Compressional origin of the Naxos metamorphic core complex, Greece: Structure, petrography, and thermobarometry

Metamorphic core complexes (MCCs) are generally believed to form during lithospheric extension (e.g., Lister et al., 1984; Teyssier and Whitney, 2002). However, the pre-extensional histories of many such examples worldwide are poorly understood, including those on the Cycladic islands, Greece. MCCs occur in a variety of tectonic settings, ranging from wholly extensional (e.g., Basin and Range Province; Coney, 1980; Armstrong, 1982; Flecher and Hallet, 1983; Wernicke, 1985; Yin, 1991) to purely compressional (e.g., North Himalayan gneiss domes; e.g., Burg et al., 1984; Lee et al., 2004, 2006) regimes, making it unlikely that they share a common mechanism of formation (Platt et al., 2014). Although the exhumation of deep-crustal metamorphic rocks by footwall unroofing beneath an extensional detachment fault is now largely accepted (Wernicke, 1981; Axen, 2007), there remains a limited understanding of the relationship between the timing of metamorphism and the onset of extensional motion (Platt et al., 2014). Furthermore, the processes driving the initiation of detachment faulting, and whether these are associated with lithospheric-scale faulting or localized extension in a compressional environment (e.g., Searle, 2010), are still debated (Fig. 1).

The transition from compression to extension is commonly recorded by the development of characteristic extensional structures and metamorphic assemblages that overprint older collisional features (Platt, 1986; Vanderhaeghe and Teyssier, 2001). Examples include overprinting of suture zones by low-angle extensional detachments, and the replacement of high-pressure–low-temperature metamorphic mineral assemblages by high-temperature–medium-pressure assemblages, often due to an increased basal heat flow associated with crustal thinning, e.g., Betics Rif, Spain (e.g., Platt et al., 2013). Nevertheless, Barrovian metamorphic overprints and MCCs are characteristic features in many active orogens, such as the Lepontine Dome and Tauern Window in the Alps (e.g., Smye et al., 2011, and references therein) and the Greater Himalayan Series (GHS) in the Himalayan Range (e.g., Jessup et al., 2006; Searle et al., 2006). Notably, a thickened crust that formed in response to continental collision would also equilibrate along a hotter geothermal gradient, resulting in widespread regional metamorphism driven largely by the conductive relaxation of perturbed isotherms (Bickle et al., 1975; England and Richardson, 1977; England and Thompson, 1984).

Although normal-sense extensional shear fabrics indeed record localized extension on their associated shear zone, they occur in zones of both overall crustal shortening and extension, and they record the relative uplift and exhumation of material, rather than a specific tectonic stress regime. Extensional microstructures typically form via one of three geodynamic mechanisms: (1) buoyancy-driven extrusion of crustal material transported to depth within a subduction zone, which occurs under a passive-roof normal fault on the interface between the subducting slab and overlying mantle wedge, in conditions responsible for creating extensional fabrics in blueschists and eclogites (e.g., England and Holland, 1979; Hacker et al., 2013); (2) exhumation and extrusion of high-grade migmatites and gneisses from a deep crustal root by coeval movement of a thrust at the base and an extensional synorogenic normal fault at the top (e.g., the top of the GHS beneath the South Tibetan Detachment System (STDS); Law et al., 2006; Searle et al., 2006); or (3) normal faulting during crustal extension and rifting (e.g., Lister et al., 1984; Jolivet et al., 2010; Teyssier and Whitney, 2002; Wernicke, 1985). Mechanisms 1 and 2 occur in compressional tectonic settings, while mechanism 3 occurs in crustal or lithospheric extensional settings. All three processes are reported herein to have occurred on Naxos.

The Cycladic islands in the Aegean Sea (Fig. 2), and the island of Naxos in particular (Fig. 3), expose a range of geological features that document the transition from a compressional to an extensional tectonic regime (Jansen and Schuiling, 1976; Buick and Holland, 1989; Urai et al., 1990). Naxos also exposes the deepest structural levels of the Attic-Cycladic Massif, which experienced high-grade metamorphism, anataxis, and leucogranite formation during the Miocene. This study documents the pre-extensional, prograde, and retrograde history of the Naxos core complex and links the timing of kyanite- and sillimanite-grade Barrovian-type metamorphism to the prevailing tectonic regime. A future study will present the absolute ages of both compressional and extensional microstructures, which were constrained via in situ U-Th-Pb isotope geochronology performed on zircon, monazite, allanite, xenotime, and rutile. Finally, we assessed various tectonic scenarios to explain the history of Naxos prior to extension, and we conclude that the core complex formed due to prolonged compression, crustal thickening, and heating prior to the switch to extension in the late Miocene.

The MCCs of the central Aegean region form part of the Attic-Cycladic Massif (Durr et al., 1978), which is a belt of thinned continental crust located to the north of the Hellenic subduction zone, where the Nubian plate is subducting northward beneath Eurasia (Wortel and Spakman, 2000; Jolivet and Brun, 2010). Within this subduction setting, the region can be subdivided into the forearc (Crete, Karpathos, Rhodes), the Hellenic arc (Santorini, Milos), and the back arc (Cyclades), which is reported to have experienced lithospheric extension in response to retreat of the Hellenic slab (Le Pichon and Anglier, 1979, 1981; Le Pichon et al., 2002; Jolivet et al., 2010, 2013, 2015). Many studies have documented extension affecting the area from the Miocene until present day, and some estimate southward propagation of the Hellenic subduction zone on the order of 1000 km over this time period (John and Howard, 1995; Seward et al., 2009). It is estimated that NE-SW–oriented crustal extension rates within the last 15 m.y. may be up to twice as fast as present-day rates (Urai et al., 1990; Jolivet et al., 2001, 2004). In response to extension, three bivergent, crustal-scale, low-angle normal fault systems developed. The North Cycladic Detachment System (NCDS), exposed on the islands of Andros, Tinos, and Mykonos, displays top-to-the-NE sense of shear (Jolivet et al., 2010). The Naxos-Paros Detachment System (NPDS), exposed on Naxos and Paros, displays top-to-the-NNE shear fabrics (Buick, 1991a, 1991b; Cao et al., 2013, 2017; Urai et al., 1990; Kruckenberg et al., 2010, 2011). The West Cycladic Detachment System occurs on the islands of Kea, Kythnos, and Serifos and shows top-to-the-SSW kinematics (Grasemann et al., 2012).

Despite extensional deformation affecting all structural levels, little is known about the geological evolution of the area prior to the initiation of extensional tectonics. Structural and metamorphic evidence from several islands suggests that the Cyclades experienced a complete cycle of mountain building (Ring and Layer, 2003; Ring et al., 2007a, 2007b; Peillod et al., 2017). Remnant ophiolites crop out at the highest level on several islands, in the hanging walls of the low-angle normal faults (Stouraiti et al., 2017). These are underlain by (M₁) Eocene high-pressure eclogite- to blueschist-facies rocks of the Cycladic Blueschists in the footwalls of the detachments, dated at 52.2–51.4 Ma by U-Pb zircon and Lu-Hf garnet geochronology from Syros (Tomaschek et al., 2003; Lagos et al., 2007) and at ca. 46 Ma from Sifnos eclogites (Dragovic et al., 2012). The area then experienced regional Barrovian metamorphism (M₂) during the Oligocene to early Miocene (40–18 Ma), followed by a sillimanite-grade event (M₃) at ca. 16 Ma that reached anatectic conditions on Naxos and Ios (Jansen and Schuiling, 1976; Buick and Holland, 1989; Keay et al., 2001). Within the high-grade dome of the Naxos core complex, there is evidence for NNE-SSW extensional fabrics coeval with M₃ conditions (Ring et al., 2007b), indicating a switch in tectonic regime during this period. Understanding the timing of this switch from compression to extension is fundamental to understanding the formation of the Aegean MCCs.

Naxos is the largest island in the Cyclades and shows evidence for having experienced multiple tectono-metamorphic events (see Figs. 3 and 4). It is composed of a NNE-SSW–elongated dome in the exhumed footwall of a major low-angle normal fault, the NPDS, which developed in response to regional Aegean extension (Cao et al., 2013, 2017). The hanging wall mainly consists of ophiolitic rocks with a mélange of disrupted Miocene–Pliocene sediments (Jansen, 1973), which are possibly synextensional (Urai et al., 1990; Gautier and Brun, 1994a, 1994b). Upper Pliocene conglomerates containing pebbles of footwall material were deposited on the flanks of the dome during unroofing (Lister et al., 1984; Buick, 1991a, 199b; Gautier and Brun, 1994a, 19914b). The lower unit represents a Barrovian-type metamorphic sequence that developed during M₂ (Jansen, 1973; Jansen and Schuiling, 1976; Buick and Holland, 1989; Buick, 1991a, 1991b; Urai et al., 1990), reaching upper-greenschist facies in the east (biotite zone: ∼5 kbar and ∼400 °C) and overprinting relict (M₁) blueschist assemblages along the southeast coastline (Avigad, 1998; Peillod et al., 2017), and reaching upper-amphibolite (M₃) anatectic conditions in the core (sillimanite zone: ∼7 kbar and ∼700 °C). In the west, an I-type granodiorite pluton with a U-Pb age of ca. 12.2 Ma intrudes the metamorphic sequence and developed a narrow metamorphic aureole (M₄; Keay et al., 2001; Koukouvelas and Kokkalas, 2003; Jansen and Schuiling, 1976).

The core region consists of highly strained M₃ sillimanite-bearing migmatites and gneisses, with intercalated marbles and leucogranites, which are separated from a lower-grade deformed sequence of Mesozoic shelf carbonates by a major ductile shear zone, forming the carapace (the attenuated metamorphic sequence structurally above the migmatite core) to the migmatite dome. Top-to-the-NNE kinematic indicators are associated with this ductile shear zone, which formed during exhumation of the core complex. Due to this complex deformation and metamorphic history, the development of most of the footwall structures and fabrics is poorly understood, and there is confusion as to whether they formed during crustal extension or predate the extensional movements and were reactivated by them (Ring et al., 2007b). Recent work has shown that the blueschist rocks on southeast Naxos experienced a contrasting metamorphic history to the underlying higher-grade rocks (Peillod et al., 2017), and there has been confusion about whether the entire island experienced (M₁) blueschist-facies conditions or whether there are discrete tectono-stratigraphic units. Furthermore, the cause of regional M₂ and M₃ kyanite-sillimanite–grade metamorphism, the pre-extensional, prograde evolution of the core complex, and the relative timing of peak metamorphism in relation to the onset of extension have remained controversial. Two contrasting models have been proposed for the initiation of extension. Wijbrans et al. (1993) and Parra et al. (2002) argued that extension commenced at 30 Ma (see also Jolivet et al., 2003) due to slab roll-back, followed by M₂ heating of the Cycladic blueschists (Bröcker and Franz, 1998). Alternatively, others argued for the concept of synorogenic extension during underthrusting and compression in the Cyclades until 21 Ma, followed by postorogenic extension and crustal thinning (Peillod et al., 2017; Ring and Layer, 2003; Ring et al., 2007a, 2007b, 2010).

Competing models have been proposed to explain the decompression of migmatites and the formation of the doubly plunging migmatite dome making up the core of Naxos. These include diapirism (Jansen and Schuiling, 1976; Vanderhaeghe, 2004), exhumation during regional extension (Gautier and Brun, 1994b), and superimposed folding from E-W shortening followed by top-to-the-NNE shearing (Urai et al., 1990; Buick, 1991a, 1991b), as well as a combination of buoyancy- and isostasy-driven flow during crustal extension (Kruckenberg et al., 2011). This study presents a systematic characterization of the tectonic and metamorphic evolution of Naxos that addresses the role of compressional and extensional tectonics in formation of the core complex.

Geological mapping was carried out (Figs. 3 and 4), building on previous detailed work (e.g., Jansen and Schuiling, 1976; Buick, 1991a, 1991b; Buick and Holland, 1989; Urai et al., 1990; Vanderhaeghe, 2004; Kruckenberg et al., 2011), but with new and important structural interpretations (Figs. 4–10). In this section, we describe each tectono-stratigraphic unit in descending structural order and discuss the major fabrics, folds, and mineral growth (Figs. 4–10) in relation to the deformation.

Retrogressed blueschists exposed in the SE corner of the island (Avigad, 1998) represent the highest structural level of the metamorphic carapace. These are composed of (M₁) blueschist- and (M₂) greenschist-facies rocks derived from proximal to distal shelf and slope metasediments including turbidites, dolomitic marbles, pelites, meta-conglomerates, and metavolcanics. The existence of deeper-sea clays is indicated by piemontite-bearing schists, which together have similarities to the Cycladic Blueschist Unit exposed on Syros, Ios, Andros, and Tinos. Additional exposures of blueschist-facies assemblages can be found toward the top of the sequence, spanning the entire eastern coastline (see Figs. 3 and 4). Within this unit, there are two types of top-to-the-NNE extensional shear fabrics. S₁ is defined by blueschist-facies assemblages including glaucophane, phengitic mica, and paragonite that are preserved within only partially retrogressed high-pressure boudins, bearing many similarities to the fabric seen throughout the Cycladic Blueschist Unit. This extensional fabric is associated with synorogenic extrusion from a NE-dipping subduction zone during the Eocene by synorogenic passive-roof normal faults such as the Vari detachment on Syros and Sifnos (e.g., Ring et al., 2003; Roche et al., 2016; Laurent et al., 2016).

Pelitic horizons have highly strained contacts with dolomitic marbles that preserve internal folding and centimeter-scale crenulation cleavage (S_2a). These are overprinted and truncated by a subparallel, highly localized extensional S-C′ shearing fabric (S₃), exclusively defined by greenschist-facies assemblages. The intensity of shearing increases toward the boundaries of the unit, where major normal-sense shear zones, the Moutsana detachment and Zas shear zone (ZSZ) (Fig. 7), have developed. Along these shear zones, some discontinuous horizons of serpentinite enclaves crop out (Katzir et al., 1999, 2002, 2007), which are also affected by the (S₃) extensional foliation. These are interpreted as fragments of serpentinized mantle from the overlying ophiolitic thrust sheet that where incorporated at low temperature during extensional shearing (S₃).

