The Cycladic Blueschist Unit on Tinos, Greece: Cold NE Subduction and SW Directed Extrusion of the Cycladic Continental Margin Under the Tsiknias Ophiolite

High pressure‐low temperature (HP‐LT) metamorphic rocks structurally beneath the Tsiknias Ophiolite make up the interior of Tinos Island, Greece, but their relationship with the overlying ophiolite is poorly understood. Here, new field observations are integrated with petrological modeling of eclogite and blueschists to provide new insight into their tectonothermal evolution. Pseudomorphed lawsonite‐, garnet‐, and glaucophane‐bearing schists exposed at the highest structural levels of Tinos (Kionnia and Pyrgos Subunits) reached ~22–26 kbar and 490–520°C under water‐saturated conditions, whereas pseudomorphed lawsonite‐ and aegirine‐omphacite bearing eclogite reached ~20–23 kbar and 530–570°C. These rocks are separated from rocks at deeper structural levels (Sostis Subunit) by a top‐to‐SW thrust. The Sostis Subunit records P‐T conditions of ~18.5 kbar and 480–510°C and is overprinted by pervasive top‐to‐NE shearing that developed during exhumation from (M1) blueschist to (M2) greenschist facies conditions of ~7.3 ± 0.7 kbar and 536 ± 16°C. These P‐T‐D relationships suggest that the Cycladic Blueschist Unit represents a discrete series of tectonometamorphic subunits that each experienced different tectonic and thermal histories. These subunits were buried to variable depths and sequentially extruded toward the SW from a NE dipping subduction zone. The difference in age and P‐T conditions between the HP‐LT rocks and the overlying metamorphic sole of the Tsiknias Ophiolite suggests that this NE dipping subduction zone was active between circa 74 and 46 Ma and cooled at a minimum rate of ~1.2–1.5°C/km/Myr prior to continent‐continent collision between Eurasia and Adria/Cyclades.

HP-LT metamorphic rocks on Tinos are investigated to constrain their pressure-temperature and deformation (P-T-D) evolution through field and microstructural observations, thermobarometry and petrological modeling. The P-T-D histories are integrated into a tectonothermal model that explains the structure of the CBU on Tinos and the relationship with the overlying Tsiknias Ophiolite.

Upper Unit (Tsiknias Ophiolite and Metamorphic Sole)
The Tsiknias Ophiolite (Figure 2) in the Upper Unit represents the highest thrust sheet in the ACM, comprising a fragmented piece of Jurassic oceanic crust and upper mantle dated at 161.8 ± 2.7 Ma by U-Pb zircon from a plagiogranite within the Moho Transition Zone . The ophiolite is incomplete and tectonically dismembered, owing to intense deformation by both out-of-sequence thrusts and normal faults (Katzir et al., 1996;. A major top-to-SW thrust (The Tsiknias Thrust) crops out structurally beneath the mantle section of the Tsiknias Ophiolite. Immediately, structurally beneath this tectonic contact is a metamorphic sole (interpreted as the subducting plate) comprising anatectic amphibolites (metamorphosed basalts and gabbros) at the top of the section and pelagic metasediments, including meta-cherts and mafic phyllites at the bottom (Mirsini Unit; . The metamorphic sole displays an inverted metamorphic gradient, with P-T conditions of~8.5 kbar and >800°C at the top, and~600°C at the base . Metamorphism extrusion mechanism and argue the rocks were reheated to 500-550°C in the midcrust. However, Behr et al. (2018) obtained lower pressures for eclogite facies metamorphism of 12.5-15.5 kbar at 450-550°C, based on quartz-in-garnet inclusion barometry, questioning whether the eclogites reached such HP. On mainland Greece, blueschist facies rocks occur in Attica and Evia but did not reach as HP and probably relate to an overlying thrust package (~12-13 kbar and 360-380°C; e.g., Baziotis et al., 2020).
Thermobarometry of the CBU on Tinos suggests peak M 1 conditions of~12-20 kbar and 450-550°C (e.g., Bröcker et al., 1993;Parra et al., 2002;. A three-stage exhumation has been suggested: first, decompression from 18-15 kbar at 500°C to 9-5 kbar at 400°C, followed by a thermal overprint (400-550°C) and further decompression from 9 kbar at 550-570°C to 2 kbar at 420°C (Bröcker et al., 1993;Parra et al., 2002). The Basal Unit comprising a dolomite-phyllite-quartzite sequence has been suggested to crop out at Panormos Bay in NW Tinos. It is postulated that these rocks did not experience HP conditions due to the absence of glaucophane, distinct deformational characteristics, and preservation of undeformed fossils within the marbles (Avigad & Garfunkel, 1989); however, this interpretation is debated. Bröcker and Franz (2005), show that Rb-Sr ages and phengite compositions are indistinguishable between both units, indicating they reached HP conditions at the same time, and oxygen isotope data suggest both units equilibrated at similar temperatures (Bröcker & Franz, 2005;Matthews et al., 1999). If there was a contrast in metamorphic grade between these two units, then the contact would have been a thrust (Panormos Thrust), placing the CBU over the Basal Unit, which did not experience such HP (Avigad & Garfunkel, 1989;Bröcker & Franz, 2005). Additionally, a mélange-like outcrop pattern with blocks of eclogite and blueschist within a host metasedimentary matrix is described by Bulle et al. (2010) and Bröcker et al., 2014 for a few locations within the Lower Unit including Kionnia Peninsular and within the interior of the island. These mélange-like outcrops differ from mélanges exposed on Syros, which are bounded by a serpentine matrix, whereas on Tinos they are hosted in a more coherent metasedimentary matrix that appears to have experienced a similar deformation and metamorphic history, although we acknowledge rare, narrow and laterally discontinuous serpentinite outcrops occur particularly near Tinos Town.
It is unclear whether these semicontinuous eclogite and blueschist horizons experienced similar pressures to the rest of the Lower Unit. The classic interpretation is that entire Lower Unit (CBU) experienced the same tectonothermal history. The apparent better preservation of HP rocks at high structural levels and more extensive retrogression at deeper levels has been attributed to the increased availability of retrograde fluids with depth (Avigad & Garfunkel, 1989;Bröcker, 1990;Bröcker et al., 2014). An alternative hypothesis to explain this observation is that the Lower Unit represents a series of tectonically bound slices that experienced discrete metamorphic histories, with lower levels not being buried as deeply within the subduction-accretion system. This is the main questions we aim to address.
The tectonic regime of the area following subduction and accretion is debated. Slab roll-back of the Hellenic subduction zone since the Eocene to present day (Lister et al., 1984;Jolivet et al., 2013Jolivet et al., , 2015Jolivet & Brun, 2010) is a popular model to explain extension and exhumation of the Cyclades. Many authors regard slab rollback as having caused isobaric heating and Barrovian metamorphism (M 2 ) of previous HP-LT rocks, due to increased basal heating from the upwelling asthenosphere (e.g., Jolivet et al., 2013;Jolivet & Brun, 2010). However, a crustal thickening model has been proposed by Lamont et al. (2019) and Searle and Lamont et al. (2019) who argue for compression as a result of continent-continent collision between Cyclades/Adria and Eurasia as the cause of M 2 -M 3 kyanite-sillimanite grade metamorphism on Naxos, based on isoclinal upright folds within the Naxos core, a clockwise prograde P-T-t paths of kyanite grade gneisses and migmatites, and evidence for horizontal constriction and vertical extension within the core of Naxos migmatite dome Virgo et al., 2018;von Hagke et al., 2018). Lamont et al. (2019) suggested a switch from overall compression to extension occurred at circa 15 Ma, timing that coincides with a two-fold decrease in the Africa-Eurasia convergence rate (DeMets et al., 2015). This tectonic reversal may have occurred due to gravitational collapse of over-thickened crust, or removal of the lithospheric mantle, which may have resulted in the onset of normal faulting that caused rapid exhumation and cooling.

Field Relationships and Petrography of the CBU on Tinos
A new structural and metamorphic map and cross sections are presented in Figures 2, 3, and S1 in the supporting information and a smaller-scale map and cross sections of the Kionnia Subunit (see below) in southern Tinos is presented in Figure 3. The new mapping has integrated the previous detailed observations (Avigad & Garfunkel, 1989;Bröcker, 1990;Bröcker & Franz, 2005;Bröcker et al., 1993;Katzir et al., 1996;Melidonis, 1980) and included the marble bands (m1-m3), which are classically used as structural markers, into a new structural framework. For a more thorough summary of methods please refer to supporting information Text S1. The following sections and Figures 4-7 describe the key field and petrological observations from structurally high to structurally low, which have new implications for the evolution of HP rocks on Tinos and the Cyclades. Blueschist, eclogite, and greenschist facies samples were collected at all structural levels to investigate spatial variations in the P-T conditions of prograde, peak, and retrograde metamorphism. Representative photomicrographs are displayed in Figures 8 and 9 and a schematic illustration of deformation and mineral growth and microstructure is displayed in Figure 10. Garnet compositions and maps are presented in Figure 11, and other phases compositions in supporting information Text S1 and Table S1. For a full petrographic description of each sample, consult supporting information Text S1.

