Oxides and sulfides were lumped together as opaque minerals in point counts of gabbros from Hole 735B (Shipboard Scientific Party, 1989, 1999b); therefore, proportions of ilmenite to magnetite in the rocks were not determined by this means. In reflected light, the proportion of ilmenite to magnetite in oxide gabbros was estimated (not point counted) to be at least 3:1 but it reaches 5:1 in some samples. Since the oxides are coarse grained in many of these rocks, no great stock should be taken in this variance; only the high proportion of ilmenite to magnetite is important. At the extreme, a few oxide gabbros are extensively hydrothermally altered. In these cases, all, even secondary, magnetite is completely replaced by green amphibole but ilmenite remains in grids, relict from an original oxyexsolution intergrowth with magnetite.
Ignoring such altered samples, however, how consistent is the proportion of the two minerals? Among oxide gabbros, bulk-rock compositions show that the proportion is very consistent. In the ternary of Figure F32A, the proportion of the two minerals in oxide gabbros can be estimated from the ratio of TiO2 to FeO, using samples in which the latter was measured by titration (Shipboard Scientific Party, 1989; 1999b). The third component in the ternary, SiO2 that is combined in silicate minerals, behaves simply as a dilutant with respect to the oxide minerals, in which TiO2 and FeO are so concentrated. Samples plotted are least-altered gabbros, having either positive or small negative loss on ignition (LOI > -2%). This procedure screens out samples with high Fe2O3/FeO, which were oxidized during alteration, plus those that have abundant secondary amphibole containing structural Fe2O3. Positive LOI indicates weight gain, mainly by addition of oxygen to FeO, for example, that is present in the oxide minerals, during ignition at 1000°C—the first step in preparation of rock powders for XRF analysis. Such samples are extremely fresh, and many oxide gabbros analyzed on board ship have positive LOI.
Least-altered olivine gabbros and troctolites (234 samples with FeO determined by titration) have so little TiO2 that they plot very close to the SiO2-FeO sideline in Figure F32A, as do average clinopyroxenes and amphiboles, taken from electron-probe analyses of minerals compiled by Dick et al. (Chap. 10, this volume). Almost all oxide gabbros fall along a single trend pointing toward the basal TiO2-FeO sideline between the plotted positions of average ilmenite and magnetite (data from Natland et al., 1991). The dashed sides of the shaded triangle define simple mixing trends between the pure oxide minerals and an apex olivine gabbro with few or no oxide minerals. The apex corresponds to the intersection of the trend for oxide gabbros with the sideline trend for olivine gabbros and troctolites. The lever rule applied to this triangle indicates that the proportion of ilmenite to magnetite in the oxide gabbros is 4:1 with small scatter. However, two least-altered samples plot well to the right of the main trend of oxide gabbros. These are olivine clinopyroxenites—cumulates rich in Fe-Mg silicates rather than plagioclase—with a very small proportion of oxide minerals. Such rocks are minor in abundance in the core (Dick et al., 2000). However, they demonstrate that to whatever extent any sample contains FeO in silicate minerals, it will plot closer to the right sideline of the ternary. Since most oxide gabbros do not do this and instead fall along a single, well-defined trend, this means that they have close to cotectic proportions of Fe-Mg silicates and plagioclase and that the oxide minerals they contain are effectively a pure cumulus addition to the rocks (cf. Natland and Dick, 2001), with ilmenite exceeding magnetite in the minimum proportion of 4:1.
Since olivine gabbros and troctolites hug the right sideline of the ternary close to the SiO2 corner, however, their bulk compositions are determined almost entirely by silicate minerals. Plagioclase plots virtually at the SiO2 corner of the ternary, of course, and olivine and orthopyroxene fall almost exactly along the sideline. Of the principal silicates, only average clinopyroxene has a small amount of TiO2 (0.61%). Average amphibole has only a bit more TiO2 than this (1.14%). Both plot between the oxide gabbros and the SiO2 corner of the ternary, thus they do virtually nothing to deflect the more titanian and iron-rich trend among oxide gabbros. That trend instead is controlled almost entirely by the total amount and fixed proportion of ilmenite to magnetite in those rocks.