The Koronos Unit is a 4-km-thick metasedimentary sequence of intercalated marble and schists that represents a proximal Mesozoic shelf carbonate cover to the Cycladic basement (Durr et al., 1978; Andriessen et al., 1979). This unit is characterized by M₂ amphibolite-facies Barrovian metamorphic rocks, in contrast to the overlying Zas Unit. In coarse-grained dolomitic marbles at the top of the unit, corundum (emery) and magnetite horizons represent paleosols that formed during sedimentation and subaerial exposure of the continental margin during the Jurassic (Altherr et al., 1982). Within the succession, meter-scale amphibolite bodies trend parallel to the regional foliation and represent doleritic sills that intruded along bedding prior to burial and metamorphism. M₂ kyanite gneisses, marbles, and schists crop out at the deepest levels, and the metamorphic grade dramatically decreases up structural section to staurolite and garnet grade. At the base of this unit, slices of continental basement are exposed and are structurally repeated in a sequence overlying semipelitic schists and marbles of the shelf carbonate sequence. This indicates structural repetition of the shelf sequence, which could only be achieved by thick-skinned thrusting of the shelf and basement over more-proximal continental shelf metasediments. Garnet- and clinopyroxene-bearing amphibolites outcrop 100 m below this horizon, within the metasedimentary sequence, as a semicontinuous layer along strike. These enclose partially serpentinized peridotite lenses (Katzir et al., 2002, 2007) that remained stronger during deformation as the amphibolites flowed and deformed around them during top-to-the-NNE ductile shearing (see Fig. 7). The intensity of top-to-the-NNE shearing within the unit increases toward the base, where the Koronos shear zone (KSZ) separates this unit from the underlying migmatites.

The Core Unit is composed of M₂ kyanite-grade to M₃ sillimanite-grade migmatites, gneisses, and leucogranites, which make up the deepest levels of the core complex. The first-order dome structure contains second-order subdomes (Kruckenberg et al., 2011), which are separated by a high-strain zone (Core High-Strain Zone [CHSZ]) of tight upright and isoclinally folded marbles, amphibolites, and migmatites that have been subsequently boudinaged. The Core Unit is separated from the overlying carapace by the KSZ along the dome margins (see below; Figs. 4 and 5). Within the anatectic core, the extent of partial melting varies with bulk composition, as well as with increasing metamorphic grade. There is lithological variation throughout the migmatite zone, with banded leucocratic gneisses representing continental basement commonly interleaved between pelitic and semipelitic protoliths and marbles of the shelf carbonate cover sequence. This structural repetition in lithology again provides further evidence for thick-skinned ductile thrusting and imbrication of the shelf-basement contact during crustal thickening. Stromatic migmatites (Fig. 9), representing the first appearance of melt, outcrop along the perimeter of the dome or adjacent to metasedimentary rafts and are characterized by a continuous foliation marked by alternating quartzofeldspathic and biotite layers with distinct leucosomes and melanosomes (see also Kruckenberg et al., 2011). Leucosomes occur along the foliation marked by biotite, muscovite, and kyanite in aluminum-rich domains, increase in abundance with depth, and become concentrated in fold hinges, where they display magmatic textures. Leucosomes also occur in boudin necks and in shear bands that crosscut compositional layering (S₀). At deeper structural levels, both coarse quartzofeldspathic horizons and pockets of melt are ptygmatically folded, with melt distributed along fold hinges, indicating it accumulated in dilatant structural sites during compression (Brown, 2002; Sawyer, 2001; Holness, 2008). Where evidence of melting becomes locally more extensive, particularly in pelitic horizons, diatexites are characterized by magmatic textures containing enclaves of the host rocks, and they are heterogeneous at outcrop and thin section scales. They include schollen, and schlieren-structured varieties (Fig. 9) with a magmatic foliation defined by schlieren of biotite or alignment of schollen (see also Kruckenberg et al., 2011). Although these features can be locally extensive, our observations and data reveal they are not as widespread as Kruckenberg et al. (2011) implied. We suggest that most of the gneisses previously described as diatexites are instead granitic basement with Variscan metamorphic ages (presented in a future study). Although some diatexites do locally occur, they are confined to fertile lithologies upon crossing the K-feldspar isograd at the deepest levels of the migmatite dome.

The large-scale structure of Naxos consists of a complex superimposition of three generations of kilometer- to meter-scale isoclinal and sheath folds, which are overprinted and truncated by a series of normal-sense shear zones (Jansen, 1973; Buick and Holland, 1989; Urai et al., 1990; Buick, 1991a, 1991b). Three-dimensional models, panoramic sketches, and accompanying interpretations are presented in Figures 4–6, while structural analyses of these folds and shear zones are displayed in Figures 6 and 7.

Isoclinal, recumbent, and noncylindrical (F₁) folds occur mainly within the metamorphic carapace (Zas and Koronos Units), up to kilometer scale, and trend along NNE-SSW fold axes. They are characterized by subhorizontal hinge lines that strike parallel to the fold axes and display a SW vergence direction and hanging-wall anticline geometries above discrete detachments (e.g., on the western flank of Mount Zas; Figs. 5 and 6). The limbs of these folds are refolded by smaller tight to isoclinal, also NNE-SSW–trending (F₂) folds and sometimes produce an overall sheath fold–type geometry with their noses trending NNE-SSW, parallel to the lineation and shear direction, and these are particularly well developed along the ZSZ (Figs. 5, 6, and 8A). Sheath folds form by progressive rotation of fold hinges and are defined by curvilinear fold axes with more than 90° of curvature (Ramsay and Huber, 1987). They initially form as isoclinal folds with axes aligned orthogonal to shear and rotate toward the transport direction under progressive non-coaxial, high-strain deformation (e.g., Cobbold and Quinquis, 1980; Ramsay and Huber, 1987; Alsop and Holdsworth, 2006). Sheath fold formation involves (1) initial buckling of layers, largely controlled by lithological contrasts such as between marble and schist; and (2) amplification and rotation of fold axes as shear strain increases (Alsop and Holdsworth, 2006). It has been shown that sheath folds can form under both constrictional (e.g., Ez, 2000) and “nonconstrictional” regimes (e.g., Cobbold and Quinquis, 1980; Alsop and Holdsworth, 2006). Under constriction, a bull’s-eye sheath fold pattern is predicted with ellipticity of the innermost ring less than that of the outermost ring, whereas under general shear, a cat’s-eye geometry is predicted with innermost ring ellipticity greater than that of the outer ring. During simple shear, innermost and outermost rings have similar ellipticities, producing an analogous eye pattern.

F₂ tight to isoclinal fold closures commonly do not show eye patterns. However, along the ZSZ, cat’s-eye geometries occur, indicative of general shear (a combination of simple shear and pure shear flattening). Along the perimeter of the migmatite dome, these folds strongly rotate into alignment with the strong (S₃) top-to-the-NNE shearing fabric in the KSZ, indicating the structures formed during top-to-the-NNE shearing (Alsop and Holdsworth, 2006). These folds preserve S₂ crenulation cleavages in their limbs with axial planar kyanite, indicating that these structures formed during M₂ conditions. Kyanite is also aligned with the pervasive S₃ NNE-SSW stretching lineation, and therefore we interpret that the folds formed in two stages: (1) F₁ isoclinal folds developed during prograde burial and thickening. (2) This was followed by non-coaxial top-to-the-NNE general shear during exhumation, synchronous with (M₂) Barrovian conditions forming F₂ folds that produce the sheath-type geometries as shown in Figure 6E.

Upright isoclinal folds (F₃; Figs. 4–10) also occur along NNE-SSW axes parallel to the NNE-SSW lineation and have been noted by several workers on Naxos (Urai et al., 1990; Buick, 1991a, 1991b) and other Aegean islands, including Tinos, Andros, Syros, Paros, and Mykonos (e.g., Virgo et al., 2018, and references therein). These folds outcrop across all units despite spatially contrasting in style. Within the migmatitic core, an earlier generation of upright, E-W–trending isoclinal folds is refolded by orthogonal to NNE-SSW–trending upright structures. These structures are defined by vertically dipping marble bands and pelitic horizons that sometimes enclose boudinaged amphibolites, which have experienced polyphase micro- and macroscale folding. Although meter-scale folds in the migmatites do occur, they do not have a consistent axial trend and were probably formed by flow of partial melts (Jansen and Schuiling, 1976). Within the migmatite dome, our observations suggest that F₃ folds are completely discordant to the overlying structures and are truncated by normal faults and the NPDS that bounds the metamorphic footwall. Within migmatite, the leucosomes and kyanite blades are folded with the foliation (Figs. 8, 13B, and 13C) by F₃ folds. This differs from the overlying carapace, where these structures range from open to tight and refold the limbs of F₁ and F₂ folds, but also have their fold axes trending parallel to the lineation (NNE-SSW; Fig. 6H). Therefore, F₃ folds must postdate peak M₂ kyanite-grade conditions and F₁ and F₂ folding, yet they must have formed in the presence of melt and simultaneous to doming as they wrapped around the core (Buick, 1991a, 1991b).

The origin of these folds has been the subject of considerable debate. Some workers have suggested that F₃ folds formed during E-W compression responsible for gentle island-scale postmetamorphic E-W doming of the isograds (Buick and Holland, 1989; Urai et al., 1990; Buick, 19911, 1991b), while other studies inferred their origin to be a result of extensional shearing, as similar features are present in other extensional terrains (Hodges et al., 1987). Rey et al. (2011, 2017) and Kruckenberg et al. (2011) argued that folds in the migmatite dome developed during convergent flow and the viscous collision of two weak channels of the lower crust during regional NE-SW crustal extension. However, this mechanism cannot explain the earlier isoclinal folding, top-to-the-NNE–related sheath folding, and upright F₃ fold geometries within the metamorphic carapace. Alternatively, Buick (1991a, 1991b) proposed F₃ folds within the carapace originated in their current orientation under greenschist-facies conditions, as rotation into alignment with top-to-the-NNE shearing would cause flattening. Because F₃ structures fold ductile mylonite fabrics, have NNE-SSW boudinaged limbs (see Fig. 10), and have upright fold axes that are truncated by the NPDS, we suggest they must have formed due to significant E-W shortening that predated NNE-SSW crustal extension.

Three-dimensional relationships of polyphase folding can be spectacularly observed in amphibolite bands within marble in the CHSZ of the Boulibas and Kinidaros Quarries, where ductile pinching and boudinage of tight upright F₃ antiforms and synforms crop out (Figs. 9 and 10). These amphibolite horizons act as strain markers and show evidence for superimposition of multiple F₃ fold geometries, vertical boudinage (i.e., boudinage with subhorizontal neck lines), horizontal NNE-SSW boudinage (i.e., boudinage with neck lines orientated subvertically), flattening, brittle fracture, and S-C shear fabrics with opposing shear senses on opposite limbs of folds. Boudinage and shear bands with opposing shear senses are orientated in the vertical and NNE-SSW horizontal planes, symmetrically arranged on the limbs of F₃ folds and aligned with their axial surfaces (Fig. 9 and 10). These indicate directions of maximum extension in the vertical and NNE-SSW horizontal planes. F₃ fold interference patterns suggest that E-W–trending upright isoclinal folds are refolded by tight to isoclinal NNE-SSW–trending folds with vertically plunging hinge lines, which in turn are refolded by several scales of upright parasitic folds with N-S–trending hinge lines. Intersection of these three fold axes about a vertical point (Fig. 10E) suggests this structure is a Ramsay type-2 fold interference feature. Vertical boudinage affects the limbs of these folds and vertically intruded pegmatites and indicates the minimum principal stress (σ₃) was vertically orientated and much less than the horizontal stresses. These observations lead us to suggest the maximum and intermediate principal stresses (σ₁ and σ₂) must have been almost equal and orientated in the horizontal plane at the time of F₃ folding and shortly thereafter. This indicates that bulk horizontal shortening and constriction coincided with sillimanite-grade partial melting. During this constrictional regime, the maximum principal stress σ₁ must have been initially oriented N-S to produce E-W–trending folds and was subsequently rotated about a vertical axis into an E-W orientation to produce NNE-SSW–trending fold axes. Finally, NNW-SSE horizontal stretching lineations, horizontal boudinage, and horizontal fracture, indicative of a horizontal NNE-SSW–oriented σ₃, postdated the folding and vertical boudinage and occurred while the amphibolite was starting to behave in a brittle manner during cooling. These NNE-SSW horizontally boudinaged amphibolites indicate NNE-SSW extensional strains of ∼50% by dominantly pure shear (Fig. 9F), with a component of simple shear as demonstrated by domino boudins (Fig. 9D). The boudins show opposing shear senses on opposite limbs of F₃ folds, suggesting the shearing postdated folding (Fig. 9G; Virgo et al., 2018). Based on this evidence, we propose that NNE-SSW stretching, which is definitively related to horizontal crustal extension, must have postdated upright F₃ folding and horizontal constriction, but it commenced while the rocks were still hot enough to deform in a plastic manner shortly following peak anatectic conditions, but at temperatures when amphibolite starts to deform in a brittle manner (e.g., <640 °C at 6.5 kbar; Cao et al., 2017).

These interpretations agree with Virgo et al. (2018) and Von Hagke et al. (2018), who demonstrated that there are five generations of boudinage and that the first vertical boudins are overprinted by later NNE-SSW horizontal boudins. Vertical boudinage includes two generations of pinch and-swell boudins, the first with a longer and the second with a shorter wavelength. These features are followed by NNE-SSW horizontal domino boudins, torn boudins, and hairline veins reflecting embrittlement of the amphibolite layers associated with retrograde cooling and exhumation. These studies also demonstrated that outcrop-scale parasitic asymmetric folds predate torn boudins and hairline veins, whereas domino boudins indicate locally deviating shear sense on opposite limbs of F₃ folds (Fig. 9G). Virgo et al. (2018) and Von Hagke et al. (2018) concluded that long-wavelength pinch-and-swell boudins (i.e., vertically orientated boudinage) are consistent with synmigmatitic flow and F₃ folding in the surrounding rocks and noted that static recrystallization in the amphibolite must have postdated vertical boudinage. Based on this reasoning and the synchronicity among upright F₃ folding, vertical boudinage, and migmatization, we also agree that static recrystallization must have postdated upright F₃ folding and horizontal constriction at the center of the migmatite dome, but annealing was due to high temperatures outlasting contractional deformation during M₃, and prior to NNE-SSW extensional deformation. We therefore propose that the CHSZ documents a major rotation in principal stress axes from bulk horizontal shortening and constriction to NNE-SSW extension during the close of the M₃ event in the presence of partial melt.

Although extensional (S₃) top-to-the-NNE shearing is penetrative throughout the entire metamorphic carapace (Zas and Koronos Units), there is petrological and geochronological evidence that the fabrics formed at different times and at different conditions across Naxos. Discrete shear zones, represented by mylonites and show microstructures displayed in Figures 11–13, separate each tectono-stratigraphic unit and are truncated by the geometry of the NPDS. These are now described in chronological order from structurally deep to shallow.

The KSZ is a normal-sense NNE-SSW–trending shear zone that wraps around and is folded by the underlying migmatite dome. It represents the basal contact of the highly strained metamorphic carapace mantling the migmatite dome and places the kyanite-grade Koronos Unit against relatively less sheared migmatites of the Core Unit. Immediately around the dome margins, the migmatitic foliation is concordant with the overlying carapace; however, less than 100 m down structural section, this foliation becomes highly discordant, where F₃ upright folds and kilometer-scale subdomes can be traced along the migmatitic foliation (Kruckenberg et al., 2011). This abrupt truncation in foliation and its gradual rotation into alignment with the enveloping Koronos Unit suggest the presence of a major tectonic boundary along the migmatite-gneiss contact, characterized by extreme grain-size reduction. High-temperature blastomylonites, feldspar plasticity, grain boundary migration, and chessboard extinction of quartz indicate deformation temperatures exceeding 600 °C (Figs. 8 and 12). Strain analysis of kinematic indicators implies overall general shear (a combination of NNE-SSW simple shear and E-W pure shear), suggesting a large component of flattening perpendicular to this shear zone coeval with definitively top-to-the-NNE kinematics on all dome margins in both northern and southern Naxos. This strongly indicates this shear zone is folded by the doming migmatites, and therefore its movement predates migmatite dome formation.

