A total of 148 samples were analyzed by inductively coupled plasma–atomic emission spectroscopy (ICP-AES) at the University of Houston as part of our study for whole-rock major and minor elements (Table T2), trace elements including Zr, V, Ba, Sr, Ni, Cr, Co, Zn, Cu, Sc, and Y (Table T3), and rare earth elements (REEs) including La, Ce, Nd, Sm, Eu, Gd, Dy, Er, and Yb (Table T4). Methods are described in Casey (1997) and Smith (1994). Adding these analyses to shipboard X-ray fluorescence (XRF) analyses of 42 samples brings the total number of samples analyzed for major, minor, and certain trace elements to 190. REEs, Sc, and Y were analyzed by preconcentration utilizing chromatographic separation techniques followed by ICP-AES analysis (see method references above). Over the 143-m cored interval, the 190 samples analyzed provide an average sampling interval of 0.77 m. In the plots of whole-rock analysis, our data were supplemented with the shipboard data when overlap in element sets occurred. Unfortunately, REEs were not analyzed on board the ship because of limitations of onboard analytical techniques, and plots are therefore restricted to 148 samples. The whole-rock data, presented here, represent the entire data set now available for Hole 1105A. Whole-rock major and trace element geochemistry generally shows that the analyzed gabbroic rocks are cumulate in origin, whereas the more silicic rocks may have meltlike compositions.
Whole-rock Mg#s (Mg/[Mg+Fe] x 100) of all plutonic rocks ranging from olivine gabbros to trondhjemites cover an exceptionally wide range from 85.4 to 6.3 (Fig. F12). This Mg# range represents a much wider range of dominantly cumulate magmatic differentiates than exhibited by Atlantis II MORB Mg#s (61.9–42.2) (Johnson and Dick, 1992). In nonoxide-bearing plutonic rocks, Mg# can be used in a general way as an index of fractionation because it largely reflects the changes in composition of the Mg# of cumulus clinopyroxene and olivine, which are very close in value. In oxide-bearing and especially oxide-rich gabbro, however, Mg# varies more rapidly downhole with increasing modal percent of opaque oxides. Whole-rock Mg# is less directly correlatable with mineral Mg# value and is not reliable as an index of fractionation in these rocks; thus, the Mg# of mafic minerals is more reliable as an indicator. Oxide-poor trondhjemites and quartz diorites appear to reflect near-liquid compositions based on their whole-rock chemistry (also see Thy, Chap. 2, this volume).
The Atlantis II basalts are characterized by Fe# (Fe/[Fe+Mg] x 100) values of 38.6–58.2 (average = ~44). The Atlantis II Transform basaltic compositional range (Johnson and Dick, 1992) corresponds to a magma density range centered on the density minimum of the tholeiitic basalt magma fractionation trend. The range is within the "window of eruptibility" along the tholeiitic trend (Fig. F12) that was defined by Stolper and Walker (1980). More primitive parental melts or more fractionated iron-rich basalts in equilibrium with many of the Hole 1105A plutonic minerals would be denser and less likely to erupt. It is not surprising, therefore, that the range of compositions and the fractionation extents are greater in the plutonic foundations of the crust than in basalts. The olivine gabbros in Hole 1105A have Mg#s commonly in excess of the maximum for basaltic liquid (69–71), which defines the most primitive possible basaltic liquids likely to be in equilibrium with oceanic residual mantle olivine. Mg#s of Hole 1105A gabbroic rocks range as high as ~85.4. Samples with Mg#s between 71 and 85, which includes 45% of those analyzed, are too primitive to represent the composition of congealed melts. These samples must represent the more magnesium cumulate products derived from crystallization of melts that were somewhat more fractionated than mantle-equilibrium melts (also see discussions in Dick et al., 2002; Niu et al., 2002), and the plutonic whole-rock values reflect the higher Mg#s of cumulus phases in equilibrium with MORB (Fig. F12). Samples with lower Mg#s in the range of basaltic liquid samples with whole-rock Mg#s <70–71, based on trace element abundances (see below), also do not appear to represent basaltic liquid compositions. The average Mg# of all samples for Hole 1105A is 62.7 because of the significant amounts of ferrogabbroic to granophyric rocks averaged within the section. As in the case of the deeper Hole 735B, the cumulate rocks analyzed are far too evolved over this limited section to mass balance back to a primitive melt composition (also see Dick et al., 2002; Niu et al., 