1. Mid-Ocean Ridge Volcanism GLY 4310 - Spring, 2015
Petrology Lecture 7
Mid-Ocean Ridge Volcanism
GLY Spring, 2015
The most voluminous type of igneous rock on the planet is MORB, a.k.a. ocean-floor basalt, abyssal basalt, or abyssal tholeiite. Mantle lherzolite rising toward the area between diverging plates undergoes adiabatic decompression followed by partial melting, with the melt filling the gap between the divergent plates.

2. MOR System
The complete mid-ocean ridge is about 65,000 kms long, reaching into the Atlantic, Indian, and Pacific Oceans. The ridge is 1-3 kms about the ocean floor, occasionally reaching above sea-level (Iceland and the Azores, for example). It averages 2000 kms in width. They are often bilaterally symmetric about the axis of the ridge, but asymmetry is not uncommon.
Divergent boundaries are usually found in the oceans. The oceanic lithosphere is much thinner. Divergent boundaries on continents are possible, but do not last long. Either they fail after in incipient rift formation, becoming failed or aborted rifts (aulacogens) or they are successful, splitting continents and creating a new ocean, such as the formation of the Atlantic Ocean after the Triassic split between the Americas and Europe-Africa
Physical observations of the ridge environment tell us that the heat flow is very high, indicating a high geothermal gradient. This results in an extensive hydrothermal system, with seawater circulating downward through fracture crust, being heated, and then ascending. Direct visual observations have shown black smoker type hydrothermal eruptions along the East Pacific Rise, the Juan de Fuca and Gorda Ridges, and the Galapagos Ridge. Rich mineralization and exotic biology characterize the hydrothermal zones.
Earthquake activity is also common along the MOR's. The earthquakes are associated with normal faulting, typical of divergent zones. Gravity studies reveal that the ridges are at isostatic equilibrium. Since the ridges are elevated, they must be compensated by a low-density region directly under the ridge, probably caused by thermal expansion of the hot rock. As plates move away from the ridge, they cool and contract. This creates subsidence. The elevation loss due to subsidence, on either side of the ridge, is proportional to the square root of the rock's age. The conductive heat loss equation indicates that cooling should be proportional to the square root of the rock's age, so this is presumably the major heat loss mechanism.
Spreading rates vary at different ridges, from a low 0.5 to a high of 8 cm a-1.

3. MOR Spreading Rates
Table 13-1 shows some of the observed rates, which are also shown on the previous figure as vectors. The rates quoted are half rates, the rate at which a point on one plate moves away from the ridge. Rates < 3 cm a-1 are considered to be slow-spreading ridges, and those over 4 cm a-1 are called fast-spreading ridges. Temporal variations in spreading rates are known to occur. The morphology of the slow and fast spreading ridges differs.
Slow-spreading ridges have a pronounced axial valley, 30 to 50 kms in width, and 1-5 kms in depth. Inward facing fault scarps are seen, a feature shared with continental rift zones. An inner rift valley is often present, 3-9 kms in width, with a flat floor. Volcanism and crustal extension is concentrated in the inner rift valley floor. Fissures open and pillow lavas erupt. They are constrained by the scarp walls, so flow parallel to the ridge axis. Volcanic activity is uneven. Volcanic mounds up to 300 meters in height are seen. The flows are elongated, reaching kms in length. Future fractures sometimes split the volcanic accumulations. Fissures and volcanic activity are concentrated near the ridge axis. Older flows are slowly carried toward the flanks of the inner rift, where they are torn apart by faulting. This is somewhat more complex than the idea of adding dikes to each side of a central rift, it does correspond in overall effect, especially in creating oceanic crust whose magnetism is aligned with the current field of the earth. This results in the magnetic stripes, first successfully explained by the Vine-Mathews-Morley hypothesis.
Like their slow-spreading counterparts, the fast-spreading ridges have been studied by deep diving submersibles. Their ridges are smoother, with less disruption by large fault displacements. The central rift valley is typical small and poorly developed, or totally absent. Axial summit calderas are common. They are about 100 meters high, a few kms across, and tens of kilometers long. Pillow lava hills are usually flanked by faster lava extrusion associated with the high spreading rate and large heat flow. Basalt becomes older and more fractured away from the spreading center. Some volcanism has been observed up to 4 km from the spreading ridge.
How much magma is produced per year? Estimates for the entire MOR system are in the range 5-20 km3/a. How long have these eruptions been occurring? Samples in the SW Pacific are Jurassic in age (about 140 m.y.), and appear identical to the present material. Of the process has been going on for 140 m.y., that is a total of 1.4 x 109 km3 of basalt. It is likely that actual eruptions have been going on at least ten times as long. This would represent 5% of the upper mantle, more than enough to create significant depletion. Subduction will recycle much of this back into the mantle, but it may not homogenize. Slabs of old crust have been shown to break off and pile up in the upper mantle. These slabs do not melt easily, and represent a significant amount of heterogeneous material.
Ridge segments themselves are offset by fracture zones, now known to be transform faults. The fracture zones often continue as aseismic extensions, disrupting the magmatic anomaly patterns, and sometimes extending across the ocean floor.