Along the east and west margins of the migmatite dome, leucogranitic sills and dikes increase in abundance and emanate from the underlying migmatites (Vanderhaeghe, 2004). These sills and dikes commonly truncate the S₃ mylonite foliation, suggesting they intruded synchronously with or postdate high-temperature shearing on the KSZ. However, on western Naxos, they rotated about a NNE-SSW horizontal axis into alignment with the NPDS (see below; Fig. 6P). These dikes appear to have been rotated by ∼45° due to later doming of all fabrics. Domains of less-strained material unaffected by penetrative (S₃) S–C′ fabrics reveal earlier folding, i.e., (S_2a) crenulation cleavages and (S_2b) cleavage domains that display top-to-sigma porphyroclasts with opposing top-to-the-SW shear senses (see Fig. 8), signifying structural reactivation at high-grade conditions. Structural imbrication and preservation of relict thrust features are inferred from thin marble bands along the KSZ that display tight to isoclinal internal folding bounded by highly strained intercalated schists that have localized the deformation due to their competency contrast with the stronger marbles (see Fig. 8). Imbricated slices of the same marble horizon can be identified by an abrupt change in foliation across strike, and they splay off the main detachment horizon.

Slices of orthogneisses basement underlying the shelf sequence are structurally repeated and overlie metasediments and amphibolites at the upper levels of the shear zone. This, combined with microstructural evidence, suggests that the KSZ possibly represents a relict SW-verging thick-skinned thrust that was one of presumably many thrusts responsible for thickening of the continental margin. These structures imply crustal shortening and thickening was extensive across Naxos and possibly led to regional Barrovian metamorphism. The compressional features and thrusts were then reactivated as normal-sense shear zones during the exhumation of the migmatites from (M₂) kyanite-grade to (M₃) sillimanite-grade conditions in the footwall.

A NE-SW–trending greenschist-facies shear zone cuts through the flank of Mount Zas, juxtaposing the greenschist-facies (relict blueschists) Zas Unit against the underlying kyanite-grade Koronos Unit, and it can be traced over 20 km along strike (Figs. 3 and 4). This structure is associated with an increase in mylonitization over a 500 m zone, overprinting and truncating F₁ and F₂ structures. These folds rotate into alignment with the shear zone, producing sheath folds with their noses pointing parallel to the lineation (NE-SW), in agreement with Buick (1991a, 1991b). This structure could represent a reactivated thrust originally placing the Cycladic Blueschist Unit onto the underlying, more-proximal sedimentary cover (Koronos Unit), as relict (F₁) hanging-wall anticlines, S₂ crenulation cleavages, and relict top-to-the-SW S-C fabrics are preserved in micaceous domains and within greenschist-facies porphyroblasts (see Figs. 6 and 11). S-C′ mylonites are associated with top-to-the-NNE (S₃) shearing and exclusively affect greenschist-facies assemblages. These structures signify extensional reactivation during the close of regional (M₂) conditions at temperatures of 280–450 °C, as demonstrated by subgrain rotation and bulging of quartz grain microstructures (Stipp et al., 2002b).

The NPDS records the evolution of a ductile shear zone from solely brittle deformation at the top to completely ductile deformation beneath a few hundred meters (Buick, 1991a, 1991b). This detachment horizon truncates all compressional structures (Fig. 3) and is gently folded along a NNE-SSW axis, bounding the entire island. The brittle fault (Moutsana detachment; Cao et al., 2013, 2017), and its associated deformation, is best exposed in the NW and central Naxos (near the towns of Melanes and Galanado; Fig. 3) and on the eastern coastline at Moutsana peninsula. Numerous brittle high-angle faults root into the fault horizon, which is associated with a 20 m cataclasite and fault breccia zone, including pseudotachylytes, that grades into ductile mylonites. The fault surface is commonly steeply dipping at ∼40° perpendicular to the transport direction and is associated with a shallowly NNE-plunging lineation, which is oblique to the foliation, consistent with a lateral ramp geometry. In central Naxos, the detachment is folded into a gently northerly plunging synform (see Fig. 7G), causing the nonmetamorphic hanging wall to be exposed in a graben-type structure that is bounded by steep late crosscutting E-W–trending normal and NE-SW– and NW-SE–trending strike-slip faults (see Fig. 3), juxtaposing the granodiorite to the west and the metamorphic sequence to the east (see Figs. 3 and 4). This doming also affects the Pliocene–Pleistocene sedimentary successions within the hanging wall of the NPDS and requires a component of E-W shortening during and after movement on this structure, which was also responsible for folding the isograds and development of minor brittle thrust faults.

Beneath this low-angle normal fault, right-way-up regional metamorphic isograds are telescoped and affected by flattening, indicating the fault cut through metamorphic stratigraphy (Fig. 7G). On west Naxos, brittle deformed cataclasites and sediments are juxtaposed against upper-amphibolite-grade migmatites, a transition of ∼700 °C within 1000 m across strike, revealing frozen-in metamorphic field gradients of up to ∼700 °C km⁻¹. Although S-C′ fabrics (S₃) affect all structural levels of the footwall, there is a larger intensity of shearing fabric development on this structure. Quartz microstructures indicate dynamic recrystallization occurred via a spectrum of deformation mechanisms, from subgrain rotation and bulging to brittle deformation, indicative of deformation temperatures less than ∼500 °C (Stipp et al., 2002a). Over 95% of samples demonstrate top-to-the-NNE kinematic indicators by asymmetrical alpha and delta porphyroclasts of biotite, kyanite, plagioclase, and garnet, in agreement with Urai et al. (1990) and Buick (1991a, 1991b). We interpret this brittle-ductile shear zone to be related to crustal extension and exhumation, as brittle normal faults root into it, and it clearly postdates, is discordant to, and truncates all compressional and metamorphic features. The west Naxos I-type granodiorite is mylonitized on its eastern margin, indicating the fault was active during crystallization at ca. 12.2 Ma (Keay et al., 2001). The NPDS presumably represents a similar structure to the North Cycladic Detachment System on Mykonos, Tinos, and Andros (e.g., Jolivet et al., 2010; Jolivet and Brun, 2010) in accommodating regional crustal extension by top-to-the-NNE shearing along a localized zone on the brittle-ductile transition.

In contrast to the NPDS, the extensional fabrics in the ZSZ and the KSZ and condensed right-way-up high-grade metamorphic isograds are purely ductile features that are concordant to the island-scale foliation. The structural discordance between these internal shear zones and the brittle-ductile NPDS (Fig. 3) suggests that movement on the KSZ and ZSZ predated movement on the NPDS and therefore records the early stages of exhumation in the mid- to lower crust prior to migmatite doming and regional NNE-SSW Aegean extension.

In western central Naxos, a largely dismembered sequence of ophiolitic material composed of hydrothermally altered pillow basalts, serpentinites, and oceanic sediments, including cherts with manganese nodules, outcrops in the hanging wall of the NPDS that transects central Naxos (Jansen and Schuiling, 1976; Vanderhaeghe, 2004). Structurally above this, there is a disturbed sequence of Nummulite-bearing dolomitic limestones/marbles, chloritized schists, and randomly orientated exotic blocks of marble and volcanic debris (Fig. 7). These field relations have been previously attributed to a “km-scale gravity slide” derived from the footwall of this detachment system (Vanderhaeghe, 2004; Kruckenberg et al., 2011). Although this could explain the wide variety of rock types, randomly orientated metamorphic boulders, and evidence for paleo-karstification, the extreme folding and entrainment of a variety of ophiolitic and distal oceanic material as blocks within a serpentine matrix bear many similarities to the ophiolitic mélanges of the Oman–United Arab Emirates ophiolite (Searle and Cox, 1999, 2002). We tentatively suggest that this sequence represents a tectonic mélange formed in a trench environment and was deformed during SW obduction of an ophiolite prior to Eocene subduction of the continental margin. Although scattered remains of upper ophiolitic material are distributed across the Cyclades at high structural levels (Katzir et al., 1996; Hinsken et al., 2017), ophiolite stratigraphy, including mantle peridotite, gabbro, plagiogranite, and a metamorphic sole, is exposed on Tinos (Lamont, 2018), implying that the ophiolite may have been widespread as the highest structural nappe across the entire Cyclades. The Miocene to Pleistocene sedimentary successions are unconformably laid on top of this mélange, and they are composed of fluvial conglomerates, in some cases with boulders and clasts of a variety of ophiolitic and metamorphic material derived from the footwall. The sedimentary rocks dip at ∼30° away from the core complex, which is inconsistent with the dip predicted by normal faulting. Clearly, these sediments must have been rotated by gentle doming of the island during the Pliocene–Pleistocene and therefore indicate a significant component of E-W shortening after the core complex was exhumed. The I-type granodiorite has aplitic sills that are folded along an N-S–trending axis (see Data Repository Item¹), also suggesting significant E-W shortening during and following crystallization at ca. 12.2 Ma.

Blueschist, gneiss, and migmatite samples used in this study were collected at various structural levels of the Naxos MCC to investigate spatial variations in the pressure-temperature (P-T) conditions of peak metamorphism. Previous thermobarometric P-T results from studies of the Naxos MCC are presented in Table 1. The analytical techniques and procedures utilized to collect mineral composition data are outlined in the Data Repository Item (see ^{footnote 1}). Mineral abbreviations follow the guidelines of Whitney and Evans (2010), and migmatite terminology is after Ashworth (1975). Anhydrous phase compositions were calculated to standard numbers of oxygen per formula unit (pfu; Deer et al., 1992), micas were recalculated to 11 oxygens, and chlorite was recalculated to 28 oxygens. Where present, H₂O content was assumed to occur in stoichiometric amounts. The proportion of Fe³⁺/Fe_total was calculated using AX (Holland, 2009), and amphibole exchange vectors were calculated using the method of Holland and Blundy (1994). Representative photomicrographs, backscattered electron images, and mineral compositions are shown for the Zas Unit in Figure 11, the Koronos Unit in Figure 12, and the Core Unit in Figure 13. A summary of the relative timings of fabric development is displayed in Figure 14, and garnet maps and line profiles from samples TL67, TL15, and TL66 are presented in Figure 15, Tables 2–4, and the Data Repository Item (see ^{footnote 1}). Summary P-T results of all samples are presented in Table 5.

All samples in this unit display microstructural evidence for M₂ greenschist-facies overprinting of older M₁ blueschist-facies mineral assemblages and fabrics, with replacement spatially concentrated along zones of fluid alteration. M₁ high-pressure precursors occur as prograde inclusions or as relict matrix phases that have been pseudomorphed or trapped inside late lower-grade minerals. Glaucophane is unzoned and is preserved within the cores of epidote grains, with phengitic and paragonitic white mica, clinozoisite, and no plagioclase, and also along unaltered microstructural domains and foliations (Figs. 11A, 11C, 11G, and 11I).

All mineral compositional trends are displayed in Figures 11J, 11K, and 11L and highlight the high Na and low Al contents of primary amphibole, high Si pfu of mica, indicative of phengitic compositions, and paragonite that is characteristic of blueschist-facies conditions. A subsidiary population of actinolite is confined to the matrix, and it characterizes the M₂ greenschist-facies overprint. Late crosscutting greenschist-facies minerals are well developed in lesser-strained domains within the preexisting foliation, presumably from the breakdown of large porphyroblasts such as garnet to form epidote, plagioclase, chlorite, tremolite, and actinolite (Figs. 11G and 11H). Sodic amphiboles occur mainly as small prismatic inclusions in epidote and commonly display decussate-shear textures that could be associated with top-to-the-NE shearing during exhumation from blueschist facies (S₁; Figs. 11A, 11C, 11G, and 11I). This is aligned subparallel to the pervasive greenschist-facies top-to-the-NNE extensional fabric (S₃), which solely affects the surrounding quartz-mica–rich matrix. In low-strain domains, blueschist- to greenschist-facies assemblages containing phengitic muscovite, paragonite, biotite, plagioclase, green amphibole, and cloudy plagioclase are deformed in crenulation cleavages (S₂) that form a continuous fabric as inclusions throughout epidotes, with preserved microlithons (S_2a) and cleavage domains (S_2b; Figs. 11B and 11D–11H). These grains are rimmed by retrograde plagioclase and chlorite, indicating they are most likely associated with decompression from high-pressure conditions. S_2a is also associated with centimeter-scale isoclinal folding, with well-developed microlithons that act as parasitic folds on the limbs of larger-scale folds (see Fig. 11F). S_2a and S_2b presumably formed by shortening along two orthogonal axes at upper-greenschist-facies conditions but predate the characteristic lower-pressure retrograde M₂ greenschist assemblage. We therefore relate S₂ fabrics to overthrusting and ductile stacking of high-pressure nappes at crustal depths that postdated high-pressure conditions.

S₂ fabrics contrast to the more localized top-to-the-NNE S-C′ shearing fabric (S₃), which purely affected the greenschist-facies assemblages, including quartz, mica, plagioclase, and chlorite–rich domains. S₃ mylonite fabrics are heterogeneous and localized along discrete shear zones. S₃ is particularly developed in samples taken from the ZSZ, in a parallel orientation to the fold axial planes, and appears to utilize S_2b cleavage domains. Microstructurally, S₃ deforms the quartz and fine-grained mica-rich matrix around pretectonic greenschist-facies porphyroclasts, including clinozoisite, epidote, and late crosscutting muscovites, creating strain shadows, asymmetrical boudinage, and rotation of the porphyroclasts (Figs. 11A, 11B, and 11C). Dynamic recrystallization via subgrain rotation and bulging is persistent within quartz-rich domains, indicating deformation temperatures must not have exceeded 450 °C (Stipp et al., 2002a, 2002b). Due to these relationships, we interpret S₃ as a purely retrograde feature postdating (M₁) high-pressure and (M₂) greenschist-facies conditions, and it was related to exhumation of the core complex through the brittle-ductile transition (Buick and Holland, 1989; Urai et al., 1990; Avigad, 1998).

