Our gravity modeling of oceanic core complexes formed at the Mid-Atlantic Ridge near 30°N suggests that their shallow, domal “cores” could be dominated by mafic intrusive rocks, consistent with recent drilling results at Atlantis Massif. The three-dimensional gravity analysis incorporates additional underway geophysics data in a new compilation and uses a higher-resolution bathymetry model to remove the gravity contribution of seafloor topography. The additional detail is required in order to confidently relate few-kilometer-scale gravity anomalies to specific morphologic/tectonic blocks. Different models of subseafloor core complex structure and density are tested to determine which minimizes the local gravity anomaly. A 3-D core with density 2900 kg/m3, as measured in the gabbroic section drilled at the central dome, and juxtaposed 3-D hanging wall of fractured basalt, density ∼2600 kg/m3, satisfactorily explains most of the Bouguer gravity anomaly at Atlantis Massif. The capping detachment fault terminates or plunges northward beneath the seafloor at the northern limit of the central dome. The southwest shoulder of the massif has lower density, consistent with an upper crustal section ∼1 km thick, whereas the summit and southeastern shoulder have overall density similar to the central dome. The older core complexes distributed along Atlantis fracture zone are similar in size, depth, and distance of their summit from the transform fault. However, weathering/alteration probably has reduced their density somewhat compared to Atlantis Massif. Bathymetric embayments occur adjacent to the fracture zone in several places on the ridge flanks and are consistently associated with core complexes.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The occurrence of oceanic core complexes (OCCs) on the flanks of mid-ocean ridges can be used to assess variability in the balance of magmatism and tectonism along an oceanic spreading center. The domal topographic highs at OCCs are capped by smooth, corrugated surfaces inferred [Cann et al., 1997; Karson, 1990; Searle et al., 2003; Tucholke et al., 1998; Tucholke and Lin, 1994] and in a few locations confirmed [Karson et al., 2006; MacLeod et al., 2002] to be exposed detachment faults. The OCCs document a period of rifting where the style of lithospheric accretion and deformation within a given portion of a spreading segment differed from the more typical style, where lineated abyssal hills with exposed volcanic rocks are formed. For the purposes of this study, a distinction is made between detachment faulting, which may occur for intervals of time along a given slow spreading axis [Smith et al., 2006], and the less frequent development of OCCs. Fully developed OCCs occur when a given detachment is active for a sufficiently long period of time (>1 Ma) that significant unroofing of deep material occurs.
 Three OCCs have formed in the past 10 Ma at the Mid-Atlantic Ridge (MAR) near 30°N where the ridge intersects the Atlantis Transform fault (ATF) (Figure 1a) [Cann et al., 1997]. Shorter detachment faulting episodes also appear to have occurred during this time without the development of full OCCs (note, for example, the “Keystone” feature just north of the transform fault at 42° 32′W). Atlantis Massif is the youngest OCC in this area, having formed at the eastern ridge-transform intersection (RTI) in the past 0.5–2 Ma [Blackman et al., 1998]. The western OCC (WOCC) also formed at this RTI, ∼9 Ma ago on the basis of magnetic anomalies identified by Pariso et al. ; it was subsequently rafted along the flank of the northern ridge segment of this ridge-transform-ridge system. The southern OCC (SOCC) formed about 3 Ma ago at the inside corner of the western RTI. The average half spreading rate at this section of the MAR is 12 mm/a [Pariso et al., 1996] so the ∼10 km width (in the direction of plate motion) of the 3 domal cores suggests that slip along each of these detachments may have represented a main locus of plate separation for ∼1–2 million years. Other domal cores (observed) within OCCs (interpreted) in the Atlantic have similar spatial and inferred temporal scales [Fujiwara et al., 2003; Reston et al., 2002; Tucholke et al., 1998].
 Prior processing of bathymetry and gravity data in this region [Blackman et al., 1998; Nooner et al., 2003] and at other OCCs [Cannat et al., 2006; Fujiwara et al., 2003; Okino et al., 2004; Searle et al., 2003; Tucholke et al., 2001] indicates that the shallow domes are not locally compensated, each being associated with a positive mantle Bouguer gravity anomaly. However, care should be taken in using mantle Bouguer anomalies to investigate subsurface density structure in the vicinity of OCCs. This approach, which consists of removing the gravity contribution of a constant thickness, constant density crust and a lithospheric cooling model from the free-air gravity anomaly [Prince and Forsyth, 1988] was specifically designed to highlight areas that deviate from such a model. However, the choice of a crust-mantle interface depth can be rather arbitrary, particularly in the vicinity of OCCs where detachment tectonics are thought to significantly thin the crust. Since the residual gravity anomalies on young ridge flanks can be as small as 5–20 mGals, the choice of crust-mantle boundary depth can be crucial, with high-relief features contributing up to ∼10 mGals. One of the objectives of this study is to assess the impact of assumptions involved in obtaining a Bouguer correction (removal of the water column contribution, above variable seafloor topography) near a slow spreading ridge-transform-ridge plate boundary. A new gravity compilation (Figure 1b) allows a more detailed study than was conducted by Blackman et al. . Insights from recent geologic mapping and drilling have guided a 3-D analysis of subsurface density at Atlantis Massif. Expanded flank coverage provides a longer-term view of segment processes, both north and south of the ATF, and allows a more robust look at the signature of the WOCC than was possible in the earlier 2-D study.
