Geological setting

The Southwest Indian Ridge (SWIR) lies at the slow end of the mid-ocean-ridge spreading-rate spectrum, with a full spreading rate of 14 mm/y. The ridge trends approximately southwest–northeast for most of its length, spreading almost due north–south. Between ~52°E and 60°E, it is offset by a series of closely spaced, long-offset transform faults. One of these, the Atlantis II Transform at 57°E, offsets the SWIR sinistrally by 200 km (Engel and Fisher, 1975) (Figure F1). Between this and the Novara Transform, ~140 km to the east, the spreading axis is divided into two segments, separated by a nontransform discontinuity at 57°35′E (Hosford et al., 2003). The spreading rate here is asymmetric: at the short westerly Segment AN-1, adjacent to the northern ridge/transform intersection of the Atlantis II Transform, magnetic anomalies reveal rates of 8.5 mm/y to the south and 5.5 mm/y to the north (Hosford et al., 2003); in consequence, the transform has been lengthening by 3 mm/y for at least the past 20 million years (Dick et al., 1991b; Hosford et al., 2003). At ~20 Ma, a 10° counterclockwise change in regional spreading direction put the Atlantis II Transform into transtension for a period of ~12 million years (Dick et al., 1991b, Hosford et al., 2003; Baines et al., 2007, 2008).

Immediately to the east of the Atlantis II Transform, parallel to it and on a flow-line directly south of spreading Segment AN-1, is an elevated transverse ridge (Figure F1) consisting of a series of uplifted blocks connected by saddles that rise to as little as 700 m below sea level (Engel and Fisher, 1975; Dick et al., 1991b). Atlantis Bank, at 32°43′S, 57°17′E, forms the shallowest and northernmost portion of this ridge, 95 km south of the axis of the SWIR. The bank consists of a raised dome ~40 km long by ~30 km wide, rising from 5700 m depth at the base of the transform wall to 700 m on a ~25 km2 flat-topped platform at its crest and then drops to 4300 m on its eastern flank across two prominent transform-parallel east-dipping normal faults (Baines et al., 2003; Hosford et al., 2003) (Figures F1, F2).

Hole 735B is located at the southwest corner of the flat surface of the Atlantis Bank platform, at 731 m water depth (32°43.395′S; 57°15.959′E) (Figures F2, F3). During ODP Legs 118 and 176, the hole was drilled to a total 1508 m below seafloor (mbsf), with 87% core recovery, all in gabbro (sensu lato) ~11 Ma in age (Robinson, Von Herzen, et al., 1989; Dick et al., 2000). Further operations during Leg 179 drilled the 160 m deep Hole 1105A in the center of the platform (32°43.13′S; 57°16.65′E) (Pettigrew, Casey, Miller, et al., 1999), also in gabbro and with similar overall core recovery rates.

Site surveys of the Atlantis Bank area have included multibeam, magnetic, and gravity surveys and rock sampling using seabed rock coring, remotely operated vehicle (ROV), submersible dives, and dredging (Dick et al., 1991b; MacLeod et al., 1998; Arai et al., 2000; Kinoshita et al., 1999; Matsumoto et al., 2002). A high-resolution bathymetric map of the ~25 km2 summit region compiled from narrow-beam echo soundings (Dick et al., 1999) shows the platform to consist of a broadly flat pavement ranging from ~750 to 689 m water depth). Proposed deep drill Site AtBk6 is located at the center of the platform, ~1 km north-northeast of Hole 1105A and 2 km northeast of Hole 735B (Figure F3). ROV survey results show the central pavement to be bare rock, locally knobby igneous outcrop surrounded on its periphery by a pavement of limestone (in some places ripple marked) and carbonate-cemented pebble conglomerate.

Using the British Geological Survey’s 5 m diamond rock drill and 1 m BRIDGE oriented corer, MacLeod et al. (1998) drilled 42 successful cores on the surface of this platform region, recovering 33 igneous rocks (Figure F3). They reveal continuous gabbro outcrop on the platform surface, including at proposed drill Site AtBk6 (Figure F4) but pass into serpentinized peridotite and pillow-basalt breccia at its southern tip, where the flat platform narrows into a north–south spine (32°44.5′S). The dredging and submersible studies, conducted over a broader ~700 km2 area, showed that gabbro is present for >35 km parallel to the spreading direction (Figure F2B), suggesting that accretion of a continuous gabbroic layer persisted for at least 4 million years (Dick et al., 1991a; Matsumoto et al., 2002).

