In reviewing our results, it is first instructive to compare the chemical discontinuities we have identified to the divisions of the igneous stratigraphy made by the Leg 118 and 176 Scientific Parties and Dick et al. (1991a) (Fig. F16). Each of the chemical discontinuities, in fact, lies close to one of their boundaries, with only small differences as to where they are drawn. This is encouraging, as it suggests that neither set is arbitrary. The mineral chemistry and lithostratigraphy do suggest that a unit boundary should have been drawn between the Unit Ia massive gabbronorite and the Unit 1b olivine gabbro and gabbronorite at the subunit boundary, as this contact is essentially the same in character as the Unit III/IV boundary (though with opposite polarity). This was not originally done, however, largely because this is also a zone of intense crystal-plastic deformation and hydrothermal alteration, which made identification and separation of the two original subunits tenuous.
A small but significant difference exists between the position of our chemical discontinuity at 941 mbsf and the Unit IX/X boundary drawn by the Leg 176 petrologists. They placed their contact to coincide with a high-temperature shear zone at 960-990 mbsf. As discussed earlier, our data suggest that the latter is a structural, not an igneous, boundary and that the base of Unit X should have been drawn near 941 mbsf. It is one of the curious features of the core that whereas there are numerous high-temperature ductile shear zones and several brittle facies faults, none on close inspection appears to coincide precisely with the major geochemical boundaries in the hole.
Perhaps more interesting is to look at lithostratigraphic boundaries where we did not identify a mineralogic one. In this light, we should note that the lithostratigraphy was intended as a nongenetic descriptive breakdown of a complex core into composite sections with like features. Thus, for example, the Subunit VIb/VIc boundary marks the appearance of small meter-scale troctolite and troctolitic microgabbro intrusions that cross-cut olivine gabbros otherwise continuous with those in Unit VIa, whereas the Unit VI/VII boundary marks the abrupt disappearance of these same troctolite and troctolitic microgabbro intrusions. Likewise the Subunit VIc/VId boundary marks the end of a series of ferrogabbros that cross-intrude the olivine gabbros in Subunits VIa-VIc with a shift to predominantly undeformed olivine gabbro and troctolite.
Noting the actual genetic significance of these different boundaries, then, it is worth considering those between Units VII through X. No breaks in the chemistry based on either mineralogy or bulk rock chemistry have been previously drawn here. Yet on close inspection, for example, one could well draw a chemical discontinuity based simply on an offset in anorthite between Units IX and X—suggesting that the lithostratigraphers may have correctly identified complexities in the core in Units VII-IX that are not evident in the mineralogy or chemistry because of undersampling. It is probable, then, that at least the first of the three lower olivine gabbro sequences identified on the basis of bulk rock chemistry and mineralogy could also be further subdivided.
Our preliminary examination of the database compiled here leads us to confirm many of the prior conclusions of the Leg 118 and Leg 176 Scientific Parties. In particular, the silicate mineral chemistry of the gabbros is consistent with two major cycles of intrusion. The earliest cycle consists of the upper two olivine gabbro sequences above 528 mbsf (Subunit Ib through Unit III and into IV and Units V-VI). The gabbros intruded during the early cycle are significantly more mafic than those intruded in the later cycle, which consists of the three lower olivine gabbro sequences (Units VII-X and XI-XII). Numerous small primitive troctolites cross-intrude the olivine gabbro above 528 mbsf but abruptly disappear below that discontinuity. This leads to the conclusion that the lower olivine gabbros postdate the upper olivine gabbros, crosscutting both the upper olivine gabbros and their cross-intruding troctolites. We suspect, however, that the discontinuity at 274 mbsf between the two upper olivine gabbros is due to the effects of chemical exchange during later intrusion of ferrobasalt melt and crystallization of the oxide and disseminated oxide gabbro sequences in Units III and IV, noting that other interpretations are possible (Natland and Dick, Synthesis Chap., this volume). The two discontinuities separating the lower three olivine gabbro sequences are of problematic origin. Both coincide with the presence of oxide gabbro and gabbronorite, and the lower of the two represents an area of significant late microgabbro intrusion. We also suspect that whereas the upper discontinuity may represent a boundary between two phases of intrusion, the lower one owes its origin to the influence of the late microgabbro intrusions—a departure from the interpretation of the Leg 176 scientists. The latter likely occurred while the host olivine gabbro was still partially molten—allowing for extensive chemical exchange between the wall rock and the crystallizing microgabbro. These issues certainly require additional study.