 A final aspect of this study uses the morphologic structure of the ridge flanks to assess the general nature of OCC development and evolution, which contrasts with the more typical formation of lineated abyssal hills during the past 10 Ma of spreading in this part of the MAR. In addition to the domes, deep bathymetric embayments are found adjacent to or on the conjugate flank of each of the OCCs along the Atlantis fracture zone, including the current nodal basin at the eastern RTI (labeled in Figure 1c), which is associated with Atlantis Massif.1
Table 1. Gravity Anomaly Calculations and Associated Model Assumptions
Seawater/Crust Δρ, kg/m3
Intracrustal Interface Model
Upper Crust/Lower Crust Δρ, kg/m3
Lower Crust/Mantle Δρ, kg/m3
2. Prior Results in the Region
 The Mid-Atlantic Ridge 30°N segments on either side of Atlantis Transform Fault were mapped with multibeam sonar in the late 1980s [Purdy et al., 1990]. The ATF is a long-lived transform fault [Klitgord and Schouten, 1986] that marks the end of a series of small-offset, en-echelon segments that extend north from the Kane transform fault near 23°N. The narrow Atlantis fault zone, and its strands within the deep transform valley, were also imaged with side-scan sonar in the 1980s [Parson and Searle, 1986]. Ridge segment-centered mantle Bouguer anomalies were reported by Lin et al.  and inferred to reflect along-strike variations in mantle melt supply. These authors concluded there was less magmatic input near the ends than at the centers of the segments. Their residual gravity anomaly map showed complexity around Atlantis fracture zone and Zervas et al.  analyzed these data in more detail. A joint U.S.-French expedition then obtained bathymetry, gravity, and magnetics data on the ridge flanks out to ∼10 Ma crust [Pariso et al., 1995, 1996] and these data are key for understanding the range of structures and processes that have characterized this part of the MAR over time.
 In 1996, a study sponsored by the UK BRIDGE program recognized the inside corner high at the eastern RTI to be an oceanic core complex [Cann et al., 1997], on the basis of morphology, side-scan sonar maps, and dredge samples. Further side-scan sonar analysis and gravity modeling [Blackman et al., 1998] supported this interpretation of Atlantis Massif, as well as documenting similar signatures for SOCC and WOCC, although the latter was only partially imaged and has still not been sampled.
 Seafloor mapping and sampling at Atlantis Massif were carried out in 2000 and 2003 [Blackman et al., 2002; Karson et al., 2006]. It was not possible to confirm the existence of a detachment fault on the central dome of the massif, due to sediment cover, but work in the uppermost part of the transform-facing wall of the southern ridge [Schroeder and John, 2004] confirmed that brittle deformation dominated the upper ∼100 m of the section exposed in headwall scarps of major slope failures. Karson et al.  reported that the normal fault exposed at the top of the southern ridge (with a sediment/breccia cover up to a few meters thick) extends a few kilometers in the spreading-parallel direction. The extent of this detachment fault zone along the strike (ridge-parallel direction) of the massif is unknown, but it is inferred to occur just below a thin, lithified sediment cover wherever bathymetric corrugations or the smaller-scale striations imaged by side-scan sonar are mapped. The fault is thus interpreted to extend over the whole central dome and the peak and southeast shoulder portions of the southern ridge (Figure 1c). No striations or corrugations have been observed on the SW shoulder of the southern ridge. The northern limit of the corrugations on the central dome (∼30°14′N) coincides with a westward jog in the eastern scarp of the adjacent hanging wall block [Blackman et al., 1998] and with a diagonally cutting small scarp that crosses the hanging wall (Figure 1c). North of this “13°14′ transition,” a series of 3 steep scarps, inferred to mark inward dipping normal faults, define the western rift valley wall.
 Most of the rocks recovered from the southern wall, where coverage is much greater in the upper few hundred meters than it is in the rest of the section, are serpentinized peridotite [Blackman et al., 2002; Boschi et al., 2006; Schroeder and John, 2004]. About 70% of Alvin and dredge samples from the transform face of the massif have this composition with the remainder being gabbroic. Circulation of seawater through, and consequent alteration of, this peridotite contributes significantly to and is perhaps the driving force [Früh-Green et al., 2003; Kelley et al., 2001] for hydrothermal venting at the Lost City field that is located just below the peak of Atlantis Massif.
 All rock samples from the hanging wall are basaltic and the sonar signature of the top of this block is typical of hummocky volcanic terrain seen elsewhere near or within the axial zone of the MAR [Cann et al., 1997; Smith and Cann, 1993].
 A set of 2-km long deep-source and longer surface-source (air gun) seismic refraction lines were acquired across Atlantis Massif in 1997 [Detrick and Collins, 1998]. Data from one of the reversed deep-source lines was well fit by a model with variable-thickness upper layers (3.0–6.5 km/s) underlain by a 7.5 km/s layer (characteristic of ultramafic rock) starting at 600–700 m depth below the seafloor of the central dome [Collins et al., 2003]. Multichannel seismic reflection data were collected in 2001 and a prominent reflector was traced across essentially all of the central dome and southern ridge [Canales et al., 2004]. Velocity analyses of the region above the reflector at the central dome suggested that this reflector could coincide with the top of the high-velocity layer modeled from the refraction data. This, together with the south wall geologic data and prior gravity modeling, supported a model of Atlantis Massif as a dominantly ultramafic core complex, with unaltered mantle rock uplifted to shallow depths within the domal high.
 Integrated Ocean Drilling Program (IODP) Expeditions 304 and 305 targeted Atlantis Massif for deep drilling in 2004/2005 [Blackman et al., 2006]. Site 1309 on the footwall to the detachment at the southern end of the central dome had high recovery (75%) of a gabbroic section from Hole U1309D, which penetrated 1415 m (location in Figure 1; lithologic section in Figure 2). Ultramafic rocks made up only ∼5% of the recovered section from Hole U1309D. Clearly, the geophysical model with the core of Atlantis Massif dominated by variably altered to fresh peridotite was not consistent with this recovered gabbroic sequence. Reanalysis of seismic refraction arrivals [Blackman et al., 2007; Canales et al., 2007] confirm that these data can be explained by velocities that are ≤6.5 km/s with a variable layer thickness in the uppermost kilometer.
