IODP

doi:10.2204/iodp.sp.335.2010

Background

Developing a comprehensive understanding of the accretion of new crust at mid-ocean ridges requires samples of cumulate gabbro, as it is these coarsely crystalline rocks that hold records of the deep magmatic, hydrothermal, and tectonic processes that dominate in the lower crust. Gabbros make up >65% of the oceanic crust but lie beneath 1000 to 2000 m of lavas and dikes in typical "Penrose"-type ocean crust. Past drilling of the upper oceanic crust (e.g., Deep Sea Drilling Project [DSDP] Hole 504B; Alt, Kinoshita, Stokking, et al., 1993) has been fraught with problems of low core recovery, poor hole conditions, and equipment failure, all because of the hard and highly fractured nature of magmas erupted onto or intruded at shallow levels into the oceanic crust. The Superfast campaign developed from the idea of minimizing the thickness of upper crustal rocks to be cored before sampling gabbros.

There is an observed relationship between spreading rate and the depth of axial low-velocity zones, thought to be magma chambers, imaged by multichannel seismic (MCS) experiments at active ridge crests (Fig. F2) (Purdy et al., 1992). Reconsideration of magnetic anomalies formed at the southern end of the Pacific/Cocos plate boundary identified crust formed at a full spreading rate of ~220 mm/y from 20 to 11 Ma (Wilson, 1996) (Fig. F1). This is significantly faster than the present fastest spreading rate (~145 mm/y) for crust forming at ~20°–30°S on the EPR. From this region created by superfast spreading, a single drill site in the Guatemala Basin, initially designated GUATB-03C (Fig. F3) and now known as Site 1256, was selected on ~15 m.y. old ocean crust. At this site, gabbroic rocks were predicted to occur between 1275 and 1550 meters below seafloor (mbsf), assuming a typical carapace of 300 m of off-axis lavas (Expedition 309 Scientists, 2005; Teagle, Alt, Umino, Miyashita, Banerjee, Wilson, et al., 2006). This approach and site was endorsed by the ODP "Architecture of the Oceanic Lithosphere Program Planning Group" (1998; available online at www.iodp.org/mission-moho-workshop/) as the optimum location to pursue major science goals in fast spreading rate crust.

ODP Hole 1256D in the eastern equatorial Pacific is the first basement borehole prepared with the infrastructure desirable for drilling a moderately deep hole into oceanic crust (~1.5–2 km). ODP Leg 206 installed a reentry cone supported by 20 inch casing with large-diameter (16 inch) casing through 250 m of sediment and cemented 19 m into basement. The cone and casing allow multiple reentries and maintain hole stability, essential for deep drilling. The wide-diameter casing allows for two further casing strings (13 and 10¾ inches) to be installed in the well, should future borehole stabilization be required and appropriate technology be available. Following the initiation of Hole 1256D, 502 m of massive lavas and sheet flows with moderate to high recovery were penetrated during Leg 206 (48%; Wilson, Teagle, Acton, et al., 2003). During Expeditions 309 and 312 (July–August and November–December 2005, respectively), Hole 1256D was successfully deepened to a total depth of 1507.1 mbsf, with the lower ~100 m hosted by a zone of gabbroic sills intruded into contact metamorphosed dikes (Fig. F4) (Wilson et al., 2006; Teagle, Alt, Umino, Miyashita, Banerjee, Wilson, et al., 2006).

In addition to the shallow depth to gabbros predicted from formation at a superfast spreading rate, Site 1256 has a number of specific attributes that indicate this site provides an excellent opportunity to sample a complete section of upper oceanic crust. The crust at Site 1256 formed near the Equator (Fig. F5), and high equatorial productivity resulted in high sedimentation rates (>30 m/m.y.) (Farrell et al., 1995) and the rapid burial of the young basement. A thick sediment blanket was needed for the installation of a reentry cone with 20 inch casing that forms the foundation for deployment of the second 16 inch diameter casing string that was cemented 19 m into the uppermost basement. At 15 Ma, Site 1256 is significantly older than the crust in Hole 504B (6.9 Ma; Alt, Kinoshita, Stokking, et al., 1993), and lower temperatures are anticipated at mid-levels of the crust. Logistically, Site 1256 has a number of advantages. It has a 12-month weather window and is ~3.5 days steaming from the Panama Canal; the short transit time allows for maximum time on site during drilling expeditions.