Further inspection of the combined mineral and lithostratigraphies suggests that the shear zone identified in Units III and IV, which apparently controlled emplacement of late iron-titanium-rich melts at the top of the section (Dick et al., 1991; Natland and Dick, 2001), had several imbrications. The Unit V/VI boundary separates the massive olivine gabbros of Unit V from similar gabbros containing numerous ferrogabbro bodies. These ferrogabbros in Subunits VIa-VIc are intercalated with olivine gabbros, as in Units III and IV, and also define a similar downward iron enrichment trend. It would appear that this trend likely has a similar origin. Subunit 1a consists largely of massive gabbronorite that defines the top of yet another downward iron enrichment trend extending into Unit II, where numerous small ferrogabbro bodies cross-intrude the olivine gabbros. Thus, based on the mineralogy and lithostratigraphy, there appears to be at least three cycles of intrusion of ferrogabbros from 0 to 500 mbsf representing probable fault-controlled intrusion of late iron-titanium-rich melts. We suggest that these likely represent imbrications of the fault on which the whole Atlantis Bank plutonic massif was unroofed, which then must have extended into or originated in a crystal mush zone beneath the sheeted dikes.
The downhole mineral variations, then, are consistent with upward enrichment of the top of the gabbro section by upward density-driven compaction of melt and intrusion of iron-titanium-rich late magmatic liquids along shear zones and fractures by deformation-controlled melt migration. The particular source of these late melts, of course, could lie out of the section. We note, however, the intriguing association of oxide gabbros and crystal-plastic deformation found in the cores—with the interesting aspect that major zones of deformation correlate with oxide abundance in the upper portion of the hole but do not correlate in the lower portion of the hole (Dick, Natland, Miller, et al., 1999). In addition to this association, there is also a correlation between magmatic foliations and crystal-plastic deformation throughout the hole, with the resulting fabrics produced by both also tending to have similar alignments. The origins of magmatic foliations are controversial but include crystal settling, preferential growth in the direction of melt flow in a crystal mush, and compaction. We view it as unlikely that the foliations in Hole 735B, which also often contain a lineation and generally lie in the plane of the crystal-plastic foliation (Dick, Natland, Miller, et al., 1999), originate from crystal settling or simple compaction. We suspect that the most likely explanation is that they represent preferred crystal growth of plagioclase in the direction of late melt flow through a crystallizing compacting olivine gabbro mush. Therefore, we speculate that these associations between oxide abundance, magmatic foliation, and crystal-plastic deformation reflect the deformation-controlled flow of late iron-titanium-rich melts compacted out of the olivine gabbros at depth and intruded to higher levels in the section where they crystallize to form oxide gabbros.
With this compilation of mineral data we can carefully examine the downhole cryptic chemical mineral variations at the outcrop scale in the olivine gabbros that presumably reflect the earliest history of intrusion and crystallization. We would like to do this in hopes that it will lead us to some insight as to what the initial scale of intrusion was and the frequency of replenishment during a cycle of intrusion. For this purpose, we screened out all other rock types and plotted the downhole variations of anorthite and forsterite in plagioclase and olivine (Fig. F17A). Although this eliminates stratigraphic complexities created by late-stage ferrogabbro intrusion, it also ignores possible contact effects and contamination of the olivine gabbros by these late melts. This screen also relies on rock names assigned by many different analysts, and thus, unavoidably, some of the variations in Figure F17A may arise from the inclusion of the odd crosscutting ferrogabbro. For most large-scale studies of cryptic variation in large intrusions, every analytical point represents an inflection point in the profile due to limitations in the number of samples that can be taken and analyzed. Even with the very large number of analyses represented by this study, we just get to where significant changes in mineral chemistry are represented by gradients in mineral composition at scales of <100 m, noting that the total number of analyzed samples (~500) is actually far fewer than the 952 discrete lithologic intervals identified in the core.
As can be seen from Figure F17A, there is a considerable amount of fine structure in the Hole 735B olivine gabbro stratigraphy at 1- to 100-m scales. The origins of this cryptic variation, defined by sometimes subtle and sometimes large swings in olivine and plagioclase composition, are unknown at present and cannot be determined on the basis of mineral chemistry alone. Instead, this requires both smaller-scale sampling of selected intervals combined with a detailed comparison of the mineral data to the cores themselves, where the compositional variations can be related to local igneous contacts, cross-cutting microgabbro and oxide gabbro intrusions, late melt flow channels, and high-temperature shear zones. Moreover, as noted above, the number of analyses is actually sufficient only to tell us that systematic cryptic variations exist but are not sufficient to really show the patterns of these variations. What we find, however, is consistent with the oft-stated hypothesis that the larger-scale gabbro units discussed to date are themselves composed of numerous related penecontemporaneous smaller intrusions derived from the same or related parental melts. Noting the rotations of the section likely during emplacement, these intrusions could be small sills, as has been suggested for the Oman section, or small inclined dikelike bodies—a lower crustal analog of a dike swarm (a microintrusion swarm).
