Drilling a complete crustal section has always been a major goal of deep ocean drilling, but because of technical difficulties and the time investments required, our sampling of the ocean crust remains rudimentary (Table T1). Hole 504B remains our only complete section of in situ upper crust and the only hole to penetrate the extrusive lavas and most of the way through the sheeted dike complex. Our poor sampling of ocean crust at different spreading rates and crustal ages compromises our ability to extrapolate observations from specific sites to global descriptions of the magmatic accretion processes and hydrothermal exchange in the ocean crust.

The transition from sheeted dikes to gabbros has never been drilled, and this remains an important objective in achieving a complete or composite crustal section by either offset or deep drilling strategies. The dike–gabbro transition and the uppermost plutonic rocks are the frozen axial melt lens and the fossil thermal boundary layer between magma chambers and vigorous hydrothermal circulation (Fig. F1). Detailed knowledge of this transition zone is critical to our understanding of the mechanisms of crustal accretion and hydrothermal cooling of the ocean crust. The uppermost gabbros and the overlying sheeted dikes and extrusive lavas provide a time-integrated record of the processes of hydrothermal exchange and fluid and chemical fluxes. The geochemistry of the frozen melt lens when compared with the overlying dikes and lavas will place important controls on crustal accretion processes and magma chamber geometry and will give a geological context to geophysical observations of low-velocity zones (Fig. F2).

Offset drilling strategies, where deeper portions of the ocean crust are sampled by drilling in tectonic windows, have recently been high priorities for ocean drilling (COSOD II, 1987; Ocean Drilling Program [ODP] Long Range Plan, 1996). Drilling at several sites has provided a wealth of new data and understanding of gabbros and peridotites from the lower crust and upper mantle, but problems still exist with drilling tectonized rocks and it is commonly difficult to relate drilled sections to the regional geology. Drilling a deep hole to obtain a complete crustal section and to more fully utilize the capabilities of the drill ship was reemphasized as an important drilling priority at the ODP-InterRidge-IAVCEI workshop in 1996 (Dick and Mevel, 1996) and was identified as one of two highest priority goals at the recent Architecture of the Ocean Lithosphere Program Planning Group meeting (held in 1998; meeting minutes available at http://joides.rsmas.miami.edu/panels/reports.html). Moreover, the ODP Long Range Plan points out that despite recent successes, offset or composite sections of the ocean crust are not substitutes for the primary goal of deep holes through the entire crustal section.

Drilling crust generated at a superfast spreading rate will provide one end-member of mid-ocean-ridge accretion (COSOD II; Long Range Plan). Recent assessment of drilling accomplishments and goals has pointed out that there has been no significant penetration (>100 m) of crust generated at a fast or superfast spreading ridge, making this fundamental objective a current high priority for drilling (Dick and Mevel, 1996). One of the major drilling objectives of the ODP Long Range Plan is to understand the architecture of ocean crust, including the lithology, geochemistry, and thicknesses of the volcanic and sheeted dike sections and the nature of the transition from dikes to gabbros, and to correlate and calibrate geological, geochemical, seismic, and magnetic observations of the structure of the crust. How does structure within Layer 2 and the seismic Layer 2–Layer 3 transition relate to alteration in the volcanics and dikes and to the dike–gabbro transition? At Site 504 in crust generated at an intermediate spreading ridge, the Layer 2–Layer 3 transition lies within the 1-km-thick sheeted dike complex and coincides with a metamorphic change (Detrick et al., 1994). Is this typical for ocean crust and for crust generated at faster spreading rates? Is the depth to gabbros shallower in crust generated at a superfast spreading rate, as predicted? Is the volcanic section thinner than that generated at slow or intermediate spreading rates? Francheteau et al. (1992) estimated a thickness of ~200 m at Hess Deep vs. >500 m at Site 504 and in the Atlantic; measurements of the thickness of seismic Layer 2A suggest 500–600 m for the East Pacific Rise (e.g., Kent et al., 1994).

