IODP Proceedings    Volume contents     Search

doi:10.2204/iodp.proc.345.101.2014

Background

Crustal accretion models for fast-spreading crust

Two end-member models for crustal accretion at fast-spreading ridges have emerged from geophysical observations in the modern oceans and geological evidence derived from the Oman and other ophiolites (e.g., Nicolas et al., 1988) (Fig. F2). Seismic reflection studies at the EPR identify axial melt lenses <1–2 km below the seafloor that are ~1 km wide and a few tens of meters thick (Detrick et al., 1987; Hooft et al., 1997; Kent et al., 1990; Singh et al., 1998). A low-velocity zone that underlies axial magma chambers (AMCs) is interpreted to be partially molten, containing <20% melt (e.g., Dunn et al., 2000). The internal structure of this region, such as the distribution of melt and its geometry, is not well constrained because of, for example, the uncertainties of converting compressional wave velocity with varying melt distribution. Locally, melt has been shown to pool at or below the Mohorovicic discontinuity (Moho), both on- and off-axis, and also within the lower crust off-axis (Garmany, 1989; Crawford and Webb, 2002; Durant and Toomey, 2009; Canales et al., 2009).

The geological framework for the crustal accretion models is largely based on the Oman ophiolite, in which there are significant contrasts between upper and lower gabbro in terms of its composition and deformation fabrics (e.g., Pallister and Hopson, 1981; Nicolas et al., 1988; Coogan et al., 2002b). The lava sequence, sheeted dike complex, and upper gabbro have mafic phases and calculated melt compositions that are not consistent with direct derivation from liquids in equilibrium with the upper mantle, as expressed, for example, by their Mg#. In the lower gabbro, compositions are more primitive and range between those of the upper gabbro and melts in equilibrium with the upper mantle.

The first crustal accretion model is the gabbro glacier model (Henstock et al., 1993; Quick and Denlinger, 1993; Phipps Morgan and Chen, 1993), in which most crystallization occurs within a shallow melt lens and the resulting crystal mush subsides downward and outward by crystal sliding to generate the full thickness of the plutonic layer (Fig. F2A). The latent heat of the plutonic crust is largely lost to the overlying hydrothermal system through the melt lens on-axis. The alternative sheeted sill model (Kelemen et al., 1997; Korenga and Kelemen, 1997, 1998; MacLeod and Yaouancq, 2000) predicts that almost all of the lower crust crystallizes in situ in a sheeted sill complex, such that melts pond repeatedly as they are transported through the lower crust, with crystallization occurring from the Moho to the upper melt lens (Fig. F2C). The heat delivery to the lower crust is so high that conductive mechanisms alone cannot remove that heat, requiring extensive hydrothermal cooling along the sides of the crystal mush zone to remove the latent heat of crystallization on-axis (Chen, 2001). Hybrid models have also been proposed, in which some crystallization occurs in the AMC and some in situ within the plutonic crust (Fig. F2B) (Boudier et al., 1996; Coogan et al., 2002b; Maclennan et al., 2004). In fact, both end-member models require some portion of each process. In the gabbro glacier model, the melt lubricating subsiding crystals, allowing them to flow, would crystallize in the deeper crust, and in the sheeted sill model, more rapid cooling at shallow levels in the crust requires some crystal subsidence to prevent the AMC from solidifying (e.g., Maclennen et al., 2004).

Predictions for a range of chemical and physical parameters for the two end-member models are shown in Figure F3. The gabbro glacier model predicts nearly constant Mg# for plutonic rock and cumulate minerals throughout the plutonic sequence, with most crystallization occurring in a nearly steady-state shallow magma chamber. The predicted increase in Mg# with depth in the sheeted sill model is based on partial in situ crystallization of magmas in the lower crust and upward migration of evolved liquids. The two models predict different changes in strain with depth, with increasing strain in the gabbro glacier model as a consequence of lower crustal flow.

