CONSTRUCTION OF THE LOWER OCEAN CRUST

There is something fascinating about science. One gets such wholesale returns of conjecture out of such trifling investment of fact.
—Mark Twain

Whereas gabbros of Series 1 and 2 may plausibly have been linked to eruptions on the floor of a rift valley, producing eruptives similar in composition and mineralogy to the usual types of abyssal tholeiite found on slowly spreading ridges, most of the gabbro locked in the lower ocean crust of Atlantis Bank experienced a more extended and profound differentiation. This in its later stages, as we have seen, was accompanied and perhaps even propelled by extensive crystal-plastic deformation (differentiation by deformation) (Bowen, 1920) to dramatic effect in the core. However, the more significant result of drilling in Hole 735B is that the entire lower two-thirds of the core and ~40% of the upper one-third crystallized in situ from magmas as least as differentiated as ferrobasalt, mainly before deformation. Comparing inferred compositions of liquids that crystallized to produce the olivine gabbro suite of Hole 735B and dredged gabbros (e.g., Bloomer et al., 1989) with glasses from basalts dredged in the vicinity (Dick et al., 1991a; Natland et al., 1991; Mahoney et al., 1992) and elsewhere (e.g., Natland, 1980; Sinton and Detrick, 1992), the economy of ferrobasalt at slowly spreading ridges thus seems to be that it forms extensively but it stays in the lower crust and only erupts extremely rarely.

Perhaps this should not be such a surprise. Rhodes et al. (1979) showed that many abyssal tholeiites from the Mid-Atlantic Ridge are hybrids, combining aliquots of both primitive and strongly differentiated basalt. One consequence of this, for example, is that Ni contents are higher in basalts of intermediate composition than in those that experience only crystallization differentiation and follow a simple closed-system line of descent. Mixing between primitive and differentiated components is evident in the compositions of phenocrysts as well, especially their zoning patterns (Rhodes and Dungan, 1979). Similar relationships exist in abyssal tholeiites from the Indian Ocean (Natland, 1991). The chemical data from the Mid-Atlantic Ridge indicate that the evolved mixing component must commonly be as differentiated as ferrobasalt (Natland, 1980). Therefore, rather than saying that ferrobasalt does not erupt at slowly spreading ridges, more accurately, some ferrobasalt does reach the seafloor, only it is in combination with more primitive material. Ferrobasalt is, to a degree, occult in the composition of many abyssal tholeiites at slowly spreading ridges.

The abundance of normally zoned phenocrysts at first suggested that many eruptions are immediately preceded by addition of primitive, phenocryst-laden magmas to shallow magma reservoirs containing extensively differentiated liquid. The precise location of mixing was presumed to be the interior of some kind of a magma chamber, within which cumulates had already lined the walls, floor, and roof (e.g., Bryan and Moore, 1977) (Fig. F2A). Indeed, entrainment of cumulates is a potential source for some of the large refractory megacrysts and phenocrysts found in many of the basalts. However, no one asked why such concentrated pods of ferrobasalt never seem simply to be displaced to the seafloor during an inflation-eruption cycle at a slowly spreading ridge as, for example, analogous lavas are along rift zones of Hawaiian volcanoes (e.g., Wright and Fiske, 1971) and the fast-spreading East Pacific Rise (Natland, 1980).

Hole 735B suggests a different sequence for the mixing. It is that it occurs prior to arrival in a magma chamber, while primitive magmas en route to high-level cupolas in pipelike conduit systems (Fig. F2E) pass through more differentiated gabbros still containing some amount of interstitial melt. Small magma chambers, or cupolas, clearly do form, produce normal successions of fairly primitive cumulates, and supply melt to the seafloor. But in Hole 735B, they formed at the top of the gabbroic section. We thus suggest that the upper two olivine gabbro series of Hole 735B represent a section through a portion of a shallow, laterally connected, body of hot, low-velocity rock analogous to that detected seismically along the Mid-Atlantic Ridge by Magde et al. (2000) and therefore that those comprise relatively primitive intrusions, whereas Series 3 represents the material below this and in between conduits (gray in Fig. F2E) and that this material is more differentiated.

Where does so much ferrobasaltic melt originate, and why does it crystallize below more primitive gabbro cupolas? We believe that the ferrobasaltic magma ultimately came from the interiors of small plutons like those cored at the top of the section but that such liquids were expelled, forced or squeezed out of those bodies once their links to mantle sources were broken and their internal overpressure dissipated (Fig. F7). Without this impulse, eruption to the seafloor became almost impossible, yet forces imparted by overburden, crustal dilation, and perhaps lateral penetration of low-angle detachment faults were still acting to drive residual magma from those cupolas into porous and partially molten surroundings. Even the minutest quantities of residual melt were driven out in the course of transformation of crystal mushes to adcumulates.