4. Oceanic Crust Cross-Section
Figure 13-5 Modified after Brown and Mussett (1993) The Inaccessible Earth: An Integrated View of Its Structure and Composition. Chapman & Hall. London.
Oceanic Crust and Upper Mantle Structure
Layers within the oceanic crust and upper mantle were originally recognized on the basis of seismic velocity discontinuities. Deep Sea Drilling Project results have sampled mainly the oceanic sediments, but has returned up to 1500 meters of volcanic samples from some holes. Ophiolite sequences have added considerable knowledge.
Figure 13-5 shows a schematic sample of the oceanic crust based on ophiolite data. The layering shown is as follows:

5. Oceanic Crust & Upper Mantle Structure
Layer 1 A thin layer of pelagic sediment
Layer 1: Pelagic sediment, absent at the spreading centers, but thicker the further away from the spreading center the sample is obtained

6. Oceanic Crust & Upper Mantle Structure
Layer 2 is basaltic
Subdivided into two sub-layers
Layer 2A & B = pillow basalts
Layer 2C = vertical sheeted dikes
Layer 2: Two sublayers, both composed of basalt. The upper layers, 2A and 2B, are pillow basalt. The second layer, 2C, is composed of vertical sheeted dikes shallowly emplaced in extensional openings. Many of the dikes have a single chilled margin, probably because latter dikes split earlier ones, intruding them. Seismic velocities in layer 2A are lower than expected. This is attributed to high porosity due to fractures and cavities in the basalt. Eventually these fractures are filled by diagenetic mineralization. The seismic velocities of the lower layer are as expected from laboratory studies.

7. Layer 3B is more layered, & may exhibit cumulate textures
Layer 3 more complex and controversial Believed to be mostly gabbros, crystallized from a shallow axial magma chamber (feeds the dikes and basalts)
Layer 3A = upper isotropic and lower, somewhat foliated (“transitional”) gabbros
Layer 3B is more layered, & may exhibit cumulate textures
Layer 3: This layer is mainly composed of gabbro. Presumably this is from shallow magma chambers near the axis, which feed the sheeted dikes. Again the layer is split into two sublayers. the upper, 3A, has isotropic gabbro on top of transitional (somewhat foliated) gabbro. Layer 3B is layered gabbros with cumulate textures. The layering ranges from horizontal to steeply dipping. Some of these gabbros have small diorite to tonalite bodies associated with them, they are thought to be magma filter pressed from the cumulate layer, squeezed up along sheeted dike contacts, sometimes even into the pillow lavas.