This unit also shows contrasting microstructural fabrics to the overlying Zas Unit. S₁ is aligned parallel to bedding and is preserved as inclusions within kyanite and garnet cores, whereas crenulation cleavages (S_2a and S_2b) can be found in less-strained enclaves and as inclusions in garnet. Isoclinal folding occurs on a millimeter scale and is associated with microlithons (S_2a) and utilizes compositional layering between quartzofeldspathic and pelitic domains (see Figs. 12A–12C and 12E). Cleavage domains (S_2b) are superimposed on S_2a and in places completely overprint the microlithons to produce a strong spaced foliation defined by biotite-muscovite foliation (Figs. 12D and 12E). Large kyanite porphyroblasts are aligned coaxial with the foliations (S_2a and S_2b) and sometimes crosscut them, indicating these are prograde features. In numerous samples (e.g., TL67), euhedral garnets overgrow kyanite-bearing S_2b fabrics preserved in aluminum-rich domains, whereas in other samples (e.g., TLN22), garnets are enclosed by large kyanite porphyroblasts, suggesting simultaneous prograde growth of kyanite at peak M₂ conditions. In sample TLN21, garnets are up to several millimeters in diameter and distributed uniquely along tight isoclinally folded pelitic layers (S_2a), whereas in TL67, euhedral garnet porphyroblasts show straight faces truncating both S_2a and S_2b foliations, indicating garnet growth in this sample postdated initial thrusting and folding (Fig. 12D). Within some of these garnet porphyroblasts, dark fluid inclusions indicate garnet growth in a mixed saline CO₂/H₂O fluid, as noted previously (Baker and Matthews, 1994; Baker et al., 1989; Bickle and Baker, 1990; Kruelen, 1980). Kyanite has commonly broken down to biotite and is confined to pelitic domains with garnet. These domains form enclaves that are partially recrystallized as an extensional blastomylonitic top-to-the-NNE shear fabric (S₃) affecting the matrix and deforming around stronger kyanite-biotite-muscovite-plagioclase-garnet–rich enclaves, indicating these M₂ kyanite-grade assemblages are pretectonic with respect to S₃ (Figs. 12D, 12G, and 12H). S₃ was accommodated via grain boundary migration of quartz and ductile deformation of plagioclase associated with temperatures exceeding 600 °C (Stipp and Kunze, 2008). This was followed by coarsening and Ostwald ripening of grain boundaries, revealing static recrystallization. Annealing postdated deformation at close to peak conditions (Fig. 13J).

Toward the base of this unit along the KSZ, many samples display evidence for a thermal and static, lower-pressure overprint on a blastomylonitic S₃ fabric. This is demonstrated by posttectonic growth of (M₃) sillimanite and garnet (e.g., TLN21 and TLN22), which truncate the top-to-the-NNE S₃ fabric, indicating that high temperatures outlasted deformation. Peak (M₂) kyanite-bearing assemblages are overprinted by (M₃) sillimanite, and biotite intergrowths occur with plagioclase and quartz in dilational strain zones, demonstrating that the transition from kyanite to sillimanite occurred via a Carmicheal-type reaction scheme (Carmicheal, 1969) that postdated shearing (Figs. 12H and 12I).

Within metapelitic samples, garnet shows a variety of compositional profiles that correlate directly with their microstructural relationships (Fig. 15; Data Repository Item [see ^{footnote 1}]). Garnet in TL67 preserves prograde growth zoning associated with a smooth transition from core to rim, increasing pyrope content, and homogeneous to slightly decreasing grossular content, as shown in Figure 15B. An inflection in the spessartine content on the outermost rim is indicative of some garnet resorption following peak conditions. In contrast, TLN22 has small garnets that crosscut the S₃ fabric and in places are engulfed by kyanite. These garnets show either a general decrease in the grossular component from core to rim, or the pyrope content slightly increases toward the inner rim followed by a decrease toward the outer rim, which is indicative of late garnet growth at lower pressure and temperature or homogeneous garnet profiles due to re-equilibration at lower pressures.

Along the KSZ, there is a band of ultramafic rock that is enclosed within foliated garnet-bearing amphibolites that are semicontinuous along strike and wrap around the dome. The meta-peridotite enclaves enclosed within the amphibolites show partial serpentinization to anthopyllite, talc, and actinolite, but they preserve in situ olivine and orthopyroxene. Katzir et al. (1999) provided a detailed description of the petrography and geochemistry of these lenses and noted bimodal temperatures of ∼700 °C and 1200 °C from pyroxene thermometry, suggesting they remained hot as they were emplaced into the shelf carbonate sequence. Within these amphibolites, there is some degree of hydrous M₂ migmatization but no petrographic evidence whatsoever to suggest they attained M₁ blueschist-facies conditions.

Samples from the migmatite dome preserve a petrologic record of kyanite- to sillimanite-grade (M₂–M₃) partial melting, and they display no evidence for having experienced Eocene M₁ conditions. Careful textural examination revealed that hydrous melting was the dominant melting process occurring within the Naxos dome, capable of producing between 1% and 7% volume melt based on reaction stoichiometry. However, at the deepest levels in pelitic lithologies, incongruent muscovite dehydration melting was the major melting process associated with growth of peritectic K-feldspar, removal of muscovite, and melt segregation from larger melt fractions (>20% melt), which eventually pooled to form S-type leucogranites that crosscut and intruded into the mantling carapace.

There is evidence for a heterogeneous degree of melting throughout the Core Unit, due to lithological and bulk compositional controls because of repetition of infertile orthogneissic basement along with more fertile metasedimentary horizons, due to prograde thrusting. Samples TL66 and TL15 outcrop as stromatic migmatites with centimeter-scale folded leucosome and melanosome banding, whereas samples TLN18 and TLN20A represent diatexites. Melanosome foliations are defined mostly by muscovite, biotite, fibrolite (fibrous sillimanite), relict kyanite, and in some samples peritectic K-feldspar (Figs. 13G–13I). Leucosome segregations contain around equal proportions of quartz and plagioclase, with lesser amounts of biotite, muscovite, fibrolite, garnet, and ilmenite, while K-feldspar occurs in other samples (e.g., TL59, TLN18, TLN34, TLN35). Accessory minerals include tourmaline, apatite, magnetite, pyrite, zircon, and monazite. Leucosome segregations comprise ∼10%–20% of the migmatite sequence in outcrop. Diatexite only occurs at the deepest levels and has foliations defined by schollen of biotite and some relict muscovite, with leucosomes characterizing larger volumes of the rock (>20%).

Garnet occurs in both the leucosome and melanosomes as small idioblastic porphyroblasts <1 mm in diameter, displaying subhedral to anhedral faces (Figs. 13A, 13E, and 13F). In sample TL15, leucosome garnet preserves rotational inclusion trails, which define the S_2a fabric, that rotate into alignment with the external matrix S_2b fabric toward its rims (Figs. 14A and 14J), indicating garnet grew synkinematically with the development of these prograde features. However, many of these porphyroblasts in sillimanite-grade samples are fractured and filled by biotite and muscovite intergrowths (Fig. 13F). In samples TLN34 and TL66, garnet is almost compositionally homogeneous (Fig. 15F; Data Repository Item [see ^{footnote 1}]). Muscovite in sample TL59 and TL66 occurs similarly as laths with associated fibrolite or as symplectic intergrowths with quartz (Figs. 13E–13H). These intergrowths are often associated with either biotite-quartz symplectite or myrmekite (plagioclase-quartz intergrowths). Secondary muscovite in leucosomes is oriented randomly relative to the foliation and fibrolitic sillimanite in the matrix, and on the margins of leucosomes, it is typically surrounded by coarse-grained, crosscutting muscovite (Figs. 14E–14F; Spear et al., 1990). Biotite occurs as individual crystals in the leucosome, as part of the melanosome foliation, and also as symplectic intergrowths with quartz, but it shows no significant compositional variation between these petrographic positions with Mg/(Mg + Fe). In kyanite-grade migmatites (e.g., TL15), kyanite is associated with the S₂ foliation, which is folded by F₃ folds (Figs. 14B–14D), indicating upright F₃ folding postdated kyanite growth and occurred after the rocks attained their peak pressures.

Several authors have observed that K-feldspar is commonly absent from sillimanite-grade migmatites despite being present at peak conditions, due to replacement by secondary muscovite in the back reaction during cooling (Ashworth, 1975; Tippett, 1980). By identifying different generations of muscovite through various petrographic characteristics, secondary muscovite can be distinguished. Tyler and Ashworth (1982) suggested that prograde (primary) muscovite that survives the partial melting process often occurs as flakes orientated in the melanosome foliation. Retrograde muscovite occurs as large lath-like porphyroblasts containing sillimanite, or as symplectic intergrowths with quartz (Figs. 13E–13H; Ashworth, 1975, 1979; Tippett, 1980). Textural features in sample TL66 include muscovite laths containing numerous fibrolite inclusions and muscovite-quartz symplectite and intergrowths of sillimanite-quartz “faserkiesel,” which are typical of retrograde replacement of K-feldspar at sillimanite-grade conditions (Figs. 13E–13G; Evans and Guidotti, 1966; Ashworth, 1975; Tippett 1980; Brown, 2002), i.e., the reversal of the muscovite dehydration melting reaction.

Although peritectic K-feldspar is commonly lacking in the Naxos migmatites due to hydrous melting textures and retrograde replacement by secondary muscovite, leucosomes in migmatites from deep levels are volumetrically abundant in outcrop, representing ∼15%–25% volume proportion of rock. This volume of melt greatly exceeds the proposed melt connectivity threshold of 7%–10% that allows melt extraction in such lithologies (Rosenberg and Handy, 2005). However, it has been shown that the observed leucocratic component of migmatites does not directly correlate with actual quantities of melting. At shallow dome levels, rocks that have been described as leucosomes, leucogneiss, and diatexites are interpreted here as orthogneiss basement that coarsened upon experiencing small degrees of partial melting (B. Dyck, 2016, personal commun.). This has major implications for the proposed mechanisms that drove migmatite doming, as the weakening effect depending on the extent of partial melting may have been extremely overestimated. At structurally high levels of the dome, the stability of primary muscovite within the melanosomes supports this interpretation, as they did not experience muscovite dehydration melting (the first major melt-producing reaction). Although diatexites occur on Naxos, they are confined to fertile lithologies at deeper levels upon crossing the K-feldspar isograd.

Some inferences about the temperature of mineral growth can be made using the compositional characteristics of different mineral generations. Prograde muscovite commonly contains higher paragonite content than retrograde muscovite that has replaced K-feldspar (Evans and Guidotti, 1966; Ashworth, 1975). Generally, low Si contents (<3.10 pfu per 11 oxygens) combined with high Ti contents (up to 0.08 pfu) are characteristic of high-grade and magmatic muscovite (Miller et al., 1981; Guidotti, 1984). In all analyzed samples, leucosome muscovite is low in Si and contains a moderate to high Ti content (∼0.04–0.10; Fig. 15), indicating that they formed during melt crystallization or retrogressively while still at high temperature. Compositional zoning profiles in garnet give further support for high-grade metamorphic conditions, as inferred from compositional and textural characteristics in muscovite. Several authors have demonstrated that garnet compositional zoning profiles form principally as a function of changes in pressure and temperature conditions during growth (e.g., Tracy et al., 1976; Spear and Selverstone, 1983; St-Onge, 1987). During prograde metamorphism in a pelitic bulk composition, garnet preferentially fractionates Mn into core regions, and therefore spessartine content decreases toward rims as a result of progressive Mn depletion in the reacting matrix (Hollister, 1966). However, in the Naxos migmatites, garnet contains spessartine profiles that are homogeneous or concave upward, with increasing Mn content from core to rim. The length scale of Mn (and Mg, Fe, and Ca) zonation within the garnets is less than the diffusional length scale (up to 3 mm) suggested by >5 m.y. residence time at peak temperatures >650 °C (Caddick et al., 2010). These factors indicate that preservation of the prograde growth history will be modified somewhat by diffusion, whereas peak conditions will be preserved by thermal re-equilibration (Woodsworth, 1977; Spear, 1993). Nevertheless, the shape and trends of garnet profiles will be somewhat preserved if peak conditions were attained for only a geologically brief time, and therefore insight into the prograde metamorphic history of the Core Unit may be obtained. Therefore, in the following, we take peak pressure estimates as a minimum depth of burial.

Kyanite-grade migmatites (e.g., TL15) occur at higher structural levels of the dome and record the earlier prograde history. Texturally, these are stromatic migmatites with distinct leucosomes highlighting centimeter-scale ptygmatic flow folding during anatectic conditions. In these samples, melanosomes preserve the S_2a crenulation cleavages and S_2b spaced foliations along cleavage domains similar to the overlying Koronos Unit (Figs. 13C, 13D, and 13H). Kyanite is bladed and lies in the folded S₂ foliation, indicating that it grew in Al-rich bands prior to anatectic deformation associated with upright F₃ folding (Figs. 13B–13C). In contrast, some biotite and secondary muscovite in the melanosome are orientated axial planar to and crosscut the folding, indicating they postdate deformation and are mainly retrogressive features associated with hydration contemporaneous with regional NNE-SSW extensional exhumation. This specimen does not exhibit any of the K-feldspar replacement features described above, suggesting that its leucosomes were never of a granitic composition, and consequently their formation must be a result of a K-feldspar–absent melting reaction. When H₂O is in excess, K-feldspar–absent reactions extend to lower temperatures (García-Casco et al., 2001, 2003; Cruciani et al., 2008); increasing pressure also favors fluid-saturated melting reactions before reaching conditions that favor dehydration melting during typical prograde metamorphism (Cruciani et al., 2008). Due to an inverse correlation between pressure and plagioclase stability, an increase in the plagioclase/muscovite ratio consumed in melting reactions with increasing pressure (Patiño Douce and Harris, 1998), and increasing H₂O activity lead to preferential solution of plagioclase components by lowering the plagioclase + quartz solidus (Conrad et al., 1988). These factors result in a significant shift of products of H₂O-fluxed melting toward tonalitic/trondhjemitic compositions and are typical of fluid-rich environments at high-pressure and low-temperature conditions.

Several forms of thermobarometry were used in this work to constrain the P-T conditions of metamorphism on Naxos, including the Ti-in-biotite thermometer of Henry et al. (2005), the garnet–aluminum silicate–plagioclase–quartz (GASP) barometer, the garnet-biotite thermobarometer (Spear, 1993; Bhattacharya et al., 1992; Holdaway, 2000), and AvPT (Powell et al., 1998) using THERMOCALC version 3.40i, which was employed using characteristic end members for each sample. The results are displayed in Figure 16. Activities of solid-solution end members were calculated (Table 5) using AX (Holland, 2009). The results systematically demonstrate that there are two distinct P-T populations on Naxos. M₁ blueschist-facies P-T conditions are exclusively confined to the Zas Unit (Fig. 16A, blue circles) and give typical P-T results of 12.3 ± 0.8 kbar and 483 ± 13 °C (TLN54). In contrast, an upper-amphibolite-facies M₂-M₃ population is recorded within the Koronos and Core Units (Fig. 16A, green and red circles, respectively). These rocks typically record increasing pressure and temperature conditions from garnet core to rim. Kyanite-bearing M₂ assemblages record higher pressures, up to 11.4 ± 1.0 kbar and 682 ± 38 °C (TL67). Sillimanite-bearing M₃ assemblages record re-equilibration at lower-pressure conditions, i.e., 7.5 ± 1.2 kbar and 742 ± 39 °C (TL66). Within these units, prograde garnet compositions show increases in both pressure and temperature from core to rim diagnostic of crustal thickening and do not represent isobaric heating, as previously inferred (Buick and Holland, 1989; Buick, 1991a, 1991b). However, the nature of the P-T path leading up to and during peak anatectic conditions is poorly constrained using this method. This is because conventional thermobarometry is largely ineffective at constraining peak P-T conditions in high-grade and/or partially melted rocks (e.g., Halpin et al., 2007; Indares et al., 2008; Palin et al., 2013) owing to changing mineral compositions due to retrograde diffusion (Kohn and Spear, 2000; Pattison and Begin, 1994, 2003), and it has large associated uncertainties (typically ±50 °C and ±1 kbar at 1 standard deviation; Powell and Holland, 2008; Palin et al., 2016). As such, phase diagram modeling (pseudosection construction) was employed herein as the main investigative tool to harness P-T information, which has significantly smaller uncertainties.

