Below, we discuss the chemistry of the major silicate phases. Though microgabbros are included in Tables T1, T2, T3, T4, and T5, we generally describe only variations in the relatively coarse grained gabbros in the plots and discussions. We do this because the plots are already cluttered and variations in the microgabbros mimic those of the gabbros and show no overall coherent stratigraphic variations.
Although a few grains of true igneous potassium feldspar were found in rare granitic veins, close to 60% of the Hole 735B gabbros are made up of plagioclase. Plagioclase ranges in abundance up to 80% in some troctolites, with most gabbros lying within the range of 50%-65% and extending down to <20% in rare oxide gabbronorite and troctolitic gabbros (Dick, Natland, Miller, et al., 1999). It is present in every rock type examined, typically in near cotectic proportions with clinopyroxene and olivine (Bloomer et al., 1991). It is generally equigranular and relatively coarse grained with a subeuhedral to anhedral shape interlocking with clinopyroxene and olivine or subophitically enclosed by clinopyroxene. Rocks composed of near euhedral plagioclase in a matrix of clinopyroxene and olivine are rare. Figure F1A shows single-spot analyses and average feldspar compositions for 1656 feldspar grains plotted in the feldspar ternary (the few analyses of K-feldspar are not included here). With the exception of a few isolated Or-rich analyses, which are likely bad data that passed our preliminary screen, all the compositions define a very tight trend close to the anorthite-albite join far from saturation with respect to orthoclase. Whereas they range from An80 to An8, there is a sharp drop in abundance at An30, with most of the remaining feldspars found in felsic veins ranging from diorite through tonalite and trondhjemite. Plagioclase is typically very uniform in composition, with the average composition close to that of its core and a locally more sodic rim. The average standard deviation for anorthite content for analyses of individual grains (generally four to six points per grain) is only 1.9%. Although both normal and reversed zoning are present, the former is far more common. Locally large core to rim variations exist, with up to 10 mol% anorthite variations common. No systematic pattern has been found downhole for zoning, with large intragranular variations in feldspar compositions locally present downhole (Fig. F1B).
In general, plagioclase is quite fresh, a characteristic of Hole 735B gabbros, where total alteration rarely exceeds a few percent. Examination of the data in Table T1 will show the reader that whereas neoblast compositions in deformed gabbros may be identical to the primocryst compositions in many cases, in others they may be significantly more sodic than relict primocrysts. In part, this is due to reaction with late iron-rich fluids migrating through the gabbros, and in other cases, it is due to reaction with hydrothermal fluids that have locally metasomatized the gabbro. Distinguishing between these two origins is tricky. Examination of the P. Robinson data set, however, taken largely from the most altered gabbros in the hole, shows that secondary or altered primary feldspar is locally quite abundant where alteration is heavy. Since this paper is primarily concerned with igneous petrogenesis, we have attempted to screen out the most obviously altered and secondary feldspar compositions. The primary criterion used is a strongly bimodal feldspar composition distribution with a high-calcium plagioclase population lying close to the cotectic trend for plagioclase-olivine or plagioclase clinopyroxene crystallization and a second population lying between An0 and An30. Analyses were excluded if this second population lay far from the inferred igneous cotectic trend (with an absence of a second population of coexisting late-magmatic igneous pyroxene and oxides representing obvious late-stage melt infiltration) and in the presence of significant amounts of hydrothermal amphibole, chlorite, or quartz.
As can be seen in Figure F2, there is a good correlation between feldspar composition and rock type, with the most calcic feldspar (up to An84) in troctolites followed successively by olivine gabbros, disseminated oxide-olivine gabbro, gabbronorites, oxide gabbros, and finally the various felsic veins. An impressive feature of Hole 735B compared to layered intrusions is the very large range in feldspar compositions found from the hand-specimen to the outcrop scale. In many layered intrusions, variations of only 4 or 5 mol% anorthite may occur over 1000 m or more. In Hole 735B, it is possible to find nearly the full range of igneous feldspar composition in a single hand-specimen scale (Table T1). This reflects the incredible diversity of rock types at all levels in the Hole 735B section.