The TiO2 contents of olivine gabbros and troctolites also correlate with MgNo (Fig. F32B), even though TiO2 is sensitive both to the proportion of plagioclase in the rocks and to the presence of even a very small amount of ilmenite. The correlation thus indicates that most of these rocks contain nearly cotectic proportions of plagioclase to Fe-Mg silicates (Shipboard Scientific Party, 1999b) and almost no magmatic ilmenite or magnetite. The average normative proportion of plagioclase in these rocks, for example, is 57 ± 7, a very tight clustering. As shown in Figure F32B, ilmenite itself can cause only a trifling shift in bulk-rock MgNo, and there is clearly insufficient magmatic magnetite to do this if it is consistently intergrown with four times its abundance of ilmenite. Instead, the rocks are adcumulates, with extremely low proportions of material crystallized from residual interstitial liquid from which the late-crystallizing oxide minerals could have precipitated (Natland et al., 1991; Natland and Dick, 2001). The oxide minerals therefore constitute no more than a few tenths of a percent of the mode of olivine gabbros and troctolites. The very low modal proportions of magmatic oxides in such rocks are a consequence of the extreme efficiency of expulsion of residual liquids during the compaction, deformation, and final stages of crystallization of these rocks.
There are thus two broad classes of rock among the gabbros of Hole 735B, one with significant and in some cases very abundant magmatic oxides and the other almost entirely without magmatic oxides. The two correspond to the division between those rocks providing susceptibility peaks and peak regions on the one hand and those with background magnetic susceptibilities on the other.
Figure F33A shows the relationship between TiO2 contents of gabbros from Hole 735B drilled during Leg 176 and their magnetic susceptibility, in each case measured at one point within the length of a sample taken from whole-round core and analyzed by XRF on board JOIDES Resolution. Each XRF sample was usually a slab or quarter-round piece of core ~5 cm long. The figure is a log-log diagram, chosen in order to amplify the scale of variability among samples with low magnetic susceptibility and low TiO2 contents, mainly olivine gabbros and troctolites. Both linear and power-law regressions have correlation coefficients of r = ~0.9. On a log-log diagram, the linear regression is curved and the power-law regression is a straight line. The two virtually coincide for rocks having TiO2 contents between 0.2% and 1.0%.
Magmatic ilmenite, in which most of the TiO2 in oxide gabbro resides, does not have a high magnetic susceptibility. However, since ilmenite is intergrown with high-susceptibility magnetite in a consistent proportion of about 4:1, this clearly provides the strong correlation between magnetic susceptibility and TiO2 contents among these rocks. Of course, the strength of the overall correlation is provided mainly by oxide gabbros. Considering gabbros with <0.7% TiO2, a linear correlation is weaker (r = 0.58). As just demonstrated, however, most of the variability in TiO2 contents among olivine gabbros is controlled by clinopyroxene, not the oxide minerals, because the rocks are adcumulates. Troctolites have very little clinopyroxene and both lower TiO2 contents and lower magnetic susceptibility than olivine gabbros (Fig. F33A). Thus all rock types contribute to the overall strong correlation, even though the amount of intergrown magmatic ilmenite and magnetite in olivine gabbros and troctolites is inconsequential.
Olivine gabbros and troctolites also show a fairly strong correlation between magnetic susceptibility and bulk MgNo of the rocks (Fig. F33B). Here, some of the scatter about the correlation results from the different volumes and geometries of rock material sampled by the Bartington sensor and for XRF analysis. Perhaps there is also a slight effect of variable but small amounts of intergrown magmatic oxides. Some other factor, however, overrides these effects among olivine gabbros and troctolites and provides the correlation. One of the other types of magnetite in the rocks mentioned earlier must be responsible. Since least-altered rocks were used to establish the correlations, magnetite associated with secondary amphibole is probably not important, although it may also contribute somewhat to the scatter about the correlation in Figure F33B and to some variability in susceptibility along the entire section drilled.
Magnetite exsolved from primary silicates, including plagioclase, is therefore left as the oxide mineral most likely causing the principal variability of background measurements of magnetic susceptibility among olivine gabbros and troctolites. The correlation between MgNo and magnetic susceptibility just among these rocks suggests that the proportion of such magnetite depends on the composition of the silicate minerals being higher in rocks with more iron-rich pyroxenes (similarly olivine) and more sodic plagioclase.