2002). Thus, parental magmas that fractionated cumulate phases were more evolved than any primary magmas from the mantle and indicate that the most primitive cumulates are not represented in the cored interval, as might be expected from the limited section cored. A similar conclusion was reached by Dick et al. (2002) for the broader cored interval in Hole 735B. Like Hole 735B, the cored section in Hole 1105A does not include the most primitive cumulates. Cannat and Casey (1995) and Cannat (1996) observed a range of gabbroic, dunitic, troctolitic, and pyroxenite bodies within residual mantle material in the 14°–16°N region of the Mid-Atlantic Ridge that suggested the existence of discrete magma bodies crystallizing in the lithospheric residual mantle. Similar rocks were recovered during Leg 153 (e.g., Cannat et al., 1997; Casey, 1997). High-Mg# cumulates found in the Bay of Islands ophiolite and other ophiolites generally occur sandwiched between the base of the layered gabbro section and the underlying residual mantle. This may indicate that deeper penetrations would be required to sample the "missing," more primitive cumulates. A comparison with the Mg#s in Hole 735B shows that even after drilling 1.5 km into the lower crust, primitive cumulates were not been sampled. It seems likely that the missing cumulates could be found by either deeper penetration at the base of the gabbroic cumulates, as plutons within the residual mantle section below (e.g., Cannat and Casey, 1995), or in dismembered lozenges of a mafic–ultramafic transition zone along a flow line within the lower crust. Predicting a drilling strategy for sampling these primitive cumulates would be, at best, difficult within a single hole.
Most gabbroic rocks with whole-rock Mg# > 71 have <0.1 mod% opaque oxide content. If oxides are present, they are typically alteration related. At Mg# < 65, most samples have >1.0 mod% opaque oxides. The appearance of oxides at high Mg# does not fit perfect fractional crystallization (PFX) modeling of Atlantis II basalts because the model predicts their appearance at much lower Mg#s. The systematic trend toward increasing modal abundances of opaque oxides with decreasing whole-rock Mg# would be expected if opaque oxides have accumulated. Opaque oxide abundances therefore, in part, control the Mg#s of more fractionated cumulate rocks. Samples with oxide abundances >1% typically have whole-rock Fe2O3 (total iron as Fe2O3) >10 wt% in Hole 1105A gabbroic rocks. Oxide abundances in oxide gabbros range as high as ~25%, and whole-rock Fe2O3 abundances reach as high as ~35 wt% (Fig. F13A). In high-Mg# samples, whole-rock abundances of TiO2 (Fig. F13B) are generally low and not well correlated with Fe2O3 abundances. When Fe2O3 < 10 wt%, TiO2 abundances tend to be <1 wt% and often significantly lower. When Fe2O3 > 10 wt%, however, TiO2 and Fe2O3 abundances are highly correlated, with both Fe2O3 and TiO2 reflecting increasing modal percentages of opaque oxides ranging from magnetite to titanomagnetite in composition. SiO2 is negatively correlated with both Fe2O3 and TiO2 in this range. TiO2 contents reach as high as 7.56 wt% in oxide gabbros, similar to the range observed in Hole 735B (Dick et al., 2002). Figure F10B shows the total range of whole-rock TiO2 contents in plutonic rocks (0.09–7.56 wt%) as well as the abundance of TiO2 within clinopyroxene in Hole 1105A gabbroic rocks and within Atlantis II basalts. In general, the Hole 1105A plutonic rocks are depleted in TiO2, with 64% of the samples having <1 wt% and 42% having <0.5 wt%. In contrast to Atlantis II basalts, which show a range of TiO2 values from 1.29 to 3.24 wt% (average = 1.96 wt%), the majority of Hole 1105A gabbroic rocks are far too depleted in TiO2 to represent any similar congealed MORB liquid compositions. This depletion generally indicates a cumulus origin for these rocks. Titanium behaves incompatibly with respect to clinopyroxene, olivine, and plagioclase, and therefore the low TiO2 in oxide-poor to oxide-free cumulate gabbroic rocks can be interpreted to indicate cumulates with somewhat low percentages of intercumulus trapped melt. Upon crystallization of Fe-Ti oxides and their obvious accumulation in the oxide gabbros, TiO2 can far exceed the abundances expected in equilibrium ferrobasaltic liquids, which reach a maximum of ~3.00–3.50 wt% when Atlantis II basalts are modeled by PFX. Thus, it is also important to note that iron- and titanium-rich gabbroic rock compositions are far removed from congealed ferrobasaltic melt compositions, and therefore the rock and oxides themselves must be, in part, cumulus in origin.