There are also significant differences in the feldspar population with depth. A strongly bimodal population is present in the upper 500 m (Bloomer et al., 1991; Dick, 1991a; Natland et al., 1991; Ozawa et al., 1991); a feature that is also observed in the whole-rock chemistry. This bimodal distribution disappears entirely in the lower 1000 m, however, with the distribution becoming strongly skewed to calcic feldspars in the range of An50-An65 in the bottom 500 m. It is noteworthy, however, that this does not simply reflect the decrease in abundance of oxide gabbros and gabbronorites, as the composition of feldspar in oxide gabbros and gabbronorites in the bottom 500 m is also more calcic than in the same rock types in the upper 500 m (Fig. F2). The lack of very sodic plagioclase compositions (<An30) in the upper 500 m, on the other hand, is not due to an absence of felsic veins, which constitute 1% of the Leg 118 core (Dick, 1991a) but is instead due to sampling bias. Whereas the Leg 118 igneous petrologists largely ignored the felsic veins, the reverse was true for the Leg 176 Scientific Party, where they were extensively (obsessively) sampled.
The anorthite content of average and spot plagioclase analyses of the Hole 735B gabbros is shown plotted against FeO in Figure F3. Here again, the strong bimodal distribution of anorthite content is clearly evident. Despite the large scatter in the data, however, there is also a strong overall increase in plagioclase FeO from troctolite to olivine gabbro, followed by a significant decrease in oxide gabbros, with gabbronorites and disseminated oxide olivine gabbro having somewhat intermediate compositions. Again, with considerable scatter of a few outlying points, a second trend of decreasing iron content with decreasing anorthite is evident for the felsic veins, offset to significantly higher iron than the gabbro trend. If this second trend is real, it might suggest that the mechanism of formation of the felsic veins was not simple fractionation of a ferrobasalt melt but could involve other processes such as liquid immiscibility.
Clinopyroxene is the second most abundant phase in the Hole 735B gabbros, ranging from a few percent in troctolite to close to 60% in a few gabbros, varying inversely with plagioclase abundance. It is generally present in near-cotectic proportions with plagioclase and olivine, however, with the large majority of Hole 735B gabbros having from 15% to 45% (Dick, Natland, Miller, et al., 1999). It is most often present as a coarse granular or subophitic phase with plagioclase and olivine but also commonly forms large oikocrysts, typically several centimeters or more in length, enclosing plagioclase and olivine chadocrysts. In addition it forms small selvages or reaction rims on the margins of plagioclase and olivine grains, where it is easily confused when in low birefringence orientations with relatively birefringent sections of hypersthene present in the same habit.
The major element composition of Hole 735B pyroxene is shown in the quadrilateral in Figure F4A along with the composition of enstatite and diopside from Atlantis II Fracture Zone residual mantle peridotites. The clinopyroxene from Hole 735B gabbros is largely augite, with a small but significant gap in composition between them and the field for the peridotites. Overall, given the large number of analyses, there is a fairly tight grouping with fewer than 20 of 861 analyses plotting away from the main cluster. Again, there is a broad trend corresponding to lithology, with the most magnesium and calcic augites present in the troctolites, followed by the olivine gabbros, disseminated oxide olivine gabbro, gabbronorites, and oxide gabbros, all trending overall to lower wollastonite contents. This trend broadly parallels the pyroxene solvus of Lindsley et al. (Fig. F4B), consistent with crystallization from a fractionating basalt liquid (Lindsley, 1983; Lindsley and Dixon, 1976). It should be noted that there is much broader overlap between the fields for different rock types than was originally found by K. Ozawa and co-workers in the upper 500 m of Hole 735B. In part, this could be due to somewhat different criteria in assigning rock names between the Leg 118 scientists and those from Leg 176, but we suggest, based on the distribution of plagioclase compositions in Figure F2, where oxide gabbros deeper in the hole have more calcic plagioclase on average than those in the upper 500 m, that the greater overlap is largely real and therefore reflects significant differences in petrogenesis of the rocks downhole. The scattered analyses lying intermediate between the main clinopyroxene and orthopyroxene trends could represent bad analyses, but without an opportunity to carefully examine the thin sections from which they were obtained, this would be difficult to prove one way or the other.
