Length scales and types of heterogeneities along the deep subduction interface: Insights from exhumed rocks on Syros Island, Greece

The subduction interface is characterized by progressive downdip changes in metamorphic conditions, rheological properties, and fluid sources and pathways. Geodetic and seismological observations in modern subduction zones suggest that these downdip changes can be accompanied by important differences in seismic style. For example, shallow slow slip—an aseismic phenomenon characterized by transient, accelerated interface strain rates—has been observed up-dip of the seismogenic zone in several modern subduction zones, including Hikurangi (Douglas et al., 2005; Wallace and Beavan, 2010; Wallace et al., 2012), Northern Japan (Ito et al., 2013), Costa Rica (Dixon et al., 2014), and Nankai (Ito and Obara, 2006). This zone of shallow slow slip transitions downdip into a locked region that has the potential to release elastic strain in large, destructive earthquakes (e.g., Kanamori, 1986; Oleskevich et al., 1999). Below the locked megathrust zone, the behavior of the deep interface is especially complex. A growing body of land-based geodetic and microseismic observations suggests that deep interface deformation is characterized by transient accelerated aseismic slow slip events (SSEs) and low-frequency earthquakes (LFEs), together forming what is known as deep episodic tremor and slow slip (ETS) (e.g., Obara, 2002; Rogers and Dragert, 2003; Ito and Obara, 2006; Shelly et al., 2006; Beroza and Ide, 2011). LFEs and corresponding tremor bursts occur during and between SSEs throughout an ETS event (Ito et al., 2007; Shelly et al., 2007b). ETS events are capable of dissipating ∼75% to nearly 100% of convergence-related stress downdip of ∼25 km accumulated over years to decades (Chapman and Melbourne, 2009; Wallace and Beavan, 2010); of this fraction, seismic tremor accounts for 0.1% to ∼20% of the total change in strain energy (Kao et al., 2010; Sweet et al., 2014). Understanding ETS events is crucial, as these events may load the megathrust up-dip and/or along strike, thereby increasing the potential for great subduction earthquakes (Dragert et al., 2004; Chapman and Melbourne, 2009; Segall and Bradley, 2012).

Each distinct zone of seismic behavior must be linked to deformation conditions, mechanisms, and types of heterogeneities within that region of the interface. For the shallow interface, these linkages can be detected through drilling (e.g., Moore, 1995), seismic imaging (Byrne et al., 1988; Abers, 1992), and continuous monitoring of seafloor deformation (e.g., Wallace et al., 2016). Combined with numerical modeling and deformation experiments, these observations suggest that shallow slow slip events occur in regions of transitional frictional properties, under low effective stress and high pore fluid pressures (e.g., McCaffrey et al., 2008; Saffer and Wallace, 2015; Wallace et al., 2016). Some studies, for example, suggest that changes from velocity strengthening to velocity weakening along a planar frictional fault surface may explain unconventional, slow seismicity (Shibazaki and Shimamoto, 2007; Rubin, 2008), while others invoke protolith-induced rigidity contrasts within macroscopically ductile interface shear zones (Fagereng and Sibson, 2010; Saffer and Wallace, 2015; Fagereng and Den Hartog, 2016). Downdip within the locked zone, the spatial extents and distributions of megathrust earthquakes can also be linked with primary structural heterogeneities such as oceanic seamounts or other topographic highs, sediment subduction and corresponding changes in deformation mechanisms, and/or heterogeneous material and mechanical properties generated at the seismogenic front during compaction and dehydration (Cloos, 1992; Moore and Saffer, 2001; Husen et al., 2002; Gulick et al., 2011; Hicks et al., 2012).

The role of primary or generated heterogeneities in contributing to seismic behavior below the megathrust locked zone is less well understood, partly because direct observations of the plate interface are not possible at these depths (e.g., ≥25 km). This is also the depth range where pressure and temperature conditions begin to favor viscous creep mechanisms in most crustal minerals, depending on fluid pressure conditions and strain rates. A transition to viscous creep along the deep interface has the potential to “smear” and/or completely homogenize heterogeneities that may have once acted as frictional asperities at shallower depths, through ductile deformation and widening of the interface shear zone. By analogy with continental shear zones, amphibolite- and granulite-facies shear zones often show such homogenization, exemplified by penetrative, continuous shear of a wide range of protolith rock types (e.g., Fountain and Salisbury, 1981; Kenkmann and Dresen, 2002; Norris and Cooper, 2003). The discovery of ETS at high-pressure conditions in subduction zones, however, suggests that rheological heterogeneities still persist in a form that may affect seismic behavior. But what kinds of heterogeneities occupy the deep subduction interface? What are their rheological properties, length scales, and spatial distributions? And how do they relate to the heterogeneities that existed in protoliths prior to subduction? In this paper, we address these questions using eclogite-, blueschist-, and greenschist-facies rocks exposed on Syros Island, Greece. Syros is interpreted as a fossilized subduction shear zone recording deformation in several rock types over a wide range of pressure-temperature conditions and therefore may shed light on processes controlling transient phenomena such as ETS.

Since the Late Cretaceous, the Aegean-Anatolian region has been characterized by ∼2000 km of northward subduction at ∼2 cm/year of the leading edge of Africa beneath Eurasia along the Hellenic subduction zone (Fig. 1). Cretaceous-to-Eocene subduction produced high-pressure/low-temperature (HP/LT) metamorphic rocks and arc volcanics that are currently exposed in parts of mainland Greece, the Cycladic Islands in the Aegean, and western Turkey. Slab retreat and regional extension began in the Eo-Oligocene and continues to the present (Royden, 1993; Jolivet et al., 2010b, 2013; Papanikolaou, 2013). Slab rollback through time has resulted in ∼700 km southward migration of the magmatic arc, southward younging of subduction-related metamorphism and flysch sedimentation (Jolivet and Brun, 2010; Ring et al., 2010; Papanikolaou, 2013), and progressive exhumation of HP/LT rocks in the wake of the retreating slab in the backarc region (Ring et al., 2001; Jolivet et al., 2003; Ring and Glodny, 2010).

In the Cycladic Islands in the central Aegean, the multi-stage tectonic history is best recorded in the Cycladic Blueschist Unit (CBU). The CBU protoliths include Carboniferous orthogneissic basement; Mesozoic sediments, volcanics, and carbonate platforms; and Late Cretaceous (ca. 80–75 Ma) oceanic lithosphere (Bonneau, 1984; Tomaschek et al., 2003; Bröcker and Pidgeon, 2007; Tomaschek et al., 2008; Papanikolaou, 2013). These rocks record Eocene (ca. 53–40 Ma) HP/LT subduction-related deformation and metamorphism (Dixon, 1976; Okrusch and Bröcker, 1990; Tomaschek et al., 2003; Lagos et al., 2007; Skelton et al., 2018). The metamorphosed Cretaceous oceanic lithosphere, which was relatively young (∼30 m.y.) and hot at the time of subduction, contains relict pillow lavas and Mn-rich metasediments characteristic of a magmatically and hydrothermally active oceanic crust (Keiter et al., 2004, 2011). However, some CBU serpentinites and chlorite schists retain oxygen isotopic and trace-element signatures that indicate serpentinization occurred on or beneath the seafloor before subduction (Cooperdock et al., 2018), suggesting they could have formed in a hyper-extended magma-poor rifted margin (e.g., Beltrando et al., 2014). Regardless of the specific geodynamic setting, the CBU juxtaposes a variety of protolith types that all experienced the Eocene high-pressure event. Subsequent greenschist-facies retrogression and deformation occurred during Oligo-Miocene exhumation along crustal-scale, low-angle normal faults of the North and West Cycladic Detachment Systems, the Paros-Naxos Detachment System, and the Santorini Detachment System (Avigad and Garfunkel, 1991; Forster and Lister, 1999; Ring et al., 2010; Jolivet et al., 2010b; Grasemann et al., 2012; Soukis and Stockli, 2013; Schneider et al., 2018). The CBU locally records an amphibolite-facies metamorphic event and migmatization of Miocene arc-related granitoid intrusions, best exposed on Naxos (Jansen, 1977; Andriessen et al., 1979).