Site surveys and the geological setting of Hole 1256D

Site 1256 (6°44.2′N, 91°56.1′W) lies in 3635 m of water in the Guatemala Basin on Cocos plate crust formed ~15 m.y. ago on the eastern flank of the EPR (Figs. F1, F5). The details of site survey operations and the reasons for the selection of this particular site are outlined in detail in Wilson, Teagle, Acton, et al. (2003). Supporting site survey data for Expedition 335 and previous drilling at Site 1256 are archived at the IODP Site Survey Data Bank. Three regions were surveyed during the 1999 Maurice Ewing site survey cruise (Hallenborg et al., 2003; Wilson et al., 2003), with Site C in the GUATB-03 region selected as the preferred location for deep drilling (Fig. F3) (Wilson, Teagle, Acton, et al., 2003). The depth of the site is close to that predicted from bathymetry models of plate cooling (e.g., Parsons and Sclater, 1977). The site sits astride the magnetic anomaly 5Bn–5Br transition in magnetic polarity (Fig. F3). This crust accreted at a superfast spreading rate (~220 mm/y full rate) (Wilson, 1996) and lies ~1150 km east of the present crest of the EPR and ~530 km north of the Cocos Ridge. The site formed on a ridge segment at least 400 km in length, ~100 km north of the ridge-ridge-ridge triple junction between the Cocos, Pacific, and Nazca plates (Fig. F5). This location was formed near the Equator within the equatorial high-productivity zone and initially endured high sedimentation rates (>30 m/m.y.) (e.g., Farrell et al., 1995; Wilson, Teagle, Acton, et al., 2003). Sediment thickness in the region is between 200 and 300 m and is 250 m at Site 1256 (Wilson, Teagle, Acton, et al., 2003).

Site 1256 has a seismic structure reminiscent of typical Pacific off-axis seafloor (Fig. F6). Upper Layer 2 velocities are 4.5–5 km/s and the Layer 2–3 transition is between ~1200 and 1500 m subbasement (Fig. F7). The total crustal thickness at Site 1256 is estimated at ~5–5.5 km. Further to the northeast of Site 1256 (15–20 km), a trail of ~500 m high circular seamounts rise a few hundred meters above the sediment blanket (Fig. F3B). Bathymetry in the GUATB-03 survey areas is generally subdued, and Site 1256 sits atop a region of smooth basement topography (<10 m relief). However, elsewhere in the region, the top of basement shows a number of offsets along northwest-striking normal faults, and an abyssal hill relief of as much as 100 m is apparent in the southwest part of the area. Relief to the northeast is lower and less organized. In the northeastern sector of the GUAT-3B region, there is evidence for a basement thrust fault with a strike approximately orthogonal to the regional fabric (Wilson et al., 2003; Hallenborg et al., 2003). This feature dips gently to the northwest (~15°) and is clearly discernible to a depth of ~1.3 km on seismic Line EW9903-28 (Wilson et al., 2003), but the feature is less pronounced on seismic Line EW9903-27, indicating that the offset on the thrust decreases to the southwest.

Additional processing (A.J. Harding, unpubl. data) of ocean bottom hydrophone (OBH) recordings indicates discernible variation in the average seismic velocity (~4.54–4.88 km/s) of the uppermost (~100 m) basement and that there is regional coherence in the velocity variations (Fig. F8A). Two principal features are apparent: a 5 to 10 km wide zone of relatively high upper basement velocities (>4.82 km/s) that can be traced ~20 km to the edge of data coverage southeast of Site 1256 and a relatively low velocity (4.66–4.54 km/s) bull's-eye centered around the crossing point of seismic Lines EW9903-21 and 25. Using the site survey MCS data (Wilson et al., 2003), we have constructed a geological sketch map of the uppermost basement in the GUATB-03 survey region (Fig. F8B). The uppermost basement at Site 1256 is capped by a massive lava flow >74 m thick (Fig. F4). This flow is relatively unfractured, with shipboard physical properties measurements on discrete samples indicating VP > 5.5 km/s (Wilson, Teagle, Acton, et al., 2003). As such, it is likely that the area of relatively high uppermost basement seismic velocities represents the extent of the massive flow penetrated during Leg 206 in Holes 1256C and 1256D. Assuming an average thickness of 40 m, this would conservatively suggest an eruption volume in excess of 3 km3, plausibly >10 km3. This volume is extremely large when compared to the size of mid-ocean-ridge axial low-velocity zones, which are thought to be high-level melt lenses with typical volumes ~0.05–0.15 km3 per kilometer of ridge axis and generally appearing to be only partially molten (Singh et al., 1998).

