Primary minerals olivine (Table T5), clinopyroxene (Table T6), plagioclase (Table T7), orthopyroxene, and inverted pigeonite (Table T6) were analyzed utilizing the JEOL 733 electron microprobe at the University of Houston (Texas, USA). Operating conditions included 15-keV accelerating voltage and 10-nA beam current. A focused 1-µm beam was used to analyze olivine and plagioclase. A broad 20-µm beam was used to analyze clinopyroxene because of exsolution. Both cores and rims were analyzed in each thin section. Zoning was generally absent in all major phases, but a subset of clinopyroxene and plagioclase showed zoning or intrasample heterogeneity within cores. Olivine core and rim analyses differences were negligible in all samples.
Of the 137 thin sections used in this study, 100 contain olivine. Olivine grains are typically anhedral to subhedral, occurring either as individual crystals or in clusters. In some thin sections the olivine grains are stringlike in appearance, whereas in others olivine shows a poikilitic relationship with plagioclase or clinopyroxene.
Olivine mineral chemistry (Fig. F27) varies from Fo78 to Fo36, with a gap between 40 and 46 (this gap was also noted by Thy, Chap. 2, this volume). Thy (Chap. 2, this volume) reported an olivine Fo range from 76.4 to 31.6 and suggested that the compositional ranges in Hole 1105A may not be similar to the Hole 735B core. A similar range of Fo84 to Fo30 was observed in nearby Hole 735B (Dick et al., 2002), except that somewhat more primitive troctolites and olivine gabbros were observed in the core analyzed. This is not unexpected when the differences in cored intervals are considered (143 vs. 1500 m). MnO content in olivine shows a systematic nearly linear increase from ~0.3 to 1.1 wt% with decreasing Fo mod%. MnO behaves incompatibly and increases in both olivine and clinopyroxene as the liquid fractionates and becomes enriched. NiO, on the other hand, exponentially decreases from ~0.12 to 0.01 wt% with decreasing forsterite. NiO is highly compatible in olivine and is rapidly depleted with increasing fractionation within the olivine gabbro. Both show systematic relationships: MnO-Fo R = 0.99 and NiO-Fo R = 0.94. The trend defined by NiO is somewhat broader than that of MnO and may reflect the effect of variable precision of the Ni analysis and postcumulus processes on the nonlinear fractionation trend. Zoning of olivine was not detected in the samples analyzed, a likely consequence of rapid diffusion and reequilibration rates in olivine when compared with clinopyroxene and plagioclase.
Augite grains were analyzed in all 137 thin sections. The studied grains have a wide range of Mg#s from 50.86 to 86.9 (average = 73.59). Thy (Chap. 2, this volume) reported a similar range of Mg#s from 48.4 to 83.7 for gabbroic rocks from Hole 1105A. Mg# also shows a wide range (42.1–89.3) in the rocks from the nearby Hole 735B (Dick et al., 2002). Cores and rims of each clinopyroxene grain analyzed were examined. The vast majority of high-Ca clinopyroxene grains lack major element zoning, although compositional heterogeneity of cores in a single thin section was observed in a number of samples. Dick et al. (2002) did not report any consistent pattern in clinopyroxene zoning in the rocks of Hole 735B. In this study, clinopyroxenes were defined as Type 1 and Type 2 for graphing purposes:
Type 2 cores are most commonly restricted to a small part of the thin section and may be the cumulate residue of small melt flow channels or localized trapped melt. Similarly, rims of the cores of Type 1 clinopyroxene can be similar in composition to Type 2 cores in the same rock. In general, Type 2 augites are more fractionated than Type 1, and the zonations are normal, always to more fractionated melt compositions. Figure F28 shows an example of a Type 1 clinopyroxene core rimmed by a normal zonation. It should be emphasized that most of the section examined lacked evidence for this type of zoning, although it was documented locally in certain rocks. Zones of heterogeneous augite cores within single thin sections typically are lower in Mg# than the dominant parts of the thin sections, perhaps indicating narrow porous zones of fractionated melt or inhomogeneous regions of trapped melt. When compared to the structure log for Hole 1105A, the samples that show two types of augite Mg# are commonly close to zones of deformation, possibly suggesting syntectonic melt flow. This may lend some support to the hypothesis of Dick et al. (2002) that melt expulsion is associated with melt-laden shear zones, although it is difficult to say that this is a general condition in the cumulate section.
