Introduction

The Cyclades in the Aegean Sea, Greece, represent a part of the Alpine-Himalayan orogenic collage and are a key area for understanding Gondwana breakup and collision between Gondwana-derived fragments and the Eurasian plate (e.g. Keay and Lister 2002). It has been generally agreed that continental rifting of northeastern Gondwana in northern Africa has been initiated during Late Carboniferous–Permian times and continued into early to mid-Triassic times, eventually leading to the creation of Neotethyan ocean basin(s) (e.g. Pe-Piper 1998 and references therein). The continental rifting resulted from extension subsequent to Hercynian subduction of the Palaeotethys ocean at an Andean-type margin (e.g. Robertson et al. 1991). However, the Triassic evolution of the western Mediterranean region is poorly understood but a significant precursor to the better documented Cenozoic history. Triassic volcanic rocks are widespread within the Hellenides and include a predominant subalkaline basalt–andesite–dacite series, and minor shoshonites and calc-alkaline rocks, alkaline basalts and mid-ocean ridge basalt (MORB)-like varieties (e.g. Pe-Piper 1998). Triassic felsic plutonic rocks are also common in different tectonic zones of the Hellenides. Such rocks intruded Permo-Carboniferous basement and record various degrees of metamorphic overprinting (e.g. Reischmann 1998; Himmerkus et al. 2009; Chatzaras et al. 2013 and references therein; Fig. 1).

Fig. 1
figure 1

Simplified geological maps of the Cycladic archipelago and the studied islands including sample locations (modified after Bulle et al. 2010): a, b regional overview maps of the Aegean region and the Cyclades; c Andros; d Syros; e Sifnos; and f Ios

Full size image

The geodynamic significance of the regionally widespread Triassic igneous and meta-igneous rocks has been a contentious issue and has either been related to rift-settings or to short-lived contemporaneous subduction processes, in association with the creation of one, two, or more Neotethyan oceanic basins (e.g. Koutsovitis et al. 2012; Bortolotti et al. 2013 and references therein). Several tectonic models concerning Triassic and subsequent magmatism have been proposed and are summarised in Robertson (2012) and Chatzaras et al. (2013), but as yet no generally accepted interpretation has been established.

The focus of this study is on Triassic and other Mesozoic igneous rocks and builds on previously and newly SIMS-dated (secondary ion mass spectrometry or ion microprobe) samples from the islands of Andros, Ios, Sifnos, and Syros (Bröcker and Keasling 2006; Bröcker and Pidgeon 2007). In contrast to the well-documented Tertiary granitoids (e.g. Bolhar et al. 2010, 2012) and the Cretaceous meta-igneous rocks (Fu et al. 2012) of the larger region, no study has addressed as yet the zircon O–Hf isotope characteristics of the metamorphosed pre-Cretaceous magmatic rocks. This is surprising because O–Hf isotope data have a considerable potential to unravel petrogenetic processes. For example, Hf isotopic ratios of zircons in crustal rocks can be used to constrain crustal residence ages, or the average time since the parental magmas were extracted, e.g. from depleted mantle, whereas the oxygen isotopic fractionation can give an indication as to the nature of processing of material. Combined zircon O–Hf isotopic systematics can distinguish between juvenile components and parental magmas derived from reworking of pre-existing crustal material and thus can help determine the magma sources (e.g. DePaolo 1981; Hawkesworth and Kemmp 2006). Such data are important because it can contribute to a better understanding of the geodynamic processes operating through geological time in the Aegean region.