Pseudosections were constructed using THERMOCALC version 3.40i and internally consistent thermodynamic data set ds-62 (Powell and Holland, 1988; Holland and Powell, 2011). Modeling of metasediments TL67, TL15, and TL66 was performed in the MnO–Na₂O–CaO–K₂O–FeO–MgO–Al₂O₃–SiO₂–H₂O–TiO₂–O (MnNCKFMASHTO) system using the following activity-composition relations: silicate melt (White et al., 2007), cordierite (Mahar et al., 1997; Holland and Powell, 1998), garnet and ilmenite (White et al., 2005), orthopyroxene (White et al., 2002), chlorite (Mahar et al., 1997; Holland et al., 1998), Ti-bearing biotite (White et al., 2007), muscovite (Coggon and Holland, 2002), K-feldspar and plagioclase (Holland and Powell, 2003), and magnetite (White et al., 2002). Modeling of metabasic lithologies in TLN54 was performed in the MnO-absent NCKFMASHTO system, with the additional usage of the omphacitic clinopyroxene and amphibole activity-composition relations of Green et al. (2016). The pure phases andalusite, kyanite, sillimanite, rutile, quartz, and H₂O were considered in all cases. The bulk-rock suprasolidus water content of migmatitic units was fixed on an individual basis in order to allow minimal fluid saturation at the wet solidus, here defined as ∼1 mol% free H₂O.

Detailed investigation into the prograde evolution and P-T conditions leading up to peak metamorphism was achieved by investigating compositional isopleths for pyrope and grossular content in garnet. Since these isopleths vary to first order with changes in pressure and temperature and commonly intersect at high angles, they specify unique intersection points with a high degree of confidence for tracking garnet composition evolution in P-T space. The intersections of isopleths representing measured compositions are represented by shaded boxes indicating uncertainties at the 1σ level calculated by THERMOCALC.

Figure 17 shows the calculated P-T pseudosection for blueschist-facies sample TLN54. Using the methods described in the Data Repository Item (see ^{footnote 1}), and the absence of carbonate phases for the modeled area, the fluid was assumed to have pure H₂O in excess. The observed peak M₁ assemblage Gl-Ms-Chl-Rt-Sph-Qtz was calculated to be stable at ∼11–15 kbar and ∼400–520 °C. Glaucophane isopleths of z(gl) = 0.96, f(gl) = 0.02–0.03, and a(gl) = 0.05–0.06 suggest that peak M₁P-T conditions reached at least 14.5 ± 0.5 kbar and 470 ± 20 °C. Model amphibole compositions match well with observed sodic amphibole analytical data from both samples TLN54 and TLN25 (Fig. 15). Ranges of AvPT results from samples TLN54, TLN25, and TLN26 from the Zas Unit record minimum P-T conditions of around 12 kbar and 450 °C, i.e., a similar result to peak metamorphic conditions of ∼13 kbar and ∼450 °C suggested by Avigad (1998) for a jadeite-bearing blueschist within the Zas Unit, and comparable to the recently published results of 15.5 kbar and 576 °C by Peillod et al. (2017).

Sample TL67 reveals the prograde and peak P-T conditions for the Koronos Unit and is shown in Figure 18. Using a pure H₂O aqueous fluid, the predicted assemblages are incompatible with the observed textures and assemblages, particularly, the lack of evidence for melting and subsolidus K-feldspar. Upon careful observation, the existence of graphite and dark CO₂ fluid inclusions in garnet and quartz indicate circulation of CO₂ fluids during peak metamorphism. A reduction in water activity, a(H₂O), causes the dehydration equilibria that underlie prograde processes in metasedimentary rocks to be displaced to lower temperature. Although estimating the actual value of water activity in a rock system is difficult, some constraints can be placed if the rock contains graphite. The coexistence of solid carbon with a C-H-O fluid has three degrees of freedom, so that one further constraint, such as oxygen fugacity, potentially allows the concentration of all C-H-O fluid species to be calculated at any given P and T. Therefore, an investigation of the effect of water activity on the calculated phase relations was conducted (Fig. 18B). Using this method, a reduced a(H₂O) = 0.5 was chosen to represent fluids during peak metamorphism, which corresponds to the findings of previous studies of metamorphic fluids on the island (Kreulen, 1980; Baker et al., 1989). This has been shown to shift the solidus to higher temperatures (Pëto, 1976; Peterson and Newton, 1989; Stevens and Clemens, 1993; Weinberg and Hasalová, 2015), stabilizing K-feldspar at subsolidus temperatures as shown in Figures 18B and 18C (Newton, 1989).

For a localized Al-rich domain, compositional isopleths representative of garnet core domains (∼10% pyrope and ∼11% grossular) intersect at ∼6.5 kbar and 550 °C in the assemblage field Grt–Ms–St–Bt–Pl–Ilm–Qtz–H₂O, and those representative of the inner rim (14%–15% pyrope and 13% grossular) intersect at 10.5 kbar and 675 °C in the assemblage field Grt–Ms–Bt–Pl–Ilm–Rt–Qtz–H₂O. This is consistent with temperature estimates of 630–710 °C calculated from the Ti concentration of biotite (Henry et al., 2005) and similar to pressures calculated by Duchene et al. (2006); therefore, we take these estimates to represent peak M₂P-T conditions. Garnet outermost rims (13%–14% pyrope and 10% grossular) have isopleth intersections at 7.5 kbar and 655 °C and cross the muscovite dehydration reaction to produce K-feldspar. This is associated with an increase in garnet mode, which is consistent with garnet rims crosscutting the S₂ kyanite-grade fabric. Garnet-rim upward inflections in Fe/(Fe + Mg) and spessartine content are also characteristic features of retrograde replacement by biotite (Kohn and Spear, 2000), and they suggest some garnet was resorbed during retrogression.

Sample TL15 records the prograde evolution of the Naxos migmatites, and the corresponding pseudosection is shown in Figure 19. Isopleths for garnet core compositions (10% pyrope and 13%–16% grossular) indicate prograde metamorphism through the Grt–Ms–Bt–Pl–Ilm–Chl–Qtz–H₂O assemblage field at ∼6 kbar and ∼550 °C. During prograde metamorphism, significant garnet is predicted to have formed via staurolite breakdown (e.g., Eq. 1). Garnet compositions just within the inner diffusion-affected porphyroblast rims provide minimum constraints on the conditions of peak metamorphism (Caddick et al., 2010). Isopleths for 17% pyrope and 9–10% grossular intersect at ∼9.5 kbar and 700 °C in the Grt–Ms–Bt–Pl–Ilm–Qtz–Liq–H₂O–Ky assemblage field, indicating that prograde metamorphism occurred along a heating path that crossed the water-saturated solidus (Eq. 6) at a minimum of 8 kbar and 680 °C. High-pressure melting is in agreement with the presence of prograde muscovite, bladed kyanites, and small proportions (1%–4%) of melt distributed along rounded grain boundaries, although peak pressures may have been higher due to diffusional re-equilibration (Caddick et al., 2010). As this rock lies within the same unit as sillimanite-grade migmatites, it is likely that this rock experienced the same prograde path as the underlying sillimanite-bearing migmatites experienced before obtaining higher temperatures. Due to the lack of sillimanite in this rock, but locations within tens to hundreds of meters of the sillimanite isograd, this rock may have crossed the kyanite-sillimanite transition at peak conditions via an indirect Carmichael reaction scheme, but it preserves no textural evidence owing to its sluggish reaction kinetics. Overall, peak P-T conditions of ∼730 ± 20 °C and 9.5 ± 0.8 kbar obtained from sample TL15 are consistent with peak conditions for sample TL59 (see Data Repository Item [see ^{footnote 1}]). Therefore, burial of the Core Unit occurred under kyanite-grade (M₂) conditions of 680–730 °C and minimum pressures of 8–10 kbar, similar to conditions of 670 ± 31 °C and 10 ± 0.5 kbar in the overlying Koronos Unit. This should be taken as a minimum pressure, and the apparent pressure inversion between the Core Unit and the Koronos Unit could be explained by diffusional homogenization between two contrasting garnet compositions at high-grade conditions. The textural relationships of kyanite aligned and within the upright folded migmatitic foliation suggest that upright (F₃) folding along a N-S axis occurred immediately following peak M₂ kyanite-bearing conditions. Therefore, the E-W–directed compressional deformation that produced these folds took place at lower pressures in the sillimanite stability field (M₃), for which the P-T conditions were constrained using garnet rim compositions affected by retrograde equilibration. These compositions (13% pyrope and 6% grossular) suggest P-T conditions of ∼7 kbar and 710 °C (Fig. 18), signifying an almost isothermal decompression P-T path during exhumation, cooling at ∼3 °C km⁻¹.

TL66 was collected from the deepest parts of the dome across the K-feldspar isograd, and bulk compositions were calculated from the observed mineral proportions combining both leucosome and melanosome. Leucocratic networks within and radiating from the Naxos high-grade core combined with the partial preservation of garnet suggest that some melt was lost from the rock system during its P-T evolution (White and Powell, 2002). Therefore, this bulk composition is appropriate for modeling P-T conditions after the loss of that melt fraction (i.e., it is valid for assessment of peak/near-peak and suprasolidus retrograde metamorphism, but not the prograde subsolidus evolution prior to melt loss). Phase equilibria with excess H₂O were calculated for subsolidus and suprasolidus conditions, representing an external flux of water to produce maximum melting. The resulting pseudosections are presented in Figures 19D and 19F, revealing complex subsolidus topologies, whereas the suprasolidus phase relationships are simple, with large areas of P-T space without a change in variance. A closed system model, more realistic melting, was calculated by analyzing the water content at the hydrous solidus. It was determined that 7.6% H₂O was required to ensure all H₂O enters melt at the solidus, leaving water undersaturated at suprasolidus conditions. The calculated bulk compositions are shown in the Data Repository (see ^{footnote 1}), along with associated pseudosections.

At subsolidus conditions, both models predict low proportions of garnet, consistent with the sample, and a garnet core composition (14% pyrope, 9% grossular) plotting within the Grt–Ms–Bt–Ky–Ilm–Pl–Qtz–H₂O field at 6.0 kbar and 610 °C. This is within 0.3 kbar of the garnet-producing reaction, probably reflecting delayed garnet nucleation. The model predicts a prograde P-T path through the kyanite stability field, accounting for kyanite inclusions in muscovite. At suprasolidus conditions, the externally open system H₂O model predicts K-feldspar absence in the presence of high melt fractions, consistent with faserkiesel textures and diatexites. Garnet rim compositions (15% pyrope, 6% grossular) plot at 4.9 kbar and 705 °C with 33% melt. It is unlikely this rock experienced such extensive melting, and the presence of kyanite preserved as inclusions within muscovite suggests little fluid infiltration above the solidus.

In this case, 11% volume K-feldspar could be attained with just 14% volume muscovite in a simple 1:1 molar ratio. Production of secondary muscovite through this reaction would also account for less sillimanite and biotite observed than predicted (8% observed compared to 10% predicted) by production of secondary garnet and muscovite. However, due to diffusional equilibration at high temperature, it is unlikely that garnet rim compositions represent peak temperatures. Ti-in-biotite thermometry also predicted most biotite homogenized at temperatures of 650–700 °C, which is unlikely to represent peak temperatures. However, AvPT predicted garnet innermost rims with representative matrix minerals compositions of 7.5 ± 1.2 kbar and 742 ± 39 °C (Table 5), in agreement with predicted garnet modal proportions. Based on these lines of reasoning, we are confident these results represent peak M₃ sillimanite-grade conditions.

To account for leucogranites that emanate from and throughout the Core Unit, it is likely that the Naxos migmatites have experienced some degree of melt loss. This occurs because once melt accumulation reaches a critical threshold, a portion may be lost, creating an open-system environment (Vanderhaege, 2004). This melt composition is dependent on the P-T conditions of formation (White and Powell, 2002), and melt extraction events may occur in an interrupted or continuous process (Sawyer, 1996). This open-system behavior, combined with the uncertainty concerning the proportion and composition of melt lost, makes recovering a bulk composition of the protolith very unlikely (Palin et al., 2013). However, White and Powell (2002) showed that various melt-loss scenarios do not change the major subsolidus and suprasolidus topologies or compositions of peritectic phases. It is therefore possible to calculate an approximate bulk composition suitable for prograde modeling. To reduce uncertainty, the melt loss was modeled as a single extraction event, representing a simple end-member case. A melt-reintegrated bulk composition was therefore calculated by the addition of a proportion of melt to the observed bulk composition to displace the solidus down-temperature to H₂O-saturated conditions, maximizing the amount of mica present prior to melting following the method of White et al. (2004). The composition of reintegrated melt was calculated at 7.75 kbar and 680 °C by using the solidus of the original bulk composition in the presence of kyanite. Although higher pressure than final melt crystallization, this P-T value was chosen because it accounts for the presence of kyanite-bearing migmatites and inclusions in secondary muscovite at higher structural levels (TL15), allowing for a clockwise P-T evolution. It was determined that 10% melt should be added to make the solidus water-saturated.

Contours for modal proportion of silicate melt are presented in the Data Repository material (see ^{footnote 1}), and they show that melt production across the wet solidus in garnet-bearing assemblages is negligible (∼1%). It is not until fluid-absent muscovite dehydration occurs that melt generation reaches significant volumes, increasing from 4% to 12% over a very small temperature interval. Contours for calculated modal proportion of garnet allow interpretation of probable conditions of garnet core growth with respect to the melting history. Although garnet cores lack prograde compositional zoning, making the early stages of the P-T evolution poorly constrained, certain inferences can be made. The presence of muscovite inclusions within garnet suggests garnet nucleated prior to the muscovite dehydration melting reaction in a muscovite-present field. The presence of kyanite relicts within secondary muscovite suggests that garnet grew when kyanite was stable. A possible interpretation involves a P-T evolution along an initial continental geotherm of 25–30 °C km⁻¹. This allows initial garnet growth to occur at ∼600 °C and 6 kbar in an assemblage containing Bt–Ms–Pl–Ilm–Qtz–H₂O. Contours for modal proportion of garnet reveal that growth in muscovite-bearing assemblages was minor, and significant growth of garnet was delayed until K-feldspar was produced as a result of muscovite dehydration along the effective solidus where the proportion of garnet increases significantly with increasing temperature (see Data Repository material [^{footnote 1}]). However, due to the presence of only ∼1.5% garnet, we conclude temperatures did not increase significantly upon passing this melting reaction, and thus it is likely that crossing the effective solidus was largely related to decompression at peak metamorphic conditions.