Figure F29A–F29E depicts various elemental variation diagrams for high-Ca clinopyroxene (augite) as a function of Mg# with rock types and Type 1 or 2 cores distinguished by different symbols. Mg#s in clinopyroxene range from 50.86 to 86.9 (average = 73.59). In general, zoning or heterogeneous cores within single thin sections show typically lower Mg#s than dominant parts of the thin section. Thus, Type 2 cores usually plot in the more fractionated portion of each diagram. Al2O3 contents of clinopyroxene (Fig. F29A) decrease linearly with decreasing Mg# from olivine gabbros to oxide-rich gabbroic rocks. The variation in Al2O3 ranges from 3.77 to 1.24 wt% in the bulk of the samples with a range of 3.44 to 0.95 wt% in heterogeneous regions of single thin sections. Overall, the concentration of Al2O3 in pyroxene is relatively low, consistent with low-pressure crustal differentiation of tholeiitic basalts. Al2O3 contents in clinopyroxene decrease linearly with decreasing Mg# from olivine gabbros to oxide-rich gabbros. Al2O3 contents in Ca-rich pyroxenes are higher than those in the Ca-poor pyroxenes, although very little data are available for Ca-poor pyroxenes. No noticeable overlap of Mg# is seen between Ca-rich pyroxene and Ca-poor pyroxene. This feature is also characteristic of both Bushveld and Skaergaard pyroxenes (Atkins, 1969; Brown, 1957). The trend above of Al2O3 is also seen in the octahedral and tetrahedral Al in the pyroxene structure. The amount of Al in both octahedral and tetrahedral positions is greater in the Ca-rich pyroxene than in the coexisting Ca-poor pyroxene; data are very limited for Ca-poor pyroxene. Hole 735B gabbroic rocks show a wider range of Al2O3 compared to the tight trend exhibited by the gabbroic rocks from Hole 1105A. The trend of decreasing Al2O3 would be consistent with decreasing Al2O3 in MORB melts caused by fractionation of plagioclase.
When the clinopyroxenes are plotted in the quadrilateral diagram showing their end-members, the samples plot very close to the Skaergaard and Bushveld trends (Fig. F30A, F30B). The oxide-free samples have higher En%, whereas the oxide gabbro and olivine gabbros have the lowest En%. The low-Ca pyroxenes are below the pyroxene solvus line, indicating that most of the low-Ca pyroxenes are inverted pigeonites or that measurements were of exsolved lamellae of orthopyroxene in clinopyroxene. A comparison of the orthopyroxenes from this study with the available orthopyroxenes from Hole 1105A and those from Hole 735B (Dick et al., 2002) reveals that the last orthopyroxenes from all the holes crystallize at En% = 43. The quadrilateral plot shows that the trend of pyroxenes from Hole 735B is similar to the trend shown by the samples from this study. Although the trend is close to the Skaergaard trend, they are not parallel (e.g., Thy, Chap. 2, this volume; Dick et al., 2002). The trends from Holes 1105A and 735B have gentler slopes than the Skaergaard trend. When plotted in the diagram of Campbell and Nolan (1974) that shows the pyroxene minima of different intrusions (Fig. F30A), the overall trend of pyroxenes from Hole 1105A is slightly different from that of the Skaergaard intrusion. The Wo content of the pyroxenes associated with oxide gabbros is higher than the pyroxenes defined by the Skaergaard or Bushveld trends. The pyroxene minimum, defined as the point where Ca-rich pyroxene is in equilibrium with the last or one of the last Ca-poor pyroxenes, is different in the current study from both the Skaergaard and the Bushveld intrusions. The pyroxene minima from Hole 1105A and 735B gabbroic rocks are at higher En% than Skaergaard and Bushveld intrusions. Instead of being prominent as in the Skaergaard and Bushveld trends, these rocks have very gentle and almost no detectable inflection, similar to that of the Palisades sill reported by Walker et al. (1973). Lindsley and Munoz (1969) suggested that the position of the pyroxene minimum depends on the physicochemical condition of the melt. Clinopyroxene En% extends to 34% and has simultaneous olivine crystallization. Olivine continues to crystallize all through the crystallization series except for a noticeable gap around En% = 39%–38%. Olivine continues to crystallize even after low-Ca pyroxene has ceased to crystallize. The apparent gap in crystallization is not seen in Hole 735B olivine, but it was reported by Thy (Chap. 2, this volume) in Hole 1105A. The Fe-rich olivines found in the uppermost 500 m of Hole 735B and those in Hole 1105A are comparable (Fig. F30B) and strengthen the correlation between the two oxide units previously proposed. Fe-rich olivines are, in turn, absent in the next 1000 m of Hole 735B.
Cr2O3 abundances (Fig. F29B) in clinopyroxene decrease exponentially with decreasing Mg# from 0.50 wt% to below detection limits. The sharpest decreases occur in clinopyroxenes with Mg# between 86.9 and 75. In oxide gabbros and olivine gabbros, Cr2O3 is generally <0.1 wt%. Cr is highly compatible in clinopyroxene and rapidly depleted during the early stages of fractionation. One unusual sample (179-1105A-19R-3, 75–80 cm) that was classified as oxide bearing plots with high-chromium oxide-free gabbro and olivine gabbroic rocks, whereas it would be expected to be plot with the oxide-bearing group at lower Cr2O3. Closer examination of the sample shows that it is heterogeneous with respect to both clinopyroxene and plagioclase; clinopyroxene Mg# varies between 77.6 and 70.2 and An content of plagioclase varies between 64.5 and 45.1. The oxide grains visible in the sample could be from a late-stage melt that was transported through the sample.
