Hole 735B provides a nearly complete 1508-m section of the lower ocean crust generated at the Southwest Indian Ridge at ~11.5 Ma. The block underlying Atlantis Bank is anomalous in that it has been unroofed and uplifted to shallow levels; however, it consists of rocks whose seismic velocities and physical properties are appropriate for oceanic Layer 3 (Iturrino et al., 1991; Dick, Natland, Miller, et al., 1999). Although the section is cut by a number of ductile shear zones and at least one major fault, it appears to be relatively coherent. Thus, it provides a unique opportunity to investigate the physical and magmatic processes involved in formation of the lower ocean crust at an ultra-slow-spreading ridge.
A total of 952 lithologic intervals, bounded by igneous or tectonic contacts, were recognized by the Leg 118 and 176 shipboard parties and Dick et al. (1991b). These were grouped into 12 major lithologic units, based primarily on modal mineralogy. Unit I, however, was defined on the basis of extensive crystal-plastic deformation and metamorphism. Highly deformed gabbros similar to those of Unit I occur extensively on the top of Atlantis Bank (MacLeod et al., 1998) and are considered to mark the detachment fault that unroofed the block.
The bulk of the cored section consists of relatively primitive olivine gabbro, troctolite, and troctolitic gabbro. These rocks are intruded by a variety of more evolved gabbros, gabbronorites and oxide-bearing and oxide-rich gabbros and gabbronorites. Most of the oxide-bearing and oxide-rich intervals are relatively thin, generally no more than a few centimeters thick. However, in lithologic Unit IV, defined on its abundance of oxide-bearing rocks, there are 28 intervals of oxide-bearing and oxide-rich gabbro and gabbronorite with an average thickness of 1.76 m (Natland and Dick, Synthesis Chapter, this volume).
Because the lithologic units were defined on a variety of criteria, they have limited genetic significance. However, downhole plots of mineral compositions and bulk-rock geochemistry reveal a number of correlations of these parameters with the lithologic units. Such correlations are most apparent in downhole plots of Mg# for clinopyroxene, An content of plagioclase, bulk-rock TiO2 contents, and bulk-rock Mg#. Plagioclase compositions and bulk-rock Mg#s must be interpreted with care because the former are very sensitive to alteration and the latter to the presence of Fe-Ti oxides. However, Mg#s for clinopyroxene (Fig. F5) correlate well with most lithologic units and exhibit similar trends within many units. In lithologic Units XII, XI, IX, VI, and V the clinopyroxene Mg# is highest at the base and decreases gradually upward. Unit IV exhibits a reverse trend, going from low Mg#s at the base to higher values at the top. Regular trends are not apparent in the other units.
Similar, but less well-defined trends are observed for An contents of plagioclase (Fig. F6). In several lithologic units, particularly Units I through V, plagioclase compositions have a bimodal distribution, reflecting extensive alteration and metamorphism.
Olivine compositions are more variable, but some trends are apparent, particularly in the lower part of the hole. In lithologic Units XII, XI, X, and IX, olivine compositions generally become more iron rich upward within each unit and within the sequence as a whole (Fig. F3). Olivine compositions are more variable in the upper part of the sequence, reflecting the intrusion of relatively evolved oxide and oxide-bearing gabbros and gabbronorites. Similar patterns are shown by orthopyroxene compositions, although the number of analyses is much lower (Fig. F4).
The Leg 176 Shipboard Party (1999), Dick et al. (2000), and Robinson et al. (2000) recognized five major geochemical cycles in the 1508-m section based largely on whole-rock Mg#s and interpreted the cycles to reflect the intrusion of five separate plutons. Upon reviewing all of the postcruise geochemical and mineralogical data, Natland and Dick (Synthesis Chapter, this volume) recognize three geochemical series within the gabbros, olivine gabbros, and troctolites, which extend from 0-232, 274-520, and 520-1507 mbsf, respectively. The gap between 232 mbsf and 274 mbsf is marked by the oxide and oxide-bearing gabbros of lithologic Unit IV. Natland and Dick (Synthesis Chapter, this volume) divided Series III into three subunits: 520-962, 962-194, and 1194-1508 mbsf, labeled 3a, 3b, and 3c, respectively. These are close to the lower three geochemical cycles recognized by Dick et al. (2000) and Robinson et al. (2000) but with slight changes in the positions of the boundaries. This does not represent a major reinterpretation of the geochemical stratigraphy but rather a re-evaluation of the significance of the lower three cycles. Whether these cycles represent individual plutons or repeated pulses of magma into a single magma chamber is a matter of interpretation. However, the overall geochemical coherence of these three cycles supports the latter interpretation.
It is clear from the downhole plot of whole-rock Mg#s (Fig. F14) that the most primitive olivine gabbros and troctolites occur in the upper parts of the section where they comprise geochemical cycles 1 and 2 (Series 1 and 2 of Natland and Dick, Synthesis Chapter, this volume). In both cases, the most primitive rocks occur at the base of the cycle and the most evolved rocks at the top with Mg#s decreasing progressively upward. We interpret each of these cycles as representing a magmatic intrusion that underwent fractionation in place. Natland and Dick (Synthesis Chapter, this volume) suggest that the small-scale variations within these cycles indicate that these plutons were constructed from numerous individual injections rather than large pulses of magma, an interpretation consistent with the general absence of layering within the sequence. However, if this was the case, the pulses must have come from a fractionating body at greater depth to account for the smooth upward trend within each cycle.
