Some 861 chemical analyses of gabbro were obtained on board ship and in several laboratories following Legs 118 and 176 (see "Appendix B"). Only a few of these were not additionally analyzed for the suite of trace elements that can be determined by XRF, and over 300 have been analyzed for rare earth elements as well. Several dozen of these have been analyzed for Sr, Nd, and/or Pb isotopes, and there are about 80 additional samples for which just the isotopes of Sr and O have been determined without additional geochemical measurements. These data indicate that the liquids parental to the gabbros were typical depleted mid-ocean-ridge basalt (MORB) and that the rocks have been modified to varying degrees by alteration (Stakes et al., 1991; Kempton et al., 1991; Hart et al., 1999; Bach et al., 2001; Holm, 2002.
To these should be added analyses of gabbros obtained during ODP Leg 179 (Shipboard Scientific Party, 1999b) and those acquired by dredging and high-speed near-surface diamond coring during the postcruise survey (Coogan et al., 2001), which allow spatial comparisons across the summit and down the flanks of Atlantis Bank. Nearly 1000 analyses of major oxides on gabbros alone are now available for this purpose.
Our strategy is to isolate the geochemical variability of primitive gabbros from the numerous seams of differentiated gabbros that crosscut them in so many places and to compare this now with the sequences of olivine gabbro having finer-grained and more olivine-rich intervals at their bases, as defined in "The Lithologic Section." We describe the core generally from the top down and take up additional topics as the descriptions warrant. We use the terms interval, composite interval, sequence, and series as they are defined in the main text of the paper.
See the shipboard site reports (Shipboard Scientific Party, 1989, 1999c) and Hertogen et al. (Chap. 6, this volume) for summaries of the geochemical attributes of the rocks drilled during Legs 118 and 176. The following criteria are used for plotting. We distinguish oxide gabbros as having >1% TiO2 contents; tonalite, trondhjemite, and diorite as having >54% SiO2 and up to 70% SiO2; differentiated gabbros, including disseminated oxide gabbros as having <54% SiO2 and Mg# (= 100 x Mg/[Mg + Fe2+]) < 72.5; olivine gabbros as the same, but with Mg# > 72.5 up to MgO = 12%; and troctolitic gabbros plus troctolites as having >12% MgO. Olivine pyroxenites are like troctolites, but with normative Di > 40%. Each of these is given a different symbol on diagrams (see caption to Fig. AF4). Some shipboard analyses obtained during Leg 176 have high totals, a consequence in large part of high SiO2 measurements (Shipboard Scientific Party, 1999a). Thus, some olivine gabbros have >54% SiO2 contents. These are taken into account. The break at Mg# = 72.5 corresponds to a clear gap in all reported data sets, evident on both variation diagrams and a histogram for Mg#. Rocks in the core with Mg# > 72.5 correspond to the general run of olivine gabbros and troctolites comprising the olivine-gabbro sequences as previously defined; rocks with Mg# < 72.5 usually intrude these.
Many of the problems of interpreting lithologic units and their relationship to chemical stratigraphy are exemplified by the upper 100 m of Hole 735B. Figure AF1 presents the variation in Mg# through this length of core. Mg# is a fair index of the fractionation stage (term of Wager, 1956) of the gabbros, since it nominally reflects the average composition of the ferromagnesian silicates in the rocks, regardless of variability in their modal proportions. Among layered intrusions such as the Skaergaard, for example, it correlates with the ratio of normative An/(An + Ab), which closely follows average feldspar compositions. Among troctolites, Mg# should be fairly close to that of the Fo content of olivines. In gabbros, which have little or no olivine and low-Ca pyroxene, the ratio should closely approximate the Mg# of clinopyroxene. Even in most olivine gabbros, clinopyroxene is far more abundant than olivine; thus, the Mg#s of such rocks should also be close to those of clinopyroxene. Mg# can obviously be lowered by the presence of the magmatic oxides ilmenite and magnetite. The extent to which gabbro Mg# can be utilized to specify original liquid Mg#s also depends on how much material frozen from trapped residual liquid the rocks still contain. If the rocks are adcumulates, however, they will contain only trivial amounts of oxide minerals and very little of material crystallized from trapped melt; therefore, this is not a great problem in these cores (Natland et al., 1991).
In any case, just considering that Mg# usefully discriminates rock types, Figure AF1 reveals an extraordinary complexity of even a short segment of Hole 735B. Lithologic units are indicated to the right, and sequences of olivine gabbro, as previously defined, to the left. There are 84 analyses of gabbro, and they divide roughly into two types, primitive and differentiated, as indicated by the two broad arrows at the top of the figure. Encircled fields indicate groups of analyses obtained from one to several adjacent lithologic intervals.
The analyses overall confirm the general impression given by the lithologic units. An upper sequence of ~20 m of differentiated gabbro (Subunit IA, gabbronorite and olivine gabbronorite) is succeeded downward by alternating lithologies of differentiated and primitive gabbro (Subunit IB), and then by assemblages dominated by olivine gabbro (Unit II). The unit boundaries were defined entirely in consideration of the prevalence of differentiated gabbro. The sequences of olivine gabbro have no relationship to them.