Predictions for the response of the hydrothermal system are critical to the distinction of the two models. The sheeted sill lower crust requires deep, on-axis convective cooling to remove the latent heat, resulting in a higher hydrothermal fluid flux and a greater overall intensity of hydrothermal alteration in the lower crust than does the conductively cooled gabbro glacier model. Theoretical studies focusing on the axial heat budget have investigated the viability of the lower crustal accretion models (Sleep, 1975; Morton and Sleep, 1985; Chen, 2001; Cherkaoui et al., 2003; Maclennan et al., 2004). Both accretion models, as well as hybrids of the two, have been shown to be viable (e.g., Cherkaoui et al., 2003; Maclennan et al., 2004); however, the lack of observational constraint on key input parameters (e.g., permeability) make the interpretation of these thermal models uncertain.

How do the predictions of the crustal accretion models compare with our current knowledge of gabbroic sequences from fast-spreading ridges? Shallow-level gabbros sampled at Hess Deep, Pito Deep, and Site 1256 all show evolved compositions (Hékinian et al., 1993; Perk et al., 2007; Wilson et al., 2006; Koepke et al., 2011). At Pito Deep, however, nearly all gabbro from >300 m below the base of the sheeted dike complex is much more primitive than at the same structural level at Hess Deep (Hanna, 2004; Perk et al., 2007). This compositional difference suggests that primitive, mantle-derived magma may be transported to shallow depths with little fractionation occurring along the way (Pito Deep) and that crystal fractionation and postcumulus reactions may produce evolved rocks at these depths as well (Hess Deep and Site 1256) (Perk et al., 2007). Tests of the predictions for hydrothermal processes are less complete because of the absence of observations of deeper level gabbros. What has been documented to date are rapid cooling rates in the shallow-level gabbros (Coogan et al., 2007) and calculated heat fluxes across a contact aureole (Gillis, 2008) that are indicative of substantial heat loss through the axial melt lens. Collectively, observations from Site 1256 and Hess and Pito Deeps suggest that the competing accretion models may both be viable, when spatial and/or temporal variability in magmatic processes along the EPR are considered (Coogan, 2007).

Geological setting of the Hess Deep Rift

The Hess Deep Rift is located in the vicinity of the Galapagos triple junction (ridge-ridge-ridge), at the intersection of the Cocos, Nazca, and Pacific plates (Fig. F4) (Hey et al., 1972, 1977; Holden and Dietz, 1972). The north-south–trending EPR is spreading at 130 mm/y, and the east-west–trending Cocos-Nazca Rift is propagating westwards toward the EPR at a rate of 42 mm/y. Reconstruction of the bathymetry in this region suggests that the current configuration of the Galapagos triple junction has been active for at least 10 m.y. (Smith et al., 2011). The Galapagos microplate was initiated at ~1.5 Ma. Although the cause is not known, this initiation may be related to the formation of seamounts in the vicinity of Dietz Deep (Fig. F4) (Lonsdale, 1989; Smith et al., 2011; Schouten et al., 2008).

The EPR at the latitude of the Hess Deep Rift is made up of many short, first-order ridge segments separated by second-order overlapping spreading centers (Lonsdale, 1988). Thus, the crust exposed at the Hess Deep Rift likely formed within a short segment, perhaps at a segment end (Lonsdale, 1988). Reconstructions of the EPR flanks indicate several recent episodes of migrating offsets, suggesting that the crust exposed in the Hess Deep Rift could have formed on the western side of the EPR (Lonsdale, 1989; Smith et al., 2011; Schouten et al., 2008). New high-precision zircon dates of between 1.420 and 1.271 Ma for gabbroic rock from the intrarift ridge (Rioux et al., 2012) support this reconstruction, as they place this gabbro to the west of the EPR at the time of its formation.

The Hess Deep Rift is a complex region that formed by deep crustal extension ahead of the westward propagation of the Cocos-Nazca spreading center. The surface expression of rifting is first evident ~30 km from the EPR where, from west to east, two 5 km wide, east–west grabens merge and the rift broadens to ~20 km and deepens to >5400 m at Hess Deep (Fig. F5). An intrarift ridge rises to 3000 meters below sea level (mbsl) north of Hess Deep. Further to the east, the tip of the Cocos-Nazca spreading center starts to build a volcanic ridge that spreads at 55 mm/y. Uplifted rift escarpments contiguous with the EPR crust rise to depths <2200 mbsl north and south of the rift.