A second objective is to understand magmatic and alteration processes, including the relationships among extrusive volcanics, the feeder sheeted dikes, and the underlying gabbroic rocks from the melt lens and subjacent sills/intrusions, as well as a comparison with abundant data for crust from slow spreading centers. Intraplate stresses can be determined, as well as the state of fracturing and permeability of the crust. Hydrothermal processes to be addressed by drilling, as outlined by the ODP Long Range Plan and the 1996 Woods Hole workshop, include fluid flow and alteration and the feedback between these and the nature of the subsurface hydrothermal "reaction zone." These will be addressed by examining the alteration "stratigraphy" within the extrusive lavas, whether disseminated sulfide mineralization and evidence for fluid mixing is present at the volcanic–dike transition (as in Hole 504B and many ophiolites), and the grade and intensity of alteration in the lower dikes and upper gabbros. In particular, the lowermost dikes and upper gabbros are predicted to be the subsurface reaction zone where fluids penetrate downward along fractures above the axial magma chamber and vent fluids acquire their final characteristics. Evidence for such fractures has previously been recovered, but an intact section has never been drilled. Drilling this lithologic transition will allow tracing of fluids and linking hydrothermal alteration in sheeted dikes and underlying gabbros to magmatic processes in the melt lens.

Although there are several questions that can be answered well with shallow holes in tectonic windows such as Hess Deep, other questions on topics from in situ permeability to alteration history will give answers that cannot be generalized to normal crust. Furthermore, at sites that are tectonized at very young ages, doubts will remain as to whether the same factors that cause the tectonic exposures also perturb the ridge axis from the normal state.

There are three factors that lead us to believe that there are very good chances of reaching gabbro in normal oceanic crust in a two-leg drilling program:

1. Purdy et al. (1992) describe an inverse relation between spreading rate and depth to an axial low-velocity zone, interpreted as a melt lens (Fig. F2). Since Purdy et al.'s compilation, careful velocity analysis, summarized by Hooft et al. (1996), has refined the conversion from traveltime to depth, and data from additional sites have been collected (Carbotte et al., 1997). The fastest-rate spreading centers surveyed with modern multichannel seismic reflection, ~140 mm/yr at 14°–18°S on the East Pacific Rise (EPR), show a reflector interpreted as the axial melt lens at depths typically of 940 to 1260 meters below seafloor (mbsf) (Detrick et al., 1993; Kent et al., 1994, Hooft et al., 1994, 1996). At 9°–16°N on the EPR where spreading rates are 80–110 mm/yr, depths to the melt reflector are mostly 1350–1650 mbsf, where well constrained (Kent et al., 1994; Hooft et al., 1996; Carbotte et al., 1997). Recent identification of magnetic anomalies formed at the southern end of the Pacific/Cocos plate boundary indicate full spreading rates of ~200 mm/yr from 20 to 11 Ma (Fig. F3), implying that these areas are underlain by relatively thin upper crust and shallow gabbro sections.
2. A possible factor in the good drilling conditions in the 6-m.y.-old crust in Hole 504B compared with very young fast-spreading crust from Legs 54 and 142 is the equatorial latitude of formation, with high sedimentation rate leading to rapid burial of the igneous crust while middle levels of the crust are still hot. Alteration under those conditions may increase cementation in the basement and increase the competency of healed fractures in young crust during drilling. The fast spreading rates highlighted in Figure F3 were formed near the equator, and rapid initial sedimentation rates (at least 35 m/m.y.) have been confirmed at ODP Sites 844 and 851 from Leg 138 and Deep Sea Drilling Project (DSDP) Site 572 from Leg 85. A sediment cover of ~240 m is estimated for the proposed drill site. This sediment cover will also facilitate the installation of the reentry cone and 40 to 80 m of 20-in casing, which together are jetted into the sediment until the reentry platform rests on the seafloor.
3. Thermal stresses that resulted in drilling-induced fracturing deep in Hole 504B should be diminished in this older crust (~15 Ma, compared with 6 Ma for Hole 504B). Although such fracturing did not prohibit deep penetration in Hole 504B, this provides some indication that deep drilling conditions can be expected to be better than at Site 504.