A disturbing aspect of Figure AF1 is the diversity of compositions found even in single lithologic intervals such as 48 and 50, which are both described as olivine gabbro by Dick et al. (1991a). The sequence to which these intervals belong consists of nearly 94% olivine gabbro based on the core descriptions, and the samples analyzed all came from those parts of the core.
Hart et al. (1999) attempted to get around the problem of short-scale lithologic diversity by obtaining 1-cm sample strips along 1.13- to 4.15-m lengths of core. These strip samples were then cleaned, powdered, and analyzed. The analyses then were construed to represent the bulk compositions of those lengths of core and to average out any fluctuations in lithology. In the upper 500 m of the core 22 such strip samples were obtained, and Hart et al. (1999) considered that the average composition of the 22 strip samples is the average composition of the Leg 118 portion of the core, down to 500 mbsf.
Strip samples in Figure AF1 are positioned at the tops of their sampled lengths; they are linked by a line and individually lettered. Five were taken from this part of the core, and all five fall in the category of differentiated gabbros. Four of them have >1% TiO2 contents and thus are oxide gabbros using the chemical breakdown described above. Two of the strip simples fall in a portion of the core dominated by olivine gabbros, yet are oxide gabbros in bulk composition. One of them (D) spanned portions of intervals 48-50, within which only one 12-cm seam of oxide gabbro, interval 49, was described. Apparently, this seam was either extremely oxide rich or there were additional thin seams of oxide gabbro nearby that were too narrow to see. Over intervals 48-50, totaling 14.87 m (expanded) thickness, narrow seams of oxide gabbro, whether detected or not, were successfully avoided by shipboard scientists in 17 out of 19 cases.
Table AT1 compares the average of the five strip samples with the average of all 76 analyses obtained through the base of olivine-gabbro sequence 4, interval 77. The strip samples have almost double the TiO2 contents of the average of 76 samples, lower Mg#, lower CaO, and higher Na2O—all trends indicating a higher proportion of differentiated gabbro among the strip samples. Ni and Zr concentrations, however, are similar.
The average strip composition of Hart et al. (1999) approximates a typical depleted and slightly differentiated abyssal tholeiite in composition (Table AT1). So does the average of the five strip samples from the upper 80 m of the core. The average of 76 individual samples, however, does not. It has a high Mg#, about that of a reasonable primary basalt and close to that expected for a liquid in composition with mantle olivine. However, the TiO2 content is substantially lower than it should be for this region of the Indian Ocean (cf., column 5, Table AT1). This and the high Mg# result from incorporation of olivine gabbro cumulates into the average of individual analyses, not one of which is represented by any one or in the average of the five strip samples. All averages are also sodic, having higher Na8 when treated as basalts than any actual basalt nearby.
The statistical basis of either means of sampling and averaging, therefore, is clearly open to question. If in the end one wants a valid assessment of the bulk composition of the core, the problem appears to lie both in identifying the populations of rock to be sampled and then properly sampling them statistically. No one has yet attempted to do this. Weighting to thicknesses of lithologic intervals (Dick et al., 2000) and obtaining average compositions of defined rock types seems proper, provided that all intervals are noted and their thicknesses is accurately determined. The great chemical variability within individual intervals and the compositions of the separate strip samples suggests that they have not been. Thus, even oxide gabbros are included among the analyses, mainly from the shipboard data set, used to determine the composition of "average olivine gabbro" (Dick et al., 2000). On the other hand, strip samples seem to be no better a statistical sampling of the core. Perhaps there are enough individual samples now to select a blind set of analyses to average. Any average, however, will be one of cumulates, with the possible exception of whatever granitic veinlets are represented. With all the new analyses, a new average should be attempted. It is still too early to say how closely a statistically valid average of a large number of cumulates over 1500 m of core will match primitive basalt (compare columns 4 and 5 of Table AT1).
For all of these difficulties, consider alternatively another manner of detailed sampling of the same rocks. During the postcruise survey aboard the British research vessel James Clark Ross, short cores (0.1-3.8 m) of gabbro were obtained using rock drills constructed at the British Geological Survey at 21 locations, most of them along a north-south transect across the summit of Atlantis Bank (Coogan et al., 2000). From these, 26 samples were analyzed. To some degree, an average of the 26 samples might be considered representative of the entire gabbro massif. The available exposure was sampled from one end to the other at an average spacing of 175 m and was done far more systematically than is possible with dredging. But the 26 samples fall far short of encompassing the full diversity of rocks in Hole 735B. They include no troctolite, no olivine pyroxenite, no tonalite/trondhjemite veinlets, and no oxide gabbros with high P2O5, reflecting the presence of abundant apatite. The redistributed incompatible components of basalt are poorly represented. This merely says that if we wish to obtain an average composition for the lower ocean crust at this location or at any other slowly spreading ridge, spot sampling will not suffice. We need to drill deep holes and to obtain continuous core, as we have done in Hole 735B.
