In this section we discuss the major implications of the data in terms of igneous pathogenesis under several headings, each of which pertains to a particular question. We agree with previous workers (e.g., Hébert et al., 1991; Bloomer et al., 1991; Ozawa et al., 1991; Natland et al., 1991; Dick et al., 1991, Dick, Natland, Miller, et al., 1999) on major aspects of Hole 735B petrology. We offer additional interpretations with demonstrations.
Figure F1 shows the correlated variations of An content in plagioclase and Mg# in olivine, clinopyroxene, and orthopyroxene of the Hole 735B gabbroic rocks. Whereas the scatter exceeds the analytical error, the significant correlations demonstrate that these minerals are, to a first order, in chemical equilibrium. Such first-order equilibrium could be achieved under subsolidus conditions, but the slowness of solid-state diffusion and the very young age (~11.5 Ma) of the system argue strongly that the observed equilibrium was largely achieved under magmatic conditions. In other words, the bulk of the coexisting minerals (plagioclase, olivine, clinopyroxene, and orthopyroxene) in each sample were coprecipitated from a common liquid undergoing cooling. For example, the compositional range of these minerals corresponds to liquidus temperatures from ~1190°C for high Mg# and An samples to ~1070°C for low Mg# and An samples (see below). The relationship Mg#cpx > Mg#opx > Mg#ol is consistent with experimental results (e.g., Grove et al., 1992), although the differences are larger, perhaps due to subsolidus reequilibration(?).
It is important to note, however, that Fe-Ti oxides exist in many of these samples, and their compositions do not in any manner (not shown) correlate with the compositions of olivine, plagioclase, clinopyroxene, and othopyroxene. A simple argument is that these oxides were not in equilibrium with the major silicate minerals and thus were not precipitated from the same melt at the same conditions. This is an important observation (see below). We also must not dismiss the fact that the scatter in Figure F1 is greater than analytical error. In some cases, the mineral compositional variation in a given crystal is as large as within-sample variation. It is possible that such compositional heterogeneity could be due to kinetics, but it could well be caused by reequilibration of mineral phases with entrapped late-stage melt (Natland et al., 1991; Natland and Dick, 1996; Coogan et al., 2001).
Assuming that the gabbros were indeed emplaced at lower crustal level, then their crystallization pressures must be ~1-1.5 kbar. Their crystallization temperatures can be precisely calculated from experimentally established basalt phase equilibria (e.g., Roeder and Emslie, 1970; Bender et al., 1978; Walker et al., 1979; Langmuir and Hanson, 1981; Nielsen and Dungan, 1983; Weaver and Langmuir, 1990), which states that the liquidus temperature of basaltic melt is proportional to the MgO content in the melt, and also to the Fo content of olivine that is in equilibrium with the melt. The relationship between the two parameters is characterized by the well-known Fe-Mg exchange relationship of Roeder and Emslie, 1970:
Kd = (XMgL/XFe2+L)/(XMgol/XFe2+ol) = 0.3 ± 0.03.Figure F2A shows the relationship between Mg# of a basaltic melt (lower shaded band) and Mg# (or Fo content) of olivine (upper shaded band) in equilibrium with the melt as a function of liquidus temperature. Also plotted in Figure F2A are whole-rock Legs 118 and 176 samples with olivine analyses available. Typical normal-type MORB (N-MORB) samples from the East Pacific Rise (EPR) and MORB samples from the eastern wall of the Atlantis II Fracture Zone (AII F.Z.) basalts are plotted for comparison.
Note that although many Leg 118 samples were studied for mineral chemistry and whole-rock compositions, not many have both mineral and whole-rock data available. In any case, most whole-rock gabbros are plotted onto or above the band defined by the liquidus olivine (Fig. F2B), which corroborates the above interpretation that the bulk of the coexisting minerals (plagioclase, olivine, clinopyroxene, and orthopyroxene) in each sample are in equilibrium and were coprecipitated from a common liquid undergoing cooling. The few samples with low Mg# plotted way below the olivine liquidus band are samples with significant amounts of excess Fe-Ti oxides, which also corroborates the above interpretation that these excess oxides were not in equilibrium with, but extraneous to, the major silicate minerals in these rocks. Most of these samples plot above the olivine liquidus because whole-rock samples contain pyroxenes that have higher Mg# than Fo in olivine (Fig. F2B; also see Fig. F1).