8. Oceanic Crust & Upper Mantle Structure
Discontinuous diorite and tonalite (“plagiogranite”) bodies = late differentiated liquids
Isolated bodies of diorite or tonalite composition represent late-stage differenates from the original mamga, filter pressed and mobilized along the gabbro-sheeted dike contact, and may extend up into the pillow layer. They are labeled plagiogranite in the figure.
Figure Lithology and thickness of a typical ophiolite sequence, based on the Samial Ophiolite in Oman. After Boudier and Nicolas (1985) Earth Planet. Sci. Lett., 76,

9. Layer 4 = ultramafic rocks
Ophiolites: base of 3B grades into layered cumulate wehrlite & gabbro
Wehrlite intruded into layered gabbros
Below  cumulate dunite with harzburgite xenoliths
Below this is a tectonite harzburgite and dunite (unmelted residuum of the original mantle)
Layer 4: The seismic velocities in this layer match that of ultramafic rock. The boundary between layers 3 and 4 is the Moho. but where exactly is the boundary? The seismic Moho is easily defined. The P-wave velocity jumps from 7.3 km s-1 (gabbro) to 8.1 km s-1 (peridotite). But there is a possible petrological Moho as well. The upper part of Layer 4 is layered, with cumulate texture. It consists of ol and px, settled from the bottom of axial magma chambers. Below this is an unlayered, residual mantle material. The seismological boundary is at the top of the ultramafic layer. The so-called petrological Moho is deeper, at the top of the unlayered ultramafics.
Ophiolites differ from true oceanic crust in several respects. The seismic velocities in ophiolite are lower than those in oceanic crust, and no magnetic anomaly stripes are seen in ophiolites. Most ophiolites have been hydrothermally altered, and all are weathered. These two processes may erase the magnetic anomalies, and perhaps lower the rock density enough to reduce seismic velocities. Seismic velocities are also reduced by the presence of faults and joints.
More fundamental differences are the thickness of the ophiolite layers, and their composition. Ophiolites layers are thinner than their oceanic counterparts. The composition is also more sialic. Such conditions are typical of back-arc spreading environments. Such environments are between the volcanic arc and the continent in subduction zones. The Sea of Japan is one modern example. The crust is typically thinner in the back-arcs. It is common for the volcanic arc to get pushed back toward the continent, thrusting the back-arc up onto the continent, a process called obduction. Real oceanic crust is thicker and older than back-arc material. Because it is older, it is colder. Combined with the greater thickness, it is stronger and more likely to simply be subducted, rather than obducted.

10. Chemical Analyses of MORB
MORB Petrography and Major Element Geochemistry
MORB’s are chemically distinct from other basalts. They are characterized by:
Composition of Olivine tholeiite
K2O < 0.2% (lower than other basalts)
TiO2 < 2.0% (lower than other basalts)
Texture: Glassy to phyric
Table 13-2 shows data for the Atlantic, Pacific and Indian Ocean Ridges, along with average values. The chemistry of the three ridges is remarkably consistent. All values are from glasses, so no magmatic differentiation processes should have affected the data. The glasses represent frozen liquid, and give the best estimate of magma chemistry. The average chemistry of MORB’s is more limited than many petrogenetic associations.
Another method of showing data is a Fenner variation diagram, where the range of various elements are plotted on the ordinate, versus the variation in a common element on the abscissa.