In Figure AF2, Mg#s and Ni concentrations of analyzed gabbros are plotted vs. depth over the upper 250 m of the core. This is to encompass the uppermost of the plutons originally identified on the basis of chemistry, now designated as Series 1. Boundaries of lithologic units are again to the right; the bases of sequences of olivine gabbros (Table AT1) are to the left. Sequences 1-19 comprise the series, and the 19 sequences comprise 39 composite intervals (Table AT1). With postcruise data, Mg#s of primitive gabbros still clearly trend toward lower values with elevation (Figure AF2A), and the scatter increases as well. Even excluding oxide gabbros, there is a distinctive bimodality to the rock compositions. Gabbronorites and olivine gabbronorites comprise the greatest number of crosscutting intervals most of the way, and between 0-18 and 190-210 mbsf they are the predominant rock types. Oxide gabbros of Unit IV predominate in the lowest 30 m, but three small screens of primitive olivine gabbro are caught up in these at ~230 mbsf.
The sequences show no consistent internal pattern of variation, except for a tendency for samples with highest Mg# to occur at or near the bases of several of them. Sequences 11 and 17 are both more differentiated upward, a tendency that is more obvious for Ni concentrations, which diminish regularly upward in each case (Figure AF2B). In this, they resemble variations in individual layers that are portions of cyclic units in layered intrusions (Irvine, 1980). High Ni indicates a high modal proportion of olivine at the base of each. The entire body of olivine gabbro therefore consists of alternations of rocks richer and poorer in olivine. The small scale of these fluctuations is additional evidence that the pluton was constructed from numerous individual injections rather than large pulses of mafic magma. Some oxide gabbros and even a few diorites have the Ni concentrations of olivine gabbros. The contradictory character of these analyses indicates that these rocks are mainly olivine gabbros within which narrow seams or veins of oxide gabbro or tonalite/trondhjemite were also sampled. Most gabbronorites and olivine gabbronorites of lithologic Subunits 1A and IIIA-IIIC and the oxide gabbros of Unit IV have consistently low Ni, thus are not hybrid samples but rock that crystallized from strongly differentiated liquid with very low Ni.
The aggregate thickness of all composite intervals of olivine gabbro and troctolite from sequences 1-19 is 140.7 m. The average thickness of 19 sequences is 7.4 m, and the thickest is 22.6 m. The latter is sequence 17, one of the two with steadily increasing Ni, thus olivine, toward its base.
Strip samples A-K of Hart et al. (1999) are indicated by letter and linked by lines in Figure AF2A. All but one fall in the range of differentiated gabbros, and seven of them have the bulk compositions of oxide gabbro. Sequences of olivine gabbro comprise 60.8% of the aggregate thickness of rock down to the base of interval 267. These are matched by only one strip sample, whereas the differentiated gabbros, making up 39.2% of the rock, are represented by ten strip samples. To this depth, primitive gabbros are strongly underrepresented by the strip samples.
The variation of Mg# and Ni with depth within Series 2 olivine gabbros and troctolites encountered in Hole 735B is shown in Figure AF3. As before, lithologic units are indicated to the right and olivine-gabbro sequences to the left. The full series comprises sequences 20-39. The top of the series is taken as the top of lithologic interval 287, an olivine gabbro only 12 cm thick that occurs as a screen near the base of the oxide gabbro of lithologic Unit IV. Since the rock is not analyzed, an "X" shows its position in Figure AF3A. The base of the series is the bottom of lithologic interval 504, a fine-grained olivine gabbro. The rocks immediately beneath these belonging to Series 3 are described as olivine gabbro and disseminated-oxide gabbro (Shipboard Scientific Party, 1999c), but the ones of these that have been analyzed are clearly more differentiated than the rocks of Series 2, having Mg#s of ~70.
The most remarkable portion of this series is sequence 20, which, apart from several narrow seams of oxide gabbro, consists mainly of one continuous body of coarse olivine gabbro more than 120 m thick; the bulk of this sequence makes up lithologic Unit V. There are eight internal igneous contacts within sequence 20, but seven of them are concentrated near its base. The sequence is more differentiated with lower Mg# at its top, and this is clearly an upward continuation of a general trend through sequences 39-21. However, there is very little scatter to the trend within sequence 20, indicating a fundamental uniformity and consistency to the compositions of the rocks. This is borne out by Ni (Fig. AF3B), which shows little scatter—no more than might result from variations in small samples of coarse-grained rock. Ni itself decreases upward within this one sequence. Again, this resembles the pattern of cryptic variation frequently encountered within individual layers of phase-layered cyclic units of layered intrusions.
The rocks of underlying sequences 21-39, however, are modally and texturally variable; thus, they have more scattered Mg#s, including some troctolites with higher Mg#s. The modal fluctuations are primarily in the proportion of olivine, as indicated by highly variable Ni contents (Fig. AF3B). Fluctuations in the proportions of intrusive differentiated gabbros define the subdivisions in lithologic units, with Unit VI being described as compound olivine gabbro, meaning that there are numerous seams of oxide gabbro in a matrix of olivine gabbro. Again, however, there is no relationship between these and the boundaries of olivine-gabbro sequences. Several analyses in this part of the core have both high Ni and high TiO2, reflective of samples containing both olivine gabbro (or troctolite) and some portion of a narrow seam of oxide gabbro. At the top of sequence 20 are several samples with the bulk composition of diorite (large red arrow). These are from the veined breccia just underlying the base of the major oxide gabbro of lithologic Unit IV. They represent differentiated olivine gabbro laced with tonalite/trondhjemite veinlets.