The Kionnia Subunit
Eclogite and blueschist facies rocks occur within an~500 m thick intensely folded metasedimentary sequence at Kionnia Peninsular (Bröcker & Enders, 1999) (Kionnia Subunit) on the south coast of Tinos, at the highest structural levels of the Lower Unit (Figures 2 and 3; N37.548015, E25.125559). This sequence comprises calcite marbles (m3) intercalated with metasediments and meta-basalts that form granoblastic and foliated eclogite and blueschist. At least three structural fabrics are preserved within the subunit. S 1 is only preserved as inclusions within garnet cores and is a planar fabric defined by prograde-peak M 1 inclusions of glaucophane, rutile, and lawsonite pseudomorphs. The S 1 fabric is folded by the S 2a fabric. S 2a is preserved as inclusions of lozenge shaped lawsonite pseudomorphs within garnet rims that show a 90°rotation from the S 1 inclusion orientation. S 2a also forms matrix crenulations (Figure 6a) with millimeter-centimetersized microlithons affecting peak M 1 matrix phases including glaucophane, omphacite, phengite and rutile (Figures 8g,8k,and 8n). In the lower half of the subunit, a top-to-SW shear fabric (S 2b ) overprints the S 2a fabrics (best seen at Kionnia Beach at the base of the subunit) and affects cleavage domains. S 2b is associated with S-C′ that deforms the peak M 1 and retrograde M 1b phases (including clinozoisite and titanite), indicating top-to-SW shearing developed at peak to immediately post M 1 and continued during earliest stages of exhumation (M 1b ) (Figure 8d). In the upper half of the subunit, a penetrative top-to-NE shear fabric (S 3 )       Figure 11. Prograde garnet maps and profiles for Kionnia and Pyrgos Subunit samples showing (a) TLTN26 EPMA maps with rotation textures in the calcium map, indicating garnet grew syntectonic with glaucophane in the matrix and strong spessartine zoning from core to rim, with (b) increasing grossular, pyrope, and almandine. (c) Eclogite TLTN24 garnet EPMA maps and line profiles (d) showing strong grossular and pyrope increases from core to rim with a small inflection in spessartine and almandine affecting the outer most rim. (e) Sharp prograde and symmetrical zoning in blueschist 17TL104 with increasing grossular toward the rim, interpreted as a single growth history. (f) TLTN33 garnet line profile, showing similar trends to TLTN24 but heavily affected by fractures (missing data) and a smoother profile than the blueschists.
The sense of shear therefore inverts across the subunit with top-to-SW shear (S 2b ) in the lower half, and topto-NE shear (S 3 ) in the upper half. This suggests the subunit is bounded by shear zones that were active during exhumation from the eclogite facies depths. The TSZ truncates the blueschist-eclogite facies structures at the top of the subunit. The TSZ is associated with NE-SW lineations (L 2 ; defined by actinolite, chlorite, titanite, and plagioclase) and M 2 greenschist facies (S 3 ) top-to-NE S-C ultra-mylonites (see stereonets in Figure 3) and places Upper Unit greenschists (possibly retrogressed amphibolite metamorphic sole) directly against eclogite facies rocks (N37.552088, E25.124767). Although the very top of the Kionnia Subunit or overlying HP rocks are not exposed, it is acknowledged that post M 1b (S 3 ) top-to-NE shearing should become more intense up structural section toward the roof (top-to-NE) shear zone responsible for the exhumation from subduction depths. In contrast, the lower half of the subunit is affected by (S 2b ) top-to-SW shear. These kinematics are associated with an inverted metamorphic field gradient upward from Kionnia Beach, suggesting the lower contact of the subunit represents a ductile thrust, referred to herein as the Kionnia Thrust (with top-to-SW kinematics affecting M 1 and M 1b assemblages) that separates it from the rest of the Lower Unit (Figures 3 and 5).
Although this sequence has been interpreted as a mélange (Bröcker et al., 1993;Bulle et al., 2010), we find the sequence comparable to the Chroussa Subunit on Syros and is much more coherent than the HP serpentine bounded mélanges on Syros (Kampos Subunit: e.g., Keiter et al., 2011;Laurent et al., 2018). The consistent and rhythmic layering of metasediment and eclogite ( Figure 6e) suggest the sequence records its primary sedimentary-volcaniclastic origin, with the eclogite representing boudinaged mafic intrusions or lava flows rather than fragments of oceanic crust.

Kionnia Blueschist
Garnet-glaucophane schists crop out as semi continuous horizons (TLTN26 Figure 6b) in the upper half of the subunit, structurally above a 10-m-thick band of metapelites and are affected by penetrative top-to-NE shear (S 3 ) that affects matrix M 1 -M 1b phases including glaucophane, rutile, titanite, and clinozoisite. Peak M 1 phases include glaucophane, comprising >50% rock volume, matrix phengite, idioblastic garnet with pseudomorphed lawsonite inclusions, rutile, and quartz, whereas M 1b phases are confined to the matrix and include clinozoisite, titanite, chlorite, and quartz. Metapelites also show blueschist facies paragenesis, comprising matrix quartz, phengite, paragonite, blue amphibole, chloritoid, pseudomorphed lawsonite, rutile, and garnet up to 2-5 cm in diameter ( Figure 6e). Highly retrogressed lawsonite schists crop out at the base of the subunit at Kionnia Beach, with 2 cm lawsonite pseudomorphs that are deformed by topto-SW (S 2b ) shear. Lawsonite schists are never in association with garnet suggesting a bulk rock/fluid control to lawsonite growth ( Figure 6d). In sample TLTN26, garnet X-ray maps and line profiles (Figures 11a and  11b) reveal a concentric spessartine zonation from core to inner rim, followed by a step decrease in the outer rim, suggesting a two-stage garnet growth history. Grossular contents remain at~20% across the cores, increase slightly toward the inner rim then increase abruptly to~24% in the outer rim, albeit~1% of the variation seems to reflect Ca inheritance from the relict amphibole matrix (S 1 -S 2a ) that garnets overgrow ( Figure 11a). Almandine increases from core to rim (54-62%), whereas pyrope only increases from 5-8%. These features suggest prograde garnet growth occurred under lawsonite stability and supports the interpretation that S 1 and S 2 are prograde fabrics, whereas S 3 developed during the second phase of garnet growth as it is retained as a rotation of inclusion trails in the outermost rim. Matrix amphibolies are glaucophane, with Na B ¼ 1.75-2.0 apfu, Si ¼ 7.70-8.00 apfu, (Na + K) A ¼ <0.05 apfu, and XMg ¼ 0.57-0.69. The occurrence of blueschist adjacent to centimeter-scale eclogite facies assemblages ( Figure 6b) suggests that the growth of omphacite is controlled by bulk composition, hydration, or oxidation state (e.g., Weller et al., 2015).

The Pyrgos Subunit
In NW Tinos, the Pyrgos Subunit crops out along strike and at a similar structural level to the Kionnia Subunit, on the southern limb of the Tinos dome (Figures 2 and 3). It comprises a series of imbricated 10.1029/2019TC005890 Tectonics and isoclinally folded (F 1 ) marbles (mainly m2 and m3) which are intercalated with metasediments and meta-volcanic sequences near the villages of Isternia and Pyrgos (Figures 5a, 7c, and 7i). These rocks preserve the same three structural fabrics as the Kionnia Subunit (with S 1 and S 2a only preserved as pseudomorphed lawsonite inclusions in garnet) but are more extensively overprinted by pervasive top-to-NE shear (S 3 ). Up to 12 marble horizons are laterally discontinuous along strike and bounded by (S 3 ) top-to-NE shear zones associated with S-C′ fabrics that affect peak (M 1 ) and retrograde (M 1b ) blueschist facies assemblages. Isoclinal and recumbent folding in the marbles (F 1 ) is best seen across Vathi Bay (N37.628424, E25.028671) (Figure 5a). F 1 folds deform the S 2 foliation (defined by glaucophane, phengite, and rutile), verge toward the SW and are displaced, and crosscut by several shear zones with top-to-NE kinematics (S 3 ). Because top-to-NE shear zones (S 3 ) deform around and truncate F 1 folds (Figure 4a), it suggests the F 1 folding predates (S 3 ) top-to-NE shearing. F 1 folds therefore formed on the prograde part of the P-T path. In contrast, (S 3 ) top-to-NE shearing occurred during exhumation (M 1b ), with the intensity of S 3 topto-NE fabrics increase toward the top of the subunit. A gentle spaced crenulation (S 4 ) is orthogonal to the L 1 lineation and affects M 1b -M 2 matrix phases including blue-green amphibole, epidote, and plagioclase, suggesting it developed during M 2 .