Among the Cycladic Islands, HP/LT metamorphism is best preserved on Syros (Ridley, 1982; Trotet et al., 2001; Putlitz et al., 2005). From structural top to bottom, the CBU on Syros comprises an uppermost metabasite unit containing slivers of serpentinite that reached peak metamorphic conditions at ca. 53–50 Ma; several nappes of intercalated metasediments and marble schists (with mafic lenses) that reached peak conditions at ca. 43–38 Ma; and greenschist-facies gneisses and schists (Tomaschek et al., 2003; Keiter et al., 2004, 2011; Lagos et al., 2007; Skelton et al., 2018) (Fig. 1). The maximum pressure-temperature estimates for the CBU are quite variable in the literature (particularly with regard to pressure), ranging from ∼14 to 23 kbar and ∼450 to 500 °C in the blueschist-to-eclogite facies (Fig. 2A) (Schliestedt, 1986; Putlitz et al., 2000, 2005; Trotet et al., 2001; Schmädicke and Will, 2003; Behr et al., 2018; Laurent et al., 2018). Reported pressures correlate to a maximum depth range between ∼45–80 km, assuming lithostatic conditions. The structural evolution on Syros comprises at least two stages of penetrative foliation development and folding during subduction to peak depths, followed by progressive localization of strain under retrogressive blueschist- and eventually greenschist-facies conditions during exhumation (Fig. 2B; Dixon, 1976; Ridley, 1982; Rosenbaum et al., 2002; Keiter et al., 2004, 2011; Bond et al., 2007; Laurent et al., 2016). Additionally, the presence of undeformed lawsonite and aragonite pseudomorphs indicates penetrative strain in some meta-mafic exposures ceased at or near peak pressures and temperatures prior to exhumation (Keiter et al., 2004; Philippon et al., 2012).

Many observations suggest exhumation on Syros occurred in two stages: an initial stage in which rocks were returned from peak depths (∼45–80 km) to the overriding forearc in a subduction channel geometry and a second stage in which rocks were exhumed from the middle crust (∼20–30 km) along semi-brittle and brittle crustal-scale detachment faults and shear zones (Avigad et al., 1997; Jolivet et al., 2003, 2010b; Ring et al., 2010; Laurent et al., 2018; Fig. 2B). These stages are commonly distinguished by metamorphic grade and by the dominant orientation of mineral lineations: whereas prograde rocks preserve N-S–trending lineations consistent with the orientation of subduction, retrograde rocks preserve NE- and E-trending lineations (Gautier and Brun, 1994; Rosenbaum et al., 2002; Jolivet et al., 2013; Philippon et al., 2013; Laurent et al., 2016). Rb-Sr ages of phengites micro-drilled from exhumation-related fabrics in metasedimentary units indicate that deformation and metamorphism occurred continuously in the subduction channel as rocks exhumed between ca. 42–30 and perhaps even to 20 Ma (Bröcker et al., 2013; Cliff et al., 2017). The latest exhumation stages on Syros are best recorded by the Miocene Vari detachment, which crops out on the southeast tip of the island (Fig. 1B) and juxtaposes greenschist-facies mylonites and gneisses in the hanging wall against CBU in the footwall (Ring et al., 2003; Soukis and Stockli, 2013). This structure has been characterized as a reactivated subduction channel roof that assisted exhumation from ∼500 to 550 °C in the eclogite facies, through ∼450–500 °C in the blueschist facies during continuous subduction and early stages of slab rollback (Jolivet et al., 2010a; Laurent et al., 2016).

For this study, we selected three Syros exposures (Kini, Agios Dimitrios, and Fabrikas) that represent different subducted protoliths and different conditions of deformation within the subduction shear zone (Figs. 1 and 2). The Kini and Agios Dimitrios exposures were selected because they comprise dominantly oceanic-affinity meta-mafic rocks and thus may provide insight into the rheological behavior of subducted oceanic crust that is sediment poor. The Fabrikas locality is a heterogeneous lens of meta-mafic and metasedimentary rocks, itself contained within a surrounding matrix of calc-schists and marbles (Fig. 1). Therefore, this locality provides insight into the rheological behavior of subducted oceanic crust blanketed by sediment. For each exposure, in Section 3, we first provide a general description of the outcrop characteristics and then summarize the interpreted pressure-temperature (PT) conditions recorded there, based on both previously published work and our own observations. We then describe the types of rheological heterogeneities preserved at each locality, where heterogeneities are defined as any features that produce outcrop-scale strain gradients or significant differences in deformation mode or mechanism. We document the types, sizes, and distributions of rheological heterogeneities based on field measurements and the deformation mechanisms that define the heterogeneous materials using microstructural observations (all measurements are provided in Supplemental Tables S1–S4¹). In Section 4, we discuss our observations in the context of (1) the sources of rheological heterogeneity that exist and/or persist at depth in generalized subduction settings; (2) the mechanical behavior of the deep interface; and (3) the compatibility of the observed rheological heterogeneities and their deformation modes with observations and models of ETS in modern subduction zones.

The Kini area, part of the structurally highest oceanic-affinity subunit, spans ∼400 m of coastal exposure where meta-mafic blueschists and eclogites are exceptionally exposed. The protolith is dominantly Late Cretaceous mid-ocean ridge basalt, with some rocks preserving a more calc-alkaline signature (Bonneau, 1984; Seck et al., 1996; Tomaschek et al., 2003; Bulle et al., 2010). Map-scale slivers of serpentinite and chlorite-talc schist are mostly localized to the Kini block’s margins (Fig. 1). In-place exposures are cliffs several tens of meters in height, comprising alternating glaucophane-garnet-epidote blueschists, minor chlorite-talc schists, and interspersed pods of eclogite containing omphacite + garnet ± glaucophane, and zoisite. Foliations dip gently to moderately S-SE and exhibit tight to isoclinal folds with axial planes subparallel to the foliation. These folds are visible on the centimeter to meter scale in outcrop and have moderately S-plunging fold axes commonly decorated by mineral lineations defined by glaucophane, epidote, and phengite (Fig. 3A). Locally, the primary foliation is cut by an upright crenulation cleavage with NE-SW–striking axial planes associated with some realignment and growth of glaucophane, epidote, and phengite.

The primary foliation at Kini is interpreted to record prograde subduction and/or peak metamorphic temperature within the subduction shear zone based on (1) preservation of N-S–trending lineations in HP minerals, which is consistent with the orientation of prograde shearing (Philippon et al., 2013; Laurent et al., 2016; Behr et al., 2018); (2) preservation of undeformed lawsonite pseudomorphs stamped over the foliation or only weakly deformed (Keiter et al., 2004, 2011; Philippon et al., 2013); and (3) a lack of retrogressive phases or extensive mineral replacement (Behr et al., 2018). We reason that the upright crenulation cleavage likely developed in the early stages of exhumation since it overprints an older foliation and represents a rotation of maximum compressive stress by ∼90°. This transition in principal stress orientation would not be expected if rocks were being sheared along a continuous path during subduction.