The aggregate thickness of olivine gabbro sequences 20-39 is 199.6 m, representing 69.1% of all the rock they span. The average thickness of a sequence is 10.0 m, and the thickest is 120.7 m. Discounting that one, the average thickness is 4.2 m with all of the remaining ones lying beneath it.
Lettered data points in Figure AF3A indicate the strip samples of Hart et al. (1999). In uniform lithologies, they closely fit the general trends. Among six strip samples below 400 mbsf, however, only two have the compositions of primitive gabbro; the others are differentiated, including two oxide gabbros and one diorite. The two oxide gabbros (strips T and V) have high Ni, thus are physical combinations of olivine gabbro and seams of oxide gabbro. Since unadulterated oxide gabbros in this part of the core are quite differentiated with very low Mg#s (0.35-0.42), the hybrid strip samples here may very well balance primitive and differentiated gabbro compositions to their approximate proportions in the core.
If the rocks of series 1 and 2 could be mapped over a large outcrop, conclusive evidence that these are essentially two intrusive bodies, or plutons as termed by Dick et al. (2000), might emerge. At this point, evidence in favor of this is the general compositional unity of each, sharp contrasts with rock above and below, sequential igneous contacts within and separating the sequences of olivine gabbro, and trends of upward increasing extent of differentiation in both.
Accepting this, however, we now speculate that these two composite masses of primitive gabbro, especially the lower one, behaved more like the magma chambers we might have expected to exist at this place than any other part of the core. At the slowly spreading Mid-Atlantic Ridge where basalts have been sampled at several places to depths of more than 500 mbsf by drilling, we know that the extrusive layer at any one place consists of sequences of hundreds of pillow basalts that break down into just a few chemically uniform basalt types (e.g. Bougault et al., 1979). These represent only a small number of actual eruptions (Natland, 1979), many of them separated by enough time (thousands of years) for significant secular variation, or even reversals, of the Earth's magnetic field to occur (Johnson, 1979). The uppermost crust appears to be an accumulation of monogenetic cones, each one erupted virtually at a point and each one consisting of pillow lavas with just one composition (e.g., Smith and Cann, 1992).
Nevertheless, the lavas are hybrids of primitive and more differentiated compositions, as reflected in both bulk compositions (e.g. Rhodes et al., 1979) and mineralogy (Dungan and Rhodes, 1978). Similar mixing histories have been deduced from compositions of phenocrysts in abyssal tholeiites from the Indian Ocean (Natland, 1991; LeRoex et al., 1989; Natland and Dick, 2001). If drilled thicknesses of these chemically uniform lava piles along the Mid-Atlantic Ridge are any indication, and assuming similar aspect ratios, the underlying storage reservoirs that supplied them held at least several meters, in some cases several tens of meters, and in one case more than 200 m of molten material, before eruption (Natland, 1979).
The sequences of olivine gabbros in the upper part of Hole 735B are the right order of magnitude in thickness to represent such magma bodies. Indeed, sequence 20, which is >120 m thick and includes one composite interval with uniform grain size some 109 m thick, evidently was held in the lower crust to freeze in its entirety without being disturbed by later injection of any other primitive magma as it froze. Whether it supplied an eruption prior to this is, of course, unknown. All other sequences in Series 1 and 2 apparently represent material of smaller crystallizing bodies that were repeatedly disturbed by injection of variably primitive and sometimes phenocryst-laden magma.
The patterns of magma flow and whether crystal mushes were invaded internally or whether there was ever something like a chamber floor in either of these bodies are extremely difficult to determine using a core 5.8 cm in diameter. Nevertheless, relationships between olivine gabbros and troctolites are reminiscent of those in troctolitic cumulates at Rum, in which repetitions of plagioclase-peridotite and troctolite cumulates were produced by sill-like injection of olivine-laden magmas into a crystal mush on a magma chamber floor (Bédard et al., 1988). However, in those rocks modal and phase layering is far more extensively developed. In Hole 735B, in the deeper of the two series so far described, the lower, more diverse half may represent a thickness of floor cumulates repeatedly disturbed by injections of dense, crystal-bearing magma. When this activity stopped, the upper half of the mass was left to crystallize slowly, on its own, nearly to completion. However, some of the fine-grained troctolites at the base of this pluton have sharp, straight contacts and apparently intruded along fractures in essentially solid rock. These are cumulates, not coarse-grained diabases, thus some avenue for the escape of intercumulus melt must have existed. The surrounding rock cannot have been so solid that it was impermeable. It contained at least a small fraction of melt. These fine-grained troctolite intrusives are thus part of the overall cycle of magmatism that produced this one pluton. They are variants of what Bowen (1920) first described as monomineralic intrusives with dikelike margins in plutonic masses, such as dunite dikes, among his examples of differentiation by deformation. Specifically, these are bimineralic (olivine-plagioclase) adcumulate intrusives with sharp margins. They point to nearly complete compaction and removal of intercumulus melt from the rocks that now surround them, without deformation, but before temperatures dropped below those of common basaltic liquids. That is when intrusion occurred. Compaction and melt expulsion then continued in the intrusion itself.