Sostis Subunit
The Sostis Subunit makes up the deepest structural levels of Tinos ( Figure 2). It comprises meta-psammites and metapelites that are intercalated with meta-volcaniclastic lithologies and several <15 m thick dolomitic marble bands (m1-m2) (Bröcker, 1990;Melidonis, 1980), which are recumbently folded on a km scale (F 1 ). Very rare, meter-scale outcrops of serpentinite occur along shear zones, particularly near Tinos Town. Unlike the overlying subunits which preserve fresh M 1 assemblages, this thick package of rocks is affected by a very strong M 2 overprint. However, rare garnet and glaucophane bearing rocks preserve some of its M 1 history (TLX and TLT60). Small occurrences of jadeitites and retrogressed eclogite have been reported on the north coast, close to the listvenites described by Hinsken et al. (2017) (M. Bröcker, personal communication, October, 2019), although were not found in this study. Several highly retrogressed mafic horizons crop out as boudinaged blocks within the metasediments (Bulle et al., 2010) but are laterally discontinuous and extensively overprinted by M 2 assemblages. It is unclear what these mafic bodies represent. They could be interpreted as (1) mafic sills that intruded the Cycladic-Adriatic continental margin prior to subduction or (2) metagabbros that represent fragments of oceanic that were entrained into the host metasediment during subduction/exhumation (Bulle et al., 2010).
The same three fabrics can be identified within Sostis Subunit samples at thin section scale (Figures 8 and  10). S 1 is only preserved within garnet cores and is defined by pseudomorphed lawsonite, rutile, and sometimes glaucophane inclusions within epidote (Figures 9 and 10). Rutile and pseudomorphed lawsonite inclusion trails in garnet cores rotate toward garnet rims (S 2a ) but are still discordant to the external (S 3 ) greenschist facies top-to-NE matrix foliation that deforms around garnet. S 2a crenulations are preserved within low-strain matrix domains represented by interlocking grains of phengite that form herringbone structures. All peak M 1 metamorphic assemblages (Grt + Gln + Ph + Lws + Chl + Rt + Qz) are prekinematic with respect to (S 3 ) top-to-NE shearing. Clinozoisite forms lath-like porphyroblasts with their long axes parallel to subparallel with the (M 1b ) S 3 foliation. Plagioclase grains are affected by S 3 producing 10.1029/2019TC005890  (1985) Krogh

Tectonics
Ravna (2000) Ed-Tr   (1985) Krogh Ravna (2000) Ed-Tr Tectonics sigma-type porphyroclasts with top-to-NE kinematics (Figures 8 and 9), although some plagioclase porphyroblasts clearly postdate shearing (Figure 9k). Plagioclase trap epidote and rutile inclusions, signifying they postdate M 1b and grew at much lower (M 2 ) pressures. Top-to-NE shearing (S 3 ) therefore developed over a range of pressures, initiating prior to plagioclase growth (M 1b ) and continued into the greenschist facies (M 2 ). The final phase of deformation developed upright crenulation cleavages (S 4 ; Figures 9j and 9m) that are associated with centimeter-millimeter-scale (F 4 ) folds trending NE-SW. These features are superimposed on S 3 and cause kinking and folding of phengites and muscovite and suggest a component of E-W shortening during exhumation (Figures 10j and 10m). Quartz grains exhibit incipient recrystallization by grain boundary migration (Figures 9k and 9l), which suggests M 2 deformation temperatures >500°C (Stipp et al., 2002).
Top-to-SW/SE sensed shear zones (see stereonets in Figure 3) crosscut the sequence and are associated with extreme grain size reduction of M 1 and M 1b assemblages and form S-C microstructures (S 2b ). These structures oppose the pervasive normal sense (S 3 ) top-to-NE kinematics and demonstrate reverse sensed offsets suggesting they are thrusts. The Sostis Thrust, at Sostis Bay (N37.534477, E25.223566), is associated with intense (S 2b ) top-to-WSW mylonitization and displaces the marble band in in its hanging-wall (Figures 7a  and 7b). A top-to-SW shear zone near Tinos Town crosscuts the subunit and juxtaposes the Akrotiri Unit (Upper Unit) against retrogressed meta-psammites and calc-schists. All these structures are truncated by the greenschist facies (M 2 ) (S 3 ) top-to-NE shear and the NE-SW L 2 lineation associated with the TSZ (see stereonet in Figure 2).
Kilometer-scale recumbent folds (F 1 ) are exposed in central Tinos on the northeast face of Mount Exomvourgo (N37.612590, E25.111204), to the SW of Kalloni, and Aetofolia villages and in east Tinos associated with the Sostis antiform ( Figure 2). These folds (Figures 7d and 7e) have an ENE vergence and have N-S striking hinge lines (see stereonet in Figure 2). F 1 folds form directly above high-strain zones associated with intense isoclinal and (F 2 ) sheath folding in marble bands. F 1 folds are interpreted as hanging-wall antiforms above ductile top-to-SW thrusts, which were responsible for thickening and structurally repeating the metasedimentary sequence. A series of dacite dykes intrude through the sequence and are related to the Tinos granite (Brichau et al., 2006(Brichau et al., , 2007Lamont, 2018). We do not find any evidence for a Basal Unit in Panormos Bay, in agreement with Bröcker and Franz (2005), although it is acknowledged that the dolomitic marbles are much thicker in this location and are affected by more intense M 2 retrogression associated with the overlying TSZ that crops out at Planitis Island.

Artemije and Efstathios Shear Zone, NW Tinos (AESZ)
The Artemije and Efstathios Shear Zone (AESZ) in NW Tinos (Figure 2; N37.655951, E24.992683) is~400 m thick and cuts through the Pyrgos Subunit. It is characterized by intense (F 1 ) SW verging parasitic folding, which affects both metasedimentary layering (S 0 ) and a planar foliation (S 1 ), whereas crenulation cleavages (S 2a+b ) develops axial planar to these folds. Well-developed (S 3 ) top-to-NE S-C′ fabric, crosscut the limbs of these folds and is associated with a blueschist and greenschist facies lineation (L 2 ) parallel to the shear direction (230/30). Although this structure dips toward the SW, its original geometry would have dipped gently to the NE after restoring the late doming. M 1b -M 2 phases are affected by top-to-NE shear (S 3 ), indicating this normal-sensed structure was active during exhumation from subduction-crustal depths. The degree of (S 3 ) top-to-NE shearing intensifies up structural section, and eventually discontinuous horizons of serpentinite and actinolite schists crop out on the northern coastline (see supporting information Text S1).

Tinos Shear Zone
The Tinos Shear Zone (TSZ) is a normal-sensed (M 2 ) greenschist facies structure~500 m thick that bounds the top of the Lower Unit (CBU; Figure 2). It is associated with top-to-NE kinematics (Mehl et al., 2005) in M 2 assemblages and is discordant to and truncates the M 1 blueschist and eclogite facies structures,

10.1029/2019TC005890
Tectonics suggesting it was not responsible for the exhumation from subduction depths. The structure outcrops on both the north and south coastlines, suggesting that it is folded around the island about an ENE-WSW axis (see stereonet of L 2 lineations on the TSZ in Figure 2). The gentle NNE dip of this structure on the island's northern coastline is responsible for the exposure of the Upper Unit including the Tsiknias Ophiolite as klippe on structurally high topography such as Mt Tsiknias (N37.581047, E25.225637). The TSZ is crosscut by the Tinos I-type granite and several smaller S-type intrusions, indicating it was active prior to granite emplacement (ca. 14.6 Ma; Brichau et al., 2007;Lamont, 2018) and may have been active at circa 21 Ma based on an Rb-Sr age close to the shear zone (Brichau et al., 2006(Brichau et al., , 2007Bröcker & Franz, 1998).
Because the Kionnia Subunit is not exposed elsewhere on Tinos and sits at a structurally high level (Figures 2  and 3), the TSZ must cut it out along strike. The TSZ also cuts out the Pyrgos Subunit in NW Tinos near the village of Marlas and at Panormos Bay places the Upper Unit directly against the deepest levels of the island (Sostis Subunit) (e.g., Avigad & Garfunkel, 1989;Bröcker & Franz, 2000;. The TSZ must truncate through all exposed M 1 metamorphic "stratigraphy." It is only on the south coast of Tinos where the thickest section of the CBU is preserved, which entrains the freshest HP rocks (Kionnia and Pyrgos Subunits), although rare occurrences of jadeitites outcrop in the Sostis Subunit on the north coast (Bröcker, 1990). The TSZ therefore postdates M 1 conditions and structures related to their exhumation, in agreement with conclusions of Bröcker and Franz (1998), who suggest that the tectonic juxtaposition occurred during M 2 (ca. 23-21 Ma). It is unclear whether the TSZ is related to crustal extension, as normal-sensed top-to-NE shear fabrics only record the relative ductile exhumation of material and cannot characterize a specific tectonic regime . Although, it is acknowledged that extensive high-angle normal faulting occurs directly above and crosscuts the structure and along the north and south coastlines, which is undoubtedly related to regional extension.