Geological background and sample description

In the context of the present study, three tectonic zones of the Hellenides are of special importance: the Vardar Zone, the Pelagonian Zone, and the Attic–Cycladic Crystalline Belt. The study area is part of the Attic–Cycladic Crystalline Belt (Fig. 1) which is built from two major tectonic units recording different pressure–temperature–deformation–time (P–T–D–t) histories (e.g. Dürr et al. 1978; Okrusch and Bröcker 1990; Ring et al. 2010). The structurally higher unit is only preserved in relatively small occurrences and includes a heterogeneous sequence of unmetamorphosed Permian to Mesozoic sediments, ophiolites, and greenschist facies rocks with Cretaceous to Tertiary metamorphic ages (e.g. Bröcker and Franz 1998, 2006), as well as Late Cretaceous granitoids and medium-pressure/high-temperature metamorphic rocks (e.g. Patzak et al. 1994 and references therein). The structurally lower unit (=Cycladic Blueschist Unit) consists of a pre-Alpine crystalline basement and a stack of tectonic subunits that is composed of metamorphosed volcano-sedimentary rocks and mélanges. The Cycladic Blueschist Unit has experienced Tertiary eclogite- to epidote-blueschist facies metamorphism and various degrees of greenschist- to upper-amphibolite facies overprinting (e.g. Okrusch and Bröcker 1990; Ring et al. 2010).

Details of the geology of the Cyclades have been described in numerous contributions and the geological and tectonometamorphic history of individual islands is well documented. Islands with many features relevant for the present study include: Syros (Dixon and Ridley 1987; Trotet et al. 2001a, b; Schumacher et al. 2008; Bulle et al. 2010; Keiter et al. 2011), Sifnos (Schliestedt 1986; Schliestedt and Matthews 1987; Avigad et al. 1992; Avigad 1993; Trotet et al. 2001a, b; Schmädicke and Will 2003; Groppo et al. 2009), Andros (Papanikolaou 1978; Mukhin 1996; Bröcker and Franz 2006; Huyskens and Bröcker 2014), and Ios (Henjes-Kunst and Kreuzer 1982; van der Maar and Jansen 1983; Baldwin and Lister 1998; Forster and Lister 1999; Huet et al. 2009; Thomson et al. 2009).

The analysed zircon populations are mostly from previously dated samples of the Cycladic Blueschist Unit that yielded Cretaceous, Jurassic, or Triassic protolith ages (Bröcker and Keasling 2006; Bröcker and Pidgeon 2007) and include rocks representing both the marble–schist sequences (Andros, Sifnos, Ios) and blocks from high-pressure/low-temperature mélanges (Syros, Andros). Cretaceous and Jurassic ages have only been documented for mélange blocks, whereas Triassic ages were reported for tectonic slabs in mélanges and for layers within schist successions. Sample-specific information such as rock type, location details, and U–Pb zircon ages is summarised in Table 1. For details of the local geology and mineral assemblages see Bröcker and Keasling (2006), Bröcker and Pidgeon (2007) and Bulle et al. (2010). Most zircon ages for the ten meta-igneous samples are interpreted to reflect the timing of magmatism at ca. 240 Ma (7 samples), ca. 160 Ma (2), and ca. 80 Ma (1); only two of them (samples 5008 and 4036) contain inherited zircon cores (ca. 1,414–273 Ma) (Bröcker and Keasling 2006; Bröcker and Pidgeon 2007).

Table 1 Samples from Andros, Ios, Sifnos, and Syros, Greece
Full size table

This sample suite is complemented by the gneissic rocks 5039 and 5041 (Electronic supplementary materials or ESM S1) from Ios, which occur as coherent layers within schist sequences. Both rocks mainly consist of quartz, plagioclase, epidote/clinozoisite, chlorite, white mica, and sphene. Relict blue amphibole is preserved in sample 5039. It is often very difficult to decipher the parent rock type of meta-felsic rocks and this is also true for these quartz and feldspar-rich samples. Original textural features of the protoliths have been completely erased by metamorphic and deformational overprinting. Bulk geochemical data are not available and likely would not be useful for identification of the protolith in any case.

Analytical methods

The available zircon mounts (MB2, MB5, MB6 and MB7: Bröcker and Keasling 2006; Bröcker and Pidgeon 2007) from SHRIMP (Sensitive High-Resolution Ion Micro-Probe) U–Pb analysis contained relatively small pieces of standard zircon CZ3 that were supplemented by one or two large grains of face-mounted standard zircon Plešovice for SHRIMP 18O/16O analysis. The pits from previous ion-probe age dating were polished out. Mount BF032 was newly prepared for samples 5039 and 5041 from Ios. For this purpose, hand-picked zircons and grains of the standards (Plešovice and Temora-2) were cast in epoxy, ground, and polished to expose the midsection of the grains, and cathodoluminescence (CL) imaging was conducted using a JEOL JSM-6610A Analytical SEM (scanning electron microscope) in Research School of Earth Sciences (RSES), The Australian National University (ANU), Canberra, equipped with a Robinson CL detector. Representative SEM-CL images of zircons from some of the studied samples are shown in Fig. 2.