 The hypothesis of Ildefonse et al.  predicts that the south wall is not lithologically continuous with most of the footwall, but rather displays a thin, deformed serpentinite sheath that surrounds a dominantly gabbroic core (both central and southern dome), whose rheologic contrast with the surrounding altered peridotite [Escartin et al., 2001] controls the geometry of uplift and faulting associated with OCC formation.
 Alternatively, it is possible that the southern ridge is dominantly peridotite, with lesser amounts of intruded gabbro, consistent with interpretation of the south wall as a cross-sectional view of the Atlantis Massif domal core [Boschi et al., 2006; Karson et al., 2006]. In this scenario, a transition or boundary must occur between the southern ridge and the central dome. The latter is about a kilometer deeper than the peak of Atlantis Massif and is more than 500 m deeper than the bulk of the southern ridge, which could indicate different composition/structure. However, the (spreading-parallel) width of the striated detachment surface is the same on both portions of the massif suggesting that exposure of the central dome and southern ridge occurred simultaneously. The corrugated surface on the central dome is offset a few kilometers to the west relative to the position of the exposed detachment on the southern ridge. This could indicate different evolutions of the central and southern domes, perhaps enabled by an accommodation zone as reflected in the steep slope on the northern side of the SE shoulder.
 In the following sections, the constraints from the recent drilling and seafloor geology studies are incorporated into an updated analysis using the new compilation of gravity and additional bathymetry data.
3. Analysis Methods
 Our general processing methodology is standard for marine geophysics, although the three-dimensional gravity modeling employs an alternate approach, using a variable-depth interface technique to remove contributions of 3-D structure. Details of the processing are available in Appendix A.
3.1. Gravity Data
 Free-air gravity anomaly (FAA) data from cruises EW9210 and the multichannel seismic cruise (EW0102) were added to the compilation of Blackman et al. . The uncertainty in the new track line anomaly values is 1.8 mGals for the full region; within a smaller area around Atlantis Massif, the uncertainty is 1.4 mGals.
3.2. Drilling/Logging Data
 Core sample measurements and borehole logging conducted during IODP Expeditions 304 and 305 provide local density and porosity [Blackman et al., 2006] (Figure 2). The upper 380 m of the section at IODP Site1309 on the central dome of Atlantis Massif has densities of 2800–2850 kg/m3 measured on core samples, while density measured in borehole logs is lower, suggesting the presence of large-scale cracks. From 400 to 800 mbsf, sample densities are consistently at the upper end of the range of logged densities, but below that depth the two techniques agree more closely. From 400 to 1300 mbsf, the density slowly and steadily increases from ∼2900 kg/m3 to ∼3000 kg/m3. The large reduction in logged density near 1100 mbsf is mainly an artifact of poor hole conditions across a fault zone [Blackman et al., 2006].
3.3. Bouguer and Residual Anomalies
 Removal from the FAA of the contribution of a constant-density layer with top boundary at the seafloor gives the (marine) Bouguer anomaly, which allows easier recognition of subsurface density variability and its relationship to specific tectonic features. In this formulation, any deeper interfaces are ignored (or are assumed horizontal). The Prince and Forsyth  implementation of Parker's method [Parker, 1972] for computing the Bouguer anomaly was applied using the 100 m bathymetry grid (Figure 1). Previous studies typically use bathymetry models with 0.5–1 km grid spacing. At the MAR 30°N, the high relief and proximity of the domes to the sea surface (≤1000 m) require a more detailed model to capture the full contribution of these features to the gravity field. In areas adjacent to the shallowest, steepest parts of OCCs along the ATF, differences in the predicted seafloor contribution to the FAA for a 500 m bathymetric grid versus the 100 m grid can be up to 8 mGals.
 In calculating the Bouguer anomaly, an assumption must be made about the density contrast between seawater (1030 kg/m3) and crust. Generally, the choice of density (within reasonable values of 2600 kg/m3 for upper crust and 2900 kg/m3 for full crust) does not strongly influence the relative anomaly pattern [Blackman and Forsyth, 1991]. However, at shallow features such as the OCC, the specific choice of density contrast across this interface can change the anomaly amplitude sufficiently to affect comparisons between various hypotheses for the crustal structure of OCCs. The 2600–2800 kg/m3 range of densities assumes that basaltic sections are heavily fractured (12–5% porosity, respectively) so that values are lower than inherent rock properties [e.g., Turcotte and Schubert, 1982]. Comparison of the results for an assumed seafloor contrast of 1700 versus 1850 kg/m3 shows that the positive anomaly associated with each of the three OCCs along ATF is 3–5 mGal greater for the former compared with the latter density contrast.
Figure 3 shows the residual gravity anomaly after removing a lithospheric thermal contribution [Phipps Morgan and Forsyth, 1988] (see Appendix A) from the Bouguer anomaly for these two assumptions of seafloor density contrast (RA1.7, RA1.85, respectively).
