A central paradigm of RIDGE studies is the hypothesis that mantle flow, or melt extraction, or both, are focused in three dimensions toward the centers of magmatic ridge segments, at least at slow-spreading ridges such as the Mid-Atlantic Ridge. This is based on:

1. Observations from ophiolites, with emphasis on the Oman ophiolite (Ceuleneer, 1991; Ceuleneer et al., 1988; Ceuleneer and Rabinowicz, 1992; Jousselin et al., 1998; Nicolas and Boudier, 1995; Nicolas and Rabinowicz, 1984; Nicolas and Violette, 1982);
2. The theory that partially molten mantle may be subject to diapirism via Rayleigh Taylor instabilities (Barnouin-Jha et al., 1997; Buck and Su, 1989; Crane, 1985; Jha et al., 1994; Parmentier and Phipps Morgan, 1990; Rabinowicz et al., 1984, 1987; Schouten et al., 1985; Sparks and Parmentier, 1993; Sparks et al., 1993; Su and Buck, 1993; Whitehead et al., 1984);
3. The observation that peridotites are commonly dredged near fracture zones along slow-spreading ridges but not near ridge segment centers (Dick, 1989; Whitehead et al., 1984); and
4. Gravity and seismic studies of the Mid-Atlantic Ridge suggesting thick crust near segment centers and thin crust at segment ends (e.g., Barclay et al., 1998; Kuo and Forsyth, 1988; Lin et al., 1990; Tolstoy et al., 1992; Tucholke et al., 1997).

In addition to a possible role for mantle diapirism, various workers have proposed that melt transport may be focused in two or three dimensions, on the basis of theoretical work and field observations (e.g., Aharonov et al., 1995; Kelemen et al., 1995a, 1995b, 2000; Magde et al., 1997; Phipps Morgan, 1987; Sparks and Parmentier, 1991, 1994; Spiegelman, 1993; Spiegelman and McKenzie, 1987). Such focused melt extraction could operate, with or without focused flow of the upwelling mantle, to produce the observed focusing of crustal accretion toward the center of magmatic ridge segments.

The idea that focused mantle upwelling at the centers of magmatic ridge segments occurs only beneath slow-spreading ridges was formulated by Marc Parmentier and his students (e.g., Lin and Phipps Morgan, 1992; Parmentier and Phipps Morgan, 1990; Turcotte and Phipps Morgan, 1992) and is supported by seismic results from the recent Mantle Electromagnetic and Tomography (MELT) experiment along the fast-spreading southern East Pacific Rise, in which no focused mantle upwelling was detected (e.g., Forsyth et al., 1998; MELT Team, 1998; Toomey et al., 1998). However, recent observations from Oman and the fast-spreading northern East Pacific Rise have called this into question (e.g., Barth and Mutter, 1996; Dunn and Toomey, 1997; Nicolas et al., 1996). Nevertheless, most investigators agree that slow-spreading ridges such as the Mid-Atlantic Ridge represent the best place to test general hypotheses for the mechanism(s) of three-dimensional (3-D) focusing of crustal accretion.

In the literature describing theories of three-dimensionally focused mantle upwelling, the terms "focused" and "3-D" receive different definitions from different authors. Thus, Parmentier and Phipps Morgan (1990), who first presented the now-famous "phase diagram" for two-dimensional (2-D) vs. 3-D mantle upwelling as a function of spreading rate and mantle viscosity, chose a detailed example that is indeed 3-D but that does not correspond well to observations of diapirs in the mantle section of the Oman ophiolite. In Parmentier and Phipps Morgan's example, the region of mantle upwelling at, for example, 40 km depth is ~200 km wide in a ridge-parallel section and widens upward; near the top it is almost as wide as their 300-km ridge segment. Along-ridge transport of upwelling mantle occurs gradually over the upper 60 km of the upwelling region. In contrast, the interpretation of Jousselin et al. (1998), loosely based on observations from Oman, is that "at any depth above 50 km there is no vertical flow outside the narrow zone of subridge upwelling." They take the zone of upwelling to be cylindrical, with a diameter of ~10 km. Furthermore, in their interpretation, all corner flow (ridge parallel and ridge perpendicular) occurs in the upper 500 m of the upwelling region. More than one-half of the shallow mantle in their 25-km-long ridge segments is fed by horizontal flow in this 500-m-thick layer just below the base of the lithosphere. Such narrow pipes of upwelling mantle may be consistent with the physical models of Buck and Su (1989; Su and Buck, 1993), which show very sharp focusing of mantle flow. Such features could conceivably have escaped seismic detection in the recent MELT experiment. However, if this is the geometry of mantle upwelling, then the amount of ridge-parallel, horizontal transport of mantle material must be very large.