11. Fenner Diagrams for MORB
Figure “Fenner-type” variation diagrams for basaltic glasses from the Amar region of the MAR. Note different ordinate scales. From Stakes et al. (1984) J. Geophys. Res., 89,
Figure 13-6 shows such a diagram for eight elements vs. Mg, with the Mg concentration decreasing from left to right. Mg is often plotted this way because lower Mg concentrations represent lower temperatures. Evolution of basaltic magmas usually involve a decrease in the Mg/Fe2+. Silica concentration varies little and is thus a poor choice for the abscissa. The change in Mg concentration is from about 9.2% down to 6.8%, or more than 25% change. Silica changes only from 49% to about 51.6%, a variation of just over 5%. Note that this data is from the Amar Valley, MAR, not the AFAR triangle, as stated in the figure caption.
Trends obvious in the diagram are an increase in silica, consistent with the magma becoming slightly more felsic, as mafic minerals crystallize. Ferrous iron concentrations are increasing, presumably as Fe2+ replaces Mg. Removal of early formed olivine would decrease the Mg/Fe2+ concentration. Calcium and alumina concentrations drop. This is consistent with the removal of early formed plagioclase. The alkalis (sodium and potassium), titanium, and phosphorous concentrations are all increasing. Sodium is included in plagioclase, but the plagioclase is anorthitic, so despite some sodium removal, it is still enriched in the melt. Potassium, titanium, and phosphorous are not removed in any crystallizing phase, and are highly enriched in the fluid phase.

12. CaO/Al2O3 vs. Mg.
Examination of the CaO/Al2O3 ratio shows a decrease as differentiation proceeds. The ratio drops from 0.90 to 0.78 (Figure 13-7). Removal of plagioclase cannot account for this. Plagioclase has a CaO/Al2O3 of 0.55, and removal of plagioclase should increase the melt CaO/Al2O3 ratio. Since olivine and opx contain neither Ca nor Al, they cannot be responsible. Cpx, which contains Ca, is the likely candidate for Ca removal.
Other data concerning cpx complicate the picture.
Figure From Stakes et al. (1984) J. Geophys. Res., 89,

13. MORB Variation Diagrams
Figure Data from Schilling et al. (1983) Amer. J. Sci., 283,
Figure 13-8 shows data from both glasses and crystalline phases for the MAR. Figure 13-8a is a plot of CaO/Al2O3 vs. Mg. No trend is evident. Figure 13-8b is a Pearce element ratio diagram, which is a method proposed to indicate extracted phases. A bivariate diagram is used, designed to test for the fractional crystallization of a particular mineral. The denominator of both ratios is the same, generally a single element, which is not contained in the fractionating minerals. The numerators are linear combinations of elements reflecting the composition of the proposed fractionating mineral. Straight line plots in Pearce diagrams indicate a mineral may be fractioning, but do not prove it. The numerators chosen in 13-8b are for olivine and plagioclase. The observed slope of +1 suggests only these two minerals are crystallizing.
Studies on samples from the East Pacific Rise showed that most samples were saturated with olivine and plagioclase. Some were also saturated with cpx. Petrographic analyses show that cpx appears to form late in MORB’s, either as the last phenocryst phase to crystallize, or more commonly as a matrix phase. How can cpx be responsible for Ca/Al reduction in all magmas, not just the more evolved ones in which it forms a separate phase? A study by Rhodes showed that cpx removal is necessary for mass balance to be maintained in models of MORB fractionation, despite the fact that cpx is not a liquidus phase in them. This has come to be called the “pyroxene paradox.”
Pressure once again may solve the mystery. Cpx is a eutectic phase at high pressures. As pressure is reduced, the eutectic shifts toward the Cpx apex, leaving the composition in the ol-plag field.
Taken together, these observations suggest several things about MORB’s.
1. They are not completely uniform magmas, as they were once said to be.
2. They show chemical trends consistent with the fractional crystallization of olivine, plagioclase, and in some cases, cpx.
3. Since fractional crystallization is involved, they cannot be primary magmas. They are derivative magmas resulting from fractional crystallization.
4. The composition of most MORB’s is near the low-pressure cotectic for olivine-plagioclase-cpx, suggesting that fractional crystallization occurred in shallow magma chambers.
5. We saw early that primary magmas, which must be in equilibrium with the mantle, have Mg#s near 70. ( )
Mg# = [(MgO/(MgO+FeO)] x 100
Since this would require Mg concentrations between 10-11%, and few MORB’s have such concentrations, these magmas cannot be primary. Less than 2% of MORB’s have Mg# >65.
To better judge the degree of fractionation in MORB’s, we need to look at the incompatible elements. Potassium, titanium and phosphorous increase by % in MORB’s. If this is due entirely to concentration in the liquid phase in an isolated magma chamber, it implies a reduction of the liquid phase by 50 to 67%. Is this related to the speed of spreading at the ridge?
Work by Grove et al. Showed that fractional crystallization between slowly spreading ridges was in equilibrium with pressures of 0.6 to 0.3 GPa, corresponding to crystallization within the mantle. The mantle must be cooler, allowing crystallization to occur. The EPR, which is a fast-spreading environment, showed equilibrium with pressures between 0.2 and GPa, corresponding to the crust.