3.4. Construction of a 3-D OCC Model
 Both the geometry of the main structural components of the OCC and the geologic constraints available at Atlantis Massif are incorporated for 3-D gravity modeling in this local area. The hanging wall does not extend the full length of the OCC, but terminates at the northern edge of the SE shoulder. All available evidence (morphology, video imaging, rock samples [Blackman et al., 1998, 2002]) indicates that the hanging wall is composed of basaltic rock (density ∼2600 kg/m3). The underlying footwall comprising the core of the massif is a mix of gabbro (density 2800–2900 kg/m3), serpentinized peridotite (variable density in the range 2600–3200 kg/m3) and, perhaps, some amount of essentially unaltered peridotite (∼3300 kg/m3). At IODP Site 1309, gabbro dominates at least the upper 1.4 km and may be representative for most of the central and southern dome. One outcome of the model proposed by Ildefonse et al.  is that the vertical scale of a gabbroic batholith within the core of an OCC is of the order of the lateral scale of the domal high. On the basis of the drilling results, basalt/diabase occurs only in parts of the upper ∼100 m of the footwall, and even these are shown, in core sample measurements, to have higher density (∼2800 kg/m3, Figure 2) than we assume for the hanging wall. Below ∼400 m below seafloor in Hole U1309D, both log and core sample data indicate a density of ∼2900 kg/m3 for the central dome.
 The structural components of the OCC can be modeled, via manipulation of the seafloor grid using Generic Mapping Tools (GMT) [Wessel and Smith, 1998], by constructing an interface that nearly coincides with the seafloor across the dome of the massif (both central and southern) and then smoothly deepens to a nominal upper crust–lower crust boundary depth of 1.5 km below seafloor in regions away from the core of Atlantis Massif. Figure 4 shows the model where a higher-density core extends almost to the seafloor at the dome (a 100 m-thick top zone of lower density is specified to simulate the greater fracturing and different composition of the uppermost rocks in Hole U1309D, Figure 2). The 2200 m seafloor contour at Atlantis Massif marks where the interface begins to drop further below the seafloor (gray line in Figure 1c). The 1.5 km depth of the interface below seafloor is reached along the 2600 m contour on the west side of the massif and along the 3200 m contour on the east. The eastern side of the hanging wall, and all of the area outside the Atlantis Massif footprint, has an interface depth of 1.5 km below seafloor. We assume density contrasts of 1600 kg/m3 at the seafloor and 300 kg/m3 at the mid-crustal interface, 1.5 km below the seafloor. The contributions of the seawater-crust interface, the “upper crust–lower crust” interface and a lithospheric thermal model are subtracted from the FAA to obtain a residual gravity anomaly (RA1.9_m1, Figure 5a).
4.1. Density Distribution
 The overall pattern in the new residual gravity anomaly maps (Figure 3) is similar to that obtained from the previous compilation [Blackman et al., 1998]. A number of artifacts at the edges of the prior compilation are eliminated in the new map. The largest positive anomalies occur along a ∼15 km wide strip on the inside-corner-created lithosphere along either side of the transform fault. This positive anomaly band continues westward along the inactive portion of the fracture zone on the flank of the northern ridge, but such a band does not extend beyond the active transform zone on the eastern flank of the southern segment. The new anomaly map shows that these strips of gravity highs are distinct, and an intervening band of lower gravity values separates them across the transform valley (Figure 3). This had not been clear in the prior studies [Pariso et al., 1995; Blackman et al., 1998]. The WOCC, like other OCCs, has a distinct positive anomaly that favors the transform and ridge side of the topographic high. Similar to the prior anomaly map, the current inside corner at the western RTI has a positive gravity signature, but it is now clear that its amplitude is less than that of the SOCC.
 Both the current western inside corner area and the section of the flank between the Keystone (Figure 1) and Atlantis Massif on the flank of the northern segment have excess mass in the subsurface, despite the fact that their topography is not particularly distinctive. Narrow lineated abyssal hills are missing in the 6–10 km region immediately adjacent to the transform fault in these areas. Instead, isolated NNE-SSW ridges spaced ∼10 km apart bound areas that slope at moderate angle down toward the active transform fault (Figure 1a). This contrasts with the morphology on the outside corner flanks and also with that on the eastern flank of the southern ridge across from and eastward of Atlantis Massif, where abyssal hills extend up to the fracture zone.
 The eastern outside corner (∼30°05′N, 41°50′W) is now recognizable, compared to the prior maps, as having a mass deficit (i.e., negative residual anomaly) relative to the assumed constant crust models (RA1.7, RA1.85; Figure 3). This is not just in comparison with the conjugate inside corner. Further east on the outside corner flank of the north segment (∼41°40′W and beyond), residual anomaly values are higher than they are in the 10–15 km interval just east of the nodal basin. The 5–10 mGal lower values of this outside corner are similar to those over lithosphere created at the southern ridge segment away from the band of positive anomalies along the active transform. However, note that the western outside corner (∼30°02′N, 42°47′W) has mid-range residual anomaly values, not the lower values seen at the eastern outside corner or on the older flanks of the southern segment. Only further west on the southern flank (beyond ∼43°05′W), at ages older than the crust conjugate to SOCC, do residual anomaly values consistently drop to levels comparable to the eastern outside corner.
 Accounting for a contribution from the structure of the core and hanging wall at Atlantis Massif, with a 3-D intracrustal interface of density contrast 300 kg/m3 (and thus lower crustal density of 2900 kg/m3) can explain most of the residual gravity anomaly there (model RA1.9_m1, Figure 5a). A relative high of a few mGals remains on the northern SE shoulder and the eastern slope of the central dome. Most of the southern ridge, central dome, and hanging wall have negligible remaining gravity signal with this simple 3-D model. The sensitivity of this result to assumed density contrast at the intracrustal boundary is illustrated in Figure 5b, where density below the interface is reduced by 50 kg/m3 to 2850 kg/m3. A larger positive residual anomaly now exists over part of the OCC. Figure 5c shows, for comparison, the estimate of the residual anomaly for a seafloor density contrast of 1850 kg/m3 and no intracrustal interface. This highlights the influence of the 3-D structure on the gravity signature. Looking at Figure 5c, it would seem that the relative anomaly between the southeastern core and the lithosphere away from the OCC is 10–15 mGals. However, if our more realistic model of the 3-D “upper” and “lower” crustal distributions is employed, the relative anomaly at Atlantis Massif is more like 4–6 mGal and its extent is much reduced. We prefer the model with core density of 2900 kg/m3 since it simulates measured values at IODP Site 1309 better than a value of 2850 kg/m3.