The large relative diversity of the samples analyzed from Hole 1105A, in terms of overall geochemistry, lithologic variation, and iron enrichment, can be illustrated on an alkali-iron-magnesium (AFM) diagram comparing the Hole 1105A data to other large oceanic data sets such as the plutonic suites from the MARK area along the Mid-Atlantic Ridge and the Bay of Islands ophiolite (Fig. F13C). Comparison shows that the Hole 1105A gabbroic samples follow a strong iron enrichment trend, like the other suites, but that the extent of oxide enrichment surpasses those of the Bay of Islands ophiolite and the MARK plutonic rocks and more closely resembles the Skaergaard enrichment extent (also see Thy, Chap. 2, this volume). In this respect, Hole 1105A samples are quite similar to Hole 735B samples (Dick, Natland, Miller, et al., 1999), as might be expected if the oxide units can be correlated.
Bulk rock SiO2 vs. Mg# or MgO plots show that SiO2 is restricted to a narrow range when olivine, clinopyroxene, and plagioclase are the dominant cumulus phases and are the main controls on bulk composition. Oxide accumulation leads to significant depletion in SiO2 (Fig. F14A, F14D) with values descending from ~51 to 33 wt% as oxide modal abundances increase, similar to samples from Hole 735B. Trondhjemitic and dioritic veins within gabbroic rocks analyzed lead to SiO2 enrichment trends, and these appear to be represented in granophyric liquid addition trends in Figure F14A, F14D, and F14E. Similar mineralogic and melt or oxide controls are exhibited when bulk rock TiO2 is plotted vs. Mg#, MgO, and Fe2O3 (Fig. F14B, F14E, F14F). Fe2O3 and TiO2 are highly correlated because of oxide content, and, likewise, the Mg#s of oxide gabbros are highly correlated with modal oxide abundances (Fig. F14C, F14F). Minor elements P2O5 and MnO show typical increasing incompatible element trends with decreasing bulk Mg#. In the more iron rich oxide gabbros, however, apatite accumulation trends are observed that drive P2O5 values to >4 wt% (Fig. F15). These apatite-bearing gabbros are also described by Thy (Chap. 2, this volume). Like TiO2, abundances of MnO and P2O5 in oxide-free gabbros and olivine gabbros are lower than in Atlantis II basalts, with which they could be in equilibrium, again reflecting their cumulate nature. Whole-rock MnO abundances are largely controlled by fractionation and the consequent increasing MnO abundances within clinopyroxene and olivine with increasing extents of fractional crystallization, but will also be affected by increasing trapped intercumulus melt contents.
Whole-rock trace element abundances (Tables T3, T4) in the gabbroic rocks depend on the composition of the parent melt from which the cumulus crystals solidified, the partitioning behavior between melt and crystalline phases, the modal mineralogy of the sample, the percentage and composition of residual intercumulus melt trapped within the crystal framework, or reactive crystallization products of slow migratory melts infiltrated through or trapped in the cumulate network or crystal mush (e.g., Irvine, 1980; Casey, 1997; Agar et al., 1997; Coogan et al., 2000; Coogan et al., 2001; Dick et al., 2002). The trapped melt abundance is best manifested generally by comparing the whole-rock abundances of incompatible trace elements to those of known MORB melts in the region and expected in equilibrium cumulus phases. The trace element patterns on extended REE diagrams generally help to distinguish plutonic rocks formed by congealing of melts (e.g., Hart et al., 1999) vs. those of dominantly cumulate origin (e.g., Casey, 1997; Hertogen et al., 2002; Niu et al., 2002).
