different repre-sentative areas were performed, and modal analysis converged at 11area % garnet for representative areas of 15 cm2 or greater. Becausemodes were determined on several slabs, we interpret the calculatedarea percent to be equivalent to volume percent. For the matrix min-eralogy, point counts were made in thin section. The modal mineralo-gy for whole rock and matrix are given in Table 4.

7.3 . Microsampling and preparation for garnet geochronology

Two large neighboring garnets from the same sample were selectedfor analysis. Onematrix sample (matrix 1)was taken immediately adja-cent to the two garnet crystals and four other matrix samples (matrix11, 12, 13, 14) were taken from other places in the hand sample, all

Table 1aAverage composition of matrix minerals.

Amphibole Epidote Phengite Paragonite Pyroxene Lawsonite Chloritoid

SiO2 56.83 38.41 50.57 49.01 56.90 39.26 24.23TiO2 0.02 0.10 0.17 0.06 0.03 0.00 0.04Al2O3 8.81 23.72 25.93 38.37 14.97 30.88 39.63Fe2O3 3.61 13.33 0.00 0.00 14.33 0.00 0.00FeO 13.28 0.00 4.59 0.92 0.00 1.38 24.29MnO 0.04 0.20 0.01 0.02 0.01 0.04 0.14MgO 7.87 0.03 2.68 0.08 0.58 0.04 3.71CaO 0.27 23.30 0.02 0.11 0.83 16.76 0.03Na2O 7.49 0.05 0.58 7.19 12.55 0.00 0.00K2O 0.02 0.02 10.09 0.70 0.00 0.00 0.01Total 98.25 99.17 94.64 96.45 100.20 88.36 92.07

13B. Dragovic et al. / Chemical Geology 314-317 (2012) 9–22

within roughly 10 cm of the analyzed garnets. Chemically contouredmicrosampling for geochronology followed the methods outlined inPollington and Baxter (2011). Roughly 2-mm thick wafers were cutsuch that the geometric centers of garnet #1 and garnet #2 were withineach wafer. The respective diameters of the garnets were 1.3 and 1.5 cm.Fe, Mg, Ca, and Mn concentrations were measured on the polished thickgarnet slices and used to create semi-quantitative chemical contourmaps(Fig. 3). At a grid spacing of 300 μm, 1676 and 1897 points were mea-sured on garnets #1 and #2, respectively. The electron microprobe map-ping work was carried out at an accelerating voltage of 15 kV, a probecurrent of 150 nA, and counting times of 5 s for each of the four majordivalent cations.

Using the chemical maps created by the MnO analysis, 3 zones fromeach garnet were defined and readied to be sampled by microdrilling,using the NewWave MicroMill drilling apparatus (Fig. 3). The garnetswere mounted on a block of graphite and two trenches were drilled toisolate the three defined garnet growth zones. Drilled garnets fromwithin the three zones created by the trenches were collected for bothof the samples, hand crushed with a tungsten carbide percussion mor-tar, sieved to a size of 75–150 μm,magnetically separated using a Frantzseparator, and handpicked in order to isolate relatively inclusion-freegarnet fragments. The fine powdered pan residue from the sieving pro-cess was also collected separately from the coarser 75–150 μm fraction,and analyzed with no additional treatment.

Upon completion of these initial sample preparation procedures, apartial dissolution technique, modified after Baxter et al. (2002) andPollington and Baxter (2011), was performed on the 75–150 μm frac-tion to remove remaining inclusions (such as other silicate, phosphates,and more refractory phases like rutile and sphene). The garnet samples

Table 1bRepresentative garnet composition.

Core Rim Average

SiO2 36.96 37.52 37.09TiO2 0.10 0.06 0.11Al2O3 21.02 21.65 21.22Fe2O3 1.92 1.62 1.58FeO 29.95 29.17 30.15MnO 5.23 1.17 2.48MgO 1.42 2.37 1.88CaO 5.08 8.03 6.57Na2O 0.00 0.01 0.00Total 101.67 101.58 101.08XFe2+ 0.666 0.639 0.669XMg 0.056 0.093 0.074XMn 0.118 0.026 0.056XCa 0.145 0.225 0.187%Alm 66.6 63.9 66.9%Pyr 5.6 9.3 7.4%Sps 11.8 2.6 5.6%Grs 10.7 20.1 15.7%And 3.8 2.4 2.9

were kept in closed beakers in concentrated nitric acid for 3 h at 120 °C,then in concentrated hydrofluoric acid for 45 min at 120 °C, and finallyin perchloric acid overnight at 170 °C. A lot of iterations of the partialdissolution techniques (varying acids, temperatures, durations, andstarting grain size) were conducted on preliminary material in orderto find the optimal combination of sample retrieval and purity ofminer-al separate prior to partial dissolution of the microdrilled samples.

After full dissolution, which included concentrated HF, 1.5 normalHCl, and concentrated HNO3, the samples (cleansed garnet, pan residue“garnet” powders, andmatrix) were treatedwith amixed 147Sm–150Ndspike prior to a three-stage column chemistry procedure (Harvey andBaxter, 2009). This procedure includes passing the samples through acation exchange resin to remove Fe (which could otherwise swampthe TRU-spec), a TRU-spec column used to isolate the rare-earth ele-ments, and a 2-methyl lactic acid column procedure to separate bothSmandNd.Using in-house distilledMLA, three-columnblanks run along-side all of the garnet samples ranged from 1 to 19 pg of Sm and 5–7 pg ofNd. Preliminary garnet samples from Sifnos (mentioned in Harvey andBaxter, 2009) were prepared without the first stage cation exchangeresin, leading to poor yields due to the high Fe-content of these garnets,and unacceptable sample-blank ratios and interference corrections.These compromised preliminary data were not used for geochronology.

Table 5 provides the isotope data collected in this study frommatrix,garnet, and powder residues from the sample. Samples were analyzedat the Boston University TIMS facility using a Thermo-Finnigan TRITON.Neodymium separates were loaded with 2 μL of H3PO4 and Ta2O5 acti-vator slurry and were collected as NdO+ in static mode with amplifierrotation, as described in Harvey and Baxter (2009). In-house Nd standardsolution (Ames metal) yielded a mean of 143Nd/144Nd=0.5121313±0.0000059 (11.6 ppm, 2 RSD, n=7) over the duration of sample analysisfor this study. Reproducibility in 147Sm/144Nd is better than 0.1% basedon repeat analyses of a mixed gravimetric normal solutionwith our cal-ibrated in-house spike.

Table 2Major element analysis of 06MSF-6C.

wt.% Whole rock Matrix

SiO2 53.15 54.70TiO2 1.07 0.99Al2O3 13.77 16.37FeO(tot) 12.92 11.16aFe2O3 4.15 4.70aFeO 9.19 6.93MnO 0.14 0.04MgO 5.65 5.22CaO 2.72 3.65Na2O 5.30 5.63K2O 0.99 0.95P2O5 0.09 0.07LOI 1.45 1.94Total 97.67 101.19

a Ferric/ferrous ratio estimated from mineral mode and chemistry, see text.

Table 3Trace element analysis for 06MSF-6C whole rock.

ppm ppm C1 norma

Li 62.8 La 4.3 La 13.0Be 0.6 Ce 9 Ce 10.5Sc 45.6 Pr 1.3 Pr 9.9Ti 0.8 Nd 5.8 Nd 9.3V 272 Sm 1.8 Sm 8.7Cr 7.8 Eu 0.6 Eu 8.2Co 40.5 Gd 2.2 Gd 8.1Ni 11.6 Tb 0.4 Tb 7.6Cu 48.6 Dy 2.5 Dy 7.4Zn 91.5 Ho 0.6 Ho 7.1Ga 9.9 Er 1.6 Er 7.0Rb 22.7 Yb 1.6 Yb 7.3Sr 111.5 Lu 0.2 Lu 7.3Y 15.5 Hf 0.1Zr 35.5 Ta 0.1Nb 0.9 Pb 4.2Cs 0.8 Th 1.2Ba 132.7 U 0.4

a Normalized to C1 chondrite.

Fig. 3. Microdrilling of garnets 1 and 2 based on MnO contours. Three concentricgrowth zones were drilled in each garnet. The drill trenches are shown in black. At bot-tom right is photo of garnet 1 mounted on carbon block after drilling.