14. Glass Composition: Slow vs. Fast Spreading Ridges
Figure Histograms of over 1600 glass compositions from slow and fast mid-ocean ridges. After Sinton and Detrick (1992) J. Geophys. Res., 97,
Figure 13-9 shows the compositions of glasses from both slow and fast spreading areas. The compositions of the fast-spreading ridges are more varied than those of the slow-spreading ridges, and seem to indicate either:
1. A larger proportion of evolved liquids or
2. A greater degree of partial melting
Additional observations show that magmas erupting off the axis are more evolved than those erupted at the axis. Some variations along the axis have also been observed.
These variations suggest great care is necessary in comparing chemical analyses of MORB’s. One way to avoid differences in fractionation effects is to compare samples with equal Mg#.

15. K2O vs. Mg for MAR MORB
Figure shows a plot of potassium versus Mg#. As this figure shows, K2O values vary considerably for nearly constant Mg#. Similar trends are seen for phosphorous and titanium. Since the incompatible elements should be very little affected by fractional crystallization, this again suggests more than one source for MORB’s. Two sources, called N–MORB and E-MORB, are proposed.
N–MORB (N = Normal) taps the depleted upper mantle, which is incompatible element poor. For N–MORB’s with Mg# > 65, K2O < 0.10, and TiO2 < E-MORB’s (enriched MORB, also called P-MORB, for Plume MORB) comes from the deeper, and incompatible rich, mantle. For E-MORB with Mg# > 65, K2O > 0.10, and TiO2 > 1.0.
MORB Trace Element and Isotope Chemistry
N and E MORB's can be further distinguished by examination of the REE chemistry.
Fig shows the variation in K2O with Mg# for the MAR data set of Schilling et al. (1983)

16. REE Patterns for MAR MORBS
Figure shows data for E and N MORB's from the MAR. The LREE patterns show a negative slope for the E-MORB's, but overall LREE enrichment. This is very similar to the enriched mantle xenoliths and basalts seen previously. The N–MORB shows a large LREE depletion, and a positive slope. The HREE patterns for both types are similar.
Since samarium is about halfway between lanthanum and lutetium, and is not plagued with anomalous behavior like europium, it has been suggested that the La/Sm ratio might be a useful distinguishing factor among basalts.
Figure Data from Schilling et al. (1983) Amer. J. Sci., 283,

17. LREE vs. Mg#
Blue = E-Morb
Red = N-Morb
Green = T-Morb
Figure shows a plot of La/Sm vs. Mg#. E-MORB's characteristically have a higher La/Sm ratio then N-MORB's, whatever the Mg#. The E-MORB La/Sm > , N-MORB La/Sm < 1.1, and T-MORB's (transitional) are in between. (Book values are wrong.) T-MORB's show all values between the end members, and are usually interpreted as the product of binary mixing, not as a distinct third type.
Isotopes do not fractionate during either partial melting or fractional crystallization. Examination of isotopic data should help to further confirm or deny the existence of two magma sources.
Figure Data from Schilling et al. (1983) Amer. J. Sci., 283,