The olivine gabbros and troctolites at the bases of the two plutons are the most primitive, or least-differentiated, rocks recovered from Hole 735B. Nothing so primitive exists in the core below this. The rocks contain the most anorthitic plagioclases and the most forsteritic olivines in the entire core (Ozawa et al., 1991; Dick et al., Chap. 10, this volume). Some even contain chromian spinel. Nevertheless, none of them contain minerals as primitive and refractory as some of the phenocrysts in abyssal tholeiites from the Indian Ocean (Natland and Dick, 2001). However, such phenocrysts are always present in the same basalts with others that are both more abundant and far more differentiated, a consequence of mixing between primitive and more differentiated magmas. The more differentiated group of phenocrysts matches compositions of minerals in the primitive olivine gabbros and troctolites of Hole 735B.
Did phenocrysts of typical porphyritic abyssal tholeiites thus collect to produce the troctolites of Hole 735B? Perhaps yes, but if they did then the refractory population of minerals has been obliterated by the extensive recrystallization and reequilibration that affected all of the rocks (Natland, Chap. 11, this volume; Dick et al., Chap. 10, this volume). More generally, however, there is no cumulate in Hole 735B, nor indeed has one yet been found at Atlantis Bank, that would correspond to a mineral assemblage crystallized from truly primitive basalt fresh out of the mantle (Dick et al., 2000; Coogan et al., 2000; Natland and Dick, 2001). Possible explanations for this are given in "Construction of the Lower Ocean Crust."
No part of the core recovered in Hole 735B has attracted as much attention as the 51 m of oxide gabbro that separates the two series of olivine gabbro so far discussed. The proportions of the two oxide minerals, ilmenite and magnetite, can be very high. Sawn surfaces and polished thin sections reveal complex patterns of distribution of the oxide minerals, which in some cases form up to 30% of the rock. Against this background, brass-colored globular sulfides are large enough and so abundant that they are readily seen by the naked eye, usually being embedded in the oxide minerals. Some of the most prominent granitic patches and veins are present in this part of the core. All investigators agree in considering that these must be the products of extended high-iron differentiation of parental abyssal tholeiite and that somehow the granitic veins are closely related to oxide gabbros, yet the precise mechanism of this final stage of differentiation has remained elusive. Ilmenite and magnetite are so abundant that they must be cumulus minerals, but another likely late-stage mineral, apatite, is only sporadically abundant and in many oxide gabbros it amounts to only very rare and tiny crystals in a given specimen (Meurer and Natland, 2001), resulting in little or no measurable P2O5 in many of these rocks (e.g., Natland et al., 1991).
Lithologic Unit IV consists of 31 lithologic intervals and is far from uniform in composition (Fig. AF4). All but three contacts between intervals are igneous, although many of them are at least partly deformed or foliated (Dick et al., 1991a). Most are quite thin and occur closely spaced in clusters; several are a few meters thick, and the thickest interval (270) is 14.2 m thick. Three narrow intervals of olivine gabbro are present at ~230 mbsf, and there is another at ~272 mbsf, near the base of the oxide gabbros. These are the screens of the two much thicker series of olivine gabbro already discussed. The oxide gabbros here constitute many small intrusives that evidently repeatedly injected preexisting rock at the same place, between two larger bodies of primitive gabbro.
Within the 51-m thickness of oxide gabbro are three compositional subdivisions, each having higher Mg# toward their bases, these being, respectively, intervals 270, 283, and 288 with depth (Fig. AF4A). The middle of these subdivisions is highlighted by the presence of the thickest trondhemite vein recovered during Leg 118, the only one large enough to be analyzed on board ship. Similar silicic material is present just below the base of the oxide gabbros in a coarse breccia of olivine gabbro at the upper part of Unit V. Bulk compositions of these rocks are hybrids of olivine gabbro and trondhjemite and thus are dioritic (Hertogen et al., Chap. 6, this volume).
The higher Mg#s at the bases of the three subdivisions do not necessarily correlate with the extent of differentiation of these rocks, as they do in the olivine gabbros. In the upper and lower subdivisions, lower Mg# is in part also a consequence of the high concentrations of oxide minerals in the analyzed rocks (Fig. AF4A). The middle subdivision, however, the one with the trondhjemite, has lower TiO2 contents toward its base (Fig. AF4B), and it also is higher in SiO2 (Fig. AF4C). Here, some of the rocks are virtually net veined with silicic material, and it is this that must account for these trends. The trondhjemite vein is only the most concentrated granitic segregation in this part of the core. Three oxide gabbros in this vicinity also have unusually high P2O5 contents. In a thin section of the sample with highest P2O5 contents (1.43%), euhedral apatite crystals up to 2 mm in diameter can be seen embedded in ilmenite.