Trace elements reported here range from compatible (e.g., Ni and Cr) to highly incompatible (e.g., light REEs [LREEs]) during crystal fractionation. Extended REE spidergrams are presented for each of the samples analyzed in Hole 1105A. Elements are listed in order of incompatibility during mantle melting (Sun and McDonough, 1989) and normalized to chondrite using values of Anders and Grevesse (1989). Spidergrams are plotted for Unit I (Fig. F16), Subunit IIA (Fig. F17), Subunit IIB (Fig. F18), Unit III (Fig. F19), and Unit IV (Fig. F20). Figure F21A–F21C shows extended REE diagrams plotted on the basis of oxide-free, oxide-bearing, and oxide-rich lithologies, respectively. Most samples plotted show large positive Sr anomalies and smaller positive Eu anomalies, indicating that the sample suites are largely cumulate in origin, with significant plagioclase accumulation. Because Sr has a sufficiently high partition coefficient (Kd) in plagioclase, the bulk distribution coefficient approaches 1 during crystallization of the basalt, creating negative Sr anomalies in the liquid and positive anomalies in cogenetic primitive cumulates and leading to nearly constant Sr abundances in each (Casey, 1997). Eventually, in fractionated oxide gabbros the adjacent REEs and other trace elements increase due to fractionation of the melt and exceed the nearly constant Sr abundance, creating negative Sr anomalies in some evolved gabbroic rocks. Eu anomalies are typically positive in gabbroic rocks, increasing with increasing modal plagioclase abundances. Frozen liquids typically have negative Sr and Eu anomalies because of plagioclase fractionation. Several granophyric rocks from Hole 1105A have major and trace element characteristics similar to those expected of highly fractionated congealed melts with negative Sr, Eu, and Ti anomalies. In addition, many of the gabbroic rocks highly enriched in oxides are characterized by low to negative Eu anomalies and negative Sr anomalies, as well as highly enriched REE abundances. These highly fractionated oxide gabbros and oxide olivine gabbros typically have positive or negative Ti anomalies, reflecting either oxide accumulation if positive or crystallization from melts that have already experienced extensive prior oxide precipitation if significantly negative. Saturation of oxides is evident from the positive Ti anomalies (oxide accumulation effects) and negative Ti anomalies (oxide depletion effects) in samples with elevated REE abundances. Whereas the analyzed samples with these positive or large negative anomalies typically show the most elevated REE and incompatible element abundances, gabbroic rocks in which REE abundances are <10x chondrite usually lack such large Ti anomalies. Small negative Ti anomalies in more primitive samples may also be a symptom of dominantly spinel facies mantle melting (Casey, 1997) or may also result from primitive magmas mixing with more evolved magmas that have undergone oxide precipitation, perhaps in a magma chamber boundary layer. Likewise, oxide gabbroic rocks typically have highly variable Zr anomalies that are dominantly positive, reflecting zircon accumulation, except in highly fractionated oxide gabbros or felsic granophyric rocks such as diorite and trondhjemite, which can have negative or positive Zr anomalies because of significant prior zircon fractionation or accumulation. The most highly evolved felsic rocks and certain evolved oxide gabbros are also typically characterized by negative Sr, Eu, and Ti anomalies and highly enriched REE signatures. The more silicic plutonic rocks based on these and other major element compositional characteristics may be close to liquid compositions.