Crystallization of the final melt fraction upon crossing the solidus is thought to mark the final retrograde textural evolution of an anatectic rock, if no further episodes of deformation or fluid influx occur (e.g., White and Powell, 2002; Indares et al., 2008; Palin et al., 2013). Nonetheless, postpeak compositional changes may still occur via intracrystalline diffusion, as seen in garnet porphyroblast outer rims in contact with matrix biotite, which have grossular and pyrope contents of ∼6% and 15%, respectively. These are lower than the concentrations recorded in core regions, consistent with equilibration at lower temperature and pressure (Spear, 1993). Isopleths for the calculated modal proportions of garnet and melt are presented in Figure 19, and they show that both phases are produced at higher temperature due to the breakdown of muscovite. However, retrograde decompression from our calculated peak P-T conditions to the solidus would have involved a reduction in the modal proportion of garnet to form biotite and sillimanite, consistent with observed textures. Taking these factors into consideration reveals that the Naxos migmatites remained at suprasolidus conditions during deep crustal exhumation and crystallized through the water-saturated solidus at ∼4.9 kbar and 690 °C.

The P-T data clearly demonstrate that two distinct and contrasting clockwise P-T loops are recorded within the Naxos core complex (Figs. 16–19). At high structural levels, the Zas Unit records M₁ conditions of ∼14.5 kbar and 470 °C followed only by retrograde greenschist-facies conditions at <450 °C. In contrast, at deeper structural levels (Koronos and Core Units), (M₁) high-pressure–low-temperature conditions are not recorded, and instead the rocks experienced a clockwise P-T path involving (M₂) prograde burial and heating along a Barrovian metamorphic thermal gradient of ∼25 °C km⁻¹ to peak (M₂) P-T conditions of ∼10 kbar and 600–730 °C (TL15, TLN22, TL67). At the deepest levels (Core Unit), the rocks crossed the water-saturated solidus and then experienced hydrous (M₂) kyanite-grade melting (Eq. 6; TL15). They then recorded near-isothermal decompression to (M₃) sillimanite-grade conditions of 5–6 kbar and 700–730 °C (TL66, TLN34, TLN20A). During this process, the hottest and deepest rocks crossed the muscovite dehydration melting reaction to produce larger melt volumes and peritectic K-feldspar. Textural evidence suggests that most melt remained in communication with the rock, and therefore crystallization through the solidus at ∼5 kbar and 690 °C released fluids, causing metasomatism and growth of secondary muscovite crosscutting the earlier extensional fabric (S₃). Juxtaposition of the hot migmatites against the overlying Koronos Unit caused a thermal (M₃) sillimanite-grade overprint and posttectonic garnet growth that truncated the S₃ top-to-the-shear fabrics and earlier (M₂) kyanite-grade assemblage (e.g., TLN13, TLN21, TLN22). The rocks then experienced a kink in the P-T path associated with cooling at <5 kbar pressure, causing resorption of garnet and cooling along a thermal gradient of ∼40 °C km⁻¹.

Although balanced and restored cross sections cannot be applied to intensely deformed rocks under ductile conditions, we attempted a palinspastic restoration of the metasedimentary and basement units to place minimum constraints on crustal shortening. A simple restoration of the Naxos MCC was performed to estimate minimum amounts of shortening during the leadup to peak M₃ conditions and postmetamorphic doming, by considering all structural, petrographic, pressure-temperature-deformation (P-T-D), and U-Th-Pb data presented herein. U-Th-Pb results will be presented in a future study. A simple two-dimensional area- and line-balance restoration along section A–A′ provided insight into the way in which the MCC formed, and how nappes of differing lithologies attained different P-T conditions at different times and were juxtaposed against each other. The restoration assumes that, in general, deformation propagated down-structural section with time, in agreement with structural, petrographic, and P-T-D data. The restoration process involved the following steps:

• (1) The late-stage E-W doming and normal faults were restored prior to pre-Miocene extension and granodiorite intrusion using the Galanado normal fault as a pinning point (pinning point 5; Fig. 20, stage 1).

• (2) Three units (Zas Unit, Koronos Unit, and Core Unit) were identified within the metamorphic footwall and treated separately; each unit is separated by a major extensional shear zone that cannot be restored due to lack of suitable pinning points (see text for more discussion).

• (3) Zas Unit (Cycladic Blueschists) was assumed to behave in a brittle-ductile manner, dominantly by thrusting, and was restored first (Fig. 20, stage 2).

• (4) Koronos Unit thrusts were restored to align the F₁ and F₂ folds. These folds were then restored using area and line-length balancing and assuming there are only three bands of marble structurally repeated. Basement-cover repetitions were assumed to represent thick-skinned thrusts and were restored to their original geometry.

• (5) The Core Unit deformation was restored in two steps. The KSZ was restored by line-length balancing of the imbricate stack that utilized basement material as a relict thick-skinned thrust. Ductile deformation and isoclinal folding were restored using line-length and area balancing as above (Fig. 20, stage 2). This represents a minimum amount of shortening, because the thick-skinned thrusts structurally repeating the basement over shelf are laterally discontinuous, overprinted, and reworked by upright folding and ductile flow.

• (6) Finally, a restoration of the continental margin was attempted in stage 3, based on structural mapping in this study. This suggests that Naxos records a thrusted and thickened continental margin sequence in a similar style to the Helvetic Nappes in the Alps and Greater Himalayan Series (Fig. 20, stage 3). The results show that the restored continental margin was intruded by numerous basaltic and doleritic sills, with intercalated tuffaceous, siliciclastic, and carbonate material.

An ∼13-km-wide section today restores to a minimum of 60 km. Approximately 400% ENE-WSW shortening across the Naxos MCC is required to explain the folding and thickening during its prograde evolution. Although accurate section balancing is precluded in terrains with ductile deformation such as the Naxos MCC, an absolute minimum shortening of 47 km cannot be rectified. Although several assumptions were applied that may be oversimplistic and violated in ductile environments, this exercise illustrates that the folding of high-grade rocks forming the core complex could not have formed by purely extensional mechanisms. This magnitude of shortening within a continental margin sequence is commonly seen across many recent mountain ranges that experienced shortening and underthrusting of rocks to kyanite-grade conditions over time scales of tens of millions of years, such as the Himalaya (Corfield and Searle, 2000) and accretionary wedges such as those in Oman (Searle et al., 2004). Therefore, we believe that most of the structures (F₁ and F₂ folding) and prograde and peak (M₂) Barrovian conditions of the Naxos MCC occurred under a compressional tectonic regime, and not under crustal extension. Upright (F₃) folding within migmatites occurred at M₃ sillimanite-grade conditions and invokes a horizontal contractional stress regime. Whatever the cause for this constriction, it highlights a significant space problem requiring lateral displacement of material from the sides, and therefore it must have formed when the maximum principal stresses (σ₁) were orientated in a horizontal plane. Deformation as a result of this contraction must have become localized within the rheologically weaker migmatites upon experiencing muscovite dehydration melting, causing them to intensely fold, thicken, and dome the overlying units.

A significant component of E-W shortening is necessary to form the structures on Naxos, and it is suggested that some E-W shortening is occurring throughout the Aegean today (McClusky et al., 2000), accommodated on active conjugate ENE-WSW– and WNW-ESE–trending strike-slip faults like those that crosscut Naxos (see Fig. 2). This could be possibly related to a 20° counterclockwise rotation of Naxos about a vertical axis since the mid-Miocene (Kissel and Laj, 1988; Avigad et al., 1998; Malandri et al., 2016), and it could explain the late-stage gentle folding of the NPDS, the sedimentary units in the upper Cycladic Nappe, and the metamorphic sequence. Although the origin of Miocene E-W shortening remains unclear, there are several possible mechanisms that are discussed below.

Due to extreme attenuation, missing metamorphic stratigraphy, and many unknowns, such as the original dip of the NPDS, restoration of the extensional features and therefore the full amount of extension cannot be accurately determined. However, NNE-SSW crustal extension is the last phase of deformation, and it is pervasive, overprinting and truncating all shortening structures. It is frozen in the metamorphic stratigraphy and the migmatite dome. Assuming the doming occurred at ∼5 kbar (17.5 km depth), and the NPDS maintained its original dip of ∼10° to the north and was responsible for the entire exhumation of Naxos, the throw on this single fault would be >100 km. This fault must have been active in its present-day shallow geometry and could not have been rotated from steeper to shallower angles, as the migmatite dome and metamorphic footwall cannot have been rotated by over 40°. It seems likely, however, that the NPDS was not responsible for the entire exhumation on Naxos, but rather represents the brittle-ductile transition, mechanically uncoupling the upper- and midcrust, and it was exhumed in its present-day geometry by uplift on several younger crosscutting brittle normal faults. The simplistic relationships and restoration for the geometry of the Naxos-Paros detachment and earlier peak M₂ and M₁ geometries based on field relations and P-T-D data are presented in Figure 21.

The field, petrological, and P-T-D data from this study imply a complicated tectonic evolution that cannot be explained by extension alone, with juxtaposition of several nappes that experienced differing metamorphic histories, as summarized by Figure 22. Previous models of Naxos, whereby all rocks experienced M₁ high-pressure conditions followed by isobaric heating and Barrovian metamorphism during crustal extension (e.g., Buick and Holland, 1989; Buick, 1991a, 1991b; Urai et al., 1990; Avigad, 1998; Vanderhaeghe, 2004; Kruckenberg et al., 2010), cannot explain the clockwise Barrovian P-T-D-time (t) paths, which involve prograde increasing pressure and temperature in the Koronos and Core Units.

The rocks of southeast Naxos (Zas Unit or Cycladic Blueschists) underwent a completely different metamorphic evolution than the underlying higher-grade units. Contrary to previous suggestions that the whole island experienced (M₁) high-pressure conditions (Avigad, 1998; Martin et al., 2006), we find no evidence in the petrography or geochronology to suggest that either the Koronos or the Core Units experienced this event. Although it is poorly understood how meta-peridotites occur as enclaves within meta-gabbro (amphibolites) and are associated with both the Koronos and Zas shear zones (Katzir et al., 1999, 2002), several mechanisms have been proposed: (1) Katzir et al. (1999) suggested that the ultramafics were emplaced during Eocene subduction of the continental margin. (2) They could have also been emplaced by thrusting of lithospheric mantle and meta-gabbros into the Naxos shelf carbonate sequence. (3) They could have been doleritic or gabbroic sills that intruded the continental margin prior to deformation, and the ultramafic component represents cumulates that formed during fractional crystallization. (4) They could be slivers of serpentinite that were incorporated from a structurally higher ophiolitic thrust sheet, although this requires over several kilometers of structural movement. To explain the thermometry data of Katzir et al. (2002), minor compositional layering within the host amphibolites, and the absence of high-pressure relicts (see 17TL75 in the Data Repository material [^{footnote 1}]), we suggest they represent doleritic-gabbroic sills that intruded the continental margin prior to deformation. This is because the host garnet-bearing meta-gabbros show no evidence of a high-pressure history, and this is the most likely rock composition to preserve evidence for blueschist- or eclogite-facies metamorphism (Heinrich, 1982). Furthermore, the mantle wedge in a subduction setting is cold, and this disagrees with two-pyroxene thermometry temperatures exceeding 1000 °C (Katzir et al., 2002); it is more likely to be associated with hot intrusion. If the Koronos and Core Units did experience M₁ conditions, they must have been reincorporated into another prograde burial Barrovian tectono-metamorphic cycle that obliterated all trace of this previous metamorphic event.

Eocene zircon ages of ca. 40 Ma have been reported from amphibolites from northern Naxos and within the Koronos Unit (Martin et al., 2006; Bolhar et al., 2017), and these could represent two possibilities: (1) The ages represent zircon growth during (M₁) high-pressure–low-temperature conditions but not in the presence of garnet (Martin et al., 2006), and the unit was extruded to the midcrust before being reburied and incorporated as part of a separate metamorphic cycle. In this scenario, prograde and peak M₂ conditions would occur ca. 25 Ma after experiencing high-pressure conditions, like the Lepontine metamorphism in the central Alps (e.g., Wiederkehr et al., 2008). (2) Alternatively, the ages do not represent M₁ conditions, which occurred on Naxos between 49 and 46 Ma and regionally at 54–45 Ma (Tomaschek et al., 2003; Lagos et al., 2007; Dragovic et al., 2012). In the absence of high-pressure relicts, the ca. 40 Ma zircon ages could be interpreted to represent the earliest M₂ Barrovian-type age due to overthrusting of the blueschists (Zas Unit). The latter of these two suggestions is our preferred choice and agrees with the petrological record on Naxos and new U-Pb data that will be presented in a future study.

To account for contrasting P-T-D paths recorded at different structural levels on Naxos, our model features subduction of the distal continental margin (Zas Unit) down a NE-dipping subduction zone to (M₁) high-pressure–low-temperature conditions that reached blueschist-facies conditions on Naxos, but also eclogite-facies conditions as seen on the islands of Syros, Tinos, and Sifnos. This was followed by SW expulsion of high-pressure rocks under a passive-roof normal fault (Vari detachment on Syros; Roche et al., 2016; Laurent et al., 2016) at the top of the subduction channel and a thrust at the base due to the positive buoyancy of continental margin material (England and Holland, 1979). Although this thrust is not convincingly exposed on Naxos due to extensional reworking, there is a strong record of thrust-related fabrics (S_2a and S_2b), and it is possible that the South Cycladic thrust on Ios (Huet et al., 2009; Mizera and Behrmann, 2015) could be this structure or a related structure. This would produce S₁ top-to-the-NE shear fabrics in the Zas Unit and the rest of the Cycladic Blueschists and would adequately explain S₂ crenulation cleavages and thrust fabrics under M₂ greenschist-facies conditions. Overthrusting of this high-pressure unit onto the more-proximal Cycladic continental margin, along with continued crustal shortening, would have led to the generation of a nappe pile (Dixon and Robertson, 1984). The proximal sedimentary cover of the continental margin (Koronos Unit) would have subsequently experienced burial, polyphase folding, and thick-skinned thrusting, as exemplified by structural repetition of basement material overlying metasedimentary protoliths during M₂ kyanite-grade conditions. This would have led to crustal thickening and (M₂) regional metamorphism during the Oligocene to Miocene (40–15 Ma). A crustal shortening mechanism would explain isoclinal F₁ folding during prograde (M₂) Barrovian metamorphism and development of the (S_2a) crenulation cleavages (S_2b) and top-to-the-SW thrust fabrics under amphibolite-facies conditions, preserved in domains unaffected by late crosscutting S₃ extensional fabrics. Burial of the continental margin to (M₂) kyanite-grade pressures of ∼10 kbar (∼35 km depth) was associated with the development of these shortening structures. Growth of garnet porphyroblasts occurred syn- to posttectonically with localized (S_2b) thrust fabrics along cleavage domains at the base of the Koronos Unit, indicating a synchronicity between peak (M₂) kyanite-grade conditions and thick-skinned thrusting.