Marlas Shear Zone
The Marlas shear zone (MSZ) crops out in NW Tinos (Figure 2; N37.654406, E25.032923) and steeply dips northward but displays subhorizontal ENE-WSW trending lineations (see stereonet in Figure 3) that could be interpreted as dextral strike slip kinematics. This structure truncates the Pyrgos Subunit and places the Tsiknias Ophiolite, greenschists, and serpentinites of the Upper Unit directly against pervasively retrogressed HP-LT rocks of the CBU. It is possible the structure represented the TSZ but has been folded from its original dip. Alternatively, it could represent a late dextral strike-slip fault that developed at during M 2 that displaces the TSZ and juxtaposes the Lower and Upper Units together.

P-T Conditions of Metamorphism
The P-T results for prograde, peak, and retrograde metamorphic conditions are presented in Table 1 and described below. For further details of the petrographic methods, electron-probe microanalysis (EPMA), mineral chemistry, thermobarometric methods, and equilibrium phase diagram modeling, please refer to supporting information Text S1.
These results overlap with the Grt-Cpx-Ph barometer and Grt-Cpx thermometer on eclogite facies samples. However, Av-PT systematically predicts lower temperatures and pressures owing to inclusion of epidote in the calculated equilibria and therefore represents a lower bound to the peak P-T conditions. Despite this, the data still show that all rocks experienced HP although the Kionnia Subunit reached slightly higher pressures (~22-23 kbar) during M 1 .

Retrograde M 1b and M 2 Conditions
Retrograde M 1b conditions were estimated using the Av-PT function in THERMOCALC. Although no symplectites occur in Tinos eclogites and blueschists, many of the large amphibole laths in eclogite sample TLTN33 exhibit thin magnesio-hornblende rims in contact with retrograde matrix albite, pyroxene and titanite and rutile. Thin coronas associated with garnet breakdown also contain fine plagioclase, green amphibole, and epidote; are in contact with pyroxene that crosscut the S 2b matrix fabrics; and were also used by this method. M 2 retrograde pressures were also determined using average pressure calculations in Av-PT using the coexisting activities of amphibole, plagioclase, titanite, epidote, and H 2 O. Temperatures of 485-570°C were derived independently using the amphibole-plagioclase thermometer of Holland and Blundy (1994) at a reference pressure of 13 kbar, based on Av-PT results. M 1b amphibole rim growth yields pressures between 12.6 and 13.7 kbar (±0.8 kbar) between 500 and 600°C. Results were combined with previously calculated temperatures to produce a P-T array for amphibole rim growth between 520-570°C and 12.6-13.7 kbar.
Av-PT was carried out for Sostis Subunit sample TLT53 using M 2 phases Grt + Ep + Chl + Ttn + Ms + Pl + H 2 O + CO 2 , and an XH 2 O of 0.9 (minor matrix carbonate present) and yielded a P-T result of 7.3 ± 0.7 kbar and 536 ± 16°C. These conditions are also consistent with the Ti in Bt thermometer (Henry et al., 2005) on adjacent sample TLT59, which yields temperatures of 480-560°C. These results imply that the Sostis Subunit decompressed while remaining at high temperature through the upper greenschist facies.
Despite these results, the nature of the prograde and retrograde P-T path is poorly constrained using these methods alone, as conventional therombarometry is largely ineffective at constraining peak P-T conditions in HP rocks (e.g., Stípská & Powell, 2005). Uncertainty in thermobarometry in HP rocks arise due the valence state of Fe in omphacite, which biases estimates up temperature, and the lack of constrained activity solution models for Fe 3+ thus giving associated large uncertainties (typically, >±50°C and ±1 kbar at 1 SD; Powell & Holland, 2008). It is also possible that phase compositions used in thermobarometry are not representative of peak conditions. Hence, phase diagram modeling (pseudosection construction) in THERMOCALC was employed herein as the major tool to harness prograde, peak, and retrograde P-T information.

Equilibrium Phase Diagram Modeling 4.3.1. TLTN26: Garnet-Glaucophane Schist (Kionnia Subunit)
A P-T pseudosection for TLTN26 ( Figure 12) reveals that the inferred peak assemblage Grt + Gln + Lws + Ph + Rt + Qz (red text) is stable between~20-24 kbar and 490-500°C. Across the peak field, the amphibole Na B content ranges between 1.90 and 1.96 apfu, in good agreement with measured compositions. Garnet core (4-5% pyrope, 20-22% grossular) and rim compositions (6-7% pyrope, 22-24% grossular; Figure 12b) constrain the prograde evolution (pink shading, Figure 12) and suggest peak P-T conditions of~24-27 kbar and~520-540°C (pink polygon) but plotting inside the omphacite stability phase field, in disagreement with lack of observed omphacite. Due to the predicted presence of omphacite in the peak assemblage, a one-step garnet fractionation was carried out, with fractionation of garnet applied along the proposed P-T path (Figures 12a and 12b) at 20 kbar and 450°C. The same trends are observed on the resultant fractionated equilibrium phase diagram (Figures 12d-12f), with H 2 O-saturated topologies; however, the omphacite producing reaction is shifted to higher temperatures by this procedure. Pyrope and grossular isopleths also shift to lower temperatures and intersect at 22-26 kbar at 500-520°C in the Grt + Gln + Lws + Ph + Rt + Qz field (red text), in good agreement with observations. TLTN26 in outcrop showed variations between omphacite present and omphacite absent domains. However, Figure 12f indicates

TLTN24: Eclogite (Kionnia Subunit)
A P-T pseudosection for TLTN24 (Figure 13) shows the peak assemblage Grt + Lws + Gln + Omp + Ph + Rt + Qz is stable at >21 kbar, and 510-560°C. The calculated assemblage does however predict~1% chloritoid, which is not observed. However, owing to the very low abundance, chloritoid can be volumetrically ignored and could possibly be related to too much aluminum in the modeled bulk composition. The measured range in matrix amphibole (Na B ¼ 0.84-0.99 apfu, Figure 12) constrains this field further, indicating that the peak assemblage is stable at~540-560°C (green shading), with minor chloritoid predicted. Consideration of garnet core (5-6% pyrope, 26-28% grossular) and rim chemistry (11-12% pyrope, 22-24% grossular) further constrains the prograde evolution of the sample (red shading, Figure 13b). Garnet cores grew at 19 kbar, 510°C, whereas peak P-T conditions of~22 kbar and 540-560°C (green polygon) are calculated from garnet inner rims. Inflection of spessartine and decrease in grossular in garnet outer rim ( Figure 11c) is interpreted to reflect garnet resorption during decompression at >560°C, consistent with decreasing garnet modes (Figure 13c). Following the methodology above, fractionation of 50% of the garnet does not affect the peak estimate ( Figure 13b). The various constraints imply a prograde P-T path (dotted black line, Figure 13) compatible with the composition of amphibole observed as inclusions in garnet ( Figure 12). The only phase that does not show a reasonable model fit is phengite, with the calculated Si range of apfu in the peak assemblage field everywhere less than the measured value of 3.50-3.52 apfu, which could be due to a porphyllitic substitution causing a vacancy in the mica.
A titration determined XFe 3+ (Fe 3+ /Fe total ) value of 0.35 was applied for TLTN24. The effect of oxidation state and the validity of titration value is analyzed using a P-M(O) phase diagram (Figure 13d). At the predicted peak pressure of~22-24 kbar, the observed assemblage is stable between 0.20 and 0.40. At lower M(O), chloritoid is lost from the assemblage (Figure 13d), and at higher M(O), hematite joins the assemblage followed by destabilization of glaucophane and eventually garnet. The same trends and limits apply in temperature space (see supporting information). As M(O) is increased, epidote and hematite join the assemblage. The presence of epidote implies the sample must have been reasonably oxidized (>0.15). The glaucophane-barroisite solvus is invariant of oxidation state, as the top of the solvus (purple line) is met consistently at~12 kbar.