Fig. 2
figure 2

Cathodoluminescence images and U–Pb ages and O–Hf isotopic ratios for representative zircon grains from meta-igneous rocks from Andros, Ios, Sifnos, and Syros, Greece: a grain no. 5 and b grain no. 1, gneissic rock sample 5039; c grain no. 26 and d grain no. 2, gneissic sample 5041, from Ios; e grain no. 19, meta-tuffaceous gneiss sample 4036, from Sifnos; f grain no. 21, felsic gneiss sample 2041; g grain no. 30, felsic gneiss sample 5008; h grain no. 11, meta-gabbro sample 5017, from Andros; and i grain no. 35, meta-plagiogranitic sample 4017, from Syros. Numbers represent SIMS U–Pb ages (±2σ; dash ellipse) and δ18O values (also ellipse) and LA-ICP-MS initial epsilon Hf or εHf(t) values (italic, underlined; circle). Data sources: ages for e to i, from Bröcker and Keasling (2006) and Bröcker and Pidgeon (2007); others, this study. Scale bars 50 µm

Full size image

The instruments employed include SHRIMP-RG (Reverse Geometry) for U–Pb geochronology, SHRIMP II and SHRIMP SI (Stable Isotope) for O isotopes, and laser ablation—multi-collector—inductively coupled plasma mass spectrometry (LA–MC-ICP-MS) for Hf isotopes, all at RSES, ANU. The analytical procedures and data processing are described in detail in ESM S1 and summarised below. U–Th–Pb isotopic analyses of zircons were carried out on the SHRIMP-RG ion microprobe. Analytical procedures are similar to those described in Williams (1998 and references therein) and Ireland and Williams (2003). Data were reduced using the SQUID Excel Macro of Ludwig (2001a) and the ISOPLOT/EX Excel Macro of Ludwig (2001b). Oxygen isotope ratios (18O/16O) of zircons were determined using SHRIMP II and SHRIMP SI ion microprobes. Analytical conditions were similar to those outlined in detail by Ickert et al. (2008). In situ Lu–Hf isotope analyses for zircons were carried out using the 193 nm excimer laser-based HELEX ablation system coupled to a Neptune MC-ICP-MS described in Eggins et al. (2005). Data were reduced offline using the software package Iolite (Paton et al. 2011). The microanalytical results are tabulated in ESM S2 to S4.

Results

Zircon U–Pb geochronology

In addition to the previously dated samples in Table 1, U–Pb geochronology was carried out on samples 5039 and 5041 from Ios, which most likely represent meta-tuffaceous protoliths. Zircon of both samples is mostly characterised by short, sharp, pyramidal terminations in morphology and oscillatory zonation as revealed by CL imaging (Fig. 2a–d). Rounded surfaces typical for detrital zircons are not seen; distinct core–rim structures are rarely developed. A total of 33 SIMS spot analyses of 32 grains from sample 5039 yielded a wide range in ages, from 567 to 204 Ma (Fig. 3a). The majority of grains contribute to a single peak in the age distribution at around 240 Ma (inset of Fig. 3a). The weighted mean 206Pb/238U age calculated from 258 to 204 Ma data points shows excess scatter with main contribution from the two youngest grains. Eliminating these two analyses gives a weighted mean age of 243.3 ± 3.3 Ma (95 % confidence level (c.l.), MSWD = 2.4, n = 25). The zircons in this group have U and Th concentrations of 106–1,017 and 31–861 ppm, respectively, and Th/U ranges between 0.29 and 0.97. Five older zircons (331–293 Ma) and two younger zircons or domains (weakly zoned rim: 210 ± 9 Ma, 2σ; bright, homogeneous core: 204 ± 9 Ma) have Th and U concentrations of 169–578 and 50–340 ppm, respectively, and Th/U ratios (0.18–0.61). Th and U concentrations in the oldest zircon (567 Ma) are 16 and 187 ppm, respectively, with Th/U ratio of 0.1, possibly of metamorphic origin.