Overall REE abundances (Table T4) show a wide range of compositions in the samples analyzed. LREEs such as La and Ce range from <1x to >110x chondrite. Similar ranges are observed in heavy REEs (HREEs) as well, reflecting the overall dominantly flat patterns to slightly LREE depleted observed in REE data, although the most enriched samples, which are typically oxide-rich gabbros and more silicic samples, tend to show slightly elevated chondrite-normalized (cn) LREE/HREE ratios (e.g., [La/Yb]cn) that range from >1 to 3. The more depleted samples, typically oxide-free to oxide-poor gabbroic samples, tend to show lower ratios (<1–0.2). One anomalous oxide gabbro pegmatite was characterized by a very large ratio of >6.0. Oxide-depleted granophyres intruding gabbroic rocks likewise can show enriched characteristics in overall abundance and LREE/HREE ratios. Gabbroic rocks that are oxide free typically have REE abundances <1x to 12x chondrite with strong positive Sr anomalies and weak negative Ti anomalies. Oxide-poor granophyric rocks have higher values from 20x to 85x chondrites, with negative Sr and Ti anomalies and positive or negative Zr anomalies. Oxide gabbros range from 5x to 120x chondrite abundances and typically have strong positive or weak negative Ti anomalies, weaker positive Sr or strong negative Sr anomalies, and positive or negative Zr anomalies. Oxide-bearing gabbroic rocks have intermediate characteristics between oxide-rich and oxide-free gabbroic samples. All three types of gabbroic rocks typically show distinctive positive Eu anomalies, although some oxide gabbros show slight negative anomalies. Granophyric rocks possess slightly negative Eu anomalies as well.
Trace elements that show compatible element trends include Ni, Cr, and Cu. These show rapid depletion with decreasing whole-rock Mg# (Fig. F22A–F22C), although Cu begins to increase at lower Mg#s, probably due to separation of late sulfide liquids and accumulation of chalcopyrite in oxide gabbros. Petrographic evidence shows abundant chalcopyrite within some oxide gabbros, which would explain the elevated abundance of Cu in more fractionated rocks. The association of chalcopyrite with oxides is likely because consumption of oxygen during oxide crystallization may force sulfide precipitation to buffer oxygen fugacities (Mathez, 1976). Rapid nickel and copper depletion in less fractionated rocks occurs because both elements are compatible during the crystallization of olivine, with probable small of amounts of pentlandite (Miller and Cervantes, 2002) contributing to Ni depletion. Chrome behaves compatibly in clinopyroxene and is likewise rapidly depleted at relatively high Mg#s. Ni and Cu may also partition somewhat into late-stage immiscible sulfide melts. Slightly increasing Cr abundances in oxide gabbros may result from accumulation of oxides in which Cr is more compatible. In primitive rocks, however, Cr depletion trends largely reflect its compatible behavior with clinopyroxene, and Ni is largely controlled by similar compatible behavior during olivine crystallization, both rapidly depleting the melt compositions and resulting in lower abundances in both cumulus minerals with increased fractionation. Of particular note on these diagrams is that only a few samples lie off the depletion trends, whereas the hypothesis of Dick et al. (1991) and Natland et al. (1991) for the origin of Hole 735B oxide gabbro seems to suggest that differentiation products along lateral and vertical channelways invaded a host gabbroic crystal mush that could be much more primitive. This should result in complexes of "mixed" or infiltrated rock whose compatible element compositions resulted from mixtures of primitive cumulus olivine, clinopyroxene, and plagioclase and fractionated melt-derived Fe-Ti oxide cumulate gabbro. Yet the bulk rock plots (Fig. F22) suggest that the rock evolutionary path lacked significant intimate commingling of disparate primitive rock types (more primitive cumulus crystals) with highly evolved melt, although the broadness of the trends observed, may, in part, be explained by mixing with slightly more evolved melts. Thus, if correct, the hypothesis would seem to suggest that the highly evolved oxide-rich channelways did not commingle with a more primitive crystal mush, but perhaps migration was through already solidified more primitive gabbroic rocks. Alternatively, there could be another explanation for the complex juxtaposition of primitive olivine gabbros (and troctolites in Hole 735B) and oxide gabbros. Vanadium and Sc abundances can usually be correlated with modal abundances. Vanadium shows a positive correlation with Ti and generally reflects oxide modal abundances, whereas Sc generally increases with modal clinopyroxene.
