High precision Lu–Hf geochronology of Eocene eclogite-facies rocks from Syros, Cyclades, Greece

Chemical Geology 243 (2007) 16 – 35 www.elsevier.com/locate/chemgeo

High precision Lu–Hf geochronology of Eocene eclogite-facies rocks from Syros, Cyclades, Greece Markus Lagos a,b,⁎, Erik E. Scherer a , Frank Tomaschek a , Carsten Münker b,a , Mark Keiter a,b , Jasper Berndt a , Chris Ballhaus b a

Institut für Mineralogie, Westfälische Wilhelms-Universität Münster, Correnstrasse 24, D-48149 Münster, Germany b Mineralogisch-Petrologisches Institut und Museum, Rheinische Friedrich-Wilhelms-Universität Bonn, Poppelsdorfer Schloß, D-53115 Bonn, Germany Received 16 June 2006; received in revised form 14 April 2007; accepted 16 April 2007 Editor: R.L. Rudnick

Abstract Garnet-bearing high pressure, low temperature (HP/LT) metamorphic rocks of the Cycladic Blueschist Unit have been investigated by Lu–Hf geochronology. Eclogites from Syros Island yield precise ages of 52.2 ± 0.3 Ma, 51.4 ± 0.4 Ma, and 50 ± 2 Ma. Preserved major- and trace element growth zoning in garnet suggests that the closure temperature of the Lu–Hf system in garnet was higher than the estimated peak metamorphic temperature of ∼500 °C. Hence, Lu–Hf ages most likely reflect garnet growth in the studied samples. Our new Lu–Hf Grt ages are in excellent agreement with previously published U–Pb SHRIMP ages of metamorphic zircon and the maximum reported white mica Rb–Sr and Ar–Ar ages from Syros. Garnet growth in all samples apparently took place over a narrow time interval because the Lu–Hf ages cluster tightly even though the distribution of Lu between cores and rims varies among samples. We find no evidence for garnet generations that are significantly older or younger than the given isochron ages. This, in conjunction with the agreement among the ages of metamorphic zircon, the maximum white mica ages, and the Lu–Hf ages, strongly suggests that the 52 Ma age dates a single HP event that affected these rocks. © 2007 Elsevier B.V. All rights reserved. Keywords: Syros; Cyclades; HP/LT metamorphism; Eclogite; Lu–Hf geochronology; Garnet

1. Introduction Determining the age of peak-metamorphic mineral assemblages is an essential part in the reconstruction of pressure–temperature–time paths of rocks from fossil subduction zones. The Cycladic Blueschist Unit (CBU) ⁎ Corresponding author. Institut für Mineralogie, Westfälische Wilhelms-Universität Münster, Correnstrasse 24, D-48149 Münster, Germany. Tel.: +49 251 8333461; fax: +49 251 8338397. E-mail address: [email protected] (M. Lagos). 0009-2541/$ - see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.chemgeo.2007.04.008

extends from the Olympos Window (mainland Greece) through the Aegean Sea to the Turkish coast (Blake et al., 1981; Okay, 2001; Jolivet et al., 2004). The CBU is part of the Hellenides, an orogenic belt that formed during the convergence of the European and African plates (Jacobshagen, 1986). Numerous studies have attempted to constrain the age of the high-pressure metamorphism in the CBU by dating white micas with K–Ar, Ar–Ar, and Rb–Sr (e.g., Altherr et al., 1979; Maluski et al., 1987; Wijbrans et al., 1990; Bröcker et al., 1993; Baldwin, 1996; Bröcker and Enders, 2001; Putlitz et al.,

M. Lagos et al. / Chemical Geology 243 (2007) 16–35

2005). The results of these studies show that white mica ages attributed to HP/LT metamorphism in the Cyclades generally range between 38 and 53 Ma. The observed range can be potentially explained by 1) metamorphic cooling (Wijbrans et al., 1990; Lips et al., 1998), 2) excess argon (Altherr et al., 1979; Sherlock and Kelley, 2002), or 3) multiple mica generations (Tomaschek et al., 2003; Putlitz et al., 2005). A similar range between 40 and 65 Ma for metamorphic zircon has been reported by Keay (1998), Tomaschek et al. (2003), and Martin et al. (2006) for the Cyclades on the basis of U–Pb geochronology. Such zircons have been interpreted to have formed under a range of prograde- to retrograde conditions (Keay, 1998). On Syros, metamorphic zircon is intergrown with garnet that displays prograde growth zoning; texturally similar metamorphic zircon from a metaplagiogranite yields ∼ 52 Ma U–Pb ages that overlap with the oldest Ar–Ar white mica ages (Tomaschek et al., 2003). These authors therefore argued that 52 Ma is the best age estimate for the high-pressure metamorphic event. Although the age coincidence between two completely different dating methods (Ar–Ar mica and U–Pb zircon) is promising, it remains unclear whether these ages actually date the peak HP/LT assemblage. Putlitz et al. (2005), for example, argued that apparent phengite ages between 52 and 42 Ma all reflect prograde mica growth, with the oldest ages being “the best estimate of minimum crystallization age.” Zircon, although usually perceived as a highly stable phase, can be susceptible to loss of radiogenic Pb by a variety of mechanisms including recrystallization via coupled dissolution–reprecipitation (e.g., Tomaschek et al., 2003; Geisler et al., 2007). In addition, Spandler et al. (2004) have suggested that zircon can recrystallize even at very low temperatures and pressures (b100 °C and b0.2 GPa), during hydrothermal seafloor alteration. Zircon growth or recrystallization is thus possible over a wide range of pressure and temperature and does not necessarily date HP/LT metamorphic conditions. To date HP/LT metamorphism, it is preferable to use minerals of the stable assemblage that forms under those conditions, such as garnet + omphacite, rather than dating accessory minerals such as zircon, whose conditions of formation are not easily constrained. This approach is not always straightforward, however. For example, Kötz (1989) attempted to date garnet + omphacite assemblages from Syros eclogites with the Sm–Nd method and obtained an imprecise age of 67 ± 28 Ma. The large uncertainty may indicate that the mineral separates contained light rare earth element (LREE)-rich inclusions such as clinozoisite, which can be a very efficient sink for

17

the LREE. Clinozoisite is a common inclusion- and matrix phase in Syros eclogites, causing measured 147 Sm/144Nd and 143Nd/144Nd to be much lower than in pure (i.e., inclusion-free) garnet, which would tend to lower the precision of Sm–Nd isochrons (e.g., Amato et al., 1999; Frei et al., 2004). In the present study, we have chosen the Lu–Hf system for dating eclogite facies assemblages because it is relatively insensitive to the presence of LREE-rich, Hfpoor inclusions in garnet (Scherer et al., 2000). Garnet strongly fractionates Lu3+ from Hf4+ (e.g., Green et al., 2000; van Westrenen et al., 2000), and therefore bulk garnet fractions generally have higher Lu/Hf–and over time a much more radiogenic Hf signature–than the whole rock, enabling precise age determinations. One potential pitfall of Lu–Hf dating is the presence of Hf-rich inclusions such as zircon, which incorporates ∼1 wt.% Hf, or rutile, which can have a ∼50 times higher Hf concentration than coexisting garnet and is often a volumetrically significant inclusion. Scherer et al. (2000) have shown that analyzing garnets that contain zircon inclusions can result in imprecise ages, or even inaccurate ages if the zircon formed significantly before or after the garnet. Unfortunately, it is difficult to remove all zircon and rutile inclusions from eclogitic garnet separates using only standard physical techniques such as crushing, magnetic separation, and hand picking. We have therefore chosen a garnet digestion method that specifically avoids the dissolution of zircon and rutile. This method yields higher measured 176Hf/177Hf- and 176Lu/177Hf values that are more representative of the pure garnet. The resulting spread among mineral separates on isochrons is often larger than that achieved when garnets are digested together with their Hf-bearing inclusions, and thus the precision of Lu–Hf ages may be improved. Excluding zircon also minimizes the potential contribution of less radiogenic Hf inherited from previous petrogenetic events, thereby providing more accurate garnet ages in some cases (Scherer et al., 2000). Our new Lu–Hf ages tightly constrain the age of the eclogite-facies metamorphism on Syros to ∼ 52 Ma (Eocene), which is indistinguishable from the U–Pb ages of metamorphic zircons from the same sample locality (Grizzas Bay; Tomaschek et al., 2003, Fig. 1). 2. Geological outline The main lithological sequence of Syros is composed of alternating marbles and schists (Fig. 1), which are part of the Cycladic Blueschist Unit. These rocks and a tectonically fragmented, meta-ophiolitic sequence were juxtaposed and intercalated during a prograde,

18

M. Lagos et al. / Chemical Geology 243 (2007) 16–35

Fig. 1. Simplified geologic map of Syros modified after Ridley (1982), Hecht (1984), and Keiter et al. (2004). The two sampling locations, one on the coast at Grizzas Bay and the other one south of Kini are indicated. Syros is dominated by metasedimentary and meta-igneous rocks of the Cycladic Blueschist Unit, which is also known as the Lower Unit of the Attic Cycladic Crystalline Complex (Dürr, 1986). The Vari Unit is a distinct tectonic body belonging to the Upper Unit of the Cycladic Crystalline Complex (Ridley, 1982).