The MnO content of clinopyroxene (Fig. F29C), like that within olivine, increases systematically with decreasing Mg#. MnO ranges from 0.15 to 0.64 wt% in homogeneous sample suites and from 0.22 to 0.79 wt% in heterogeneous samples. MnO is an incompatible element during silicate fractionation. Oxide gabbroic rocks show the greatest extent of MnO enrichment. TiO2 also behaves as an incompatible element in basaltic melts until Ti-rich oxides precipitate. TiO2 contents of clinopyroxene (Fig. F29D) increase from ~0.37 to 1.10 wt% with decreasing Mg# until Mg# = ~73. At this point, TiO2 in clinopyroxene reaches a maximum and begins to descend. This inflection point corresponds to the point at which oxide-bearing gabbroic rocks and oxide gabbros are encountered with decreasing Mg#. This point acts as a lithologic divider between the "oxide-free" and "oxide-bearing" gabbro and olivine gabbro. Anomalous samples, such as 179-1105A-19R-3, 75–80 cm, which plots on the oxide-free domain but contains oxides in its modes, typify rocks with mixed heritage. Generally, inspection of these thin sections and mineral analysis show that the samples are heterogeneous in clinopyroxene and plagioclase compositions and have wide ranges of composition, suggesting possible allochthonous melt infiltration.
The TiO2 trend observed in clinopyroxene is very similar to a typical tholeiitic basalt fractionation trend; however, fractionation of basalts (e.g., modeled Atlantis II basalts or Galapagos basalts) lead to peak TiO2 followed by depletion trends at Mg#s of ~25–30 in the melt before oxides initiate crystallization. The Mg#s of clinopyroxenes in equilibrium with this peak in the melt would be much lower than when the appearance of oxides is observed in Hole 1105A data. The appearance would be in the Mg# range of 52–59 rather than 70–73, as observed. Thus, the peak in clinopyroxene TiO2 seems to be displaced by ~20 units if Ti melt evolution trends observed in natural basalts caused the inflection. Clearly, the peak Fe-Ti-rich melt would not be in equilibrium with the clinopyroxenes with these higher Mg#s, possibly suggesting an early onset of oxide fractionation that led to the depletion of TiO2 in clinopyroxene, unrelated to typical PFX trends. This roughly approximated Mg# threshold of ~70–73 in clinopyroxene is essentially the same Fe and Ti enrichment and Si depletion point observed in the bulk rock trends (Fig. F14). Similar trends have also been observed in rocks from Hole 735B (by Ozawa et al., 1991; Dick et al., 2002) and Hole 1105A (by Thy, Chap. 2, this volume). Other authors have interpreted these trends to result from mixtures of primitive olivine gabbros and Fe-Ti-rich interstitial melts.
Another possibility for these trends is that they result from coupled substitution in clinopyroxene and compositional effects on partitioning behavior as the melt evolves. At first glance this seems unlikely in that the dominant controls on distribution coefficients are overall relatively small in terms of variations in major element concentrations (Blundy and Wood, 1994; Wood and Blundy, 1997). Hill et al. (2000) reported the influence of Ca-Tschermaks (calcium Tschermaks or CaTs) content of clinopyroxene on the partitioning of trace elements between clinopyroxene and melt. Al exists in pyroxene both in the octahedral and tetrahedral valence states. In a Ca-Tschermak molecule, Al exists in both the octahedral and the tetrahedral form. Lundstrom et al. (1998) demonstrated positive dependency of clinopyroxene liquid partition coefficients of high-field-strength cations Ti, Zr, Hf, Nb, and Ta on the Al(IV) content of clinopyroxene. Hill et al. (2000), with data from their studies, further confirmed the trend. Figure F31A shows the already established relationship between DTi and Al(IV), as shown in Lundstrom et al. (1998). This graph precisely suggests that as Al(IV) decreases DTi also decreases, and so does the ease with which Ti can enter the clinopyroxene structure. Figure F31B is an enlarged view of Figure F31A. The data are from Ray et al. (1983). At lower concentrations of Al(IV) in clinopyroxene the relationship between the two variables holds and DTi decreases with decreasing Al(IV). The concentrations of Al(IV) in clinopyroxene as shown in Figure F31B are included in the range of the data from Holes 1105A and 735B. Figure F32A shows the relationship between Al(IV) and TiO2 for clinopyroxenes from the study area. The graph shows a kink, with TiO2 increasing or variable with decreasing Al(IV) when it is greater than ~0.06 and TiO2 decreasing with decreasing Al(IV) below ~0.06. Figure F32B shows the relationship between Al(IV) and Mg# for clinopyroxene from the study area. Al(IV) decreases systematically with Mg#. Figure F32B shows an Mg# of ~73–75 separates the oxide-free from oxide-bearing samples, although the range is broad due to some scatter in the data. The correlation between the TiO2 content and trends in clinopyroxene and the appearance of opaque oxides may suggest that melt was saturated with oxides and evolving in contact with relatively primitive clinopyroxene. Alternatively, the TiO2 content and trends in clinopyroxene may indicate that variable Kd may play a role in clinopyroxene as a function of evolving melt composition. Walker et al. (1973), in their study on the pyroxenes of the Palisades sill, New Jersey, reported similar behavior of Ti in clinopyroxene with respect to clinopyroxene Mg#, and we observed this behavior in all suites of gabbroic rocks studied. Walker et al. indicated that the decrease in Ti in pyroxene appeared to correspond with the reduction in the Ti content in the liquid at the same time that Ti was being incorporated in ilmenite and titanomagnetite.