18. 143Nd/ 144Nd vs. 87Sr/ 86Sr
Figure is a plot of 143Nd/144Nd vs. 87Sr/86Sr ratios Atlantic and Pacific N–MORB's, and Southwest Indian Ocean E-MORB. The N–MORB's are clustered, with all values of 143Nd/144Nd > (book mistake, , and 87Sr/86Sr < Both sets of values are consistent with a depleted mantle source. The E-MORB's show more enriched values (lower Nd, higher Sr). T-MORB's, not shown in the diagram, show intermediate values.
Petrogenesis of MORB's
Potential parents of MORB's, which are controversial, are saturated with ol, opx, and cpx, in the range 0.8 to 1.2 GPa (25-35 km). This is mainly in the spinel lherzolite field (slight overlap with plagioclase). The lack of HREE depletion tends to exclude garnet lherzolites, since they would preferentially remove HREE. In addition, the data (such as 13-11) show no europium anomaly, expected if plagioclase lherzolite were the source. Multiple saturation indicates the point of separation of magma and source rock, not necessarily the point of magma generation, which might be much deeper. Major element (and thus mineralogical) data are controlled by the equilibrium between melt and residual mantle phases during the magma's ascent toward the surface. Incompatible trace element and isotope data reflect equilibrium between the melt and the ultimate source rock. The ultimate source depth is certainly greater than kms. N-MORBs may reach down to 80 kms, E-MORBs even deeper.
Divergent plate boundaries create an opening, allowing diapirs of mantle material to move upward. Decompressive partial melting beings as the diapir ascends. For N-MORBs, the onset of melting is in the km depth range. Experimental work indicates partial melting percentages of percent are achieved. Partial melting terminates when conductive heat loss to the surface becomes too high. The disappearance off cpx may trigger the exact end of melting, since the melting temperature jumps after the cpx disappears.
Figure Data from Ito et al. (1987) Chemical Geology, 62, ; and LeRoex et al. (1983) J. Petrol., 24,

19. Generation of N-MORB and E-MORB
Figure
After Zindler et al.
(1984) Earth Planet.
Sci. Lett., 70, 175
-195. and Wilson
(1989) Igneous
Petrogenesis,
Kluwer.
Melt blobs separate from their residuum at about km (see Figure 13-14). Upward migration probably continues to an axial magma chamber 1-2 km below the surface.
Plumes may be initiated by either of two events:
1. Enriched mantle material moving upward beneath the rising N-MORB material (Figure 13-14).
2. The plume may form independently, but in a geographical coincidental manner, so that it emerges near the ridge axis.
The ultimate source region for a plume should be below 660 km, in enriched mantle material. The plume may stall, becoming entrapped in the upper mantle, or may rise until partial melting due to decompression occurs. The rising melt may or may not interact with N-MORB, thus producing a transitional MORB.
Axial Magma Chambers
Sets of persistent magma chambers just below the axis offer a number of advantages in explaining the eruption of magma at the ridge.

20. The Axial Magma Chamber
Original Model
Semi-permanent
Fractional crystallization ® derivative MORB magmas
Periodic reinjection of fresh, primitive MORB
Dikes upward through extending/faulting roof
Hekinian et al. (1976)
Contr. Min. Pet. 58, 107.
They were postulated as narrow (5 km) and not too deep (up to 9 km). Magma was injected and began to fractionally crystallize to produce derivative MORB's. Dikes could emanate upward to create sheeted dike complexes, and feed the pillow flows. Crystallization along the walls, top, and bottom of the chamber could create gabbros seen in layer 3. Periodic injections of fresh magma, together with divergent plate movement, would continually expand the chamber, preventing the crystallization from completely filling the chamber.
Figure From Byran and Moore (1977) Geol. Soc. Amer. Bull., 88,

21. Semi-Permanent Axial Magma Chamber
Infinite onion model, since it resembled an infinite number of onion shells created from within and added to the walls
Indeed, Cann called this model the "Infinite Onion", since layers would being added from the inside. The dense ol and opx would settle toward the floor, producing the ophiolite layers of level 3, and suspected to be present in layer 4.
Such an "open-system" model (figure 13-16), with periodic injections of fresh magma, appeared to nicely explain features seen at and below the mid-ocean ridge, and was widely accepted. One problem existed. Such shallow magma chambers, full of liquid, should be readily detectable by seismic wave studies. But seismic studies have failed to locate magma chambers of any significant size beneath the axial ridges. So, as any model which fails to fit new data must, this model was discarded.