M. Lagos et al. / Chemical Geology 243 (2007) 16–35

penetrative deformation event characterized by isoclinal folding and thrusting (Ridley, 1982; Dixon and Ridley, 1987; Keiter et al., 2005). The meta-ophiolite sequence is most prominent in the North at Kampos, at the west coast of Syros between Kini and Finikas, and to the south of Hermoupolis (Hecht, 1984; see Fig. 1). The dismembered meta-ophiolite includes 1) former ocean-floor basalts, some with well-preserved pillow structures, that have recrystallized to Grt + Omp + Gln + Czo + Phe ± Pg ± Rt ± Qtz eclogite boudins within a glaucophane-rich, schistose matrix (Fig. 2a); 2) cumulate-textured metagabbro fragments in which magmatic pyroxene and plagioclase were replaced statically by Omp ± Gln and Czo ± Pg aggregates, respectively (Fig. 2b); 3) meta-igneous breccias with fragments and veins of basaltic- to plagiogranitic bulk composition, that have been recrystallized to Grt + Omp + Gln + Czo + Phe ± Pg ± Rt ± Qtz and Jd + Qtz + Pg ± Czo assemblages, respectively (Fig. 2c); and 4) Mn-rich ocean-floor metasediments now present as spessartine chert layers (e.g., Ridley, 1984; Dixon and Ridley, 1987; Seck et al., 1996; Putlitz et al., 2000; Bröcker and Enders, 2001; Marschall et al., 2006; mineral abbreviations after Kretz, 1983). The protoliths of these lithologic units were derived from a hydrothermally active oceanic crust that was most likely formed in a backarc setting (Seck et al., 1996; Lagos et al., 2002). Today they are found as isolated blocks and fragments within a highly strained, serpentinite or metasedimentary schist matrix, interpreted to be a tectonic mélange (e.g., Keiter et al., 2004). It is now generally accepted that the U–Pb SHRIMP zircon ages of ∼ 80 Ma that were obtained from metaplagiogranites and metagabbros represent the igneous protolith age of these rocks (Keay, 1998; Tomaschek et al., 2003; Bröcker and Keasling, 2006). The metamorphic history of the main sequence on Syros is characterized by blueschist- to eclogite-facies metamorphism, M1, and a retrograde greenschist overprint, M2 (e.g., Okrusch and Bröcker, 1990 and references therein). Most workers have concluded that the HP/LTmetamorphic rocks on Syros experienced a clockwise P– T-path, and reached peak conditions of approximately 1.5 GPa and 500 °C (Dixon, 1976; Ridley, 1984; Okrusch and Bröcker, 1990; Putlitz et al., 2005). The retrograde path was characterized by near-isothermal decompression during exhumation (Trotet et al., 2001; Putlitz et al., 2005; Marschall et al., 2006). 3. Sample description Three eclogite-facies rock samples were selected for this study. They all have the mineral assemblage Grt +

19

Omp + Gln + Czo + Phe + Qtz ± Pg ± Rt. The garnet porphyroblasts are unstrained, and the rock foliation, which is defined by omphacite, glaucophane, and mica, is deflected around the garnets, suggesting pre- to syndeformational garnet growth. The rims of both omphacite and garnet appear to have been in textural equilibrium. Samples Ag85 and Ag31 are eclogite fragments from a meta-igneous breccia at Grizzas (Fig. 1). Both are relatively fresh, but Ag31 also contains zoned epidote, actinolite, titanite, and chlorite

Fig. 2. Outcrop relations of meta-igneous rocks at the northeastern coast of Syros near Grizzas. a) Eclogite boudins in a dark, schistose, glaucophane-rich matrix. b) Cumulate-textured metagabbro with magmatic pyroxene and plagioclase statically replaced by omphacite and clinozoisite, respectively. c) Light metaplagiogranite fragments in a blue, glaucophane-rich, metabasaltic matrix. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

20

M. Lagos et al. / Chemical Geology 243 (2007) 16–35

as retrograde phases. The omphacite in this rock has been partially replaced by retrograde Ab + Qtz + Hem. The third sample, Ap21, is an eclogite from the Poulli peninsula south of Kini (Fig. 1) that is slightly more mafic and contains more garnet, glaucophane, and rutile than the ones from Grizzas. All samples were characterized by electron microprobe techniques to aid in interpreting the significance

of Lu–Hf ages. Particular attention was paid to garnet because it has the highest Lu/Hf of all major phases present and thus controls the spread of the mineral concentrates along an Lu–Hf isochron. Electron microprobe measurements indicate that garnets from all samples are Alm 0.49–0.77 Sps 0.01–0.23 Grs 0.15–0.38 Prp0.01–0.17 solid solutions characterized by bell-shaped manganese profiles (Figs. 3a, 4a–c), which are generally

Fig. 3. X-ray element maps for Mn, Ti, Y, and Zr of garnets from sample Ag85. Warm colors indicate higher amounts of an element. a) Garnet has Mnrich cores and decreasing Mn contents toward the rims. b) Ti-rich areas indicate the presence of rutile or titanite. c) Y content is elevated at the rimsand perhaps also within the garnet, where fluids may have gained access (indicated by white arrows). Garnets from the other samples lack these features. We note, however, that the apparently interior Y-rich zones might be just an artifact of the thin section plane intersecting an irregular 3-dimensional zoning pattern. d) Red areas are zircons which are commonly present as inclusions in garnet. Note that some Y-rich areas in c) also correspond to zircon in both the garnet and the matrix. All electron microprobe maps were obtained on a JEOL JXA 8600. Operating conditions were 50 nA and 15 kV. Y-Lα (TAP), Zr-Lα (PET), Ti-Kα (PET) and Mn-Kα (LIF) were monitored with counting times of 1 s on peak positions. e–f) backscattered electron images of Ag85 garnets. Lighter grey areas correspond to (Na,Y,HREE)–garnet. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

M. Lagos et al. / Chemical Geology 243 (2007) 16–35

21

Fig. 4. Chemical zoning across Alm–Grs rich garnets of a) Ag85, b) Ag31, and c) Ap21. Note that XFe decreases towards the rims. The profile length across the garnets is approximately 500 μm. Operating conditions for quantitative electron microprobe (EMP) measurements were 20 nA and 15 kV. Counting times were 20 s on peak positions and 5 s on background positions for Mg-Kα (TAP), Si-Kα (TAP), Al-Kα (TAP), Ca-Kα (PET), Fe-Kα (LIF) and Mn-Kα (LIF). Data reduction was achieved with ZAF corrections.

interpreted to indicate original growth zoning in garnet (Hollister, 1966; Spear, 1993; Kohn, 2003). Additionally, XFe values (molar Fe2+ / [Fe2+ + Mg2+]) decrease from core to rim, (Fig. 4a–c).

The Zr and Ti maps (Fig. 3) illustrate the distribution and frequency of zircon and rutile inclusions in garnet. Other frequent inclusions are clinozoisite, glaucophane, quartz, apatite, and omphacite. Most inclusion-rich garnets

Table 1 Representative analyses of outer rim (Na,Y, HREE)–garnet and inner rim of sample Ag85 garnet Outer rim (Na,Y, HREE–garnet) 1

2

Weight percent oxides 37.92 38.20 SiO2 TiO2 0.03 0.04 21.98 21.86 Al2O3 Y2O3 0.66 0.88 FeO⁎ 33.3 32.37 MnO 0.61 0.68 MgO 1.26 1.18 CaO 6.60 7.35 Na2O 0.28 0.30 Total 102.68 102.85 Cations on the basis of 12 oxygens Si 2.976 2.989 Al 2.033 2.016 Ca 0.555 0.616 Mg 0.147 0.138 Fe⁎ 2.189 2.118 Mn 0.041 0.045 Ti 0.002 0.002 Y 0.028 0.036 Na 0.043 0.046 Total 8.013 8.006

Inner rim

3

4

5

6

7

8

9

10

11

38.27 0.06 21.81 0.76 32.36 0.79 1.27 7.06 0.28 102.66

37.92 0.05 21.72 1.09 31.47 0.84 1.04 7.49 0.34 101.95

37.97 0.07 21.96 0.83 32.57 0.62 1.28 7.00 0.35 102.64

37.69 0.05 21.73 0.78 32.73 0.77 1.22 7.06 0.28 102.30

37.61 0.08 21.54 1.02 32.78 1.41 1.17 6.24 0.40 102.24

37.87 0.04 21.65 0.51 33.59 0.71 1.45 6.40 0.17 102.38

38.40 0.15 21.90 bDL 31.40 1.10 0.59 9.59 bDL 103.20

38.30 0.07 21.70 bDL 27.50 6.51 0.49 9.11 bDL 103.60

37.90 0.17 21.70 bDL 31.60 1.46 0.59 9.44 bDL 102.80

2.987 2.011 0.801 0.069 2.047 0.074 0.009 – – 7.997

2.984 1.991 0.760 0.057 1.790 0.430 0.004 – – 8.015

2.969 2.006 0.793 0.069 2.070 0.097 0.010 – – 8.014

3.002 2.026 0.477 0.166 2.215 0.048 0.000 0.022 0.040 7.994

2.991 2.019 0.633 0.122 2.076 0.056 0.003 0.046 0.052 7.999

2.977 2.030 0.588 0.149 2.136 0.041 0.004 0.035 0.054 8.013

2.972 2.020 0.596 0.143 2.158 0.051 0.003 0.033 0.042 8.020

2.976 2.009 0.529 0.138 2.169 0.095 0.005 0.043 0.061 8.025

2.983 2.010 0.540 0.170 2.213 0.047 0.002 0.021 0.026 8.013

Operating conditions during EMP measurements were 40 nA and 15 kV. Y-Lα (TAP) and Na-Kα (TAP) were monitored with counting times of 30 s both on peak as well as on background positions. Counting times for all other elements including Si-Kα (TAP), Mg-Kα (TAP), Al-Kα (TAP), Ca-Kα (PET), Ti-Kα (PET), Fe-Kα (LIF), and Mn-Kα (LIF) were 20 s on peak and 10 s on background positions. The ZAF method was applied for matrix corrections.

22

M. Lagos et al. / Chemical Geology 243 (2007) 16–35

Fig. 5. Element profiles across the rim of Ag85 garnet. EMP operating conditions are the same as described in Table 2. Note the abrupt change in element concentrations at the contact between inner and outer rim.

were eliminated by hand-picking. Any remaining inclusions except zircon and rutile were dissolved along with garnet in our digestion procedure, but do not significantly affect garnet ages because of their low Hf concentrations. A distinct Y-rich garnet zone is observed in sample Ag85 at the outermost 15–40 μm of garnet rims and perhaps associated with some cracks and channels in the garnet interiors (Fig. 3c, e, f). The sharp boundary between the two types of garnet is not more than ∼2 μm wide. Quantitative microprobe analyses of the Y-rich garnet (Table 1) show that Y is correlated with Na (Figs. 5, 6a). The negative correlation between Na + Yand Mg (Fig. 6b) indicates the substitution of Y3+ + Na+ for 2Mg2+ .