TLTN33: Eclogite (Kionnia Subunit)
A P-T pseudosection for TLTN33 ( Figure 14) reveals that the interpreted peak assemblage Grt + Gln + Omp + Lws/Ep + Ph + Rt + Qz (red text) is stable from~18 to 28 kbar, and 480-600°C, with this upper temperature boundary representing the appearance of kyanite. At~20 kbar, lawsonite (blue line) is predicted to replace epidote, and at higher temperatures, glaucophane is lost from the assemblage and is replaced by kyanite. Glaucophane becomes less sodic with decreasing pressure until the glaucophane-barroisite solvus extension is met, across which Na B rapidly decreases from 1.60-1.10 apfu. Two areas are highlighted by shading on Figure 14, which correspond to the range in Na B of the glaucophane cores and barroisite-magnesiohornblende rims with the former only present at HP within the peak field or within the epidote stability region. With these constraints, the composition zoning in large amphibole matrix laths must be retrograde (M 1b ).
Further constraints on the prograde evolution are provided by consideration of the garnet pyrope and grossular content. Garnet cores are characterized by 6-7% pyrope and 28-32% grossular, whereas garnet rims contain 12-13% pyrope and 25-27% grossular ( Figure 14). The colored shaded regions on Figure 14 correspond to where these constraints are met with cores growing at 22-23 kbar 540-550°C, whereas rims correspond to~24 kbar 570-575°C. All phases show a good model fit except for white mica. In this case, the model does not recreate paragonite found within the matrix and instead paragonite is not predicted until much lower pressures (>13 kbar).

10.1029/2019TC005890
Tectonics peak field, the amphibole Na B content ranges between 1.85 and 1.96 apfu, consistent with measured compositions (Figure 12). Garnet core (4% pyrope, 20-22% grossular) and rim chemistry (6-7% pyrope, 30-32% grossular; Figure 15b) further constrain the prograde evolution of the sample (pink shading, Figures 15b and 15c). Intersection of garnet core isopleths indicate garnet nucleation occurred at 27 kbar and 490°C in the peak assemblage field. In contrast the grossular rich rims intersect at 22 kbar and 510°C (pink polygon) but just plotting inside the omphacite stability phase field, in disagreement with lack of observed omphacite in the sample, although minimal omphacite is predicted (<0.02 mode). These results imply 17TL104 experienced decompression as garnet grew from~27-

10.1029/2019TC005890
Tectonics 22 kbar while temperatures increased by only~20°C and is consistent with an increase in garnet mode from 0.06-0.10 between core and rim intersections.
If the garnet core P-T result of~27 kbar is correct, it indicates significant overstepping of the garnet producing reaction and UHP conditions were achieved. However, no relict coesite inclusions were found in garnet, and the presence of paragonite and lack of kyanite in metapelitic rocks within the CBU raises ambiguity as to whether this estimated~27 kbar was experienced. A minimum pressure of 22 kbar can be rectified based on the predicted assemblage, lack of titanite inclusions in garnet and glaucophane chemistry between 1.90 and 1.96 apfu. This decompression path during garnet growth also suggests S 2a and S 2b rotational fabrics where associated with the earliest stages of exhumation, thus implying top-to-SW shearing was active during decompression. Further constraints can be placed on the retrograde history associated with growth of matrix epidote, chlorite and titanite. These join the assemblage and coexist together between 470 and 525°C at <17 kbar and coexist until~12 kbar when the sample crosses the glaucophane-barroisite solvus and plagioclase becomes stable.

TLX: Garnet-Glaucophane Schist (Sostis Subunit)
A P-T phase diagram (Figures 16a and 16b) reveals the peak assemblage Grt + Gln + Lws + Ph + Chl + Rt + Qz (red text) is stable between~18-26 kbar and~460-540°C. Across the peak field, the amphibole Na B content ranges between 1.90 and 1.96 apfu, in good agreement with measured compositions. Intersection of garnet core (4% pyrope, 27-29% grossular) and rim (6% pyrope, 27-29% grossular; Figure 16b) isopleths provides further constraints on the sample's prograde evolution (pink shading, Figure 16b) and suggests that garnet cores grew at P-T conditions of~21 kbar and~480°C (pink polygon), whereas rims grew at 22 kbar 510-520°C in the same field. This implies TLX underwent a dominantly heating P-T trajectory during garnet growth, consistent with increases in garnet mode. Although garnet rim isopleths also intersect at~12 kbar, the predicted assemblage of garnet, barroisite, chlorite biotite, epidote, rutile, and titanite is not consistent with observations of pseudomorphed lawsonites in the garnet cores or Na B compositions of amphibole. Constraints on the retrograde history can be determined by the lack of biotite in the matrix, implying the rock decompressed through the garnet, glaucophane, epidote, chlorite, and titanite phase field (green text) between~400-510°C and 10-15 kbar and remained cool during exhumation.

TLT60: Garnet-Mica Schist (Sostis Subunit)
A pseudosection of TLT60 (Figures 16c and 16d) reveals intersection of un-zoned garnet isopleths of pyrope (3-4%) and grossular (28-30%) occurred at~18 kbar 510-520°C in the Grt + Gln + Ph + Lws + Omp + Rt + Qz phase field. The inferred presence of lawsonite at peak conditions is consistent with relict pseudomorphed inclusions within garnet cores. The model predicts consistent phengite compositions of 3.51 apfu to observations, although no omphacite or glaucophane are present in the current assemblage. This peak P-T result is in good agreement with an Av-PT result of 18.5 ± 0.8 kbar and 498 ± 10°C. Upon reaching its peak pressures, the rock decompressed to the M 2

Tectonic Evolution and P-T-D Paths on Tinos
Diverse P-T paths recorded within the Lower Unit on Tinos (Figure 17) indicate that the CBU is not a single package of rocks, but rather a series of tectonometamorphic slices. Each subunit experienced a different tectonothermal evolution and were juxtaposed against each other during exhumation by a series of top-to-NE ductile extensional shear zones at the top of each subunit, and top-to-SW directed thrusts along their base. The exhumation mechanism of the CBU on Tinos is thus consistent with a model of SW directed synorogenic extrusion from a NE dipping subduction zone, whereby buoyant continental margin derived rocks return from subduction depths via return ductile flow in a channel or extruding wedge due to the positive buoyancy contrast compared to the overriding mantle (Agard et al., 2009;Chemenda et al., 1995;England & Holland, 1979;). An inverted metamorphic gradient (decreasing peak pressure with structural depth) is recorded across Tinos. Although we acknowledge a limitation with the calculated P-T is the uncertainties on activity-composition models and the role of local disequilibrium at peak and retrograde P-T conditions . However, this result has also been shown to occur on many other Cycladic islands including Syros and Sifnos (e.g., Laurent et al., 2016Laurent et al., , 2018Roche et al., 2016) and lower pressure blueschists in the Zas Unit at deeper levels of the CBU on Naxos .
Although the apparently better preservation of HP rocks toward the top of the CBU succession has been previously interpreted to more severe retrograde overprinting/availability of fluids in the lower parts of the CBU during the M 2 event (Avigad, 1998;Avigad & Garfunkel, 1989;Bröcker, 1990;Bröcker et al., 1993Bröcker et al., , 2004Katzir et al., 2002), our thermobarometry and field data questions this model. We argue for a major topto-SW thrust (Kionnia Thrust) and several other structures throughout the CBU on Tinos and importantly, peak M 1 pressures decrease with structural depth. This is consistent with the lack eclogite exposed at deeper levels of the CBU, the pervasive internal shearing, and formation of discrete HP mélanges throughout the higher structural levels of the CBU. Both these observations and our P-T data are inconsistent with the coherent unit hypothesis. Furthermore, the >6 km structural thickness of the CBU on Tinos is inconsistent with the narrow subduction/exhumation channel (otherwise referred to as the plate interface; Agard et al., 2018) predicted by numerical models for exhumation of HP rocks (England & Holland, 1979;Gerya et al., 2002), and geophysical observations of active subduction zones (Abers, 2005;Hilairet & Reynard, 2008). Additionally, many other Tethyan HP terranes are not exhumed as coherent units but as discrete thrust sheets or mélanges (e.g., Agard et al., 2018;Brovarone & Herwartz, 2013;Plunder et al., 2016;Searle et al., 2004;Smye et al., 2011). Finally, peak M 1 P-T conditions vary across the Cyclades (~26-22 kbar on Tinos vs.~14.5-12 kbar within the Zas Unit on Naxos; Lamont et al., 2019) supporting a model of progressive (diachronous) stacking of HP nappes at slightly different times and depths within the subduction channel or wedge.
In this subduction/extrusion model, underplating would be the dominant mode of accretion, as the active top-to-SW thrust stepped structurally downward with time. This would be combined with normal-sensed shearing also stepping structurally downward on the roof shear zone, allowing for exhumation of earlier subducted rocks up a subduction channel coeval with underthrusting of newly subducted rocks. This is consistent with the occurrence of the South Cycladic Thrust on Ios, which bounds the base of the CBU (Huet et al., 2009), recent interpretations from Syros (Laurent et al., 2018;Philippon et al., 2011;Ring, et al., 2020), the classical underplating model of Platt (1986), and refinements discussed in Agard et al. (2018). The evolution of each subunit is summarized from structurally high-to-low and where possible combined with available geochronology. 5.1.1. Kionnia Subunit Rocks at the highest structural level of the Kionnia Subunit, reached blueschist facies conditions of 22-26 kbar and 490-520°C during M 1 , significantly higher pressure than previous estimates from Tinos. In contrast, eclogite boudins >200 m structurally below record slightly lower pressures of~20-23 kbar but higher temperatures of 530-570°C. Although we acknowledge the transition of blueschist to eclogite 10.1029/2019TC005890 Tectonics Figure 16. Equilibrium phase diagrams of samples TLX and TLT60 from the Sostis unit, showing the phase relations (a, c) with isopleths of garnet pyrope and grossular (b, d), which show unique points of intersection, red text is the observed phase field for TLX, whereas the green text is the matrix retrograde assemblage.