A systematic study of zoning in clinopyroxene in the Leg 176 section is shown if Figure F5, where core and rim compositions are plotted downhole. No discernible consistent pattern of clinopyroxene zoning, however, can be seen downhole. Clinopyroxene generally shows normal zoning with rims consistently more iron, sodium, and titanium rich than the cores. The extent of zoning is relatively limited for magnesium and iron, rarely amounting to more than 5 mol% of the iron-rich end-member. Reversed zoning is rare, identifiable in only a few oxide and olivine gabbros downhole. An exception is found for titanium, which is frequently lower on the rims of oxide-rich gabbros, even while the latter are more iron rich and sodic. This is easily explained, however, by the appearance of iron-titanium oxides on the liquidus. The latter would cause titanium in the liquid to rapidly drop with additional crystallization after reaching a peak prior to their appearance. Overall, the greatest zoning is seen for titanium, which commonly exceeds a factor of two in concentration from core to rim. This is consistent with the formation of late iron-titanium-rich interstitial liquid during crystallization of the gabbros and its migration by porous flow through the section.
Minor element (Cr, Al, Ti, and Na) concentrations in the Hole 735B gabbro clinopyroxenes define rough differentiation trends correlating with decreasing Mg# (Fig. F6). Chrome exhibits a striking strong exponential drop in abundance with decreasing Mg#. This corresponds to the presence of cumulus chrome spinel in the troctolites and the strong preference of chrome for pyroxene relative to melt with the onset of cotectic pyroxene crystallization. Alumina shows a strong linear decrease with decreasing Mg#, whereas sodium, if anything, exhibits a slight enrichment. The most interesting element is titanium, which, with decreasing pyroxene Mg#, first increases sharply, reflecting its preference for the melt during cotectic crystallization of olivine, plagioclase, and pyroxene, then abruptly decreases, reflecting the appearance of iron-titanium oxides on the liquidus with crystallization of the oxide gabbros. What is particularly interesting is that at any particular pyroxene Mg#, there is a factor of two range in typical (not outlier) titanium concentrations. Moreover, magmatic oxides are also found in greater than accessory amounts in gabbros with pyroxene Mg# much greater than 75—the point at which the overall trend suggests oxides first appeared on the liquidus. This latter oxide, the presence of oxide gabbros with relatively high Mg# clinopyroxene, and the dispersion of the titanium contents all argue for extensive postcumulus melt-rock interaction with reequilibration of crystals and trapped melt and precipitation of additional oxides both from trapped melt and from relatively iron-rich melts migrating through the cumulus crystal mass.
Orthopyroxene is ubiquitous as an accessory phase throughout the Hole 735B gabbros, appearing as thin selvages or reaction rims, generally between olivine and plagioclase and minor overgrowths on clinopyroxene. It is also present as a relatively coarse grained granular phase in rocks of intermediate composition, including both orthopyroxene-olivine gabbro and gabbronorites of various descriptions. It does not generally persist as a granular, presumably cumulus, crystallization phase to the more extreme iron-rich compositions found for many oxide gabbros. The composition field for orthopyroxene is shown in the pyroxene quadrilateral in Figure F4A. Orthopyroxene defines a very tight trend parallel to the pyroxene saturation surface of Lindsley and co-workers (Lindsley, 1983; Nabelek et al., 1987), with slightly increasing wollastonite content with decreasing enstatite (Mg#) content. This broad range in composition, from En86 to En43, was previously noted by Ozawa (Ozawa et al., 1991). What can be seen here, however, is that the trend passes through the infamous Skaergaard trend, where pigeonite is supposed to replace hypersthene as a cumulus phase. In the Skaergaard Intrusion, however, pigeonite is actually present as an intercumulus phase only in the marginal border group (S. Morse, pers. comm., 2001) and this replacement of hypersthene is actually better seen in the Bushveld Complex (G. Cawthorn, pers comm., 2000). Pigeonite is present in the Hole 735B gabbros, where it has been reported by Ozawa et al. (1991), but is largely absent in the lower kilometer of Hole 735B. Because of the difficulty in analysis of this latter phase, which generally shows coarse exsolution, there are only a couple of pigeonite analyses in Table T2. The scarcity of pigeonite in Hole 735B could reflect relatively low pressure crystallization of the Hole 735B gabbros in an ocean ridge environment and differences in melt composition relative to the rare pigeonite-bearing layered intrusions.