Analyzed sodian–yttrian garnet contains up to 3 mol% of the Na1.5(Y, HREE)1.5Al2Si3O12 end-member. Such peculiar garnet has been reported from UHP metamorphic terrains (Enami et al., 1995; Carswell et al., 2000). However, it is deeply entrenched in the petrological literature that the paragonite-out reaction has not been crossed by the lithological units of the CBU (Okrusch and Bröcker, 1990). We tentatively attribute the Na–Y-rich zone to an abrupt change in fluid composition near the end of continuous garnet growth. Alternatively, it could have resulted from a reaction between the original garnet rims and a fluid that infiltrated during a later event. (Below, we will argue that such an event could not have occurred more than a few Myr after the primary garnet growth.) Fluid inclusion studies have repeatedly shown that fluids in high-pressure rocks can be rich in Na+ and Cl− (e.g., Barr, 1990; Philippot and Selverstone, 1991; Giaramita and Sorensen, 1994; Philippot et al., 1998; Scambelluri et al., 1998), and Cl− is known to complex and mobilize Y and the REE quite efficiently (e.g., Flynn and Burnham, 1978; Wood, 1990). Sodian–yttrian garnet was not observed in sample Ag31, even though it came from the same metaigneous breccia as Ag85, suggesting that the formation of sodian–yttrian garnet was a highly localized phenomenon. Laser ablation inductively coupled plasma mass spectrometry (LA-ICPMS) element profiles through garnet crystals of all three samples indicate that Lu is generally positively correlated with Y (Table 2; Fig. 7a–c). For the Ag31 garnet, concentrations of both elements and Tb as well are highest in the garnet core and decrease towards the rims. This relatively simple pattern can be potentially explained by Rayleigh fractionation between growing garnet and the surrounding rock matrix (Otamendi et al., 2002). The Y and Lu profiles for Ag85

Fig. 6. a) Atomic correlation between Na and Y in rims of Ag85 garnet. b) Correlation between Na + Y and Mg points to the substitution mechanism Na+ + Y3+ = 2Mg2+.

M. Lagos et al. / Chemical Geology 243 (2007) 16–35 Table 2 LA-ICPMS Y, Tb, and Lu concentration profiles of Ag85, Ag31, and Ap21 garnets Sample

Spot

Distance [μm]

Y [ppm]

Tb [ppm]

Lu [ppm]

Ag31 grt

1 2 3 4 5 6 7 1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8

0 65 141 212 288 400 453 0 84 142 226 310 362 433 517 569 646 691 756 0 80 240 318 376 540 596 646

1810 1950 2530 4960 2580 1580 2290 4480 393 347 196 622 622 434 156 384 610 937 3870 91.4 67.0 33.7 132 91.8 35.5 57.8 54.1

7.2 7.8 6.5 12.1 6.0 7.5 10 18 2.0 2.1 1.4 2.7 2.1 2.0 1.1 2.0 3.9 5.3 15 bDL bDL bDL 2.2 bDL bDL bDL bDL

30 32 49 240 44 20 21 90 9.5 9.5 16 46 46 27 11 13 15 25 85 1.2 4.3 2.2 16 11 1.5 2.6 3.1

Ag85 grt

Ap21 grt

Concentrations were measured with a Finnigan Element II sector field ICPMS coupled to a 193 nm ArF-excimer laser. Runs were performed in low resolution mode (M/ΔM = 300) with laser settings of 5 Hz, ∼ 10 J/cm2, 30 μm spot size, and ∼ 1 min total acquisition time per run (30 s gas blank, 30 s data). NIST 612 glass was used as an external standard. NIST 610 was measured several times and treated as an unknown to monitor drift. Concentrations were obtained from 89Y, 159 Tb, and 175Lu background corrected signal intensities using 44Ca as an internal standard. Ca concentration was checked close to each laser pit with EMP to account Ca zoning in the garnet. Uncertainties (2 s.d.) of element concentrations were estimated to be 5%, 13%, and 11% for Y, Tb, and Lu, respectively.

and Ap21 are more complex, with secondary concentration maxima at or near the rims. Such patterns may form as the result of open-system behavior (of the rock matrix) during garnet growth, breakdown of an Y + HREE phase, or, as recently proposed by Skora et al. (2006), a changing relationship among garnet growth rate, diffusion rates of these elements through the bulk rock matrix, and the size of the matrix diffusion domains. In the Ag85 garnet profile, a pronounced increase in Lu and Y is observed at the contact between the inner rims and the outer rims of sodian–yttrian garnet. This feature is restricted to the outermost 15–40 μm of the grains and more clearly resolved on the Y map and profile in Figs. 3c and 5.

23

4. Sample preparation and digestion The aim of the sample preparation was to obtain relatively pure mineral separates to ensure a maximum spread in the Lu/Hf of isochrons. All rocks were crushed in a steel mortar. Mineral separates (garnet, omphacite, titanite, and zircon) were obtained by sieving and concentrating with a Frantz magnetic separator. To avoid biasing the bulk garnet toward core or rim components, garnet should be subjected to magnetic separation only to remove non-magnetic minerals, taking care not to split the garnet into different magnetic fractions (Lapen et al., 2003). For enhanced purification and to exclude mineral phases visibly rich in inclusions, the separates were handpicked under a binocular microscope. The mineral separates were washed in cold 2.5 M HCl for about 10 min to remove any surface contamination, and then rinsed several times in deionized water. Before digestion, samples were spiked with a mixed 176 Lu–180Hf tracer for isotope dilution determinations of 176 Lu/177Hf. Omphacite, titanite, and zircon separates were digested with an HF–HNO3 acid mixture (5:1) in Savillex® vials placed inside steeljacketed Parr® Teflon® bombs. Omphacite and titanite were held for 24 to 48 h at 200 °C, whereas the more refractory zircon fractions were bombed for 5 days at 200 °C to ensure complete digestion. To selectively dissolve the garnet fractions while leaving the Hf-bearing phases zircon and rutile largely intact, they were digested in closed Teflon® vials on a hotplate rather than in high-pressure Parr® bombs. This method, hereafter referred to as “tabletop” digestion, has been routinely used by many workers for recently erupted basalts and is essentially the same procedure used by Blichert-Toft et al. (1997) for ancient samples, and by Duchêne et al. (1997) specifically for Lu–Hf analyses of Eocene- to Early Oligocene eclogites and their mineral separates. Although Blichert-Toft et al. (1997) implied that spike-sample equilibration would be better achieved in high-pressure bombs rather than by tabletop digestion, these authors were referring to digestion of the entire sample, including refractory minerals such as zircon, and its equilibration with the spike. Here, we specifically try to avoid digesting zircon (e.g., Scherer et al., 2003; Connelly, 2006; Choi et al., 2006), while still obtaining spike-sample equilibration. The “sample” in this case refers to garnet plus inclusions except zircon and rutile (i.e., omphacite, glaucophane, clinozoisite, quartz, and micas). Minor changes were made to the tabletop procedure to increase the likelihood of spike-sample equilibration and to test that such equilibration has occurred. Typically, 100–200 mg of

24

M. Lagos et al. / Chemical Geology 243 (2007) 16–35

Fig. 7. LA-ICPMS Y, Lu, and Tb profiles across garnets. The symbol width corresponds to the diameter of the laser ablation spots (30 μm). Ag31 garnets (a) and Ap21 garnets (c) show elevated Lu and Y concentrations in the cores with decreasing concentrations toward the rims. The Ag85 garnet (b) shows an additional elevation in Y + Lu + Tb concentration at the rims that exceeds that in the core. (This grain is not shown in Fig. 3.) The outermost rims comprise a generation of (Na,Y,HREE)–garnet marked by vertical grey shading. d–f) Distribution of Lu in concentric shells representing 10% volume steps from core to rim. These were approximated from the LA-ICPMS data (a–c). Note that the Lu distribution in Ag85 (b, e) is skewed toward the rims, whereas that in samples Ag31 (a,d) and Ap21 (c,f) is slightly skewed toward the core.

garnet were digested in HF–HNO3–HClO4 and then 10 M HCl (cf. 6 M used by Blichert-Toft et al., 1997), drying down at high temperature (fuming HClO4)

between steps. If any fluoride-based precipitates remained after the HCl step or if the solution was cloudy, the digestion procedure was repeated until the

M. Lagos et al. / Chemical Geology 243 (2007) 16–35

solution was clear and no solids were visible except perhaps a few rutile grains, at which point it was assumed that garnet and its soluble, low-Hf inclusions were in equilibrium with the spike. As discussed later, this assumption is tested by analyzing at least three garnet fractions for every dated rock. Clear solutions were typically obtained after two digestion cycles, but as many as four may be required. After digestion, the solutions were dried down and the residues were redissolved in 2–3 M HCl and centrifuged to remove undigested zircon and rutile inclusions and any newly formed (post sample-spike equilibration) precipitates before loading onto ion exchange columns. For the whole-rock aliquots, fractions of crushed material were powdered in an agate mill. To ensure the complete dissolution of all phases, including refractory zircon and rutile, the whole-rock powders were melted with lithium-tetraborate flux (1 part wholerock powder to 5 parts Li2B4O7) in graphite crucibles at 1150 °C. The melts were quenched and dissolved in 3 M HCL, spiked, and then equilibrated on a hotplate in closed Savillex® vials for 24 h. The solutions were then evaporated to dryness and the residues treated four times by adding concentrated HNO3 and evaporating to dryness. Finally, the samples were dissolved in 3 M HCl and loaded onto ion exchange columns. 5. Chemistry and mass spectrometry The separation of Lu and Hf from the matrix was accomplished in a single-column chemistry step using Eichrom Ln-spec resin, following the procedure of Münker et al. (2001). This routine was slightly modified to handle high Lu/Hf mineral separates, such as garnet, as follows. After obtaining the Lu (+Yb) cut, the columns were rinsed with 20- to 40 resin-bed volumes of 6 M HCl to more effectively remove remaining Lu and Yb before collection of the Hf cut. In addition, the Hf cut itself was eluted with 2 M HF instead of 6 M HCl–0.2 M HF. These modifications to the separation procedure were found to be essential for eliminating interferences of 176 Lu and 176Yb on 176Hf during multiple collector (MC)-ICPMS analyses of garnet. The Hf and Lu isotope ratios of all samples were determined by MC-ICPMS using a Micromass Isoprobe in static mode. For Lu isotope dilution measurements of whole-rock and bombed mineral separates carried out early in the study, Lu cuts were doped with Re to a concentration of 50 ppb and externally normalized to 187 Re/185Re = 1.6738 (De Bièvre and Taylor, 1993). The corresponding (i.e., 187Re/185Re-normalized) 176 Yb/