Apart from the complexities of coupled substitution in clinopyroxene, the apparent systematic relationships between Mg# of clinopyroxene and the appearance of oxides may imply early crystallization of Fe-Ti opaque oxides and the possible operation of boundary layer fractionation and in situ crystallization and fractionation processes (Casey and Karson, 1981; Langmuir, 1989; Nielson and DeLong, 1992).
Likewise, Na2O in clinopyroxene shows a similar but somewhat less pronounced trend as TiO2. Na2O abundances are variable but elevated, on average between 0.32 and 0.54 wt%, with decreasing Mg#s in primitive olivine gabbros. Na2O peaks at Mg# = 70 and then generally decreases as Mg# decreases to <70. The compositions are most variable in single thin sections of oxide gabbros where zoned rims or heterogeneous clinopyroxene populations tend to be the most depleted in Na2O, extending to 0.25 wt% in evolved samples (Mg# = 50–65). Ozawa et al. (1991) noted an increase in Na2O with increasing grain size and interpreted this to indicate both magma composition and subliquidus postcumulus growth and reaction with trapped melt. It, however, is not clear how these processes could lead to reductions in Na2O content in clinopyroxene, as trapped melt would tend to increase in Na2O content as crystallization advances. Natland et al. (1991) also pointed out a possibility that immiscible Fe-rich melts will have low Na2O contents.
Orthopyroxene is present locally in some samples as an accessory phase (see also Thy, Chap. 2, this volume). It usually occurs as a granular phase in highly evolved rocks. The grains are subrounded to rectangular in habit and are pleochroic. It also occurs as thin reaction rims, generally between olivine and plagioclase. It occurs extensively as an exsolved phase in clinopyroxene. The exsolution lamellae are of different types. Some are very fine, and blebs of orthopyroxene are also visible. In rare instances, inverted pigeonite is observed replacing orthopyroxene or as an intercumulus phase. Typical herringbone twinning is visible in inverted pigeonites. Again, inverted pigeonite is found in oxide-rich rocks. Both orthopyroxene and pigeonite occur in Unit I, Subunits IIA and IIB, and Unit IV, where local oxide abundances are high.
Plagioclase varies in composition from An67Ab32 to An14Ab84. Thy (Chap. 2, this volume) showed slightly higher anorthite contents up to An70. Increases in albite content are coupled with systematic increases in K2O (Fig. F33) with a range of Or0.18 –2.85. Around An% (30%–35%), the slope suddenly changes (i.e., a small change in An% shows a greater change in K2O). This trend is also seen in plagioclase from Hole 735B. The slope change takes place as oxides become more prevalent in the mode and the rate of increase in K2O in plagioclase accelerates. Plagioclase compositions show the widest extent of variability in single thin sections, and zoned rims are generally more abundant, especially in more fractionated oxide gabbros. Overall, however, the extent of major element zoning in most samples is low. The lowest An compositions (<An30) occur in felsic veins, which are not part of the cumulate assemblage but probably formed from late-stage melts migrating through the local cumulate pile. Also, the composition of the plagioclase correlates broadly to the lithology of the rock. Anorthite content decreases with progress from "oxide-free" gabbro and olivine gabbro to "oxide-bearing" and "oxide" gabbro and olivine gabbro. Plagioclase compositions are usually very uniform with some local heterogeneity. Some samples show multiple domains of plagioclase, and those samples are the same samples that show multiple domains of clinopyroxene. As previously stated, these samples were probable pathways for melt migration. As pointed out by Ozawa et al. (1991), the plagioclase in oxide-rich gabbroic rocks has a very dirty appearance as a result of the presence of very fine exsolved lamellae of opaque iron oxides. The exsolution lamellae are very thin and mostly detectable in backscatter images, suggesting that the plagioclase in the oxide gabbro and olivine gabbros was Fe rich and they exsolved the Fe as an oxide as cooling proceeded. FeO ranges from 0.1 to 0.4 wt% in Hole 1105A samples. A poorly defined trend of increasing FeO with decreasing An% is visible. This increasing FeO supports the fact that the melt became progressively Fe rich with fractionation. Dick et al. (2002) reported decreasing FeO in plagioclase in the oxide gabbros and also described a second trend of decreasing FeO starting at a higher Fe content and lower An (<20%) content. This trend is not visible in the data from Hole 1105A, and it does not appear to be present in the data from the upper 500 m of Hole 735B.