In almost all troctolites and olivine gabbros, tiny dark inclusions are present in all three of the principal silicates, olivine, pyroxene, and plagioclase. Some are secondary inclusions arrayed along fractures (Fig. F34A-F34F), some are along cleavage planes, and some follow crystallographic partings (Fig. F34E-F34G), but on the usual scale of observation they were rarely noted in shipboard descriptions. Migrating fluids or melts may have precipitated the inclusions along the fractures; alternatively, these were produced by reaction with those fluids. The other two types are present along cleavage planes and partings. They more likely formed by simple exsolution. Tiny opaque grains are also present at grain boundaries between plagioclase neoblasts and also evidently formed by exsolution. The tiny inclusions are usually completely surrounded by host or adjacent silicates in the 30-m thickness typical of thin sections; only those few intersecting the polished surfaces of the thin sections were brought out as pin-point reflective surfaces by routine polishing (Fig. F34B, F34D, F34F, F34G). Magnetite inclusions in plagioclase are usually rodlike in form, parallel to crystallographic axes, and pale brown in color, which is appropriate for very tiny magnetite grains. In some plagioclases, exsolved magnetite forms spectacular grids paralleling crystallographic axes (Fig. F34H).
Plagioclase may contribute significantly to the total amount of magnetite in olivine gabbros and troctolites. Plagioclase phenocrysts in abyssal tholeiites from the Indian Ocean contain as much as 0.2%-1.2% FeO, increasing as An decreases (Fig. F35). Presumably similar amounts of FeO originally partitioned into plagioclases of gabbro cumulates of Hole 735B. At similar An values, plagioclase in gabbros from Hole 735B now contains only 0.04%-0.30% FeO. The feldspar therefore appears to have lost between 0.2% and 0.6% of iron as FeO during subsolidus reequilibration, with the greater amount being lost from more sodic plagioclase. By themselves, the electron-probe analyses do not say where the missing iron went, but from petrography, it probably is now tied up in exsolved magnetite either within the crystals or at the interfaces between them. There should also be more exsolved magnetite in plagioclases of the more differentiated olivine gabbros than in those of troctolites. Acicular iron oxide exsolved from plagioclase has described from some layered intrusions and elsewhere (Sobolev, 1990; Tegner, 1997; Selkin et al., 2000).
MgNo, however, is a ratio and independent of the variations in the modal proportion of plagioclase to olivine and pyroxenes. For it to correlate with magnetic susceptibility among troctolites and olivine gabbros, either the amount of exsolved magnetite in a given rock does not differ between Fe-Mg silicates and plagioclase, or it does, but the proportion of plagioclase in the rocks is nearly constant, and exsolution has proceeded to the same degree in all rocks. Average modal proportions of olivine gabbros measured during Leg 176 (Shipboard Scientific Party, 1999b) are ol/plag/cpx = 9.8/59.2/29.2. These are very close to experimentally determined proportions (11/59/30) (Grove and Baker, 1984; Grove et al., 1992; Toplis and Carroll, 1995). The proportion of normative plagioclase (57 ± 7, as mentioned earlier) is virtually the same as the proportion of plagioclase in the mode. In troctolites, the cotectic proportions are ol/plag = 30/67, also very similar to the modal proportions measured in such rocks during Leg 176. There is thus no simple way of telling whether plagioclase or Fe-Mg silicates have more exsolved magnetite. Both mineral groups have some, so they both must contribute to the signal of magnetic susceptibility of these rocks.
In summary, whether or not there is more exsolved magnetite in Fe-Mg silicates than in plagioclase, the overall amount of it correlates with a cryptic compositional parameter of the olivine gabbros and troctolites, namely their MgNo, and this provides a signal of magnetic susceptibility increasing with the extent of differentiation of these rocks. Second, because most the rocks also have nearly cotectic proportions of silicate minerals, this also provides a somewhat weaker correlation between magnetic susceptibility and the bulk TiO2 content of the rocks, this being provided by their nearly fixed proportion of clinopyroxene. In rocks with >0.7% TiO2 contents, the susceptibility signal is dominated by magmatic ilmenite and magnetite, intergrown in nearly fixed proportions in the rocks.