Sheet flows (<3 m thick) and massive flows (>3 m) make up most of the lava stratigraphy cored so far at Site 1256, and such lava morphologies have been shown to dominate crust formed at fast spreading rates, away from segment tips (e.g., White et al., 2000, 2002). Subordinate pillow lavas are present in Hole 1256D, and because of the large number of fractures and pillow interstices, seismic velocities in these units are generally lower than those in more massive lava flows. We speculate that the bull's-eye of relatively low seismic velocities is a thick pile of dominantly pillowed lava flows.

Highlights of deep drilling in Hole 1256D

The uppermost crust at Site 1256 comprises a ~100 m thick sequence of lava dominated by a single massive lava flow up to 75 m thick, requiring at least this much seafloor relief to pond the lava. On modern fast spreading ridges, such topography does not normally develop until 5–10 km from the axis (Macdonald et al., 1996). Although this lava flow cooled off-axis, it may have originated at the ridge axis before flowing onto the ridge flanks, as is observed for very large lava flows on the modern ocean floor (Macdonald et al., 1989). The lavas immediately below include sheet and massive flows, along with minor pillow flows. Subvertical, elongate flow-top fractures filled with quenched glass and hyaloclastite in these lavas indicate flow lobe inflation, requiring cooling on a subhorizontal surface off-axis (Umino et al., 2000). From this shipboard evidence, we estimated a total thickness of 284 m of lavas that flowed and cooled off-axis. Sheet flows and massive lavas erupted at the ridge axis make up the remaining extrusive section down to 1004 mbsf before a lithologic transition is marked by subvertical intrusive contacts and mineralized breccias. This stratigraphy contrasts slightly with the volcanic stratigraphy for Hole 1256D developed from analysis of wireline geophysical imaging (Tominaga et al., 2009; Tominaga and Umino, 2010), which suggests <50% of the lavas drilled crystallized within 1000 m of the axis but that the majority of the lava pile had formed within 3000 m of the ridge crest.

Below 1061 mbsf, subvertical intrusive contacts are numerous, indicating the start of a relatively thin ~350 m thick sheeted dike complex dominated by massive basalts. Some basalts have doleritic textures, and many are cross-cut by subvertical dikes with common strongly brecciated and mineralized chilled margins. There is no evidence from core or from geophysical wireline logs for significant tilting of the dikes, consistent with subhorizontal seismic reflectors in the lower extrusive rocks that are continuous for several kilometers across the site (Hallenborg et al., 2003). Measurements of dike orientations by wireline imaging are in close agreement with direct measurements on recovered cores and indicate that the dikes at Site 1256 are slightly tilted away from the paleospreading axis (dip and dip direction are 79° ± 8°/N053 ± 23, respectively) (Tominaga et al., 2009).

The secondary mineralogy of the rocks indicates a stepwise increase in alteration grade downhole from the lavas to the dikes, with low-temperature phases (<150°C phyllosilicates and iron oxyhydroxides) in the lavas giving way to dikes partially altered to chlorite and other greenschist facies minerals (at temperatures greater than ~250°C) (Fig. F4) (also see Alt et al., 2010). Within the dikes, alteration intensity and grade increase downhole, with actinolite more abundant than chlorite below 1300 mbsf and hornblende present below 1350 mbsf, indicating temperatures approaching ~400°C (Alt et al., 2010). The dikes have significantly lower porosity (mostly 0.5%–2%) and higher P-wave velocities and thermal conductivity than the lavas, and porosity decreases and P-wave velocity increases with depth in the dikes.