 Models similar to the 3-D Atlantis Massif higher-density core would account for much of the SOCC and WOCC residual gravity anomalies. In these two cases, large hanging wall blocks do not abut the core but both these OCCs have a domal shape, which presumably reflects a similar structural (and compositional) core. Rather than generating ad hoc subsurface models that lack geologic constraint, we note that the results of a simple model with density contrast at the seawater-crust interface of 1850 kg/m3 and no intracrustal interface (RA1.85) removes much of the signal over WOCC and SOCC (Figure 3b). By comparison, the case with a seawater-crust interface of 1700 kg/m3 (RA1.7, Figure 3a), results in a poorer fit with larger positive anomalies over each OCC.
 Although a layered crustal model, with specified crust-mantle interface, is probably not appropriate at a slow spreading ridge, the predicted gravity for such a model can provide an estimate of the overall volume of mafic crust within portions of young lithosphere. An area with a history of less intrusive activity, such as might typically characterize segment ends, can be approximated as having a greater volume (shallower interface depth) of peridotite (density ∼3300 kg/m3) below the interface, compared to a lower volume (deeper interface) or lower density (2900–3200 kg/m3) of mixed gabbro and variably altered periditote. Since the actual geometry of mafic intrusions within the mantle lithosphere cannot be uniquely determined by gravity modeling, it is more useful to assess possible limits on the range of structures consistent with the data. We test a model (Figure 4b) where the crust-mantle interface (density contrast 400 kg/m3) parallels the shape of the intracrustal interface (density contrast 300 kg/m3) at an average depth of 6 km subseafloor [Kuo and Forsyth, 1988; White et al., 2001]. Beneath the domal core of Atlantis Massif, the crust-mantle interface shoals to 4.6 km subseafloor. Removing the modeled crust/mantle contribution from the residual anomaly with 3-D upper crustal structure (RA1.9_m1, Figure 6a) gives the residual mantle Bouguer anomaly map (RMA1.9_m1) shown in Figure 6b.
 The RMA1.9_m1 (variable depth mantle interface, Model 1) map shows that the high-density core of Atlantis Massif OCC is confined to the central and southern domes. With this model a large negative gravity anomaly results north of the 13°14′N transition (Figures 1c and 6b) for the elongated core of Model 1 (Figure 4b). This misfit was not obvious when just a variable upper crustal model was considered (Figure 6a). The combined interfaces have clearly removed too much mass from this northern part of the range. A model in which the core of Atlantis Massif only extends as far north as the 13°14′N transition (Model 2, Figure 4c) provides a better match to the gravity data, either without (RA1.9_m2, Figure 6c), but especially if we include a variable mantle interface (RMA1.9_m2, Figure 6d). Both 3-D OCC models with a mantle interface (RMA1.9_m1, RMA1.9_m2) over estimate the contribution of the SW shoulder, as evidenced by the resulting negative anomaly in that area. This mass deficit indicates the presence of more lower-density material there, either as a thicker upper crustal component or a somewhat thicker mafic intrusive section. Given the limited lateral extent of the anomaly, the former is more likely. No rock samples have been obtained from the SW shoulder, but an ArgoII video imaging survey in 2000 [Blackman et al., 2002] mapped a number of basaltic features atop the southwestern part of this shoulder. Only at the edge, descending to the uppermost south wall (∼30°07′N, 42°12′W), did this Argo survey image outcrop characteristic of serpentinized peridotite.
4.2. OCC Morphology
 The gross morphology of the three OCCs along the Atlantis fracture zone is similar. Peak topographic heights are on the order of 1000 m, distance of the peak from the fracture zone is ∼15 km (Figure 7), spreading-parallel width of the corrugated surface is ∼10 km, and the ridge-parallel extent of the domes is ∼15 km if we accept Model 2 (Figure 4c) to set the extent of “anomalous” lithosphere and thus the width of Atlantis Massif. Atlantis Massif displays more prominent headwall scarps on the transform side than do the other OCCs, and this probably reflects degradation of formerly steeper slopes on the timescale of 1–2 Ma (the difference in age between Atlantis Massif and SOCC).
 The transform side of WOCC has a ∼5 km wide, stepped bench that may comprise a series of slumps (Figure 8a). Similar slumps are not as clearly recognizable on the transform side of SOCC. The inferred slumps at WOCC formed while this feature was still part of the active transform zone, since there is a clear scarp along their southern edge that is transform-parallel. Only a small debris lobe on the western edge of the lowest bench may extend across the trace of the fracture zone.
 Deep, flat, presumably sediment filled embayments are associated with each OCC in this area. The current nodal basin at the eastern RTI is paired with Atlantis Massif. A smaller embayment occurs on the western flank of the southern ridge segment at the location conjugate to SOCC (longitude ∼43° 02′W). This is the clearest embayment observed south of the fracture zone along the stretch covered in this study. Another, less clear example is centered ∼42°50′W. The conjugate to WOCC is outside the study area but embayments occur on the eastern flank of the northern ridge starting at a position that would have been conjugate to the Keystone block.