Cryptic variation is a term coined by Wager and Deer (1939) to describe variations in gabbro composition that are independent of the modal proportions of minerals. Mineralogically, the Fo content of olivine, En/Fs of pyroxenes, and An content of plagioclases are all cryptic parameters. So also in a general way are bulk-rock MgNo and normative An/(An + Ab). If a cumulus mineral assemblage is quite pure with a small proportion of material crystallized from trapped residual liquid, then its mineralogical and even bulk-chemical cryptic parameters can be related to the liquid compositions that produced the cumulates. At issue now is whether magnetic susceptibility is a measure of one of the most important cryptic parameters of the gabbros from Hole 735B, namely MgNo. We know that the high magnetic susceptibilities of oxide gabbros reflect their proportion of magmatic oxides. This is not a cryptic property of the rocks. What about olivine gabbros and troctolites?
One test is to find out whether susceptibility correlates with the variability of a silicate mineral, for example, the Fo content of olivine. Figure F36A plots compositions of olivine vs. magnetic susceptibility from the Leg 176 portion of the core (olivine data from Dick et al., Chap 10, this volume). Although there are weak linear and power-law correlations, magnetic susceptibility still spans two orders of magnitude at some olivine compositions (e.g., Fo70). The two do not correlate. Thus there are many susceptibility peaks representing oxide gabbros in which the oxide minerals have apparently simply been added to an underlying and preexisting matrix of olivine gabbro. Some of these, identified in the core descriptions, contain so many magmatic oxides that they were not even identified as olivine gabbro. There are several cases of two samples only a few centimeters apart in the core, one being an oxide gabbro that provided a spike in magnetic susceptibility of several thousand x 10-6 MU, the other being an olivine gabbro with background magnetic susceptibility of only ~100 x 10-6 to 300 x 10-6 MU. Yet the olivine in the oxide gabbro is even more forsteritic than in the adjacent olivine gabbro. On the contrary, some rocks with olivine as iron-rich as Fo55 have very low magnetic susceptibility.
This result is so widespread that it can be used to outline a general argument about the petrogenesis of the rocks cored during Leg 176. First, the oxide minerals overall had to crystallize from liquids more differentiated—more iron rich—than the liquids that produced even the most iron-rich olivine in any of them (Fo46). Such liquids are represented by some of the even more differentiated oxide gabbros from the upper 500 m of the hole, drilled during Leg 118, which have olivine as fayalitic as Fo30 (Ozawa et al., 1991). Second, iron-rich liquids almost always were efficiently expelled from the intercumulus matrix of all gabbros that crystallized olivine ranging from Fo84 to Fo46 to produce adcumulates with very little iron-rich interstitial melt remaining. Third, not only did the expelled liquids aggregate and reintrude the section en masse at those places that now show up as peak regions in magnetic susceptibility, some of them nearly 2 m thick, but they also injected the intercrystalline porosity structure of primitive gabbros at hundreds of places up and down the core, producing a host of narrow spikes in magnetic susceptibility but without significantly changing the mineralogy of their new hosts. Preexisting olivines, for example, were not completely destroyed by reaction with the penetrating melts. Reaction with the injected liquids may have left those olivines embedded in the common orthopyroxene coronas seen adjacent to them in many of these rocks. In this way, the original cryptic variability of the olivine was maintained, even if the rock itself became infested with oxide minerals and cannot now be identified as an olivine gabbro. Some very unusual rock types, such as the disseminated-oxide troctolite mentioned earlier, resulted from this process.