173

25

Yb used to correct for the interference of 176Yb on Lu in these early runs was 0.7939 (Nebel-Jacobsen et al., 2005). Without further removal of Yb from the Lu cut, the external reproducibility of Lu concentrations and 176 Lu/177Hf measured with this Re-doping method is normally 1%, 2 s.d. (cf. 0.2% if an additional column is used to separate Yb from Lu, e.g., Scherer et al., 2001). Later in the study (i.e., for all tabletop-digested samples), we used the trend of ln(176Yb/171Yb) vs. ln(174Yb/171Yb) from several Yb-only standards run during the Lu measurement session to improve the accuracy of the 176 Yb interference correction (e.g., Maréchal et al., 1999; Blichert-Toft et al., 2002; Barfod et al., 2003; Albarède et al., 2004; Vervoort et al., 2004). This procedure typically results in an external 176Lu/177Hf reproducibility of ∼0.1% (2 s.d.) based on replicate analyses of ideally spiked sample solutions that have Yb/Lu of ∼2 after chemistry. For these analyses, the effects of error magnification due to non-ideal spike-sample ratios are also included in the estimated 2 s.d. external reproducibilities reported in Table 3. For some Grt separates from Ag85 and Ag31, the Lu concentrations were unexpectedly high, and Lu was inadvertently underspiked, resulting in uncertainties in 176Lu/177Hf of up to 4%. Hafnium isotope ratios were corrected for instrumental mass bias by using the exponential law and a 179 Hf/177 Hf = 0.7325. For MC-ICPMS analyses, the exponential law does not completely correct instrumental mass bias. We therefore applied a secondary massbias correction to the exponentially-corrected 176 Hf/177 Hf of unknowns that is based on the average exponentially-corrected 176 Hf/177 Hf of ten or more analyses of our in-house Ames Hf standard made during the same measurement session. The isotope composition of our in-house standard is indistinguishable from that of the JMC-475 standard, so all 176Hf/177 Hf values for the run session are multiplied by a factor such that the average for standards is 0.282160. The mean withinsession adjustments to measured 176 Hf/177Hf values were small, averaging + 40 ppm, and ranging between − 21 and + 140 ppm. This correction, and an analogous one for 180 Hf/177Hf, were applied before simultaneous spike-stripping and exponential mass bias correction. The external reproducibility of 176 Hf/177Hf for samples was estimated using the relationship between internal run statistics (% s.e.) and within-measurement-session reproducibilities (% 2 s.d.) of Hf standard solutions that cover a wide range of concentrations (8 to 80 ppb), following the general approach of Bizzarro et al. (2003). Typical 2 s.d. within-session reproducibilities for 8, 24, 40, and 80 ppb Hf solutions are ± 1.3, ± 0.7, ± 0.5, and ± 0.3 ε-units, respectively. There was no significant 176

26

M. Lagos et al. / Chemical Geology 243 (2007) 16–35

Table 3 Lu–Hf isotope data set of the examined HP/LT-metamorphic rocks from Syros Sample

Digestion

Lu (ppm)

Hf (ppm)

176

Lu/177Hf

% 2 s.d.

176

Hf/177Hf

% 2 s.d.

Ag31 Wr a Grt 1 a Grt 2 a Grt 3 a Grt 4 a Grt 5 a Grt 6 a Grt 7 a Leftover Zrn Ttn

flx bmb bmb bmb tt tt tt tt tt bmb bmb

1.96 67.5 76.7 72.7 68.9 70.5 51.8 66.9 0.726 446 9.29

18.3 22.1 13.0 8.51 0.568 0.299 0.186 0.195 0.155 7720 8.71

0.0152 0.434 0.835 1.21 17.3 33.6 39.78 49.2 0.6640 0.00820 0.151

1 1 1 1 3.2 4.0 0.16 3.4 0.66 1 1

0.283106 ± 7 0.283548 ± 10 0.283935 ± 7 0.284234 ± 9 0.300066 ± 7 0.315364 ± 23 0.321870 ± 45 0.332742 ± 11 0.283786 ± 14 0.283052 ± 13 0.283251 ± 14

0.0035 0.0058 0.0035 0.0047 0.0137 0.0167 0.0577 0.0167 0.0096 0.0090 0.0096

Ag85 Wr a Omp a Grt 1 a Grt 2 a Grt 3 a Grt 4 a Grt 5 a Grt 6 a Grt 7 a Leftover a

flx bmb bmb bmb bmb tt tt tt tt tt

2.07 1.72 30.4 30.6 34.3 35.3 35.4 39.1 39.8 0.606

22.7 18.7 14.2 9.88 3.86 0.266 0.189 0.137 0.116 0.257

0.01297 0.01299 0.304 0.440 1.26 18.9 26.7 40.81 49.1 0.3348

1 1 1 1 1 1.9 2.0 1.7 1.0 0.97

0.283120 ± 10 0.283125 ± 8 0.283427 ± 10 0.283551 ± 11 0.284285 ± 7 0.301548 ± 21 0.308933 ± 11 0.322477 ± 14 0.329949 ± 28 0.283429 ± 15

0.0058 0.0047 0.0063 0.0069 0.0036 0.0150 0.0063 0.0080 0.0194 0.0107

Ap21 Wr a Omp a Grt 1 a Grt 2 a Grt 3 a Grt 4 a Grt 5 a Grt 6 a Leftover a

flx bmb bmb bmb bmb tt tt tt tt

0.514 0.136 2.65 2.62 2.56 2.41 2.34 2.38 0.231

4.37 2.15 3.29 3.09 2.90 0.293 0.264 0.268 0.242

0.0167 0.00898 0.115 0.120 0.125 1.167 1.258 1.258 0.1356

1 1 1 1 1 0.14 0.16 0.18 0.25

0.283075 ± 11 0.283091 ± 17 0.283188 ± 9 0.283154 ± 20 0.283223 ± 20 0.284135 ± 12 0.284248 ± 32 0.284257 ± 14 0.283213 ± 16

0.0074 0.0129 0.0052 0.0156 0.0156 0.0080 0.0270 0.0101 0.0123

Reported uncertainties in the 6th digit of Hf isotope ratios are 2 s.e. (2σ/√n) in-run statistics. The estimated external reproducibilities of 176Hf/177Hf values are indicated as % 2 s.d. and were estimated from the observed relationship between the in-run and external reproducibility of Hf standard solutions following the approach of Bizzarro et al. (2003). For samples that were inadvertently underspiked for Lu, the resulting error magnification factor has been included in the 176Lu/177Hf uncertainty. The “leftover” fractions comprise minerals that remained after most of the garnet and omphacite had been separated from the bulk crushed sample (N180 μm grain size). Digestion types: flx = flux digestion, bmb = bomb digestion, tt = table top digestion. a Samples included in the regressions.

difference among the measured 176Hf/177 Hf values for different concentrations of Hf standard. Typical analytical blanks for the tabletop digestion method were 30 pg Hf and 10 pg Lu. Hafnium and Lu blanks for the bomb digestions averaged 15 and 12 pg, respectively. Blanks for Li-tetraborate digestions were 65 pg Hf and 12 pg Lu. 6. Results and discussion All isotope ratios and concentrations are listed in Table 3. Ages and their uncertainties were calculated

using the Isoplot/Ex 2.06 program of Ludwig (1999) and a 176 Lu decay constant of λ = 1.867 × 10− 11 yr− 1 (Scherer et al., 2001, 2003; Söderlund et al., 2004). Isochrons (Fig. 8) were constrained by the flux-digested whole rock, a bomb-digested omphacite fraction, and several garnet aliquots, except for Ag31, in which the omphacite was too altered by a retrograde overprint to be included in the regression. In the left-hand column of Fig. 8, the garnets on the isochrons were digested with the tabletop method, whereas in the right-hand column, the garnets were all digested in bombs. In addition, the isochrons of the samples Ap21 and Ag85 were also complemented by

M. Lagos et al. / Chemical Geology 243 (2007) 16–35

27

Fig. 8. Lu–Hf isochron plots for eclogite samples Ag31, Ag85, and Ap21. Isochrons for which garnets were digested using the tabletop method are shown in the left column (a–c) and those for which garnets were digested in bombs are shown in the right column (d–f). Model 1 isochron ages are reported at the 95% confidence level. Error bars indicate the estimated 2σ uncertainties given in Table 3. Where no error bars are shown, the estimated uncertainties are less than or equal to that represented by the size of the plot symbols. Data point labels correspond to the fractions listed in Table 3.