Overall, there is a fair first-order correlation between olivine and plagioclase chemistry evident in Figure F17A at wavelengths >200 m. This is impressive, as the diffusion timescales for olivine and plagioclase reequilibration are very different, with plagioclase having very long equilibration times at magmatic temperatures. It is clear then that at this scale melt transport by permeable flow through the section has not been sufficient to disrupt the large-scale chemical stratigraphy. However, when we examine the chemistry on the finer scales shown in Figure F17A and discussed above, we find that the correlations often become quite poor. Local peaks and lows in anorthite and forsterite content are often offset from one another and frequently show opposite polarity. This is illustrated by the very large scatter found in a plot of the difference in coexisting Fo and An compositions against anorthite in Figure F17B. If anorthite and forsterite were well correlated, one would expect either that most of the data would plot near a single point or define a correlation with variable anorthite content.
In the absence of significant late-stage permeable melt flow across lithostratigraphic boundaries, original cryptic variations, reflecting progressive cotectic crystallization of a magma, would be preserved and the compositions of coexisting phases would be well correlated. The poor correlation of coexisting mineral compositions for Hole 735B olivine gabbros seen in Figure F17, then, would be expected as the result of reequilibration of the matrix of a crystal mush with melts being transported through it by permeable flow. This would result due to the very different solid-state diffusion coefficients for olivine, pyroxene, and plagioclase and differences in reaction rates for different elements within individual phases (e.g., Korenaga and Kelemen, 1997a). In such a situation, even as cryptic peaks in mineral composition profiles are subdued by reequilibration with a migrating melt, the differences in reaction rate will cause the peaks in mineral composition to shift relative to one another. Korenaga and Kelemen (1997a) recently showed in a detailed study of a 600-m layered section of the Oman Ophiolite excellent covariations of the silicate minerals that precluded large-scale melt migration through the section. In particular, they found that the nickel content of olivine correlated well with the forsterite content, with a correlation coefficient of R2 = 0.89. By contrast, for Hole 735B gabbros there is no correlation at all between olivine nickel and forsterite content other than an increasing upper bound for Ni with increasing Mg# (Fig. F7). Similarly, they found a correlation coefficient for anorthite and forsterite in their gabbros of R2 = 0.59, whereas the same correlation coefficient for the Hole 735B olivine gabbros is only R2 = 0.42. Strong evidence for extensive late melt-rock reaction and permeable flow can also be seen in the pyroxene mineralogy, where chrome, alumina, sodium, and most particularly, titanium all show huge variations in concentration at a given Mg#.
In this light, it is interesting to compare the relative offsets of the mineral compositional trends defined by the olivine gabbro sequences at the chemical discontinuities in the upper and lower olivine gabbos. At the 528-mbsf discontinuity, the offset in forsterite content is ~18 mol%, anorthite content ~11 mol%, and clinopyroxene Mg# ~8 mol%. Comparing this to the mineral covariation plots (Figs. F9, F10), we see that this is consistent with differences predicted for cotectic crystallization of a single magma. Whereas clinopyroxene Mg# ranges from ~90 to 55 (35 mol%) along the cotectic trend, olivine Fo ranges from 84 to 30 (54 mol%) and plagioclase An ranges from 75 to 30 (45 mol%). Thus, for the relative magnitude of the offsets of the mineral trends, Fo > An > Cpx, Mg#s are about right for the intrusion of less primitive magmas, represented by the lower olivine gabbros into olivine gabbros derived from more primitive parental magmas (upper olivine gabbros). By contrast, the relative offset at the 941-mbsf discontinuity is ~7% Cpx Mg#, ~8% An, and only ~7% Fo, and at the 1321-mbsf discontinuity, ~3% Cpx Mg#, ~4% An, and only ~2% Fo! This is not consistent with simple intrusive contacts because the offset of forsterite content is smaller than those of clinopyroxene and plagioclase. Interestingly, clinopyroxene and plagioclase do preserve appropriate offsets relative to each other at the 941-mbsf discontinuity. This, combined with the coherent upward decrease in chrome in clinopyroxene from the bottom of the hole to 941 mbsf followed by a jump in chrome content (Fig. F5D), suggests that this may have originally been a major contact between two different cycles of intrusion.