The most important
conclusion from Figure F2
is that the bulk of Hole 735B gabbros were formed from evolved melts. The
olivine Fo84.2
in the most primitive troctolite (Ozawa et al., 1991) corresponds
to a parental liquid of Mg# 
0.637 (0.590-0.637, for Kd = 0.27-0.33) with a liquidus
temperature of 1190° ±
10°C. This is not
surprising because of the simple fact that clinopyroxene appears on the
tholeiite liquidus at temperatures no higher than ~1180°C at lower pressures
(<2 kbar) and that the bulk of Hole 735B lithology is gabbroic with 40%-50%
modal clinopyroxene.
Bowen (1928) first suggested that gabbros were formed by crystal sorting and accumulation from cooling basaltic melts. Four decades later, Wager and Brown (1968) illustrated in great detail various types of cumulate textures developed during the formation of gabbros in response to cooling of basaltic magmas. Perhaps most, if not all, modern igneous petrologists would accept the concept by Bowen and the conclusions by Wager and Brown, but ironically, some modern igneous petrology textbooks maintain "gabbroic rocks are intrusive/compositional equivalents of basalts but have phaneritic grain size" (e.g., Best, 1995), implying that the slow cooling, and thus greater grain size, is the sole difference between gabbros and basalts. A simple conceptual argument is that if gabbros are indeed the product of fractional (vs. equilibrium) crystallization, then we cannot avoid the conclusion that gabbros are cumulate, not melt, equivalents, although gabbros could trap a significant amount of interstitial melt (e.g., Coogan et al., 2001). Natland et al. (1991) and Natland and Dick (1996) argue for the cumulate nature of Hole 735B gabbros drilled during Leg 118 and gabbros drilled from Hess Deep during Leg 147. Casey (1997) and Ross and Elthon (1997) also demonstrated the cumulate nature of gabbros drilled at the Mid-Atlantic Ridge Kane Fracture Zone (MARK) area during Leg 153. However, Hart et al. (1999) used compositions of strip samples from Leg 118 and Dick et al. (2000) used reconstructed major element compositions of bulk Hole 735B to interpret that the Hole 735B gabbros are meltlike.
Figure F3 compares Hole 735B gabbroic samples with MORB melts and model melt compositions on Ca# - Mg# and CaO/Al2O3 - Mg# plots. Note that AII F.Z. basalts are plotted at the low-Ca# and low-CaO/Al2O3 end of the global MORB data array, which is consistent with the interpretation that these melts result from the lowest extent of melting associated with slowest spreading rate (Niu and Hékinian, 1997). The major conclusions from this comparison are as follows:
The most relevant question
here is whether the bulk hole composition resembles a basaltic melt. Hart et al.
(1999) conclude that the bulk composition of the upper 500 m of core is in fact
meltlike and interpret that the large downhole modal compositional variations
result from "local separation of melt and solids, but no large scale
removal of melts." Dick et al. (2000) conclude "a feature of the bulk
[hole] compositions is that they are all close to those of various primitive to
moderately differentiated basalts. Thus, the interpretation of most Hole 735B
gabbros as cumulates is based on trace rather than major elements." This
conclusion can be tested by a simple phase equilibrium analysis. The bulk hole
composition reconstructed by Dick et al. (2000) has FeOt = 7.31 wt% and MgO =
9.21 wt%, which gives Mg# = 0.714 (assuming 10% total Fe as Fe3+). If
this were indeed the Mg# of the parental melt, then the most primitive olivine
in these rocks must have Fo
0.883 (0.883-0.893, for Kd
= 0.33-0.27), but this is not observed. In fact, the most primitive olivine in
Hole 735B has Fo = 0.842 (troctolite) (Sample 176-735B-83R-7, 77-81 cm) (Ozawa
et al., 1991). The parental melt from which this olivine crystallized must have
Mg#
0.637 (0.590-0.637, for Kd
= 0.27-0.33). Using this Mg# value as a constraint and the AII F.Z. basalts as a
reference, the melt parental to Hole 735B would have the composition (circle
with plus in Fig. F3)
that contrasts with the reconstructed bulk composition (square with cross in
Fig. F3) of Hole
735B as shown in Table T9.
