Chemical compositions of primary basaltic magmas beneath the AAD (i.e., magmas equilibrated with the upper mantle) were back calculated from basalt compositions based on the "olivine maximum fractionation" model (e.g., Tatsumi et al., 1983; Yamashita and Tatsumi, 1994; Yamashita et al., 1996; Tamura et al., 2000). In this model, the chemical composition of olivine in equilibrium with basalt was first calculated based on both Fe-Mg and Ni-Mg exchange partitioning between olivine and silicate melt. The calculated olivine composition was then added to the original whole-rock composition in a weight ratio of 1:99, and this was repeated until the equilibrium olivine had an NiO composition equivalent to that of mantle olivine. The olivine composition that was calculated in each step was fractionated from an original primary magma to yield the present basalt composition. Furthermore, whole-rock composition generated by the same number of calculation steps as olivine composition that reached mantle composition represented the melt composition that was finally equilibrated with mantle olivine (i.e., primary magma).
Assumptions involved in the calculation are as follows:
This calculation assumes that only olivine was fractionated from magma prior to eruption. If plagioclase was fractionated, the calculation gives an overestimation of the value of MgO in primary magma. On the other hand, if clinopyroxene was fractionated, the calculation gives an underestimation of the value of MgO in primary magma. Therefore, a starting composition that is likely to have only crystallized olivine must be chosen before such a calculation can be performed.
Figure F4 shows major element variation diagrams for basaltic glass from the AAD collected during Leg 187 (C. Russo, unpubl. data). On a CaO/Al2O3 vs. MgO diagram (Fig. F4C), CaO/Al2O3 values of Pacific-type MORB are scattered, whereas CaO/Al2O3 values of Indian-type MORB are relatively constant at >8 wt% MgO and increase at <8 wt% MgO. On an Na2O vs. MgO diagram (Fig. F4B), Indian-type MORB has relatively constant values at >8 wt% MgO and increases with decreasing MgO; Na2O increases with decreasing MgO for Pacific-type MORB. The FeO vs. MgO diagram (Fig. F4A) shows similar tendencies as the Na2O vs. MgO diagram. FeO of Indian-type MORB is relatively constant at >8 wt% MgO, then increases with decreasing MgO; Pacific-type MORB shows a systematic increase in FeO with decreasing MgO.
In accordance with the assertions of Klein and Langmuir (1987), major element variations for Indian-type MORB indicate plagioclase fractionation prior to eruption for basalt with <8 wt% MgO, whereas Indian-type MORB with >8 wt% MgO is considered to fractionate only olivine. Therefore, only basalts containing >8 wt% MgO are suitable for calculation of Indian-type MORB based on the olivine maximum fractionation. For Pacific-type MORB, it is concluded that all basalts had fractionated plagioclase. However, rare earth element (REE) chondrite-normalized patterns for Pacific-type MORB reported by C. Russo (unpubl. data) indicate that basalts from Site 1160 have the lowest REE concentrations. Furthermore, basalts from Site 1160 have smaller FeO/MgO (close to 1) and relatively high mg# (= Mg/[Mg+total Fe]) than other Pacific-type MORB (Table T2). These lines of evidence suggest that basalts from Site 1160 are not as highly evolved as other Pacific-type MORB. In this manuscript, the olivine maximum fractionation calculation can be applied to only basalts from Site 1160 for Pacific-type MORB, with the assumption that the effect of fractionation of plagioclase is lowest for these samples. It should be recognized that any estimates of primary MgO contents for these samples are slightly overestimated.
Based on the criteria mentioned above and proximity of analyzed samples in this study and Shipboard Scientific Party (2001) and/or C. Russo (unpubl. data), five samples were chosen from Indian-type MORB and two samples were chosen from Pacific-type MORB as starting compositions for the calculations. Each chosen sample belongs to the same lithologic unit as that containing the analyzed olivine grains. The chemical compositions of these samples are listed in Table T2.