Thy (Chap. 2, this volume) showed that the Fe-Ti oxide minerals in the oxide gabbros and oxide olivine gabbros of Hole 1105A are granular intergrowths with a composition between magnetite and titanomagnetite. Equilibrium temperatures were calculated and reside between 850° and 675°C, indicating more extensive subsolidus reequilibration than in silicate phases, which generally show more elevated equilibrium temperatures (700°–1100°C). Our ilmenite and magnetite analyses showed equilibrium temperatures of 686° to 872°C. Magmatic oxides are present in greater than accessory amounts in gabbros with pyroxene Mg# <70–73, the point at which the overall trend of TiO2 in clinopyroxene begins to decrease with decreasing Mg#, which may suggest oxides first appeared on the liquidus. The presence of oxide gabbros with relatively high Mg# clinopyroxene may argue for early multiple saturation with oxides or a postcumulus melt-rock interaction and partial reequilibration between crystals and an evolved trapped melt. The additional precipitation of oxides both from trapped melt and from relatively iron rich melts migrating through the cumulus crystal mass may also help explain their coexistence with high-Mg# pyroxenes. The overall abundance in opaque oxides in some rocks can be >30%, strongly suggesting a cumulus process for origin of the oxides and that oxides were on the liquidus at the time of formation of the cumulus minerals. If the oxides were the result of trapped liquid, oxide modal percentages would be significantly lower.
We observed zircon, apatite, and various sulfide accessory phases in the samples, but they were not analyzed. In some of the samples, the accessory phases were along small veinlets a few grains thick that appear to penetrate the gabbroic rocks. In these cases, they usually resulted in trace element spikes in downhole plots. In other cases, they were observed as intergranular phases in silicic granophyres or oxide gabbroic rocks.
A summary of the major silicate phase compositions is presented in histograms of Fo content of olivine, Mg# of clinopyroxene, and anorthite content of plagioclase (Fig. F34). Both clinopyroxene and plagioclase show a small bimodal distribution of samples, but the general and predictable trend is toward decreasing numbers of more fractionated analyses. A remarkably good correlation exists between the average composition of forsterite-anorthite, Mg# clinopyroxene–anorthite, and Mg# clinopyroxene–forsterite (Fig. F35). The correlation is significant because of the different diffusion timescales for reequilibration for olivine, plagioclase, and clinopyroxene (e.g., Korenaga and Kelemen, 1998). Olivine has very short diffusion times, and plagioclase has long reequilibration times near magmatic temperatures. The first-order correlation between the Mg# of clinopyroxene and anorthite content is especially clear on the downhole plot (Fig. F36). The correlation is aided by the fact that plagioclase and clinopyroxene pairs were sampled in almost all samples and the peaks and troughs in the downhole cryptic chemical variation plots are well correlated. This illustrates that the primary control on rock composition was simple magmatic differentiation along a typical MORB liquid line of descent (LLD) (also see Thy, Chap. 2, this volume; Dick et al., 2002).
Like the whole-rock trends, downhole variations show both normal and inverse variation trends. The forsterite-anorthite or Mg# clinopyroxene vs. forsterite correlation is still strong, although not as robust because of the sparser sampling of olivine. However, general first-order downhole trends are easily correlated where sampling of olivine is adequate. It appears that the original cryptic variation is well preserved in the section and not muted significantly by extensive late-stage permeable melt flow through a crystal matrix. However, there may be considerable hypersolidus reequilibration between evolved trapped melt, olivine, and possibly clinopyroxene and/or subsolidus reequilibration between olivine and clinopyroxene based on the lack of agreement with equilibrium Fe-Mg partition coefficients between clinopyroxene and olivine and melt. Coexisting clinopyroxene and olivine (Fig. F35E) fall well off Kd(ol/liq) = 0.29 and Kd(cpx/liq) = 0.23 predicted correlation lines (e.g., Grove and Bryan, 1983). However, it is clear that permeable melt flow through the cumulate pile has not significantly altered the fine-scale chemical stratigraphy and covariations between mineral assemblages.
For example, in addition to major elements, the cryptic variations of compatible minor elements such as Cr2O3 in clinopyroxene and NiO in olivine also show a strong correlation in terms of the positions of major peaks and troughs, indicating that the original magmatic variation on the scale of a meter or less is well preserved in each of these primary phases (Fig. F37). Cr2O3 abundance in clinopyroxene appears to more rapidly deplete to lower levels at higher Mg#s when compared to NiO in olivine. Likewise, plots of Fo in olivine vs. Cr2O3 in clinopyroxene and Mg# of clinopyroxene vs. NiO in olivine show strong correlation expected of covariation during fractional crystallization. Incompatible elements such as MnO in clinopyroxene and olivine show very strong correlations with Fo content in olivine. If diffusion rates are so different between olivine and clinopyroxene we would not expect such good correlations if migrating melts were strongly reequilibrating with primary assemblages throughout the section. Such strong correlations between Fo-NiO in olivine (Fig. F37) and Fo-Cr2O3 in clinopyroxene (Fig. F37) were interpreted by Korenaga and Kelemen (1998) to indicate the absence of significant late-stage permeable melt flow. Poor correlation of coexisting mineral compositions in rocks with different modal abundances would be expected as the result of reequilibration of the matrix of a crystal mush with melts being transported through it by permeable flow on a large scale. Although each diagram has outliers that appear to represent small channels or melt conduits, as we have previously discussed, and are related to fracturing or concentrated areas of porous flow of unrelated melts, the dominance of highly correlated covariation between the primary phases olivine, clinopyroxene, and plagioclase may support Korenaga and Kelemen's interpretation. Their interpretation follows from the very different solid-state diffusion coefficients for olivine, pyroxene, and plagioclase and differences in reaction rates for different elements (e.g., trivalent vs. divalent cations) within individual phases (e.g., Korenaga and Kelemen, 1998; Kelemen et al., 1997). However, more rapid expulsion of migratory interstitial melts within a narrow boundary layer undergoing in situ heterogeneous nucleation and in communication with a central homogeneous magma chamber, some postcumulus reequilibration with trapped melt, and subsolidus equilibration may explain some of the scatter within each diagram. When the few outliers are eliminated from the covariation diagrams, the covariation correlation coefficients are generally as high as or higher than those depicted by Korenaga and Kelemen (1998) for the Oman ophiolite and tend to be higher than those described by Dick et al. (2002).