22. Axial Magma Chamber, Fast-Spreading Ridge
Figure After Perfit et al. (1994) Geology, 22,
A new model has been developed for fast-spreading ridges. The liquid occupies a lens, from 10's to 100's of meters thick, less than 2 km wide, and 1-2 km below the ocean floor. It creates a sub-horizontal seismic reflector, compatible with seismic observations along and across the EPR. The melt region is surrounded by a much larger "mush" region. (See Figure 13-17) The mush is solid enough to transmit shear waves, and is in turn surrounded by a transition zone. Both regions are seismic low-velocity zones. The magma chamber consists of both the melt and mush regions. The mush has enough liquid so that liquid circulation through it is possible. Crystallization occurs, solidifying the mush, and turning the melt to mush. Primitive magma is periodically reinjected, very similar to the Infinite Onion model.
The amount of liquid in the mush is not well known, but may be up to 30%. This zone is consistent with an in situ crystallization model, first proposed by Langmuir.

23. Crystal Mush Zone
The crystal mush zone contains perhaps 30% melt and constitutes an excellent boundary layer for the in situ crystallization process proposed by Langmuir
The crystal-laden mush near the walls is near the solidus temperature. It would grade to pure liquid at the center, which is hotter. Liquid composition would range from low-temperature eutectic near the walls to the original bulk composition near the center. Cpx would crystallize along the walls. Melts in equilibrium with the cpx might be expelled, mixing with the cpx-free melts of the interior. By removing cpx, then creating a mix where cpx is not in a liquidus phase, this may explain the pyroxene paradox, which required removal of cpx for mass balance, but did not have any cpx crystallization. Elevated concentrations of incompatible elements are also produced in the resulting mixed liquid. Langmuir's model is consistent with laboratory and theoretical studies of magma chambers.
The transition zone is necessary because seismic velocities are still lower than expected outside of the mush zone. The mush/transition boundary may represent a rigidus, a zone where solidification > 50%, and contact between crystals makes the entire assemblage behave like a crystalline aggregate.
The small sill model of the magma chamber seems to present difficulties for the fractional crystallization/crystal settling model traditionally used to explain the layered gabbros and cumulates. Recent research may have provided an answer. Leg 735B of the DSDP in the SW Indian Ocean penetrated 500 m of gabbro. The textural and chemical variations observed in these gabbros are consistent with evolution at the crystal-rich margin of a small magma chamber. Lithologic variations in the gabbros (layers) may result from isolated melt pockets formed during deformation. Layering of oceanic gabbros in ophiolite has been attributed to deformation during spreading, not to crystal settling.
Dike structures are usually thought to be limited to the immediate axial area. Lava fields and pillow ridges, which appear very young, have been found up to 4 km from the axis.
Figure From Winter (2001) An Introduction to Igneous and Metamorphic Petrology. Prentice Hall

24. Discontinuous Axial Magma Chamber
Available chemical data also show the off-axis lavas to be much more heterogeneous, when compared to axial summit caldera (ASC, F13-17) lavas. Ephemeral off-axis magma chambers, of either N-MORB or E-MORB, could be responsible. Off-axis intrusions and extrusions may also explain a rapid thickening of layer 2A away from the spreading center.
The melt region has been seismically traced for several kms along the ridge axis. Gaps appear at fracture zones. Small gaps also are seen between fracture zones, as shown in Figure There are chemical variations on the scale of 100's off meters to km, indicating poor mixing along the axis. The high heat flow of fast-spreading ridges can maintain persistent magma chambers.
Slow-Spreading Ridges
The heat flow at slow-spreading ridges is much less, and the persistence of magma chambers is thus quite doubtful.
Figure After Sinton and Detrick (1992) J. Geophys. Res., 97,