10.1029/2019TC005890
Tectonics facies, defined by the appearance of omphacite, is sensitive to bulk composition and fluid availability, thus allowing both metamorphic facies to be present within the same rock volume at the outcrop scale (e.g., Weller et al., 2015; Figure 5b). The discrepancy in apparent P-T conditions across such a narrow deformed zone could be explained by either (1) A temporally evolving geothermal structure with depth within the subduction-accretion system, and subsequent tectonic juxtaposition of eclogite and blueschist during Tectonics exhumation. In this model, the upper levels of the Kionnia Subunit were buried deeper under a cooler temperature regime (2-4 kbar equating to~6-12 km) and were juxtaposed against the underlying (warmer) eclogites by a top-to-SW thrust. This is consistent with top-to-SW fabrics throughout the eclogite facies rocks in lower half of the Kionnia section corresponding to the Kionnia Thrust. This is, however, difficult to justify, as the blueschists are located <200 m away without a clear structural break. (2) Both packages of rocks were buried to similar pressures of~22-26 kbar and~500-520°C. At this point during their evolution, the blueschists and eclogites detached from the down-going slab and began buoyancy-driven exhumation toward the Earth's surface. The rocks in the upper half of the Kionnia Subunit remained sufficiently cool to maintain their blueschist assemblages, suggesting that they were located in the middle of the extruding channel sufficiently far away from localized shear zones that would otherwise heat the channel boundaries due to dissipative heating. In contrast, eclogite in the lower half of the Kionnia Subunit (close to the channel boundaries) increased in temperature upon exhumation possibly due to shear heating on the Kionnia Thrust. Although it is acknowledged that fluids may buffer out any extremely localized thermal gradients. This slight increase in temperature during exhumation may also explain the smoother garnet zoning profiles from core to rim in eclogite (TLTN24 and TLTN33) as compositions used to determine maximum pressures may have been affected by diffusion and therefore may only give an apparent prograde P-T trajectory. Eclogite garnets also show a spessartine inflection and decreasing grossular in their outer rim, interpreted as garnet resorption with decreasing pressure and increasing temperature. This could be consistent with shear heating during exhumation on the Kionnia Thrust. Irrespective of these speculative scenarios, the heating episode recorded in the eclogite caused lawsonite to decompose, and subsequently rehydrate the rock to form coarse, zoned, and crosscutting epidote and glaucophane after deformation in the eclogite boudin ceased.
U-Pb geochronology from the Kionnia Subunit includes a jadeitite dated at~63-61 Ma (Bröcker & Enders, 1999), although this likely reflects a mixed age (Bulle et al., 2010), whereas eclogite yields a lower intercept age of 78.2 ± 1.4 Ma (Bröcker & Keasling, 2006), overlapping with the circa 80-76 Ma ages from Syros HP metagabbro and plagiogranite in the Kampos mélange that are interpreted to represent magmatic crystallization in a small section of oceanic crust (Tomaschek et al., 2003). If these Late Cretaceous dates are interpreted as metamorphic ages, it would imply a separate HP event predating or associated with subduction initiation and ophiolite obduction at circa 74 Ma, which is difficult to fit in with this geodynamic framework. The compatibility of these ages and geochemistry between Syros and Tinos eclogite could be interpreted as the eclogite represent fragments of circa 78 Ma oceanic crust subducted slightly earlier (Bröcker & Keasling, 2006;Bulle et al., 2010). This may explain the warmer eclogite M 1 temperatures. However, this is not supported by our field data, which suggest the Kioinnia subunit is coherently layered and does not show evidence for other fragments of oceanic crust (e.g., meta-cherts, gabbro) and greater than tens of kilometers displacements between the eclogite and its metasedimentary host required by this model. An alternative explanation is the Kionnia eclogite represents a boudinaged sill or dyke that intruded the Cycladic/Adriatic continental margin associated with the same circa 78 Ma magmatic event, which formed a fragment of subducted oceanic crust exposed on Syros, immediately prior to subduction initiation at circa 74 Ma. Two ages on oscillatory zoned zircon from Kionnia eclogite are the best constraints of peak M 1 eclogite facies metamorphism as 53.5 ± 1.6 and 56.7 ± 3.9 Ma (Bulle et al., 2010).
Structural constraints show the Kionnia Subunit was exhumed by a combination of top-to-NE normal-sensed shearing concentrated on the roof shear zone, and top-to-SW thrust sensed shearing at the base (the Kionnia Thrust) (Figure 18). Both normal and thrust sensed shear fabrics show a continuum of deformation from peak to retrograde (M 1 -M 1b ) conditions. The exhumation and deformation history therefore follow an extrusion type mechanism with the Kionnia Subunit overthrusted onto the underlying Sostis Subunit that was not buried to as great depths (Figure 18). The exhumation was associated with the development of F 2 folds and isothermal decompression to conditions of~13 kbar 520-540°C (e.g., TLTN33), as represented by retrograde matrix phases including coexisting titanite and plagioclase and growth of coarse crosscutting barroisite amphibole rims. Upright open folds (F 3 ), with their axes trending parallel to the NE-SW shear direction developed during exhumation and were probably related to constriction within the extruding channel, owing to along-strike variations in channel thickness. These features have also been documented across the CBU (Gerogiannis et al., 2019;Xypolias & Alsop, 2014). If there were no variation in along-strike exhumation rates, narrower zones within the channel would produce a component of 10.1029/2019TC005890 Tectonics convergent flow orthogonal to the main transport direction (e.g., Lévy & Jaupart, 2011;Mancktelow & Pavlis, 1994) toward unrestricted zones.

Pyrgos Subunit
The Pyrgos Subunit records the stacking and folding of the proximal shelf carbonates under peak M 1 blueschist facies conditions of~22 kbar, 510°C. This result, however, represents a minimum pressure estimate as garnet cores yield up to 27 kbar, 480-490°C. These P-T conditions are comparable to that recorded by Kionnia Subunit blueschists (TLTN26) that sits at a similar structural level, although no eclogites have been found in the Pyrgos Subunit. Deformation was comparable to the Kionnia Subunit along strike, with S 1 and S 2 fabrics and (F 1 ) isoclinal folds forming during burial of the continental margin. This was later overprinted by pervasive top-to-NE shearing (S 3 ) that developed continuously during exhumation (M 1b -M 2 ) as the unit was extruding toward the SW, which overprints all previous prograde structures.