 The Keystone block (Figure 1a) has a surface character that is similar to that of the domal cores of the OCCs (Figure 8b) although it is quite different in terms of position relative to the transform fault. Preservation of the corrugated domal topography of the Keystone block may have been aided by the lack of an over-steepened slope, which would have made it conducive to slope failure. The northern side of the Keystone block is significantly more abrupt than the southern side and both are less steep than the east and west scarps.
 The presence of intrusive mafic rocks essentially just below the seafloor at OCCs can account for much of the positive gravity anomaly observed at these features in the MAR 30°N area. The lack of lower-density upper crustal sequences in the footwalls of OCCs, together with the 3-D structure of the detachment and basaltic hanging wall, provides enough contrast to explain most of the residual gravity anomaly highs. This was determined explicitly for Atlantis Massif and is inferred for WOCC and SOCC on the basis of basic similarities in morphology and gravity signature. A result where all the OCCs in this area have mafic cores is consistent with the model of Ildefonse et al. , in which the core of an OCC is a gabbroic batholith, and subsequent detachment faults localize in the surrounding serpentinized peridotite that deforms more easily.
 Our 3-D gravity model is a simplest-case test; it has constant density throughout the domal core of Atlantis Massif. The simplest geologic interpretation of the fact that such a model can explain almost all of the gravity anomaly is that the central and southern domes are similar in composition, both dominantly gabbroic and of similar genesis. This is what the model of Ildefonse et al.  would predict. However, a mix of significantly serpentinized peridotite and lesser gabbro could also combine to an average density of 2900 kg/m3. The latter case would support the model put forward by Karson et al.  where the Southern Ridge is predicted to be dominantly ultramafic in contrast to the mafic Central Dome. Along similar lines, although at somewhat larger along-strike scale, seafloor mapping studies of OCC along the Kane fracture zone lead Dick et al.  to prefer a model where domal cores can coincide with either dominantly gabbroic or dominantly serpentinized peridotite sections. We retain a measure of caution in our current endorsement of the simplest model (single, dominantly mafic core) since each time an OCC inferred to have ultramafic footwall on the basis of models and seafloor mapping nearby has been drilled, a gabbroic sequence has been recovered. However, we recognize that the average density of 2900 kg/m3 would allow the presence of a considerable amount of serpentinized peridotite within the core; the average density is low enough that the alteration level would have to be significant throughout at least the upper few kilometers.
 The presence of unaltered peridotite with only minor amounts of mafic intrusives (effective density 3300 kg/m3) at depths of ∼4.5 km within the core of Atlantis Massif would account for the full residual mantle anomaly (Figure 6). The success of this model depends on the core structure being confined to the central and southern dome, not extending north beyond the 13°14′N transition. The several-mGal negative residual gravity anomaly (RMA1.9_m2) over the SW shoulder implies generally less high-density material in that part of the southern ridge, as had been suggested by prior 2-D gravity profile modeling [Blackman et al., 1998, 2002] and MCS interpretation [Canales et al., 2004].
 Alternatively, if unintruded, unaltered mantle is not shallower than average at Atlantis Massif, the presence of oxide-rich mafic intervals could be responsible for at least a portion of the positive residual gravity anomaly (RA1.9_m2, Figure 6c). Recovered core from Hole U1309D that was classified as oxide gabbro was 7% of the total. In the 200–420 mbsf interval, this rock type is 10–15% of the total. Three other intervals that are each ∼100 m thick contained 20–25% oxide gabbro (650, 900, and 1250 mbsf) in the recovered core. The average density of discrete core samples of oxide gabbro was 3000 kg/m3 but about half had density 3100 kg/m3 or greater. The coarsest-grained of these samples had density as high as 3400 kg/m3. To contribute measurably to the sea surface gravity signature, an oxide-gabbro-rich interval(s) would need to extend laterally several hundred meters. A buried, vertical cylinder model provides a guide to what would be required. A 200-m thick interval (cylinder height) with density of 3100 kg/m3 and radius several hundred meters, at a depth of several hundred meters will produce a signal of 3–6 mGals (values of radius = 500 m, depth = 500 m, and height = 200 m give 3.9 mGal peak anomaly) if buried within a background material of 2900 kg/m3. At present, we cannot rule out such a scenario, but data available suggest that oxide gabbros intrude mainly as individual (sub)meter-scale lenses. Therefore, developing a ∼100-m interval of oxide-rich intrusions would require some sort of sustained zone (of weakness?) into which fluid is repeatedly injected over a period of time. Our preferred explanation is the former one, where unaltered mantle peridotite tends to be somewhat shallower, on average, beneath OCC, but without further seismic constraints we cannot confirm this model.
 Gravity constraints on the along-strike extent of the Atlantis Massif core suggest that the size of the core at all three OCCs in this area is about the same (on the basis of morphology and gravity signature for WOCC and SOCC). They also raise a question about the along-strike character of the hanging wall at Atlantis Massif. This easternmost block of the normal fault system for the full massif had been considered to be a single structural element [Cann et al., 1997; Blackman et al., 1998, 2002] but this may not be the case. The depth to the top and possible slight back tilt is constant along the length of the hanging wall block, but there is a westward step and steepening of the bounding eastern scarp at the 13°14′N transition where the corrugated central dome ends. Flat-topped seamounts are observed in a number of places on the southern part of the hanging wall but not north of the transition. Canales et al.  interpret a discontinuous reflector along the full length of this eastern block as the possible extension of the detachment fault beneath the hanging wall. However, the reflectivity character changes significantly throughout the imaged section along their N-S line. Stronger and more continuous reflectors characterize the portion of the line south of the transition. Although increased scattering due to rough seafloor on the northern part of the line may play a role, this rough seafloor extends south of the 13°14′N transition so it is likely that actual differences in the subsurface structure exist between the northern and southern hanging wall blocks. Mechanically, this could accommodate the differential faulting and uplift that is observed.