25. Axial Magma Chamber, Slow-Spreading Ridge
Distance (km) 10 5 2 4 6 8
Depth (km)
Moho
Transition
zone
Mush
Gabbro
Rift Valley
Figure shows a schematic cross-section of a slow-spreading ridge. A dike-like mush zone, and smaller transition zone, replace the steady state magma lens. The temperatures are well below the liquidus, severely limiting convection and mixing. It has been proposed that small, ephemeral magma bodies occur along the slow-spreading ridge by Nisbit and Fowler. They dubbed this model the Infinite Leak, intending it as a variation of the Infinite Onion. These magma bodies probably concentrate in the mush area. Seismic velocity studies are consistent with, at most, a few percent melt.
Slow-spreading ridges generally show less differentiation than fast-spreading ridges. Magmas entering the shallow melt zones are apt to erupt directly to the surface, with less time for any interaction. Polybaric fractionation or plagioclase accumulation in magma blobs does occur along slow-spreading ridges, leading to some heterogeneity. Slow-spreading ridges show closed-system behavior, rather than the open-system behavior of fast-spreading ridges.
Studies in the laboratory of C.H. Langmuir have suggested a series of conclusions. Some of their work was based on what they called "global" results, averages for ∼100 km segments of the ridges, designed to average out local variations. Their conclusions:
1. Low average Na8.0 (wt. % Na at 8.0 wt.% Mg) was found to exist with high average Fe8.0. Sodium is an incompatible element, concentrating in early melts. Therefore, low Na indicates extensive melting. Iron correlates with depth of melting, with more Fe at greater depths. Thus low Na, high Fe indicates that the mean degree of melting increases with depth within the melting column. The melting column is the depth between the onset of melting and the point nearer the surface where conditions are too cool for any further melting to occur. Their major conclusion was that global correlations are controlled by differences in thermal regime between segments, much more than by magma composition differences between segments.
2. The thermal regime between a ridge segment exerts a major control on the quantity and composition of MORB's. This in turn controls crustal thickness, and the depth below sea-level.
3. Melts are extracted from depth without re-equilibrating at lower pressures. Their work shows a correlatable mean-pressure signal from the averaged MORB data. This high-pressure signal would be wiped out by low-pressure re-equilibration. They suggest, "a process of fractional melting, in which incremental melt blobs escape from the mantle matrix rapidly and efficiently, as soon as they reach some small fraction sufficient to permit extraction". This does not preclude different batches of melt from mixing at depth, prior to eruption.
4. Local trends vary from the global average trends. Local trends in slow and fast environments differ. Less temperature variation, corresponding with more efficient processing, are seen in fast environments. Slow environments are less efficient, and more heterogeneous.
5. Hot spots play an important role in MORB chemical compositions. Iceland and the Azores, which rise above the MAR, are combinations of divergent plate boundaries and hot spots. The combination produces more magma, accounting for their greater highs. It also accounts for the transitional isotopic data seen in Figure
Figure After Sinton and Detrick (1992) J. Geophys. Res., 97,

26. Oceanic Basalt
Figure (a) Initial 143Nd/144Nd vs. 87Sr/86Sr for oceanic basalts. From Wilson (1989). Igneous Petrogenesis. Unwin Hyman/Kluwer. Data from Zindler et al. (1982) and Menzies (1983).

27. Ultramafic Xenoliths
Figure (b) Initial 143Nd/144Nd vs. 87Sr/86Sr for mantle xenoliths. From Wilson (1989). Igneous Petrogenesis. Unwin Hyman/Kluwer. Data from Zindler et al. (1982) and Menzies (1983).