Sostis Subunit
In contrast to the eclogite and blueschist facies rocks preserved at higher structural levels, the interior of Tinos (Sostis Subunit), achieved less extreme peak pressures and temperatures of~18-19 kbar, 490-500°C during M 1 . It should be noted however, small occurrences of jadeitites and retrogressed eclogite occur on the north coast close to the listvenites reported by Hinsken et al. (2017) (M. Bröcker, personal communication, October 2019, although were not found in this study. Ar-Ar geochronology of white mica from this subunit suggest the minimum age of M 1 conditions were experienced at circa 44-40 Ma (Bröcker et al., 1993) and were affected by extensive M 2 retrogression through the upper greenschist facies at circa 31-21 Ma (Bröcker et al., 1993(Bröcker et al., , 2004Bulle et al., 2010). Rocks from this subunit underwent a similar prograde style to the overlying blueschists and eclogites and display lawsonite pseudomorphs as inclusions within garnet and within the matrix, and >3.5 Si apfu of phengites. Kilometer-scale (F 1 ) recumbent eastward verging folds are related to the burial process and the stalling of relatively buoyant metasediments. This presumably occurred upon reaching the point of neutral buoyancy due to entrainment of sufficient continental material. NE trending (F 2 ) sheath folds within all units are associated with the (S 3 ) top-to-NE fabrics and presumably also developed during its exhumation from mantle depths. The (top-to-SW) Kionnia Thrust separates this subunit from the overlying (and higher pressure) Kionnia Subunit. The~4-5 kbar pressure difference between the subunits (~12 km) suggests that the Sostis Subunit was underplated beneath the extruding Kionnia Subunit and achieved its peak pressures (equating to~60 km depth) at the same time as the overlying Kionnia Subunit was returning back toward the surface. This is because we should expect higher pressures in the Sostis Subunit if underplating occurred at the same time without any differential exhumation. The Zas Unit on Naxos records even lower pressures (~14.5-12 kbar) and represents the more proximal Cycladic/Adriatic continental margin . It may be expected that the Zas Unit structurally underlies the Sostis Subunit ( Figure 18) and was not buried as deeply.
Retrograde M 2 P-T conditions of~7.3 kbar, 500-540°C, indicate that rocks were exhumed at a minimum rate of~4 km/Myr and experienced a component of heating once they reached midcrustal conditions. The cause of this heating remains debated. While Avigad and Garfunkel (1991) suggest the heating is related to crustal extension, we argue Barrovian metamorphism is a result of crustal thickening as observed on Naxos, following continental collision between Eurasia and Adria/Cyclades .
In summary, the geometry of the CBU on Tinos does not represent the architecture of the subduction/ accretion system at peak M 1 conditions, but rather a snapshot of the lower half of the subduction channel once the subunits were juxtaposed against each other during exhumation (Figures 18 and 19). The P-T-D relationships within the CBU suggest that prograde and retrograde deformation progressed down structural level with time as the rocks were sequentially buried and extruded toward the SW from the subduction zone. This implies the extruding channel became wider at shallower depths as more deeply buried extruding subunits became coupled with subunits that did not reach such depths (Figures 18 and 19). This is also consistent with blueschists in Attica that structurally overlie the CBU and did not reach as HP but were exhumed from subduction depths by the same extrusion event (Baziotis et al., 2020).
Upon reaching midcrustal (M 2 ) conditions, the TSZ (associated with top-to-NE kinematics) truncates the HP (M 1 and M 1b ) structures at circa 21-14.6 Ma (Brichau et al., 2007;Bröcker et al., 1993). Deformation on the TSZ was responsible for exhumation of rocks from midcrustal depths much later in the orogenic history. It is unclear whether the top-to-NE extensional fabrics on the TSZ represents overall crustal extension or relative

10.1029/2019TC005890
Tectonics extension along a passive roof fault in an overall compressional regime (Means, 1989;. The TSZ was responsible cutting out~25 Myr of subduction history by juxtaposing the Upper Unit containing the Tsiknias Ophiolite, metamorphic sole, Mirsini Unit (pelagic metasediments) and Akrotiri Units in the hanging wall against the CBU in its footwall.
Thermal modeling of subduction zones shows that considerable conductive cooling occurs in the mantle wedge within the first few million years following initiation, eventually converging on a steady-state system (e.g., Kincaid & Sacks, 1997;Molnar & England, 1990Warren et al., 2008). Immediately following subduction initiation, temperatures on the slab top reach~800-900°C at 70 km depth, enough to trigger melting of the sediments under water-saturated conditions (Kincaid & Sacks, 1997). After this early thermal maximum, slab surface temperatures decrease to a steady state within a few Myr by continued subduction of the cold slab (Kincaid & Sacks, 1997;Molnar & England, 1990. The thermal evolution of the slab surface is influenced by (1) the subduction rate, (2) the age of the subducting plate, and (3) factors controlling the flux of hot material into the wedge to oppose the cooling effects of the slab (England & Wilkins, 2004;Molnar & England, 1990, although the age (or thickness) of the overriding plate and the slab dip angle are most influential (e.g., Molnar & England, 1990. For a fixed dip angle, rapid subduction beneath a thicker overriding plate results in increased cooling of the mantle wedge and generates the coolest slab surface temperatures (England, 2018;Molnar & England, 1990. In contrast, the maximum slab surface temperatures are recorded in slow subduction zones, where both the subducting and overriding plates are young and thin (England, 2018;Molnar & England, 1990. Dissipative heating likely plays a role (England, 2018;England & Molnar, 1993;Kohn et al., 2018;Turcotte & Schubert, 1973), although some disregard it as a driving metamorphism and melting (e.g., Abers et al., 2017;Syracuse et al., 2010). Shear stresses of 10-100 MPa are required to explain additional heat flow observations of 40 mW m −2 (England, 2018), corresponding to a coefficient of friction on subduction zone interfaces as~0.05-0.07 (England, 2018). Shear heating thus acts to increase temperatures near subduction zone interfaces with faster convergence rates and has been shown through thermal modeling to be necessary to explain P-T conditions recorded in HP rocks globally (Kohn et al., 2018;Peacock, 1992).
The cool geotherm during M 1 of~6-7°C/km, contrasts with >30°C/km recorded in the Tsiknias Ophiolite metamorphic sole associated with subduction initiation . This difference can be easily explained if both events were related to the same subduction zone but separated by circa 25 Myr (i.e., ca. 74 Ma vs. 53-46 Ma). During this time, the Upper Unit, including the Tsiknias Ophiolite, metamorphic sole, and Mirsini Unit, must have been accreted in an upper-plate position with the active subduction thrust stepping back and down structural level beneath these features (Figures 18 and 20).
The subducted oceanic plate (i.e., metamorphic sole amphibolites) formed during the early Jurassic (ca. 190 Ma;, similar to the age of the Cycladic continental margin (Altherr et al., 1982;Andriessen et al., 1979). During subduction initiation at circa 74 Ma, the subducted ocean crust (amphibolites) on the plate interface (The Tsiknias Thrust), reached higher temperatures due to a combination of shear heating and a possible excursion of the mantle wedge . In contrast, at the time of M 1 (ca. 53-46 Ma), the subducted oceanic crust would have been circa 140 Myr old and cold, thus cooling the thermal structure of the subduction zone. A minor complication is the circa 78 Ma ages recorded in Syros and Tinos eclogite. If these are interpreted as magmatic protolith ages, it suggests a small fragment of subducted oceanic crust and ocean-continent transition may have been warmer at the time of M 1 . However, the heating effect of these rocks is unlikely to be significant if most of the subducted oceanic crust was~140 Myr older and the upper plate was~110 Myr old. Using these ages, and the shortest possible time between sole formation (subduction initiation) and HP-LT metamorphism of circa 25 Myr, the apparent geothermal gradient in the subducting slab cooled at a rate of~1.2-1.5°C/km/Myr. Figure 19. Extrusion channel flow model after England and Holland (1979) and Chemenda et al. (1995) for Tinos and As-Sifah in Oman showing the sequence of exhumation for various parts of the subduction channel. (a) Sequential extrusion where Package C is extruded adjacent to Packages B and D. At this point the packages become coupled and are extruded adjacent to Packages A and E, note thrusts at the base of the extruding package and normal-sensed faults at the top. (b) Schematic velocity distribution diagrams for Poiseuille and Couette flow. (c) Predicted strain and fabrics recorded from symmetrical extrusion within the channel and (d) asymmetrical extrusion within the channel with more localized thrusting deformation at the base of the channel and more pervasive normal-sensed shearing throughout the channel.

10.1029/2019TC005890
Tectonics Because Molnar and England (1990) demonstrated that most cooling occurs in the first few Myr following subduction initiation, this cooling rate represents an absolute minimum.
A final consideration to explain low temperatures within the CBU is whether the HP-LT rocks remained within the interior of the subducting slab or channel and therefore had a conductive boundary layer separating them from the hotter slab surface. This mechanism may preserve cooler temperatures within the slab interior (e.g., England, 2018), whereas higher temperatures would occur on the slab-mantle wedge interface or on major shear zones due to both shear heating and conduction from the overlying mantle wedge (England, 2018). Following these arguments, the first-order control on the cool geotherm was likely the cold and old subducting lithosphere as well as the old and cold upper plate (ca. 190 Ma subducted oceanic lithosphere and the ca. 162 Ma Tsiknias Ophiolite in the upper plate), although fast subduction may have played an important role with cooling, it would also increase the magnitude of shear heating, which may become locally important.
In Oman, eclogites and blueschists crop out at the deepest exposed levels of the Arabian continental margin, 6-7 km structurally beneath the Semail Ophiolite and metamorphic sole (Searle & Cox, 1999, 2002Searle et al., 1994Searle et al., , 2004Warren et al., 2003). Eclogite in As-Sifah occur as boudins that represented basaltic sills that intruded the Arabian shelf carbonates (Searle et al., 2004) similar to eclogite within the Kionnia Subunit. These HP rocks reached similar P-T conditions to the CBU (20-22 kbar, 530-570°C) at circa 78 Ma (Warren et al., 2003;Warren & Waters, 2006) as the Arabian continental margin attempted to subduct. As-Sifah eclogites are also associated with extensive top-to-ENE extensional shear fabrics that formed during extrusion of buoyant continental margin back up the subduction channel under a passive roof fault during the final stages of ophiolite obduction, similar to the process we propose for Tinos (Searle et al., 2004). However, in As-Sifah the highest-pressure rocks are located at the deepest levels (right-way up metamorphism) and each tectonometamorphic unit are separated by normal-sensed shear zones associated with large P-T jumps (Searle et al., 2004). As-Sifah therefore represents the upper half of the extruding subduction channel, whereas the CBU on Tinos likely represents the lower half of the extruding channel/wedge and hence shows inverted metamorphism associated with firstly thrusting and underplating, overprinted by normal-sensed shearing during exhumation ( Figure 19). The subduction-extrusion event also occurred some circa 15 Myr following subduction initiation and obduction of the Semail Ophiolite (Searle et al., 2004;Warren et al., 2003). The timing, occurrence and P-T-D constraints of HP metamorphism following ophiolite obduction in Oman are almost identical to that recorded in the Cyclades. The only difference is the HP sequence is cut by the later TSZ, which juxtaposes the Tsiknias Ophiolite and Metamorphic Sole directly against the CBU, which was subducted some~25 Myr later.