We find an explanation for this enigma by suggesting that the lower two contacts were modified by exchange with late-magmatic liquids migrating upward through the section. Because of the relatively rapid rate of diffusive exchange and reequilibration of olivine relative to plagioclase and pyroxene with a migrating liquid during permeable flow, contrasts in phase composition across layer boundaries and igneous contacts will be eliminated much faster for olivine than for plagioclase or pyroxene (Korenaga and Kelemen, 1997a). What is significant here is that whereas the 941- and 1321-mbsf discontinuities separating the three lower olivine gabbro sequences have been affected, the 528-mbsf discontinuity separating the upper and lower olivine gabbros has not. This is consistent with a significant phase of late melt flow through the three lower olivine gabbros, which did not affect the two upper olivine gabbro units. Thus, the upper gabbros were likely largely solidified and impermeable at the time of the latter intrusive cycle(s). Hence, we view the lower three and the upper two as representing two different major cycles of intrusion and suggest that the appropriate length scale for large-scale permeable melt flow in the lower crust at a slow-spreading ridge is that of the individual intrusion or cycle of intrusion. Thus, the evidence for permeable flow we find is not for transport of melts from the mantle over long distances or through the entire crustal section, but for local transport of late iron-rich melts through and out of an individual intrusion.
Whereas studies of the Oman Ophiolite and at Hole 735B suggest that the lower crust accretes by the intrusion of innumerable small bodies of gabbro at both fast- and slow-spreading ridges, our data suggest fairly large-scale permeable melt flow occurred locally in the Hole 735B gabbros in sharp contrast to the Oman situation documented by Korenaga and Kelemen (1997a). Given the presence of a long-lived melt lens at fast-spreading ridges (and by inference at the paleo-Oman spreading center) and its absence at slow spreading ridges, one might expect the opposite conclusion. The presence of fairly large-scale permeable melt flow in the lower crust at slow-spreading ridges, however, is likely due to the influence of coarser grain size and deformation on transport of late interstitial melt at the scale of individual intrusions. Its absence in the lower crust at fast spreading ridges, in turn, is likely due to the finer grain size and a relatively static crystallization environment. Whereas deformation may considerably enhance compaction and melt migration, this alone may not be sufficient to explain the differences. In a small sill at a fast-spreading ridge, due to the slow cooling rate, grain growth and nucleation would both be slow relative to the time for crystal settling. This would result in the isolation of mineral grains in a rapidly compacted, and therefore impermeable, fine-grained cumulate. By contrast, with more rapid cooling at a slow-spreading ridge but possibly similar scales of intrusion, crystals would grow to larger size, resulting in cumulates with much longer compaction lengths. This would in turn support larger initial porosity and permit larger-scale permeable flow and late-stage melt migration through the cumulate than would have occurred in the fine- to medium-grained gabbros seen in Oman and drilled in old East Pacific Rise lower crust (e.g., Natland and Dick, 1996).
Comparing the mineral chemistry of the Hole 735B section to that of several well-known ophiolites and the Skaergaard Intrusion reveals several other interesting features. In Figure F18, we show the composition field for Hole 735B olivine together with mineral data for the CY4 drill hole from Malpas et al. (1989) on a plot of the downsection variation of plagioclase and olivine for three different sections of the Oman Ophiolite from Pallister and Hopson (1981). The latter authors noted the striking difference of the Oman section from the variations found in layered intrusions like the Skaergaard. The latter are generally believed to be the product of in situ differentiation of a single magma body. Obviously, the Oman section has to have been constructed in an entirely different manner, with no systematic vertical chemical variation from one section to the next, except for a concentration of extreme differentiates at the top of the section immediately below the sheeted dikes. This is generally believed to reflect multiple intrusions of new magma throughout its crystallization history, in contrast to the Skaergaard Intrusion. Another feature worthy of note is that both the Oman section and the Troodos section are far more forsteritic and anorthitic than Hole 735B. This is consistent with crystallization from more refractory parental liquids, the products of higher degrees of mantle melting, and a very different oceanic environment from that of the very slow spreading Southwest Indian Ridge.