As complicated as this might seem, all is not lost. The reaction process may have occurred in portions of rock that are discrete enough that we can still consider the cryptic variation of olivine among just olivine gabbros and troctolites, even including some rocks that experienced oxide enrichment, and compare it with the background magnetic susceptibility. Figure F36B shows both background magnetic susceptibility and average composition of olivine per sample vs. depth. The plotted olivine compositions are the same as those within the box labeled "Selected" in Figure F36A. They are screened to include all olivine compositions more magnesian than Fo60 with magnetic susceptibility <1000 x 10-6 MU. This removes rocks as differentiated as olivine gabbronorites from the comparison. Among all remaining rocks, it assumes that the differential between background susceptibilities and 1000 x 10-6 MU resulted from addition of a small proportion of oxide minerals to the rocks, without destroying the host olivine completely. In other words, the rocks originally had a lower proportion of oxide minerals, and after freezing, would have produced magnetic susceptibilities within background limits. Instead, they lay in the path of a percolating, iron-rich melt. Figure F36B includes 10% weighted curves for both magnetic susceptibility and olivine compositions, scaled so that the two curves are superimposed.
Because vertical scales were chosen so that the amplitudes of fluctuations along the two weighted curves are similar, the scatter of olivine compositions falls almost entirely within the band of background magnetic susceptibility. Although there are far fewer olivine compositions than measurements of magnetic susceptibility, when similar window averages are used then some of the similarities of the two weighted curves appear to be significant. The first is that the fault at 1100 mbsf, marked by a sharp drop in background susceptibility among the rocks deeper than this, is also reflected in more forsteritic olivine compositions. There are also similar general fluctuations in the two curves, notably rises in amplitude between 900 and 1100 mbsf, downturns in both curves to very low susceptibilities and high Fo contents at ~830 mbsf, and perhaps a small upturn in both curves at ~1350 mbsf. Note that the downturn in the olivine curve at ~820 mbsf probably would not exist without inclusion of forsteritic olivine from just two samples. This underscores the difficulty of determining the true pattern of cryptic mineralogical variation in this complicated core from mineral data using samples spaced on the average 1-2 m apart.
Between 500 and 700 mbsf, the two curves do not correspond. Partly this resulted from selected sampling of troctolites used for microprobe analysis at ~510 mbsf. These provided several magnesian olivines, but adjacent, more differentiated, and typical gabbros from this part of the core were not sampled. From 550 to 650 mbsf, core recovery was somewhat low because of the presence of fractured and hydrothermally altered rock in two closely spaced fault zones. The alteration particularly affected olivine; those samples in which it was preserved may not represent the typical lithologies in this part of the core.
I conclude, somewhat guardedly, that background magnetic susceptibility in general reveals the pattern of cryptic variability among olivine gabbros and troctolites along most of the core recovered during ODP Leg 176. Since almost all of these rocks are adcumulates, this means that expulsion of intercumulus melts, the residual from which the oxide minerals precipitated, was both efficient and fairly uniform throughout the core. Some of the local scatter in background magnetic susceptibility, but not its weighted variability on a scale of 50 m or more, may indicate variations in the percentage of interstitial melt that was present when the oxide minerals crystallized. Although the core was riven with many late-stage oxide-rich seams, magnetic susceptibility reveals the occurrence and width of those seams and the maximum extent of the core that was refertilized with iron-rich melt. All such rocks have been screened from Figure F36B. The figure is thus a map of the distribution of adcumulates in the core rocks from which interstitial melt was efficiently squeezed, amounting to nearly 63% of the section. It also gives a general idea of how differentiated those rocks happen to be. Some fairly strongly differentiated gabbronorites are also adcumulates with low percentages of oxide minerals and background levels of magnetic susceptibility. These are included in Figure F36B.
For oxide-bearing and oxide-rich gabbros, the utility of magnetic susceptibility as a geochemical index is much more straightforward. Simply considering TiO2 contents, a crude breakdown is quite meaningful. Thus olivine gabbros and troctolites have <0.7% TiO2; disseminated-oxide gabbros have 0.7% < TiO2 < 2.0%; and oxide gabbros have >2.0% TiO2. The bulk of analyzed samples within these limits have clearly different magnetic susceptibilities. Presence or absence of the different silicate minerals allowed discrimination of particular lithologies. In gabbros with <0.7% TiO2 content, most of the TiO2 resides in clinopyroxene and is simply influenced by olivine, orthopyroxene, and plagioclase, which have very low TiO2 contents, acting as dilutants. The proportion of clinopyroxene to these other minerals, however, oscillates only a small amount from cotectic proportions, and this ultimately allows us to use magnetic susceptibility as a geochemical log even among these rocks.