Bulk rock incompatible trace elements such as Ce and Yb in most of the more primitive gabbros are likely controlled by the modal percent of clinopyroxene and plagioclase and interstitial trapped melt (e.g., Irvine, 1970; Casey, 1997) or migratory melt cumulate crystallization products (e.g., Coogan et al., 2000). Figure F23A and F23B depicts Ce and Yb concentrations in Atlantis II basalts (Johnson and Dick, 1992), their concentration in equilibrium clinopyroxene and plagioclase calculated from partition coefficients (see Casey, 1997), and the gabbroic rocks analyzed as part of this study. Figure F23C and F23D illustrates Ce/Yb vs. Yb and La, respectively, showing the mixing lines between average basalt and equilibrium clinopyroxene and plagioclase. These three components should exert dominant controls over REE content of the average cumulate (Casey, 1997). Except for highly fractionated (i.e., zircon-apatite bearing) samples, whole-rock concentrations in Hole 1105A plutonic rocks generally plot within or close to the volume defined by the mixing of the three components, average basalt and calculated compositions of average clinopyroxene, and average plagioclase. In particular, the majority of samples analyzed plot parallel to mixing lines depicted between clinopyroxene and plagioclase compositions and indicate 10%–20% or more of the average basalt could be contained within most primitive samples, suggesting a dominantly adcumulate to mesocumulate nature for most of the samples analyzed. The translation of petrographically determined textural types adcumulate, mesocumulate, and orthocumulate to equivalent trapped melt content classification for partial or complete major element reequlibration between trapped melt and cumulus or primary framework crystals is based on the classification and terminology of Irvine (1980, 1982):
We use this classification assuming highly incompatible trace elements can be used as a proxy for estimating minimum trapped melt contents. Alternatively, the trace element enrichment could result from migratory melts and more transient cumulate crystallization from a variety of compositions including ferrobasaltic and silicic melts driven through the cumulate pile; however, the migration rates would likely be very small or entrapment would have to occur to allow extensive reequilibration with plagioclase and clinopyroxene. The nature of cryptic chemical variations tends not to support extensive reequilibration of minerals with migrating melt, as minerals such as olivine and clinopyroxene have very different diffusion rates. If more fractionated liquids or their cumulate crystallization products (e.g., apatite and zircon) were involved as interstitial solidified melt, the calculated "trapped" melt abundance could actually be much lower. Clear subsets of gabbroic rocks are much more enriched in incompatible trace elements and appear to contain higher proportions of more evolved congealed melt components or evolved crystallization products like accessory minor phases (e.g., zircon, apatite, and oxides) that elevate incompatible element abundances. These subsets of rocks are pulled toward the Atlantis II basalt field or beyond to more enriched compositions as depicted in Figure F23. Some of these more enriched rocks are significantly higher in incompatible element concentrations than basalts because of more extensive fractionation (e.g., granophyric melts) and/or accumulation of accessory phases such as zircon and apatite in thin veins (samples with positive Zr anomalies and P2O5 enrichment) within oxide gabbros. Zircons tend to have high HREE/LREE ratios (~10,000:1–1500:1), and apatites have high LREE/HREE ratios (~2:1) (e.g., Gillis, 1996). Certain trace element patterns show strongly positive Zr anomalies symptomatic of zircon accumulation on extended REE diagrams (e.g., Figs. F16, F17, F18, F19, F20). Accumulation of either zircon or apatite as an accessory phase can significantly elevate certain incompatible element abundances.
Hart et al. (1999) used the composition of strip samples from nearby Hole 735B to reconstruct major element compositions of the bulk hole. They surmised that the major element composition of the average gabbro from Hole 735B is meltlike. Based on our analysis of the bulk rock data set, we simply averaged all plutonic rock compositions (N = 196) and concur that the average of all plutonic rocks including gabbroic, dioritic, and trondhjemitic rocks in Hole1105A are meltlike in terms of major element abundances (Table T2). In fact, there is an amazing 1:1 correlation between the average major element abundances of the Atlantis II basalt (Johnson and Dick, 1992) and the average for Hole 1105A plutonic rocks (Fig. F24A). The average Atlantis II basalt represents a moderately differentiated melt, and therefore an equivalent plutonic average is likely caused by averaging of the more extensive fractionation range from primitive gabbroic samples to granophyric samples represented by the plutonic assemblage (i.e., when compared with the more limited basaltic range). Obviously, the average of all plutonic rocks is likewise much more evolved than any melt composition that could be in equilibrium with oceanic mantle.