14 B. Dragovic et al. / Chemical Geology 314-317 (2012) 9–22

8 . Data

8.1 . Bulk rock analyses

06MSF-6C is a mafic blueschist probably from an altered basalticprotolith common in the EBU of Sifnos (Schliestedt, 1986). It has moder-ate SiO2 content (53 wt.%), elevated Na2O (5.3 wt.%), low CaO (2.7 wt.%),and high total iron (12.9 wt.% expressed as FeO). The rock is similar inbulk composition to the mafic blueschists analyzed by Schliestedt andOkrusch (1988) andMocek (2001). The high Na2O and low CaO contentsare consistent with spilitization of a basaltic protolith. The REE chemistryfor 06MSF-6C is similar tomafic blueschist data fromMocek (2001),witha slight enrichment of the LREE that is typical of back-arc tholeiite.Table 2 also shows the bulk chemistry of thematrix only (without garnetand garnet mineral inclusions). Most notable are the increase in Al andCa and decrease in total Fe in the matrix (relative to the whole rock)which reflect the removal of garnet and its inclusions from the bulk ma-trix. In addition, we usedmeasured averagemineral chemistry (Table 1aand 1b) and measured matrix modal analysis (Table 4) to estimate thebulk ferric/ferrous content in both the whole rock (Fe3+/Fetot=0.29)and thematrix (Fe3+/Fetot=0.38) as reported in Table 2. This differencein ferric/ferrous content brought about by removal of garnet from theremaining matrix proved an important aspect of pseudosection model-ing described below.

8.2 . Microprobe analysis

Garnet is almandine-rich (Table 1b). A core to rim transect of spotanalyses in a 15 mmgarnet reveals a smoothly decreasingMn concentra-tion from core to rim,with the exception of a small “spike” near the rim ofthe garnet, as confirmed in other garnets analyzed (e.g. Fig. 3). Other

Table 4Mineral modal analysis (from point count).

Phase Whole rock Matrix only

Glaucophane 0.37 0.42Garnet 0.11 0Pyroxene 0.13 0.14Epidote 0.05 0.06Paragonitea 0.13 0.145Phengitea 0.13 0.145Quartz 0.06 0.07Rutile 0.02 0.02

a White mica estimated to be 50% phengite and 50% paragonite based on microprobeobservation.

major cations show similar smooth variation as shown in Fig. 4 with Mgand Ca increasing from core to rim, and Fe remaining fairly flat until aslight decrease near the rim. These patterns imply undisturbed progradegrowth for the garnetwithno significant diffusive re-equilibration, exceptfor the outermost analysis whichwas not used in our calculations. Andra-dite component (averaging about 5 mol%) was estimated by normalizingto 8.00 cations and 12.00 oxygens. In comparison to the higher MnO ob-served in the garnet chemical map (Fig. 3) this thin section transectseems to have missed the exact chemical core of the garnet, implyingthat the actual core had somewhat higher MnO and probably slightlylower MgO than shown in Fig. 4. We also report the rim chemistry inTable 1b from the average of six spot analyses in two garnets just insidethe extreme edge of the garnet.

The sodic amphibole in thematrix and included in garnet lies betweenglaucophane and riebeckite (average composition Na2.04(Mg1.65Fe2+1.55Fe3+0.35Al1.47)Si8O22(OH)2) using the approach of Holland and

Table 5Sm–Nd isotope data.

Sample Sm (ppm) Nd (ppm) ng Nd loaded 147Sm/144Nd 143Nd/144Nd ±2 S.E. (abs) ±2 S.E. (ppm)

Matrix 1 2.763 9.163 18 0.1824 0.5126123 0.0000066 13Matrix 11 1.222 4.091 ~25 0.1807 0.5126060 0.0000037 7.2Matrix 12 1.701 5.598 ~25 0.1838 0.5126108 0.0000047 9.2Matrix 13 2.904 9.936 ~25 0.1768 0.5126007 0.0000080 16Matrix 14 2.261 7.582 ~25 0.1804 0.5126102 0.0000053 10Gt1 core 0.3415 0.04824 0.9 4.282 0.513862 0.000025 48Gt1 mid 0.4949 0.06236 2.1 4.800 0.514012 0.000017 32Gt1 rim 0.1352 0.02680 2.1 3.051 0.513481 0.000011 21Gt2 core 0.3350 0.05474 1.1 3.618 0.513637 0.000037 72Gt2 mid 0.5441 0.07173 1.4 4.589 0.513963 0.000051 99Gt2 rim 0.0637 0.03777 1.9 1.019 0.512853 0.000025 49Gt1 core pwd 0.5854 1.112 19 0.3186 0.5126045 0.0000050 10Gt1 mid pwd 0.6526 0.8863 18 0.4454 0.512644 0.000011 21Gt1 rim pwd 0.9100 2.813 17 0.1935 0.5125947 0.0000048 9.4Gt2 core pwd 0.6849 1.481 26 0.2797 0.5126053 0.0000045 8.7Gt2 mid pwd 0.6686 0.9358 14 0.4322 0.5126530 0.0000059 11Gt2 rim pwd 0.6074 1.926 19 0.1908 0.5125774 0.0000076 15

15B. Dragovic et al. / Chemical Geology 314-317 (2012) 9–22

Blundy (1994). Paragonite (average composition (Na0.88 K0.06Ca0.01)(Al1.94Fe0.05 Mg0.01)(Al0.91Si3.09)O10(OH)2) and phengite (average com-position (K0.87Na0.08)(Al1.51Fe0.26 Mg0.27)(Al0.57Si3.43)O10(OH)2) coexistin this sample in roughly equal proportions. Sodic pyroxene (jadeite–acmite with average composition (Na0.86Ca0.03)(Mg0.03Fe3+0.34Al0.63)Si2O6) is present in anhedral patches in the matrix, and also occurs as in-clusions in garnet. Epidote is abundant throughout the sample and showsa wide range of compositions: inclusions within garnet vary from 0.78 to0.87 of the FeAl2 end member, and matrix grains are zoned from corecompositions of ~0.72 FeAl2 to rim compositions of 0.76–0.97 FeAl2. Ma-trix epidotes clearly overprint the fabric and represent late stage growth.Epidotes included within garnet show irregular zoning and morphologyindicative of a replacement texture (perhaps after lawsonite; see below).

Fig. 4. Representative core to rim traverse of garnet chemistry in sample MSF-6C.

8.3 . Geochronology

Three zones from each of the two adjacent garnets were analyzedfor isotope compositions (Table 5). All of the garnets in this studycontain less than 0.1 ppm Nd (0.03–0.09 ppm). Given the inclusiondensity, the chemical steps necessary to remove impurities from garnetresulted in such high mass losses (90–95%) that very little Nd (between0.9 and 2.1 ng) remained in the cleansed garnets for analysis. High147Sm/144Nd values ranging from 3.0 to 4.8 for five of the zones and 1.0for the rim zone of Gt2 were achieved, indicating success in eliminatingadverse inclusion contamination effects (see below). The variation inNd concentration and 147Sm/144Nd between garnet zones could be dueto differing chemistry of the garnet itself, the nature of the inclusion pop-ulation, or the effectiveness of the partial dissolution processwithin eachzone. Analysis of the five matrix samples showed them to be essentiallyidentical throughout the hand sample; small variations reflect subtlemineralogical differences. Thus, each of the six garnet zone sampleswas paired with the same five matrix samples, yielding six 6-point iso-chrons (because the five matrix values are essentially identical theseare still effectively two-point isochrons). All age calculations used the Iso-plot program (Ludwig, 2003). The ages are shown in Table 6 and Fig. 5.The third (rim) zone from garnet #2 has a lower 147Sm/144Nd than theother zones analyzed and thus poorer age precision. This could be dueto the incorporation of somematrixmaterial or incomplete removal of in-clusions during the physical or chemical separation techniques. Nonethe-less, all six garnet-matrix isochron ages yield identical results withinuncertainties (Fig. 5).

Table 6Garnet ages.