We document two primary types of heterogeneities at Kini: (1) coarse-grained blueschist-to-eclogite facies gabbroic pods; and (2) fine-grained, massive to weakly foliated eclogite lenses. The coarse-grained meta-gabbro pods have massive textures that resemble igneous cumulates, characterized by blueschist-facies minerals glaucophane, epidote, garnet, ± omphacite (Figs. 4A and 4B). The surrounding matrix rocks are fine grained, strongly foliated, epidote-blueschist–facies meta-basalts that wrap around the meta-gabbros, producing boudin-matrix structures. Meta-gabbro pods range from ∼5 cm to 5 m in equivalent circle diameter, with aspect ratios (stretching-parallel length to foliation-perpendicular width) on average ∼1.5:1 (Fig. 3B). The interiors of meta-gabbros are unstrained to weakly foliated; weak fabrics are visible in preferentially aligned epidote. Some metamorphosed pods preserve weak shearing of relict intrusive relationships in which finer-grained mafic rocks are juxtaposed with the coarser-grained gabbros (Fig. 4A). The second type of heterogeneity at Kini is associated with local concentrations of eclogite-facies minerals in discrete lenses, pods, and boudins embedded within the blueschist matrix (Figs. 4C–4G). The largest eclogite pods are fallen blocks along the coastline up to tens of cubic meters in volume. The majority of eclogite lenses that are observed in place, however, range from 0.5 to 20 cm in equivalent circle diameter with average aspect ratios of 2.5:1–3.5:1 (Fig. 3B). The smaller eclogite pods are typically distributed on the centimeter to tens of centimeters scale throughout the local matrix, whereas larger ones are commonly separated from one another by several meters. Several of the eclogitic lenses can be classified as fractured boudins, where the fractures are dilational veins opened perpendicular to the foliation (Figs. 4F,4G, and 5). Dilational shear fractures are also common, occurring at 30°–55° angles to the foliation, and some filled with quartz, phengite, glaucophane, and/or chlorite (Figs. 3C, 4C, 4D, 4F, 4G, 5A, and 6A). Displacements on shear fractures range from 0.1 to 2.0 cm (Fig. 3D and Table S3 [^{footnote 1}]). In some places, an earlier generation of veins appears sheared into parallelism with the surrounding matrix and is cut by later-generation veining oriented at higher angles to the foliation (Figs. 4F and 4G). Some mineralized veins filling boudin necks are themselves boudinaged (Fig. 4G). The veins and shear fractures are most prominent in omphacite- and garnet-rich eclogitic lenses, whereas eclogite pods with greater concentrations of glaucophane and/or epidote are instead weakly folded at a greater wavelength than the surrounding blueschist matrix, and fractures are less well developed. In nearly all examples, the dilational veining and shear displacement that occur within the eclogitic lenses are accompanied by ductile flow in the surrounding blueschist matrix (Figs. 4C–4G).

In thin section, Kini blueschists consist of alternating glaucophane- and epidote-rich layers with porphyroblasts of omphacite, phengite, garnet, rutile, and minor quartz, titanite, and chlorite. Glaucophane and epidote grain sizes range from 50 to 500 µm, and porphyroblasts of garnet and omphacite range from 300 µm–5 mm (Fig. 6A). Omphacite porphyroblasts comprise two chemically distinct types: a high-Mg, low-Na core, fringed by a low-Mg, high-Na tail. In Figure 6B, a low-Na omphacite porphyroblast is brittlely sheared and offset by micro-domino fractures. Brittle fractures have a component of dilation and are filled with glaucophane and high-Na omphacite. In Figure 7, we show electron backscatter diffraction (EBSD) data to quantify intracrystalline strain and lattice-preferred orientation (LPO). Glaucophane and epidote preserve evidence of intracrystalline plasticity in the form of undulose extinction, subgrain walls oriented both parallel and perpendicular to the foliation, bulging grain boundaries, newly formed strain-free grains (∼100 µm), and strong LPOs. For description of electron micro-beam methods, please see Supplemental Text (^{footnote 1}). Behr et al. (2018) also presented EBSD data from two Kini blueschists with similar results. The eclogites at Kini are primarily massive in texture, with grain sizes ranging from 300 µm–1 mm. As observed at the field scale, the eclogites are commonly cut by multiple generations of veins filled with white mica ± chlorite and quartz (Figs. 5A–5D).

Agios Dimitrios, just north of Ermoupoli on the east coast of Syros, is another exposure of the structurally highest oceanic-affinity blueschist-eclogite subunit. Similar to Kini, the outcrop spans ∼900 m of coastline and is continuous over several tens of meters in height and length. The primary foliation is characterized by intercalated fine-grained blueschist and chlorite-actinolite greenschist, with lenses of omphacite-garnet ± glaucophane, epidote eclogite. Blueschist and greenschist foliations are folded, with a shallowly E-plunging fold axis. Mineral lineations in glaucophane and epidote plunge dominantly E, but S- and SE-plunging lineations are also present (Fig. 3A).

The dominant phase of deformation and metamorphism at Agios Dimitrios is interpreted to record return flow in the subduction shear zone from eclogite facies down pressure and temperature into greenschist-facies conditions. This interpretation is based on (1) preservation of some S-trending, but dominantly SE- and E-trending lineations in HP and greenschist-facies minerals (Fig. 3A), which we interpret to represent a progressive change in stretching direction from subduction to exhumation; (2) preservation of foliation-parallel, coexisting blueschist-facies and greenschist-facies mineral assemblages (cf. Fig. 8); (3) partial replacement of eclogites by epidote and glaucophane (Fig. 9).

Similar to Kini, the primary types of heterogeneities at Agios Dimitrios include: (1) coarse-grained blueschist-to-eclogite–facies meta-gabbros and (2) eclogitic lenses embedded in the surrounding matrix (Fig. 8D), but as discussed in Section 3.2, this unit has been partially retrogressed under blueschist-greenschist–facies conditions. We document a distinct meta-gabbro lens, which, like those described at Kini, preserves a coarse-grained massive texture resembling an igneous cumulate with omphacite, garnet, glaucophane, ± epidote (Fig. 8A). The surrounding matrix wraps around the eclogitic meta-gabbro and consists of fine-grained greenschists. This meta-gabbro is the largest heterogeneity of its type that we have yet found, standing ∼5 m tall, ∼10 m wide, and at least several m in the stretching-parallel direction, equating to a minimum volume of ∼150 m³. Because the view in Figure 8A is looking roughly parallel to lineation, it is difficult to determine a true aspect ratio, but the ratio of long:short axes in this view is ∼2:1. This eclogitic meta-gabbro lacks a pervasive internal fabric and is dissected by multiple generations of dilational veins filled with a variety of hydrous phases (Figs. 8B and 8C). Crosscutting relationships indicate that epidote ± garnet veins rimmed with glaucophane are oldest; they are crosscut and in some places offset by later epidote ± glaucophane veins that are commonly rimmed by actinolite. All early veins are crosscut by actinolite- and/or quartz-filled veins, suggesting these are the youngest generation (Fig. 8B). Locally, splaying epidote-filled veins rimmed with actinolite appear to have been ductilely sheared to low strain during or after emplacement (Fig. 8C). The second type of heterogeneity at Agios Dimitrios, like at Kini, are local concentrations of eclogite-facies minerals in pods and lenses smeared and boudinaged within the mafic blueschist matrix (Fig. 8D). Pod sizes typically range from ∼1–50 cm in equivalent circle diameter with average aspect ratios of 3:1–5:1; some elongated lenses are traceable parallel to the stretching direction for several meters with aspect ratios up to 15:1 (Fig. 3). Eclogite pods in the tens of cm size range are typically distributed within the matrix on a sub-cm to cm scale, whereas elongated eclogite lenses are distributed on the tens of cm to several meters scale. Brittle fractures in eclogite pods are observed (e.g., Fig. 8D), and ductile strain in eclogites is also common. Many of the lenses preserve pinch-and-swell textures or are folded at wavelengths slightly greater than the folds within the matrix blueschists.