28

M. Lagos et al. / Chemical Geology 243 (2007) 16–35

analyses of the bulk sample that remained after garnet and omphacite were separated from the crushed whole rock. This fraction, designated “leftover” in Table 3, was digested using the tabletop method. Zircon and titanite separates were not included in the regressions because they were interpreted to have formed before (the inherited fraction of the zircon) or after (titanite) the eclogitic assemblage. The “leftover” fractions may contain retrograde minerals, but zircon would have been effectively excluded from the analyses during digestion. A five-point isochron from sample Ag31 from the Grizzas meta-igneous breccia, defined by a whole-rock aliquot and four tabletop-digested garnet separates, yields an age of 52.2 ± 0.3 Ma with an MSWD of 1.7 (Fig. 8a). A close-up view of the low-Lu/Hf points (Fig. 8a inset) shows additional data that are not included in the regression. The zircon fraction falls below the isochron, suggesting that at least some of the zircons contained an older, inherited Hf component. Titanite is a retrograde phase in eclogites of Syros and was therefore excluded from the regression, as was the “leftover” fraction, which is rich in titanite, white mica, chlorite, and altered omphacite. A 4-point isochron from the same sample, in which the garnets were dissolved in bombs (Fig. 8d) gives an age of 52.1 ± 8.4, which is identical to–but less precise than–the age obtained when digesting garnets using the tabletop procedure. The maximum 176Lu/177Hf of the bombed garnet is ∼1.2 (Grt 3, Fig. 8d) as compared to 49 achieved by tabletop digestion (Grt 7, Fig. 8a). The initial 176 Hf/177Hf value of both isochrons agree to within error, but the MSWD of 24 of the bombed garnet isochron (Fig. 8d) is much higher than that of the isochron based on tabletop-digested garnets (MSWD = 1.7; Fig. 8a). We attribute the low Lu/Hf of the bombed garnets and the pronounced excess scatter of the resulting isochron to zircon (± rutile) inclusions that were digested along with the garnets. Mixing between pure garnet (no inclusions) and Hf-rich inclusions can reduce the Lu/Hf of the bulk digested garnet. This shortens isochrons, but does not necessarily introduce scatter unless some of the inclusions were not in initial isotopic equilibrium with garnet. This is clearly possible for sample Ag31, because its bulk zircon fraction lies below the Grt–WR isochron, most likely indicating that at least some of this zircon is older than the Grt–WR age. The whole rock will also contain this component, but its Hf composition is buffered by other phases in the rock. In fact, the inherited zircon probably does not affect the Ag31 Lu–Hf age much because the flux-digested whole rock, which contains zircon, lies on the isochron defined by the tabletop-digested garnet and “leftover” fractions, which should be minimally affected by zircon.

A 7-point isochron from sample Ag85, defined by four tabletop-digested garnet separates, omphacite, the “leftover” fraction, and the whole rock gives an age of 51.4 ± 0.4 Ma, with an MSWD of 1.3 (Fig. 8b). The corresponding age of 49.8± 2.9 Ma (Fig. 8e) obtained using bomb-digested garnets is again less precise and has a higher MSWD (5.3), but the initial 176Hf/177Hf of both isochrons are identical within error. The highest 176Lu/ 177 Hf among bombed garnet fractions of sample Ag31 is ∼ 1.3 (Fig. 8e) which is distinctly lower than the 176 Lu/ 177 Hf of 49 achieved by tabletop digestion (Fig. 8b). In the case of the bombed garnet isochron, low 176Lu/177Hf values for garnet most likely points to the presence of zircon inclusions in the Ag85 garnets. On an isochron based upon the four tabletop-digested Ag85 garnets, using the tabletop-digested “leftover” fraction as the low-Lu/Hf point yields the same age and initial Hf isotope composition as using the zircon-bearing whole rock and omphacite fractions, suggesting that the inherited component in the zircon is minor. A 6-point isochron for eclogite Ap21 from Poulli peninsula, defined by the whole rock, omphacite and “leftover” fractions, and three tabletop-digested garnet separates, yields an age of 50 ± 2 Ma with an MSWD of 1.7 (Fig. 8c). The bombed garnet separates of Ap21 yield a poorly constrained age of 57 ± 26 Ma (Fig. 8f). The MSWD of 1.9 is only slightly higher than that obtained by the tabletop procedure. The precision in this case seems to be limited because the isochron is “short,” such that the uncertainties in 176Hf/177 Hf are relatively large as compared to the total range in these values. The initial 176 Hf/177Hf values of both isochrons agree to within error. Maximum 176 Lu/177Hf values are ∼ 10 times lower for bomb-digested garnets of sample Ap21 than for garnet separates digested by the tabletop procedure. Again, this strongly points to the presence of zircon in the garnet separates of Ap21. If complete removal of Hf-rich inclusions was achieved by the tabletop digestions, then the 176Lu/177 Hf values of the Ap21 garnets (∼ 1.2) are much lower than those of the Ag31 and Ag85 garnets (17–49), demonstrating that such inclusions are not the only control on the Lu/Hf of bulk garnet. The supply of trace elements to the surface of growing garnets also plays an important role as discussed by Skora et al. (2006). When including the tabletop- rather than bombdigested garnets on Lu–Hf isochrons, the ages of all three samples are more precise, and still nearly identical within error, with only a small (0.1 to 1.5 Myr) difference between the ages of Ag31 and Ag85. Likewise, the initial 176Hf/177Hf values of all three samples are almost identical within error. For samples

M. Lagos et al. / Chemical Geology 243 (2007) 16–35

Ag31 and Ap21, the slight excess scatter indicated by MSWD values of 1.7 could conceivably stem from analyzing garnets that have a small difference in age (probably not more than a few Myr) between core and rim. Different fractions of crushed and separated garnet most likely contain different proportions of core and rim material and may therefore exhibit slight differences in average age. The high 176Lu/177Hf values of garnet reported here, up to 49.2, are distinctly higher than those normally achieved with bomb digestion methods (e.g., Scherer et al., 2000; Blichert-Toft and Frei, 2001; Philippot et al., 2001; Lapen et al., 2003) and are among the highest ratios ever reported for this mineral (e.g., Duchêne et al., 1997; McClelland et al., 2005). To a large extent, these high Lu/ Hf values can be attributed to the tabletop digestion technique employed here. We suggest that the high Lu/Hf values approach those of the true, inclusion-free composition of the garnet and are not artifacts of incomplete garnet digestion or spike-sample equilibration. This is because at one point in the digestion (in hot 10 M HCl) a totally clear solution is achieved, with the visible residue– if present–comprising undigested rutile and zircon, which can be readily identified using a microscope and crossed polarizers. At this point, both the Hf and Lu of the digested garnet sample are presumed to be in isotopic equilibrium with the spike. This assumption was tested for every sample by analyzing three or four garnet fractions.

29

Failure of the spike in one or more tabletop-digested samples to equilibrate isotopically with the dissolved garnet would be expected to produce significant excess scatter on isochrons. This test is analogous to that applied by Lapen et al. (2004) to demonstrate that spike-sample equilibration had occurred during bomb digestions of various physical mixtures of 1 Ga garnet and amphibole. Precipitates or gels sometimes form after the digestion when samples are evaporated, re-dissolved in ∼3 M HCl, and centrifuged at room temperature to prepare them for ion-exchange chemistry. However, spike-sample equilibration has already been achieved by this point, so the isotope dilution determinations of Lu and Hf concentrations would not be affected. The garnets of each rock sample have 176Lu/177 Hf values that correlate strongly with 1/Hf regardless of digestion type. For the bombed garnets, we attribute this mostly to digesting mixtures of pure garnet and their Hfrich inclusions. For the tabletop digestions, this could indicate variable amounts of eclogite facies inclusions that have low Lu/Hf such as omphacite that were completely digested along with garnet. Partial dissolution of zircon (or rutile) is also a possibility, but again, these minerals will only affect age accuracy if they are inherited or retrograde phases. In any case, any adverse effects of the zircon or rutile components on ages should be lessened in tabletop digestions as compared to bomb digestions. Another explanation for the observed range

Fig. 9. Compilation of age data from HP/LT-metamorphic rocks in the Cycladic Blueschist Unit (CBU) on Syros. The vertical dashed line and grey bar indicate the weighted mean and 95% confidence limits, respectively, of the three “tabletop” Lu–Hf eclogite ages from this study. Note the overlap among the Lu–Hf ages, the younger group of U–Pb zircon ages, and the maximum white mica ages, which we interpret to date the HP/LT metamorphism in the CBU.

30

M. Lagos et al. / Chemical Geology 243 (2007) 16–35

in 176Lu/177Hf among different garnet fractions from the same sample could simply be that Hf concentration in the pure garnets varies, either between porphyroblasts or among growth zones within porphyroblasts. Such heterogeneity may not be eliminated during the mineral separation process. Unfortunately, the Zr and Hf contents in the pure garnet were always below the LAICPMS detection limit for both elements, so we could not directly measure the Lu/Hf of inclusion-free spots or investigate how Lu/Hf changes from core to rim. However, the low implied Hf concentrations (b1 ppm) are consistent with those of tabletop-digested garnets (0.12–0.57 ppm) but not those of bomb-digested garnets (2.9–22 ppm). This strongly suggests that pure garnet in the studied rocks really does have sub-ppm Hf contents and that relatively insoluble, Hf-rich inclusions are the primary cause for the elevated Hf contents, low 176 Lu/177 Hf, and low 176 Hf/ 177 Hf of the bombed garnets. Of the identified inclusion phases in the garnets, only zircon and perhaps rutile have high enough Hf concentrations to produce such a pronounced effect on the garnet compositions. To some degree, however, differences in garnet 176 Lu/177 Hf values among the three rock samples also reflect the average Lu contents of these garnet populations (Table 3), which in turn depend on the Lu concentration of the whole rocks and the modal amount of garnet. For example, the Ap21 whole rock has only one fourth of the Lu concentration found in the other two samples, and much of that Lu resides in garnet, which is more abundant in Ap21 than in the other samples. This could easily explain the relatively low Lu concentrations of the Ap21 garnet. Lutetium concentration varies within single garnet crystals as well, as seen in the trace element profiles (Fig. 7), in which Lu varies by a factor of 9 to 13 over rim-to-rim traverses. Garnets from Ag31 and Ag85 both have generally high Lu concentrations in their inner cores (∼ 240 and ∼45 ppm respectively) that decrease in the outer cores (∼ 20 and ∼ 5 ppm). High Lu concentrations (∼ 90 ppm) also occur in the sodian–yttrian garnet generation at the outermost rims of the Ag85 garnet (Fig. 7a). Lapen et al. (2003) showed that Rayleigh fractionation during garnet growth could concentrate Lu in the cores of garnet relative to the rims, which would bias the Lu–Hf ages of bulk garnet separates toward the age of the early-grown cores. The garnets in our present study have Lu profiles that are more complex than would be expected from simple Rayleigh fractionation during constant porphyroblast growth (Fig. 7a–c, cf. Fig. 4a of Lapen et al., 2003), and may have been influenced by other processes such as those described by Skora et al. (2006). The

