Hole 735B was drilled in part to test developing models of magma chambers at slowly spreading ridges, derived over the past 25 yr by successive integration of geophysical observations, seafloor sampling of basalt, observations on ophiolites, and basalt stratigraphy deduced from drilling. Early on, distinctions between unrifted fast-spreading ridges and deeply rifted slow-spreading ridges provided a duality to general consideration of accretion of the ocean crust. Initially, the concept that large magma chambers span the widths of rift valleys and the thickness of the entire lower ocean crust (e.g., Fig. F2A) (Bryan and Moore, 1977) ran afoul of the failure of seismic refraction experiments to detect any melt bodies beneath the axial rifts of slow-spreading ridges. Thus Nisbet and Fowler (1978) proposed that the crust of the Mid-Atlantic Ridge accretes not by the crystallization of layers of cumulates about a large crystallizing magma chamber, likened to an "infinite onion" (Fig. F2A), but by injection and crystallization along narrow vertical fractures from the base to the top of the crust (an "infinite leek") (Fig. F2B). If "infinite onions" exist anywhere in the ocean crust, they would have to be beneath the East Pacific Rise (Rosendahl, 1976).
In the leek hypothesis, Nisbet and Fowler (1978) proposed that melt collects in pods below a rheological barrier beneath the width of the axial rift; each pod may be the root of one or more fractures that vent lava to the seafloor. The small bodies in effect should crystallize in the form of "nested plutons," each with a separate history of crystallization and differentiation. Nisbet and Fowler (1978) also proposed that ascending magma comes temporarily to rest in a shallow high-level reservoir, where much of the mixing between comparatively primitive magmas, evidenced by the compositions and zonation of phenocryst minerals, takes place.
Sinton and Detrick (1992) again seized on the lack of seismic evidence for molten material in the crust beneath the axes of slow-spreading ridges. With little to go on from interpretation of ophiolites for this type of crust, they borrowed heavily from Marsh's (1989) concept, derived from linked sill complexes in continental crust, to postulate that there are also linked sills in the midst of a nearly consolidated crystal mush in the lower ocean crust beneath rift valleys (Fig. F2C). The concept suggests that magmatic differentiation should march in step from one sill to the next one above, as long as they are interconnected or resupplied with melt. The most differentiated magma would likely concentrate toward the top of the linked plumbing system nearer the cold ocean floor, at least until different parts of the system become physically isolated. Marsh (1989) defined mush, however, as material that behaves like a liquid even though it is laden with crystals. As such, it must contain at least 50% melt, which would be seismically detectable if present in any volume beneath a ridge axis. Therefore, either mush zones beneath slow-spreading ridges are too narrow to detect by seismic refraction techniques or they are ephemeral. Thus in 1992, the linked-sill hypothesis was entirely conjectural and could only be tested by drilling. Citing some prior studies including Leg 118 drilling results, Sinton and Detrick (1992) did indicate that steep faults curving from the rift valley and its walls might root in hot cumulates adjacent to the mush zone.
Drilling in Hole 735B began in 1987, but the 504 m of core recovered during Leg 118 was not really sufficient to provide much insight into these problems. The drilling verified the importance of high-temperature crystal-plastic deformation in guiding transport of strongly differentiated liquids along zones of shear. The bulk of the core, however, consists of primitive (high temperature) olivine gabbros and even some troctolite, arranged in two bodies, each more differentiated upward. Were these the manifestations either of nested plutons or of linked sills? Renewed drilling in 1997 carried the hole to about three times its original penetration and resulted in a new postulate, based on shipboard data, that the olivine gabbros there represent as many as five nested plutons (Fig. F2D), each on the order of 200-400 m thick (Dick et al., 2000; Robinson et al., 2000) and each more differentiated upward. Even though the lower two-thirds of the core still contains hundreds of thin, crosscutting, and oftimes deformed oxide gabbros, the essential underlying stratigraphy of the olivine gabbros is intact.
Now we have some new seismic results to link to this picture. Magde et al. (2000) report a tomographic experiment on a rift valley of the Mid-Atlantic Ridge. They did not actually detect melt bodies or lenses comparable to those found along the East Pacific Rise but rock with somewhat attenuated seismic velocities that they interpret as defining the geometry of dispersed melt flow beneath this ridge axis (Fig. F2E). Essential features of their zone of hot rock are (1) concentration of low-velocity material beneath several small volcanoes on the seafloor, (2) lateral linkage of these low-velocity zones along the rift axis, and (3) the presence of two pipes of such material extending from the mantle to the laterally linked near-surface masses and directly beneath two of the volcanoes. Magde et al. (2000) propose that the other volcanoes are supplied melt by lateral diking in the shallow low-velocity zone along the rift axis from these two sources to distances of ~20 km in either direction. Such dikes should be present in the gabbroic portion of the ocean crust, even if they crystallized to coarse grain size.
Although this is certainly a more detailed picture of the shallow seismic structure of the lower crust at a slow-spreading ridge than we have ever had before, it does not explain how the entire mass of the lower crust is formed. Specifically, where does the large mass of material surrounding the vertical pipes and underlying the laterally linked shallow low-velocity material (light gray in Fig. F2E) originate? Also the scale of any one zone of hot rock detected seismically beneath an axial volcano is on the order of the entire thickness of rock cored in Hole 735B. The experiment does not resolve whether there are linked sills or nested plutons at any scale smaller than 1-2 km.
The several models we have so briefly summarized have each contributed to a developing conception of crustal accretion at slow-spreading ridges, and although some may have been largely discredited on the basis of one line of evidence or another, there are conceptual threads that link them all which may still turn out to be important. The crystallization of cumulates, the role of deformation, the progress and physical locus of extended igneous differentiation; the likely small dimension and ephemeral existence of bodies of magma; the concept that such bodies of magma are linked in a continuity of space and process from the mantle to the seafloor, perhaps also along the rift axis; the question of whether there are small sills or larger plutons that are relieved of their contents by means of ascent of buoyant melt along vertical cracks; all are still a part of the story. Finally, how the totality of the ocean crust accretes in this environment remains substantially unknown.