Similar to Kini, the Agios Dimitrios blueschists consist of alternating glaucophane and epidote-rich layers and porphyroblasts of garnet, phengite, rutile, chlorite, and minor quartz and titanite (Fig. 9). Local concentrations of eclogitic minerals omphacite and garnet are contained within the blueschist foliation and appear to be in textural equilibrium with the surrounding blueschist in some samples, whereas in others, glaucophane (±white mica) appears to be replacing and/or overgrowing brittlely distended crystals of omphacite (Figs. 9A–9C). Glaucophane and epidote grain sizes range from 100 µm–2 mm, and porphyroblasts of garnet and omphacite range from 400 µm–2 mm (Fig. 9). Glaucophane and epidote preserve evidence for intracrystalline plasticity similar to Kini blueschists, including weak undulose extinction, grain-size reduction to a uniform size of ∼250 µm, and subgrain boundaries in larger ∼1-mm-sized crystals oriented parallel and perpendicular to the foliation (Fig. 9).

Fabrikas is an exposure of the middle subunit and is composed of blueschist-to-greenschist–facies metabasites, quartz-rich schists, and marbles. This mixed marble-metabasite-schist sequence is derived from Triassic–Jurassic rift-related volcanics and passive margin sediments (Bonneau, 1984; Bröcker and Pidgeon, 2007; Papanikolaou, 2013). The outcrop we focus on spans 25 m along the southeast Syros coastline, but similar exposure continues for ∼600 m. Outcrops similar to Fabrikas are peppered throughout a dominantly marble and calc-schist matrix that occupies the majority of the island (Fig. 1). The Fabrikas locality is preferentially eroded by waves, resulting in a unique exposure of relatively resistant, coarse-grained meta-mafic pods that physically stand out from the surrounding, more strongly eroded fine-grained matrix of alternating blueschist and quartz-mica schist. Blueschist and quartz-schist foliations dip gently to moderately to the N, and poles to foliations define girdles that are roughly N-S, with fold hinges shallowly plunging E-W (Fig. 3A). An outcrop-scale extensional fabric is well developed, such that meta-mafic lenses and matrix rocks both preserve strong E-trending, shallowly plunging glaucophane mineral lineations (Fig. 3A), and all rock types preserve strong, asymmetric top-to-the-E senses of shear (Figs. 10A–10D). Similar lithologies and structural relationships are seen at the Delfini peninsula (west-central coastline), although matrix rocks at Delfini are dominantly greenschist.

As at Agios Dimitrios, Fabrikas records deformation during subduction to eclogite facies now preserved in local pods and lenses, but the majority of the outcrop-scale structural features likely record significant strain under retrograde blueschist-to-greenschist–facies conditions. This interpretation is based on: (1) preservation of eclogite-facies lenses encased in a blueschist matrix; these lenses are affected by incipient greenschist-facies mineral replacement (Ridley, 1982; Rosenbaum et al., 2002; Bond et al., 2007); (2) E-W–trending mineral lineations and fold hinges consistent with Eocene exhumation-related kinematics (Gautier and Brun, 1994; Jolivet et al., 2013; Laurent et al., 2016); (3) retrogressive blueschist-facies dilational shear veins crosscutting eclogite pods (this study, Figs. 10A and 10B); and (4) its structural position immediately beneath the Vari detachment, interpreted as the “subduction channel roof” during early stages of exhumation (Laurent et al., 2016).

Similar to Kini and Agios Dimitrios, heterogeneity at Fabrikas occurs as meta-mafic lenses encased within the surrounding matrix, except in this case, the matrix is dominantly metasedimentary. As at Agios Dimitrios, the rocks at Fabrikas are retrogressed to blueschist-greenschist facies. Because of the unique exposure at Fabrikas, heterogeneities could be measured in three dimensions. The meta-mafic lenses range in equivalent circle diameter from ∼5 cm to 4 m, corresponding to tens of cm³ to tens of m³ in volume, with the majority of pods in the ∼1–15 m³ range (Fig. 3D), and aspect ratios typically ranging between 2:1:1 and 3:1:1 (l:w:h) (Fig. 3B and Fig. 10). Spatial distributions commonly have two modes: pods in the less than tens of cubic centimeters range are separated by cm to several m, and pods in the less than tens of cubic meters range are separated by tens of centimeters to several meters, both along the stretching direction and within the plane of the foliation (Figs. 10D and 10E). Many of the meta-mafic lenses are fractured boudins. Mineralized boudin necks are oriented nearly perpendicular to the surrounding matrix foliation, and pods are crosscut by multiple generations of dilational veins and mineralized shear fractures, oriented at 35°–65° angles to the foliation and filled with quartz, glaucophane, phengite, ± actinolite and chlorite (Figs. 3C, 3D, and 10). Displacements on dilational shear fractures range from ∼1 cm to ∼25 cm (Fig. 3D) but are usually in the cm range. In most pods, an earlier generation of quartz veins has been transposed into parallelism with the internal eclogite foliation; these older veins are crosscut by the younger, higher-angle veins (Fig. 10B). Brittle fractures in eclogites are commonly continuous with viscous shear bands in the surrounding blueschist- and quartz-schist matrix (Fig. 10B). In some places, quartz veins cut through matrix rocks near the margins of eclogite pods and preserve at least three generations of vein injection followed by progressive rotation into the foliation (Fig. 10C).

The matrix at Fabrikas comprises layers of garnet-mica-quartz schists and garnet-omphacite blueschists that alternate on the centimeter to decimeter scale. The quartz-rich schists preserve alternating ∼1-mm-thick layers of white mica and quartz, with grain sizes ranging from 40 to 400 µm (Fig. 11). C′ shear bands decorated with recrystallized phengite, paragonite, and chlorite are commonly developed in more micaceous regions. Porphyroblasts of glaucophane and epidote 200 µm–5 mm are aligned within the foliation, and garnets 400 µm–12 mm in size are dispersed throughout (Figs. 11A and 11B). Glaucophane and epidote crystals are commonly brittlely boudinaged and partially replaced by green amphibole and/or chlorite (Figs. 11A and 11B). In Figure 12, we present EBSD data for one quartz-mica schist and one blueschist from Fabrikas. Quartz aggregates preserve moderate shape-preferred orientations and abundant evidence of intracrystalline plasticity in the form of sweeping undulose extinction, subgrain formation at multiple orientations, highly irregular or “amoeboid” grain boundaries, new, small, recrystallized grains, and a moderate-strength lattice-preferred orientation indicative of basal <a>, rhomb <a>, and prism <a> slip under plane strain conditions. The blueschists are dominantly glaucophane, omphacite, and epidote, and minor phengite, rutile, and calcite, with grain sizes between 150 µm–1 mm. Poikiloblastic garnet and epidote range in size from 1 to 3.5 mm (Figs. 11D and 11E). Epidote porphyroblasts are both aligned parallel to the matrix and rotated obliquely within the foliation. Garnet and epidote porphyroblasts contain inclusions of omphacite and glaucophane. As at Kini and Agios Dimitrios, the matrix preserves evidence for intracrystalline plasticity in glaucophane and epidote in the form of undulose extinction, subgrain walls oriented both parallel and perpendicular to the foliation, bulging grain boundaries indicative of grain boundary migration, and strong LPOs (Fig. 12).

Our observations from Syros highlight two sources of rheological heterogeneity: (1) variations in rock type in the subducted protoliths; and (2) more minor textural and geochemical variations in a single oceanic-affinity protolith that are amplified during metamorphism (Fig. 13).