Many of the gabbroic rocks in the shallow drill cores and seen in ROPOS ROV footage from the surface of the Atlantis Bank platform are mylonitic, displaying intense crystal-plastic deformation with subhorizontal fabrics. In Hole 735B, amphibolite facies crystal-plastic deformation of oxide gabbro and olivine gabbro was likewise found to be strong in the uppermost 500 m but diminished markedly downhole (Robinson, Von Herzen, et al., 1989; Cannat et al., 1991; Dick et al., 1991a; Miranda and John, 2010). Considered together, it is apparent that the upper surface of Atlantis Bank represents a detachment fault zone, effectively a high-temperature “plutonic growth fault” active (for >4 million years) during the accretion of the gabbroic lower crust and responsible for its exhumation. This “hot”’ detachment must have rooted in or near a melt lens at the top of a continuously replenished magma chamber/mush zone, close to the midpoint of the AN-1 spreading ridge (Dick et al., 1991a, 2000; Natland and Dick, 2001, 2002).

Oceanic detachment faults are now known to be widespread along slow- and ultraslow-spreading ridges (e.g., Escartín et al., 2008; Sauter et al., 2013), often responsible for exhuming elevated oceanic core complex massifs that expose mantle peridotite and/or gabbro on the seafloor. Oceanic core complexes at the Mid-Atlantic Ridge (MAR) typically display prominent spreading-direction-parallel corrugations on their flat-topped or domal upper surfaces (Cann et al., 1997; Tucholke et al., 1998). Sampling of fault rock from the detachment fault zones themselves shows that deformation typically localized on weak, low-temperature secondary minerals such as talc, derived primarily from alteration of peridotite (Dick et al., 2001, 2008; MacLeod et al., 2002; Escartín et al., 2003; Schroeder and John, 2004). Where gabbro is present in the detachment footwall it may be unaffected or barely affected by crystal-plastic deformation (e.g. 15°45′N on the MAR: MacLeod et al., 2002, and Escartín et al., 2003; Atlantis Massif: Ildefonse et al., 2007), whereas elsewhere it is heavily deformed, as at Kane Megamullion on the MAR (Dick et al., 2008) and at Atlantis Bank.

Submersible observations at Atlantis Bank (Kinoshita et al., 1999; Matsumoto et al., 2002) show that the detachment fault surface is preserved over large regions in the deeper waters on both sides of the platform. The damage zone and underlying gabbro are well exposed by high-angle normal faulting on the eastern side of the complex and in headwalls of large landslips on the western flank. Here, lower temperature fault rocks are present in addition to the amphibolite facies mylonite. These chloritized and weathered fault gouge and talc-serpentine schists, preserved locally on the fault surface (Dick et al., 2001; Miranda and John, 2010), are closely comparable to fault rocks found on the MAR core complexes (e.g., MacLeod et al., 2002; Escartín et al., 2003, Schroeder and John, 2004; Dick et al., 2008).

The absence of the low-temperature fault rocks on the flat surface of the Atlantis Bank platform compared to its flanks suggests they have been removed, a deduction entirely consistent with the long-held belief (Dick et al., 1991b; Palmiotto et al., 2013) that Atlantis Bank was once an ocean island, and its flat top results from erosion, before later subsiding to its present level. Shallow drilling and dredging on the summit and uppermost flanks of the Atlantis Bank platform amply verified this. Indurated carbonates were drilled at 24 sites from the periphery of the Atlantis Bank platform and were also recovered in dredges from its flanks (MacLeod et al., 1998, 2000; Palmiotto et al., 2013). Although some (probably recent) carbonate sand was recovered, most of the sediment is indurated bioclastic limestone (skeletal packstone to wackestone; MacLeod et al., 2000) of Miocene to Pleistocene age (Palmiotto et al., 2013). This sediment contains an abundant macrofauna, primarily bryozoans, mollusks, algal nodules, and echinoids but also some solitary corals. Benthic and, in some instances, large planktonic foraminifers are common. Whereas Palmiotto et al. (2013) suggest a water depth of ~100–200 m based upon assemblages in dredged carbonates from the flanks of the platform, green algal (dasyclad) assemblages in drill cores from the platform summit (G. Della Porta and V.P. Wright, pers. comm., 1999) indicate water depths at wave base or shallower, demonstrating that the platform was at sea level, and probably above. How much material has been removed by subaerial erosion is not known, potentially 100–200 m with reference to the inferred thickness of the detachment fault damage zone. On the basis of pitted gabbro outcrops to ~1500 m below sea level that may reflect preferential subaerial weathering of plagioclase, Palmiotto et al. (2013) speculate that the Atlantis Bank island could have been elevated to as much as 1 km above sea level. ROV observations of steep sides up to 10 m, locally even 50 m, at the flanks of the gabbro pavement on the summit of the platform may represent coastal cliffs and small sea stacks resulting from the dispersal of wave energy on the flanks of the island during erosion.