Isochron Age (Ma) 2σ age uncertaintya MSWD

Gt1 core (gt1 core+5 mtx) 46.74 0.91 1.3Gt1 middle (gt1 mid+5 mtx) 46.44 0.55 1.3Gt1 rim (gt1 rim+5 mtx) 46.48 0.59 1.3Gt2 core (gt2 core+5 mtx) 45.8 1.6 1.3Gt2 middle (gt2 mid+5 mtx) 47.0 1.7 1.3Gt2 rim (gt2 rim+5 mtx) 44.8 4.5 1.3Multi-point Gt core (gt1 core+gt2 core+5 mtx)

46.50 0.80 1.2

Multi-point gt middle (gt1 mid+gt2 mid+5 mtx)

46.49 0.53 1.1

Multi-point gt rim (gt1 rim+gt2 rim+5 mtx)

46.46 0.59 1.1

Average garnet age (11 point:all gt+all mtx)

46.49 0.36 0.78

In bold are the ages we consider most robust for core, middle, and rim.a Age calculations use the poorer (higher) of the internal 2 SE (reported in Table 4)

or the external 2 SD (0.0000059).

Fig. 5. Summary of ages from core, middle, and rim of garnets 1 and 2. The multipointages shown include both garnet data as well as matrices. See text.

16 B. Dragovic et al. / Chemical Geology 314-317 (2012) 9–22

Because the chemically similar core, middle, and rim of each garnetyield consistent ages, we elect to group like garnet zones into 7-point iso-chrons, each of which includes the same zones from each of the two gar-nets and the five matrices. Table 6 and Fig. 5 show these multi-pointisochron ages for each of the three garnet zones. The ages of the core torim of the garnets from this sample are 46.50±0.80 Ma (MSWD=1.2),46.49±0.53 Ma (MSWD=1.08), and 46.46±0.59 Ma (MSWD=1.14),respectively. These ages for core, middle, and rim growth are statisticallyindistinguishable. An 11-point isochron age, including all six cleansedgarnets and the five matrix analysis, gives 46.49±0.36 (MSWD=0.78)for the average age of garnet growth. This isochron is shown in Fig. 6.The low MSWD of this combined isochron supports the conclusion thatboth of these garnets grew at the same time. However, there is morevalue in recognizing the statistical significance of the simple growth ge-ometry and separately calculating ages for core, middle, and rim of thegarnets to constrain the actual growth duration.

Fig. 6. Sm–Nd isochron diagram showing all cleansed garnet data (red diamonds for garnet(open diamonds for garnet 1; open squares for garnet 2). The reference isochron shown is thhow the powders fall well off the isochron due to the effect of contamination from inherite

Finally, the previously mentioned pan residue powders (or pwd, asin Table 5) from gt1 and gt2 are also plotted with the multi-point iso-chron shown in Fig. 6. All of the pan residues fall significantly off ofthe garnet-matrix isochron, suggesting that they represent contamina-tion from a component of the inclusion population (likely inheritedmonazites with very low 147Sm/144Nd ratios) not in age equilibriumwith the garnet-matrix system. If these untreated “garnet” powders –which thus represent mixes between clean garnet and inherited inclu-sions – are used in our calculations instead of the matrices, multi-point “isochron ages” for each zone yield poorer MSWD (5.1, 2.5, and6.1), poorer precision (±2.2, ±0.58, and ±2.7 Ma) and skewed ages(48.1, 47.72, and 47.8 Ma). If the powders are added to the 11-pointgarnet-matrix isochron shown in Fig. 6, MSWD further worsens to 24.As the inclusion-rich powders reflect age inheritance and do not yieldtrue isochrons (evidenced by high MSWD; Wendt and Carl, 1991),they are not used as part of themulti-point isochrons nor in our age in-terpretations. We show them here to illustrate the great importance, interms of garnet age precision and accuracy, of removing the effects ofthese inclusions as well as our ability to have successfully done so.

8.4 . Overall garnet growth span

The most significant aspect of the geochronologic results presentedhere is the concordance in age between all zones of the two garnets(Figs. 5 and 6). This suggests that the duration of blueschist–facies garnetgrowth in this lithology on Sifnoswas very brief. Thedifference in age be-tween the core and rim multi-point isochron ages is 0.04 Myr, thoughwith larger uncertainty. Propagating the 2σ age errors provides a maxi-mum duration of garnet growth of 1.03 Myr. Instantaneous growth isalso allowable within our uncertainties. Overall, we conclude that garnetgrowth occurred in less than 1.0 Ma, andmore likely spanned only a few100,000 s or even 10,000 s of years.

1; red squares for garnet 2), matrices (blue circles), and uncleansed “garnet” powderse 11-point isochron including all cleansed garnets and all matrix points. The inset showsd inclusions with low Sm/Nd mixing with pure garnet.

17B. Dragovic et al. / Chemical Geology 314-317 (2012) 9–22

8.5. Garnet growth rate

By combining the garnet growth durations determined here withthe observed modal abundance of garnet in the sample, we can calcu-late a bulk-rock garnet growth rate. This time-averaged rate may beexpressed in terms of moles of garnet produced per year. Althoughgrowth rates may be expected to change in response to fluctuations inreaction (or reactions) progress and/or conditions amenable to garnetgrowth (e.g. Pollington and Baxter, 2010), our data permit only a time-averaged growth rate. Because we are interested here in reconstructingthe total amount of reaction that occurred over the growth interval, thetime-averaged rate is appropriate. For the maximum (2σ) growth spanof 1.0 Ma, given a volumetric garnet abundance of 0.11 cm3 garnet/cm3

rock, and taking a density of 4.17 g/cm3 and a molecular weight of482 g/mole for the almandine-rich garnet (as calculated from the garnetmineral chemistry), we arrive at a constraint for the minimum (2σ)garnet growth rate of 9.4×10−10 mol/cm3/year. The same calculationusing the absolute garnet growth duration of 0.04 Ma yields a garnetgrowth rate of 2.4×10−8 mol/cm3/year. Infinitely fast garnet growthrates (spanning zero time) are permissible (though obviously not likely)within our age constraints.

9 . Thermodynamic analysis

9.1 . Pressure–temperature span of garnet growth

To constrain the pressure–temperature span of garnet growth, wecompare measured garnet core and rim chemistry with garnet com-positional isopleths predicted from pseudosection analysis. We con-structed two pseudosections: one for thewhole rock chemistry (Table 2)in Fig. 7A (appropriate for the initiation of garnet core growth) and onefor matrix chemistry (Table 2) in Fig. 7D (excluding garnet chemistrywhich is effectively fractionated out of the system, appropriate for thevery final growth of garnet rim). Pseudosections were computed withthe program Perple_X (Connolly, 2009) using full solution models (in-cluding ferric iron end members where appropriate) for the followingphases: garnet (White et al., 2005), Ca- and Na-amphibole (Dale et al.,2000), white mica, (Smye et al., 2010), omphacite (Green et al., 2007),epidote (Holland and Powell, 1998), chloritoid (Smye et al., 2010), chlo-rite (Holland et al., 1998), feldspar (Fuhrman and Lindsley, 1988) andFe-oxides (Andersen and Lindsley, 1988). Lawsonite, rutile, quartz,and sphene were considered as pure phases. Calculations were forwater-saturated conditions. Note the significant differences between thewhole-rock (Fig. 7A) and matrix (Fig. 7D) pseudosections, highlightingthe importance of accounting for garnet fractionation during growth(e.g. Marmo et al., 2002; Konrad-Schmolke et al., 2008a, 2008b; Gaidieset al., 2008).