In cases where there is widespread melt flow through the cumulates and cryptic peaks in downhole mineral composition profiles are subdued by reequilibration with a migrating melt, the differences in reaction rate cause the peaks in mineral composition to shift relative to one another. Korenaga and Kelemen (1998) showed in a detailed study of a 600-m layered section of the Oman ophiolite excellent covariation of the silicate minerals that precluded large-scale melt migration through the section. They found that the nickel content of olivine correlated well with the forsterite content (R2 = 0.89). By contrast, for Hole 735B gabbros, Dick et al. (2002) suggested no correlation at all between olivine nickel and forsterite content other than an increasing upper bound for Ni with increasing Mg#. Similarly, Korenaga and Kelemen (1998) found the correlation coefficient for anorthite and forsterite in their gabbros to be R2 = 0.59, whereas Dick et al. (2002) found a significantly lower correlation coefficient for the Hole 735B olivine gabbros. Dick et al. interpreted their results to indicate extensive late melt-rock reaction and permeable flow. Our results suggest that large-scale melt flow has failed to significantly perturb the chemical stratigraphy in Hole 1105A.
Many authors who have described the characteristics of igneous layering in continental layered intrusions have interpreted these characteristics to originate, in part, by in situ crystallization in a boundary layer of graduated solidification (Jackson, 1961; McBurney and Noyes, 1979; Campbell, 1978, 1996). Casey and Karson (1981) proposed in situ crystallization as an important mechanism of layer formation in ophiolite gabbroic sections and, by extension, within the oceanic crust based on the complex and steep (sidewall) layer geometries and observations of clear in situ crystallization features such as harrisitic textures. We noted the occurrence of similar complex, variably inclined (see "Hole 1105A Deformation Extent and Core Microstructure"), and laterally discontinuous layering in Hole 1105A. The layering is generally defined by modal abundance and grain size changes. Modal changes relate to the appearance and disappearance of oxide or olivine or by modal proportion changes.
Langmuir (1989) and Nielson and DeLong (1992) explored the geochemical consequences of in situ crystallization utilizing equilibrium and Rayleigh crystallization models, respectively. They showed that having a boundary layer with a graduated temperature gradient between the solidus at the wall of the chamber and convecting primitive melt within the interior of the chamber could result in complex mixing between fractionated, multiply saturated melts formed at the outer wall of the boundary layer and primitive melts in the convecting part of chamber interior. Multiply saturated melts may be expelled and mixed into the more primitive main part of the chamber, a process called compositional convection (e.g., Tait et al., 1984; Sparks et al., 1984). A characteristic of in situ crystallization is that the minerals solidifying in the boundary layer are not necessarily on the liquidus of the convecting crystal-free interior of the chamber. For example, the boundary layers may crystallize clinopyroxene, oxides, or other accessory phases in addition to olivine and plagioclase, but they are not necessarily on the liquidus in the interior of the chamber. Melts in the interior of the chamber would be continually enriched in incompatible elements and also show evidence of clinopyroxene or oxide crystallization. Langmuir (1989) suggested in situ crystallization was a possible explanation for the well-known "clinopyroxene paradox" in MORB (also see Langmuir et al., 1992; Nielson and DeLong, 1992), where melts with olivine and plagioclase on the liquidus show evidence of extensive clinopyroxene fractionation (phantom crystallization). The mineral assemblage in the boundary layer is dependent on the extent of crystallization prior to the point where the liquid escapes (Nielson and DeLong, 1992). We expect this process depends on the temperature/density gradient on the margin of the chamber, which may vary in time and space along the chamber margin. Small gradients between melt and solid gabbro would produce small amounts of fractionation in the melts delivered to the convecting interior of the chamber, whereas large gradients would produce highly fractionated crystallization products and melts delivered to the convecting portion of the chamber. The cycling between olivine gabbros and oxide gabbros may be in part related to periodic replenishment or narrow or small magma chambers or chambers near the roof zone of the plutonic crust.