The theoretical basis for expecting an inverse relation between spreading rate and melt lens depth is quite simple. The latent heat released in crystallizing the gabbroic crust must be conducted through the lid of the melt lens to the base of the axial hydrothermal system, which then advects the heat to the ocean. The temperature contrast across the lid is governed by the properties of magma (1100°–1200°C) and thermodynamic properties of seawater (350°–450°C where circulating in large volumes) and will vary only slightly with spreading rate. The heat flux through the lid per unit ridge length will therefore be proportional to the width of the lens and inversely proportional to the lid thickness. For reasons which are not understood, seismic observations show uniform width of the melt lens, independent of spreading rate. With width and temperature contrast not varying, the extra heat supplied by more magma at faster spreading rates must be conducted through a thinner lid (dike layer) to maintain steady state (see Phipps Morgan and Chen [1993] for a more complete discussion). To reach the dike–gabbro transition in normal oceanic crust with minimal drilling, it is therefore best to choose the fastest possible spreading rates. A setting similar to the modern well-surveyed area at 14°–18°S could be expected to reach gabbro at a depth of ~1400 m, based on 1100 m to the axial magma chamber (AMC) reflector and subsequent burial by an additional 300 m of extrusives (Kent et al., 1994). At faster rates, depths could possibly be hundreds of meters shallower. In contrast, seismic velocity inversions at the axes of the Juan de Fuca Ridge and Valu Fa Ridge, Lau Basin, are at depths of ~3 km (Purdy et al., 1992) at intermediate spreading rates comparable to Site 504.

Although perhaps only 20% of the global ridge axis is separating at fast spreading rates (>80 mm/yr full rate), this end-member style of the ocean spreading produced ~50% of the present-day ocean crust and ~30% of the total Earth's surface. At least in terms of seismic structure (Raitt, 1963; Menard, 1964), crust formed at fast spreading rates is relatively simple and uniform. Hence, the successful deep sampling of such crust in a single location can reasonably be extrapolated to describe a significant portion of the Earth's surface.

Drilling of the fast spreading crust has been mostly unsuccessful (e.g., DSDP/ODP Legs 34, 54, and 142). Apart from surface sampling of recent basalts at ridge axes, little is known of the shallow and intermediate depth structure of fast-spreading crust (Table T1). One recent exception is the coring completed at Site 1224 during Leg 200 in the eastern North Pacific, which sampled a 146.5-m-thick section of basaltic oceanic crust created by fast seafloor spreading (142 mm/yr full rate). Studies of Site 1224 are just getting underway but will be limited to the extrusive basalt flows recovered. A continuous section through the upper oceanic crust and ultimately into mid-crustal gabbros is imperative to calibrate geophysical observations and numerical models of the ocean crust.

A recent synthesis of magnetic anomaly data for the central Cocos plate and corresponding regions of the Pacific plate demonstrated that the spreading rate on the southern Cocos/Pacific plate boundary during the middle Miocene was ~200 mm/yr, ~30% to 40% faster than the fastest modern spreading rate (Wilson, 1996). This episode of fast spreading ended fairly abruptly in a plate-motion reorganization at 10.5–11.0 Ma; subsequent motions have been similar to present-day motions. The southern limit of crust formed at the superfast rates is the trace of the Cocos-Nazca-Pacific triple junction, as Nazca-Pacific and Cocos-Nazca spreading rates were not as fast. The lower age limit of this spreading episode is hard to determine with the limited mapping and poor magnetic geometry of the Pacific plate. It is at least 18 Ma and could reasonably be 24–25 Ma. The northern limit of this province is entirely gradational, with rates dropping to ~150 mm/yr somewhat north of the Clipperton Fracture Zone. By apparent coincidence, the fastest spreading rates occurred within a few degrees of the equator.