It is now believed that Oman represents a composite body of gabbro composed of numerous gabbro sills, one injected into another (Boudier et al., 1996; Korenaga and Kelemen, 1997b). Similarly, there is no simple upward sequence seen in the Hole 735B gabbros; rather, it too seems to be a composite section composed of several small olivine gabbro intrusions on scales of 200 to 700 m, which themselves are each likely the compound products of a cycle of numerous even smaller penecontemporaneous intrusions. The relatively uniform subvertical chemistry profiles of the upper two and lower three olivine gabbro sequences argue that there was only minor in situ vertical differentiation on initial crystallization. Rather, they must have at first cooled fairly rapidly to form a crystal mush. We would believe that the extensive evidence for late-stage permeable flow we have found demonstrates that following this initial burst of crystallization, subsequent cooling and final solidification of these intrusions took a considerably longer time—perhaps contributing to their unusually coarse grain size. Of all the gabbro sections we have examined in ophiolites (Troodos, Leka, Karmoy, Oman, Josephine, and the Liguride ophiolites), only the Ligurian gabbros exhibit similar coarse grain size, abundant crystal-plastic shear zones, and bimodal distribution of oxide and olivine gabbros. The Ligurian gabbros crystallized in a very different environment than a mid-ocean ridge: continental breakup and intrusion into old upwelling mantle during the formation of the Ligurian Tethys (e.g., Rampone et al., 1998). It would appear, then, that the strongly bimodal chemistry observed in Hole 735B and late-stage melt removal and reintrusion to higher level is characteristic of the slow-spreading and rifting tectonic environments.
Several Leg 118 scientists (Dick et al., 1992; Natland et al., 1991) calculated the Mg# of liquids in equilibrium with the olivine gabbros and oxide gabbros from Hole 735B and compared these to the Mg# of dredged basalt glasses from the Atlantis II Fracture Zone. These authors noted that whereas there was correspondence between the basalts and the liquids in equilibrium to the olivine gabbros, there were no erupted liquids corresponding to the majority of the ferrogabbros. This, and that the bulk composition of Hole 735B is close to a moderately differentiated MORB (Dick et al., 2000) provides compelling evidence that the Hole 735B section represents crystallization and in situ tectonic differentiation of small bodies of moderately evolved melt.
This raises an important question—what is the source of these liquids and where are the missing cumulates needed to mass balance them back to a primary MORB composition? One possibility is that they represent the products of crystallization and fractionation of primary melts intruded from the mantle directly below the hole. Gabbros, particularly primitive gabbros, are anomalously scarce compared to their supposed abundance in the ocean crust at the Atlantis II Fracture Zone and other large Indian Ocean transforms where mantle sections are exposed (Dick, 1989; Bloomer et al., 1989; Coogan et al., 2001). This led to the hypothesis that melt flow out of the mantle is highly focused toward the midpoints of ridge segments, from whence it is subsequently intruded and erupted down-axis to form the ocean crust at slow-spreading ridges (Dick, 1989; Muller et al., 1999; Smith and Cann, 1999; Whitehead et al., 1984). Abundant peridotites are found along the wall of the transform to the west of Hole 735B. At the same time, dunites marking melt-flow channels through the mantle are rare in these peridotite suites. This indicates that relatively little shallow mantle melt transport has occurred near the transform and that the missing cumulates are not likely to be in the mantle section there (Dick, 1989, Dick et al., 1991b). The magma source therefore is away from the transform, and while it could be out of the section to the east of Hole 735B (Dick et al., 2000), this depends on where the midpoint of the paleoridge segment lay. If it was centered near Hole 735B, there could be a considerable thickness of lower crust yet to be drilled there.
This still leaves the question as to why primitive melts close to equilibrium with the mantle do not erupt along most of the Southwest Indian Ridge. Where and how are primary magmas nearly universally buffered to such moderate compositions? The scarcity of gabbroic rocks and dunites with appropriate composition in the mantle sections in ophiolites suggests that this does not happen in the mantle beneath the ridge. This then leaves the lower crust beneath the ridge axis—but not in the Hole 735B section. We suspect then, seismological evidence to the contrary, that this requires some form of long-lived crystal mush zone in the lower crust near the midpoint of slow-spreading ridges.
In this light, we note that the suite of gabbros studied by Meyer et al. (1989) from the midpoint of a ridge segment at 7°16´E on the Southwest Indian Ridge is strikingly different from the Hole 735B suite, containing strongly modally layered gabbros, anorthosite, troctolite, and dunite, with olivine forsterite contents ranging up to Fo89 in the dunite. This suggests that the appropriate bit of the magmatic plumbing system does exist in the crust beneath the midpoints of slow-spreading ocean ridges, whereas the lack of erupted primitive MORBs suggests that they must be fairly universal. If the plumbing system in which this occurs is not long lived, then how does one explain the rather universal lack of erupted primary or even relatively primitive magmas along the length of the Southwest Indian Ridge (e.g., le Roex et al., 1983, 1989, 1985, 1992)?