We, however, outlined above that most of the plutonic samples have incompatible minor and trace element abundances that are significantly lower than trace element abundances expected within basalt and that they have trace element patterns that demonstrate their cumulate origin (e.g., positive Sr, Eu, and Ti anomalies). REEs, which typically behave as incompatible elements during mantle melting and basaltic fractionation, demonstrate the differences between melt and cumulate gabbroic compositions. A plot of the average REE abundances for Atlantis II basalts vs. the average for all Hole 1105A plutonic rocks shows that REEs constitute approximately one-half the abundances of the basaltic average, even when highly fractionated trace element–enriched dioritic to trondhjemitic rocks are included in the average (Fig. F24B). Without these highly fractionated rocks that are probably close to melt compositions, the average would be significantly lower in the plutonic assemblage. Hart et al. (1999) interpreted the variability in Hole 735B to be caused by local separation of melt and crystals, but that the bulk average indicated no removal of melt. Thus, on average they interpreted the cored Leg 118 section to represent a melt composition, with little large-scale melt removal. We cannot concur with this conclusion or that the cored interval of Hole 1105A is meltlike because incompatible trace elements are so significantly depleted on average when compared with basaltic melts and for reasons summarized above (also see discussions of Hole 735A in Dick et al., 2002; Niu et al., 2002). The mass balance based on trace element bulk average indicated removal of melt from the magma chamber is a necessity. At the same time, it appears that the majority of Hole 1105A cumulates, apart from highly fractionated oxide gabbroic and silicic rocks, would be close to equilibrium assemblages with the moderately evolved Atlantis II basalts analyzed by Johnson and Dick (1992).
Downhole modal abundance data indicate significant variations in mineral proportions on small scales (<1 m) within Hole 1105A. Bulk rock geochemical data (Fig. F25) likewise show significant downhole variation in both major and trace elements and good correlation with lithologic unit definitions based on detailed shipboard descriptions. Units I and III are typically characterized by low Fe-Ti and high-Mg# gabbroic rocks with short intervals of more Fe-Ti rich gabbroic rocks that typically correspond to oxide-rich intervals noted in the core. Units II and IV are characterized by shorter wavelength variations between Fe-Ti-rich (i.e., oxide rich) and Fe-Ti-poor (oxide poor or oxide free) gabbroic rocks. The scale of sampling downhole for bulk rock measurements varies from <0.5 to 0.8 m; however, the average sampling is ~0.77 m over the 143-m cored interval. In coarsely sampled regions, peaks and troughs in the cryptic chemical variations of Mg#, Fe2O3, TiO2, and incompatible elements Cecn and Zr show that inflections in trends are often centered on a single sample in both peaks and troughs. Where sampling intervals are denser, however (e.g., at the Unit III/IV boundary and within select intervals in Unit II), smoother transitions delineate peaks and troughs. This seems to indicate that, where resolved by a higher sampling rate, boundaries between cryptic units are not abrupt but vary smoothly. In addition, Mg# trends show both "normal" fractionation (Mg# decreasing uphole) and "inverse" fractionation (Mg# increasing uphole) (e.g., base to top of Unit IV and 62–72 mbsf). This is an interesting observation because downhole variations indicate Fe-Ti-rich gabbros are intimately mixed with relatively primitive olivine gabbroic rocks on the scale of a few meters or less, similar to variations observed in Hole 735B. Sampling in Hole 1105A was generally conducted on a finer scale than sampling through the oxide gabbro interval of Hole 735B during Leg 118 (Dick et al., 1991). Natland and Dick (2002) and Dick et al. (2002) described the complex interlayering of oxide and olivine gabbro in Hole 735B, and this is very obvious on the FMS image logs (Miller et al., Chap. 3, this volume; Zarian, 2003) in Hole 1105A. However, these mixtures in Hole 735B have typically been interpreted to indicate that ferrobasaltic melts infiltrated through a preexisting more primitive olivine gabbroic section rather than being cogenetic local derivatives and products of fine-scale crystallization (e.g., thermomechanical boundary layer and in situ crystallization phenomenon). Natland and Dick (2002) interpreted the formation of oxide gabbro to be the result of "buoyant intercumulus melts and their coalescence and ascent in open fractures and shear zones." They interpreted oxide gabbros to result from numerous intrusions along permeable pathways, often associated with ductile deformation zones. Why presumably massively oxide-precipitating and dense melts (ferrobasaltic) rise buoyantly within shear zones with respect to other less dense primitive or highly fractionated iron-depleted granophyric interstitial melts is unclear. The smooth transitions revealed by small-scale sampling in Hole 1105A could indicate that in situ crystallization processes may also be a viable mechanism to produce the fine-scale interlayering observed in both holes. It is clear that an even finer scale of sampling will be necessary to more fully resolve the true nature of cryptic chemical variation and the nature of the contacts between oxide-rich and oxide-poor gabbroic rocks. The transitional nature of these contacts will be further explored below in discussions of mineral chemistry and structural relationships.