 The ∼5 mGal negative residual mantle anomaly (Figure 6b) on the outside corner of the eastern RTI likely corresponds to a thicker extrusive section than is typical for the northern ridge segment during the past ∼10 Ma. Canales et al.  document Layer 2A thicknesses of 600–900 m on the outside corner, which is 200–250 m greater than that observed in other sections of the MAR. With just their single MCS line in the area, it is not possible to assess changes with time that might be associated with the evolution of the OCC on the conjugate lithosphere. In a study of the southern end of a nearby segment (∼29°N), Allerton et al.  used deep-tow magnetic anomalies to determine that crustal accretion there is highly asymmetric, with a greater proportion of magmatic crust accreting to the outside corner.
 The positive residual gravity anomaly over the WOCC is not as great as that over SOCC and Atlantis Massif. This may reflect processes associated with evolution of an OCC, although differences in the core density cannot be ruled out. Mechanical erosion probably plays a role in lowering the overall density of OCC through time, with talus being less dense than solid rock. The linear edge of the slump benches that step down toward the fracture zone valley (as opposed to scalloped headwalls and lobate terminal ends) suggest fault control of their formation process. Cracking would introduce seawater into the core that would contribute to weakening but also, in olivine-rich rock, alter the host to a lower-density mineral assemblage. The fact that the toe of the slump retains a clean scarp exactly along the fracture zone trace indicates that most of the evolution took place in the 4–5 Ma following initial development of WOCC, while it was rafted from the inside corner along the active transform fault.
 The consistent ∼15 km distance of the peak of each OCC from the transform fault may reflect the flexural properties (flexural wavelength, effective elastic thickness) of young lithosphere that undergoes detachment faulting. The presence of the Keystone block as a domal surface extending continuously up from the transform fault suggests that an active detachment fault can plunge in the ridge-parallel, cross-slip direction all the way into the transform fault at an RTI. If this is a scenario that consistently contributes to OCC formation, evidence of it is often rapidly eroded (within 0.5–2 Ma, based on the lack of transform-adjacent corrugated surface on the south side of Atlantis Massif).
 Inside corner lithosphere is often characterized by higher than average density (Figures 3 and 6), but in reflecting on the overall level of magmatism at segment ends, the average of both inside and outside corner crust needs to be considered. The distribution of intrusive and extrusive magmatism with respect to the spreading axis may be important. The currently available gravity and seismic data suggest that segment ends near Atlantis transform receive an equivalent of about 1 km less melt than the segment center. Preferential accretion of extrusives to the outside corner crust is suggested to occur when detachment faulting is persistent [e.g., Canales et al., 2004; Searle et al., 2003]. However, current data are limited (especially seismic and outside corner geology). Additional coverage is required to address detailed relationships between OCCs, the distribution of magmatic material, and the tectonic evolution of the plates that initially form in the axial rift zone.
 The new compilation of gravity and bathymetry data analyzed in this study includes more detail over oceanic core complexes and has expanded coverage on the Mid-Atlantic Ridge flanks near 30°N. The positive seafloor Bouguer gravity anomalies of the three OCCs along the Atlantis fracture zone can be attributed to the presence of mafic intrusives essentially at the seafloor. Three-dimensional effects of variably distributed extrusive upper crustal rocks contribute to the gravity anomaly pattern in the vicinity of these shallow, high-relief features. The high-density, gabbroic core of Atlantis Massif, exposed via detachment faulting, is found to terminate within a short distance of the northern extent of the corrugated domal surface. The southern ridge at Atlantis Massif has overall density that is similar to that of the central dome, which suggests a similar, dominantly mafic composition. However, the data cannot distinguish between this interpretation and the possibility that significant amounts of highly serpentinized peridotite occur within the southern ridge. Although the maximum gravity anomalies in this area occur at OCCs, gravity highs characterize a broader spreading-parallel band along lithosphere formed at the inside-corners at both ends of the transform. This suggests that crustal accretion at this part of the plate boundary is often less consistent at the inside corners than it is at outside corners, whether due to biased magmatic injection or the geometry of tectonism. The positive anomaly bands are strongest on either side of the active transform fault, indicating that weathering becomes a dominant factor in changes of shallow density structure of the flanks along the inactive fracture zone trace. The 9 Ma WOCC has a gravity signal ∼10 mGals lower than the younger OCCs (3 and ∼1 Ma, respectively), yet the topographic signal is similar. This suggests that the density may have been altered over time, due to a combination of mechanical weathering and chemical alteration (serpentinization).
 Free Air Anomaly (FAA) data from cruises EW9210 and the MCS cruise (EW0102) were obtained from the National Geophysical Data Center and the PIs, respectively. Weather was poor during EW0102 and the higher noise level along those tracks reflects higher sea-state conditions. The data were not smoothed beyond the normal shipboard processing, so in the detailed anomaly maps, the level of measurement uncertainty for this cruise is evident as small-scale jitter about the local trends.
 The XSystem crossover error (COE) analysis package [Wessel, 1989] was used on individual and combined cruises. Mean COE within each cruise was less than 0.5 mGals. Standard deviation was 3.1 mGal for EW0102 and ≤1.85 mGal for the other cruises. Best fit drift corrections and offsets determined for the full matrix of COE were applied to the gravity data from each cruise.