fractions of the total garnet Lu that contained each successive 10% volume shell from core to rim are shown in Fig. 7d–f. For the Ag85 garnet, the bulk of the Lu is skewed strongly toward the rim, whereas garnets from Ag31 and Ap21 have a more evenly distributed Lu, skewed slightly toward the core. This observation will be important in our interpretation of the Lu–Hf ages below. 6.1. Interpretation of Lu–Hf ages and comparison with other metamorphic ages The preservation of Mn growth zoning in the garnets suggests that the original distributions of Lu3+ and Hf 4+ have not been disturbed by diffusion because 2+ ions diffuse significantly faster than 3+ (and presumably 4+) ions in garnet (e.g., Van Orman et al., 2002). Thus the closure temperature of the Lu–Hf system in these garnets is apparently greater than the peak metamorphic temperature of ∼ 500 °C. This is broadly consistent with the closure temperature of ∼ 600 °C or higher inferred for granulite garnets that have radii similar to those in this study (∼ 0.2–0.3 mm; data of Scherer et al., 2000, but using the revised 176 Lu decay constant of 1.867 × 10− 11 yr− 1). We therefore interpret the Lu–Hf ages to date garnet growth rather than closure of the Lu– Hf system upon cooling from high temperature. Textural relationships suggest simultaneous or overlapping growth intervals of garnet and omphacite. This is supported by the fact that the omphacite fractions of Ag85 and Ap21 lie on isochrons defined by the whole rock and the tabletop-digested garnet- and leftover fractions as discussed above. (The Ag31 omphacites have undergone retrograde reactions and do not plot on that sample's isochron.) We conclude that the Lu–Hf ages date the formation of the garnet + omphacite assemblage on Syros. The Lu–Hf ages of the three eclogite samples, when calculated using tabletop-digested garnets (Fig. 8a–c), tightly cluster at 51.9 ± 1.4 Ma (weighted mean of the three ages, 95% confidence limits). Adding bombdigested garnets to the isochrons increases MSWD values as compared to the tabletop-only isochrons. As explained previously, this effect is likely caused by Hfbearing inclusions, components of which may be inherited. We therefore consider the ages calculated from the tabletop-digested garnets to be more robust and are referring these when discussing Lu–Hf ages below. A compilation of ages determined from Syros eclogitefacies rocks in Fig. 9 shows that our garnet growth ages are indistinguishable from the U–Pb–SHRIMP ages of metamorphic zircons dated by Tomaschek et al. (2003).

M. Lagos et al. / Chemical Geology 243 (2007) 16–35

They also agree well with the oldest Ar–Ar and Rb–Sr ages of white mica (Maluski et al., 1987; Baldwin, 1996; Bröcker and Enders, 2001; Putlitz et al., 2005). Dating the HP/LT metamorphism at ∼ 52 Ma with Lu–Hf in garnet has important consequences for the interpretation of these other ages. For example, the Ar–Ar ages of white micas, which range from 52 to 42 Ma (Fig. 9), were interpreted by Putlitz et al. (2005) to document ∼ 10 Myr of mica growth along the prograde segment of the P–T–t path. These authors pointed out that this interval was consistent with the one defined by the crystallization of metamorphic zircon at 52.4 ± 0.8 Ma (Tomaschek et al., 2003) and the 52–47 Ma spread in Lu–Hf ages from our preliminary, bomb-digested garnet results (Lagos et al., 2003). Putlitz et al. (2005) also noted these mica and garnet growth intervals are similar in duration to the ∼ 12 Myr of garnet growth proposed for a UHP rock from Lago di Cignana (Western Alps; Lapen et al., 2003), implying that such a long-lasting prograde segment could have also affected the HP/LT rocks of Syros. In light of the more accurate and precise Lu–Hf ages reported here, however, our preferred interpretation is that 1) garnet growth at Syros occurred only over a short time span, and 2) the formation of metamorphic zircon and the oldest white mica coincide with the establishment of the eclogitic assemblage. 6.2. HP/LT metamorphism on Syros: Single- vs. multiple events Several lines of evidence indicate that the CBU on Syros underwent HP/LT metamorphism in the Eocene. The Lu–Hf garnet ages of eclogites presented here coincide with the maximum ages of white mica from high-pressure assemblages (Maluski et al., 1987; Tomaschek et al., 2003; Putlitz et al., 2005) and U–Pb SHRIMP ages of metamorphic zircon from meta-igneous rocks (Tomaschek et al., 2003). Zircon crystallization itself does not necessarily indicate HP/LT conditions. However, metamorphic zircon that is intergrown with garnet in an eclogite was interpreted, on the basis of grain morphology and the presence of secondary inclusions of xenotime and (Y, HREE, Th)-silicates, to be of the same generation as one from an associated metaplagiogranite that was dated at 52.4 ± 0.8 Ma (Tomaschek et al., 2003). The presence of an older zircon generation in the Grizzas meta-igneous suite, yielding U–Pb SHRIMP ages around 80 Ma (Keay, 1998; Bröcker and Enders, 1999; Tomaschek et al., 2003; Bröcker and Keasling, 2006), points to another event preserved in these rocks. Textural and chemical features indicate that these zircons most likely crystallized from a melt, and therefore they have

31

been interpreted to date the magmatic protolith (see Keay, 1998; Tomaschek et al., 2003 for detailed discussion). However, Bröcker and Enders (1999), who studied metasomatic rocks from the same meta-ophiolite belt, found high-pressure metamorphic minerals, such as rutile, as inclusions in ∼80 Ma old zircons. They interpreted these inclusions to be primary, concluded that the zircons are metamorphic in origin, and argued for a protracted, high-pressure metamorphism, perhaps comprising multiple events, extending from the Upper Cretaceous to the Eocene. More recently, Bröcker and Keasling (2006) have provided additional arguments for an ∼80 Ma HP metamorphic event. These workers presented U–Pb SHRIMP data of zircons from jadeitites (N 90% Jd) and surrounding metasomatic alteration zones in the metaophiolitic mélange near Kampos. Citing the work of Harlow and Sorensen (2005), they argued on the basis of bulk rock composition and texture that the jadeitites could not have formed directly by isochemical transformation of magmatic or sedimentary protoliths during metamorphism. Instead, they suggested that the jadeitites and their ∼ 80 Ma zircons precipitated from aqueous fluids during subduction. However, it is not entirely clear whether these zircons differ chemically and texturally from the 80 Ma magmatic zircons found in the metaplagiogranites and metagabbros. Our new Lu–Hf data, combined with Lu concentration profiles, can offer further insights into the number and duration of HP/LT events that affected the Syros eclogites. As mentioned earlier, the garnets in this study show two different Lu distribution patterns. Garnet Ag85 apparently has the majority of its Lu located near the rim, whereas garnets from the other two samples have more evenly distributed Lu, with a slightly higher concentration at the core. The distribution of radiogenic Hf follows that of Lu, such that the volumes of garnet containing the most Lu will have the biggest influence on the Lu–Hf age (Lapen et al., 2003). For example if most of the Lu resides in garnet cores, the age of the bulk garnet fraction may not reflect the average age of garnet growth but may instead be biased toward the core age. Thus in the case of constant garnet growth (volume / time = constant) the Lu–Hf ages of the Ag31 and Ap21 would show a slight bias toward the core age, whereas that of Ag85 will be biased toward the rim age. That the three eclogites still yield essentially the same Lu–Hf age suggests that the garnets grew over a relatively short time interval, perhaps only a few million years. We found no evidence for a previous (∼80 Ma) HP metamorphic event in these samples. Such a component in the cores of garnet would more strongly influence the

32

M. Lagos et al. / Chemical Geology 243 (2007) 16–35

bulk garnet age of Ag31 than that of Ag85, and would most likely produce a larger age difference than the one we observe. Furthermore, the preserved growth- and trace element zoning profiles of garnets imply that the Lu– Hf system in their cores has not been thermally reset, so it is unlikely that evidence of an 80 Ma component could have been erased in this manner. With the possible exception of Na–Y-rich rims of Ag85, there is also no evidence for recrystallization of garnet via a dissolution–reprecipitation process. There are, for example, no dips or spikes in trace element profiles that point to the presence of a reaction front between original and recrystallized garnet (e.g., Kohn, 2003). The sodian– yttrian rims (and perhaps some interior volumes) of the Ag85 garnet may be the result of a sudden shift in fluid composition during continuous garnet growth or they may represent garnet that reacted during a late fluid flux event. In either case, we interpret the sodian–yttrian garnet to have crystallized within a few Myr of the garnet cores because the bulk Lu–Hf age of this sample is similar to the core-dominated age of sample Ag31, which is from the same sample locality. Thus, a later fluid event–if it did occur–did not substantially lower the Lu–Hf age of Ag85 relative to neighboring rocks. These interpretations obviously depend on how well the HREE profiles of the garnets represent the entire garnet populations of the samples. Ideally, the profile paths would intersect the exact centers of the garnets, so the growth zoning pattern may be observed in its entirety and at its true length scale (Skora et al., 2006). For the LAICPMS analyses, we attempted to sample the garnets near their centers by selecting the largest diameter grains in the thin section whose shapes did not resemble garnet crystal faces parallel to the slide. Even so, it is possible that our profiles did not intersect the garnet nuclei. Skora et al. (2006) observed very narrow, high-Lu zones in the cores (inner ∼25% of garnet radius) of eclogite garnets. We did not observe such sharp Lu peaks in our profiles. If this is because our profiles missed the garnet centers, we may have underestimated the amount of Lu in the cores. (The effect should be minor, however, because of the small volume represented, i.e., ∼ 2%, assuming roughly spherical porphyroblasts.) This would only affect our interpretation for the Ag85 garnet, where we suggest that the Lu–Hf age is biased toward that of the rim. However, none of the several dozen Ag85 grains checked using back-scattered electron imaging appeared to have a Y + HREE enrichment in cores comparable to that seen in the rims. We therefore consider it very unlikely that we have missed a substantial Lu peak in the Ag85 core. Locating the exact geometric centers of garnets can now