The cumulate-textured gabbroic pods at Kini and Agios Dimitrios are examples of textural variations that are inherited from the oceanic protolith and that persist to at least 50 km depth. Prior to subduction, fine-grained extrusive volcanic rocks and sheeted dikes close to the seafloor are highly altered by high-temperature, amphibole-bearing hydrothermal veins (up to ∼10% by volume) and weakened by low-temperature clay minerals and (sub) greenschist-facies metamorphism (Bach et al., 2003; Schramm et al., 2005). In contrast, coarse-grained lower-crustal gabbros record significantly less alteration by high-T phases than the upper volcanics (up to ∼2% by volume), and only achieve low-T alteration in local zones of high fracture intensity (Bach et al., 2001). Therefore, the weaker uppermost volcanics likely localize strain early on in their subduction history (e.g., Rubie, 1983; Hacker, 1996), permitting the stronger coarser-grained gabbros to persist as rigid clasts within the weaker matrix (Fig. 13, insets C and D). Examples of meta-gabbros in a more deformed, fine-grained mafic or ultramafic matrix are preserved in many other exhumed subduction shear zones globally, suggesting that this form of heterogeneity on the deep interface may be quite common (e.g., Kurz et al., 1998; Rawling and Lister, 2002; Angiboust et al., 2011). Because gabbroic rocks are intrusive and by definition form deeper (≥2 km) within an oceanic crustal column than basaltic volcanics, this source of heterogeneity is likely to be most common in subduction zones where the volcanic and pelagic component of the oceanic crust is relatively thin and trench sediment supply is low, such that slightly deeper sections of the oceanic crust can be incorporated into the subduction interface shear zone. Alternatively, in a hyper-extended margin or magma-poor rift setting, gabbros may be exposed on the seafloor through exhumation along oceanic core complex detachments (Péron-Pinvidic et al., 2007; Masini et al., 2013). In this scenario, discontinuous igneous pockets are surrounded by exhumed, hydrated subcontinental mantle (Beltrando et al., 2014) (Fig. 13, cross sections). The juxtaposition of different rock types on the seafloor sets up another type of rheological contrast, such that weaker serpentinites may localize strain around stronger gabbro pods during subduction (Fig. 13, insets C and D). Incorporation of gabbros into the subduction shear zone can be achieved, therefore, by detachment of a normal crustal sequence from the subducting plate along a rheologically weak layer such as serpentinite and/or a zone within the lower crust affected by fracturing and alteration prior to subduction, or by subduction of a hyper-extended margin. Alternatively, mafic intrusive igneous rocks can potentially be incorporated into the subduction channel via erosion of the forearc in subduction erosion environments (Cloos and Shreve, 1988; von Huene and Scholl, 1991; Stern, 2011).

The eclogitic lenses in a blueschist matrix at Kini demonstrate that even in a single basaltic protolith, substantial rheological heterogeneity develops as a result of partial dehydration to form eclogite (cf. Behr et al., 2018). Behr et al. (2018) interpreted the Kini rocks to record deformation at the blueschist-to-eclogite transition, with the transition between the two likely reflecting minor differences in hydrothermal seafloor alteration prior to subduction (e.g., Spandler et al., 2004). More generally, however, several other factors can lead to “patchy” dehydration and/or preferential metamorphism such that the eclogite reaction is incomplete, including differences in (1) basaltic protolith grain size (Hacker, 1996); (2) bulk chemistry or water content due to magmatic differentiation (El-Shazly et al., 1997); and/or (3) fluid persistence due to the presence of an impermeable upper plate (Abers, 2005). This source of rheological heterogeneity is likely to be most common in sediment-poor subduction zones where a chemically and texturally heterogeneous volcanic upper crust becomes incorporated into the subduction interface shear zone (Fig. 13, cross sections).

The mixed-lithology heterogeneities preserved at Fabrikas reflect coeval subduction of sediments and mafic oceanic crust. Interestingly, however, differences in these protoliths do not appear to produce rheological heterogeneity in the viscous regime unless, or until, the mafic rocks are metamorphosed to form eclogite. We reason this to be the case based on the uniform strain exhibited by layered quartzites and mafic blueschist at the outcrop scale at Fabrikas (Fig. 10), which suggests the different rock types have similar viscosities at blueschist-facies conditions. This is in contrast to exhumed shear zones from the shallow subduction interface, where sediments and basaltic lenses show strong rheological contrasts (e.g., Kimura and Mukai, 1991; Meneghini and Moore, 2007; Fagereng and Diener, 2011). Because the Fabrikas locality records stages of exhumation, it also demonstrates the persistence of eclogitized metabasites as rheologically strong features in the subduction channel during retrogression and return flow. The length scales and distributions of these mixed-lithology heterogeneities may vary substantially, however, as a function of: (1) the ratio of subducted sedimentary to mafic rocks; (2) the viscosity of the matrix; and/or (3) the amount of finite strain accumulated (Lloyd and Ferguson, 1981; Jessell et al., 2009; Abe and Urai, 2012; Schmalholz and Maeder, 2012). Eclogite pods contained within metasedimentary matrix rocks have been described for several other HP terranes; their sizes range from tens of centimeters to tens of meters in diameter, similar to the length scales observed on Syros (Lardeaux and Spalla, 1991; Gao et al., 1999; Davis and Whitney, 2008). This type of mixed-lithology heterogeneity is likely quite common in subduction zones, since many show abundant evidence for sediment subduction to the depths of arc magma genesis (Tera et al., 1986; White and Dupré, 1986; Plank and Langmuir, 1993).

Geophysical models describing the mechanical behavior of the deep subduction interface commonly approximate the plate boundary as a planar frictional fault (e.g., Shimamoto, 1986; Shibazaki, 2003; Liu and Liu, 2005; Rubin, 2008). However, the structural and microstructural observations from Syros suggest that the bulk-rock creep mechanism in the subduction shear zone was distributed power-law viscous flow accommodated by dislocation creep in both mafic oceanic-affinity blueschists and quartz-rich metasediments. Evidence for this includes the abundant observations of intracrystalline plasticity, grain boundary migration, dynamic recrystallization, and lattice-preferred orientations discussed in Section 3.

In many other subduction complexes, especially those deformed at lower temperatures, carbonate and quartz-rich metasedimentary rocks preserve evidence for Newtonian viscous creep during subduction accommodated by pressure solution and/or granular flow (Stöckhert et al., 1999; Stöckhert, 2002; Behr and Platt, 2013; Wassmann and Stöckhert, 2013; Fagereng and den Hartog, 2016; Platt et al., 2018). On Syros and farther south in Crete, the preservation of unstrained aragonite pseudomorphs in meta-carbonate packages, well-developed dissolution seams, and strong shape-preferred orientations of calcite-aragonite marbles indicate that pressure solution creep was the dominant deformation mechanism during subduction (Stöckhert et al., 1999). Strong c-axis preferred orientations in aragonite pseudomorphs indicate that at or near peak conditions, marbles entered the dislocation creep deformation regime (e.g., Brady et al., 2004). Similarly, quartz-rich rocks on Syros described here display pressure-solution fabrics that are subsequently overprinted by dislocation creep. Behr and Platt (2013) suggested, based on subduction complex rocks exposed in southern Spain, that as quartz-rich rocks approach peak PT conditions, thermally activated plastic mechanisms may be favored due to an increase in grain size and/or water fugacity. Additionally, as quartz-rich rocks are cycled back to shallower crustal levels during return flow (e.g., as at the Fabrikas locality) higher shear stresses and lower temperatures may favor dislocation creep, while simultaneous partitioning of water into retrogressive minerals may suppress diffusion creep mechanisms such as pressure solution (Behr and Platt, 2013; Xia and Platt, 2017). In contrast to quartz-rich rocks, flow-law parameters for common minerals in blueschists such as glaucophane and epidote are poorly constrained. However, several natural observations, including our data from this study, suggest that power-law viscous creep is a dominant deformation mechanism in blueschists during both subduction and exhumation (Kim et al., 2013; Behr et al., 2018) (Figs. 7 and 12).