Fig. 3
figure 3

Concordia diagrams for zircons of gneiss samples 5039 and 5041 from Ios. Insets are probability density diagrams. Data-point error ellipses denote 2σ errors

Full size image

Fifty analyses of 39 zircon grains from sample 5041 yielded ages between 620 and 204 Ma although most zircons are around 320 Ma. The weighted mean 206Pb/238U age calculated by trimming tails for a major group of age varying from 334 to 299 Ma, as seen in a probability density diagram (inset of Fig. 3b), is 318.6 ± 2.7 Ma (95 % c.l., MSWD = 1.5, n = 39). The zircons in this group have U and Th concentrations of 72–2,029 and 22–344 ppm; Th/U ratios vary between 0.12 and 0.59. A second age group of zircons or distinct domains within individual grains (286–268 Ma) are characterised by a weighted mean 206Pb/238U age = 279 ± 12 Ma (95 % c.l., MSWD = 0.19, n = 7), U = 155–2,075 ppm, Th = 55–275 ppm, Th/U = 0.11–0.61. U and Th concentrations in two older zircons or domains (620, 517 Ma) are: 25 and 166 ppm, 33 and 45 ppm, respectively, and their Th/U ratios are 1.4 and 0.3. Two zircons have much younger zoned rims, 210 ± 8 Ma (2σ) and 204 ± 8 Ma, similar to the above sample, than the bright cores. Their Th/U ratios are identical at 0.2, with U = 2,275 ppm and 1,705 ppm and Th = 486 ppm and 346 ppm, respectively.

Oxygen isotopic compositions

The δ18O and εHf(t) values for all zircon spot analyses plotted against ages are shown in Fig. 4 and averages for individual samples are summarised in Table 2. The zircon O–Hf isotopic data, available for many of the dated zircons with concordant age, are presented in Figs. 2, 4, and 5.

Fig. 4
figure 4

Ages plotted against a ion microprobe δ18O values of zircons from meta-igneous rocks from Andros, Ios, Sifnos, and Syros, Greece; and b LA-ICP-MS initial epsilon Hf values. Dashed horizontal lines in a indicate the compositional range of igneous zircons from the mantle and in equilibrium with primitive magmatic compositions (δ18O = 5.3 ± 0.6 ‰, 2σ; Valley et al. 2005). Published data for zircons from Cretaceous and Cenozoic igneous or meta-igneous rocks in the region are also plotted for comparison: RB, Bolhar et al. (2010, 2012); BF, Fu et al. (2010, 2012)

Full size image
Table 2 Oxygen and hafnium isotope compositions of zircon in selected samples from Andros, Ios, Sifnos, and Syros, Greece, analysed by SHRIMP II and calibrated against Plešovice (8.19 ± 0.08 ‰; J.W. Valley, unpublished data in Fu et al. 2012)
Full size table
Fig. 5
figure 5

δ18O versus εHf(t) values for zircons from meta-igneous (mig) rocks from Andros, Ios, Sifnos, and Syros, Greece: a all ages (for data sources and abbreviations, see Fig. 4); b Jurassic; and c Triassic (others: samples 1828, 2038, 2041, 5000, 5001 and 5034). Data for inherited zircon grains are excluded. See text for endmembers A and B and explanations

Full size image

The δ18O values from the peak 320-Myr-old zircons of sample 5041 (Ios) range from 7.0 to 9.0 ‰. The ca. 240 Ma zircons from sample 5039 indicate comparably high δ18O values, ranging from 6.2 to 8.2 %. It is noted that δ18O values for two Carboniferous zircon grains of this sample (7.4 and 7.8 ‰) fall within the range for sample 5041. The Triassic zircon populations of 6 meta-igneous samples from Andros (2041, 1828 and 2038) Ios (5034), and Syros (5001 and 5000) are characterised by δ18O values varying from 2.7 to 6.0 %; δ18O (zircon) for sample 4036 from Sifnos ranges between 4.1 and 6.5 ‰, excluding one outlier of −0.4 ‰ and inherited cores. δ18O for Jurassic zircons from two meta-igneous samples range from 5.2 to 7.2 ‰ (sample 5008) and from 6.9 ‰ to 10.1 ‰ (sample 5017), respectively. Five inherited zircon grains or zircon cores of Proterozoic to Permian age from samples 5008 and 4036 record a wide range in δ18O (5.6–9.6 ‰).