Contrasting peak P-T conditions between the Koronos and Core Units indicate these packages equilibrated along an evolving geotherm at slightly different times (U-Pb data will be provided in a future study). The most likely explanation in accordance with field evidence is that thick-skinned thrusting was responsible for burying the Core Unit under the Koronos Unit. Repetition of basement rocks interleaved with metasedimentary schists and marbles along the base of the Koronos Unit and within the Core Unit is consistent with this idea. Further evidence for a southwestward-verging thrust also occurs on the island of Ios as the South Cycladic thrust, which places a thin package of amphibolite-grade metasedimentary cover over Cycladic basement (Huet et al., 2009; Mizera and Behrmann, 2015).

Following thick-skinned thrusting, conductive relaxation of the footwall geotherm would have heated and driven partial melting over a time span of several million years (England 1978; England and Thompson, 1984). Footwall rocks first underwent prograde hydrous melting in the kyanite field, producing kyanite- and garnet-bearing microstructural domains and leucosomes (Eq. 6). Upon experiencing melt weakening, the rocks then record decompression through the muscovite dehydration melting reaction to sillimanite-grade conditions. This decompression was associated with non-coaxial (S₃) top-to-the-NNE shearing, flattening, and the development of F₂ sheath folds on the Koronos and Zas shear zones. This reaction is the first major melt-producing reaction and would allow deformation to become localized along weakened zones of melt segregation and facilitate the development of upright folding. Kyanite-grade fabrics are folded with this foliation, placing tight constraints on the timing of upright (F₃) folding after (M₂) kyanite-grade conditions and following decompression to (M₃) sillimanite-grade conditions.

Several models for isothermal decompression and migmatite dome formation have been suggested, but many fail to explain all structural, petrological, and P-T-D constraints. In a subsequent study, it will be shown that decompression of migmatites from ∼10 to 5 kbar occurred in less than 2.5 m.y., a time scale much shorter than previously thought. From the structural and petrological features described in this paper, a relative chronology and framework for the sequence of deformation events leading to doming can be firmly established, and therefore the various dome forming mechanisms can be assessed.

Crosscutting relationships on the KSZ bounding the migmatites demonstrate that (S₃) top-to-the-NNE shearing predated doming and leucogranite intrusion. This is because the shear zones are folded around the dome with consistent top-to-the-NNE kinematics on all margins. This normal-sense shearing must have been associated with SW-directed exhumation of the footwall Core Unit relative to the overlying Koronos Unit. S-type leucogranite dikes and sills clearly postdate shearing, because they crosscut the (S₃) mylonite fabrics on the KSZ (refer to Figs. 7–8; see also Data Repository material [^{footnote 1}]), and they show emanating geometries from the migmatite core (see Figs. 6O and 8H). The crosscutting dikes dip at ∼45° toward the center of the dome and intrude perpendicular to the S₃ foliation, particularly on the eastern margin. Because these dikes and the shearing foliation are tilted around the dome, we infer that both the dikes and shear zones have experienced rotations of ∼45° about a NNE-SSW horizontal axis from their original subhorizontal orientation. The leucogranites dikes probably intruded vertically when the KSZ and migmatites where subhorizontal and were passively rotated as the dome formed. Because leucogranites are the product of melt extraction, and this only occurs upon reaching the critical melt threshold (7%–10%; Rosenberg and Handy, 2005), this volume of melting only occurred upon decompression through the muscovite dehydration melting reaction (Eq. 4), at lower-pressure (M₃) sillimanite-grade conditions, also coeval with the timing of upright folding. Upright F₃ folds in the CHSZ have three fold axis populations that intersect around a vertical point, representing a type-2 Ramsey fold interference pattern. Vertical boudinage of folds and leucogranite dikes (see Figs. 8–10) suggests overall horizontal constriction while also at M₃ sillimanite-grade conditions. The intense upright E-W–trending and N-S–trending isoclinal folds in the center of the dome require that the maximum principal stresses σ₁ and σ₂ must have been almost equal and orientated E-W and N-S in the horizontal planes, whereas the minimum principal stress, σ₃, must have been orientated vertically. This horizontal contraction scenario can be easily explained by overall shortening of the crust, and although diapirism would predict similar features (Ramberg, 1980), the low overall melt fractions in migmatites (see Fig. 19) are inconsistent with diapirism being the dominant doming process. Similar features have been documented in Archean granite-greenstone terranes, such as the Pilbara North Pole Dome in Australia (e.g., Nijman et al., 2017). Diapirism or crustal turnover has been proposed as a mechanism to explain these tonalite-trondhjemite-granodiorite (TTG)–cored domes (e.g., Collins et al., 1998), which are associated with larger melt fractions than Naxos migmatites. The pattern of doming and basins with other observed shortening phenomena indicates the importance of regional compression during the formation of these domes, which may have caused a first Raleigh instability, leading to an egg-box pattern of sinking greenstones around the TTG domes (Nijman et al., 2017). N-S horizontal boudinage, and a range of brittle deformation features overprinting these folds, suggests the principal stress axes rotated to a NNE-SSW–orientated σ₃ and E-W–oriented σ₁ direction shortly after these folds formed.

Migmatite doming must have therefore occurred as a late feature at low-pressure (M₃) conditions (∼5–6 kbar and 700–730 °C) after top-to-the-NNE shearing on the KSZ, and upon crossing the muscovite dehydration melting reaction. Although lower-crustal isostasy-driven flow during crustal extension has been suggested as a mechanism to produce such upright contractional features at the center of migmatite domes (Kruckenberg et al., 2011; Rey et al., 2011, 2017), the predicted sequence of deformation is inconsistent with our observations. Isostasy-driven flow predicts that decompression is caused by convergent flow of a partially molten lower crust under a horizontally extending upper crust. In this scenario, we would expect high overall melt fractions and doming to occur simultaneously with shearing and opposing and divergent shear senses on all margins of the dome due to a mechanical decoupling of the migmatites with the overlying rocks. Instead, we suggest the doming, upright folding, vertical boudinage, and constriction are late features that formed under compression, and they postdate decompression and top-to-the-NNE shearing from peak M₂ pressures. The migmatite dome folds the isograds, intruding S-type granite dikes and the entire overlying metamorphic sequence, including the top-to-the-NNE shear zones (KSZ and ZSZ), and it was therefore not mechanically decoupled from the overlying rocks at the time of its formation. Doming therefore represents a separate process to the decompression of rocks from kyanite- to sillimanite-grade conditions. To explain the high-temperature N-S boudinage overprinting the upright folds and vertical boudinage, we suggest that early stages of doming resulted from constrictional strain localization in the midcrust associated with the rheological weakening that occurred upon crossing the muscovite dehydration melting reaction, whereas the latter stages of doming coincided with a change in boundary conditions, most notably, a decrease in the N-S principal stress, presumably associated the transition from overall compressional to extensional tectonics (see Fig. 23).

Isothermal decompression of migmatites and gneisses from 10 to 5 kbar (∼35–17.5 km depth) was rapid and associated with S₃ top-to-the-NNE shearing on the normal-sense Koronos and Zas shear zones. If this normal-sense shearing occurred synchronously with SW-directed thrusting within or under the migmatites, it would result in synorogenic ductile extrusion in an overall compressional setting (Law et al., 2006; Searle et al., 2006, 2010; Searle, 2010). Ductile extrusion can explain the decompression P-T path and extensional fabrics recoded in several other Barrovian metamorphic terranes in compression, including the Greater Himalayan Series, the Lepontine dome, and the Tauern Window in the Alps. There is an abundance of evidence for SSW-verging thick-skinned thrusting and compression across Naxos, and it seems likely that more thick-skinned thrusts are preserved at depth under the migmatites.

We propose that normal-sense shearing on the Koronos and Zas shear zones occurred due to extrusion in an overall crustal thickening scenario based on five key reasons: (1) If the crust was extending, we would not expect horizontal constrictional strains associated with doming. NNE-SSW crustal extension clearly produced NNE-SSW pure shear boudinage under M₃ conditions, which postdated upright F₃ folding in the migmatite dome. Because F₃ folding predated NNE-SSW extension but postdated movement on the KSZ, extrusion and top-to-the-NNE shearing on the KSZ and ZSZ must also have predated extension. (2) The KSZ and ZSZ record only top-to-the-NNE kinematics on all margins of the dome and were therefore active prior to doming. They are not multidivergent shear zones as predicted by extensional dome models (Kruckenberg et al., 2010, 2011; Rey et al., 2011, 2017). (3) The KSZ and ZSZ are completely discordant to the NPDS at a higher structural level, which truncates them (see Figs. 3 and 4), and the S₃ foliation associated with the Zas and Koronos shear zones is refolded by F₃ folds, which in turn are cut by the NPDS. Therefore, these shear zones formed earlier than and are separate structures from the NPDS, which exhibits both brittle and ductile deformation features. (4) If movement on the KSZ represents the deep lower-crustal signature of a detachment fault associated with crustal extension, then we would expect rapid cooling and migration of the brittle-ductile transition (England and Jackson, 1987). There is no evidence to suggest this occurred. Instead, isothermal decompression of migmatites to ∼5 kbar led to cooling by <30 °C upon their exhumation to the midcrust. This isothermal decompression P-T path would be predicted if the KSZ and ZSZ represent stretching faults or ductile shear zones that progressed up structural section with time (Means, 1989). (5) Hydrous melting is the major anatectic process for the Naxos migmatites that was responsible for initial melting and weakening at M₂ kyanite-grade conditions. This hydration reaction acts as a large sink for water. To balance the net water budget of the deep crust, calculations require a large external supply of water. A potential water source that could drive the partial melting process could be the result of prograde dehydration reactions from underthrusted crust. Underthrusted rocks would have been experiencing prograde metamorphism and dehydration reactions at the time of migmatization in the overlying units. This could alleviate this conundrum because liberated water could migrate into the overlying migmatites and act as a source to drive anatexis. In contrast, during crustal extension, underthrusting of hydrated rocks would not occur, and we would expect dry granulite-facies rocks to underlie Naxos migmatites, which would be strong and less likely to flow or act as a source of water to drive anatexis.

If indeed the Koronos and Zas shear zones represent passive-roof stretching normal faults in an overall compressive and E-W horizontal constrictional regime, these faults would form at the top of the SW-extruding wedge, as material was returned via ductile flow toward the foreland, and by possible erosion due to melt weakening and the buoyancy and viscosity contrasts of migmatites. Together, these shear zones preserve a sequence of right-way-up metamorphic isograds, similar to the South Tibetan Detachment System (STDS) in the Himalaya, by a process of general shear (simple shear plus pure shear flattening) that telescoped the isograds (Jessup et al., 2006; Cottle et al., 2007; Law et al., 2006; Searle et al., 2006, 2010). Top-to-NNE shearing (S₃; Fig. 23, stage 1) first developed along the rheological boundary marking the solidus. Deformation temperatures in the KSZ exceeded 650 °C, and it was therefore responsible for the earliest and deepest stages of exhumation. With subsequent ductile flow, sheath folds developed under general shear due to high non-coaxial shear strain followed by E-W constriction (Fig. 23, stage 2).

Emplacement of hot migmatites at midcrustal levels would have provided the thermal perturbation in the hanging wall of the KSZ to nucleate posttectonic garnets that crosscut S₂ kyanite-grade and S₃ top-to-the-NNE shearing fabrics and M₃ sillimanite-biotite intergrowths with strong annealing textures (Fig. 23, stage 2). This thermal pulse would also explain the intrusion of S-type leucogranites into the Koronos Unit that crosscut the S₃ blastomylonite fabric. All structures were then gently domed about a N-S axis that resulted in the frozen bull’s-eye isograd pattern.

Following extrusion of migmatites to shallower crustal levels and F₃ upright folding, initiation of NNE-SSW regional extension produced retrograde greenschist-facies, top-to-the-NNE (S₃) mylonite fabrics associated with the NPDS, which truncated, reactivated, and utilized subparallel (S_2a) crenulation cleavages and top-to-the-SW (S_2b) fabrics. NNE-SSW boudinage of amphibolites in the CHSZ indicate a rapid switch from crustal thickening to NNE-SSW regional extension during the closure of the M₃ event.

Regional extension subsequent to 15 Ma (Fig. 23, stage 3; a future paper will present the U-Th-Pb ages) was accommodated by normal faulting and horizontal NNE-SSW–directed pure shear ductile boudinage. In the upper crust, normal faults sole into and crosscut a low-angle detachment horizon (NPDS). This structure, which is discordant to, and truncates, the folding, migmatite dome, and metamorphic stratigraphy within the core complex, caused extreme telescoping of previously frozen-in isograds beneath it, and rapid footwall cooling at rates of 60–90 °C m.y.⁻¹ along a peak thermal gradient of 40 °C km⁻¹. The onset of extension is recorded by the kink in the P-T path while the migmatites were at the close of the M₃ event at ∼5 kbar and 690 °C (Fig. 22). The Naxos I-type granodiorite intruded into the metamorphic sequence at ca. 12.2 Ma, and it is cut by the NPDS, with brittle-ductile deformation features including pseudotachylytes, suggesting the arrival of the core complex at the brittle-ductile transition at this time. In contrast, in the ductile crust, extensional strain during the latter stages of the sillimanite-grade event was accommodated by mainly NNE-SSW–directed pure shear with extensional strains of up to ∼50%, resulting in horizontal boudinage and stretching that overprinted earlier compressional features (e.g., Figs. 9 and 10) and that formed the elliptical dome structure.