be performed with X-ray tomography (Skora et al., 2006). Even so, LA-ICPMS Lu profiles of garnet can probably only serve as a qualitative guide when interpreting ages of multi-grain garnet fractions. This is because the point of garnet nucleation might not be located in the geometric center (asymmetric garnet growth) and individual garnets could have different zoning patterns influenced by their local environments such that even a few well-centered LA-ICPMS Lu profiles might not be representative of the population of garnets dated. 7. Conclusions The Lu–Hf ages of three high-pressure metamorphic rocks from Syros directly date the formation of the eclogitic assemblage (Grt + Omp) at 51.9 ± 1.4 Ma. The garnet ages overlap with the U–Pb ages of metamorphic zircon and the maximum white mica ages. We therefore regard this 52 Ma age as the best approximation for the high-pressure overprint on Syros. There is no indication of multiple- or protracted HP/LT events in the Lu–Hf age data. Thus, the garnet growth in these eclogites most likely records a single, HP/LT event, which occurred in the Eocene. Acknowledgments We would like to thank Heidi Baier for technical support in the laboratory and Cora WohlgemuthUeberwasser for assistance during LA-ICPMS measurements. Markus Lagos, Frank Tomaschek, and Mark Keiter acknowledge generous financial support by the Deutsche Forschungsgemeinschaft through DFG grants Ba 964/10-2 and Ba 964/7 to Chris Ballhaus. We thank Christine Putnis for sharing her ideas on distinguishing continuous garnet growth- vs. fluid reaction textures. We also extend thanks to two anonymous reviewers whose constructive comments helped to improve the manuscript. References Albarède, F., Telouk, P., Blichert-Toft, J., Boyet, M., Agranier, A., Nelson, B., 2004. Precise and accurate isotopic measurements using multiple-collector ICPMS. Geochimica et Cosmochimica Acta 68, 2725–2744. Altherr, R., Schliestedt, M., Okrusch, M., Seidel, E., Kreuzer, H., Harre, W., Wendt, I., Wagner, G.A., 1979. Geochronology of highpressure 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 of the Zermatt–Saas ophiolite deduced from high-precision Sm–Nd and Rb–Sr geochronology. Earth and Planetary Science Letters 171, 425–438.

M. Lagos et al. / Chemical Geology 243 (2007) 16–35 Baldwin, S.L., 1996. Contrasting P–T–t histories for blueschists from the western Baja terrane and the Aegean: effects of synsubduction exhumation and backarc extension. Geophysical Monograph, American Geophysical Union 96, 135–141. Barfod, G.H, Otero, O., Albarède, F., 2003. Phosphate Lu–Hf geochronology. Chemical Geology 200, 241–253. Barr, H., 1990. Preliminary fluid inclusion studies in a high-grade blueschist terrain, Syros, Greece. Mineralogical Magazine 54, 159–168. Bizzarro, M., Baker, J.A., Haack, H., Ulfbeck, D., Rosing, M., 2003. Early history of earth's crust–mantle system inferred from hafnium isotopes in chondrites. Nature 421, 931–933. Blake, M.C., Bonneau, M., Geyssant, J., Kienast, J., Lepvrier, J., Maluski, H., Papanikolaou, D., 1981. A geologic reconaissance of the Cycladic blueschist belt, Greece. Geological Society of America Bulletin 92, 247–254. Blichert-Toft, J., Frei, R., 2001. Complex Sm–Nd and Lu–Hf isotope systematics in metamorphic garnets from the Isua supracrustal belt, West Greenland. Geochimica et Cosmochimica Acta 65, 3177–3187. Blichert-Toft, J., Chauvel, C., Albarède, F., 1997. Separation of Hf and Lu for high-precision isotope analysis of rock samples by magnetic sector-multiple collector ICP-MS. Contributions to Mineralogy and Petrology 127, 248–260. Blichert-Toft, J., Boyet, M., Télouk, P., Albarède, F., 2002. 147Sm–143Nd and 176Lu–176Hf in eucrites and the differentiation of the HED parent body. Earth and Planetary Science Letters 204, 167–181. Bröcker, M., Enders, M., 1999. U–Pb zircon geochronology of unusual eclogite-facies rocks from Syros and Tinos (Cyclades, Greece). Geological Magazine 136, 111–118. Bröcker, M., Enders, M., 2001. Unusual bulk-rock compositions in eclogite-facies rocks from Syros and Tinos (Cyclades, Greece): implications for U–Pb zircon geochronology. Chemical Geology 175, 581–603. 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-Eocene subduction. Journal of Metamorphic Geology 24, 615–631. Bröcker, M., Kreuzer, H., Matthews, A., Okrusch, M., 1993. 40Ar/39Ar and oxygen isotope studies of polymetamorphism from Tinos Island, Cycladic blueschist belt, Greece. Journal of Metamorphic Geology 11, 223–240. Carswell, D.A., Wilson, R.N., Zhai, M., 2000. Metamorphic evolution, mineral chemistry and thermobarometry of schists and orthogneisses hosting ultra-high pressure eclogites in the Dabieshan of central China. Lithos 52, 121–155. Choi, S.H., Mukasa, S.B., Andronikov, A.V., Osanai, Y., Harley, S.L., Kelly, N.M., 2006. Lu–Hf systematics of the ultra-high temperature Napier Metamorphic Complex in Antarctica: evidence for the early Archean differentiation of Earth's mantle. Earth and Planetary Science Letters 246, 305–316. Connelly, J.N., 2006. Improved dissolution and chemical separation methods for Lu–Hf garnet chronometry. Geochemistry, Geophysics, Geosystems 7, Q04005. doi:10.1029/2005GC001082. De Bièvre, P., Taylor, P.D.P., 1993. Table of the isotopic compositions of the elements. International Journal of Mass Spectrometry and Ion Processes 123, 149–166. Dixon, J.E., 1976. Glaucophane schists of Syros, Greece. Bulletin de la Société Géologique de France, Série 7 (18), 280. Dixon, J.E., Ridley, J., 1987. Syros. In: Helgeson, H.C. (Ed.), Chemical transport in metasomatic processes. NATO Advanced Study Institutes Series C, vol. 218, pp. 489–501.

33

Duchêne, S., Blichert-Toft, J., Luais, B., Télouk, P., Lardeaux, J.M., Albarède, F., 1997. The Lu–Hf dating of garnets and the ages of the Alpine high-pressure metamorphism. Nature 387, 586–589. Dürr, S., 1986. Das Attisch-Kykladische Kristallin. In: Jacobshagen, V. (Ed.), Geologie von Griechenland. Beiträge zur regionalen Geologie der Erde, vol. 19, pp. 116–149. Enami, M., Cong, B., Yoshida, T., Kawabe, I., 1995. A mechanism for Na incorporation in garnet: an example from garnet in orthogneiss from the Su-Lu terrane, eastern China. American Mineralogist 80, 475–482. Flynn, R.T., Burnham, C.W., 1978. An experimental determination of rare earth partition coefficients between a chloride containing vapor phase and silicate melts. Geochimica et Cosmochimica Acta 42, 685–701. Frei, D., Liebscher, A., Franz, G., Dulski, P., 2004. Trace element geochemistry of epidote minerals. Reviews in Mineralogy and Geochemistry 56, 553–605. Geisler, T., Schaltegger, U., Tomaschek, F., 2007. Re-equilibration of zircon in aqueous fluids and melts. In: Harley, S.L., Kelly, M. (Eds.), Zircon: Tiny but Timely. Elements, 3, pp. 43–50. Giaramita, M.J., Sorensen, S.S., 1994. Primary fluids in lowtemperature eclogites: evidence from two subduction complexes (Dominican Republic, and California, USA). Contributions to Mineralogy and Petrology 117, 279–292. Green, T.H., Blundy, J.D., Adam, J., Yaxley, G.M., 2000. SIMS determination of trace element partition coefficients between garnet, clinopyroxene and hydrous basaltic liquids at 2–7.5 GPa and 1080–1200 °C. Lithos 53, 165–187. Harlow, G.E., Sorensen, S.S., 2005. Jade (nephrite and jadeitite) and serpentinite: metasomatic connections. International Geology Review 47, 113–146. Hecht, J., 1984. Geological Map of Greece 1:50000, Syros Island. Institute of Geology and Mineral Exploration (IGME), Athens. Hollister, L.S., 1966. Garnet zoning: an interpretation based on the Rayleigh fractionation model. Science 154, 1647–1651. Jacobshagen, V., 1986. In: Jacobshagen, V. (Ed.), Geologie von Griechenland. Beiträge zur regionalen Geologie der Erde, vol. 19, pp. 1–10. Jolivet, L., Rimmelé, G., Oberhänsli, R., Goffé, B., Candan, O., 2004. Correlation of synorogenic tectonic and metamorphic events in the Cyclades, the Lycian nappes and the Menderes massif. Geodynamic implications. Bulletin de la Société Géologique de France 175, 217–238. Keay, S., 1998. The geological evolution of the Cyclades, Greece: constraints from SHRIMP U–Pb geochronology. Ph.D. thesis, Australian National University, Canberra, pp. 335. Keiter, M., Piepjohn, K., Ballhaus, C., Lagos, M., Bode, M., 2004. Structural development of high-pressure metamorphic rocks on Syros island (Cyclades, Greece). Journal of Structural Geology 26, 1433–1445. Keiter, M., Tomaschek, F., Ballhaus, C., 2005. Lithostratigraphic and structural news on the history of Syros (Cyclades, Greece). International Symposium on the Geodynamics of Eastern Mediterranean: Active Tectonics of the Aegean Region 15–18 June 2005. Kadir Has University, Istanbul, TURKEY, p. 43. Kohn, M.J., 2003. Geochemical zoning in metamorphic minerals. In: Rudnick, R.L. (Ed.), The Crust. Treatise on Geochemistry, vol. 3, pp. 229–261. Kötz, J., 1989. Zur Geochemie der Metabasite von Syros, Kykladen, Griechenand. Ph.D. thesis, Universität Köln, pp. 182. Kretz, R., 1983. Symbols for rock-forming minerals. American Mineralogist 68, 277–279.