In the ensuing discussion, we take the Jousselin et al. geometry as the end-member example of 3-D focused mantle flow and passive corner flow to be the end-member example of 2-D mantle flow with no focusing. The Jousselin et al. scenario may seem extreme at first, but it does provide a clear description of an upwelling geometry that could produce a variation in igneous crustal thickness from ~10 km at a segment center to ~0 km near the segment ends, as interpreted on the basis of geological and geophysical observations in the 14° to 16°N region of the Mid-Atlantic Ridge. These observations are typical of the first-order features of slow-spreading ridges, which are thought to reflect three-dimensionally focused magmatic accretion.

In contrast, available 3-D physical models of diapric mantle upwelling beneath ridges cannot account for these observations because the upwelling is not sufficiently tightly focused. As stated by Barnouin-Jha et al. (1997), "short wavelength segmentation of slow-spreading centers requires some process not included in our models of mantle flow." This missing process might be tightly focused upwelling, as in the schema of Jousselin et al., or focused melt migration.

Despite the difficulties with 3-D physical models (outlined above in "Really Focused, Sort of Focused, Unfocused, Blurry: Terms for Mantle Upwelling"), the hypothesis that mantle flow, or melt extraction, or both, are focused in three dimensions toward centers of magmatic segments at slow-spreading ridges has essentially reached the status of accepted theory. However, these ideas have never been subject to a direct test. A strike line of oriented mantle peridotite samples extending for a significant distance within such magmatic segments offers the possibility of directly testing hypotheses for focused crustal accretion.

The primary aim of drilling in the 14° to 16°N area along the Mid-Atlantic Ridge was to characterize the spatial variation of mantle deformation patterns, residual peridotite composition, melt migration features, and hydrothermal alteration along axis. Published hypotheses for focused solid or liquid upwelling beneath ridge segments make specific predictions regarding the spatial variation of mantle lineation or the distribution of melt migration features.

Models of focused solid upwelling require ridge-parallel, subhorizontal flow of residual mantle peridotites from segment centers to segment ends (Fig. F6A) (Barnouin-Jha et al., 1997; Buck and Su, 1989; Crane, 1985; Jha et al., 1994; Parmentier and Phipps Morgan, 1990; Rabinowicz et al., 1984, 1987; Schouten et al., 1985; Sparks and Parmentier, 1993; Sparks et al., 1993; Su and Buck, 1993; Whitehead et al., 1984). Futhermore, because focused mantle upwelling must be faster than plate spreading, these models predict zones of substantial shear strain at the top of the upwelling "fountain" of mantle peridotite, through which all the upwelling solid material must pass. These theories are substantiated to some extent by patterns of mantle flow inferred from ductile fabrics in residual peridotites in the Oman ophiolite (Fig. F6B) (Ceuleneer, 1991; Ceuleneer et al., 1988; Ceuleneer and Rabinowicz, 1992; Jousselin et al., 1998; Nicolas and Boudier, 1995; Nicolas and Rabinowicz, 1984; Nicolas and Violette, 1982), although as already noted in "Really Focused, Sort of Focused, Unfocused, Blurry: Terms for Mantle Upwelling" the scale of focused upwelling in Oman (~10 km) is different from that in current 3-D models of mantle diapirism (~100 km). In Oman and other ophiolite massifs, mantle flow direction can be determined by measurement of spinel shape fabrics (lineation at high strain is parallel to ductile flow), measurement of the orientation of olivine crystal shape fabrics relative to subgrain boundaries (subgrain boundaries are oblique to the long sides of elongate crystals, indicating the sense of shear), and measurement of olivine crystal lattice preferred orientation (olivine a-axes are aligned parallel to ductile flow directions at high strain).