Fig. 7A shows garnet chemical isopleths for Ca, Fe, Mn equivalentto measured garnet core chemistry (Table 1b) superimposed on thewhole-rock pseudosection. Mg content is low and it is highly sensitiveto uncertainties (especially including the likelihood that the true coreMg was lower than that in the analyzed traverse; see above) and isnot considered. AnalyzedMn in the traverse core is also probably slight-ly too low, butMn isopleths are very tight in this part of the diagramandlead only to slight shift towards lower T within the bounds shown.These isopleths converge on an average P–T of garnet core growth of~2.0 GPa and ~460 °C, which falls just within the assemblage chlorite–chloritoid–glaucophane–garnet–phengite–lawsonite–quartz–rutile.This is consistent with garnet inclusions of all these minerals exceptchlorite which is predicted to be in very low abundance. Fig. 7D shows adiagram for the garnet rim chemistry on thematrix pseudosection. Theseisopleths converge on an average P–T of garnet rim growth of ~2.2 GPaand ~560 C, which falls in the assemblage omphacite–glaucophane–garnet–phengite–paragonite–lawsonite–quartz–rutile. The appearanceof omphacite and paragonite is consistentwithmatrixmineral observa-tions. The only notable discrepancy between predicted and observed

mineralogy is the absence of epidote in the pseudosections at the P–Trange of garnet growth. Epidote was observed as an abundant inclusionin garnet as well as a major euhedral overprinting phase in the matrix.Notable also is the almost complete absence of lawsonite in the matrix.Both can be explained by examining the pseudosection in Fig. 7D, whichshows lawsonite breakdown upon decompression after garnet growthand the resulting growth of epidote in the final observed assemblage.All epidote is therefore considered to reflect breakdown of lawsonite,both included in garnet and in the matrix where the large euhedral ep-idotes were able to form.

The P–T constraints on the span of garnet growth are also consistentwith the P–T envelope determined for Sifnos blueschists by Groppo etal. (2009) shown for reference in Fig. 8 which summarizes the P–T con-straints. Note that garnet growth spanned less than 1 Ma during nearisobaric heating from 460 to 560 °C. This equates to aminimum heatingrate of 100 °C/Ma.

9.2 . Constraining a dehydration rate

The broad aim of this study is to provide a method to constrain therate of dehydration during subduction zone metamorphism. Thus, gar-net growth and water production must be directly linked. To a firstorder, note that predicted growth of garnet corresponds to predictedloss of water for both starting (Fig. 7B,C) and ending (Fig. 7E,F) effectivecompositions. If we were to ignore the fractionating effect of garnetgrowth and simply follow the P–T path to the end of garnet growth(11 vol.%) on the whole-rock pseudosection (Fig. 7A,B,C), we wouldconclude that the final bulk water content was ~2.6 wt.%, indicatingliberation of ~1.0 wt.% water during garnet growth. However, as vividlyindicated by the matrix pseudosection in Fig. 7D,E,F, this conclusionwould be erroneous. In fact, fractionation of garnet (and garnet inclu-sion) chemistry changes thematrix chemistry during growth. Most im-portantly, this has the effect of stabilizing lawsonite during the entiregarnet growth interval, which in turn sequesters much of the waterinto its structure, limiting total dehydration during garnet growth. It iscrucial for the reader to appreciate that one cannot directly interpretnet dehydration from a single pseudosection analysis that ignores thefractionating effect of garnet growth. The whole rock pseudosection isneeded to constrain the conditions at start of garnet growth, and thematrix pseudosection is needed to constrain the conditions at the endof garnet growth.

A comparison between total structurally bound water at the startand finish of garnet growth permits the dehydration flux calculation.To accomplish this, we use calculated bulk rock water contours fromthe pseudosections for whole rock (garnet core growth) and matrix(garnet rim growth) compositions (Fig. 7C,F). Since the pseudosectionanalysis predicts that ~1.7 vol.% garnet should already have grown atthese P–T conditions, we recalculate the initial bulk rock water contentby forcing garnet modal proportion to zero arriving at ~3.7 wt.% water(virtually identical to that shown in Fig. 7C). Such an offset betweenpredicted and observed P–T for garnet nucleation is not a surprise andlikely relates to the critical overstepping required for nucleation andgrowth. Uncertainty in P–T of garnet core growth has very little effecton predicted starting water content because the water contours arewidely spaced in this range of P–T space and predicted garnet mode isvery low.

Using the final matrix bulk composition to account for garnet frac-tionation shows that ~3.8 wt.% water remains in thematrix only at theend of garnet growth, as shown on Fig. 7F. Indeed, as lawsonite wasstabilized, the water content of the matrix actually increased duringgarnet growth. However, this does not include the garnet (11 vol.%)that is present in the rock. When garnet is added back into the totalrock assemblage (forcing the total garnetmode to 11 vol.% as observed)thefinal bulk rockwater content is 3.3–3.4 wt.% (given uncertainty is PTestimate), indicating a loss of 0.3–0.4 wt.% during garnet growth. Fig. 8

Fig. 7. Pseudosection analysis for whole rock bulk chemistry (A,B,C) and matrix-only bulk chemistry (D,E,F). (A) Pseudosection diagrams for the whole rock chemistry. Garnetchemical isopleths corresponding to observed core garnet chemistry (Table 1b) are plotted (almandine=red, grossular=green, spessartine=blue); their intersection constrainsthe P–T of garnet nucleation. Relevant phase fields along the PT trajectory are labeled. [(1) pl namph ph pa gt q ru; (2) pl namph ph gt q ru; (3) ep camph namph ph pa gt q ru; (4) chlnamph ph pa gt law sph q; (5) chl camph namph ph pa gt sph q; (6) camph namph ph pa gt sph q ru] (B) garnet modal isopleths in 1 vol.% increments. Garnet growth appears to havebegun slightly after the predicted garnet-in indicating some critical overstepping for nucleation. Bold black line is the 11 vol.% contour for reference. (C) Structurally bound water wt.%contours in 0.1 wt.% increments for unfractionated whole rock chemistry. (D) Pseudosection diagrams for chemistry of the matrix only (not including garnet). Garnet chemical isoplethscorresponding to observed rimgarnet chemistry are plotted; their intersection constrains the P–T at the end of garnet growth. Relevant phasefields along the PT trajectory are labeled. [(1)jd namph phgt law q ru; (2) jd namphph gt ky law q ru; (3) jd ep namph ph gt ky q ru; (4) ep namph ph pa gt law sph q; (5) ep namph ph pa gt sph q; (6) ep pl namph phpa gt sph q; (7)ep pl namph ph pa gt q ru; (8) pl camph namph ph pa gt q ru] (E) Garnet modal isopleths in 1 vol.% increments for the matrix composition. Because this diagrammodels the garnet-freematrix only, we expect the predicted garnet mode to be near zero which it is. (F) Structurally bound water wt.% contours in 0.1 wt.% increments for the matrix composition. [gt =garnet; jd = jadeite; namph = Na-amphibole; camph = Ca-amphibole; ph = phengite; pa = paragonite; pl = plagioclase; q = quartz; ru = rutile; ep = epidote; law =lawsonite; chl = chlorite; sph = sphene; ky = kyanite].

18 B. Dragovic et al. / Chemical Geology 314-317 (2012) 9–22

Fig. 8. Summary of P–T-H2O history for sample 06MSF-6C. Text boxes denote the volume abundance of garnet and the wt.% of water at both garnet core and rim P–T, as well as afterlawsonite is completely consumed upon exhumation. For comparison, three recently published P–T paths for Sifnos are shown. The thick, light gray P–T “envelope” is from Groppoet al. (2009) which we use to infer exhumation trajectory. Solid, dark gray P–T path is from Schmadicke and Will (2003). Dashed, light gray P–T path is from Trotet et al. (2001b).Note that the thin dashed, black line denotes the lawsonite-out/epidote-in phase boundary at which point a significant amount of water is lost (see also Fig. 7C,F).

19B. Dragovic et al. / Chemical Geology 314-317 (2012) 9–22

summarizes the constraints we can place on the timing and amount ofdehydration from this lithology.

Pseudosection analysis, accounting for bulk chemical fractionationof garnet by direct measurement of matrix and whole-rock chemistry,indicates that the 11 vol.% of garnet grown in b1 Ma yielded 0.3 to0.4 wt.% water. Pseudosection analysis also indicates the modal min-eralogy at start and finish of garnet growth (with garnet mode forcedto zero at the start and 11 vol.% at the end of garnet growth). By sub-tracting the final modal mineralogy from the initial modal mineralogy(recast into molar quantities), we can calculate the net reaction stoi-chiometry to identify the major matrix phases contributing to water

release (and uptake). The following reaction describes this net reactionfrom start to finish of garnet growth:

0:02 chloriteþ 0:97 glaucophaneþ 0:10 phengiteþ 0:06 chloritoidþ 0:10 rutile ¼ 1:07 omphaciteþ 0:18 paragoniteþ 0:16 lawsoniteþ1:36 quartzþ 1:00 garnetþ 0:69 water

Here, we see that glaucophane is themajor reactant phase contribut-ing to water release, with chlorite, chloritoid and phengite contributingsmaller amounts. Lawsonite (in particular) and paragonite absorb someof the water that might otherwise have been liberated from the rock.

