| Journal of Petrology | Pages |
© 1998 Oxford University Press |
|
|
Isotopic studies of MORB (mid-ocean ridge basalts) have established the existence of a vast mantle domain in the Indian Ocean distinct from the sources of Pacific and North Atlantic MORB. Along the present Indian Ocean spreading axes, this domain includes the entire Central Indian and Carlsberg ridges and most of the Southeast and Southwest Indian ridges, stretching from about 126°E on the Southeast Indian Ridge (Klein et al., 1988; Pyle et al., 1992) to about 26°E on the Southwest Indian Ridge (Mahoney et al., 1992) and northward into the Red Sea (e.g. Schilling et al., 1992; Volker et al., 1993). Indian MORB are characterized, in particular, by lower values of 206Pb/204Pb relative to [epsilon]Nd and 208Pb/204Pb than Pacific and North Atlantic MORB, and also tend to have comparatively high 87Sr/86Sr (e.g. Hedge et al., 1979; Dupré & Allègre, 1983; Hamelin et al., 1986; Michard et al., 1986; Price et al., 1986; Dosso et al., 1988; Mahoney et al., 1989, , 1992; Hall et al., 1995). When isotopic data for samples from the fringes of the Indian Ocean domain are removed from consideration, there is remarkably little overlap of the Indian MORB data set with the isotopic field defined by >95% of published Pacific and North Atlantic MORB data in either the [epsilon]Nd vs 206Pb/204Pb or 208Pb/204Pb vs 206Pb/204Pb diagrams (Fig. 1). Furthermore, recent studies of Western Pacific back-arc and marginal basin lavas (e.g. Hochstaedter et al., 1990; Loock et al., 1990; Hickey-Vargas, 1991, , 1998; Tu et al., 1992; Crawford et al., 1995; Hickey-Vargas et al., 1995; Spadea et al., 1996), and of island-arc lavas in the Philippines (Mukasa et al., 1987; Castillo, 1996), show that isotopically Indian-MORB-like asthenosphere also underlies this region and thus appears to extend far to the east of the Indian Ocean proper. However, despite its great size, the history and origins of this domain (i.e. why it is different from Pacific and North-Atlantic-type mantle) are understood only poorly. Figure
Two general classes of hypotheses have been proposed to account for the Indian Ocean asthenospheric domain. One is that it was created shortly before and during the breakup of Gondwana in the processes that formed the Indian Ocean itself. Possible causes involve either upwelling of deep, isotopically unusual plume-related mantle or widespread introduction of small amounts of continental lithospheric or old, shallowly subducted sedimentary material into the MORB source mantle (promoted by Gondwanan rifting, subduction-erosion, and/or the erosive action of the Kerguelen, Marion, Crozet, and Bouvet starting-plume heads) (e.g. Castillo, 1988; Klein et al., 1988; le Roex et al., 1989; Mahoney et al., 1989, , 1992; Storey et al., 1989, , 1992; Pyle et al., 1992, , 1995; Weis et al., 1992; Hickey-Vargas et al., 1995; Rehkämper & Hofmann, 1997). In either case, this material is postulated to have been dispersed to its present extent along asthenospheric flow paths as the Indian Ocean opened. In this class of hypotheses, older Indian MORB would be expected to show more variable isotopic signatures than modern ones because less time would have been available for intermixing of `normal' (i.e. Pacific-North Atlantic type) and contaminated asthenosphere. Depending upon their location and age, and the nature of asthenospheric dispersal patterns, some older lavas might not have Indian-MORB-type isotopic signatures at all; the same would probably be true of seafloor formed north of Greater India in the eastern Tethyan Ocean (Mahoney et al., 1992), the Mesozoic ocean that existed in much of the same region now occupied by the Indian Ocean before opening of the latter. An alternative hypothesis is that the Indian MORB mantle domain is a much longer-lived asthenospheric feature that existed well before the Indian Ocean started opening up (e.g. Hart, 1984; Crawford et al., 1995). In this case, the characteristic Indian-Ocean-type isotopic signature would be expected to be typical of both old Indian MORB and seafloor erupted in a widespread region of the Tethyan Ocean. Recently, Lanyon, (1995), Pyle et al., (1995), and