Cretaceous zircons from meta-plagiogranite 4017 yielded δ18O values of 4.9 to 6.0 ‰. These largely overlap with the ‘mantle-like’ values (δ18O: 4.7–5.5 ‰) reported for 80-Myr-old zircons of other mélange blocks and matrix rocks from Tinos and Syros (Fu et al. 2012).

Hafnium isotopic compositions

Fifteen out of 17 analyses (by LA–MC-ICP-MS) of zircons from sample 5041 (Ios) yielded 176Hf/177Hf ratios of 0.282328–0.282458, corresponding to εHf(t) values of −9.0 to −4.7 (Fig. 4b). Measured 176Lu/177Hf and 176Hf/177Hf ratios of ca. 240 Ma zircons from sample 5039 range from 0.000631 to 0.002460 and from 0.282439 to 0.282574, respectively, corresponding to εHf(t) values of −5.3 to −1.9 (Fig. 4b). One older zircon grain has an εHf(t = 322 Ma) value as low as −5.0, within the range given for the ca. 320 Ma zircons from sample 5041.

The Triassic zircons from meta-igneous samples have εHf(t) values of −2.6 to −0.3 for one sample (4036) and +5.5 to +15.7 for the other 6 samples. εHf(t) for Jurassic zircons from two meta-igneous samples range from −4.9 to −2.5 (sample 5008) and from −12.5 to −9.0 (sample 5017). Seven inherited zircon grains or zircon cores (ca. 273–1,414 Ma) of samples 4036 and 5008 indicate a wide range in εHf(t) (−19.0 to +5.0).

The εHf(t) values for Cretaceous zircons from the meta-plagiogranite 4017 vary from +12.7 to +14.2. The results are close to that reported for other contemporaneous mélange blocks and matrix rocks in the same region (Fu et al. 2012), interpreted to be derived from depleted MORB mantle or similar sources.

Discussion

There are four major zircon U–Pb age groups in the investigated samples: ca. 320 Ma, ca. 240 Ma, ca. 160 Ma, and ca. 80 Ma (Table 1). The group of Cretaceous zircons can clearly be distinguished by their contrasting O–Hf isotopic signatures from the Late Carboniferous, Late Jurassic, and most of the Triassic zircons (Fu et al. 2012; this study).

Zircon isotope characteristics through time

The morphological representation of zircons from the two Ios samples (5039, 5041), that is, sharp, pyramidal terminations with rarely developed core–rim structures, and their U–Pb ages can best be reconciled by assuming a largely igneous protolith (i.e. tuffaceous rock), derived from a proximal magmatic source, with a distinct detrital component. Hafnium depleted mantle model ages (TDM) for zircon of the main Hercynian age group from sample 5041 (Ios) range from 1.4 to 1.0 Ga (ESM S4), consistent with Nd-depleted mantle model ages of ca. 1.3–1.1 Ga for the Thessaly Hercynian plutons on mainland Greece (e.g. Pe-Piper 1998). Together with the presence of inherited zircons of Late Precambrian age in two Triassic (meta-) igneous rocks (samples 5008 and 4036; Bröcker and Pidgeon 2007), this is characteristic of the West African craton (e.g. Keay and Lister 2002; Meinhold et al. 2008). Combined high δ18O values and negative εHf(t) values indicate that the related parental igneous rocks record reworking of continental crust in the context of Variscan (or Hercynian) orogenic processes, as also documented in the Late Carboniferous intrusions exposed in the External Hellenides on the island of Kithira (Xypolias et al. 2006). Magmatism at such a large scale was caused by Late Paleozoic continental collision between Gondwana and Laurasia to form the supercontinent Pangaea (e.g. von Raumer 1998; Murphy et al. 2009).