20 B. Dragovic et al. / Chemical Geology 314-317 (2012) 9–22

Finally, we see that the average garnet:water stoichiometric productionratio is ~1.0:0.7. Fig. 7 shows that this stoichiometric production ratiosurely varies along the garnet growth path as the exact garnet-formingreaction changes, though we cannot constrain the relative timing andduration of each segment of the garnet-forming reaction given our ageconstraints. Combining the geochronology with the total water releaseyields an average minimum dehydration rate during prograde garnetgrowth of 5 to 7×10−10 moles H2O per cm3 of rock per year.

To place the dehydration flux related to garnet formation in thisrock into a subduction-zone context, we consider the additionaldehydration that followed garnet growth. Although we do not haveP–T constraints beyond garnet growth for this rock, we can use theP–T envelope of Groppo et al. (2009) to estimate the post-garnetgrowth P–T trajectory. The inferred P–T path shows nearly isothermaldecompression during the initial stages of exhumation, following gar-net growth. This is important because it means that the post-garnetgrowth water release did not occur during prograde subduction(as the garnet-related dehydration did), but rather during exhuma-tion. As shown in Fig. 8, at some point during exhumation the rockcrossed the major lawsonite-out reaction (lawsonite is replaced byepidote which we observe as an overprinting phase in the finalassemblage), which liberated another ~1.4 wt.% water, leaving thefinal rock assemblage with ~2.0 wt.% water. We observe 1.5% LOI inthe rock today (Table 2).

9.3 . Possible causes of rapid garnet growth rates

The Sifnos garnets grew rapidly. To our knowledge, these are the firstsubduction-zone garnets that have been dated from core to rim thus con-straining growth rate. However, several similarmeasurements exist fromBarrovian metamorphic settings (e.g., Christensen et al., 1989, 1994;Vance and O'Nions, 1992; Vance and Harris, 1999; Pollington andBaxter, 2010). With the exception of the latter (see below), each ofthese regional metamorphic studies indicates growth spans of like-sized garnets of severalmillion years, at average rates significantly slowerthan what we observe on Sifnos. Therefore, relative to Barrovian meta-morphism, conditions conducive to more rapid garnet-forming reactionsexisted during the brief interval of garnet growth within the subductedlithology preserved on Sifnos. We offer some speculation as to why gar-net growth (and associated dehydration) was so rapid in this case. Onepossibility involves rapid overstepping (in P and/or T) of closely spacedgarnet-in reaction isopleths. Another possibility is the influence of akinetic catalyst — perhaps via infiltration of an externally derived fluid.The former (rapid P–T shift) seems most likely given the evidence forrapid heating from 460 °C to 560 °C in less than 1 Ma (Fig. 8). However,the subduction descent rate for these rockswas very slow (andwith shal-lowdip angle), requiring perhaps 10–20Ma to gradually reach blueschistfacies conditions (Ring and Layer, 2003). If the garnet-forming reactionoccurred during nearly isobaric heating (as implied by the P–T span ofgarnet growth; Fig. 8), perhaps after detachment from the slowly de-scending slab into the subduction channel where temperature gradientsare severe, such heating may be geodynamically feasible. Or, the rapidheating could relate to a slab-mantle decoupling depth of ~70–80 km(very close to the estimates of maximum Sifnos pressures and depthsfrom our work and from Groppo et al., 2009) where sharp thermalgradients driven by hot mantle flow have been shown in some models(e.g. Wada and Wang, 2009; van Keken et al. 2011). Forster and Lister(2005), on the basis of geospeedometric analysis of 40Ar/39Ar data onSifnos, also suggest the possibility of brief excursions in temperatureand/or pressure representing a more complex P–T-t path than iscurrently resolvable in thermodynamic modeling. We could indeed beseeing supporting evidence for these ideas in our garnet growthgeochronology. Brief thermal pulses have also been documented inregional metamorphism associated with advective heat sources (e.g.,Ague and Baxter, 2007; Vorhies andAgue, 2011). Kinetic fluid triggeringhas also been documented in subduction zones (e.g. Austrheim, 1987;

John and Schenk, 2003; Camacho et al., 2005) and regional orogeny(e.g. Pollington and Baxter, 2010). In this scenario, an externally(or internally) derived fluidwets the system and allows an oversteppedreaction to burst to completion by eliminating kinetic barriers (e.g.,Baxter, 2003). A fluid triggering mechanism during subductionwould require a further accounting of the source of the catalyzingfluid, proposed by some to derive at least in part from mantle de-serpentinization (e.g., John et al., 2004; Halama et al., 2011; Spandleret al., 2011) or by dehydrating assemblages within the subductionchannel. Also, a fluid kinetic triggering would not necessarily lead tothe rapid heating observed (though it could be linked to such heating).While the brief duration of garnet-growth-related dehydration in thisSifnos blueschist is clear, further analysis will be required to testhypotheses regarding the cause of the rapid growth pulse in this andother subducting lithologies.

9.4. Implications for net water-release in subduction zones

Based on the modal abundance of garnet in the rock (11 vol.%), andthe thermodynamic reaction analysis above, we calculate that 0.3–0.4 wt.% H2O was released from the rock due to the prograde garnet-forming reaction(s) during this brief time interval (0.04±0.99 Ma).Thoughmany variables (rock composition, subduction rate, subductiondip, thermal regime) can alter the thermodynamic–geodynamic pre-dicted dehydration flux, our 0.3 to 0.4 wt.% H2O release may be com-pared to the range of available model predictions (e.g., Kerrick andConnolly, 2001; Hacker et al., 2003; Hacker, 2008; van Keken et al.,2011) for subduction to comparable depths (~2.2 GPa) and maxi-mum temperatures (~560 °C). For example, Kerrick and Connolly(2001) modeled metamorphic devolatilization of subducting basaltthat began with 2.68 wt.% water. Their results for water release alonghigh, intermediate, and low temperature PT-paths range from completedevolatilization for high-temperature gradients (i.e. release of 2.68 wt.%water at b2.5 GPa) to negligiblewater-release for low-temperature gra-dients. Although direct comparisons to these calculations are hamperedbydifferences in bulk rock composition and initialwater content, for theprograde Sifnos P–T trajectory shown in Fig. 8, Kerrick and Connolly(2001) would predict 0.1 to 0.2 wt.% water release, just slightly lowerthan our constraint. Hacker et al. (2003) and Hacker (2008) averageddata from a wide array of natural metabasalts to predict a total watercontent of 0.7 wt.% at the “peak” of the Sifnos P–T trajectory, comparedto ~3 wt.% before garnet growth; they thus predict a release of~2.3 wt.% during the time interval of garnet growth. Note, however,that our constraint does not include any water released before or afterthe garnet-forming reaction, and indeed when the additional waterloss due to lawsonite breakdown is added, total water loss from ourrock is ~1.7 wt.% much closer to the Hacker et al. (2003) estimates. Itis important to realize that this lawsonite breakdown occurs not duringsubduction for sample 06MSF-6C, but only upon the onset of exhumationto the surface. Because Hacker et al. (2003) use final assemblages to con-strain bulk water their analysis will incorporate post-subduction waterrelease. In any case, our analysis focused solely on garnet-forming reac-tions has captured a significant amount of the total predicted waterrelease at blueschist facies conditions.

It is important to emphasize that our result applies strictly to onlythis lithology in this particular P–T path for Sifnos. Mafic blueschists withsimilar mineralogy and chemistry are fairly common in the eclogite–blueschist unit of Sifnos, but subtle differences in chemistry may changethe dehydration specifics even so. Further analysis of other Sifnos litholo-gies, beyond this test case, would be required to determine overall trendsand to reconstruct a broader dehydration flux from diverse subductingrock types for this P–T trajectory. Still, the brevity of significant garnetgrowth and related dehydration and heating in this lithology is strikingwithin the context of the overall subduction descent timescale of ~10–20Ma (Ring and Layer, 2003).

