The terrane stack outcropping on Ios involves thin tectonic slices separated by island-scale deformation structures such as detachment faults and/ or ductile shear zones (e.g., Forster and Lister, 2009; Forster et al., 2020). All terranes experienced Early Oligocene stretching (Fig. 2) and were variably affected by the south-directed South Cyclades Shear Zone (SCSZ, D2 in other publications) (e.g., Forster and Lister 1999a, b, 2009; Forster et al., 2020; Huet et al., 2009, 2011; Ring et al., 2007). An Eocene–Oligocene high-pressure terrane (the Cycladic blueschist unit, upper plate of the Ios Detachment) overlies the Gondwanan Ios basement terrane. There are three tectonic slices in the basement, each recording slightly different metamorphic histories (refer to maps in Supplement). The top-most Port Beach tectonic slice contains two lithologies (garnet–mica schist above augengneiss) and lies immediately beneath the Ios detachment (Forster and Lister 1999a). Beneath this tectonic slice is the ∼500 m thick garnet–mica schist unit (of variable thickness across the island) at mid-structural level and the structurally deepest augengneiss core of the Ios metamorphic core complex (e.g., Andriessen et al., 1987; Baldwin and Lister, 1998; Forster and Lister, 2009; Vandenberg and Lister, 1996).

Figure 1 shows the areal extent of the Asteroussia terrane based on Be'eri-Shlevin et al. (2009), with the shaded area indicating the revised extent according to what we report in this paper. Importantly, this study recognises the exhumed Asteroussia terrane north of Crete, in the Ios basement terrane across the terrane stack, which enabled the identification of a subduction jump that is impossible without a tectonic mode switch. If correct, this is a significant modification implying a widespread metamorphic event in northern Tethys during Late Cretaceous time, including most of the Cycladic islands as reported in this paper and across the Mediterranean region (Altherr et al., 1994; Be'eri-Shlevin et al., 2009; Langosch et al., 2000). Moreover, if the extent of the basement terrane is as large as indicated in Fig. 1, its accretion to the modern terrane stack in the latest Eocene time implies a southward jump of the subduction zone megathrust exceeding 250–300 km (Figs. 2 and 3). Potentially these terranes were autochthonous, with their final accretion involving a period of flat slab subduction followed by the initiation of a new subduction zone (Fig. 3c–d). Subsequent rollback can then stretch the Cycladic crust, explaining the variation in crustal thickness from ∼32 km thick beneath the Cyclades, to 18 km beneath the Sea of Crete, and ∼30 km beneath Crete (based on Makris and Vees, 1976 and Makris et al., 2001). Moreover, the subduction jump is able to explain the formation of core complexes during extreme extension caused by rollback after accretion, and the later Oligocene–Miocene magmatic event that is observed across the Cyclades.

The Port Beach tectonic slice just beneath the Ios detachment represents the structurally shallowest level of the Ios basement terrane and is made up of two lithological units: a thin slice of structurally above garnet–mica schist and the underplating augengneiss with quartz porphyroclasts. Both units preserved numerous recumbently folded veins, isoclinal folds and boudinage structures overprinted by the south-directed SCSZ. A section of altered, greenschist facies garnet–mica schist with chloritoid replacing garnets was observed in the garnet–mica schist tectonic slice near the tectonic contact between Ios upper and lower plate. Beneath the altered zone, garnet porphyroblasts in the garnet–mica schist overgrew pervasive white mica fabrics and were rotated to form and δ-type clasts during SCSZ operation (Fig. 5a, cf. Passchier and Simpson, 1986).