Tectonic Model and Conclusions
Integration of P-T-D data with existing geochronology, the tectonothermal evolution of the CBU on Tinos, involves the following stages ( Figure 20 Tectonics . These rocks were accreted to the base of the ophiolite and became part of the upper plate during the obduction process and are today exposed in the Upper Unit that did not reach HP (M 1 ). During obduction, the subduction interface must have stepped back, allowing steady-state subduction of oceanic lithosphere under the Upper Unit (in the upper plate) for~25 Myr. Late Cretaceous UHP metamorphic rocks also occur along strike in the Rhodope Massif (Collings et al., 2016) suggesting the subduction zone was a regional feature. 2. Steady-state NE dipping subduction of circa 190 Ma oceanic lithosphere continued between 74 and 53-46 Ma (i.e., for ca. 25 Myr). During this period, the subduction zone cooled, possibly owing to fast plate motions and advection of cold 190 Ma oceanic crust down to mantle depths. We believe Tinos eclogite represent boudinaged mafic intrusions and meta-volcanics within the leading edge of the Cycladic continental margin. Although eclogite in the Kampos and Kini mélanges on Syros, probably represent fragments of subducted oceanic crust and include metagabbro, serpentinites, plagiogranites, and metacherts (Keiter et al., 2004(Keiter et al., , 2011. On Tinos, U-Pb data suggest the M 1 event occurred at circa 57-53 Ma (Bulle et al., 2010) and possibly as old as circa 63-61 Ma (Bröcker & Enders, 1999) although the latter is likely a mixed age. This contrasts with slightly younger Lu-Hf garnet ages across the CBU of circa 52-46 Ma suggesting that the M 1 event was diachronous. 3. The arrival of the leading edge of the Cycladic continental margin at the trench resulted in burial of the CBU under a cold geothermal gradient (6-7°C/km) through the lawsonite stability field to peak (M 1 ) blueschist-eclogite facies conditions at circa 53-46 Ma (Lu-Hf garnet; Dragovic et al., 2012;Lagos et al., 2007). This burial process produced (S 2b ) top-to-SW thrusts and isoclinal (F 1 ) folding as rocks were continuously underplated, causing the subduction wedge to thicken. Blueschist and eclogite exposed on the Kionnia peninsular achieved P-T conditions of~22-26 kbar, 490-520°C and 20-23 kbar, 500-570°C, which equates to depths of~70-80 km. Such depths can only be achieved due to the slab-pull force from already subducted and eclogitized dense oceanic crust at greater depths (e.g., Kampos Subunit on Syros; e.g., Keiter et al., 2004Keiter et al., , 2011Laurent et al., 2018). 4. Upon reaching peak pressures, and possibly due to a change in the boundary conditions from subduction termination as a result of continent-continent collision between Cyclades/Adria and Eurasia. The Kionnia and Pyrgos Subunits detached from the subducting slab and were buoyantly extruded back up an overriding channel (plate interface; Agard et al., 2018) producing top-to-NE (S 3 ) extensional fabrics along the roof shear zones, and top-to-SW fabrics (S 2b ) along the base, and (F 2 ) sheath folds due to noncoaxial SW directed ductile flow. During exhumation, (M 1b ) deformation was localized at the margins of the subunits. Eclogite remained unaffected by the exhumation related deformation (S 3 ), due to strain localization at the margins of the boudins, owing to the competency difference with the host metasediments. Lawsonite breakdown during decompression, released fluid, hydrating the matrix of these rocks at~18-20 kbar and 550°C, producing crosscutting coarse and zoned epidotes and glaucophane-barroisite. Upright open folds (F 3 ) with their fold axes parallel to L 1 developed during the early stages of exhumation (M 1b ), within the extruding channel due to lateral constriction conserving flow, likely owing to the channel having a variable along-strike thickness. Localized constriction occurred in the narrowest parts of the channel due to converging ductile flow in response to the obstruction at the margins. 5. Following peak M 1 pressures, the Kionnia Subunit was overthrusted onto the Sostis Subunit, by a top-to-SW thrust (Kionnia Thrust). The timing of this thrusting event (M 1b ) is best constrained by Ar-Ar geochronology from phengite deformed with the foliation which yield ages of circa 44-40 Ma (Bröcker et al., 1993). The opposing kinematics on the top and bottom of the Kionnia Subunit and the lower pressures of the Sostis Subunit imply an extrusion mechanism, whereby the Sostis Subunit achieved its peak pressures at a slightly later time than the overlying Kionnia Subunit (which was on its return journey toward the surface). The Sostis Subunit experienced a similar prograde evolution, through the lawsonite stability field to less extreme P-T conditions of~18.5 kbar and 490-500°C, followed by isothermal decompression associated with widening of the extrusion channel at shallower depths. Top-to-NE shearing (S 3 ) from M 1 blueschist-M 2 greenschist facies conditions was coeval with localized (S 2b ) top-to-SW thrusting, suggesting the zone of underplating and subsequent extrusion stepped down structural level with time. The Sostis Thrust is the best exposure of a thrust that developed during this extrusion stage. The sequential juxtaposition of material during extrusion postdating peak metamorphism was when the architecture of the CBU on Tinos was aggregated. Exhumation from the subduction zone was complete by circa 31 Ma (Bulle et al., 2010), implying a minimum exhumation rate of~4 km/Myr.

10.1029/2019TC005890
Tectonics 6. Upon reaching crustal depths (<12 kbar) the rocks were affected by extensive greenschist facies metamorphism (M 2 ), at P-T conditions of~7.3 kbar and 530°C at circa 31-21 Ma (Bröcker et al., 1993(Bröcker et al., , 2004Bröcker & Franz, 1998;Hinsken et al., 2016). Top-to-NE (S 3 ) shearing became localized on the TSZ, which was responsible for juxtaposing the Lower Unit against the ophiolitic Upper Unit that was already at a much higher structural level. Across the TSZ,~25 Myr of subduction history is missing. The TSZ truncates all M 1 -M 1b fabrics and folds and therefore postdates synorogenic extrusion of HP rocks from subduction depths. 7. During M 2 , upright crenulation cleavages (S 4 ) and open upright folds (F 4 ) formed across all subunits on Tinos with NE-SW axes parallel to the extensional (L 2 ) lineation. A component of E-W shortening must have occurred during the Miocene, prior to final exhumation. 8. All structures on Tinos are affected by doming and cut by the I-type granite pluton and smaller S-type intrusions, suggesting M 2 top-to-NE shearing ceased by circa 14.6 Ma (Brichau et al., 2007;Lamont, 2018). The final exhumation to the surface was due to footwall uplift associated with high angle brittle normal faults along the northern and southern coastlines.
Further geochronological investigation to constrain the rates of deformation and exhumation of the shear zones and subunits identified in this study are required to successfully correlate them across other Aegean islands. Such work would greatly improve understanding of deep accretionary processes that occurred during, or immediately prior to continent-continent collision between Cyclades/Adria and Eurasia resulting in the Aegean Orogeny.

Data Availability Statement
Data used in this study are deposited in the Oxford University Research Archive under the search name "The Cycladic Blueschist Unit on Tinos, Greece: Cold NE Subduction and SW Directed Extrusion of the Cycladic Continental Margin under the Tsiknias Ophiolite," (Lamont, Palin, 2020) available via https://10.5287/bodleian:rbaYPaMy7.