A composite section of EPR crust and uppermost mantle is exposed along the rift escarpments and valley floor of the Hess Deep Rift. Multichannel reflection profiling along the EPR flanks north of the Northern Escarpment indicates a crustal thickness of 5–5.5 km (Zonenshain et al., 1980). The thickness of the upper crust ranges from 0.7 to 1.2 km (Karson et al., 2002), so the plutonic crust is ~3.8–4.8 km thick.

The subseafloor geology of the western end of the intrarift ridge and its southern slope has been investigated by on-bottom seismic and gravity surveys (Ballu et al., 1999; Wiggins et al., 1996). Results of these geophysical surveys, in combination with the regional geology, have led to a model whereby coherent crustal blocks were unroofed by detachment faulting and block rotation on listric normal faults (Francheteau et al., 1990; Wiggins et al., 1996; MacLeod et al., 1996b; Pariso et al., 1996). The viability of this model has been called into question based on new high-resolution bathymetric data collected as part of the JC21 site survey cruise for Expedition 345 (C.J. MacLeod et al., unpubl. data). Revised models consider the role of mass wasting linked with rifting (Ferrini et al., 2013) and dynamic uplift associated with the Cocos-Nazca spreading center (MacLeod et al., 2008).

The results of Expedition 345 substantiate the role of mass failure as being a dominant process along the southern slope of the intrarift ridge (see below).

Previous research at the Hess Deep Rift

The Hess Deep Rift exposes contiguous sections of the mid- to upper crust along the Northern and Southern Escarpments and sections of the mid- to lower crust along the rift valley floor. The EPR upper crustal sections include laterally extensive outcrops of the lava and sheeted dike complex that are normal, depleted MORB (Stewart et al., 2002). Lateral variation in the thickness and internal structure of the volcanic sequence and sheeted dike complex are attributed to temporal variations in magma supply (Karson et al., 2002). Hydrothermal alteration of the upper crust is largely focused in the sheeted dike complex, where the dikes are variably altered to amphibole- or, locally, chlorite-dominated assemblages (Gillis, 1995; Gillis et al., 2001).

Shallow-level gabbros are exposed along the Northern Escarpment and at the summit of the western end of the intrarift ridge. The northern scarp exposures, which directly underlie the sheeted dike complex, are dominated by gabbronorite, with lesser amounts of Fe-Ti oxide and amphibole gabbro, varitextured gabbro, olivine gabbronorite, Fe-Ti oxide gabbronorite, and rare tonalite (Hanna, 2004; Kirchner and Gillis, 2012; Natland and Dick, 1996). The shallow-level gabbro has a wide range in Mg# (mean Mg# = 0.56; range = 0.30–0.76), with gabbronorite being the most primitive and Fe-Ti oxide ± amphibole gabbro being the most evolved end-members (Hanna, 2004; Kirchner and Gillis, 2012; Natland and Dick, 1996). Rock types exposed at the summit of the western end of the intrarift ridge are similar to the Northern Escarpment, with gabbronorite, oxide gabbronorite, gabbro, olivine gabbro, and patches of pegmatitic amphibole gabbro (Fig. F6) (Gillis, Mével, Allan, et al., 1993; Hékinian et al., 1993: Lissenberg et al., 2013). This shallow-level gabbro is slightly less evolved (mean Mg# = 0.58; range = 0.35–0.68) than Northern Escarpment shallow gabbro, with the most evolved samples located along the northern margin of the intrarift ridge, close to dolerite interpreted as sheeted dikes (Blum, 1991; C.J. MacLeod, pers. comm., 2009; Pedersen et al., 1996; Gillis, Mével, Allan, et al., 1993). In both locations, shallow gabbroic rock lacks modal layering but locally contains lithologic boundaries identified by grain size variation (Gillis, Mével, Allan, et al., 1993; Hékinian et al., 1993).