Two generations of white mica were observed in this unit, including pre-mylonite porphyroblasts (now present as muscovite fish with dynamically recrystallised rims) and the younger, recrystallised phengite (separated into wm2 and wm3 based on their overprinting relations) that intergrew with dynamically recrystallised K-feldspar and quartz (Fig. 5b). Silicate content of the phengite deformation fabric is ∼3.40–3.45 Si a.p.f.u. (atoms per formula unit) (Fig. 5c). This, along with the mineral assemblage of quartz ± garnet ± potassium feldspar ± hornblende ± white mica (phengite) ± biotite, suggests P–T conditions of 1.8–2.2 GPa and 500–600∘C based on calculations in Massonne and Schreyer (1987), Patrick (1995), Velde (1967) and Kamzolkin et al. (2016). The small garnet blasts are preserved in low-strain zones, particularly adjacent to pull-aparts marked by quartz-filled voids (Fig. 5b). Amphibolite facies may have taken place in Hercynian time, but during the Late Cretaceous it appears that the Port Beach tectonic slice was subjected to high-pressure eclogite facies conditions.

Field observations in this garnet–mica schist slice and in the top part of the underlying augengneiss core identified evidence for multiple alternating and overprinting deformation events such as recumbent folds overprinted by extensional shear zones. The effects of the (here) N–S striking SCSZ fabric is pervasive, and most of the early fabrics recrystallised during this extensional episode. Nevertheless, relicts of earlier fabric are observed in low-strain zones beside large porphyroclasts. Samples collected from the structurally deeper level of this tectonic slice in the north Mylopotas headland preserved the most complex mineral growth and micro-deformation history. We note that metabasite was observed sporadically in the augengneiss basement, with mineral assemblages that suggest it was subject to transitional greenschist–blueschist metamorphism during the Δ1D event of Forster et al. (2020). The protolith for such metabasite pockets is likely to have been intermediate-mafic intrusive dykes that folded and deformed with the country rock (see structural maps in the Supplement). A late-developed, intense north-sense shear zone defines the structural contact between the garnet–mica schist and the augengneiss in this locality. Fluid associated haematite nodes are found in the top 3 m of an intense shear zone at the contact between the juxtaposed garnet–mica schist tectonic slices and the augengneiss core.

The four samples presented in Table 2 have garnet porphyroclasts recording multiple mineral growth events and preserved earlier fabrics as inclusions. Samples IO17-03 and IO17-04 retained the earliest formed garnets (some of which are large, exceeding 2–3 cm in diameter). Samples IO17-03, IO17-04 and IO17-05 preserved different stages of micro-tectonic events during SCSZ operation. The larger, first-generation (1–2 cm diameter) garnet porphyroblasts are intact in IO17-03, fragmented during shear zone operation in IO17-04 and acting as porphyroclasts during deformation in IO17-05. Relicts of earlier fabrics are preserved in the low-strain zone behind garnets in IO17-03 and IO17-04 and are microstructurally distinguishable, whereas fabrics in sample IO17-05 are almost completely reset by the SCSZ. IO18-05 is collected from a fold hinge of a recumbent fold (M1 folding) that is overprinted by the SCSZ (D2 crustal stretching), and it represents the structurally lowest level of this tectonic slice. Haematite nucleating on the deformation fabric is observed, and relicts of earlier deformation fabrics with rutile are preserved as inclusions in the garnet porphyroblasts.

Mineral chemistries with the three generations of garnet growth recognised in sample IO17-03 are chemically similar to almandine (see Supplement – electron microprobe analysis (EPMA)) but are iron enriched and calcium depleted with slightly higher magnesium content compared to end-member almandine. Dynamically recrystallised, south-sense white mica fabric wraps around the larger (2–3 cm diameter) garnet porphyroblasts, with second-generation garnets growing over this fabric. The two younger garnet growth events are close in time, producing crystals of different size: the 2–3 mm diameter crystals with uniform colour and the 5–8 mm diameter garnets have a zoned mineral growth (Fig. 6a and b). Observation on the two types of crystals identify the light red (Ca-depleted, Mn-enriched) garnets as the first growth event, followed by the second growth event producing black (Mn-depleted, Ca-enriched) garnets (Fig. 6c; see Supplement). Although the later greenschist facies is pervasive across the outcrop, with chlorite overprinting the south-directed white mica deformation fabric, traces of rutile crystals “floating” in the relicts of earlier (pre-SCSZ) fabric in the low-strain zone are preserved in IO17-03 (Fig. 6d). This also implies a higher-pressure history (potentially eclogite facies) than previously recognised.