The long section of olivine gabbro in the lower part of the section from 520 to 1508 mbsf also shows a relatively continuous upward evolution from primitive to less primitive compositions but with small breaks at the boundaries of the lithologic units. Again, small variations in composition within this sequence suggest intrusion of many magma pulses (Natland and Dick, Synthesis Chapter, this volume), but this interpretation is difficult to reconcile with the relatively consistent upward fractionation trend. We suggest that the section was formed from three major magmatic pulses (Series 3a, 3b, and 3c of Natland and Dick, Synthesis Chapter, this volume) which underwent fractionation in place. The small-scale compositional variations may reflect pulses of more mafic melt into these bodies or perhaps redistribution of interstitial liquids within the magma chamber.
As pointed out above, the most mafic rocks in Hole 735B occur at the top of the section and the most differentiated rocks at the base. The relative ages of the different bodies is unknown, but Natland and Dick (Synthesis Chapter, this volume) suggest that the crustal section was formed from the top down by underplating. In this model melts rising from the mantle move upward until they encounter a permeability barrier, at which point they coalesce to form a magma chamber. The primitive olivine gabbros and troctolites at the top of the section represent small, solidified magma chambers from which melts may have erupted to the seafloor to form the basaltic carapace. The less differentiated gabbros lower in the section would have formed after the high-level plutons by injection of ferrobasaltic melts into the lower crust. Natland and Dick (Synthesis Chapter, this volume) suggest that the ferrobasaltic melts were produced by fractionation of the more mafic magmas that formed the small plutons at the top of the section. This is a reasonable model for construction of the cored section in Hole 735B, but more information is needed on the relative ages of the plutons and on the nature of the unsampled section beneath the hole before it can be adequately tested.
The compositions of the melts from which the Hole 735B gabbros crystallized can be estimated using the method of Roeder and Emslie (1970). Calculated melt compositions for the olivine gabbros and troctolites are all quite highly differentiated with Mg#s generally between 35 and 60. The olivine gabbros from the lower 987 m of Hole 735B crystallized from ferrobasalt liquids with Mg#s generally between 35 and 45. Even the most primitive olivine gabbros and troctolites in the upper part of the section crystallized from relatively evolved abyssal tholeiites with Mg#s between ~50 and 60. Such ferrobasalts are virtually unknown from the Southwest Indian Ridge, suggesting that such liquids are trapped and crystallized in the lower crust before they can be erupted.
The bulk composition of the Hole 735B rocks is much too evolved to mass balance any reasonable crustal section back to a primitive melt composition (Dick et al., 2000). Thus, the parental magmas for the Hole 735B gabbros were considerably more evolved than any primary magma from the mantle suggesting an earlier stage of fractionation (Dick et al., 2000; Coogan et al., in press), which presumably would have produced primitive cumulates. If such cumulates exist they must lie either beneath the section sampled in Hole 735B or perhaps closer to the center of the transform-bounded ridge segment in which Atlantis Bank was formed. An alternative suggestion is that the relatively evolved melts were formed by reaction within the upper mantle and that the term "missing cumulates" is misleading (Coogan et al., in press).
One of the more surprising results of Legs 118 and 176 is the abundance of Fe-Ti oxide and oxide-bearing gabbros and gabbronorites in Hole 735B. These occur in layers and bands, a few millimeters to several meters thick, which appear to have intruded the dominant gabbro, olivine gabbro, and troctolite. Most, but not all of the oxide gabbros, are associated with zones of crystal-plastic deformation and, in many cases, the oxide minerals themselves are sheared and deformed. The origin of the melts from which the oxide gabbros and gabbronorites crystallized is unclear, but presumably they formed by extreme differentiation of the ferrobasaltic magmas. These are clearly highly evolved rocks as indicated by their modal and chemical compositions and their common association with trondhjemetic veins. However, the oxide gabbros are far too abundant to be mass balanced with the crustal section to any reasonable parental liquid. Thus, the Fe-Ti-rich melts must have formed beneath or adjacent to Hole 735B and have migrated into the section. It seems unlikely that such melts would move laterally along the ridge crest for any significant distance so they probably formed in the upper mantle or at the base of the crust beneath the drilled section. The Fe-Ti oxide melts would have been buoyant within a crystalline matrix (Natland and Dick, Synthesis Chapter, this volume) and would have risen to be concentrated in the upper parts of the crust. The extreme differentiation necessary to produce such melts must have formed an abundance of mafic cumulates beneath the ridge crest. Unlike the ferrogabbroic magmas mentioned above, the oxide-rich magmas could not have been produced by melt-rock reaction in the upper mantle. The mafic cumulates produced during formation of the oxide-rich melts may not lie directly beneath the drill hole because the section was emplaced along a detachment fault, which separated it from the underlying mantle. A major question to be resolved by future drilling is whether the abundance of Fe-Ti oxide gabbros is unique to this portion of the Southwest Indian Ridge or whether such rocks are widely present in the lower ocean crust.