The studied sample suite includes 8 (out of 12) meta-igneous rocks with Triassic protolith ages. Prominent Triassic age clusters are also a striking feature of detrital zircon populations from the Cyclades (Keay 1998; Löwen et al. 2014; Bröcker et al. 2014a) further emphasizing the importance of magmatic activity at that time for the wider region. Petrogenetic processes involved in forming igneous rocks, such as assimilation and fractional crystallization, can be examined through a plot of δ18O versus εHf(t) (DePaolo 1981; Hawkesworth and Kemmp 2006). Triassic igneous zircons yielded both positive and negative εHf(t) values and variable δ18O values (Figs. 4, 5). The slightly lower-than-depleted mantle εHf(t) values for zircons of some Triassic samples (2041, 1828, and 2038 from Andros, 5034 from Ios, 5001 and 5000 from Syros) may be explained by contributions of juvenile components such as recycled lower crust of an ancient island arc and/or a slab component (oceanic crust or sediments) in the mantle wedge (e.g. Nowell et al. 1998; Woodhead et al. 2001). In some modern arcs, a deviation from depleted mantle in εHf(t) as well as δ18O can be related to source contamination processes in the mantle induced by subducted oceanic crust and overlying terrigenous sediments, distinguishable from assimilation of continent crust into arc magmas (e.g. Roberts et al. 2013). A melt source close to MORB-type composition is inferred for island arc magmatism, whereas subduction-related magmatism is characterised by lower εHf(t) values than MORB. Here, the mantle-like oxygen isotopic signature for the above described Triassic igneous zircon samples suggests that the high εHf(t) values are related to a less depleted melt source than MORB-type mantle, e.g. enriched mantle (EM) and high-μ (HIMU) mantle reservoirs (Pfänder et al. 2007; Geldmacher et al. 2011), with little involvement of older crustal material. The EM components may have higher δ18O values than the HIMU mantle components, due to contamination of subducted sediments with the basaltic portion in recycled oceanic crust, that is, MORB-type basalts (Eiler 2001).

In contrast, the negative εHf(t) values for the zircon sample 4036 from Sifnos can be interpreted to reflect different magma sources such as existing continental crust. Inherited zircons in Triassic meta-igneous rocks indicate a composite nature of the zircon populations recording different phases of melting and crystallisation since the Precambrian. These inherited zircons are attributed to either a Precambrian continental crust or clastic sediments subducted below an ancient volcanic arc. The O–Hf isotopic ratios suggest variations in melt composition somewhere between a crustal end member and a depleted (MORB-type) mantle source and this favours a continental margin setting during Triassic time. It is noted that both positive and negative εHf(t) values reported for zircons from the Trans-Himalayan batholiths (i.e. Kohistan–Ladakh–Gangdese granites) on the southern margin of the Lhasa Terrane, southern Tibet, are attributed to island arc sources (Ji et al. 2009 and references therein). Therefore, we propose that those Triassic felsic rocks from the Cyclades characterised by high and positive εHf(t) values and mantle-like (and/or slightly lower than mantle) δ18O values formed from metasomatised mantle (cf. Perkins et al. 2006), whereas those Triassic felsic rocks characterised by negative εHf(t) values and high δ18O values formed either by magma mixing or from lower continental crust.

Partial melting of subducted oceanic crust, which would result in the formation of adakite, is unlikely because there are no Triassic adakites or adakitic rocks reported in the region (e.g. Pe-Piper and Piper 2002). It is noted that, in some cases, upper oceanic lithosphere may have δ18O values of >7 % (Bindeman et al. 2005; Martin et al. 2011), distinct from typical peridotitic mantle (and melts thereof) that exhibits a narrow range in δ18O values of 5.5 ± 0.4 ‰ (Eiler 2001). The estimated Hf two-stage model ages (T 2DM ) are mostly Mesoproterozoic, for zircons from two samples (5039 and 4036), or Neoproterozoic, for other Triassic samples (ESM S4). This further suggests that Triassic magmas in the region may be derived from different source regions.