Plutonic rock exposed along the central and eastern region of the southern slope of the intrarift ridge, between 4400 and 5400 mbsl, include gabbro, gabbronorite, olivine gabbro, and lesser troctolite (Fig. F6) that are, on average, more primitive than the summit of the intrarift ridge (mean Mg# = 0.71; range = 0.39–0.85) (Blum, 1991; C.J. MacLeod, pers. comm., 2009). As with the shallow-level gabbro, modal layering has not been documented for the deeper gabbro. It is important to note that the stratigraphic depth of this more primitive gabbro is not known, as these rocks likely migrated downslope by mass wasting processes.

A weak to strong magmatic foliation defined by the shape-preferred orientation of plagioclase laths is found in some samples from throughout the gabbroic sequence (Hékinian et al., 1993; Coogan et al., 2002a; Gillis, Mével, Allan, et al., 1993; MacLeod et al., 1996a). Reorientation of sections of the shallow gabbros in ODP Hole 894G to geographical coordinates shows that foliation is steeply dipping (mean dip = 69°) with a nearly north–south orientation parallel to the EPR, and that there is a steeply plunging lineation (MacLeod et al., 1996b). Measurements of the anisotropy of magnetic susceptibility in samples also show that a magnetic fabric is parallel to the plagioclase fabric (Richter et al., 1996). Crystal-plastic deformation is rare throughout the gabbroic sequence, with minor undulose extinction and very rare deformation twins in plagioclase.

Hydrothermal fluid flow throughout the entire gabbroic sequence was dominated by pervasive fluid flow along grain boundaries, microfractures, and fractures. The shallow-level gabbro and the limited suite of deeper gabbro show that incipient flow occurred at amphibolite facies (average temperature = 720°C) conditions and is manifest by amphibole veins that display no preferred orientation (Manning et al., 1996; Coogan et al., 2002a) and replacement of pyroxene by amphibole-dominated assemblages (Früh-Green et al., 1996; Gillis, 1995; Kirchner and Gillis, 2012). Whole-rock samples unaffected by a lower temperature stage of brittle deformation (see below) are depleted in δ18O relative to fresh values (Agrinier et al., 1995; Lecuyer and Reynard, 1996) and show minor enrichment in 87Sr/86Sr (data for shallow-level gabbro only) (Lécuyer and Gruau, 1996; Kirchner and Gillis, 2012). Calculated fluid/rock ratios using both isotopic systems range from 0.1 to 1 (Lécuyer and Gruau, 1996; Lecuyer and Reynard, 1996; Kirchner and Gillis, 2012). The rate of cooling of the shallow-level gabbro is rapid (1,000° to 60,000°C/m.y.) and comparable to the upper gabbro section in the Oman ophiolite (Coogan et al., 2007; Faak et al., 2011).

When the shallow-level gabbro exposed along the intraridge rift cooled to ~450°C, it became influenced by the effects of Cocos-Nazca rifting, creating a dense array of east–west tensile fractures filled with greenschist to zeolite facies assemblages and local cataclasis cemented with the same assemblage (Früh-Green et al., 1996; Manning and MacLeod, 1996). The consequences of Cocos-Nazca rifting on deeper gabbro have not yet been constrained.

Mantle peridotite is exposed in the vicinity of ODP Site 895, southeast of Site 894 (Fig. F5). Clinopyroxene-poor harzburgite, at the most depleted end of the range for abyssal peridotite, is interpreted to be the residue of melting of a normal (N)-type MORB source (Natland and Dick, 1996). The association of dunite–troctolite-olivine gabbro with depleted harzburgite records the interaction of migrating melt with depleted harzburgite wall rock in the shallowest mantle (Natland and Dick, 1996; Arai and Matsukage, 1996; Arai et al., 1997). The geometry of these associations suggests that at least some melt transport was fracture-controlled. Significant modal layering in peridotite-hosted gabbroic rocks, a feature of the Oman mantle transition zone (Korenga and Kelemen, 1997), was not observed.