21B. Dragovic et al. / Chemical Geology 314-317 (2012) 9–22

10 . Conclusion

This work presents a method for constraining the rate and durationof water release from subducted lithologies that is based on con-straining rates of garnet-forming reactions occurring during subduc-tion. We produced precise Sm/Nd geochronological data on concentricgrowth zones in individual garnets with extremely low concentrationsof Nd. Garnet growth in this mafic blueschist took place over a briefspan between growth of the core at 46.50±0.80 (at 2.0 GPa and460 °C) and growth of the rim at 46.46±0.59 Ma (at 2.2 GPaand 560 °C), with a maximum (2σ) growth duration of 1.0 Ma. By com-bining precise geochronological datawith thermodynamic constraints onthe garnet-forming dehydration reaction, we calculate a time-averagedminimum (2σ) dehydration rate of 5 to 7×10−10 mol H2O/cm3 rock/year. The samemethod should also allowquantification ofwater releaserates for other sampled lithologies and conditions fromwhich integrat-ed fluid fluxesmay be compiled to testmodel predictions. The release of0.3 to 0.4 wt.% water over a very brief period of time (most likely10,000 s to 100,000 s of years) during subduction suggests that a signif-icant amount of dehydration from subducted lithologies may occur in abrief focused pulse, rather than more slowly and continuously duringdescent as in some other models (e.g. Schmidt and Poli, 1998; Kerrickand Connolly, 2001). Future testing of this hypothesis is necessaryto determine whether such dehydration pulses occur in othersubduction terranes and lithologies, and when detected, whetherthey are driven by heating within the subduction channel, sharpthermal gradients related to slab-mantle decoupling (e.g. Wadaand Wang, 2009; van Keken et al. 2011), by externally derivedfluids triggering the overstepped reaction, or by some other kineticor geodynamic factor.

Acknowledgments

We thank Jason Harvey and Jeremy Inglis for their assistance withthe aspects of mass spectrometry. We thank Michael Bröcker for hisinformation and suggestions about field sampling on Sifnos. We thankMark Caddick for his discussions about the aspects of thermodynamicanalysis. We thank Neel Chatterjee for his help with electronmicroprobeanalysis.We thank two anonymous reviewers for their constructive com-ments. We acknowledge the support from NSF grant EAR-0547999(to EFB). Sm–Nd analysis was conducted in the Boston UniversityTIMS Facility which has been supported by NSF grants EAR-0521266and EAR-0949390.

References

Ague, J.J., Baxter, E.F., 2007. Brief thermal pulses during mountain building recorded bySr diffusion in apatite and multicomponent diffusion in garnet. Earth and PlanetaryScience Letters 261, 500–516.

Altherr, R., Schliestedt,M., Okrusch,M., Seidel, E., Kreuzer, H., Harre,W., Lenz, H.,Wendt, I.,Wagner, G., 1979. Geochronology of high-pressure rocks on Sifnos (Cyclades, Greece).Contributions to Mineralogy and Petrology 70, 245–255.

Amato, J.M., Johnson, C.M., Baumgartner, L.P., Beard, B.L., 1999. Rapid exhumation ofthe Zermatt-Saas ophiolite deduced from high-precision Sm–Nd and Rb–Sr geo-chronology. Earth and Planetary Science Letters 171, 425–438.

Anczkiewicz, R., Thirlwall, M.F., 2003. Improving precision of the Sm–Nd garnet datingby H2SO4 leaching—a simple solution to phosphate inclusions problem. In: Vance, D.,Müller, W., Villa, I.M. (Eds.), Geochronology: Linking the Isotopic Recordwith Petrologyand Textures: Geological Society Special Publication, 220, pp. 83–91. London.

Andersen, D.J., Lindsley, D.H., 1988. Internally consistent solution models for Fe–Mg–Mn–Tioxides. American Mineralogist 73, 714–726.

Austrheim, H., 1987. Eclogitization of lower crustal granulites by fluid migrationthrough shear zones. Earth and Planetary Science Letters 81, 221–232.

Avigad, D., 1993. Tectonic juxtaposition of blueschists and greenschists in Sifnos Island(Aegean Sea)— implications for the structure of the Cycladic blueschist belt. Journal ofStructural Geology 15, 1459–1469.

Avigad, D., Matthews, A., Evans, B.W., Garfunkel, Z., 1992. Cooling during the exhuma-tion of a blueschist terrane — Sifnos (Cyclades), Greece. European Journal of Min-eralogy 4, 619–634.

Baxter, E.F., 2003. Natural constraints on metamorphic rates. In: Vance, D., Müller, W.,Villa, I.M. (Eds.), Geochronology: Linking the Isotopic Record with Petrology andTextures: Geol. Soc. Special Publication, 220, pp. 183–202. London.

Baxter, E.F., Ague, J.J., DePaolo, D.J., 2002. Prograde temperature–time evolution in theBarrovian type-locality constrained by Sm/Nd garnet ages from Glen Cova, Scotland.Journal of the Geological Society of London 159, 71–82.

Bebout, G.E., 1991. Field-based evidence for devolatization in subduction zones: implicationsfor arc magmatism. Science 251, 413–415.

Breeding, C.M., Ague, J.J., Bröcker, M., Bolton, E., 2003. Blueschist preservation in aretrograded, high-pressure, low-temperature metamorphic terrane, Tinos, Greece:implications for fluid flow paths in subduction zones. Geochemistry, Geophysics,Geosystems 4, 1–11.

Bröcker, M., 1990. Blueschist-to-greenschist transition in metabasites from TinosIsland, Cyclades, Greece: compositional control or fluid infiltration? Lithos 25,25–39.

Bröcker, M., Franz, L., 2006. Dating metamorphism and tectonic juxtaposition on An-dros Island (Cyclades, Greece): results of a Rb–Sr study. Geological Magazine143, 609–620.

Bröcker, M., Keasling, A., 2006. Ionprobe U–Pb zircon ages from the high-pressure/low-temperature mélange of Syros, Greece: age diversity and the importance of pre-Eocenesubduction. Journal of Metamorphic Geology 24, 615–631.

Bröcker, M., Pidgeon, R.T., 2007. Protolith ages of meta-igneous and metatuffaceousrocks from the Cycladic blueschist unit, Greece: results of a reconnaissance U–Pbzircon study. Journal of Geology 115, 83–98.

Camacho, A., Lee, J.K., Hensen, B.J., Braun, J., 2005. Short-lived orogenic cycles and theeclogitization of cold crust by spasmodic hot fluids. Nature 435, 1191–1196.

Christensen, J.N., Rosenfeld, J.L., DePaolo, D.J., 1989. Rates of tectonomorphic processesfrom rubidium and strontium isotopes in garnet. Science 244, 1465–1469.

Christensen, J.N., Selverstone, J., Rosenfeld, J.L., DePaolo, D.J., 1994. Correlation by Rb–Srgeochronology of garnet growth histories from different structural levels withinthe Tauern Window, Eastern Alps. Contributions to Mineralogy and Petrology118, 1–12.

Connolly, J.A.D., 2009. The geodynamic equation of state: what and how. Geochemistry,Geophysics, Geosystems 10, Q10014, http://dx.doi.org/10.1029/2009GC002540.

Dale, J., Holland, T., Powell, R., 2000. Hornblende–garnet–plagioclase‐thermometry: anatural assemblage calibration of the thermodynamics of hornblende. Contributionsto Mineralogy and Petrology 140, 353–362.

Ducea, M.N., Ganguly, J., Rosenberg, E.J., Patchett, P.J., Cheng, W., Isachsen, C., 2003.Sm–Nd dating of spatially controlled domains of garnet single crystals: a newmethod of high-temperature thermochronology. Earth and Planetary Science Letters213, 31–42.

Durr, S., Altherr, R., Keller, J., Okrusch, M., Seidel, E., 1978. The median Aegean crystallinebelt: stratigraphy, structure, metamorphism, magmatism. In: Cloos, H., Roeder, D.,Schmidt, K. (Eds.), Alps, Appenines, Hellenides 38. I.U.G.S. Rep, Stuttgart, pp. 455–477.