Following the interpretation that subduction mechanics are best described by power-law viscous flow, in Figure 14A, we plot viscosity as a function of temperature for a variety of strain-accommodating rock types on Syros using experimentally-derived constitutive flow laws to gain insight into bulk interface rheology. Viscosities are plotted for a strain rate of 10⁻¹² s⁻¹, which is generally consistent with a shear zone thickness between ∼300 m–3 km and subduction velocities ∼1–5 cm/yr. Bulk shear zone rheology, strength, and viscosity will depend on the phases, their relative proportions, and their spatial distributions in the shear zone (Handy, 1990, 1994; Ji, 2004; Beall et al., 2019). Our observations from Syros suggest that two length scales of heterogeneities exist—at the map scale and at the outcrop scale—each defined by distinct phase mixtures, proportions, and distributions. Therefore, the “bulk” mechanical behavior of the deep subduction interface should be described as a scale-dependent property.

At the map scale, the juxtaposition of different lithologies leads to rheological heterogeneity that affects bulk rheology. In the uppermost structural unit, slivers of serpentinite and/or chlorite-talc schist several meters to tens of meters thick are located throughout and at the margins of meta-mafic blocks (e.g., Kini and Agios Dimitrios; Fig. 1); in contrast, in the intermediate structural unit, mixed quartz-schist and meta-volcanic exposures (e.g., Fabrikas) are peppered throughout a matrix of marbles and calc-schists (Fig. 1; Keiter et al., 2004, 2011). The spatial associations and relative distributions of rock types are reminiscent of framework-controlled (weak inclusions of serpentinite in a strong matrix of blueschist-eclogite; Fig. 13, box C) and matrix-controlled (strong inclusions of meta-volcanics in a weak matrix of marble; Fig. 13, box B) mixtures, respectively. However, according to phenomenological equations describing the steady-state rheology of two-phase mixtures, bulk shear zone viscosity will rapidly approach that of the volumetrically dominant phase (if the inclusions comprise ∼30% by volume or less) (e.g., Handy, 1990, 1994; Ji, 2004).

These inferences are consistent with structural observations in the Syros rocks (cf. Section 3). For example, serpentinites and chlorite and/or talc schists likely did not exert primary control on interface rheology for the uppermost unit. Serpentine has a plastic yield strength and effective viscosity ∼1–2 orders of magnitude lower than eclogite (and presumably blueschist) (Fig. 14A) and would preferentially localize strain if possible, yet blueschists accumulated multiple generations of penetrative strain during subduction indicating they were the rheologically dominant phase. Therefore, dislocation creep in framework meta-basalts (e.g., glaucophane) and a viscosity between that of quartz and eclogite (but much closer to quartz, e.g., 8 × 10¹⁸ < η << 8 × 10²⁰ Pa-s at ∼500 °C) may best approximate the rheology of subducted oceanic crust under blueschist-facies conditions. Exceptions presumably exist for the extreme circumstance where serpentinite is present in much greater proportions than seen on Syros (e.g., a subducted hyper-extended margin). In contrast, quartz and marble have similar viscosities during subduction to peak conditions (Fig. 14A), and strain would be partitioned locally into both phases. However, the volumetric dominance suggests that dislocation creep in matrix marble may best approximate the bulk rheology and viscosity in the intermediate unit on Syros (approaching ∼5 × 10¹⁸ Pa-s at ∼500 °C).

At the outcrop scale, the juxtaposition of different metamorphic rock types and lithologies produces heterogeneity that will affect bulk rheology. Here, with the goal of a more generalized comparison between Syros and active subduction zones, we envision the three exposures discussed in Section 3 as analogs for subducted oceanic crust (Kini and Agios Dimitrios) and subducted oceanic crust and sediments (Fabrikas) (cf. Section 4.1). At Kini and Agios Dimitrios, eclogite pods distributed in a blueschist matrix are traceable throughout the outcrops (cf. Sections 3.1 and 3.2; Figs. 4 and 8); the proportion of the strong phase (eclogite) is ∼1%–10%. At Fabrikas, however, eclogite pods are distributed more densely through a mixed blueschist-quartz schist matrix (cf. Section 3.3; Fig. 10), approaching ∼30% on average. A significant proportion (∼30% or greater) of strong, competent clasts relative to weak matrix rocks may lead to the development of force chains and clast interactions, effectively amplifying shear stresses, and significantly increasing the bulk shear zone viscosity (e.g., Fagereng and Sibson, 2010; Beall et al., 2019). Stress amplification within “clasts” (eclogites) may lead to the onset of brittle failure in the clasts, even in the absence of elevated pore fluid pressures (Beall et al., 2019). Therefore, concentrations of eclogitic heterogeneities in subducted oceanic crust with or without sediments may approach critical thresholds wherein shear zone jamming and stress amplification occurs across the interface, leading to higher bulk viscosities and perhaps complex transient behavior.

Geodynamic models of subduction dynamics have illustrated the profound effects of non-Newtonian rheology and absolute viscosity on subduction channel morphology, rates and styles of exhumation, and even subduction plate speeds. For example, Gerya et al. (2002) show that non-Newtonian rheologies (and relatively high effective viscosities, ∼10¹⁸ Pa-s) in the channel favor the formation of large-scale, self-organizing, coherent flow patterns, which are reminiscent of nappe structures characteristic of many exhumed subduction terranes, including Syros. Additionally, exhumation by forced return flow in a subduction channel is predicted to occur much faster—on the order of plate velocities—for Newtonian, low-viscosity shear zones (Gerya and Stöckhert, 2002). Furthermore, non-Newtonian, higher-viscosity shear zones are capable of transferring stress from the subducting to the overriding lithosphere more efficiently, leading to greater degrees of mechanical coupling (e.g., Androvičová et al., 2013; Agard et al., 2018) and relatively slow subducting plate speeds (∼2–4 cm/yr) compared to a Newtonian, lower-viscosity shear zone (>10 cm/yr) (King and Hager, 1990; Behr and Becker, 2018). In all of these cases, the predominance of dislocation creep in quartz schists and mafic schists is likely to contribute to complex evolution in subduction zone structure and dynamics, as different materials with evolving rheological properties cycle through the subduction shear zone.

In addition to steady-state dynamics and bulk interface viscosity, power-law viscous creep—as opposed to rate-and-state friction along a planar fault—along the subduction interface is important to understanding rock response to transient stress fluctuations, such as intermediate depth earthquakes and ETS, for several reasons. Firstly, power-law creep mechanisms will affect the rates and spatial distributions of host-rock stress loading and postseismic relaxation, owing to their high stress exponents (n = 3–4) (Poirier, 1985; Hirth and Kohlstedt, 2003). Secondly, the mechanisms of “healing,” or damage reduction, associated with viscous creep differ from those characteristic of frictional interfaces. For example, damage reduction on faults is achieved by fluid-assisted precipitation of new mineral phases on the fault plane (e.g., Sheldon and Ord, 2005; Sheldon and Micklethwaite, 2007), whereas damage reduction in the viscous regime is commonly achieved by grain growth processes (Handy, 1989; Kameyama et al., 1997). Thirdly, distributed viscous shear in the bulk rock should influence the rates and pathways of fluid migration, through, for example, different diffusion mechanisms and/or through the development of fabric anisotropy (e.g., Kirby, 1985).