Above 0.7% TiO2 contents, however, the trend produced simply by the low-TiO2 silicate minerals is deflected by an increase in the mode of oxide minerals (Shipboard Scientific Party, 1999b). The proportion of TiO2 in the oxide minerals exceeds the amount tied up in clinopyroxene. Addition of cumulus ilmenite and magnetite becomes paramount and is independent of the underlying composition of silicate minerals. Natland and Dick (2001) describe this as a process of accumulation of the two oxide minerals as they precipitated in small-scale porosity structure in a matrix, or mush, of silicate minerals during flux of highly differentiated liquids rich in iron and titanium. The effectiveness of formation of cumulus oxide minerals was strongly controlled by patterns of porosity structure that developed in response to deformation of the rocks. Some aspects of this have been developed in this paper.
The linear regression of Figure F33A can be used to estimate aspects of the bulk composition of the core. This obviously applies only to the three general classes of gabbroic rocks just mentioned. Neither Figure F33A nor F33B includes compositions of felsic veins (quartz diorites to granites) or hybrids between felsic veins and country rocks (diorites). However, these comprise only ~0.3% of the core; they do not significantly affect the result outlined below.
Dick et al. (2000) calculated average compositions for each 500-m portion of Hole 735B. Their procedure corrected the lengths of lithologic intervals for instances of both low recovery and >100% recovery and used average compositions of each lithologic type, weighting them properly for differences in density. Recovery of >100% occurs when the drill recores a stump of rock left from a prior core and then adds a full 9.7 m of additional rock to it. I wished to check the validity of this approach using magnetic susceptibility, given that this measurement allows a greatly refined estimation of the proportions of rocks having different compositions throughout the core recovered during Leg 176. My analysis only concerns TiO2 contents, since this oxide is one that I can relate most simply and directly to magnetic susceptibility. In the end it fairly strongly validates the estimates of Dick et al. (2000).
The procedure I used for magnetic susceptibility was to obtain the average value per core, relate this as quantitatively as possible to the TiO2 contents of analyzed rocks, and consider that the average estimated TiO2 contents represent the depth range for that core given in the site report (Shipboard Scientific Party, 1999b, table T1). The average TiO2 content of each core was calculated using the linear regression in Figure F33A. For the density correction, I calculated a linear regression for the relationship between normative density calculated from bulk XRF compositions (based on Niu and Batiza, 1991 with some additional mineral densities from Deer et al., 1992) and TiO2 contents. The relationship for 323 chemical analyses is D = 0.68 x TiO2 + 3.013 (r = 0.87). The range of normative densities is 3.013-3.120 g/cm3, slightly higher than measured densities for minicores because as measured on deck, the latter usually include tiny cracks and other microporosity structure, reducing their densities. Only relative differences in normative density among samples have any bearing on this calculation, however, and normative densities follow differences in measured densities very closely. The average density-corrected TiO2 content for each core was then divided by the total depth range under consideration (part or all of the 1003.2 m cored during Leg 176) and the values summed over that interval to provide the average TiO2 content for the interval.
The average TiO2 content calculated in this way for the interval from Cores 176-735B-88R through 153r (504.8-1005.3 mbsf) is 0.71%, nearly the same as the value (0.70%) obtained by Dick et al. (2000) for lithologies from 500 to 1000 mbsf. From Cores 176-735B-154R through 210R (1005.3-1508.0 mbsf), the average is 0.41%, a bit lower than the 0.50% obtained for the lowermost 500 m of the hole by Dick et al. (2000). The values for the two parts of the hole should correspond to differences in other correlative attributes of composition (Mg#, Ca#, etc.) cited by Dick et al. (2000) as being related in general to the degree of differentiation of basaltic liquids from which these cumulates crystallized. The deepest 500 m of the hole, and certainly the body of rocks below the fault at 1100 m, is a slightly more primitive body of gabbro than suggested by the estimate based on identification of lithologic intervals in the shipboard descriptions. Given the uncertainties implicit in either of the estimates by lithology or magnetic susceptibility, I consider that the overall agreement is quite good and that the estimates of Dick et al. (2000) for TiO2, all the other oxides, and the several trace elements are not likely to differ significantly from a more extensive treatment based on magnetic susceptibility.