Weis & Frey, (1996) have studied old seafloor basalts from drill and dredge sites in the eastern Indian Ocean east of about 90°E. Their results indicate that lavas with ages ranging from 15 to 125 Ma exhibit rather typical Indian-MORB-type isotopic signatures; that is, values lying to the low-206Pb/204Pb side of the Pacific-North Atlantic MORB fields in Fig. 1. However, ~150 Ma samples from Deep Sea Drilling Project Site 261 (off the northwestern corner of Australia) lack clear Indian-MORB-type characteristics, and the same may be true of similar-age MORB from nearby Site 765 (see
Weis & Frey, 1996). Although preserved within the northeastern corner of the Indian Ocean, these basalts were formed around 30-35°S at a Tethyan spreading center some 15 my before significant spreading began in the eastern Indian Ocean between Greater Indo-Madagascar and Australia-Antarctica (e.g. Ogg et al., 1992). In this paper, we present results for 46-140 Ma lavas from drill sites in the western Indian Ocean. Also, although the eastern Tethyan Ocean no longer exists, fragments of Tethyan seafloor are preserved in MORB-type ophiolites along the Tethyan suture belt in southern Asia (Fig. 2). Here, we discuss results for two such suites: 110 Ma basalts from the Yarlung-Zangpo suture of Tibet, and a group of 150 Ma and 120 Ma rocks from the Masirah ophiolite off the Arabian peninsula. As a comparison, we also present isotopic data for Jurassic Pacific MORB from Alexander Island, Antarctica. Figure
Several of the samples studied were fairly fresh but most were affected by seawater-mediated alteration, ranging from mild `brownschist' to zeolite, prehnite-pumpellyite or, in some cases, lower greenschist facies (see
Davies et al., 1974; Fisher et al., 1974; Simpson et al., 1974; Whitmarsh et al., 1974; Moseley & Abbotts, 1979; Abbotts, 1981; Girardeau et al., 1985; Pearce & Deng, 1988; Doubleday et al., 1994). Rarely, we were able to pick enough fresh glass or clear-looking plagioclase for isotopic and isotope-dilution analysis; glass and plagioclase separates were cleaned ultrasonically in ultrapure, 1 M HCl and water before dissolution and further processing. For other samples, we followed a preparation procedure closely similar to that used in our previous isotopic studies of old, non-glassy submarine basalts. Chips from the least-altered interior portions of samples were broken to pieces of 3-5 mm in size, which were handpicked under a microscope to avoid visible alteration products (veins, vesicle fillings, and more altered patches of groundmass). The pieces selected were briefly cleaned ultrasonically in ultrapure, weak HF-HNO3 and H2O (in sequence) and then broken into smaller (~2 mm) pieces, after which the picking and cleaning procedure was repeated. The pieces chosen were ground in a boron carbide mortar, dissolved, and analyzed for isotopic ratios of Nd, Pb, and Sr and isotope-dilution abundances of Nd, Sm, Pb, Th, U, Sr, and Rb at the University of Hawaii. Several of the samples from Masirah (those lacking MSX or MA prefixes in the tables) were analyzed for Nd and Pb isotopes and Nd, Sm, Pb, and U abundances at the University of Bern following a generally similar preparation procedure; Th was analyzed for these samples at the University of Hawaii. In addition, isotopic ratios and parent-daughter element abundances were determined for some samples on splits of powder subjected to a multistep, HCl-dominated acid-leaching procedure effective at removing low-temperature alteration phases (carbonates, clays, chlorite, phosphate, zeolites, ferromanganese oxides; some fresh material is also removed in the process; see
Mahoney, 1987; Mahoney & Spencer, 1991). It should be noted that because the picking procedure, as well as acid leaching, when employed, variably modifies a sample's mineralogical composition relative to that of the bulk rock, the isotope-dilution data do not strictly represent bulk-rock elemental abundances and are used here for isotopic age-corrections only. The results are given in Tables 1 and 2. Table 1. Nd and Sr isotopic ratios and isotope-dilution abundances.