A2. IODP Core and Logging at Hole U1309D
 As part of IODP Expeditions 304 and 305, measurements of bulk density and porosity were obtained on ∼8 cm3 samples cut every 2–3 m from Core U1309D (methods described by Blackman et al. ). Running averages over 10 m sections of core were computed from the individual core sample measurements (Figure 2), to smooth out variability arising from differences in rock type where, for example, thin lenses might contain oxide gabbro which commonly was more dense than other types of gabbro. Porosity measured on the core samples ranged from 1% to 6% in the upper 750 m of the hole. Below that depth, core sample porosity was small, 0.6% on average.
 Downhole logging was conducted three times in the two drilling expeditions. In each case a main and a partial repeat pass run was done with a suite of tools that included the high-resolution Hostile Environment Litho-density Sonde (http://iodp.ldeo.columbia.edu/TOOLS_LABS/TRIPLE/hlds.html). All density measurements were averaged between the passes and then a 10-m running average was applied (Figure 2). Caliper measurements indicate that the hole was generally in very good condition but intervals with local breakouts did occur, commonly in the upper few hundred meters. In those sections, logged density measurements are inaccurate so only data below 200 m are shown in Figure 2. As expected, core sample density maps at the upper limit of the logged density, since the samples were chosen to avoid fractures and highly cracked intervals. Further, logging measurement integrates over a scale of tens of centimeters that may include in situ porosity/cracks.
A3. Topography Model
 The topography compilation of [Blackman et al., 1998] included Simrad EM12S data from RRS Charles Darwin cruise CD100 and 200 m gridded data from the RIDGE Multibeam Synthesis (http://www.marine-geo.org/rmbs/), which is mainly based on SeaBeam data from R/V Conrad cruise RC2912. Gridded Hydrosweep data from R/V Ewing cruise EW9210 have been added. Additional SeaBeam 2000 data from R/V Atlantis cruise AT3-60 was obtained in 2000. Figure A1 shows the location of the AT3-60 and CD100 track lines that provide complete coverage of the ridge-transform-ridge area and WOCC. The 200 m grid data fills the remainder of the box shown in Figure 1. The swath sonar beams have a footprint that varies with seafloor depth. For most of the survey area this is on the order of 100 m; the deep transform valley floor and along its steep walls have coverage at lower resolution. The track lines for both CD100 and AT3-60 were acquired at an angle to the dominant morphologic grain, to optimize insonification of the ridge- and spreading-parallel features by the swath sonar system.
A4. Thermal Model
 Cooling and contraction of the lithosphere with age probably contributes ∼30 mGals to the Bouguer gravity anomaly from the spreading axes to the oldest crust in the study area. An estimate of this contribution is obtained from a 3-D ridge-transform-ridge flow and thermal model [Phipps Morgan and Forsyth, 1988]. This method accounts for cooling across the transform as well as the cooling with distance from the axis, which is similar to that predicted for square root of plate age [Parsons and Sclater, 1977]. The gravity contribution of the predicted thermal model is computed using a coefficient of thermal expansion of 3.55 × 10−5/°C, as determined by Pariso et al.  to best predict the observed seafloor deepening for plate ages 3–10 Ma along the 29°–31°30′N section of the MAR.
 The predicted plate cooling contribution varies slowly across the study area so details of the relative residual anomaly (after the plate cooling signature is subtracted from the Bouguer anomaly) are not affected by inaccuracies in details of the thermal model. The main reason for applying this adjustment to the Bouguer anomaly map is to accentuate the signature of the WOCC, for comparison to that of Atlantis Massif and SOCC. In the Bouguer anomaly map (BA) (Figure 1b and Table 1), it is difficult to determine whether the WOCC has a distinct anomaly or if it is included in a broader feature extending to the west with positive Bouguer signature. In the residual anomaly map (Figure 3), the WOCC gravity anomaly is clear.
 In the previous study [Blackman et al., 1998], a transform length of 70 km was used to compute the plate cooling contribution to the gravity based on the reports using magnetic anomaly patterns to define the axis of the spreading zone [Zervas et al., 1995]. A more accurate location of the ridge axes can be obtained from the CD100 bathymetry data (Cann et al., 1997). A small volcano with a distinct caldera is evident at the RTI of the southern segment so that is taken as the position of the spreading axis. The center of the rift valley is chosen for the axis of the northern segment. The offset between these locations is 62 km so this value for the ATF length was used to obtain an updated prediction of the plate cooling contribution to the Bouguer gravity anomaly.
A5. Determination of Residual Gravity Anomaly
 Predicted seafloor topography and lithospheric thermal contributions are computed for a grid that corresponds to the bathymetry map (Figure 1b). Five terms in the power series expansion of the Parker  solution are required, since changes in the predicted topographic contribution near shallow, high-relief seafloor is more than a fraction of a mGal up to the fifth-order term. The predicted contributions are interpolated onto the track positions at the points where the FAA is archived. The track line values, adjusted by the amount of the predicted contribution are then passed through GMT's blockmedian program for 100-m grid spacing. A minimum curvature surface is fit to these values, with tension parameter of 0.25, to create a map grid.
 After the first author, author order is alphabetical with Karner and Searle contributing equally to the work reported in the paper. Data acquisition for this study was supported by NERC/BRIDGE program, NSF/OCE-9712164, and IODP Expeditions 304/305. Postcruise efforts were supported by Joint Oceanographic Institutions grants to D.K.B. (JOI T304B22 and T305A22) and G.D.K. (JOI T305A33). Thanks go to Brian Tucholke for providing gravity data along the Ewing 2001 seismic tracks. Maps and gravity model generation used GMT [Wessel and Smith, 1998]. Suzanne Lyons ping-edited the AT3-60 SeaBeam data that were added to the topography compilation for this study.