Perhaps the most significant aspect of the orthopyroxene distribution in Figure F4A is its very low wollastonite content and the tightness of the trend. The latter can in part be attributed to the steepness of the pyroxene solvus relative to clinopyroxene (see Fig. F4B) but also implies relatively low equilibration temperatures. This is not likely an analytical problem due to exclusion of clinopyroxene lamellae from the analysis, as several investigators specifically used techniques designed to include these. As seen from Figure F4, hypersthenes in Hess Deep gabbronorites have significantly higher wollastonite content yet were analyzed using the same techniques as many of the orthopyroxenes in Table T2 (Natland and Dick, 1996). Thus, the difference in orthopyroxene compositions here reflects a difference in equilibration temperature. The Hess Deep gabbronorites are generally fine-grained rocks, believed to have crystallized from the melt lens below the sheeted dikes at the East Pacific Rise (Natland and Dick, 1996). The lower equilibration temperature of the relatively coarse grained Hole 735B orthopyroxenes could reflect reequilibration during relatively large scale percolation of late-magmatic liquids through the cumulates or relatively slow cooling and massive subsolidus recrystallization. The former possibility is presented by the gabbronorites, which contain coarse granular orthopyroxene, believed to originally represent cumulus grains precipitated during initial crystallization of the gabbro. If rapid cooling occurred in the absence of the widespread percolation and reequilibration with late-magmatic liquids, one would expect that the orthopyroxene compositions would be similar to those in the Hess Deep gabbronorites.
Orthopyroxene shows very similar composition trends for aluminum, sodium, and titanium with Mg# as those previously described for clinopyroxene, so they will not be described in detail here. Unlike clinopyroxene, chrome shows no trend—probably because its concentration is very low and is likely close to the detection limit for most probes.
Olivine is typically present in the Hole 735B gabbros as rounded anhedral grains but varies from subhedral to amoeboid and locally may enclose plagioclase crystals in an oikocrystic or poikilitic habit. Its abundance varies from 0 to >50 vol% in a few rare examples, with the large majority of olivine-bearing gabbros having <20 vol% olivine, with 5% to 10% representing the most typical values. There is only a rough correlation of volume percent modal clinopyroxene with olivine abundance, which generally decreases with increasing clinopyroxene (Dick, Natland, Miller, et al., 1999).
Olivine composition shows a large continuous range in the Hole 735B gabbros ranging from <Fo30 to Fo84. Overall, there is a fair relationship with rock type, with the most magnesian olivine present in the troctolite microintrusions above 550 mbsf and becoming successively less magnesian from troctolitic gabbro, to olivine-gabbro, disseminated oxide olivine gabbro, gabbronorite, oxide gabbro, and, finally, oxide olivine gabbro. Overall, it is again noteworthy that there is more overlap among these fields for the entire Hole 735B section than was found higher in the hole for the Leg 118 section (e.g., Ozawa et al., 1991). Whereas nickel concentration is highly variable and there is no correlation with forsterite content, the limiting value of nickel decreases sharply with fayalite, consistent with the strong preference of nickel for olivine and its rapid depletion in the melt with fractional crystallization (Fig. F7). The broad range of nickel concentration, ranging down to the detection limit at all forsterite values, suggests that the concentration of nickel in olivine has been strongly affected by postcumulus processes and reequilibration with late melt compositions.
Brown hornblende is nearly ubiquitous in the Hole 735B gabbros but only rarely exceeds trace amounts (<1%). It is commonly present as an intergranular selvage between olivine, pyroxene, and plagioclase with much the same growth habit as orthopyroxene. It appears to have a particular affinity to oxides and is commonly found intergrown with or enclosing ilmenite and titanomagnetite. It also reaches its greatest abundance in the oxide-rich gabbros, where it may be present in more than trace amounts (1%-5%). The olivine gabbros, with a single exception, generally have significantly <1% brown hornblende, usually present in only trace amounts. Based on textural criteria, most of the brown hornblende appears to be igneous. However, there is clear petrographic evidence that it can grade into metamorphic amphibole, with dark reddish brown hornblende locally grading into ragged green amphibole. Brown hornblende exhibits only weak compositional patterns with a tendency toward lower aluminum, calcium, titanium, and chromium and somewhat higher potassium with decreasing Mg# (Fig. F8). There is also considerable downward scatter from the main cluster of aluminous titaniferrous amphibole compositions (10%-13% Al2O3 and 2%-4% TiO2), likely due to the inclusion of metamorphic amphibole compositions in the database.