Using the fastest possible spreading rate as a proxy for shallowest occurrence of gabbro still allows a range of possible drilling sites. There is no reason to expect a difference in crustal structure between the Cocos and Pacific plates, but logistics favor a site on the Cocos plate. Transits from a variety of Central American ports would be only 2–4 days, and sediments are about 200 m thinner than on the Pacific plate. It seems prudent to choose an anomaly segment at least 100 km long and a site at least 50 km from the end of the segment. For ages 12–16 Ma (anomalies 5AA–5B) these criteria are easy to satisfy because the southernmost segment of the Pacific-Cocos Ridge had a length of at least 400 km. For ages 17 Ma (anomaly 5D) and older there is a fracture zone to avoid, but the length of anomaly segments is at least 150 km. A possible option would be to reoccupy Site 844, near the young edge of anomaly 5D about 100 km from a fracture zone and roughly 150 km from the trace of the triple junction with the Cocos/Nazca boundary. The sedimentary section was redundantly cored during Leg 138 in 1991, with basalt chips recovered from 290 mbsf. Concordance between the age of the deepest sediments and the magnetic anomaly age (Wilson, 1996) indicates that there are no sills significantly above the base of the sediment column that would reflect magmatic rejuvenation of the site.

The only serious drawback to this area for a crustal reference section for fast spreading rates is the low original latitude, which makes magnetic polarity determinations impossible from azimuthally unoriented core samples and, given the nearly north-south ridge orientation, makes the magnetic inclination insensitive to structural tilting. The polarity problem could be solved with a reliable hard rock orienting device, but development efforts for such a tool have been abandoned. Magnetic logging with either the General Purpose Inclinometry Tool (GPIT) fluxgates that are part of the Formation MicroScanner/Dipole Shear Sonic Imager (FMS/sonic) tool string or preferably a separate magnetic tool with gyroscopic orientation, such as the Bundesanstalt für Geowissenschaften und Rohstoffe (BGR) borehole magnetometer, should also be adequate for polarity determination, as demonstrated in Holes 504B and 896A with logs collected during Leg 148 (Worm et al., 1996). Alternative sites have other, often more serious, drawbacks. Sites flanking the EPR south of the equator generally have poor accessibility and for the age range 10–25 Ma have a complicated tectonic setting and often uncertain spreading rates. North of the equator, sites are available in the same Cocos-Pacific system, which is better understood and more tectonically stable, but there is a severe tradeoff between latitude and spreading rate. A magnetically desirable latitude of 20° would reduce the spreading rate to ~60% of the rate for the sites we propose, which may significantly reduce the chances of reaching gabbro in limited drilling time. To detect structural rotations about a nearly north-south ridge axis, paleolatitude should probably exceed 25°, which means that no site satisfying this criterion will also offer fast spreading rate and short transit to common ports.

The site survey cruise for this proposal took place in March and April 1999, aboard the Maurice Ewing, led by D. Wilson, A. Harding, and G. Kent. At the urging of the Architecture of Oceanic Lithosphere Program Planning Group, we modified our original plan for four sites in the Guatemala Basin to instead cover three sites there and a separate site near Alijos Rocks west of Southern Baja California (Figs. F4, F5, F6, F7). The principal advantage of the Alijos site is higher paleolatitude, allowing determination of magnetic polarity with azimuthally unoriented cores. The other significant difference recognized before survey work is lower spreading rate, ~120–130 mm/yr instead of 200–210 mm/yr.

The site survey work focused on seismic reflection and refraction. Multichannel seismic reflection (MCS) and refraction work to ocean bottom hydrophones (OBHs) were conducted separately because of differences in desired shot intervals. MCS work used a tuned array of 10 air guns shooting to a new 480-channel, 6-km streamer, with a nominal shot interval of 37.5 m (15–18 s). With a hydrophone spacing of 12.5 m, this geometry gives 80-fold coverage with 6.25-m midpoint spacing. Refraction shooting to grids of 10–11 OBHs using 20 air guns was at a shot interval of 90 s (130–180 m) for most of the grid and 150 s (300 m) for the outermost shots. The grid geometry was designed for well-constrained measurements of velocities in both across-strike and along-strike directions to depths of 1.5–2.0 km and to cover to Moho depths in the across-strike direction only (Figs. F5, F6, F7). Because of time constraints and delays from several causes, refraction surveying was only done at two of the three Guatemala Basin sites.