The earliest microstructure observed was within garnet porphyroblasts, rotated during shear zone operation. This could be inferred from the oblique angle between white mica–rutile inclusions (in IO18-05) and the recrystallised groundmass (Fig. 7a, showing the core of a garnet porphyroblast). As the inclusions were fine grained, energy-dispersive X-ray spectroscopy (EDS) analysis was used to confirm the presence of rutile in the included fabric. Electron microprobe analysis (EPMA) identified the chemical composition the garnet porphyroblasts (5–8 mm diameter) as between almandine and grossular, nucleated on Al-rich white mica during operation of an early shear zone, and continuing to grow during deformation until they reached Al-depleted, foam-textured quartz in pressure shadows (Supplement – EPMA analysis results). Upon reaching the foam-textured quartz, the fluids then corroded grain boundaries, allowing the new-grown garnet to develop a skeletal structure in which the original foam texture in the incorporated quartz grains can still be recognised (Fig. 7b, bottom right corner). Late-stage (first-order) grey albite grew in exsolution trails preserved across the garnet porphyroblasts, implying decompression as the garnet porphyroclast fractured during shear zone operation. (Fig. 7b, Supplement – EPMA analysis results).

Dynamic recrystallisation of white mica (phengite) and quartz in the groundmass of IO17-03 and IO18-05 occurred synchronously during shear zone operation. The Si content of phengite in samples IO17-03 and IO18-05 suggests that the phengite grew under P–T conditions up to 450–500∘C (based on the presence of garnet and biotite) with pressure in the range 0.7–1.7 GPa based on calculations in Massonne and Schreyer (1987), Patrick (1995), Velde (1967) and Kamzolkin et al. (2016) (Fig. 7c). We therefore suggest that the structurally mid-level garnet–mica schist tectonic slice also recorded a complex history of deformation and metamorphism with evidence of high-pressure transitional amphibolite–eclogite facies metamorphism.

The structurally deepest augengneiss core of the Ios basement terrane is characterised by large (0.3–1.0 cm) K-feldspar xenocrysts preserved as porphyroclasts. Rocks in this locality were deformed by the south-directed South Cyclades Shear Zone (SCSZ) then variably overprinted by narrow north-directed shear zones. Occasionally, single hornblende porphyroclasts wrapped by a south-sense white-mica shear fabric could be observed (e.g., Fig. 8a, in a thin-section cut parallel to the stretching lineation). Pre-deformation hornblende was also observed in low-strain zones adjacent to these K-feldspar porphyroclasts (Fig. 8b). K-feldspar porphyroclasts surrounded by dynamically recrystallised white mica and quartz in these sample were fractured by shearing, with recrystallisation at the edges (Fig. 8b). The youngest microstructures observed in these samples are quartz filled cracks. The augengneiss basement records evidence of recumbent folding during crustal shortening, followed by ductile stretching under a south-directed shear zone. All of this occurred before the augengneiss was juxtaposed against the garnet–mica schist slice by an intense north-directed shear zone.

Microstructurally, our study has conclusively identified the presence of more retentive phengite in a fabric that was later overprinted by dynamically recrystallised white mica and quartz. The earlier metamorphic fabrics formed under conditions that potentially reached eclogite facies. Our UHV 39Ar diffusion experiments show that this phengite is highly retentive, allowing preservation of the growth ages of the white mica that formed during these earlier events. Thus, despite intense overprinting during Alpine deformation events, the Late Cretaceous argon populations were retained. This is consistent with the concept of an Argon Partial Retention Zone in which mineral grains undergo some partial resetting by diffusion, but where recrystallisation causes the most effects (Baldwin and Lister, 1998). However, the concept of a partial retention zone is appropriate only for systems with single diffusion domains and no variation of activation energy, which is not the case here.