Shown in Figure F9 are plots of plagioclase anorthite content vs. Mg# (100 x Mg/[Mg + Fe]) of olivine, orthopyroxene, clinopyroxene, and brown hornblende. Overall, there are excellent correlations between plagioclase and all the other primary silicate phases, consistent with crystallization along the olivine-plagioclase, olivine-plagioclase-clinopyroxene, and finally plagioclase-clinopyroxene-oxide ± olivine cotectics. There are significant kinks in the trends, reflecting the initial appearances of clinopyroxene and oxides on the liquidus during cumulus crystallization. This demonstrates that the primary control on rock composition was simple magmatic differentiation along a tholeiitic liquid line of descent. An important feature of these trends, however, is the very large scatter along most of them, with a 10 mol% or greater variation in plagioclase composition at any particular value of Mg# for olivine, clinopyroxene, and brown hornblende. This lies far outside analytical scatter, even when comparing analyses from many different laboratories. Only orthopyroxene exhibits a significantly tighter trend. This scatter could be due to the intrusion and crystallization of a range of basalt liquids representing different degrees of mantle melting (e.g., Meyer et al., 1989) or it could be due to postcumulus processes such as incomplete reequilibration of the crystal matrix with different liquids percolating through the crystal mush. Whereas the basalt glass compositions reported for the Atlantis II Fracture Zone (Dick, 1991a) have sufficiently large scatter in Ca/(Ca + Na) at constant Mg# to explain the scatter along the trends in Figure F9, this cannot explain the large scatter in TiO2 contents seen for pyroxene (Fig. F6) or the range of nickel values seen in Figure F7. Thus, we prefer the latter hypothesis that most of the scatter is due to reaction with late-magmatic liquids and not due to a broad spectrum of primary mantle melt compositions.
Shown in Figure F10 are Mg# covariation diagrams for clinopyroxene, olivine, orthopyroxene, and brown hornblende. Clear curvature can be seen in the trends for olivine with clinopyroxene and orthopyroxene, showing the complex controls on Fe-Mg partitioning between these phases. These diagrams again show excellent correlations similar to those seen in the previous figure, with the worst correlations existing for brown hornblende. The latter likely reflects the inclusion of metamorphic compositions in the hornblende data set as well as the very late magmatic character of the amphibole, which likely crystallized out of equilibrium with the bulk of the matrix mineralogy. The correlation between olivine and clinopyroxene shows the largest scatter (~10 mol% for any particular value of olivine or clinopyroxene Mg#, suggesting that these minerals preserve considerable local disequilibrium produced by postcumulus processes). By contrast, the tightest correlation exists between orthopyroxene and olivine, which likely reflects the ease of olivine reequilibration and the steep slope of the orthopyroxene solvus combined with its relatively late magmatic interstitial character.
Figure F11A and F11B are downhole plots of single-spot and average analyses of individual primary igneous mineral grains. Shown for comparison in Figure F11C is a similar plot of whole-rock Mg# reproduced from Dick et al. (2000). As can be seen, many of the characteristics of the igneous stratigraphy first identified from the whole-rock chemistry are clearly evident in the downhole mineral plots. These include the strongly bimodal chemistry of the gabbros in the upper 500 m of Hole 735B, the sparsity of ferrogabbro and gabbronorites downhole, and the disappearance of the bimodal chemical distribution at ~1000 mbsf. In addition, the strong downward iron and sodium enrichment trend from ~160 mbsf down to the major chemical discontinuity at 274 mbsf found by the Leg 118 scientists (Bloomer et al., 1991; Dick, 1991a; Natland et al., 1991; Ozawa et al., 1991) is clearly defined by the synthesis of mineral compositions. The sequence of five broad trends of upward iron and sodium enrichment in the olivine gabbros is also clearly seen. These latter features define four significant chemical discontinuities in the stratigraphy at 274, 528, 941, and 1299 mbsf, respectively. The origins of these iron enrichment trends in the olivine gabbros and the strong bimodality found in the oxide gabbros in the upper 500 m is the subject of considerable debate among the Leg 118 and Leg 176 scientists.