In the case of the gabbroic rocks sampled in Hole 1105A, fractionation would have to proceed in some instance to the point where Fe-Ti oxide crystallizes and continuously mixes back into the convecting interior of the magma chamber. Variations in temperature gradient presented by replenishment, variable cooling rates, or episodes of assimilation of altered gabbro may have allowed cycling between olivine gabbro, gabbro, and oxide gabbro solidification and smooth transitions between them. The entire variation on small scales would be aided by periodic replenishment (or recharge) of the interior of the magma chamber with hotter, more primitive basaltic liquids and convective instability of the boundary layer (periodic boundary layer erosion) (e.g., Campbell, 1996). Reestablishment of the crystal mush boundary layer would then renew the ability of the boundary layer to produce more diverse magma types, leading to precipitation of oxides and perhaps other minor accessory phases. The appearance of oxide gabbros may denote that the magma system was locally saturated in Fe-Ti oxides and crystallizing cumulus oxides at the colder edge of the boundary layer and that local melt flow/expulsion and crystallization of these evolved melts or trapped restite melt may have led to increases in the percentages of oxides.
One of the consequences of boundary layer fractionation is that apparent magma liquid lines of descent in the boundary layer do not necessarily follow those expected by homogeneous fractionation within the interior of the magma chamber by simple Rayleigh distillation fractionation processes. Nielson and DeLong (1992) showed that boundary layer fractionation trends are particularly sensitive to the relative volume (thickness) of the boundary layer compared with that of the interior of the magma chamber and the extent of fractionation in the boundary layer. In Figure F38 we show the calculated liquids in equilibrium with clinopyroxene in Mg#-TiO2 space using constant clinopyroxene/melt partition coefficients of 0.23 for FeO/MgO (Roeder and Emslie, 1970) and 0.41 for TiO2 (Johnson and Kinzler, 1989) and assume these parameters in modeling. Utilizing Nielson and DeLong's modeling approach, we show the LLD appropriate for one of the most primitive basalts recovered from the Atlantis Fracture Zone (Johnson and Dick, 1992) that would result from perfect fractional (Rayleigh) crystallization. This approach generally matches the Atlantis II Fracture Zone basalt trend. The basalt range, however, is not as extensive as the plutonic equilibrium melts. The LLD derived from PFX modeling does not match the apparent LLD illustrated by the calculated liquids in equilibrium with clinopyroxene in Hole 1105A. Specifically, the broad trend derived from Hole 1105A pyroxenes shows a decrease in TiO2 in the liquid starting at Mg# = ~45 in the liquid. Modeling shows that TiO2 depletion should not take place until Mg# = ~23 under reasonable redox conditions. Galapagos basaltic glasses, which show extreme fractionation to ferrobasalts, andesites, and dacite, illustrate FeOT and TiO2 depletion and SiO2 enrichment trends at Mg#s of ~28–29. To match the trend illustrated by clinopyroxene data using constant partition coefficients, we conducted boundary layer fractionation modeling (Nielson and DeLong, 1992) of one of the most primitive Atlantis II basalts in which we assumed a boundary layer represented 5% of the magma chamber volume and the extent of fractionation in the boundary layer was as high as 70%. We were able to match the trend illustrated by calculated melts in equilibrium with clinopyroxene. In boundary layer fractionation, the extent of fractionation determines the extent of multiple saturation. In the case of 60%–70% fractionation, the melt can saturate even in minor phases. Most basaltic magmas would readily saturate with opaque oxides in the solidification zone under the modeling conditions imposed, leading to the TiO2 depletion trend at the high Mg#s observed. Unlike the saturation achieved in the boundary layer, the central homogeneous magma chamber would not be saturated in these late crystallizing phases (e.g., Atlantis II basalts are not). TiO2 in this case would show enrichment in the homogeneous convecting part of the magma chamber but a more limited or flattened extent of enrichment in the boundary layer. The basalt trend observed follows the homogeneous magma chamber trend, whereas the cumulates seem to follow a boundary layer trend with respect to the clinopyroxene mineral chemistry and modal relationships. Apatite fractionation evident in some of the oxide gabbros further supports the idea of saturation in minor phases within the boundary layer. The extent of fractionation over small intervals (1 m) suggests strong thermal gradients across the boundary layer, consistent with small or high-level magma chambers.
An important observation in our analysis is that the appearance of magmatic (or abundant) oxides in the mode of the gabbroic and metagabbroic rocks analyzed corresponds to calculated liquid compositions of Mg# = ~45 or below. Fifty-two samples below this value contained magmatic oxides; only three did not contain oxides. One sample that contained oxides had an Mg# in the equilibrium liquid > 45; the remainder of the samples with liquid Mg#s > 45 (57 samples) lacked magmatic oxides. This apparent boundary between oxide-bearing and oxide-barren gabbroic samples (Mg# = ~45 in liquid) also corresponds to the point where the liquid and equilibrium clinopyroxene (Mg# = ~70–73) show the initiation of a TiO2 depletion trend. It is important to note that PFX modeling at or near quartz-fayalite-magnetite (QFM) buffer would predict oxide precipitation at or below liquid Mg# = 30.
