At ocean islands, crust is thickened directly beneath high volcanoes by intrusion of primitive hot-spot magmas at the crust-mantle transition, a process termed underplating (e.g., McNutt and Bonneville, 1999). The process that we have described is similar, except that it occurred just off axis at a slowly spreading ridge and it probably was not symmetrically distributed about the ridge axis. Nevertheless, a wedge of plutonic material, gabbro, is also driven between shallow igneous masses and the mantle.
Deep intrusion near the crust-mantle transition has recently been proposed to occur at fast-spreading ridges. However, it is different from the case of Hole 735B. Crawford et al. (1999) suggest that the lowermost crust just at the mantle transition along the East Pacific Rise is a zone of rheological contrast that favors lateral injection of primitive magma supplied from the mantle. These produce high-temperature and perhaps sill-like cumulates of the type found at Hess Deep in the eastern Pacific (Hékinian et al., 1993) and Oman (Boudier et al., 1996; Kelemen et al., 1997; Kelemen and Aharanov, 1999). Magma appears to follow the path of least resistance at the mantle transition, perhaps because it is charged with olivine phenocrysts and is thus fairly dense (Natland and Dick, 1999).
Along the slowly spreading Southwest Indian Ridge, however, imagine that crustal construction is centered on small and often isolated cupolas arrayed along the ridge axis. Imagine also that these are emplaced infrequently and in disconnected fashion, as befits a slowly spreading ridge (Flower and Robinson, 1979). Now consider the next substantial packet of magma to arrive from the mantle. Where will it find the path of least resistance? We argue that in this situation it should be near the top of the crystallizing cupolas, not the mantle transition.
Neutral buoyancy has little to do with this. The strength of the sidewalls to conduits has everything to do with it. Water under pressure flowing through a sprinkler system does not care whether the tubing is made of dense copper or light plastic. It only cares where the holes are or where it can make a hole if there is enough pressure. In a magma plumbing system, if the surrounding rock contains no melt, even if it is still hot, it is impermeable, whether it is made of peridotite or gabbro. Magma will rise through it until overpressure overcomes the vertical load, and then it can flow sideways, lifting the rocks overhead. With enough overpressure, magma can leak sideways at many points all along the length of a long column (drip irrigation) as long as there is an impermeable cap at the end of the path. Obviously, wherever there already is significant melt porosity, that is an easy place to make a hole or to expand one already there.
Formation of adcumulates, which was undoubtedly important in the rocks of Hole 735B (Natland et al., 1991; Natland and Dick, 2001), is thus a critical process that determines the size and distribution of permeability pathways at spreading ridges. At the East Pacific Rise, melt porosity in cumulates right at the ridge axis is very low, only a few percent, and this is an important factor in melt transport through the lower crust and in the formation of shallow melt lenses (Natland and Dick, 1996). In the lower crust at a slowly spreading ridge, where the magma recharge rate is much smaller than along the East Pacific Rise, residual melt right beneath the ridge axis under the rift valley floor is expelled even more nearly to completion between each inflation-injection cycle. The rock, especially that deep in the crust, and though still quite hot—well above the basalt solidus—becomes virtually impermeable beyond distances of more than a few meters. It can fracture, but magmas ascending in the fractures sense only barriers on either side. The mantle transition therefore is transparent. It cannot be a preferred place for lateral magma injection. The only place where ascending magma may ultimately find some horizontal leeway is where any bit of melt is puddled, and the most likely place for this is near the tops of slowly crystallizing magma cupolas, where the crustal load is also less and where there is also a permeability cap. This in general explains why central conduits through nearly crystalline gabbroic material form at slowly spreading ridges (Magde et al., 2000) (Fig. F2E).
A fine point must come during formation of adcumulates when they effectively become solid and impermeable, yet they can still contribute some strongly differentiated liquid to an ascending magma column. Perhaps mixing between primitive magma and ferrobasalt is not strictly between melts but consists of ascending magma reacting with and partially assimilating engulfed blocks containing bits of melt in the diameter of a conduit network. The effect of this on foundered blocks at the Skaergaard intrusion, for example, was that they lost total iron as FeO*, TiO2, and P2O5 to liquids in surrounding cumulates (Sonnenthal and McBirney, 1998).
The East Pacific Rise, in one final contrast, is a place where a considerably higher rate of magma recharge ensures that there will still be melt pockets deep in the lower crust. Residual melt is expunged from cumulates, but not so efficiently and completely as at a slowly spreading ridge. Here, the wall is weak. Thus, a rising packet of magma will first encounter potential porosity in the column at the base of the crust and is able to spread laterally in sill-like masses. Crustal accretion above this may be mainly by straightforward diking or by drip irrigation from intrusions that are dikelike in character. The upper melt lens forms at a permeability barrier, the cap at the end of the system.