Table 2. Pb isotopic rations and isotope-dilution abundances.
Very few trace element data have been published for the western Indian Ocean drillhole lavas. Thus, we analyzed a subset of the bulk-rock samples for a broad suite of trace elements; we also analyzed several Masirah samples. Slabs of fresher portions of rock with a minimum of veins and amygdules were chosen (note that for some of the smaller drillhole samples, the freshest looking material had already been reserved for isotopic work). To avoid possible drilling- and/or handling-related contamination, the slabs (typically 5-30 cm3) were taken from sample interiors, lapped with SiC, cleaned briefly (~5 min) in ultrapure, weak HF-HNO3 (each ~0·2 M) and water in an ultrasonic bath and powdered in alumina; experience has shown that this procedure does not significantly modify bulk-basalt compositions for the elements analyzed. The resulting bulk-rock powders were prepared and analyzed by inductively coupled plasma-mass spectrometry at the University of Hawaii following techniques similar to those described by
Jain & Neal, (1996). The data appear in Table 3. Table 3. Bulk-rock incompatible element abundances (ppm).
Because seawater has a fairly high concentration of Sr (~8 ppm; e.g. Li, 1991) and high 87Sr/86Sr (today ~0·709) relative to oceanic mantle, 87Sr/86Sr values in basalts altered by seawater-derived solutions are typically elevated above values in pristine samples; in constrast, Nd and Sm abundances are extremely low in seawater (4 * 10-6 and 8 * 10-7 ppm, respectively) and Nd isotopes are resistant to modification by even rather high amounts of alteration (e.g. McCulloch et al., 1981; Staudigel et al., 1995). Like Nd, Pb abundances in seawater are very low (2 * 10-6 ppm), so that Pb isotope ratios are affected little by seawater interaction. Pb can be mobile in hydrothermal systems, but redeposition of Pb from one part of a volcanic system to another will normally not be isotopically distinguishable (unless the system possesses significant local-scale isotopic heterogeneity). However, unlike 147Sm/144Nd values, 238U/204Pb ratios (and to a lesser extent, 232Th/204Pb) can be affected markedly by mobility of U and uptake of U from seawater (3·2 * 10-3 ppm), in particular, as well as Pb (and sometimes Th) mobility (e.g. Tatsumoto, 1978; Macdougall et al., 1979; Chen & Pallister, 1981). If alteration of an oceanic basalt occurs within several million years after eruption, as typically appears to be the case (e.g. Staudigel et al., 1981), and the rock remains a nearly closed system thereafter, then age-correction of Pb isotope ratios will result in values close to the initial magmatic values even for an old specimen. Indeed, previous studies have demonstrated that good initial-Pb isotope information can be obtained on a range of crystalline magmatic rocks with hydrothermal overprints as high as greenschist facies (e.g. Chen & Pallister, 1981; Göpel et al., 1984). However, if alteration of parent-daughter ratios occurs long after eruption, or repeatedly over many millions of years, then age-adjusted Pb isotopic values will be erroneous. As a hypothetical illustration, let us consider a basalt erupted at 100 Ma whose 238U/204Pb ratio has been elevated from an original value of, say, 10 to 40 by some very recent seawater-alteration process; using the measured value of 40 to calculate an `initial' 206Pb/204Pb ratio leads to a substantial overcorrection of 0·47. The basalts of Alexander Island represent faulted slices of Jurassic seafloor formed by spreading between the Pacific and Phoenix plates, and preserved in an accretionary wedge complex (Doubleday et al., 1994). Ages are well determined at 150 Ma in one location (Sullivan Glacier; samples KG 3513-27 and KG 3513-4) and relatively poorly known at a second (Herschel Heights), where a 150 