Independent of the host-rock creep mechanisms, the three exposures on Syros each preserve structural relationships that record multiple generations of dilational, mineralized shear fractures (e.g., “dilational hydro-shears” documented by Fagereng, 2011) in meta-mafic heterogeneities in the presence of fluids. Key lines of evidence for this include crosscutting vein generations filled with a variety of high-pressure minerals. At Kini, for example, glaucophane, phengite, and chlorite occupy veins forming under blueschist-facies conditions. Additionally, vein-filling phengites at Kini have relatively high Si per formula unit values (Behr et al., 2018) that are comparable to those reported for blueschist-facies metamorphism elsewhere in the Cyclades (e.g., Putlitz et al., 2005; Bröcker et al., 2013), which indicate they are growing (and veins are opening) under relatively high pressures consistent with blueschist-facies conditions. Semi-brittle, low-Na omphacite porphyroblasts are fringed by asymmetric pressure shadows with higher-Na omphacite, and Riedel shear fractures are decorated with high-Na omphacite and glaucophane, indicating coupled brittle-viscous shear occurred in Kini blueschists under increasing pressure conditions during subduction (cf. Figs. 6A and 6B). At Agios Dimitrios, the oldest blueschist-facies veins (glaucophane + epidote) are crosscut by younger greenschist-facies veins (actinolite ± epidote), which indicate repeated vein cycles during exhumation (cf. Section 3.2, Fig. 8). At Fabrikas, glaucophane-filled veins indicate eclogites are transiently brittle under retrogressive blueschist-facies conditions (cf. Section 3.3, Fig. 10). At all localities, older generations of veins are commonly rotated into parallelism with the dominant matrix and crosscut by younger, higher-angle veins indicating fracture and vein-fill events occurred in multiple cycles through time (cf. Section 3).

The structural relationships also suggest that transient brittle deformation in meta-mafic heterogeneities is synkinematic with viscous shear in the surrounding host rock. At Kini, for example, evidence for coeval brittle-viscous deformation in eclogitic lenses is documented by brittle shear that decays into viscous shear in surrounding blueschists. Mixed-mode I–III veins open perpendicular to the surrounding blueschist foliation, indicating the fractures are synkinematic with respect to blueschist fabric development. Additionally, foliation-parallel thinning of mineralized boudin necks indicates brittle veining is followed by continuous fabric development in the viscous regime (cf. Figs. 4F and 4G). At Agios Dimitrios, brittle deformation is followed by viscous deformation, as exemplified by some vein sets that are sheared after emplacement (cf. Fig. 8C). At Fabrikas, evidence for coeval brittle-viscous shear is the continuity of high-angle fractures with components of both dilation and shear within eclogite pods and viscous shear bands in the surrounding host rocks (cf. Fig. 10B). At the outcrop scale, the asymmetry of structures in eclogites suggests that they were affected by both viscous (e.g., 10A) and brittle deformation (e.g., Figs. 10B and 10D) during top-to-the-east shearing. Evidence for continuous viscous flow bracketing transient brittle failure is also demonstrated in Figures 4F and 10C, which illustrate several generations of Mode-I veins injected perpendicular to the foliation and progressively transposed parallel to the foliation through time.

The mechanics of this mixed-mode (brittle-ductile and/or frictional-viscous) system can be approximated using experimentally derived constitutive relationships for meta-mafic rocks and metasediments, respectively. In Figure 14B, we plot fields showing deformation modes as a function of temperature and pore fluid factor (ratio of pore fluid pressure to lithostatic pressure, λ) for a quartz-eclogite system deforming by dislocation creep for quartz (Hirth et al., 2001) and eclogite (Zhang and Green, 2007), in a thrust-dominated regime. The plot shows that for the peak temperature conditions on Syros, best represented by the Kini outcrop, elevated pore fluid factors of >0.7 are required to produce brittle deformation in eclogite. Additionally, the presence of mineralized veins suggests that the pore fluid pressures must be nearly lithostatic, to keep veins and fractures open long enough for fluids to infiltrate and mineralize under extremely high confining pressures. For lower temperatures associated with exhumation of eclogite, as represented by the Agios Dimitrios and Fabrikas outcrops, mixed frictional-viscous deformation of quartz metasediments and persisting eclogites is anticipated for nearly any fluid pressure conditions (except for precisely lithostatic). Thus, this mixed frictional-viscous behavior is predicted for a wide range of conditions within the subduction shear zone, particularly for systems in which mafic eclogites are abundant.

To establish whether our observations from Syros may reflect heterogeneous deformation representative of the deep ETS source region, we consider three aspects of ETS as based on modern subduction zones, including: (1) the physical conditions under which tremor occurs, (2) conceptual models for ETS mechanisms, and (3) tremor and LFE source areas and magnitudes.

Episodic tremor and slow slip (ETS) appears to occur over a range of PT conditions, most commonly spanning 15–35 km depth within individual subduction zones (Dragert et al., 2001; Obara, 2002; Kao et al., 2005, 2007) but in some places extending as deep as 45–55 km (Kao et al., 2005; Obara and Hirose, 2006; Ito et al., 2007). Geophysical interpretations of the conditions of the ETS source region suggest that tremor and LFE signals are derived from within a low-velocity layer (LVL), or a zone with seismically high Vp/Vs, interpreted as hydrated meta-stable oceanic crust and/or the presence of fluids (Shelly et al., 2006; Wang et al., 2006; Abers et al., 2009; Bostock et al., 2012). Large uncertainties in LFE relocations produce wide apparent vertical distributions of seismic tremors throughout the forearc crust and might suggest tremor does not occur exclusively along the down-going subducting slab, but also within the overriding plate (e.g., Kao et al., 2005, 2007; Peterson and Christensen, 2009). Furthermore, ETS is sensitive to external tidal forces, suggesting the system is critically stressed and under high pore fluid pressures (Rubinstein et al., 2008), and P-wave first arrivals demonstrate that most tremor involves shear slip at or near the plate interface (Obara 2002; Rogers and Dragert, 2003; Dragert et al., 2004; Ide et al., 2007; Shelly et al., 2007a).

Our observations from Syros are consistent with these interpretations. The three different outcrops on Syros demonstrate that rheological heterogeneities exist over a range of PT conditions and protolith rock types. We interpret the strong heterogeneities (eclogites and metagabbros) to have formed within a subduction channel during subduction to eclogite-facies depths (e.g., Kini); we suggest they then persisted as heterogeneities during exhumation along the interface under blueschist- to greenschist-facies conditions (e.g., Fabrikas and Agios Dimitrios outcrops). The discrepancies in reported peak pressure estimates for Syros correspond to a wide range in potential subduction depths. Maximum pressures of 14–16 kbar correspond to ∼45–55 km, whereas pressures of ∼22–23 kbar correspond to ∼75–80 km depth. The former would place these rocks just downdip of the typical ETS zone (∼15–35 km), while the latter is >30 km in excess. However, regardless of the peak pressures and maximum depths these rocks reached, the rheological heterogeneities and structural features we report here were acquired during subduction to peak depths (e.g., Kini), and in some cases were continuously modified during exhumation (e.g., Fabrikas and Agios Dimitrios). Therefore, we suggest that eclogites may be capable of contributing to ETS as they pass through typical “tremorgenic” depths during both subduction and exhumation within a subduction channel along the top of the slab. Additionally, the rocks represent partially hydrated oceanic crust and sediments that have retained hydrous phases to ∼50–80 km depth, consistent with characteristics of the LVL. Finally, as discussed in Section 4.2, the deformational style in the heterogeneous lenses is consistent with brittle shear slip under high pore fluid pressures in the presence of fluids.