The extensive intercalation of extreme differentiates with the moderately primitive olivine gabbros in the upper 500 m of Hole 735B is generally considered to be the result of intrusion of late iron-titanium-rich differentiates into the olivine gabbros by percolation along shear zones and intrusion along fractures in the olivine gabbros during active deformation of the section (Bloomer et al., 1991; Dick, 1991a; Natland et al., 1991; Ozawa et al., 1991). There is, in particular, often a striking association between ferrogabbros and disseminated oxide olivine gabbro and high-temperature crystal-plastic and cataclastic deformation throughout the hole. The source of these liquids, given the strong drop in abundance of ferrogabbros in the lower kilometer of Hole 735B, would likely be the lowermost crust—either liquids produced during the crystallization of the Hole 735B olivine gabbros or from similar rocks nearby.
A simple mass balance calculation using the logged proportions of different lithologies and their average whole-rock compositions shows that whereas the bulk composition of the entire Hole 735B section is close to that of a moderately differentiated mid-ocean-ridge basalt (MORB) (Mg# = 69.2 wt% and TiO2 = 0.87 wt%), the bulk compositions of each of the 500-m increments are inconsistent with a single upwardly differentiating sequence as identified in some layered intrusions (Dick et al., 2000). The uppermost increment, for example, has the most primitive Mg# (71.4), but has high TiO2 of 1.41 wt%, inconsistent with a reasonable mantle melt composition. Whereas bulk Mg# does decrease upward for each 500-m sequence, bulk anorthite remains constant and bulk TiO2 increases nearly threefold, changes inconsistent for progressive residues of fractional crystallization. Throughout the core, there are numerous small patches of undeformed oxide-rich gabbro that appear to represent local puddling and crystallization of late melt in the gabbro, showing that such liquids were formed and segregated locally. Taken by itself, however, the olivine gabbro must have lost considerable late iron-titanium-rich liquid, as it lies far from the composition of any reasonable MORB liquid with respect to incompatible element concentrations. It is thus believed that the distribution of late-magmatic melts produced by crystallization of MORB magmas during intrusion and solidification has been strongly controlled by the ongoing deformation of the crust beneath the rift valley floor, characteristic of slow-spreading ridges (Bloomer et al., 1991; Dick, 1991a; Natland et al., 1991).
The tectonically controlled igneous differentiation of the lower ocean crust found in Hole 735B is strikingly different from the processes believed to have formed the large stratified layered intrusions and is relatively new to much of the igneous community. However, it is the initial igneous stratigraphy of the section, represented by the olivine gabbro stratigraphy, that is probably the most controversial feature of the igneous petrogenesis of Hole 735B. The five upward enrichment trends seen in the whole-rock chemistry and mineralogy of the olivine gabbros could be explained as representing a series of small upwardly differentiating intrusions or cycles of intrusions. As can be seen from inspection of Figure F11, many of these cycles overlap, with what appears to be the most extreme differentiates of an underlying sequence intercalated with olivine gabbros at the base of the next (e.g., the 941-mbsf discontinuity). This suggests that the discontinuities do not represent the boundaries between simple plugs that have differentiated in situ to form discrete intrusions. Rather, it would appear that there are four or five cycles of intrusion, each representing multiple injections of a related magma that are locally interspersed with screens of older rocks from a previous cycle. In detail, however, each of these discontinuities is different. Thus, we wish to consider each one separately, for as we will show, there is substantial evidence that each of the discontinuities has a different petrogenesis and they may not all represent simple breaks in the stratigraphy produced by the intrusion of new magma batches.
The 274-mbsf discontinuity lies at the base of the unique downward iron enrichment trend in lithologic Units III and IV disseminated oxide olivine gabbro and oxide gabbros. Below it is the top of an upward iron enrichment trend running through the underlying Unit V olivine gabbro. It is shown in more detail in an expanded downhole plot in Figure F12. The Unit III and IV ferrogabbros above the boundary contain a foliation that flattens systematically with depth, defining a simple zone of downward-increasing shear. Such shear zones are well known as "ductile faults" (Ramsay, 1976; Ramsay and Graham, 1970) and represent the progressive flattening of a foliation with increasing strain from parallel to the principal stress direction to parallel to the plane of faulting. It is common to find an abrupt discontinuity at the base of these shear zones, as is the case here, where fault displacement has juxtaposed rock from where the shear zone nucleated over undeformed rock. Recrystallized rocks, due to reduction of grain size, are normally relatively impermeable compared to coarse-grained rocks. However, where shear is active, increased permeability occurs due to the formation of microcracks and grain boundary sliding. Thus, an active ductile fault can localize fluid flow, including late-magmatic melts in a compacting gabbro massif. Such flow would be greatest where strain is largest. Thus, the downward iron enrichment trend is consistent with increasing melt-flux and melt-rock reactions between deforming olivine gabbro protoliths and late iron-rich melts along a shear zone. Significantly, it can be seen that near the top of this series of oxide gabbros and in the middle of the sequence at around 225 mbsf there are undeformed screens or shear polyhedra of olivine gabbro that lie on the upper olivine gabbro mineral composition trend. These olivine gabbros contain extension cracks filled with undeformed oxide-rich gabbro, demonstrating the presence of iron-rich melts during deformation in the shear zone. Thus, the downward iron enrichment trend is likely due to migration late iron-rich melts and melt-rock reaction along a large composite shear zone (Dick, 1991a; Dick et al., 1992).