This may not, however, be a unique solution because it is possible that the partition coefficient for TiO2 in clinopyroxene is not constant, as previously discussed. In Figure F38B, we conducted the same type of PFX and boundary layer fractionation (BLF) modeling with varying Kd. As discussed in "Mineral Chemistry," TiO2 concentrations in the clinopyroxene structure may depend on Al(IV) in the structure. Thus, as shown in Figure F31 (Al[IV] vs. Ti), as Al(IV) decreases, DTi decreases and so does the ease with which Ti can enter the clinopyroxene structure. So Al(IV) was calculated for all the samples using stoichiometric relationships. The Al(IV)-DTi correlation equation obtained from Figure F31 was used to calculate DTi in each individual sample using the previously calculated Al(IV), giving us the DTi for each individual sample; thus, DTi was a variable and not a constant value. This variable DTi was used to calculate the amount of TiO2 in the liquid in equilibrium with the clinopyroxene in the gabbros. Figure F38B shows a plot between calculated Mg# of melts and calculated TiO2 in melts (using variable DTi) in equilibrium with the clinopyroxene grains. The plot is significantly different from Figure F38A. The distinct peak observed in the melt in equilibrium with cumulates is absent, and a more gentle BLF curve almost parallel to the PFX LLD is observed. A very subtle kink is observed in the BLF curve. The boundary separating the oxide free from the oxide bearing and oxide rich is not a well-defined peak, but it is still clearly seen. Although the BLF curve is not very distinct and runs almost parallel to the PFX curve, maximum TiO2 occurs much earlier compared to PFX under same conditions of fractionation. Like oxygen fugacity modeling that we conducted, BLF modeling was completed to find a suitable model to match the melt compositions. In our modeling the boundary layer was 10% of the total chamber volume, oxygen fugacity was decreased to 2 log units below the QFM buffer, and the amount of fractionation in the boundary layer was taken to be 40%. The plot shows the curve, and it is a reasonable fit, but as suggested by Nielson and DeLong (1992), at lower extents of fractionation in the boundary layer it is difficult to observe the effects of BLF. Rather, the results are similar to homogeneous fractionation. The concentrations of TiO2 in the equilibrium melts along the LLD are different in the models (PFX and BLF). PFX predicts much more TiO2 enrichment than does BLF. Also, the BLF estimation seems to be much closer to the melts in equilibrium with the clinopyroxene grains in the gabbroic rocks. Although choosing between variable Kd for TiO2 or constant Kd creates a degree of ambiguity on which models appropriately describe the process, in both cases BLF remains a viable process for the formation of the gabbroic rocks and, in particular, an early appearance of opaque oxides.
Natland and Dick (2002) and Thy (Chap. 2, this volume) interpreted oxides in gabbroic rocks of Holes 735B and 1105A, respectively, and their association with primitive olivine gabbros to indicate that gabbros were impregnated by evolved intercumulus melts that were saturated in oxides and that the association of oxides with zones of ductile deformation demand that these zones were hypersolidus mush zones impregnated by oxide-saturated melts. Our analysis shows that the most primitive olivine gabbros (with high Fo, An, and Mg# in clinopyroxenes) in Hole 1105A are generally not oxide bearing, with one exception, and that only more evolved gabbroic rocks and mineral chemistries are associated with oxides. The model of Dick et al. (2002) predicts random impregnation of both primitive and more evolved intervals, but clearly oxides in Hole 1105A core do not necessarily appear randomly in primitive gabbros, except locally in those sections noted, and there is a systematic progression of phase chemistry and bulk rock compositions leading to the appearance of oxides. Alternatively, both models may have some commonality and shared consequences.
Our analysis may suggest that in situ crystallization processes may help to explain the proximity and interlayering of more primitive olivine gabbros and oxide gabbros. In part, the models of Dick et al. (2002) and Thy (Chap. 2, this volume) may be compatible because boundary layer fractionation would predict removal or extraction of much of the interstitial fractionated boundary layer melt and delivery to the main body of the magma chamber as the cumulates texturally matured. Certain observations of impregnation zones of oxide-saturated melts and higher incompatible element abundances thus may be consistent with this type of short-distance transport, melt reequilibration with cumulus minerals, in situ fractionation within pore space, and melt extraction to the interior of the chamber. The relative enrichment in incompatible elements over those expected in perfect adcumulate crystal growth in many of the gabbroic rocks coupled with lack of zoning in most samples may indicate a certain amount of reequilibration in many samples. However, the strong and intricate cryptic chemical covariations among cumulus phases outlined above do not support widespread porous melt transport and reequilibration throughout the cumulate pile significant enough to alter peak and trough covariations between mineral phases with long and short equilibration times. Localized transport along fractures or shear zones may be supported by the fact that some samples show significant differences in core to core and core to rim compositions.
Other mechanisms that have been proposed for the complex juxtaposition of oxide gabbros and more primitive gabbroic rocks such as in the Skaergaard intrusion (e.g., McBirney, 1995; also see Thy and Dilek, 2000) seem to involve boundary layer fractionation along the strongly cooled walls and roof zone of a magma chamber and leakage of dense ferrobasaltic liquids back into the magma chamber that en mass sink in the less dense, more primitive magma in the interior of the chamber to the chamber floor. This process would cause complex juxtaposition of primitive and highly fractionated gabbroic rocks. In fact, upper isotropic gabbroic rocks within ophiolites are commonly oxide bearing and contain abundant trapped melt. This suggests that the strongly chilled and hydrothermally cooled upper contacts of magma chambers may be factories for the production for ferrobasaltic liquids that could be density driven to escape into the magma chamber, resulting in dense fingers that sink to the chamber floor.