This study and several others highlight a significant component of E-W shortening on Naxos (e.g., Virgo et al., 2018). This shortening occurred during migmatite dome formation and F₃ folding and occurred prior to and synchronous with regional NNE-SSW extension. All structures, including the NPDS, aplite sills within the 12.2 Ma granodiorite, and even Pliocene–Pleistocene sediments, are folded or tilted around the island, which cannot be explained by isostatic footwall rebound alone. It is unclear what the origin of this shortening was during the Miocene, as it would have predated movement on the North Anatolian fault, which commenced at ca. 11 Ma (e.g., Menant et al., 2013). Convergent E-W lower-crustal flow during NNE-SSW extension (Kruckenberg et al., 2011) could possibly explain E-W shortening in the middle of the migmatite dome, but this should have been simultaneous with NNE-SSW extensional boudinage and would exclusively affect migmatites, in disagreement with our observations of F₃ folding also affecting the entire carapace. Furthermore, many islands in the Cyclades document a significant component of E-W shortening (e.g., Virgo et al., 2018, and references therein), indicating it was a regional compressive stress and not related to ductile convergent flow of migmatites under crustal extension (Kruckenberg et al., 2011). It is possible that Miocene E-W shortening was related to the westward escape of Anatolia into the Cyclades (e.g., Phillipon et al., 2014). This would have been accommodated via dextral strike-slip reactivation of the eastward Vardar suture zone to the north of the Cyclades, and possibly related to strike-slip movement on the Mid-Cycladic Lineament (e.g., Malandri et al., 2016), prior to rollback on the Hellenic subduction zone. This scenario could produce horizontal constriction-type conditions due to shortening in all directions, like those recorded in the Naxos migmatite dome. This would have been followed by NNE-SSW extension following a change in boundary stresses, either due to the onset of slab rollback on the Hellenic subduction zone or removal of lithospheric mantle during the mid-Miocene. Doming of Pleistocene sediments on the margins of Naxos is probably related to a component of recent E-W transpression, which possibly resulted in a 20° counterclockwise rotation about a vertical axis (Kissel and Laj, 1988) and the series of conjugate NNE-SSW–trending and NNW-SSE–trending strike-slip faults that crosscut the island.

Naxos records the entire evolution of an orogeny, from ophiolite obduction to postorogenic collapse, and it displays similarities to many well-studied geological areas, including the following:

• (1) Oman/United Arab Emirates: The relatively unmetamorphosed hanging wall of the NPDS displays an ophiolitic mélange similar to the Haybi oceanic unit, directly underlying the sole of the Oman/United Arab Emirates ophiolite (Searle and Cox, 2002). This unit is interpreted to represent off-axis alkali seamounts with carbonate caps that formed on the lower plate and were scraped off into a mélange during the subduction/obduction process. Oman also records blueschist- and eclogite-facies rocks that are bounded by normal-sense extensional shear zones within the continental margin sequence under the ophiolite. These formed ∼15 m.y. after the initiation of obduction as the leading edge of the Arabian continental margin attempted to subduct. A comparable ophiolite obduction scenario could also have taken place in the Cyclades prior to high-pressure metamorphism of the Cycladic Blueschists, which also represent the leading edge of the Cycladic continental margin.

• (2) The Tauern Window, Austrian Alps: The metamorphic footwall of the Naxos MCC records Barrovian metamorphism due to the stacking of various tectonic slices. Many comparisons can be drawn to the Tauern Window in the Austrian Alps, one of the most intensely studied overthrust terranes on the planet (e.g., Schmid et al., 2013; Smye et al., 2011, and references therein). It is composed of an amphibolite-facies metamorphic core that underlies discrete slices of eclogite- and blueschist-facies thrust sheets. Barrovian metamorphism was a result of overthrusting Austroalpine nappes derived from the Adriatic continental margin (upper plate), in some respects like the Upper Cycladic Nappe (Jolivet et al., 2013) and Cycladic Blueschists.

• (3) The Greater Himalayan Series (GHS) in the Himalaya: The metamorphic core of the Himalaya records an almost identical P-T-D path to that observed on Naxos, with kyanite-grade conditions followed by isothermal decompression through the muscovite dehydration melting reaction to lower-pressure sillimanite-grade conditions and S-type leucogranite formation. Decompression occurred during SW-directed extrusion of the GHS in an overall compressional setting, bounded by the Main Central thrust at the base and the South Tibetan Detachment System at the top, which were active at the same time. The extensional top-to-the-NE shear fabrics on the STDS formed due to exhumation of footwall material beneath a fixed hanging wall in an overall compressional setting, and they preserve a sequence of right-way-up metamorphic isograds that are strikingly similar to the Koronos and Zas shear zones on Naxos. Naxos has gone one stage further than the Himalaya and represents a collapsed orogen that beautifully documents the transition from an overall compressional to extensional environment that coincided with the latter stages of migmatite dome formation.

• (4) Betic-Rif, Spain: This structure records the transition from overall compressional to extensional tectonics. Eocene to Oligocene south-directed subduction was associated with high-pressure metamorphism of the Iberian continental margin and was followed by significant crustal thinning and coeval heating during the early stages of arc formation (e.g., Platt et al., 2013). The mechanisms that caused extension within the hinterland of this orogen are debated, but removal of lithospheric mantle seems to be the most likely scenario and would result in upwelling of the asthenosphere, causing heating, a rise in topography, crustal extension, and rapid westward rollback on the West Mediterranean subduction zone (Platt et al., 2013, and references therein). Comparable processes could have caused extension within the Aegean during the late Miocene. An important difference between Naxos and the Betics is that there is substantial kyanite-grade metamorphism and no subcontinental lithospheric mantle exhumed in the Cyclades, as the Moho lies at ∼25 km depth and deepens to the north (Cossette et al., 2016).

The tectonothermal evolution of the Naxos metamorphic core complex is more complicated than can be explained purely by crustal extension. The data presented in this study demonstrate that Naxos records a prolonged history of compression, resulting in crustal thickening and regional metamorphism, followed by extension. Eight key points can be made:

• (1) The footwall of the Naxos MCC represents three tectono-metamorphic units that each record contrasting P-T-D-t paths, and each represents a series of nappes that were stacked due to prolonged crustal thickening. During this process, rocks were buried and metamorphosed diachronously as the geothermal gradient conductively increased over time, and eventually resulted in Barrovian metamorphism and partial melting. The sequence records M₁ blueschist-facies conditions at the top to M₂–M₃ upper-amphibolite-facies migmatites at the base.

• (2) These nappes were juxtaposed against one another by top-to-the-NNE normal-sense ductile extensional shear zones that coincided with peak M₂–M₃ Barrovian conditions at the deepest levels. This resulted in telescoping of previously frozen-in isograds during extrusion of footwall rocks, similar to the normal-sense shear zone of the STDS bounding the top of the Greater Himalayan Series. Deformation migrated from deep to shallow structural levels with time and was associated with general shear. These shear zones predated doming and are discordant to, and predated, normal faulting related to crustal extension. In contrast, the low-angle normal faults truncate all structures and metamorphic stratigraphy in their footwall.

• (3) The hanging wall of the NPDS dominantly consists of trench mélange formed along the base of the obducting ophiolite. This may record further evidence for a relict ophiolite obduction event, as seen on Tinos (Lamont, 2018), which preceded M₁ high-pressure metamorphism in the Cyclades. Ophiolitic rocks are exposed at the highest structural level across the Cyclades. Miocene–Pliocene fluvial sediments and conglomerates were deposited unconformably on top of this, suggesting that topography was much higher during the Miocene.

• (4) At the deepest levels, M₂ kyanite-grade hydrous melting (Eq. 6) was the dominant melting process and was followed by isothermal decompression from ∼10 kbar and 680–750 °C to 5–6 kbar and 700–730 °C. With small degrees of partial melting, the migmatites weakened and decompressed under a passive-roof normal fault (KSZ). The hottest and most deeply buried rocks decompressed through the muscovite dehydration melting reaction (Eq. 4), locally producing up to 20%–25% melt. This advective process may have acted as substantial source of heat applicable to many Barrovian metamorphic terranes.

• (5) The migmatite dome is a late feature, and it was formed in a constrictive stress regime at lower-pressure M₃ sillimanite-grade conditions. This postdated top-to-the-NNE shearing on the Koronos and Zas shear zones and folded the entire metamorphic sequence. Structural analysis of polyphase deformation within the center of the dome indicates a major change in stress regime from NNE-SSW compression to dominantly E-W contraction and later NNE-SSW extension, coinciding with doming.

• (6) Naxos records the transition from compression to extension in the Cyclades. The formation of the sillimanite-grade migmatites marks the climax of high-grade metamorphism in the Cyclades, and no evidence for crustal thickening postdates this event. Within the dome, N-S pure shear horizontal extensional boudinage occurred during the end of the M₃ sillimanite-grade event and overprints earlier upright F₃ isoclinal folds and vertical boudinage. This corresponds to the final stage of migmatite dome formation and marks the transition from regional compression to extension in the Cyclades.

• (7) Regional extension and crustal thinning were accommodated by brittle-ductile normal faulting that crosscut and is discordant to all footwall structures. It attenuated and telescoped the frozen-in metamorphic isograds to produce metamorphic field gradients of up to 700 °C km⁻¹, particularly in west Naxos. In the ductile crust, horizontal N-S pure shear boudinage accommodated extensional strains of up to 50%. These structures postdate all earlier collisional features, and brittle-ductile normal faulting was responsible for the final stage of exhumation from 5 kbar (∼17.5 km depth) to the surface along an apparent geothermal gradient of 40 °C km⁻¹.

• (8) Three types of extensional fabrics are recorded on Naxos: (i) SW extrusion from the subduction zone, as recorded in the Zas Unit with top-to-the-NE fabrics, (ii) synorogenic SSW extrusion of migmatites and high-grade gneisses under a passive-roof fault (Koronos and Zas shear zones), with top-to-the-NNE kinematic indicators, and (iii) top-to-NNE extensional fabrics due to late brittle-ductile crosscutting normal faults related to crustal extension and cooling (NPDS). This movement was superimposed on earlier fabrics along the western side of the core complex. In particular, extensional fabrics on the Koronos and Zas shear zones within the metamorphic footwall were responsible for synorogenic extrusion of kyanite- to sillimanite-grade gneisses and migmatites that predate extension. These are truncated and crosscut by later S-type granite dikes and lower-grade fabrics and structures related to crustal extension.

Because the timing constraints and rates of deformation, metamorphism, and partial melting are crucial for resolving the complete P-T-D-t history of Naxos to a high resolution, we will provide new U-Th-Pb age data in a future paper. We also present a fully integrated tectonic model for the development and exhumation of the Naxos MCC and show that Naxos documents the entire evolution of what we term the “Aegean orogeny.” This mountain belt records a full orogenic cycle from ophiolite obduction, to high-pressure metamorphism, through subduction-related exhumation, crustal thickening, and regional metamorphism to postorogenic extension. We demonstrate that the transition from compression to extension on Naxos coincided with the latter stages of migmatite formation at ca. 15 Ma, corresponding with the timing of a major slowdown in the convergence rate between Nubia and Eurasia (DeMets et al., 2015). We build on this structural and metamorphic framework and show that crustal extension was associated with rapid cooling and brittle-ductile normal faulting and was responsible for the final stages of core complex exhumation, which postdated peak metamorphism, partial melting, and synorogenic extrusion.

This work forms part of T.N. Lamont’s doctoral project, which was funded by the Natural Environment Research Council (NERC; grant NE/L0021612/1). Analytical work at the NERC Isotope Geosciences Laboratory was funded by NERC Isotope Geosciences Facilities Steering Committee grant IP-1597-1115. We thank Philip England, Anthony Watts, Dave Wallis, Lars Hansen, and Tyler Ambrose for thought-provoking and critical discussion. We extend thanks to Callum Higgins, William Nash, and Anna Bidgood for assistance in the field, Adrian Wood for assistance with mineral separation, Owen Green and Jonathan Wells for thin section preparation, and Gren Turner, Jeremy Rushton, and Jon Wade for analytical assistance. Chris Ballhaus and Christoph Von Hagkne are thanked for thorough reviews, which greatly improved the quality of the manuscript, and Nick Timms, Chris Kirkland, and Aaron Cavosie are thanked for handling the manuscript and providing further comments. We finally thank University College, Oxford for travel bursaries covering fieldwork expenses and the University of Oxford and NERC for generously covering publication and open access costs.

MCC—metamorphic core complex.

Moutsana detachment—The brittle low-angle normal fault that was partly responsible for exhuming the island.

NCDS—North Cycladic Detachment System, a brittle-ductile low-angle normal fault and series of shear zones that display top-to-the-NE kinematics exposed on the north coast of Mykonos, Tinos, and Andros, juxtaposing I-type granites and blueschists in the footwall from relatively less metamorphosed ophiolitic rocks and sediments in the hanging wall.

NPDS—Naxos-Paros Detachment System, the brittle-ductile low-angle normal-sense extensional shear zone that cuts all previous metamorphic stratigraphy and produces telescoping of isograds on the western side of the island.

Carapace—the metasedimentary units that overlie the migmatite dome, including the Zas Unit and Koronos Unit.

CHSZ—Core high-strain zone, a zone of upright folded marbles and amphibolites that are affected by vertical and horizontal boudinage at the center of the migmatite dome, best exposed in the Boulibas and Kinadaros quarries.

Upper Cycladic Nappe—The relatively unmetamorphosed, ophiolitic and sedimentary hanging wall of the Naxos metamorphic core complex and the hanging wall of the North Cycladic detachment system on Mykonos, Tinos, and Andros.

GHS—Greater Himalayan Series.

STDS—South Tibetan Detachment System.

TTG—Tonalite-trondhjemite-granodiorite.

Zas Unit—A blueschist-facies unit that can be regionally correlated with the Cycladic Blueschist Unit, representing the distal Cycladic (Adriatic) continental margin.

Koronos Unit—A kyanite-grade amphibolite-facies unit that represents the more-proximal continental margin, forming the deeper carapace to the metamorphic core complex.

Core Unit—The kyanite- to sillimanite-grade core of the island, with migmatites, orthogneisses, and structurally repeated and uprightly folded and boudinaged marbles and amphibolites that form a doubly plunging migmatite dome.

Main ultramafic horizon—A semicontinuous band of amphibolites directly above the Koronos shear zone that wraps around the Core Unit, locally containing enclaves of ultramafic lithologies.

ZSZ—Zas shear zone, the top-to-the-NE greenschist-facies shear zone separating blueschist-facies rocks of the Zas Unit (on top) from amphibolite-facies rocks of the Koronos Unit (below).

KSZ—Koronos shear zone, the top-to-the-NE, upper-amphibolite-facies shear zone bounding the migmatite dome (Core Unit), which is folded and cut by S-type leucogranites.

M₁—High-pressure–low-temperature blueschist-facies metamorphism that locally reached eclogite-facies conditions on Syros, Tinos, and Sifnos.

M₂—Regional metamorphism that reached greenschist-facies conditions in the Cycladic Blueschists (Zas Unit) and kyanite-grade conditions in the Koronos and Core Units.

M₃—Regional metamorphism that is associated with sillimanite-grade conditions, partial melting, and S-type leucogranite intrusion at the deepest levels of Naxos in the Core Unit.

P-T-D-(t) paths—Pressure, temperature, deformation, and time paths.

F₁—First generation of folds, which show isoclinal to recumbent geometries and developed during burial and SW-directed thrusting.

F₂—Second generation of folds to form, which display sheath and isoclinal geometries that refold the F₁ folds, and that developed during top-to-the-NNE shearing, during exhumation.

F₃—Last generation of folds that formed under lower-pressure conditions. These are upright folds that trend along NNE-SSW axes across all units and are boudinaged both vertically and horizontally in the Core High-Strain Zone.