In the lower ~60 m of the sheeted dikes (1348–1407 mbsf), basalts are partially to completely recrystallized to distinctive granoblastic textures characterized by granular secondary clinopyroxene and lesser orthopyroxene resulting from contact metamorphism by underlying gabbroic intrusions (Figs. F4, F9A) (Koepke et al., 2008; Alt et al., 2010). Aside from the granoblastic contact metamorphic assemblages in the basal dikes, hydrothermal mineralogy and inferred alteration temperatures of the lower dikes in Hole 1256D are generally similar to those in the lower dikes of Hole 504B (as high as ~400°C). The much thinner dike section at Site 1256 than at Site 504 (~350 versus ~1000 m), however, indicates a much steeper hydrothermal temperature gradient at Site 1256 (~0.5°C/m versus 0.16°C/m in Hole 504B).

Gabbro and trondjhemite dikes intrude into sheeted dikes at 1407 mbsf, marking the top of the plutonic complex (Fig. F9). Two major bodies of gabbro were penetrated beneath this contact, with the 52 m thick upper gabbro (Gabbro 1) separated from the 24 m thick lower gabbro (Gabbro 2) by a 24 m screen of granoblastic dikes (Figs. F4, F9). The textures and rock types observed in Hole 1256D are reminiscent of varitextured gabbros thought to represent a frozen melt lens beneath the sheeted dike complex and above the cumulate rocks in many ophiolites (e.g., MacLeod and Yaouancq, 2000; France et al., 2009).

Gabbro 1 is mineralogically and texturally heterogeneous, comprising gabbros, oxide gabbros, quartz-rich oxide diorites, and small trondjhemite dikelets. Oxide abundance decreases irregularly downhole, and olivine is present in significant amounts only in the lower part of the upper gabbros. These rocks are moderately to highly altered by hydrothermal fluids to actinolitic hornblende, secondary plagioclase, epidote, chlorite, prehnite, and laumontite.

The intervening dike screen is an interval of sheeted dikes captured between the two intrusions of gabbros. The dike screen consists of fine-grained metabasalts similar to the granoblastic dikes overlying the upper gabbros. The dike screen is cut by a number of small quartz gabbro and tonalite dikelets of variable thickness (1–10 cm), grain size, and composition.

Gabbro 2 comprises gabbro, oxide gabbro, and subordinate orthopyroxene-bearing gabbro and trondjhemite that are similarly altered to Gabbro 1 and has clear intrusive contacts with the overlying granoblastic dike screen. Partially resorbed stoped dike clasts are entrained within both the upper and lower margins of Gabbro 2 (Fig. F9G). Gabbro 2 is characterized by an absence of fresh olivine, high but variable orthopyroxene contents (5%–25%), and considerable local heterogeneity. Oxide abundance generally diminishes downhole. The predominant rock type is orthopyroxene-bearing gabbro with gabbronorite in the marginal units. The lowermost rock recovered from Hole 1256D is a highly altered actinolite-bearing basaltic dike that lacks granoblastic textures and hence is interpreted to be a late dike that postdates the intrusion of the lower gabbro.

Relative to other well-studied upper ocean crust sections (e.g., Karson, 2002), Site 1256 shows a thick lava sequence and a thin dike sequence. Steady-state thermal models require that the conductive lid separating magma from rapidly circulating seawater thin as spreading rate increases, indicating that the thin dike sequence is a direct consequence of the high spreading rate. A thick flow sequence with many massive individual flows and few pillow lavas is a reasonable consequence of short vertical transport distance from the magma chamber and is similar to observations from the midsegments of the fastest spreading ridges in the modern ocean (White et al., 2002). This model is in direct contrast to spreading models developed from observations of tectonically disrupted fast spread crust exposed in Hess Deep (Karson et al., 2002), which suggest regions of high magma supply should have thin lavas and thick dikes. There is little evidence for tilting (at most a few degrees) in Hole 1256D and no evidence for significant faulting. In contrast, the upper crust exposed at Hess Deep shows significant faulting and rotations within the dike complex (Karson et al., 2002), indicating that observations from that tectonic window might not be widely applicable. The massive lava flow at the top of the Site 1256 basement indicates that faults with ~50–100 m throws must exist relatively near to the ridge axis, even in superfast spreading rate crust, to provide the necessary relief to pond of the lava. At fast spreading ridges such relief is typically developed ~5 to 10 km from the axis.