These differences and the phase equilibrium constraints presented above argue
strongly that bulk Hole 735B is not meltlike but is cumulate in nature: cumulate
clinopyroxene and olivine give the high Mg#, cumulate plagioclase and
clinopyroxene give the high Ca#, excess of clinopyroxene over plagioclase gives
the high CaO/Al2O3 ratio, and the removal of Fe-Ti-rich
melt gives the low FeOt and TiO2 in the bulk Hole 735B composition.
If the above estimated
parental melt is reasonable, then a significant amount of more differentiated
products is needed to balance the high Mg# (0.714) of the bulk hole composition.
The AII F.Z. basalts, which have an average Mg# = 0.561 ± 0.031, may be such
more differentiated products. Assuming this is the case, then 25%-45% of average
AII F.Z. basalt is needed to combine with 55%-75% of bulk Hole 735B gabbros to
give a melt (Mg#
0.637) parental to olivine of Fo = 0.842. It is noteworthy, however, that the
most primitive AII F.Z. basalt has Mg# = 0.616, which in fact corresponds to the
most primitive olivine of Fo = 0.842 ± 0.14 (for Kd = 0.30
± 0.03) in the troctolite.
The above deduced melt
with Mg#
0.637 is parental to all the lithologies in Hole 735B including the volcanics.
This parental melt is not the same as primary mantle melt (or primary magma)
conceived by petrologists. The latter, which is considered to be in equilibrium
with mantle residual olivine (Fo
0.89), must have Mg#
0.71. Assuming that the mantle melt passing the crustal/mantle boundary does
have Mg#
0.71, then "a considerable thickness of primitive cumulates complementary
to the lavas, dikes, and Hole 735B gabbros" (Dick et al., 2000) remains in
the deep crust to be revealed. We suggest that the mantle melt passing the
crustal/mantle boundary is already cooled and evolved. That is, the more
primitive cumulates (i.e., chromitite/dunite, troctolite, etc.) may be found in
the mantle. This is because below slow-spreading ridges, slow mantle upwelling
allows conductive cooling to the surface to extend to a greater depth against
the adiabat and hence leads to a thickened, cold thermal boundary layer atop the
mantle (Niu and Hékinian, 1997; Niu, 1997). Melt ascending/migrating through
this cold boundary layer must cool and crystallize out minerals on the liquidus,
which in this case are mostly chromite/Cr spinel, forsteritic olivine, and plus
anorthitic plagioclase (see Fig. F3E,
F3F) before
reaching the crust/mantle boundary (Niu, 1997; Niu et al., 1997). This latter
process may also be termed high-pressure (vs. crustal level) fractionation
(e.g., Bender et al., 1978; Elthon et al., 1982; Grove et al., 1992).
This suggests that bulk igneous crust is not equivalent to primary mantle melt in composition but is more evolved as the result of cooling/crystallization during ascent in the thermal boundary layer atop the mantle. We further add that the thin igneous crust at slow-spreading ridges (vs. thick crust at fast-spreading ridges) results from two actions: lower extent of melting (Niu and Hékinian, 1997) and the loss of melt mass during ascent through the thickened thermal boundary layer (Niu et al., 1997; Niu, 1997). This concept is not a speculation but a testable hypothesis, which is already supported by observations that exist in the literature (e.g., Cannat, 1996, and by Y. Niu's recent observations during the JR63 expedition at 15°20´N Mid-Atlantic Ridge.
About 40% of rock types recovered from Hole 735B are named with "oxide" as modifier (e.g., oxide gabbro, oxide gabbronorite, oxide olivine gabbro, disseminated oxide gabbro, etc.) (Dick, Natland, Miller, et al., 1999; Dick et al., 2000). These oxide-rich or oxide-bearing rock types take up ~17 vol% of the bulk Hole 735B (Dick et al., 2000). The oxides are mineralogically dominated by ilmenite and ilmenite-hematite-magnetite solid solutions (Table T6); thus, they are undoubtedly magmatic products. Although most of these oxide-rich/bearing rocks are also cumulate (Dick et al., 2000), they are nevertheless genetically closely associated with Fe-Ti-rich melt as a natural consequence of tholeiitic magma evolution (Wager and Brown, 1968). As noted in "Attainment of Phase Equilibrium," what is important here is the disequilibrium coexistence of Fe-Ti oxides with clinopyroxene, plagioclase, and olivine in most of these gabbroic rocks.