El-Shazly, Liou, J.G., 1991. Glaucophane chloritoid-bearing assemblages from NE Oman:petrologic significance and a petrogenetic grid for high P metapelites. Contribu-tions to Mineralogy and Petrology 107, 180–201.

Forster, M.A., Lister, G.S., 2005. Several distinct tectono-metamorphic slices in the Cy-cladic eclogite–blueschist belt, Greece. Contributions to Mineralogy and Petrology150, 523–545.

Fuhrman, M.L., Lindsley, D.H., 1988. Ternary-feldspar modeling and thermometry.American Mineralogist 73, 201–215.

Gaetani, G., Grove, T., 1998. The influence of water on melting of mantle peridotite.Contributions to Mineralogy and Petrology 131, 323–346.

Gaidies, F., de Capitani, C., Abart, R., 2008. THERIA_G: a software program to numericallymodel prograde garnet growth. Contributions to Mineralogy and Petrology 155,657–671.

Ganguly, J., Tirone, M., 1998. Diffusion closure temperature and the age of a mineralwith arbitrary extent of diffusion: theoretical formulation and applications. Earthand Planetary Science Letters 170, 131–140.

Gautier, P., Brun, J.P., 1994. Crustal-scale geometry and kinematics of late-orogenic exten-sion in the central Aegean (Cyclades and Evvia Island). Tectonophysics 238, 399–424.

Green, E., Holland, T., Powell, R., 2007. An order–disorder model for omphacitic pyrox-enes in the system jadeite–diopside–hedenbergite–acmite, with applications toeclogitic rocks. American Mineralogist 92, 1181–1189.

Groppo, C., Forster, M., Lister, G., Compagnoni, R., 2009. Glaucophane schists and asso-ciated rocks from Sifnos (Cyclades, Greece): New constraints on the P–T evolutionfrom oxidized systems. Lithos 109, 254–273.

Hacker, B.R., 2008. H2O subduction beyond arcs. Geochemistry, Geophysics, Geosystems 9,1–24.

Hacker, B.R., Abers, G.A., Peacock, S.M., 2003. Subduction factory 1. Theoretical miner-alogy, densities, seismic wave speeds, and H2O contents. Journal of GeophysicalResearch 108, 2029.

Halama, R., John, T., Herms, P., Hauff, F., Schenk, V., 2011. A stable (Li, O) and radiogenic(Sr, Nd) isotope perspective onmetasomatic processes in a subducting slab. ChemicalGeology 281, 151–166.

Harvey, J., Baxter, E.F., 2009. An improved method for TIMS high precision neodymiumisotope analysis of very small aliquots (1–10 ng). Chemical Geology 258, 251–257.

Holland, T.J.B., Blundy, J.D., 1994. Non-ideal interactions in calcic amphiboles and theirbearing on amphibole-plagioclase thermometry. Contributions to Mineralogy andPetrology 116, 433–447.

Holland, T., Powell, R., 1998. An internally consistent thermodynamic dataset for phases ofpetrologic interest. Journal of Metamorphic Geology 16, 309–343.

Holland, T., Baker, J., Powell, R., 1998. Mixing properties and activity-composition rela-tionships of chlorites in the system MgO–FeO–Al2O3–SiO2–H2O. European Journalof Mineralogy 10, 395–406.

Husen, S., Quintero, R., Kissling, E., Hacker, B., 2003. Subduction-zone structure andmagmatic processes beneath Costa Rica constrained by local earthquake tomogra-phy and petrological modeling. Geophysical Journal International 155, 11–32.

22 B. Dragovic et al. / Chemical Geology 314-317 (2012) 9–22

John, T., Schenk, V., 2003. Partial eclogitisation of gabbroic rocks in a late Precambriansubduction zone (Zambia): prograde metamorphism triggered by fluid infiltration.Contributions to Mineralogy and Petrology 146, 174–191.

John, T., Scherer, E., Haase, K.M., Schenk, V., 2004. Trace element fractionation duringfluid-induced eclogitization in a subducting slab: trace element and Lu–Hf/Sm–Ndisotope systematics. Earth and Planetary Science Letters 227, 441–456.

Kerrick, D.M., Connolly, J.A.D., 2001. Metamorphic devolatilization of subducted oceanicmetabasalts: implications for seismicity, arc magmatism and volatile recycling. Earthand Planetary Science Letters 189, 19–29.

Kohn, M.J., 2009. Models of garnet differential geochronology. Geochimica et CosmochimicaActa 73, 170–182.

Konrad-Schmolke, M., O'Brien, P.J., de Capitani, C., Carswell, D.A., 2008a. Garnet growthat high- and ultra-high pressure conditions and the effect of element fractionationon mineral modes and composition. Lithos 103, 309–332.

Konrad-Schmolke, M., Zack, T., O'Brien, P.J., Jacob, D.E., 2008b. Combined thermody-namic and rare element modeling of garnet growth during subduction: examplesfrom ultrahigh-pressure eclogite of the Western Gneiss Region, Norway. Earthand Planetary Science Letters 272, 488–498.

Lagos, M., Scherer, E.E., Tomaschek, F., Munker, C., Keiter, M., Berndt, J., Ballhaus, C.,2007. High precision Lu–Hf geochronology of Eocene eclogite-facies rocks fromSyros, Cyclades, Greece. Chemical Geology 243, 16–35.

Lapen, T.J., Johnson, C.M., Baumgartner, L.P., Mahlen, N.J., Beard, B.L., Amato, J.M., 2003.Burial rates during prograde metamorphism of an ultrahigh-pressure terrane: anexample from Lago di Cignana, western Alps, Italy. Earth and Planetary Science Let-ters 215, 57–72.

Lister, G.S., Raouzaios, A., 1996. The tectonic significance of a porphyroblastic blueschistfacies overprint duringAlpine orogenesis: Sifnos, AegeanSea, Greece. Journal of Struc-tural Geology 18, 1417–1435.

Ludwig, K.R., 2003. User's Manual for Isoplot 3.00. Berkeley Geochronology Center Spe-cial Publication, 4. Berkeley.

Marmo, B.A., Clarke, G.L., Powell, R., 2002. Fractionation of bulk rock composition dueto porphyroblast growth: effects on eclogite facies mineral equilibria, Pam Peninsula,New Caledonia. Journal of Metamorphic Geology 20, 151–165.

Matthews, A., Schliestedt, M., 1984. Evolution of the blueschist and greenschist rocksof Sifnos, Cyclades, Greece. Contributions to Mineralogy and Petrology 88,150–163.

Mibe, K., Fujii, T., Yasuda, A., 1999. Control of the location of the volcanic front in islandarcs by aqueous fluid connectivity in the mantle wedge. Nature 401, 259–262.

Mocek, B., 2001. Geochemical evidence for arc‐type volcanism in the Aegean Sea: theblueschist unit of Siphnos, Cyclades (Greece). Lithos 57, 263–289.

Okay, A.I., 2002. Jadeite–chloritoid–glaucophane–lawsonite blueschists in northwestTurkey: unusually high P/T ratios in continental crust. Journal of Metamorphic Geology20, 757–768.

Okrusch, M., Brocker, M., 1990. Eclogites associated with high-grade blueschists in theCyclades archipelago, Greece: a review. European Journal of Mineralogy 2, 451–478.

Okrusch, M., Seidel, E., Davis, E., 1978. The assemblage jadeite-quartz in the glaucophanerocks of Sifnos (Cyclades archipelago, Greece). Neues Jahrbuch fur Mineralogie,Abhandlungen 132, 284–308.

Peacock, S.M., 1990. Numerical simulation of metamorphic pressure–temperature–timepaths and fluid production in subducting slabs. Tectonics 9, 1197–1211.

Peacock, S.M., 1993. The importance of blueschist to eclogite dehydration reactions insubducting oceanic crust. Geological Society of America Bulletin 105, 684–694.

Plank, T., Langmuir, H., 1998. The chemical composition of subducting sediment and itsconsequences for the crust and mantle. Chemical Geology 145, 325–394.

Pollington, A.D., Baxter, E.F., 2010. High resolution Sm–Nd garnet geochrononolgy revealsthe uneven pace of tectnometamorphic processes. Earth and Planetary Science Letters293, 63–71.