The tripartite subdivision of the oxide gabbros is also evident in the compositions of olivine and plagioclase, although there is no clear trend toward more differentiated compositions in the uppermost rocks (Fig. AF5). Instead, oxide minerals simply seem to be more concentrated toward its base. Both olivine and plagioclase in the two underlying subdivisions, however, track more evolved yet similar stages of differentiation, with the most sodic plagioclase and fayalitic olivine present at the bases of the subdivisions. The difference between the two subdivisions thus rests on the upper being silicic in overall aspect and not especially rich in oxide minerals, whereas the lower is equally differentiated, but not silicic, and instead is rich in oxide minerals. Two rocks also contain unusually abundant apatite.
The same three subdivisions can be discerned in the trace elements Y and Zr, both of which increase in concentration downward in each subdivision (Fig. AF6). They vary overall by an order of magnitude just among the oxide gabbros. The highest concentrations of both are matched by those measured in the veined olivine gabbro breccias at the top of lithologic Unit V. Because of the presence of globular, magmatic sulfides, oxide gabbros also have the highest S and Cu contents of all the rocks (Fig. AF7). These respectively mirror SiO2 and mimic TiO2, increasing downward in the uppermost subdivision, decreasing downward in the more siliceous rocks below that, and increasing again in the basal oxide-rich gabbros. Silicic rocks in the veined olivine gabbro and elsewhere have both low S and low Cu. The reduced S and Cu of the intermediate subdivision of oxide gabbros again indicate a generally granitic aspect to the bulk analyses of these rocks, although this might also be the result of fine-scale veining or even lit-par-lit injection of silicic material into a more primitive host.
In summary, we note three particular aspects of these rocks. The first is where they are in the core, namely between two much larger bodies of primitive gabbro, and particularly just at the top of the single thickest sequence of unalloyed olivine gabbro in the entire recovery. The second is that whereas among the larger masses of olivine gabbro the more differentiated rocks are present at the tops, among these oxide gabbros, on the other hand, the most differentiated rocks are present at the bottom and individually toward the bottom of each of the three subdivisions we have defined. The third aspect is the seemingly divergent contrast between the second, generally silicic, and third, generally oxide-rich, subdivisions that nevertheless share a common range in stage or degree of differentiation, as indicated by the compositions of olivine and plagioclase.
Olivine gabbro sequences comprise 82.3% of the expanded section recovered below 520 mbsf during Leg 176, and they make up 92.2% of the expanded section below 1000 mbsf. They provide a very different impression than the two series of olivine gabbro and troctolite so far described. Whereas those rocks are diverse, these are uniform. Whereas those rocks have definite chemical trends with sharply defined boundaries, these show only subtle shifts in composition. Whereas some of those rocks include Cr spinel, that mineral never reappeared during Leg 176.
Clearly, this is a different domain of the lower ocean crust and it is quite substantial. The thickness and uniformity of the rocks indicate something fundamental about the organization of the lower crust. Although the impression of a continuing sequence of plutons has disappeared with all of the new chemical analyses, the data nevertheless suggest that this series can be divided into three zones, numbered 3a-3c in the remainder of this discussion, with the middle one, Zone 3b (961-1194 mbsf), being slightly more differentiated and uniform than the others. Consequently, we describe the zones in sequence downward using the same procedures and types of diagrams as before.
Zone 3a spans 520-962 mbsf, lithologic intervals 508-716, and olivine gabbro sequences 40-67 (Fig. AF8A). Leaving out differentiated rock, the sequences average 11.3 m thick, the thickest is 40.5 m, and their aggregate thickness is 283.1 m. Poorly recovered, altered, fractured, and brecciated rock marks the upper 50 m and corresponds to the upper of two seismic reflectors detected by vertical seismic profiling (VSP) in an experiment at the end of Leg 118 (Swift et al., 1991). The base of Zone 3a is within a shear zone centered at ~960 mbsf.
Although olivine gabbro is the predominant rock, there are numerous prominent seams of oxide gabbro and some substantial thicknesses of gabbronorite, especially between 520 and 700 mbsf. Crystal-plastic deformation is particularly intense in rock down to 700 mbsf and is especially associated with the seams of oxide gabbro (Natland, Chap. 11, this volume). This is a continuation of a pattern that begins at 400 mbsf and thus that overlaps a portion of the lower of the two plutons cored during Leg 118. This is high-temperature crystal-plastic deformation that commenced while the rocks were still partially molten. It preceded the brecciation, fracturing, and attendant hydrothermal alteration of the rock at the level of the upper reflector detected by the VSP experiment.
All of this no doubt has confused the original stratigraphy of the underlying olivine gabbro. Both removal and repetition of rock sequences probably occurred, and penetrative infiltration of some of the rock by highly differentiated melts may have infused some rocks with magmatic oxides and otherwise modified certain chemical signatures. However, we deal with what exists.