Flow of ferrobasalt magma was both along and across the ridge axis and probably was mainly directed beneath a permeability barrier at the brittle-ductile transition. Along the axial zone of hot, low-velocity rock, this barrier was quite shallow. Off axis to either side of the rift valley floor and over time with continued spreading, however, the brittle-ductile transition probably descended into the gabbroic section as heat was drawn off and crystallization proceeded. However, tectonic forces related to uplift of the plutonic section from beneath the rift valley floor and its eventual exposure (Natland and Dick, History Chap., this volume) may actually have served to warp the transition upward and even to lower the pressure of accumulated rocks above it by removing them along high-angle detachment faults (Natland and Dick, History Chap., this volume). In this way, buoyant magma could continue to aggregate beneath the brittle-ductile transition, even as its locus shifted deeper into the gabbros. Thus, we envision a process of thickening of the lower crust by injection of differentiated magma beneath inactive shallow cupolas as they are carried away from the center of the rift valley by spreading, while the total crustal thickness lessens with the removal of extrusives and dikes. This likely would sever vertical conduits from their sources in the mantle and probably dismember them in the process.

Students of layered intrusions commonly visualize them to be contained by impermeable country rock—the ultimate natural crucible. At a ridge axis, however, new pulses of magma released from the mantle encounter a range of hot, partially molten, and variable permeable material, namely cumulates that are not yet completely solid, and crystal mushes (Sinton and Detrick, 1992). At a slowly spreading ridge, we picture this process as being intermittent rather than continuous, at least over a span of tens of kilometers along axis. Ascending magma therefore must penetrate a sheath of partly molten cumulates before it can erupt. It will be buoyant even in a crystalline matrix containing as much as 20% pore melt (Natland and Dick, 2001), thus will tend to pass through whatever is already there, including the fraction of melt retained by the hot cumulates, and coalesce at the top of the stack at the brittle-ductile transition. The gauntlet from the mantle to this barrier is consequently where mixing with ferrobasaltic liquid occurs. Ascending magma is not plunged into a ferrobasalt bath; instead, ferrobasalt is added to it bit by bit along a tortuous passage.

Since uplift and exposure of plutonic sections at transverse ridges is an asymmetrical process, with extrusives and dikes being faulted away from the ridge axis in the opposite direction (Natland and Dick, History Chap., this volume), so this process also is likely to be axially asymmetrical. Rupture of cupolas and expulsion of their residual ferrobasaltic magma into surrounding partly molten cumulates will be easier to the one side where crustal overburden is being relieved than to the other. The seismic thickness of the crust carried into rift mountains on either side of a rift valley might wind up being similar, but on one side the gabbroic portion of the crust will be thinner and, on the average, fairly primitive in composition; it will also contain almost all of the dikes and pillow lavas. On the other side, the dikes and pillows are largely missing, but the gabbro is thicker and carries a preponderance of more strongly differentiated compositions.

Figure F7 represents the cupolas, or eventual plutons, as the cross-sections of funnels. This is a shape inferred for some layered intrusions (e.g., Muskox; Irvine, 1980; Great Dyke, Worst, 1958; Wilson, 1996), but we have no evidence that the two cored at Hole 735B were any particular shape at all. Funnels, however, are useful to make a point about the possible distribution of primitive cumulates, those derived from near primary magmas, in the lower ocean crust—what Coogan et al. (2001) term the missing cumulates of Atlantis Bank. Earlier, we mentioned that such cumulates might not be present in a layer between the bottom of Hole 735B and mantle peridotite, wherever that is, but instead could lie out of the section. Bloomer et al. (1989) and Dick et al. (2000), for example, suggest that primitive cumulates are concentrated nearer the centers of spreading segments along slowly spreading ridges; thus, they may not even be present at transverse ridges near transform faults. Figure F7, however, is drawn to suggest that primitive cumulates do not have to be very far away. They could be in narrow columns marking the feeders to every shallow, flatter cupola and thus are surrounded, undercut, and probably crosscut and distorted by subsequently emplaced differentiated gabbro. Perhaps as well, most conduits are separated asymmetrically during spreading to the other side of the rift valley. In such a case, we would rarely, if ever, encounter remnants of these conduits on transverse ridges. This may be why differentiated gabbros similar to those of Hole 735B rather than primitive gabbros are so commonly recovered in the same dredge hauls as peridotite from transverse ridges in the Indian Ocean and elsewhere (Francheteau et al., 1976; Dick, 1989; Bloomer et al., 1989).

Magde et al. (2000) identified two columns in their seismic experiment that they described as active vertical magma conduits (Fig. F2E). The two columns are places where injections of magma from the mantle "travel upward until they intersect the brittle-ductile transition, where they are then diverted along-axis to supply shallow intrusive bodies and seafloor eruptions along much of the ridge segment (Magde et al., 2000, p. 55)."

Although this interpretation has appeal, the geophysical data by themselves do not demonstrate either that the two columns are long lived or that they are responsible for the entire interconnected shallow zone of low velocity. The columns may instead be the only two currently detectable, the roots of older ones having solidified completely by formation of adcumulates and then cooled after perhaps thousands of years of inactivity. The shallow along-axis low-velocity region then is simply a series of coalesced cupolas, their sources now completely choked off, which are not yet completely frozen. Along-axis flow is likely restricted to some few kilometers between these; it need not extend for 20 km from the ones that are currently active. Whatever the case, the seismic images lend credence to the notion that early high-temperature cumulates may be restricted to columns beneath individual volcanoes, and they demonstrate that there is a large volume of lower crust with higher seismic velocities through which these columns extend.