34

M. Lagos et al. / Chemical Geology 243 (2007) 16–35

Lagos, M., Münker, C., Tomaschek, F., Ballhaus, C., 2002. Geochemistry and Lu−Hf−geochronology of the metavolcanic Grizzas sequence in northern Syros (Cyclades, Greece). Berichte der Deutschen Mineralogischen Gesellschaft, Beihefte zum European Jounal of Mineralogy 14 (1), 97. Lagos, M., Münker, C., Tomaschek, F., Ballhaus, C., Scherer, E.E., 2003. The age of oceanic crust and of HP/LT-metamorphism on Syros (Cyclades, Greece) based on Lu–Hf geochronology and geochemistry. Geophysical Research Abstracts 5, 12851. 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 ultra-high pressure terrane: an example from Lago di Cignana, western Alps, Italy. Earth and Planetary Science Letters 215, 57–72. Lapen, T.J., Mahlen, N.J., Johnson, C.M., Beard, B.L., 2004. High precision Lu and Hf isotope analyses of both spiked and unspiked samples: a new approach. Geochemistry, Geophysics, Geosystems 5, Q01010. doi:10.1029/2003GC000582. Lips, A.L.W., White, S.H., Wijbrans, J.R., 1998. 40Ar/39Ar laserprobe direct dating of dicrete deformational events: a continuous record of early Alpine tectonics in the Pelagonian Zone, NW Aegean area, Greece. Tectonophysics 298, 133–153. Ludwig, K.R., 1999. Isoplot/Ex 2.06 a geochronological toolkit for Microsoft Excel. Berkeley Geochronology Center Special Publication 1a, 1–49. Maluski, H., Bonneau, M., Kienast, J.-R., 1987. Dating the metamorphic events in the Cycladic area: 39Ar/40Ar data from metamorphic rocks of the island of Syros (Greece). Bulletin de la Société Géologique de France 8, 833–842. Maréchal, C.N., Télouk, P., Albarède, F., 1999. Precise analysis of copper and zinc isotopic compositions by plasma-source mass spectrometry. Chemical Geology 156, 251–273. Marschall, H.R., Ludwig, T., Altherr, R., Kalt, A., Tonarini, S., 2006. Syros metasomatic tourmaline: evidence for very high-δ11B fluids in subduction zones. Journal of Petrology 47, 1915–1942. Martin, L., Duchêne, S., Deloule, E., Vanderhaeghe, O., 2006. The isotopic composition of zircon and garnet: a record of the metamorphic history of Naxos, Greece. Lithos 87, 174–192. McClelland, W.C., Vervoort, J.D., Oldow, J.S., Watkinson, A.J., Sha, G.S., 2005. Grenville-age metamorphism on the western margin of Laurentia, northern Idaho: evidence from Lu–Hf garnet geochronology. Geochimica et Cosmochimica Acta 69, A305. Münker, C., Weyer, S., Scherer, E.E., Mezger, K., 2001. Separation of high field strength elements (Nb, Ta, Zr, Hf) and Lu from rock samples for MC-ICPMS measurements. Geochemistry, Geophysics, Geosystems 2, 10.1029. doi:10.1029/2001GC000183. Nebel-Jacobsen, Y., Scherer, E.E., Münker, C., Mezger, K., 2005. Separation of U, Pb, Lu, and Hf from single zircons for combined U–Pb dating and Hf isotope measurements by TIMS and MCICPMS. Chemical Geology 220, 105–120. Okay, A.I., 2001. Stratigraphic and metamorphic inversions in the central Menderes massif: a new structural model. International Journal of Earth Sciences 89, 709–727. Okrusch, M., Bröcker, M., 1990. Eclogite facies rocks in the Cycladic blueschist belt, Greece: a review. European Journal of Mineralogy 2, 451–478. Otamendi, J.E., de la Rosa, J.D., Douce, A.E.P., Castro, A., 2002. Rayleigh fractionation of heavy rare earths and yttrium during metamorphic garnet growth. Geology 30, 159–162. Philippot, P., Selverstone, J., 1991. Trace-element-rich brines in eclogitic veins: implications for fluid composition and transport during subduction. Contributions to Mineralogy and Petrology 106, 417–430.

Philippot, P., Agrinier, P., Scambelluri, M., 1998. Chlorine cycling during subduction of altered oceanic crust. Earth and Planetary Science Letters 161, 33–44. Philippot, P., Blichert-Toft, J., Perchuk, A., Costa, S., Gerasimov, V., 2001. Lu–Hf and Ar–Ar chronometry supports extreme rate of subduction zone metamorphism deduced from geospeedometry. Tectonophysics 342, 23–38. Putlitz, B., Matthews, A., Valley, J.W., 2000. Oxygen and hydrogen isotope study of high-pressure metagabbros and metabasalts (Cyclades, Greece): implications for the subduction of oceanic crust. Contributions to Mineralogy and Petrology 138, 114–126. Putlitz, B., Cosca, M.A., Schumacher, J.C., 2005. Prograde mica 40 Ar/39Ar growth ages recorded in high pressure rocks (Syros, Cyclades, Greece). Chemical Geology 214, 79–98. Ridley, J., 1982. Tectonic style, strain history, and fabric development in a blueschist terrain, Syros, Greece. Ph.D. thesis, University of Edinburgh, pp. 283. Ridley, J., 1984. Evidence of a temperature-dependent ‘blueschist’ to ‘eclogite’ transformation in high-pressure metamorphism of metabasic rocks. Journal of Petrology 25, 852–870. Scambelluri, M., Pennacchioni, G., Philippot, P., 1998. Salt-rich aqueous fluids formed during eclogitization of metabasites in the Alpine continental crust (Austroalpine Mt. Emilius unit, Italian western Alps). Lithos 42, 151–167. Scherer, E.E., Cameron, K.L., Blichert-Toft, J., 2000. Lu–Hf–garnet geochronology: closure temperature relative to the Sm–Nd system and the effects of trace mineral inclusions. Geochimica et Cosmochimica Acta 64, 3413–3432. Scherer, E., Münker, C., Mezger, K., 2001. Calibration of the Lutetium–Hafnium clock. Science 293, 683–686. Scherer, E.E., Mezger, K., Münker, C., 2003. Dating metamorphism with the Lu–Hf garnet method. Geophysical Research Abstracts 5, 12851. Seck, H.A., Kötz, J., Okrusch, M., Seidel, E., Stosch, H.-G., 1996. Geochemistry of a meta-ophiolite suite: an association of metagabbros, eclogites and glaucophanites on the island of Syros, Greece. European Journal of Mineralogy 8, 607–623. Sherlock, S., Kelley, S., 2002. Excess argon evolution in HP–LT rocks: a UVLAMP study of phengite and K-free minerals, NW Turkey. Chemical Geology 182, 619–636. 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–Hf and Sm–Nd geochronology. Contributions to Mineralogy and Petrology 152, 703–720. Söderlund, U., Patchett, P.J., Vervoort, J.D., Clark, E.I., 2004. The 176 Lu decay constant determined by Lu–Hf and U–Pb isotope systematics of Precambrian mafic intrusions. Earth and Planetary Science Letters 219, 311–324. Spandler, C., Hermann, J., Rubatto, D., 2004. Exsolution of thortveitite, yttrialite, and xenotime during low-temperature recrystallization of zircon from New Caledonia, and their significance for trace element incorporation in zircon. American Mineralogist 89, 1795–1806. Spear, F.S., 1993. Metamorphic phase equilibria and pressure– temperature–time paths. Monograph, 1. Mineralogical Society of America, pp. 799. Tomaschek, F., Kennedy, A.K., Villa, I.M., Lagos, M., Ballhaus, C., 2003. Zircons from Syros, Cyclades, Greece — recrystallization and mobilization of zircon during high-pressure metamorphism. Journal of Petrology 44, 1977–2002. Trotet, F., Vidal, O., Jolivet, L., 2001. Exhumation of Syros and Sifnos metamorphic rocks (Cyclades, Greece). New constraints on the P–T paths. European Journal of Mineralogy 13, 901–920.

M. Lagos et al. / Chemical Geology 243 (2007) 16–35 Van Orman, J.A., Grove, T.L., Shimizu, N., Layne, G.D., 2002. Rare earth element diffusion in a natural pyrope single crystal at 2.8 Gpa. Contributions to Mineralogy and Petrology 142, 416–424. van Westrenen, W., Blundy, J.D., Wood, B.J., 2000. Effect of Fe2+ on garnet–melt trace element partitioning: experiments in FCMAS and quantification of crystal-chemical controls in natural systems. Lithos 53, 189–201. Vervoort, J.D., Patchett, P.J., Söderlund, U., Baker, M., 2004. Isotopic composition of Yb and the determination of Lu concentrations and

35

Lu/Hf ratios by isotope dilution using MC-ICPMS. Geochemistry, Geophysics, Geosystems 5, Q11002. doi:10.1029/2004GC000721. Wijbrans, J.R., Schliestedt, M., York, D., 1990. Single grain argon laser probe dating of phengites from the blueschist to greenschist transition on Sifnos (Cyclades, Greece). Contributions to Mineralogy and Petrology 104, 582–593. Wood, S.A., 1990. The aqueous geochemistry of the rare-earth elements and yttrium. 2. Theoretical predictions of speciation in hydrothermal solutions to 350 °C at saturation water vapor pressure. Chemical Geology 88, 99–125.