Marine seismologists have long been dividing the ocean crust into seismic layers: Layer 1 comprises low-velocity sediments, Layer 2 has low-velocity and high-velocity gradient, and Layer 3 has high-velocity (generally at least 6.7 km/s) and low-velocity gradient. There is a widespread perception that Layer 3 is equivalent to gabbro, even though Hole 504B has penetrated Layer 3 but not gabbro (Alt et al., 1996; Detrick et al., 1994). From regional seismic refraction data, the transition from seismic Layer 2 to Layer 3 at Site 1256 occurs at ~1450–1750 mbsf (1200–1500 meters subbasement [msb]) (Wilson et al., 2003) (Fig. F7). Shipboard determinations of seismic velocities of discrete samples are in close agreement with in situ measurements by wireline tools, and gabbro velocities are <6.5 km/s (also Swift et al., 2008). Contrary to expectation, porosity increases and P-wave velocities decrease stepwise downward from lowermost dikes into uppermost gabbro in Hole 1256D as the result of the contact metamorphism of the granoblastic dikes and the strong hydrothermal alteration of the uppermost gabbros (Fig. F4). Porosity and velocity then increase downhole in the gabbro but are still <6.5 km/s. Wireline velocity measurements end at the top of gabbro, but we interpret the gabbro intervals as being within Layer 2 because a smoothed extrapolation of the downhole velocities will either have velocities <6.5 km/s, still characteristic of Layer 2, or an exceptionally high gradient to higher velocities, also characteristic of Layer 2. Encountering gabbro at a depth clearly within Layer 2 reinforces previous suggestions that factors including porosity and alteration are more important than rock type or grain size to controlling the location of the Layer 2/3 boundary. The position of the dike/gabbro boundary therefore appears to have little control over the seismic velocity structure of the crust (Alt et al., 1996; Detrick et al., 1994). Unfortunately, because the transition from Layer 2 to Layer 3 lies beneath Hole 1256D, we cannot yet determine what controls this transition at Site 1256.

Geochemistry of magmas from Site 1256

Flows and dikes from Hole 1256D show a wide range of magmatic fractionation, from fairly primitive to evolved (Figs. F4, F10). Shallower than 600 mbsf, magma compositions are bimodal, with relatively evolved thick flows and more primitive thin flows. The lava pond includes rocks with the highest incompatible element compositions (Zr, TiO2, Y, and V) and lowest compatible element concentration (Cr and Ni), suggesting that it is more evolved than other basalts from Site 1256. The initial division (based on Leg 206 lavas only) into high Zr-TiO2, low Zr-TiO2, and high Zr disappears with more data from Expedition 309 and 312. Rare samples from the lava pond fall off the dominant Y versus Zr and TiO2 versus Zr trends, suggesting possible minor variation in source compositions.

Downhole geochemical compositions of dikes are variable and do not define trends. Primary and evolved compositions are closely juxtaposed as would be expected for vertically intruded magmas. The range of major element compositions in the dikes is similar to the overlying rocks, and the average composition of the dikes is indistinguishable from the average composition of the lavas (e.g., Fig. F5).

The gabbros have highly variable bulk compositions. The uppermost gabbros have geochemical characteristics similar to the overlying dikes with mid-ocean-ridge basalt (MORB) chemistries (MgO ~7–8 wt%; Zr ~47–65 ppm). Conversely, deeper in Gabbro 2 the rocks are significantly less fractionated, and there are general downhole trends of increasing MgO, CaO, and Ni and decreasing FeO, Zr, and Y. The uppermost rocks of Gabbro 2 are fractionated with MgO contents of 6.1 wt%, but lower in the sequence MgO reaches 9.3 wt%. Decreasing concentrations of FeO and TiO2 downhole suggest that Gabbro 2 is fractionated similarly to Gabbro 1. The intrusive nature of both gabbro bodies and the chemical variations within them suggest that they intruded into the base of the sheeted dikes and underwent minor internal fractionation in situ, resulting in the observed general geochemical stratification.