Both Mg# and Ni concentrations have fairly restricted ranges over the entire zone (Fig. AF8A, AF8B), and there are no sustained trends in either comparable to those higher in the core. Between 520 and 720 mbsf, compositions of olivine gabbros merge with those of gabbronorite and olivine gabbronorite; below this they are fairly distinct. The pointed bar at the top of Figure AF8A shows the range in compositions of the more differentiated gabbros originally thought to characterize Zone 3a. Lithologic Unit VII, for example, is described as consisting primarily of gabbronorite and olivine gabbronorite and Unit IX of gabbro and gabbronorite. New analyses, however, show that both of these units include olivine gabbro that is both as magnesian and rich in Ni as the predominant olivine gabbros lower in Unit X.
What emerges from chemical trends now is the existence of several short sequences of olivine gabbro with decreasing Mg# and Ni upward (green arrows in Fig. AF8). Some of these are present within individual olivine gabbro sequences and others span several sequences. However, Mg# and Ni concentrations are rarely as high as in the lower sequences of either Series 1 or Series 2. This is as far as we can go in noting any similarity to trends in the two plutons in the upper third of the section, and these are present in much thinner sequences of rock. At best, these trends represent rather subtle fluctuations in the proportion of olivine, from a few percent in one rock to ~10%-15% in another. In sequences 59-63, the variations may be related to grain size and modal layering, which is better developed in sequence 60 over 9 m of rock than in any other part of the core. These fluctuations in Mg# and Ni, however, occur throughout Zone 3a, and there is no general tendency for the rocks at the top to be more differentiated than the ones at the bottom. Thus, whereas these olivine gabbros are an aggregate of cumulates crystallized from numerous injections of magma, those injections appear to have been extremely localized in scale and influence. There was no pulsing into some larger body of magma, no interpolation of material into a thick crystal mat, and no pooling of residual melt into an eventual thick, stagnating sill. The igneous layering was likened in the shipboard report to that seen in layered intrusions. However, there is no indication that any individual molten mass ever exceeded more than a few meters in thickness. The layering therefore may have been produced hydraulically, by flow differentiation, as magma perhaps passed forcibly through a tortuous porosity structure. Compared with the olivine gabbros in the upper 520 m of the core, the scale of magma injection was even more diminutive, and the overall tenor of differentiation more advanced.
As in rocks above, however, there are a number of oxide gabbros with fairly high Ni concentrations, matching those of olivine gabbro. Again, these appear to combine both olivine gabbro and narrow oxide-rich seams in samples taken for chemical analysis. Similarly, rocks with high SiO2 have the compositions of diorite because they combine olivine gabbro and narrow granitic veinlets. These samples should sound a note of caution in the interpretation of trends. How many other samples, which otherwise fall into the chemical categories for olivine gabbro or troctolite, defined earlier, have some trace of either oxide gabbro or trondhjemite in their makeup? In what way do such traces blur or obscure the chemical trends we are trying to perceive in scatter diagrams?
Zone 3b spans 962-1194 mbsf, olivine gabbro sequences 68-79, and intervals 718-833 (Fig. AF9). The average sequence thickness is 18.0 m, and the aggregate is 216.1 m, or 93.1% of the expanded recovery. This leaves only 6.9% of Zone 3b to the differentiated gabbros, including oxide-rich seams. This is so small a proportion that the rocks are included in lithologic Unit XI, olivine gabbro. Deformation of the rocks is much reduced, although again there is a general association between zones of deformation and the location of oxide seams. A fault zone is centered at ~1100 mbsf and consists of a series of narrow interfaces showing both normal and reversed senses of shear.
Miller and Cervantes (Chap. 7, this volume) observe differences in compositions of sulfides and corresponding contrasts in Ni contents of gabbros across a boundary that they place at 1170 mbsf. This is very close to the boundary between Zones 3b and 3c. Based on data in Niu et al. (Chap. 8, this volume), the boundary in Ni contents is more precisely between 1173.02 and 1175.65 mbsf. This is within olivine gabbro Sequence 79 and, indeed, is within a single interval (831). The boundary in terms of Ni is ~2.3 m from the top of the interval, which is 20.15 m thick. One more interval (832) lies below this in sequence 70; it is 1.42 m thick. For the sake of consistency, we place the boundary between Zones 3b and 3c at the base of interval 833, sequence 79, but note that it might justifiably be 19.2 m higher, at the dashed line between sequences 79a and 79b in Figure AF9.
Throughout the rest of Zone 3b, as nowhere else in the core, the term "monotonous" is appropriate. Both Mg# and Ni concentrations (Fig. AF9) have quite narrow ranges, and both are at the lower ends of ranges in the rocks of Zones 3a and 3c above and below. In fact, Ni in most samples is <100 ppm. Sequences 68-73 (above the dotted line in Fig. AF9A) have a somewhat higher percentage of slightly more differentiated gabbros than is present in sequences 74-79, but this is barely a contrast. At the base of sequence 79, Ni creeps higher. A few oxide gabbros have Ni concentrations comparable to those in their surroundings, thus appear once again to be samples that mixed olivine gabbro and portions of oxide-rich seams. Other differentiated rocks are also clearly physical mixes of olivine gabbro and granitic veinlets.