Figure F4B shows that during cooling and evolution of tholeiitic magmas, there is a significant Ti (also Fe) enrichment until the temperature of the system falls down to 1080°-1110°C (depending on fO2 and H2O content in the system), at which temperature Fe-Ti oxides appear on the liquidus and begin to crystallize. Therefore, production of Fe-Ti oxides is genetically associated with Fe-Ti-rich melt at relatively low temperatures. Figure F4C plots the modal abundances of Fe-Ti oxides in MS samples (see Table T1) against the liquidus temperatures calculated from olivine composition of these samples. As olivine is, to a first order, well in equilibrium with clinopyroxene, orthopyroxene, and plagioclase in these rocks (Fig. F1), it is these Fe-Ti oxides that are out of equilibrium. In other words, these oxides should not be in these rocks because the melts that produced olivine, plagioclase, and clinopyroxene were too hot to produce the oxides. This observation suggests that the Fe-Ti oxides in most of these rocks must be physically added to their host rocks below the liquidus temperatures of host minerals. As solid-state mass transport on such macroscopic scales is difficult, the oxide carrier must be silicate melt. There are two possibilities. One is that these oxides, particularly those in "disseminated" oxide-bearing rocks, represent crystallization of trapped Fe-Ti liquids that cooled and evolved in isolated pockets with the residual felsic liquid expelled during compaction and consumed locally during subsequent recrystallization. This could be the cause of mineral compositional variation (scatter) in a given crystal or between crystals of a given sample (see Fig. F1 and "Attainment of Phase Equilibrium"). The other possibility is that these oxide-rich rocks may represent a snapshot of melt migration (Dick et al., 2000). Fractional crystallization/crystal sorting results in residual melt enriched in Fe and Ti (Fig. F4B). These residual melts must be continuously expelled out of already formed crystal piles and transport/coalesce toward zones of low pressures. Continuous cooling during transport leaves oxides behind as traces of the passageways of melt transport. This latter interpretation is consistent with the observation that oxides are clearly enriched in shear zones of varying size, and synmagmatic deformation may play an important role (Dick et al., 2000)—a process envisaged as "differentiation by deformation" by Bowen (1928).
Within the bulk gabbroic rocks of Hole 735B are numerous small (a few millimeters to several centimeters thick) felsic veins (Dick, Natland, Miller, et al., 1999). Most veins are leucodiorite dominated by plagioclase plus some green amphiboles. Other lithologies include diorite, trodhjemite, and tonalite with variable amounts of amphibole, quartz, and biotite. Granitic veins were also recovered with up to 28% quartz and 24% alkali feldspars plus hornblende, biotite, pyroxene, Fe-Ti oxide, apatite, and zircon. Despite the small volume of these veins (~0.5% of the bulk of Hole 735B), their occurrence throughout the entire 1.5-km section of Hole 735B has important implications for processes of melt emplacement and evolution at slow-spreading ridges. While oxide-silicate liquid immiscibility (e.g., Natland et al., 1991) and Fe-Ti-rich melt reaction with wall rock (e.g., Dick et al., 2000) may be invoked to interpret the origin of the felsic veins, we believe that simple fractional crystallization is adequate. Figure F4A and F4B demonstrates that SiO2 enrichment in residual melt is the natural consequence of oxide removal/crystallization at a late stage of tholeiitic magma evolution. This is evident from the close association of oxides with felsic veins/veinlets throughout Hole 735B (Dick, Natland, Miller, et al., 1999).