With the synthesis of all the available data, it now appears that at least two additional downward iron-enrichment trends can be seen in the upper 500 m of Hole 735B. These are defined largely by the gabbronorites intercalated with older olivine gabbros from 0 to 136 mbsf and by ferrogabbros intercalated with older olivine gabbros from 356 to 450 mbsf (Fig. F11). This suggests that there was considerable imbrication of the faulting and formation of shear zones in the olivine gabbros controlling late melt distributions in this section, with a significant variation in the intruding melt compositions.
A striking feature of the olivine gabbro underlying the 274-mbsf discontinuity is a sharp swing and broad spread of plagioclase and clinopyroxene compositions right at the top of the sequence adjacent to the contact with the ferrogabbro sequence. By contrast, olivine, the most easily reequilibrated phase, shows more uniform composition below the contact and a small but steady increase in Mg# from 319 mbsf (Fig. F12). The rest of this lower olivine gabbro sequence can be interpreted as lying on a simple extension of the upper olivine gabbro trend. Moreover, it is striking that these two sequences are on average more magnesian and calcic than the lower three olivine gabbro sequences. This raises the possibility that these two olivine gabbro sequences actually represent a single intrusive cycle disrupted by cross-intrusion of the ferrogabbro sequence along a shear zone. Plagioclase and clinopyroxene are the two phases most resistant to rapid reequilibration with exotic magmas and thus can preserve the early magmatic record where more rapidly equilibrated phases such as olivine cannot. Thus, the broad range of plagioclase and clinopyroxene compositions for 100 m below the 274-mbsf discontinuity argues for equilibration with a large range of liquids rather than a single upwardly differentiating magma. This and the preservation of calcic-plagioclase and magnesian-clinopyroxene at the top of the lower olivine gabbro is most simply explained due to melt-rock reaction driven by percolation of a small quantity of melt from the overlying ferrogabbro intrusion into the underlying olivine gabbro. This, then, suggests that originally both the upper and lower olivine gabbros above the 528-mbsf discontinuity were part of a single olivine gabbro sequence.
The alternative hypothesis, that the upper two olivine gabbro sequences represent separate upwardly differentiated sequences, is deemed inadequate here as it fails to explain the huge spreads in plagioclase anorthite and clinopyroxene Mg# immediately below the contact and the contrasting small range in olivine forsterite content. Moreover, the large offset of forsterite at the 274-mbsf contact from the overlying olivine gabbro sequence compared to anorthite and clinopyroxene Mg# is also inconsistent with this being a simple intrusive contact, as discussed elsewhere in this paper. The apparent upward iron enrichment trend of the olivine gabbros in Unit V olivine gabbro sequence is explained then as due to reequilibration with iron-rich melts intruded along the overlying shear zone and disruption of the sequence below 400 mbsf by reaction with the primitive melts that produced the cross-intruding troctolites there.
The lowermost gabbros in the ferrogabbro sequence are mylonites representing subsolidus deformation. However, our hypothesis requires significant chemical exchange between the underlying olivine gabbro and the overlying ferrogabbro sequence. The lower olivine gabbro shows significant iron enrichment down to ~36 m below the 274-mbsf discontinuity. As solid-state diffusion is far too slow to account for this by subsolidus reequilibration of the lower sequence with the overlying ferrogabbro, this requires significant permeability in the lower olivine gabbro at the time of intrusion of the ferrogabbro. This may have occurred because of remelting of the olivine gabbro during faulting accompanying intrusion of the ferrogabbro up the shear zone or because the underlying olivine gabbro was still partially molten at the time of faulting, or some combination of the two.