Identification of the Late Cretaceous age in the Ios lower basement has been interpreted as a result of mixing (e.g., Andriessen et al., 1987): in other words, defining these dates as “intermediate” ages due to excess radiogenic argon or simultaneous degassing of the Alpine mica and the older Hercynian micas. However, here we have shown that these ages represent a period of Late Cretaceous deformation and metamorphism. Therefore, the Ios basement may indeed be part of the Asteroussia terrane. However, pressure–temperature estimates from phengites in the Ios lower plate record high-pressure conditions, contrary to what has been observed in Asteroussia klippen across the Cyclades, albeit preserved at different structural levels. This suggests that more than one set of tectonic slices may have preserved the Asteroussian ages, and we have already pointed to the role that tectonic shuffling may play in producing such variation. It is important in this aspect that the Ios data are the first report of Asteroussia ages in a terrane of unmistakeably Gondwanan affinity (Keay and Lister, 2002).

There may be an earlier Hercynian history: the earliest reported argon age in Ios is a single K/Ar hornblende date reported to be post-Hercynian (268±27 Ma) by Andriessen et al. (1987) and Flansburg et al. (2019). However, based on the peak metamorphic P–T conditions documented across the Cyclades, it may be that the Asteroussian terrane slices record a variety of metamorphic pressure conditions (Table 3, and references therein). Rocks from the basement slices on Ios suggest the occurrence of high-pressure–medium-temperature conditions based on the microchemistry preserved in relicts of earlier deformation fabrics.

The nature of the tectonic processes that affected the evolution of the terranes accreted by the Hellenic subduction zone remains controversial, e.g., comparing the papers by Forster and Lister (2009), Forster et al. (2020) to that written by Huet et al. (2009) and (2011). However, the polemic seems misguided. The architecture of Tethyan orogenic belts, the Hellenides included, invariably involves a nappe stack or a terrane stack, and all terrane stacks are created by thrusting. However, most if not all terrane stacks are also modified by later episodes of extension (e.g., as in Forster and Lister, 2009) leading to tectonic shuffling. It is no different in the Cyclades. The Cycladic archipelago preserves the results of the destruction of an extensive terrane stack that extended from the Hellenides in Greece to the Taurus Mountains in Turkey (Gautier and Brun, 1994a, b; Kempler and Garfunkel, 1994; McKenzie, 1977; Taymaz et al., 1991). The debate as to the nature of exhumation processes will not be resolved by a sole focus on the Cycladic eclogite–blueschist belt, as demonstrated in this paper.

The key questions surround the evolution of the terrane stack overall, rather than the details of the exhumation of an individual tectonic slice. The extrusion wedge (or forcible eduction) model suggests constant compression, resulting in the squeezing of softer material, so that it is extruded to the surface (Forster and Lister, 2008; Xypolias and Koukouvelas, 2001). The competing hypothesis, known as the tectonic mode switch or tectonic shuffle zone model, considers that thrust slices are exhumed by periods of crustal extension that take place in between episodes of crustal shortening caused by individual accretion events (Forster and Lister, 2009). Dispute arises because of the focus on the exhumation of the Cycladic eclogite–blueschist terranes, whereas the continuing nature of the orogenic process means that (without question) the subduction megathrust had to have episodically leapt southward every time a new terrane was accreted (e.g., Lister et al., 2001; Ring et al., 2007; Huet et al., 2009; Forster and Lister, 2009). As the African plate migrated northward, terranes were first subducted, then sliced from the subducting lithosphere by the advancing subduction megathrust and thus accreted to the terrane stack (e.g., Lister et al., 2001; Lister and Forster, 2009).

For Ios, the question is how rollback of the subducting slab was able to throw the overriding terrane stack into horizontal extension immediately after the accretion of the Cycladic blueschist onto the Gondwanan basement from which the Asteroussian terranes were derived (Fig. 4), in particular given the requirement thereafter of a massive southwards leap of the outcrop of the active subduction megathrust. Previous work (e.g., Forster et al., 2020) has suggested that the Cycladic blueschist belt had already been largely exhumed before it was thrust over the Ios basement terrane in Late Eocene time (from ∼38 Ma, Fig. 4a). A first period of extensional tectonism formed the Ios metamorphic core complex, and this had commenced by ∼35 Ma, accelerating by the time of the Eocene–Oligocene transition. A second period of extensional tectonism then ensued, after the Oligocene–Miocene transition, with extreme lithospheric extension triggering a major magmatic event, with intrusions in and through the core of younger metamorphic core complexes across the Cyclades.