There are also fine-grained equigranular microgabbros that crosscut the coarse-grained gabbros/olivine-bearing gabbros throughout the core (Dick, Natland, Miller, et al., 1999). Many of these microgabbros are small in size (1-5 cm thick) but span nearly the full compositional range of the host gabbros (Dick et al., 2000). Whereas some of the microgabbros show intrusive contact with the host, gradational, or sutured, contacts dominate. Most microgabbros exhibit shallow to moderate dips, but some, particularly those near the base of the hole, occur as vertical to subvertical "veins" or "veinlets." Dick et al. (2000) interpret this latter type as representing channels of melt transport through the crystallizing intrusions. Whereas this interpretation is sensible, the question that remains is why these microgabbros are finer grained than their host. It is possible that the finer grain size may simply result from the "quench effect." That is, the hot melt migrates in an already cooled (>30°C cooler?) host. This interpretation would suggest a significant time difference between the solidification of the host gabbro and the intrusion of the microgabbros. This time difference would imply different parental melts with different cooling and evolution histories. If this interpretation is correct, then the microgabbros and their host should have different compositions in their respective mineralogy. This is not observed. Figure F5 shows that, although scattered, mineral compositions of the microgabbros are broadly similar to those of the host gabbros. Neither group is necessarily more primitive than the other (straddle about the 1:1 line in Fig. F5) with the exception of Sample GS77-11. This suggests that there is a fairly good degree of compositional equilibrium between microgabbros and their host. Subsolidus equilibration may be invoked to explain such compositional equilibrium, but this cannot explain why such compositional equilibration (diffusion controlled) is not accompanied by crystal growth (recrystallization). Alternatively, both the microgabbros and their host gabbros may be different products of the same parental melts. Microgabbros may represent zones of melt expelled from the host gabbros as a result of compaction or filter pressing. The expelled melt must therefore be in equilibrium with both the minerals in the host (already formed coarse crystals) and the minerals crystallizing out of the melt (currently forming fine crystals). We can imagine that the zones of the expelled melt (1) must be present as crystal mushes and (2) must migrate/move in response to continuous compaction and filter-pressing. This motion/stirring (vs. static growth of the host minerals) may not impede nucleation but can prevent crystal growth of the system. This latter process may also contribute to the finer grain size of the microgabbros.
The foregoing discussion indicates that whereas trapped melt may be present, the bulk gabbroic rocks are not meltlike but are crystal cumulates. This is further demonstrated by modal and compositional control of minerals on bulk rock composition, particularly for compatible and less incompatible elements that prefer to stay in minerals over possible interstitial melts, if any. Figure F6, for example, shows that Sc abundance in whole-rock gabbros is largely controlled by clinopyroxene and plagioclase, the two major constituent minerals of gabbros. The deviation of the data trend away from the mixing lines in Figure F6A and F6B result from dilution of CaO- and Al2O3-poor phases (i.e., olivine, oxides, and, perhaps, orthopyroxene) in the rocks. This deviation disappears when CaO/Al2O3 (Fig. F6C) is used because these CaO- and Al2O3-poor phases (1) have low Sc and (2) do not affect bulk rock CaO/Al2O3 ratio. The hyperbolic curve in Figure F6C is a typical mixing line involving ratios. This mixing curve can be used to calculate relative mass proportions of clinopyroxene (i.e., cpx# = cpx/[cpx + plag]) and plagioclase (1 - cpx#) in the bulk rock from whole-rock CaO/Al2O3 ratio (RC/A) and Sc abundance as illustrated in Figure F6D. As bulk rock Sc abundance could be affected by ilmenite, RC/A would be a more reliable parameter for calculating cpx# and for estimating whole-rock clinopyroxene mass proportion for rocks without significant amounts of oxides, olivine, and orthopyroxene: cpx# = [1.3835 RC/A - 0.5163]/[1.2571 RC/A + 0.4587]. As plagioclase takes essentially no Mg and Fe, whole-rock Mg# would be an efficient measure for the composition of mafic minerals (clinopyroxene, olivine, and orthopyroxene), and thus the liquidus temperatures of the parental melt, provided that the whole-rock samples do not have excess cumulate oxides.
Figure F7A shows that there is no correlation between Mg# and cpx# even excluding samples with excess oxides (i.e., those with TiO2 > 0.6 wt%). This is expected, as Mg# reflects liquidus composition and temperature of the parental melt, but cpx# reflects modal heterogeneity on sampling scales. Whereas Ca# should be proportional to the liquidus temperature, as might be inferred from the correlation of An in plagioclase with Fo in olivine (Fig. F1), the weak correlation in Figure F7B demonstrates that Ca# in the whole rock is also controlled by mineral modes because Ca in the whole rock is dominated by clinopyroxene, whereas Na is dominated by plagioclase. The relationships among several compositional parameters in Figures F6 and F7 are important for understanding the abundances and distributions of trace elements in gabbroic lower crust. Figure F8 gives some examples, but a full discussion and account of the topic will be addressed elsewhere (Y. Niu et al., unpubl. data).