The downhole plot in Figure F25 also shows that Mg# trends are typically inverse to Fe2O3 and TiO2 trends, as expected. In addition to major element downhole variations, incompatible elements Cecn and Zr are displayed downhole. Note that oxide-rich gabbroic rocks (high Fe2O3 and TiO2) appear to correlate with Ce and Zr abundances, but the correlations are far less regular than major element variations. In fact, even in intervals where the major element abundances are nearly constant, trace element abundances of Zr and Ce vary significantly. In addition, values of these trace elements in some oxide-rich gabbroic rocks as well as oxide-poor high-Mg# gabbro can exceed minimum values of the Atlantis II basalts (pink line on Cecn plot). These rocks typically have positive Zr anomalies, suggesting zircon accumulation is common. Most of the more primitive high-Mg# gabbroic rocks, however, are commonly significantly more depleted in these incompatible trace elements than minimum basaltic values (Johnson and Dick, 1992), again reflecting their cumulate nature. The plot does, however, illustrate that the major element cryptic chemical variation is not perfectly correlated and is sometimes decoupled from highly incompatible trace element abundances, perhaps consistent with variable percentages of trapped melt components, boundary layer processes, or small late vein systems. In one case within an olivine primitive gabbro that showed a strong positive Zr anomaly, we identified a small silicic vein with a network of zircon crystals that has intruded as a late feature and that obviously significantly altered the bulk rock trace element chemistry. From the abundance of zircon in the vein, it is clear that the zircon accumulates in the vein as a cumulate mineral and imparts a strong positive anomaly to the rock.
Finally, whole-rock Mg# is plotted downhole over roughly similar intervals and at the same scale (Fig. F26) through the oxide-rich Unit IV of Hole 735B, in which 37 samples were analyzed, and a depth-adjusted downhole Mg# plot of Hole 1105A, in which 109 samples were analyzed within oxide-rich Subunits IIA and IIB. The units in each hole are potentially correlatable. Data for Hole 735B were compiled from all available data reported in the Leg 176 Scientific Results volume (Natland, Dick, Miller, and Von Herzen, 2002) and are similar to those depicted in Dick et al. (2002). However, the scale of variability may be deciphered more precisely by the increased sample density in Hole 1105A. The differing sample scales makes direct comparisons between Hole 735B and Hole 1105A more difficult. The more closely spaced sampling intervals in Hole 1105A allow us to more simply identify the scale of cryptic chemical units, although even finer scale sampling may be required to fully document all of the forward and inverse trends between oxide-rich and oxide-poor cryptic chemical units. These units appear cyclic on scales of 5 to 10 m through Subunits IIA and IIB of Hole 1105A. One peculiarity throughout the hole is that the Mg#s of the most primitive gabbroic rocks are similar from top to bottom, perhaps corresponding to the average composition of the chamber interior. Similar scales of variability were suggested by Dick, Natland, Miller, et al. (1999) for Hole 735B, but oxide-rich cryptic units were interpreted to be sill-like intrusions that split preexisting olivine gabbro as they were sequentially injected into the olivine gabbro. It is important to note also that Dick, Natland, Miller, et al. (1999) were able to recognize much larger scale cryptic chemical units, at a scale of hundreds of meters, which they termed rock "series." These series were defined by clear offsets in the whole-rock Mg#s of the most primitive rocks and apparent upward normal fractionation trends. The normal and inverse cryptic variation described in Hole 1105A on a smaller scale would likely superimpose on this larger scale of variation.