Pollington, A.D., Baxter, E.F., 2011. High precisionmicrosampling and preparation of zonedgarnet porphyroblasts for Sm–Nd geochronology. Chemical Geology 281, 270–282.

Ring, U., Layer, P.W., 2003. High‐pressure metamorphism in the Aegean, easternMediterranean: underplating and exhumation from the Late Cretaceous until theMiocene to Recent above the retreating Hellenic subduction zone. Tectonics 22, 1022.

Rosenbaum, G., Avigad, D., Sanchez‐Gomez, M., 2002. Coaxial flattening at deep levelsof orogenic belts; evidence from blueschists and eclogites on Syros and Sifnos(Cyclades, Greece). Journal of Structural Geology 24, 1451–1462.

Rüpke, L.H., Morgan, J.P., Hort, M., Connolly, J.A.D., 2004. Serpentine and the subductionzone water cycle. Earth and Planetary Science Letters 223, 17–34.

Scherer, E., Cameron, K.L., Blichert-Toft, J., 2000. Lu–Hf garnet geochronology: Closuretemperature relative to the Sm–Nd system and the effects of trace mineral inclu-sions. Geochimica et Cosmochimica Acta 64, 3413–3432.

Schliestedt, M., 1986. Eclogite–blueschist relationships as evidenced by mineral equilibriain the high pressure rocks of Sifnos (Cycladic Islands), Greece. Journal of Petrology 27,1437–1459.

Schliestedt, M., Matthews, A., 1987. Transformation of blueschist to greenschist faciesrocks as a consequence of fluid infiltration, Sifnos (Cyclades), Greece. Contributionsto Mineralogy and Petrology 97, 237–250.

Schliestedt, M., Okrusch, M., 1988. Meta-acidites and silicic meta-sediments related toeclogites and glaucophanites inNorthern Sifnos, Cycladic Archipelago, Greece. In: Smith,

D.C. (Ed.), Developments in Petrology. : Eclogites and Eclogite-Facies Rocks, vol. 12.Elsevier, Amsterdam.

Schmadicke, E., Will, T.M., 2003. Pressure–temperature evolution of blueschist faciesrocks from Sifnos, Greece, and implications for the exhumation of high‐pressurerocks in the Central Aegean. Journal of Metamorphic Geology 21, 799–811.

Schmidt, M.W., Poli, S., 1998. Experimentally based water budgets for dehydratingslabs and consequences for arc magma generation. Earth and Planetary Science Letters163, 361–379.

Skora, S., Baumgartner, L.P., Mahlen, N.J., Johnson, C.M., Pilet, S., Hellebrand, E., 2006.Diffusion-limited REE uptake by eclogite garnets and its consequences for Lu–Hfand Sm–Ndgeochronology. Contributions toMineralogy and Petrology 152, 703–720.

Skora, S., Lapen, T.J., Baumgartner, L.P., Johnson, C.J., Hellebrand, E., Mahlen, N.J., 2009.The duration of prograde garnet crystallization in the UHP eclogites at Lago diCignana, Italy. Earth and Planetary Science Letters 287, 402–411.

Smye, A.J., Greenwood, L.V., Holland, T.J.B., 2010. Garnet–chloritoid–kyanite assemblages:eclogite facies indicators of subduction constraints in orogenic belts. Journal of Meta-morphic Geology 28, 753–768.

Spandler, C., Pettke, T., Rubatto, D., 2011. Internal and external fluid sources foreclogite-facies veins in the Monviso meta-ophiolite, Western Alps: implicationsfor fluid flow in subduction zones. Journal of Petrology 52, 1207–1236.

Spear, F.S., 1993. Metamorphic Phase Equilibria and Pressure–Temperature–TimePaths. Mineralogical Society of America, Washington, DC.

Spear, F.S., Wark, D.A., Cheney, J.T., Schumacher, J.C., Watson, E.B., 2006. Zr-in-rutilethermometry in blueschists from Sifnos, Greece. Contributions to Mineralogy andPetrology 152, 375–385.

Staudigel, H., Hart, S.R., Schmincke, H-U., Smith, B.M., 1989. Cretaceous ocean crust atDSDP Sites 417 and 418: carbon uptake from weathering versus loss by magmaticoutgassing. Geochimica et Cosmochimica Acta 53, 3091–3094.

Straub, S.M., Layne, G.D., 2003. Decoupling of fluids and fluid-mobile elements duringshallow subduction: evidence from halogen-rich andesite melt inclusions fromthe Izu arc volcanic front. Geochemistry, Geophysics, Geosystems 4, 9003.

Thöni, M., 2002. Sm–Nd systematics in garnet from different lithologies (Eastern Alps):age results, and an evaluation of potential problems for garnet Sm–Nd chronometry.Chemical Geology 185, 255–281.

Tirone, M., Ganguly, J., Dohmen, R., Langenhorst, F., Hervig, R., Becker, H.-W., 2005. Rareearth diffusion kinetics in garnet: experimental studies and applications. Geochimicaet Cosmochimica Acta 69, 2385–2398.

Trotet, F., Jolivet, L., Vidal, O., 2001a. Exhumation of Syros and Sifnos metamorphicrocks (Cyclades, Greece). New constraints on the P–T paths. European Journal ofMineralogy 13, 901–920.

Trotet, F., Jolivet, L., Vidal, O., 2001b. Tectono‐metamorphic evolution of Syros andSifnos Islands (Cyclades, Greece). Tectonophysics 338, 179–206.

Ulmer, P., 2001. Partial melting in the mantle wedge — the role of H O in the genesis ofmantle-derived ‘arc-related’ magmas. Physics of the Earth and Planetary Interiors127, 215–232.

Ulmer, P., Trommsdorff, V., 1995. Serpentine stability to mantle depths and subduction-related magmatism. Science 268, 858–861.

van Keken, P.E., Kiefer, B., Peacock, S., 2002. High-resolution models of subductionzones: Implications for mineral dehydration reactions and the transport of waterinto the deep mantle. Geochemistry, Geophysics, Geosystems 3, 1–20.

van Keken, P.E., Hacker, B.R., Syracuse, E.M., Abers, G.A., 2011. Subduction factory: 4.Depth‐dependent flux of H2O from subducting slabs worldwide. Journal of Geo-physical Research 116, B01401, http://dx.doi.org/10.1029/2010JB007922.

Vance, D., Harris, N., 1999. Timing of prograde metamorphism in the Zanskar Himalaya.Geology 27, 395–398.

Vance, D., O'Nions, R.K., 1990. Isotopic chronometry of zoned garnets: growth kineticsand metamorphic histories. Earth and Planetary Science Letters 97, 227–240.

Vance, D., O'Nions, R.K., 1992. Prograde and retrograde thermal histories from the cen-tral Swiss Alps. Earth and Planetary Science Letters 114, 113–129.

Vorhies, S.H., Ague, J.J., 2011. Pressure–temperature evolution and thermal regimes inthe Barrovian zones, Scotland. Journal of the Geological Society London 168,1147–1166, http://dx.doi.org/10.1144/0016-76492010-073.

Wada, I., Wang, K., 2009. Common depth of slab-mantle decoupling: reconciling diver-sity and uniformity of subduction zones. Geochemistry, Geophysics, Geosystems10, Q10009, http://dx.doi.org/10.1029/2009GC002570.

Wallace, P.J., 2005. Volatiles in subduction zone magmas: concentrations and fluxesbased onmelt inclusion and volcanic gas data. Journal of Volcanology andGeothermalResearch 140, 217–240.

Wallmann, K., 2001. The geological water cycle and the evolution of marine δ18Ovalues. Geochimica et Cosmochimica Acta 65, 2469–2485.

Wendt, I., Carl, C., 1991. The statistical distribution of the mean squared weighted de-viation. Chemical Geology 86, 275–285.

White, R., Pomroy, N.R., Powell, R., 2005. An in-situ metatexite–diatexite transition inupper amphibolites facies rocks from Broken Hill, Australia. Journal of Metamor-phic Geology 23, 579–602.

Wijbrans, J.R., Schliestedt, M., York, D., 1990. Single grain argon laser probe dating ofphengites from the blueschist to greenschist transition on Sifnos (Cyclades, Greece).Contributions to Mineralogy and Petrology 104, 582–593.