Although there is scatter in the shipboard data, there are linear trends between MgO versus TiO2, FeO, CaO, Na2O, and Zr, most likely resulting from fractional crystallization of a gabbro, with the fractionating assemblage consisting of clinopyroxene and plagioclase, as expected for relatively evolved basaltic magmas. The predicted composition of the cumulate gabbros generated during this fractional crystallization is very different to that of the gabbros drilled so far. Simple mass balance calculations indicate that the average basalt has lost >30% of its original liquid mass as solid gabbro, implying the presence of at least 300 m of cumulate gabbro in the crust below the present base of the hole.

Gabbro compositions span a range similar to the flows and dikes but are on average more primitive, with higher MgO and lower FeO, albeit still within the range of EPR basalts (Fig. F10). Even though less fractionated, the average gabbro composition is evolved relative to candidates for primary magma in equilibrium with mantle olivine. Possible primary mantle melt compositions should have Mg# of 70–78 and MgO of 9–14 wt%. All flows and dikes and most gabbros are too evolved to be candidates for primary magmas. Therefore, the residue removed from primary magma to produce the observed gabbro and basalt compositions must occur below the present base of Hole 1256D.

The compositional ranges of fresh lava and dike samples correspond to typical values for MORB for most major elements and many trace elements (Su and Langmuir, 2003) and are similar to those observed for the northern EPR (Fig. F10). A few incompatible elements, including Na and Zr, have lower concentrations than those observed for modern EPR lavas, but the generally substantial overlap of compositions indicates that a similar process operated at the superfast spreading ridge that formed Site 1256 and the modern EPR. When trace elements are compared to EPR MORB, they are within one standard deviation of average, albeit on the relatively trace element depleted side of MORB. For example, compared with first-order mid-ocean-ridge segments along the EPR, basalts from Site 1256 have low Zr/TiO2 and Zr/Y. Although there is overlap among the segments and a large scatter in the data for each segment, Zr/TiO2 and Zr/Y appear to decrease with increasing spreading rate. The origin of this relationship remains unclear. Spreading rate may affect the extents of magma fractionation or partial melting of the mantle, or it may instead reflect regional-scale mantle heterogeneity. To decrease trace element ratios to this extent would require ~30% more melting at the superfast ridge, but this appears unlikely because of normal crustal thickness (~5.5 km) at Site 1256 (see Park et al., 2008).

The simplest model for mid-ocean-ridge magma plumbing is that the melt lens imaged by MCS experiments is the magma chamber where crystal-rich residues are separated from the evolved lavas that reach the seafloor (Fig. F11). Hole 1256D gabbros are texturally and compositionally similar to varitextured gabbros at the base of the sheeted dike complex in Oman interpreted to represent axial melt lenses (MacLeod and Yaouancq, 2000; France et al., 2009). If the gabbro bodies so far encountered in Hole 1256D were intruded on-axis and if they extended roughly horizontally for at least hundreds of meters, at 52 and 24 m thick they would have dimensions appropriate for axial low-velocity zones imaged by MCS experiments at intermediate to fast spreading ridges (Singh et al., 1998). However, these bodies could not have been the sites of primary magmatic fractionation. Chilled margins against the underlying dike screens preclude segregating a crystal residue that subsides to form the lower crust as in the gabbro glacier model, and its overall fractionated composition requires that crystals have been segregated elsewhere. This implies that sills or other bodies containing cumulate materials must exist deeper in the crust and/or below the crust/mantle boundary, consistent with recent models based on lower crustal sections of ophiolites (e.g., Boudier et al., 1996; Kelemen et al., 1997; MacLeod and Yaouancq, 2000) and some marine geophysical experiments (Crawford and Webb, 2002; Dunn et al., 2001; Garmany, 1989; Nedimovic et al., 2005; Canales et al., 2009). However, the gabbro glacier mode of accretion, or a combination of the gabbro glacier and sheeted sills models, cannot yet be rejected, as fractionated gabbros in the dike–gabbro transition are not unexpected and the predicted region of cumulate rocks could still exist just below the present maximum depth of Hole 1256D. Deepening Hole 1256D by as little as a few hundred meters would provide the critical samples that enable answering this outstanding basic question.