Shown in Figure F13 is an expanded view of the 528-mbsf discontinuity. We believe that this represents the most significant discontinuity in the stratigraphy of the Hole 735B olivine gabbros, with an average offset of some 13 mol% anorthite and forsterite content between the upper two olivine gabbro sequences and the lower three at this point. It should be noted that the olivine gabbro at the base of the upper olivine gabbro sequence is cross-intruded by an unrelated series of fine- to coarse-grained troctolites. Close inspection of the clinopyroxene profile in Figure F13, however, shows that a case can be made that the contact is actually gradational over some 30 m, with clinopyroxene Mg# and plagioclase anorthite content increasing sharply upward toward the boundary over this interval.
At the 941-mbsf discontinuity there are significant offsets of the trends for plagioclase anorthite as well as clinopyroxene and orthopyroxene Mg#, with generally more calcic plagioclase and magnesian pyroxene in the overlying olivine gabbro (Fig. F11). There is a somewhat smaller, more irregular offset in forsterite (as is also the case for the 1299-mbsf discontinuity). A potentially important feature of this boundary is that there is a series of gabbronorites and oxide gabbros that appear to lie on an upward extension of the lower olivine gabbro trend into the upper olivine gabbro sequence at this point. This raises the interesting question of whether or not these represent intrusion of late differentiates related to the lower olivine gabbro sequence into the overlying gabbros produced.
An expanded view of the 941-mbsf discontinuity is shown in Figure F14. There is a major high-temperature reverse-sense crystal-plastic shear zone at this horizon (Dick, Natland, Miller, et al., 1999), which extends from 930 to 960 mbsf, with a very sharp rise in intensity of deformation at 945 mbsf just below where we have drawn the 941-mbsf discontinuity. The shear zone ends abruptly at 960 mbsf, where the rock passes from mylonite into undeformed olivine gabbro over <10 cm. We analyzed gabbro on each side of this lower contact and found no change in mineral chemistry across it. The question arises as to whether there is any relationship between this 941-mbsf discontinuity and the shear zone. Despite the proximity of some brittle faults, none of the other chemical discontinuities we have identified in the hole correlate directly with the location of a fault or a major crystal-plastic shear zone. However, there is a strong association between crystal-plastic deformation and oxide gabbros through much of the hole, with abundant evidence that deformation was localized in these in the presence of melt (Dick, 1991a; Dick et al., 2000; Natland and Dick, 2001). Thus, it is possible that late-magmatic liquids concentrated at the top of the olivine gabbro, represented by the oxide gabbros and gabbronorites, may have localized deformation here by reducing the strength of rock due to liquid pore pressure. However, the zone of most intense deformation, and hence the region of the greatest strain localization, lies within the olivine gabbros, not the oxide gabbros and gabbronorites below the 941-mbsf discontinuity. Thus, whereas there is a coincidence here of strain localization and a major cryptic chemical boundary, in detail it is not convincing that this is more than coincidental.
The 1321-mbsf chemical discontinuity, shown in Figure F15, appears to be a relatively minor disruption of the overall upward trend in the olivine gabbros, with the olivine composition profile passing transparently through it. Pyroxene seems to show the greatest disruption, whereas plagioclase shows a smaller discontinuity than at the other boundaries we have discussed. However, the 1321-mbsf discontinuity does coincide with a zone of microgabbro intrusion with accompanying evidence of local melt-rock reaction with the olivine gabbro, producing a variety of oxide gabbros in the wall rock. For this reason, we have included in Figure F15 the mineral chemistry of the late microgabbro microintrusions that we have excluded elsewhere (as they show no vertical stratigraphic coherence and add unnecessary detail to a complex stratigraphy). This was one of two horizons selected for detailed study of the microgabbros and their influence on the wall-rock chemistry by Dick and co-workers. At this point this would seem to be an equally good explanation for this discontinuity as a boundary between two separate intrusive cycles in the olivine gabbros. We note that only slightly deeper in the core there is a second such reaction zone where the products of reaction are somewhat different and include oxide clinopyroxenite, websterite, and gabbronorite. Olivine gabbro on either side of this intrusion appears unaffected, and we suspect that the differences between these two zones of microintrusions reflect differences in the chemistry of the intruding melts and the degree of solidification of the wall rock at the time of intrusion.