Monotonous or not, these rocks are like their chemical counterparts elsewhere in the core in that they consist of repetitions of olivine gabbro sequences comprising alternations in grain size that adjoin mainly on sutured igneous contacts. Within sequences 68-76, there are 20 internal contacts, all but 6 of them being gradational or sutured and thus igneous in character. This was not one large mass of magma that crystallized to a homogeneous composition. It is the crystalline residue of many injections of magma that were very similar in composition and presumably also were similar in phenocryst load and composition from the start.
With olivine gabbro sequence 80 at the top of Zone 3c, bulk compositions of the core shift back to higher Mg# and a greater range of Ni concentration and remain that way to the bottom of the core (Fig. AF10). The distinction between lithologic Units XI and XII is that the latter includes troctolitic gabbros and troctolites. The average sequence thickness is 17.0 m, and the aggregate thickness of sequences is 306.8 m, or 92.4% of the expanded recovery for these sequences. From 1400 to 1508 mbsf, many of the troctolitic gabbros and troctolites have the aspect of narrow, nearly vertical pipes within a matrix of olivine gabbro. These contribute to the large number of gradational or sutured igneous contacts in these sequences (Fig. AF10A). The criterion for defining olivine-gabbro sequences, namely grain-size variation across a nearly horizontal contact, was not easy to apply in these rocks. However, whereas flat or slightly dipping contacts between similar rocks higher in the core suggest either layers or lateral sill-like bodies of olivine-rich rock, here some kind of upward (or downward) protrusions of crystal-laden magma into or through crystal mushes evidently occurred. This resulted in the rather scattered pattern of Ni variability below 1400 mbsf.
Shipboard scientists aboard engineering Leg 179 noted similarities between fairly thick oxide gabbros cored in the lowest 50 m of Hole 1105A and those of Hole 735B (Shipboard Scientific Party, 1999b). Postcruise studies suggested to Casey et al. (1999) and Banerji et al. (2000) that the two may in fact correlate over the 1.2-km distance that separates the holes. Although we agree that the two sections may share common intrusive intervals of oxide gabbro, shipboard data (Shipboard Scientific Party, 1999b) show that the oxide gabbros near the base of Hole 1105A are not in general as rich in TiO2, as correspondingly low in SiO2, or as high in Zr or Y as most of the rocks of Unit IV from Hole 735B. At best, only the uppermost of the three sequences of lithologic Unit IV was reached in Hole 1105A, and its rocks are spread along 54 m rather than 20 m of core, being separated by several substantial intervening thicknesses of more primitive gabbro. That gabbro is bimodal in Mg#, like that from 200 to 220 mbsf in Hole 735B (Figs. F3, F5). Neither the lower two subdivisions of lithologic Unit IV, oxide gabbro, of Hole 735B nor the massive olivine gabbro of lithologic Unit V were reached in Hole 1105A.
Coogan et al. (2000) reported compositions for two oxide gabbros among 26 gabbros analyzed from short drill cores obtained from a generally north-south transect along the summit of Atlantis Bank (Fig. AF11A). The two are adjacent samples on the seafloor, respectively, 2.8 and 3.5 km east-northeast of Hole 735B. Rock magnetic data suggest that the section in Hole 735B has been tilted southward by an average of 19° ± 5°, thus that the oxide gabbro should crop out on the platform summit somewhere to the north. Oxide gabbro Unit IV below 232 mbsf in Hole 735B, its possible equivalent below 102 mbsf in Hole 1105A, and the two outcropping oxide gabbros to the north may therefore be one and the same. Assuming a simple southward rotation and that the upper surface of the oxide gabbro is planar, the angles between the outcropping samples and 232 mbsf in Hole 735B and 102 mbsf in Hole 1105A are shallower than 19° ± 5° (Fig. AF11B), although the projected angle between the two ODP drill holes (17°) is statistically indistinguishable from this, and that between 232 mbsf in Hole 735B and the outcropping oxide gabbros (12.4°) is barely different. In detail, however, the tops of the cored oxide gabbro units and the outcropping oxide gabbros are not coplanar on a southward-dipping surface, although they could lie on a plane striking to the northwest and dipping slightly to the southeast. Or they could be on a curving surface that falls off somewhere toward the south and west. Since there could be an error of 50 m in the relative placement of the two outcropping samples in the equivalent drilled sections, the details of this are assuredly speculative. The data simply indicate the existence of a generally southward-dipping mass of oxide gabbro that originally was approximately horizontal.
Taking 12.4° as the actual amount of southward rotation of an originally horizontal igneous structure beneath the rift valley floor, the locations of surface gabbro samples reported by Coogan et al. (in press) can be projected onto a vertical extension of Hole 735B as shown in Figure AF11B. On this basis, 174 m of section that lies to the south was not drilled but is represented by three rock cores. The most northerly of the gabbro cores projects to a depth of 637 mbsf in Hole 735B. The total section is 1682 m, of which the upper 48.2% is exposed on the nearly flat surface of Atlantis Bank. Hole 735B sampled 89.7% of the section. To the south, extrusive basalt either buries gabbro higher in the section or is in fault contact with it (Fig. AF11B).
A basalt dike obtained by surface coring almost due west of Hole 735B may be the same as one cored in the hole at 105 mbsf. If the two are the same, then the dip of the west-striking dike is S63.6° (Fig. AF11B).