Figure F5 shows Mg#-NiO plots of analyzed "actual" olivine composition (phenocryst center/peripheral, microphenocryst center/peripheral, and groundmass grains) for selected basalt samples. Also shown in each diagram is the Mg#-NiO variation for mantle olivine (mantle olivine array) as well as the "back-calculated" trend of calculated "equilibrium" olivine compositions. If the olivine phenocrysts in a given basalt represent crystals that crystallized from a melt of the same composition as a given basalt, they should coincide with the low-Mg# end of the back-calculated trend. This appears to be the case for Pacific-type MORB, whereas it clearly not the case for most of Indian-type MORB, which contains many olivine crystals that have more magnesian compositions than expected. In particular, Sample 187-1155A-5R-1, 13-17 cm, exhibits a significant disequilibrium between actual olivine phenocrysts and calculated olivine compositions. This disequilibrium could be due to (1) precision of shipboard whole-rock analysis, (2) whole-rock compositions that are not representative of the bulk composition of the lava, or (3) another origin for phenocrysts. Whole-rock composition for Sample 187-1155A-5R-1, 20-23 cm, was analyzed by the inductively coupled plasma-atomic emission spectrophotometer (ICP-AES) on the JOIDES Resolution during Leg 187. Because the total composition for Sample 187-1155A-5R-1, 20-23 cm, exceeds 100 wt% (102.20 wt%) (Shipboard Scientific Party, 2001), the problem of accuracy during analysis still remains. However, the difference between actual olivine phenocrysts and calculated olivine compositions is too large to be explained solely by analytical precision. Because of rapid quenching of the margins, flow differentiation, crystal setting, and many other factors, small subsamples taken from different parts of the lava flow can have very different compositions. It is very difficult to determine what part of the flow was sampled during Leg 187 because of low recovery. Therefore, although samples for whole-rock analysis belong to the same lithologic unit as that containing the analyzed olivine grains, it is plausible that they do not completely have the same compositions. Basalts from other sections of Hole 1155A (187-1155A-2R-1, 14-16 cm) have larger plagioclase phenocrysts with compositionally reversed zoning. Relatively smaller and abundant plagioclase phenocrysts have normal compositional zoning. Therefore, larger plagioclase phenocrysts in basalts (Sample 187-1155A-2R-1, 14-16 cm) are considered to be xenocrystic. Large olivine and plagioclase phenocrysts form glomeroporphyritic assemblages in Sample 187-1155A-2R-1, 14-16 cm. Thus, the olivine phenocrysts are also considered to be xenocrysts. Compositional variation in olivine grains in Sample 187-1155A-2R-1, 14-16 cm, is very similar to that in Sample 5R-1, 20-23 cm, suggesting that olivine phenocrysts in the latter sample, are also xenocrysts. These observations may explain disequilibrium between "real" and calculated olivine compositions in Indian-type MORB, suggesting that many of the Mg-rich olivine phenocrysts in Indian-type MORB are, in fact, xenocrysts.
Because the length of the back-calculated trend indicates the extent of olivine fractionation from an original primary magma required to produce a given basalt, Pacific-type MORB has undergone much more extensive preeruption fractionation than Indian-type MORB. For Indian-type MORB, calculated olivine compositions reach the "mantle olivine array" after zero to four steps of calculation, whereas four to seven steps are needed for Pacific-type MORB. Groundmass olivine grains of Sample 187-1160B-4R-1, 48-52 cm, have different compositions from calculated olivine equilibrated with melt at the final stage of fractionation. This disequilibrium can be explained if crystallized olivine phenocrysts and/or microphenocrysts with compositions of Fo85-87 were removed from the magma just before eruption.
The diagrams in Figure F5 can be used to estimate the major element composition of the original, prefractionation, primary melt for each basalt as mentioned above. Compositions of the estimated primary melt and olivine equilibrated with mantle are listed in Table T3. Although differences in olivine compositions between Indian- and Pacific-type MORB are small, olivine compositions calculated for Indian-type MORB have lower Mg# and NiO contents (86.8 ~ 88.5 and 0.296 ~ 0.339, respectively) than those calculated for Pacific-type MORB (88.9 ~ 90.2 and 0.342 ~ 0.351, respectively).
Estimated primary magma compositions for both Indian- and Pacific-type MORB were plotted on an olivine-plagioclase-quartz (Ol-Pl-Qz [silica]) diagram of Walker et al. (1979) overlain with the isobaric liquid compositional trend of lherzolite determined by Hirose and Kushiro (1993) (Fig. F6). The locations of these primary magmas in Figure F6 are considered to represent the pressure (i.e., depth) where melt was fully equilibrated with mantle material or where melt was produced. Calculated compositions of primary magma in this study indicate that Indian-type MORB was equilibrated with mantle materials (or was produced) at pressures of ~10 kbar (~30 km depth), whereas Pacific-type MORB was equilibrated at ~15 kbar (~45 km depth).