The Guatemala Basin sites have 200–300 m of sediment cover resulting from their formation near the paleoequator. Referring to the sites in the order MCS was collected, grid 1 was chosen to include ODP Site 844 at a line crossing and to be centered near the C5D(y) magnetic anomaly boundary, and grids 2 and 3 were centered on anomalies C5C(y) and C5B(o) along a flow line perpendicular to anomaly strike (Fig. F4). Grids 1 and 3 have refraction data. Grid 1 is quite shallow for its ~17-Ma age at 3400–3500 m, and basement at Site 844 is at 3705 m (Fig. F6). Relief on basement as seen in MCS is extremely low, with largest scarps having ~30-m amplitude. Subtle horizontal reflections ~1.6–1.7 s below basement suggest Moho. A cluster of seamounts with minimum depth of 2790 m is present near the southern tip of the grid.

Grid 3 is deeper than grid 1 at 3600–3700 m, and basement at ~3900 m is near normal depth for the ~15-Ma age (Fig. F5). In the southwest half of the grid, abyssal hill fabric is visible through the sediment cover, and larger scarps approach 100-m amplitude. The northeast half of the grid has low relief, comparable to grid 1. Reflection data here commonly show complex reflectors at 1.3–1.8 s below basement, indicating dipping interfaces in the lower crust or upper mantle, probably including some Moho reflections. Upper crustal reflectors at ~0.4–0.8 s into basement are often bright and tend to have shallow (~20°) apparent dips in the isochron direction, with more horizontal apparent dips in the spreading direction (Figs. F8, F9). Analysis of refraction data in this grid shows crustal structure that is fairly typical for off-axis Pacific seafloor. Upper Layer 2 velocities are 4.5–5 km/s, a gradual transition between Layers 2 and 3 is at ~1.5 km below basement, and total crustal thickness is ~5–5.5 km (Fig. F10). Velocities of the uppermost crust are slowest in the southwestern part of the grid where the abyssal-hill relief is greatest.

In contrast to the Cocos plate sites, grid 4 near Alijos Rocks has thin (50–100 m) sediment, slightly deep water (3800–4300 m) for the ~16.5-Ma age, and extremely high relief for the fast spreading rate (Fig. F7). Individual scarps are commonly 150 m, and up to 400 m. MCS data show no coherent reflections below Layer 2 in preliminary stacks. Receiver gathers for refraction data are broadly similar to the Cocos plate sites, perhaps suggestive of slightly slower velocities around 1 km below basement.

Magnetic data at the Cocos plate sites show trends parallel to the previously mapped regional trend, with no evidence for isochron offsets at ~1-km detection limit within the grids (Fig. F11) and perhaps 3- to 5-km detection limit outside the grids. The Alijos grid is located within an area where magnetic and topographic features are linear for 30–40 km, but right-stepping offsets of a few kilometers leave the local trend a few degrees counterclockwise of the regional trend.

Of the three survey grids with refraction data, we have chosen grid 3 (the southwesternmost and youngest of the Cocos plate grids) as our primary target because its depth and relief are closest to normal. Within this grid, several factors affected the final site selection. The slower seismic velocities in southwestern part of the grid indicate more porous and possibly more rubbly material that may lead to poorer drilling conditions. OBH failure on line 23 along the southeastern part of the grid led to limited constraints on velocity determinations there. A very bright upper crustal reflector is observed on much of line 21 in the northeastern part of the grid and for short distances on lines 27 and 28 where they cross line 21. The reflector dips northwest and projects updip to a hill ~50 m high with northeast strike, which is perpendicular to the normal abyssal-hill trend. The character of the reflection and its relation to the nearly vertical velocity gradient determined by refraction analysis are both more consistent with a narrow low-velocity zone rather than a simple interface between materials of different velocity. All of these relations suggest that the reflector might be a thrust fault, possibly driven by thermal contraction of the lower lithosphere. Because such a fault might lead to very poor drilling conditions at about the depths gabbro might be encountered on a return leg, we have chosen to avoid this area as well. The remaining area near the northern corner of the grid appears very suitable for deep drilling, and we have chosen the intersection of lines 22 and 27, where the velocity control is best, as the primary drilling site, GUATB-03C.