Cores from a series of drill holes in mantle peridotites along a slow-spreading ridge axis can, in principle, be used to test the prediction that shallow ductile flow of residual mantle at the ends of segments is ridge parallel and subhorizontal. There are two problems with this approach: (1) the core must be restored to a geographical reference frame and (2) tectonic rotations of the peridotite that postdate ductile flow must be considered before the orientation of ductile fabrics can be interpreted in terms of large-scale mantle flow. Work on cores of partially serpentinized mantle peridotite from the East Pacific (Boudier et al., 1996) and the Atlantic (Ceuleneer and Cannat, 1997) has shown that they can be reoriented into the geographical reference frame using remanent magnetization, provided that tectonic rotations were minor or that their effects can be removed (Hurst et al., 1997; Kelso et al., 1996; Kikawa et al., 1996; Lawrence et al., 1997; Richter et al., 1996).

There are several assumptions involved in the reorientation process, namely

1. The remanence in individual samples accurately records the geomagnetic field direction (there is little magnetic anisotropy);
2. The average remanence in a given hole or at a given site reflects the time-averaged field direction (magnetization is slow enough that paleosecular variation is averaged;
3. The type of remanence and the timing of remanence acquisition must be known (the remanence need not be primary, so that, for example, a late viscous remanence would be adequate for geographic reorientation
4. The polarity of the remanence must be known; and
5. Independent constraints on the rotation axis—if any—must be available.

It is not clear whether all these assumptions are actually valid in all previous studies that attempted reorientation of core from ODP drilling, but all of them are implicit in the reorientation process.

It is noteworthy that there are two distinct types of reorientation studies. In the first, one simply reorients all the remanence directions to a common azimuth to look at the clustering of the structural features. For convenience, this common azimuth may be chosen as 360° or 180°, depending on the polarity of the remanence. During drilling, individual core pieces are rotated to different extents around a near-vertical axis, and this process restores them all to a common orientation, presuming that all had a common remanence azimuth prior to drilling. This process may also be interpreted as crudely placing structural features in a geographical reference frame. However, this interpretation is only valid if the remanence is a relatively recent one (e.g., as at ODP Site 920) or if the amount of rotation is sufficiently small or about an axis such that little change in declination has occurred.

A second reorientation process is needed when the amount of remanence deviation from the expected dipole direction is large, or if rotations occurred about axes that will substantially change both the inclination and declination of the remanence (as is generally the case). This reorientation requires selecting a plausible rotation axis (on geological grounds) and calculating the amount of deviation in both inclination and declination. For specific examples of this type of process, with a quantification of the effect of different rotation axes on both the declination and inclination of the remanence direction, see the "Paleomagnetism" sections in the "Site 1268" and "Site 1270" summaries below.

After "undoing" tectonic rotations prior to drilling, the improved declination can, in principle, be used for the common azimuth in reorientation to yield something closer to a geographical reference frame. One could go further and undo the full rotation of the remanence vector to provide the attitude of structural features at the acquisition.

Where the magnetic inclination in the core after horizontal rotation is not parallel to the inferred magnetic inclination at the time of lithospheric formation, tectonic rotations may be inferred and then "removed." However, an important caveat is that the remanent magnetization in partially serpentinized peridotites is hosted in magnetite that is produced during serpentinization, so tectonic rotations of the peridotite prior to serpentinization cannot be detected.