The Ios basement has been argued to be autochthonous, moving with Africa, and part of Gondwana (Flansburg et al., 2019; Keay et al., 2001; Keay and Lister, 2002). Its accretion to the terrane stack is therefore likely to have been an event with considerable tectonic significance. The magnitude of the southward leap of the subduction megathrust is thus unlikely to have been accomplished without the development of a new lithosphere-scale structure. There are two end-member options: one requiring that the slab peels free from the subduction megathrust (Fig. 3, using the slab peel hypothesis discussed by Brun and Faccenna, 2008) while the other requires a subduction jump and slab break-off (Fig. 4, cf. von Blanckenburg and Davies, 1995). Although the slab-peel model is consistent with enhanced heat flow during crustal stretching after the accretion event, such a model requires the asthenosphere to be exhumed to such shallow levels as to require significant partial melting of the uplifting asthenosphere, which would a period of widespread basaltic volcanism, with volumes comparable to those observed in some large igneous provinces. Such effects were not observed in the Cyclades. Sizova et al. (2019) also showed the “peel-off” model (Brun and Faccenna, 2008) to be unlikely in the Aegean region.

An alternative model involving slab necking (or boudinage) and break-off must therefore be considered (e.g., Fig. 4). This (provisional) three-staged “slab break-off” model more accurately describes Aegean tectonics by addressing how the terrane stack was subjected to overall stretching with some evidence of melting such as plutonic intrusions in the centre of metamorphic core complexes. This model also requires significant magmatism, but in consequence of fluids rising from a devolatising slab which would lead first to crustal magmatism, such as the I-type granite of Ios, and later to the appearance of arc volcanoes, as on Thera. The necking and eventual break-off of the subducting slab and formation of a new subduction zone (Fig. 4d) might be of sufficiently small scale to escape observation in models based on P-wave tomography.

We do not understand why the Asteroussia event is recorded in the top-most slices of the terrane stack outcropped in other Cycladic islands but is found only in the lower slice in Ios. Such architecture implies that the Cycladic eclogite–blueschist tectonics slices are “sandwiched” between tectonic slices affected by the Asteroussia event, whereas on Crete the Asteroussia units are juxtaposed above the Vatos unit, the Arvi unit, the Pindos unit and the Tripolitza unit (e.g., Bonneau, 1984; Flansburg et al., 2019; Kneuker et al., 2015; Langosch et al., 2000; Martha et al., 2017; Martha et al., 2016; Palamakumbura et al., 2013; Seidel et al., 1976; Zulauf et al., 2002). This must have occurred sometime between mid-Oligocene–early Miocene time. Laterally, the unit is connected to the eastern Alps in the west and the Lycian ophiolite nappes, the Menderes Massif and the Sakarya Zone in Turkey (van Hinsbergen et al., 2020). Further work is required to validate the tectonic-shuffling hypothesis which is capable of explaining these observations.

Some authors suggest that tectonic slices outcropping on islands in the northwest (Andros, Tinos, Syros) are different to those on other islands such as Anafi, Nikoria, Donoussa, Ikaria and Crete (Altherr et al., 1994; Langosch et al., 2000; Martha et al., 2016). Arguments arise due to the difference in dates obtained (despite all being Late Cretaceous) and different results for geothermobarometry across islands with different lithologies and metamorphic facies (Kolodner et al., 1998; Langosch et al., 2000; Patzak et al., 1994; Seidel et al., 1976, 1981; Yeung, 2019). Research in the upper and middle tectonic units in Tinos produced dates at 90–100 Ma and a peak metamorphic P–T estimate of 120 MPa at 450–500 ∘C (Avigad and Garfunkel, 1989; Avigad and Garfunkel, 1991; Bröcker and Franz, 1998; Patzak et al., 1994), whereas studies on Donoussa and Crete produced younger ages at 70–80 Ma and peak metamorphic P–T conditions at 300–600 MPa and 600–730∘C (Be'eri-Shlevin et al., 2009; Keay and Lister, 2002; Kolodner et al., 1998; Langosch et al., 2000; Seidel et al., 1976).