It is intriguing that Jurassic zircons from the meta-gabbro 5017 have higher δ18O values but lower εHf(t) values than those determined for felsic gneiss 5008 from the same high-pressure mélange on Andros (Figs. 4, 5). These rocks are believed to represent SSZ-type (supra-subduction zone) rather than MOR-type (mid-ocean ridge) ophiolites, linked to the Vardar Ocean or a different coeval oceanic basin (Bröcker and Pidgeon 2007). The Jurassic ophiolites formed either within the Pindos (-Mirdita) oceanic basin between the Apulian (or Adria) and Pelagonian continental unit or within the Vardar ocean further east (see review in Robertson 2012). However, the high δ18O values of the meta-gabbro zircons may be explained by assimilating high δ18O older continental crust components into mafic magmas possibly derived from the mantle wedge above subducting oceanic lithosphere.

There are two similar cases reported elsewhere. For instance, high-δ18O(Zrn) values (7.0 to 7.8 ‰) of Cretaceous gabbroic rocks from southern Sierra Nevada Batholith, California, USA, were interpreted to be subduction-related (Lackey et al. 2005). It has been suggested that these mafic magmas may be derived from normal sublithospheric mantle of continental arc type contained in the mantle wedge above subducting oceanic lithosphere, with assimilation of either subducted, hydrothermal altered oceanic crust (Lackey et al. 2005) or continental crust (Nelson et al. 2013). Another example is the Zhengga diorite-gabbro suite from the Gangdese area, southern Tibet, where gabbros have δ18O(Zrn) values of 5.9 to 7.2 ‰ (Ma et al. 2013). These authors conclude that the parental magmas of the Zhengga gabbros were generated by the hydrous partial melting of lithospheric mantle metasomatised by sediment melts/fluids and are therefore also subduction-related. The δ18O signatures of the meta-gabbro and felsic gneiss from Andros are in accordance with an interpretation suggesting that these rocks are not part of the typical Jurassic ophiolite successions of the Balkan Peninsula but instead formed by partial melting of a metasomatised mantle wedge with significant assimilation of supracrustal material, either hydrothermally altered oceanic crust subducted beneath the volcanic arc or continental crust material. It is noted that the Andros samples have higher radiogenic Nd–Sr isotopic compositions than contemporaneous meta-ophiolites in the region (Bröcker et al. 2014b). However, a more comprehensive dataset would be necessary to robustly constrain this conclusion.

The interpretation for the Late Cretaceous zircons of meta-plagiogranite 4017 is straightforward (Figs. 4, 5). These zircons have εHf(t) values typical for MORB-type mantle, similar to results reported for other Cretaceous zircon populations from Syros and Tinos (Fu et al. 2010, 2012), excluding the group of low-δ18O zircons with cauliflower-like internal structures and/or porous or dark-CL weakly zoned rims (Fig. 2 of Fu et al. 2012). The average δ18O(Zrn) of 5.5 ± 0.8 ‰ of the pristine grains is also consistent with other mantle-like components (Valley et al. 2005; Cavosie et al. 2009; Grimes et al. 2010).

Such large variations in both δ18O and εHf(t), especially for older rocks, are not only attributed to complex magmatic processes, but also diverse magma sources. Simple mass balance calculations can be used to assess magma mixing of two endmembers in the region, that is, deeply derived and supracrustal sources (A and B in Fig. 5). The following parameters were used (Fig. 5; see DePaolo 1981): r = (mass of assimilated wallrock into a magma/unit time)/(mass of fractionated crystalline phases effectively separated from the magma/unit time) = 0.6, D = the bulk solid/liquid partition coefficient for the element between the fractionating crystalline phases and the magma = 1, thus z = (r + D − 1)/(r − 1) = −1.5. Two endmembers are assumed: Endmember A, δ18O = 5.3 ‰, εHf = −2; Endmember B, δ18O = 12 ‰, εHf = −16; and HfA/HfB = 0.25. We estimate about 20–30 % by mass supracrustal material was incorporated into mafic magma derived from enriched mantle or lower continental crust during the Late Jurassic. It appears that more and more supracrustal material may be involved in episodic magmatism in the Cyclades over time. Both highest δ18O values and lowest-εHf(t) values for Triassic to Cenozoic igneous zircons from the Cyclades define an evolution trend consistent with average continental crust, with the exception of Cretaceous zircons formed from depleted MORB mantle sources (Fig. 4).