Figure F8 plots the abundances of representative elements against cpx# (left) and Mg# (right) (where both felsic veins/veinlets and N-MORB data are plotted for comparison). For gabbroic samples, for example, TiO2, V, and, to some extent, Y are controlled by both modes (cpx#) and mineral compositions (Mg#). Sr is largely controlled by cpx#. Highly compatible elements like Ni and Cr are controlled by mineral compositions (Mg#). Incompatible elements like Zr and Nb are not controlled by cpx# but are determined by the amount of excess Fe-Ti oxides in the rock (samples with TiO2 > 0.6 wt%), which is consistent with the fact that Fe-Ti oxides are genetically associated with trapped melt or late-stage Fe-Ti basaltic melts (Fig. F4).
As noted by Dick et al. (2000), there is no systematic downhole compositional change, as would be expected if the stratigraphic sequence resulted from differentiation of a single large magma body, nor is there a simple evolutionary sequence, as seen in layered intrusions like the Skaergaard. In fact, the igneous stratigraphy of Hole 735B is characterized by extremely small-scale chemical and textural variability with products of highly evolved melts (e.g., Fe-Ti oxides, ferrogabbro, felsic veins, etc.) occurring throughout the hole as enclaves of variable size within the less-evolved gabbroic lithologies (Fig. F9). Such a "composite" sequence cannot be fully explained by simple models of magma chamber processes (Sinton and Detrick, 1992) but requires processes of multiple melt injection on highly localized scales (e.g., as sills) and in situ cooling and crystallization (Dick et al., 2000; Kelemen et al., 1997; MacLeod and Yaouancq, 2000).
Dick et al. (2000) divided the Hole 735B stratigraphy into 12 polygenetic units considering detailed changes in principal lithologies and intrusive events. We show in Figure F9 downhole whole-rock compositional variations. Except for felsic veins/veinlets and rocks with excess Fe-Ti oxides (TiO2 > 0.6 wt%), the bulk Hole 735B shows two major units below and above the ~550-meters below seafloor (mbsf) discontinuity reflected by many compositional parameters such as Mg# (= Mg/[Mg + Fe]), Ca/[Ca + Na], Cr, Ni, TiO2, Zr, Y, and Zn. This discontinuity, which is coincident with a shear zone (Dick, Natland, Miller, et al., 1999), may in fact be a fault contact separating the two units. It is possible that the upper 550-m and the lower kilometer sections may be genetically unrelated. As these parameters largely reflect mineral compositional control (see Fig. F8 and discussion above), they thus record the temperature and compositions of liquids parental to the cumulate sequence. The melts parental to the upper ~550 m (excluding oxide-rich zones) are clearly more primitive and hotter than melts parental to the lower kilometer.
Using Mg#, Dick et al. (2000) separated the upper ~550 m into two chemical units and the lower ~1000 m into three chemical units, each being characterized by an upward trend of decreasing Mg#, with a sharp increase in Mg# at the beginning of each overlying unit, as representing some form of cyclic intrusion. Such cycles are not so obvious for the lower kilometer but are better developed for the upper ~550 m (Fig. F9). What is clear is the small but systematic upward decrease in Mg# (i.e., the liquidus temperature) and the corresponding changes of other parameters for units both above and below the ~550-m discontinuity. Note that Sr and Cu do not show such systematics because these two elements are not controlled by mineral compositions but by mineral modes (e.g., Sr is proportional to the plagioclase/clinopyroxene ratio and Cu is largely controlled by sulfide globules). Such downhole liquidus temperature variation is better reflected by the mineral compositional data as shown in Figure F10. In Figure F10, the liquidus temperature is calculated from olivine composition and represents the maximum values. It is obvious that the melts parental to Hole 735B cumulate sequence are rather evolved. To further illustrate this point, the Mg# of the parental melts in equilibrium with olivine is shown. The simple implication is that the parental melt to the bulk of Hole 735B is too evolved to be in equilibrium with residual mantle mineral assemblages. A significant amount of more refractory cumulate must be hidden as discussed above and by Dick et al. (2000).