Accounting fully for possible tectonic rotations of exposed mantle peridotite is a daunting prospect, but there was hope for a definitive result for the following reasons. Tectonic rotations resulting from normal faults are likely to occur mainly around axes parallel to the ridge axis. Thus, subhorizontal, ridge-parallel flow lineation is likely to be affected very little, if at all. Furthermore, rotations are likely to be away from the ridge axis, increasing the angle between lineations and the ridge axis. Thus, if ridge-parallel lineations had been consistently observed, this could have been taken as good evidence that shallow ductile flow of the mantle was parallel to the ridge. In the best case, observation of systematically varying ductile flow lineations in mantle peridotites, ranging from nearly ridge-perpendicular lineation near segment centers to ridge-parallel lineations near segment ends would have constituted evidence that focused mantle upwelling did occur near segment centers.

Models of focused crustal accretion predict different patterns of mantle depletion due to melt extraction as a function of distance from magmatic segment centers. For strongly focused 3-D mantle flow, there should be no variation in the degree of mantle depletion along axis, since all of the shallow mantle peridotites originate within a narrow, pipelike upwelling zone. For purely passive corner flow, with no other factors considered, again there should be no variation in depletion along axis. However, when passive flow is coupled with cooling of the ends of ridge segments against a fracture zone wall, then the degree of melting is predicted to decrease along axis away from segment centers. This has been termed the "transform edge effect" (Ghose et al., 1996; Langmuir and Bender, 1984; Magde et al., 1997; Phipps Morgan and Forsyth, 1988). Provided that melt extraction is equally efficient throughout the melting region, this variation in melt production should be observed in shallow mantle samples. If partial crystallization of melt migrating into conductively cooled mantle lithosphere occurs, forming "impregnated peridotites" (e.g., Ceuleneer et al., 1988; Ceuleneer and Rabinowicz, 1992; Dick, 1989; Elthon et al., 1992; Seyler and Bonatti, 1997), then this should occur primarily near fracture zones, enhancing the chemical signal of the transform edge effect in mantle peridotites. Furthermore, impregnated peridotites often preserve structural relationships indicative of the nature of melt migration. Impregnated peridotite samples from the western ridge/transform intersection (RTI) of the Kane Fracture Zone (Ishizuka et al., 1995) show evidence for migration of melts into localized ductile shear zones, suggesting that melt migration extended into the active transform fault.

In general, geochemists have searched for the transform edge effect in lavas, which is complicated by the difficulties of seeing through variations in crustal differentiation processes and mantle source composition. Detailed analysis of our suite of peridotite samples, collected from a two ridge segments at various distances from a fracture zone, will provide an independent evaluation of the presence and importance of the transform edge effect.

Models of focused melt migration toward ridge segment centers predict various different spatial distributions and orientations of melt transport features. Before discussing the various predictions, we will introduce some of the melt transport features that can be recognized in mantle peridotite samples. For reviews of the literature on these features, please see papers by Nicolas (1986, 1990) and Kelemen et al. (1995a, 1997a). Melt transport features include the following:

1. Dunites are rocks composed almost entirely of the mineral olivine, with minor spinel; pyroxene generally forms <1% of these rocks. Dunites occur in tabular to cylindrical bodies in ophiolite peridotites. Few, if any, are tabular dikes filled entirely with magmatic olivine. Instead, most or all form by dissolution of pyroxene and crystallization of a smaller amount of olivine in olivine-saturated melt migrating by porous flow. Some have an origin entirely via focused porous flow, either in dissolution channels or within ductile shear zones, whereas others form in porous reaction zones around cracks. The relative importance of entirely porous- vs. fracture-related origins for dunites is controversial, but is not crucial to this report. The main point to be made here is that dunites commonly form in the region of adiabatic mantle upwelling beneath spreading ridges, though they also form within the region of transition between adiabatic upwelling and conductively cooled lithosphere.
2. Pyroxenite and gabbro mantle dikes are highly elongate, generally parallel-sided features that almost certainly form as fracture-filling magmatic rocks. Their compositions, where they have been studied in detail, are indicative of crystal fractionation from magma that was cooling within conductive "lithosphere." However, this is debated, and Nicolas and coworkers have interpreted them to be representative of melt-filled fractures that form within the adiabatically upwelling mantle.
3. Large gabbroic plutons, perhaps generally with high aspect ratios but distinguished from dikes by having horizontal extents of tens to thousands of meters, may intrude mantle peridotite. These are abundant in the "crust–mantle transition zone" of the Oman ophiolite (Boudier and Nicolas, 1995; Boudier et al., 1996; Kelemen et al., 1997b; Korenega and Kelemen, 1997), where they are interpreted as "ponds" of melt accumulated beneath a permeability barrier created by the onset of crystallization from cooling magma entering the thermal boundary layer (Kelemen and Aharonov, 1998). At slow-spreading ridges, such plutons might form at much greater depth, perhaps as deep as 20 km, since thermal models and metrological data suggest that the thermal boundary layer is that thick at half-spreading rates of ~10 mm/yr (Sleep, 1975; Reid and Jackson, 1981; Braun et al., 2000; Michael and Chase, 1987; Meurer et al., 2001; Grove et al., 1992; C. Xia et al., unpubl. data).

We now consider predictions of spatial distribution and orientation of melt migration features, with an emphasis on dunites formed within the adiabatically upwelling mantle. Most models predict that such dunites are transposed into a subhorizontal orientation in the shallow mantle, at least by 2-D corner flow and perhaps also by 3-D diapiric flow. (Dunites that are not subhorizontal may have formed in the region of transition from adiabatically upwelling mantle to conductively cooled lithosphere).

1. If melt migration and crustal accretion are focused mainly because of diapirism, as proposed by Nicolas (1990), then no systematic variation in dunite abundance along the ridge axis is predicted. Futhermore, if melt-filled fractures form within mantle diapirs and these are represented by mantle dikes (Nicolas, 1990), then mantle dikes may be nearly vertical near segment centers and progressively transposed into a horizontal orientation toward segment ends.
2. If melt migration and crustal accretion are focused mainly because of melt migration beneath permeability barriers parallel to the base of the lithosphere, as proposed by Sparks and coworkers (Magde et al., 1997; Sparks and Parmentier, 1991, 1994) and Spiegelman (1993), then dunites should be shallowest (and most commonly sampled by drilling) near the centers of segments.
3. If melt migration and crustal accretion are focused mainly as a result of coalescing porous flow within the upwelling mantle, as proposed by Phipps Morgan (1987), Spiegelman and McKenzie (1987), Kelemen and coworkers (Aharonov et al., 1995; Kelemen et al., 1995a, 1995b, 2000), and Daines, Zimmerman, and Kohlstedt (Daines and Kohlstedt, 1997; Kohlstedt and Zimmerman, 1996), then dunite abundance in the shallow mantle should increase toward segment centers. Thus if porous flow mechanisms predominate in producing focused crustal accretion, then dunite abundance should increase toward segment centers, whereas if diapiric upwelling is the predominant reason for focused crustal accretion, then dunite abundance should be relatively constant along axis.

On a smaller scale, the detailed size/frequency and spatial distribution statistics of a large number of dunite veins in outcrops of mantle peridotite can be used as indicators of the geometry of melt extraction conduits (Kelemen et al., 2000; Braun and Kelemen, 2002). Dunites in mantle outcrops in the Ingalls and Oman ophiolites show a negative power-law relationship between size and abundance, with many small dunites and only a few large ones. This is consistent with the hypothesis that dunites form an interconnected channel network in which many small conduits feed a few large ones. The systematics of the spatial distribution can be used to distinguish between dunites that originate as reaction zones around cracks and dunites that form entirely as a result of porous flow mechanisms.