Tectonic implications

A plausible geodynamic model must account for the following geochemical and isotopic signatures. Combined Pb and Nd isotopic data indicate that the widespread subalkaline basalts of the Hellenides were predominantly derived from previously melt-depleted peridotite within subcontinental lithospheric mantle on the northern margin of Gondwanaland (e.g. Pe-Piper 1998). Unusually high contents of LILE (large ion lithophile elements) in shoshonite and calc-alkaline rocks suggest an additional ‘crustal’ component that might be associated with a pre-Triassic, possibly Hercynian subduction in heterogenous subcontinental lithosphereic mantle and lower crust (e.g. Pe-Piper 1998; Tsikouras et al. 2008). Triassic rift-related mafic volcanic rocks, abundant in the Hellenides to the west, show enriched mantle and within-plate (rift-related) chemical signatures (E-MORB and OIB: enriched MORB and ocean island basalt) (Bonev et al. 2012).

Taking these aspects into consideration and based on Pe-Piper (1998) and Robertson (2012), we suggest the following scenario for the magmatic evolution of the larger study area (Fig. 6). At ~320 Ma, Paleotethys oceanic crust subducted southwards to the northern margin of Gondwana (e.g. Xypolias et al. 2006), which caused partial melting of lower continental crust. At ~240 Ma rift-related magmas were produced from ‘enriched’ mantle that had been metasomatised by the preceding Hercynian slab melts and/or fluids, and continental crust. The Triassic melts record the separation of the Apulia and (Korabi–) Pelagonia microcontinents from Gondwana. A rift basin further developed and the Pindos (–Mirdita) ocean was formed. At ~160 Ma, Neotethys oceanic lithosphere subducted northwards below Pelagonia and produced Sierra-type gabbros (cf. Lackey et al. 2005). This stage is represented by ~160 Ma meta-gabbros with distinct O–Hf isotopic characteristics. Group B hornblende diorite (hornblende K–Ar age: 154 ± 14 Ma, 2σ) of the calc-alkaline series found as clasts in flysch at Amphissa and characterised by enrichment in REE (rare earch elements) and LILE (Pe-Piper and Koukouvelas 1992) might be another example of Jurassic gabbroic rocks of the same type. At ~80 Ma, MORB formed at spreading centres of small Neotethys ocean basins. The remnant Pindos oceanic crust subducted and closed progressively southwards during Cretaceous and Tertiary times. During the Late Cenozoic, orogeny I and S-type granites were formed (Bolhar et al. 2010, 2012).

Fig. 6
figure 6

Tectonic model proposed for the larger Aegean region, modified after Pe-Piper (1998) and Robertson (2012). See text for discussion

Full size image

Conclusions

Zircon U–Pb, O, and Hf isotopic data of meta-igneous rocks representing mélanges and marble–schist sequences from the islands of Andros, Ios, Sifnos, and Syros were used to define the crustal evolution of the Attic–Cycladic Crystalline Belt. The following conclusions can be reached:

  1. 1.

    Middle Triassic igneous zircons characterised by variable δ18O and εHf(t) values could have formed from different sources, that is, metasomatised mantle and lower continental crust. The results of this study are in accordance with the interpretation that Triassic magmatism in the Hellenides is not subduction- but more likely rift-related (e.g. Pe-Piper 1998).

  2. 2.

    Late Jurassic igneous zircons of a meta-gabbro from Andros have higher δ18O but lower εHf(t) values than zircons from an associated felsic gneiss with the same protolith age, arguing against MORB affinity. This supports the conclusion that the Jurassic gabbroic rocks from Andros do not belong to the group of SSZ-type ophiolites that are widespread along the margin of the Pelagonian zone (Pe-Piper and Koukouvelas 1992), but represent a distinct group recording partial melting of metasomatised mantle wedge with significant assimilation of supracrustal material.

  3. 3.

    Late Cretaceous igneous zircons are characterised by MORB-like δ18O and εHf(t) values, suggesting that the Late Cretaceous magmas were produced by partial melting of depleted MORB mantle (see also Fu et al. 2010, 2012).

  4. 4.

    The Mesozoic magmatic rocks of the Attic–Cycladic Crystalline Belt record reworking of older continental crust with input of multiple mantle components or juvenile components over certain periods of time.