Our study reports metamorphic conditions with higher pressure, despite producing similar dates. Although the presence of phengite is widespread across the Ios lower plate, the highest pressures are inferred only in the garnet–mica schist unit and the Port Beach tectonic slice. With no evidence of higher pressures in the underlying augengneiss unit, it is possible that a more complex deformation and metamorphic history has been recorded in these intermediate slices in the Ios terrane stack. These observations also reflect on a possible distinction between European and Gondwanan terranes, with evidence mostly preserved in the Alps and in the Pelagonian zone of Greece (Pourteau et al., 2013; Porkoláb et al., 2019; Regis et al., 2014; Thöni, 2006). Brown et al. (2014) reports evidence of Late Cretaceous intracontinental shear zone deformation across Africa, thus demonstrating that the ∼70–80 Ma age is not limited to northern Tethys. Detrital zircon (DZ) analysis on pre-plutonic metasedimentary rocks in the Ios lower plate by Flansburg et al. (2019) pushes tectono-magmatic histories of the southern Cyclades further in time to the early Cenozoic. They noted a striking resemblance between their DZ age spectra from Ios lower plate to exposures on Crete, northern and central Peloponnese, the northern Hellenides, and the siliciclastic cover sequence of the Menderes massif in western Turkey (Flansburg et al., 2019). Comparing these Ios DZ age spectra to those from northeast Africa and Arabia, they confirmed that the Cycladic basement terrane (outcropped in the Ios lower plate) has a distinct peri-Gondwanan affinity (Flansburg et al., 2019). This led them to propose a tectonic model where the terrane was located along the northern margin of Gondwana in the early Paleozoic and experienced pluton emplacement between ∼335 and ∼305 Ma in an arc setting (Flansburg et al., 2019).

Tectonic reconstructions by van Hinsbergen et al. (2020) demonstrated that major continental-scale events occurred across Eurasia and Gondwana from Late Jurassic–Late Cretaceous time. These global tectonic events involve continental-scale deformation such as the formation of the Alpine Tethys with microcontinents tearing from the south coast of Europe. In their reconstruction model, the Ios basement, along with other tectonic units in the Cycladic islands and Crete, are all part of a subducted Greater Adria continental ribbon. While our island-scale study cannot contribute to the discussion on whether Greater Adria was a single continental landmass or made up of several large islands, it is evident that Late Cretaceous deformation is widespread in both the European and Gondwana terranes.

Distinguishing European versus Gondwanan terranes in the Central Aegean and greater Mediterranean area will remain difficult. One central argument is the number of oceans present in the “greater Tethys seaway” between Europe, Africa and potentially Adria at Mesozoic and associated tectonic evolution (e.g., Channell and Kozur, 1997; Kilias et al., 2010; Robertson et al., 2013). This argument mainly concerns the paleogeography of the Pelagonian unit outcropped in mainland Greece. It is thus of interest that evidence for Late Cretaceous ages is reported from white mica deformation fabrics isolated from the northern end of the upper Pelagonian unit (e.g., Kilias et al., 2010; Robertson et al., 2013). The Pelagonian unit may have been a continental ribbon (or micro-continent) separating two Tethyan realms: the Vardar Ocean in the northeast and the Pindos (or even Cyclades) Ocean in the southwest (e.g., Channell and Kozur, 1997; Robertson et al., 2013). Such models imply high-pressure metamorphism in the Pelagonian unit as the result of the attempted subduction of the continental ribbon (Robertson et al., 2013). Other reconstructions consider the Pelagonian unit as the eastern-most unit of a continental Adria terrane, adjacent to a single northeastern oceanic basin (the Vardar Ocean) (e.g., Bortolotti et al., 2013; Ferriere et al., 2012; Kilias et al., 2010; Palamakumbura et al., 2013). These researchers disagree with the concept of a distinct Pindos Ocean both in Triassic and in Jurassic time (Kilias et al., 2010).