The major elements presented here agree closely with the shipboard measurements and fairly clearly reflect the igneous evolution of the region surrounding the hole. The Mg numbers (Mg2+/[Mg2+ + Fe2+]) of the gabbros reflect the basic magmatic evolution of the cumulate pile (Fig. F1). They show the interfingering of differently evolved lavas and the reintrusion of the crystal mush by highly evolved iron-rich magmas. Figure F1 shows the evolution of the Mg number with depth in the hole. The data from the current study agree well with the shipboard measurements and reflect as well the major magmatic segmentation seen in the shipboard results. Also visible in Figure F1 are the ferrogabbros, with significantly lower Mg numbers. In Figure F1, only ferrogabbros from this study are shown. Samples with >0.5 wt% TiO2 have been filtered out of the other two data sets for clarity. The ferrogabbros occur in two varieties: as gabbros bearing disseminated oxide grains and as massive oxides infiltrating the grain matrix of the gabbro.
Ferrogabbro intrusion into the crystal mush sequence is documented by variations in Fe2O3 and TiO2, as shown in Figure F2. The correlation of Fe2O3 with TiO2 is a sign of the involvement of ilmenite and titanomagnetite as crystallizing phases. The lack of a perfect correlation is because of multiple intrusion at different Fe2O3/TiO2 ratios in the intruding magma, intrusion into country rocks of differing Fe2O3 contents, and differing liquidus mineralogy of the intruding magmas.
Trace elements in the Hole 735B cumulate sequence reflect the same two themes just mentioned: magmatic differentiation as a cotectic cumulate followed by reintrusion of highly evolved liquids that crystallize oxides. This sequence can be best illustrated by plotting TiO2 against Ni, as in Figure F3. In all of the Hole 735B gabbros, significant deposition of TiO2-rich minerals appears only to have occurred in rocks with less than ~180 ppm Ni (Fig. F3). This observation has two main implications for the evolution of the magmatic sequence.
The ferrogabbros show a variety of intrusive relationships with the country rock. These range from graded or sutured contacts with more evolved units to sharper contacts with more primitive, more olivine-rich rocks. This suggests that there was a "crystallization wall" at ~180 ppm Ni that represented the state of solidification of the magmatic body as the reintrusion. Thus, the primitive cumulates must have been solid at the time of intrusion of the oxides, whereas the more evolved cumulate assemblages (with a lower solidus temperature) were still partially molten and amenable to infiltration.
The melts must have been autogenous. Had the evolved melts been generated elsewhere and only penetrated the cumulate sequence by chance, then it is unlikely that only evolved units would have been intruded. The nonrandom association of oxide-rich reintrusion with more evolved rocks suggests that instead the oxide-bearing magmas are derived from the cumulate sequence they intrude.
Other oxides than Fe2O3 also correlate with TiO2, some in fact much better. Here such elements are termed titanophile elements and may be able to deliver clues to the origin of the magmas that deposited oxides in the rock. A case in point is V, where the highest concentrations in ferrogabbros are a factor of 8-10 higher in the most Ti-rich rocks than they are in the background gabbros. Figure F4 shows the relationship between TiO2 and V in the Hole 735B gabbros. Although the two elements are clearly correlated through a very large concentration range, the V/Ti ratio is clearly variable. This suggests that the evolved magmas had variable Ti/V or that different minerals played a role in the oxide ores that were deposited. If so, this may have had a variety of causes.
Additional titanophile elements include Zr and Y. Interestingly, Sc correlates well with all of these elements in the less evolved gabbros and microgabbros. This is probably because of the compatibility of all of these elements in clinopyroxene. There is no correlation, however, between Sc and Mg, Mg number, or Ca number. Apparently, Sc is not compatible in oxides, as it is not correlated with Ti (Fig. F5).
In order to assess the contributions of different magmas to the formation of the oxide gabbros, it is instructive to consider the titanophile elements V, Y, and Zr as a "fingerprint" of the melt that crystallized the oxide phases. These elements are all compatible in oxides; thus, just as incompatible elements in basalts record the trace element ratios in their source, these three elements should largely retain their relative concentrations present in the liquid they crystallized from. A constant and regular variation in these elements is visible in the non-ferrogabbros, and it is assumed that this can be subtracted to yield only the excess contained in oxide.
Sc is a particularly useful element for assessing the baseline concentration for the titanophile elements. Like them, Sc is compatible in clinopyroxene, but is not enriched in the oxide phases. Figure F5 shows Sc plotted against TiO2. Sc was measured only in samples from this study, not in either shipboard data set. There is no particular enrichment of Sc in any of the high-Ti samples. Even in the oxide-richest sample at ~15% oxide, the Sc content is nearly within the range of nonoxide gabbros. For example, this means that the TiO2 baseline can very clearly be established. The same is true of the other titanophile elements. When this is done, nearly constant silicate baselines can be established for TiO2 (0.5 wt%), Zr and Y (both 15 ppm), and V (200 ppm).
Some trends in the data can be observed by looking at the ratios of the elements Zr, Y, and V together. The reasoning behind doing so is that different groupings of these elements in the oxide-bearing magmas would tend to "fingerprint" the magmas, possibly resulting in the identification of multiple magmatic events.
A sensible way to do this is to first subtract out the silicate "baseline" of 200 ppm V and 15 ppm each Zr and Y from each sample. Since the variations of these elements are small along the silicate evolution trend, this does not introduce significant scatter in the data. Then each data point is divided by the range (maximum - minimum) of that element in the data set: 1524 ppm for Zr, 130 ppm for Y, and 235 ppm for V. That way each analysis can have a value between 0 (the minimum in the data set) and 1 (the maximum in the data set). These values can thus be considered dimensionless. Figure F6 shows a ternary diagram of the data normalized in this way. In this diagram, a major broad grouping is visible between the bulk of the samples in a broad band across the top of the diagram and a group of four high-V samples. As is clear from Figure F4, these are not the only samples with high V content but they are the only high-V samples where Zr and Y are also low. As there is little fractionation of these elements to be expected from the crystallization of oxide phases, it is possible that these four samples represent a completely separate oxide gabbro magma.
The remainder of the samples are plotted in Figure F7. These samples are grouped by depth in the section in order to see if any trends in relative titanophile content are depth dependent. The normalized analyses group in a broad band that stays close to constant relative V*/Y* (Fig. F7). The samples with the highest relative Zr contents tend to occur in the upper 300 m (lithologic Units X-Z), whereas in the lower part of the section Y and V tend to dominate over Zr. However, the overlap between the different depth intervals makes it difficult to draw firm additional conclusions about potential magmatic sources.
It is important to note that the phases precipitated from the reintruding liquids are themselves cumulates. A residual melt bearing much of the incompatible elements such as K and Rb has clearly been efficiently extracted from the ferrogabbros as well. Few of the ferrogabbros have more than a few tenths of a percent K2O or P2O5. Even though they precipitated from highly evolved liquids, the oxide-rich gabbros must themselves have lost this component to still more highly evolved magma. This magma may be the one recorded in leucocratic and even granitic veins that cut the whole sequence.
A total of, 19 microgabbros were sampled for this study. Particular care was taken sampling the microgabbros, as their texture suggests that they may be liquid compositions and they are overall more mafic than the average for the core. Indeed, several of the microgabbros have elevated contents of P2O5 and K2O, suggesting that they are not cumulates but instead represent liquid compositions. These may thus prove useful starting points for the evaluation of the overall liquid line of descent of the sequence.