The Atlantis II Transform transverse ridge, on which Atlantis Bank lies, is clearly analogous to oceanic core complex massifs found on the MAR (e.g., Cann et al., 1997; Tucholke et al., 1998; Dick et al., 2008), although here, as elsewhere along the SWIR (Cannat et al., 2006; Sauter et al., 2013), the prominent spreading-direction-parallel corrugations that are characteristic of the flat surfaces of the Atlantic oceanic core complexes are not so obvious and potentially present only on the down-dropped terrace on the eastern side of Atlantis Bank (Figure F2A). The Atlantis II Transform transverse ridge has, furthermore, clearly been uplifted far above the surrounding seafloor. Whereas flexural uplift of detachment fault footwalls to form oceanic core complexes is typically 1 ± 0.5 km (Tucholke et al., 1998; Lavier et al., 1999), the Atlantis II Transform transverse ridge has been uplifted by 3 km relative to surrounding seafloor of similar age (Baines et al., 2003). Dick et al. (1991b) and Baines et al. (2003) propose that the original detachment-related uplift at the ridge/transform intersection was accentuated by an additional phase of flexural uplift, imparted upon the transform in response to the change of spreading direction on the SWIR at 19.5 Ma, and accommodated by the reactivation as normal faults of originally transform-related north–south structures.

One consequence of the relative uplift of the Atlantis II transverse ridge is that the crust/mantle boundary, as mapped by dredging and submersible traverses, is exposed along its western wall for a distance of nearly 40 km (Figure F2B). This boundary was traversed at two locations by Shinkai 6500 Dives 466 and 458 at 4500 and 4650 m water depth, respectively. Dives 467 and 459 immediately above each of them found the detachment surface at 3000 and 1755 m, respectively, potentially indicating gabbro layer thicknesses as little as 1500 and 2895 m. Elsewhere along the western flank of Atlantis Bank, serpentinized peridotite was recovered in dredges from depths as shallow as 2000 m water depth (Dredge JR31-DR8), though in other nearby places only at >3000 m (Figure F2B). At the southern end of the platform itself, serpentinized harzburgite was drilled at 839 mbsf (Site JR31-BGS12). Peridotite sampled at several locations above the contact mapped on the transform wall consists largely of talc-serpentine schist that likely overlies the gabbro massif. This and serpentine pebbles found in beach conglomerates overlying gabbro near the crest of the platform indicate that a discontinuous talc-serpentine sheet was associated with the detachment fault at shallow levels and is now locally preserved on the detachment surface. The serpentinite drilled at the southern end of the platform, however, is relatively massive harzburgite with well-preserved pseudomorphs of pyroxene. Possible origins of these include a peridotite screen in gabbro, similar to those drilled at Atlantis Massif in Hole U1309D, an enclave of less deformed and altered serpentinized peridotite in the original detachment fault shear zone, or juxtaposition through a northward-dipping ridge-parallel normal fault that demarcates the southernmost end of the flat platform at 32°44.4′S (Figure F3).

A wide-angle seismic refraction survey of the Atlantis II Transform region found the Moho at 5 ± 1 km beneath Atlantis Bank (Muller et al., 1997, 2000; Minshull et al., 1998) (Figure F5). The direct geological constraints outlined above offer strong support to the inference that the seismic Moho cannot therefore represent the crust/mantle boundary in this region. Whereas Muller, Minshull, and colleagues also concluded that it was likely a serpentinization front, they based this conclusion on the geochemical argument that the original igneous crustal thickness there was originally ~4 km (based upon rare-earth element inversions), and with the basaltic carapace removed by detachment faulting, the remainder was likely to be ~2–2.5 km thick. However, they admit this interpretation is nonunique, primarily because of the overlap in P-wave seismic velocity between gabbro and ~20%–40% partially serpentinized peridotite (e.g., Miller and Christensen, 1997).

At what level is the crust/mantle boundary likely to lie beneath Atlantis Bank? Projecting the detachment surface to the locations of the traversed crust/mantle boundary described above indicates that the crustal thickness at these points prior to mass wasting on the transform wall was significantly <2000 m, whereas the depth to Moho below the transform wall remains ~5 km (Figure F5). Given that the elevated core complex massif is produced by flexure during uplift into the rift-mountains due to a spreading direction change (Baines et al., 2003), it seems reasonable to conclude that the mapped boundary is most likely to project approximately subhorizontally beneath the center of Atlantis Bank, consistent with the igneous crustal thickness of 2–2.5 km suggested by Minshull et al. (1998).