A common conceptual model for the production and repetition of LFEs and tremor is coseismic failure of small interface patches contained within a larger shear zone patch that otherwise creeps aseismically (e.g., Shelly et al., 2007a; Fagereng and Sibson, 2010; Beeler et al., 2014; Hayman and Lavier, 2014; Chestler and Creager, 2017b). These patches, or “asperities,” fall into two end-member categories invoking the theme of heterogeneity. The first involves contrasts in frictional properties, wherein conditionally velocity-weakening patches are surrounded by velocity-strengthening areas along a planar interface (Ito et al., 2007; Ando et al., 2012; Ghosh et al., 2012). This model is supported by friction experiments and rate-and-state–dependent friction laws (Shimamoto, 1986; Liu and Rice, 2005, 2007; Shibazaki and Shimamoto, 2007; Rubin, 2008). The second model, derived from geologic observations of exhumed shear zones, involves brittle-viscous heterogeneities in distributed shear zones with competency contrasts reflecting mixed protoliths (Fagereng and Sibson, 2010; Hayman and Lavier, 2014; Malatesta et al., 2018). High pore fluid pressures and low effective stress due to metamorphic dehydration are a likely cause of co-seismic brittle failure, corroborated by dehydration experiments on high-pressure minerals producing low-frequency seismic emissions (Burlini et al., 2009). Some specific proposed evidence for ETS in the rock record includes dilational microcracking and healing by quartz precipitation accompanied by shear slip on micaceous cleavage planes (Platt et al., 2018), mutually crosscutting foliation-parallel and foliation-oblique, quartz-filled veins that accommodate both shear and dilation (Fagereng, 2011; Fagereng et al., 2014, 2017), and mixed-mode dilational-shear fractures that offset competent lenses within dynamically recrystallized matrices (Hayman and Lavier, 2014). All of these potential records of tremor and slow slip are alike in that they preserve localized, transient (semi-) brittle deformation assisted by fluids, coeval with viscous flow, from submillimeter to outcrop scales.

Our observations of heterogeneities on Syros are consistent with the brittle-viscous model and also support the area-limited seismic moment concept: the map areas of individual mafic units on Syros can be interpreted as former interface shear zone patches (see for example the mapped lenses for Kini, Agios Dimitrios, and Fabrikas in Fig. 1). Within each mapped patch, there are brittlely deformed heterogeneities (eclogite and meta-gabbro lenses) that could potentially represent tremor-producing sub-patches (Fig. 13). According to the conceptual model described above, the remainder of the geodetically observed slip on the surface would be accommodated by accelerated viscous creep in surrounding blueschist and quartz schist matrix rocks.

(Kanamori, 1977). We can use this relationship and our observations from Syros to estimate the maximum moment magnitudes that would be expected from the interface patches analyzed at Kini, Agios Dimitrios, and Fabrikas, if they were capable of producing tremor. These calculations require the following assumptions.

• 1. The present-day map patterns of different rock types on Syros approximate the length scales of heterogeneous patches within the interface shear zone during subduction, and the percentages of heterogeneities observed in subregion exposures are characteristic of the entire shear zone patch. Our field observations suggest that the types and distributions of documented heterogeneities are characteristic of the map areas shown in Figure 1. This allows us to estimate source area (A) by multiplying the mapped area of the lithological unit represented by the outcrop (cf. Fig. 1) by the ratio of brittle heterogeneities documented for a subregion within that outcrop (cf. Supplemental Fig. S1 and Table S4 [^{footnote 1}]).

• 2. Brittle shear slip of mafic lenses within these interface shear zone patches occurred simultaneously or within close succession. Although we cannot confirm that this was the case through field observations, it is not an unreasonable assumption because there are several mechanisms by which shear zone deformation can propagate transiently within a viscous interface shear zone, including via the generation of porosity waves (e.g., Skarbek and Rempel, 2016), the development of stress bridges among competent clasts (e.g., Webber et al., 2018; Beall et al., 2019), or viscoelastic interactions among clasts. Furthermore, this assumption is also compatible with observations of modern tremor bursts, which consist of swarms of small, low-frequency earthquakes rather than single events (Ito et al., 2007; Shelly et al., 2007b). The assumption of simultaneous failure then allows us to equate our observed displacement amounts on individual brittle lenses to total displacement over the whole interface area, A, discussed above. This assumption also means that our moment magnitude estimates are maximums.

• 3. For rock rigidity, we assume 30–40 GPa for oceanic crust at eclogite-facies depths; as reasoned by Sweet et al. (2014), 30 GPa splits the difference between normal oceanic crust at these depths (40 GPa) and values for highly anomalous upper oceanic crust with Vp/Vs ratios ∼2.4 (Audet et al., 2009).

Additional details about these calculations and how uncertainties were treated are presented in the Supplemental Text (^{footnote 1}).

Figure 15 compares our calculations of source area and observed displacement to those estimated from tremor-producing regions in active subduction zones. The gray data points in Figure 15 include LFE source areas estimated using high-resolution spatial relocation of LFE events, and displacement magnitudes calculated through inversion of the seismic moment equation, where seismic moments were estimated by stacking LFE events and comparing their amplitudes with local small earthquakes of known magnitudes (Kao et al., 2010; Sweet et al., 2014; Bostock et al., 2015; Thomas et al., 2016; Chestler and Creager, 2017b). Our displacements in Figure 15 are limited by the smallest size of documented eclogite pods and lenses for each locality. We provide equivalent circle diameters (ECDs) of calculated map and source areas in Table S5 (^{footnote 1}) to compare with similar measurements derived from geophysical data; our patch and source area ECDs range from ∼400 to 1000 m and ∼10 to 300 m, respectively, the same order of magnitude as areas calculated by Chestler and Creager (2017a), Sweet et al. (2014), and Thomas et al. (2016). Both the Kini and Agios Dimitrios localities have displacement versus source area scaling that is consistent with observations from Cascadia. The Fabrikas locality is inconsistent, perhaps suggesting that heterogeneities there record the superposition of smaller slip events. Our calculations from Syros suggest that the simultaneous or successive brittle failure of heterogeneous lenses with the documented length scales could yield equivalent seismic moments of 4.5 × 10⁹ to 4.6 × 10¹⁴ N m (Mw ∼0.3–3.7), entirely compatible with observed seismic signals in the ETS source region (Fig. 14B).

Exhumed rocks on Syros Island provide insight into the mechanical behavior of a heterogeneous subduction interface shear zone. Rheological heterogeneities associated with lithological and/or textural differences in the subducting protoliths persist to peak depths and during viscous return flow, are accentuated by eclogite-facies metamorphism, and produce cyclical brittle deformation events triggered by high pore fluid pressures. Brittle deformation is accompanied by viscous power-law flow in surrounding quartz-schist and blueschist host rocks. The Syros rocks exhibit several similarities with the physical conditions of episodic tremor and slow slip in active subduction zones as inferred from geophysical studies—namely: (1) fluid-assisted brittle shear slip in a variety of hydrous rock types over a wide range of PT conditions; and (2) tremor source areas and equivalent seismic moments. Our observations support a conceptual model of tremor-genesis in which coseismic failure occurs in small sub-patches distributed within a larger, viscously creeping interface patch.

This work was supported by a National Science Foundation (NSF) Graduate Research Fellowship awarded to A.J. Kotowski and an NSF CAREER grant (EAR-1555346) awarded to W.M. Behr. Thanks to James Maner for help with the microprobe and environmental scanning electron microscope at University of Texas at Austin. We are grateful to Danny Stockli, Kostis Soukis, Jaime Barnes, Mark Cloos, Luc Lavier, and Laura Wallace for discussions and to four anonymous reviewers for constructive comments that improved this manuscript.