Ma age is assumed; a third area (Lully Foothills) is well dated at Early Jurassic (~200 Ma). Chemically, the basalts include both normal-type and incompatible-element-enriched MORB (N- and E-MORB); ocean-island-type compositions are also present (Doubleday et al., 1994). The samples we analyzed isotopically are altered to zeolite and prehnite-pumpellyite facies. In an initial [epsilon]Nd(t) vs (87Sr/86Sr)t diagram (Fig. 3), most of the Alexander Island data lie to the high-87Sr/86Sr side of the MORB field: although [epsilon]Nd(t) is between +8·6 and +5·4, within the range of modern N- and E-MORB, the (87Sr/86Sr)t of unleached splits varies from 0·70295 to 0·70439. In contrast to many of the drillhole and Masirah samples (see below), acid-leaching of the two Alexander Island samples whose unleached splits had the highest age-adjusted Sr isotope ratios produced only modest decreases in (87Sr/86Sr)t, not enough to move their data points into the MORB field in Fig. 3. This result probably reflects the fact that the main repository of Sr, plagioclase (which along with clinopyroxene typically makes up most of the residue for leached tholeiites; e.g. Mahoney, 1987), was largely replaced by secondary feldspar in these rocks during alteration (Doubleday et al., 1994) and that little material with pristine 87Sr/86Sr remains. Figure
Unlike Sr isotopes, the age-corrected Nd and Pb isotopic data all fall in a restricted field within the Pacific-North Atlantic MORB mantle array in Fig. 4c and d, particularly when the modern array is adjusted to the approximate position it would have occupied at 150 Ma (assuming the only changes have been caused by radioactive decay of parent nuclides in the MORB source mantle). The data plot in the low-[epsilon]Nd, high-206Pb/204Pb, high-208Pb/204Pb region of this array. Values of (207Pb/204Pb)t lie within the range for MORB and oceanic islands as well (Fig. 5a). The one sample for which we determined Nd and Pb isotopes on both acid-leached and unleached splits (KG 3513-27) has identical [epsilon]Nd(t), within errors, for each split, at +7·8 and +7·9; leaching also produced only small changes in the age-adjusted Pb isotope ratios of this sample [e.g. (206Pb/204Pb)t of 18·98 and 18·91; see linked light and dark squares in Fig. 4]. Because the age-corrected Pb isotopic values of all the samples (1) plot in a small field within the narrow Pacific-North Atlantic MORB array in both panels c and d of Fig. 4 and (2) have a lesser total spread in 208Pb/204Pb and 206Pb/204Pb relative to the measured, present-day range (0·33 vs 0·70 and 0·28 vs 0·47, respectively; the range in 207Pb/204Pb is the same within analytical error), and because (3) values for the leached and unleached splits of KG 3513-27 agree well with each other, it appears that alteration affecting U/Pb or Th/Pb ratios mainly occurred within a few million years after eruption. The same appears to be true of most samples from the other localities studied (see below). Figure
Figure
The drillhole samples display a wide range in elemental composition. Primitive-mantle-normalized incompatible element patterns of several samples are illustrated in Fig. 6, and can be seen to vary from typical N-MORB type (sloping generally downward to the left; e.g. Site 235-20-5) to E-MORB type (moderate upward slope to the left; e.g. Site 250A-26-5). The basalt from Site 690C (on Maud Rise) has an ocean-island-like pattern resembling those of Gough Island lavas (South Atlantic), including a spike at Ba and trough at Th and U; this rock is an unusual xenocryst-bearing alkalic basalt that
Schandl et al., (1990) concluded originated from a hydrous source containing small amounts of phlogopite and apatite. The Site 224 sample shows somewhat similar, but less extreme, ocean-island-like features. In contrast, the pattern for a Site 249 lava has a sizeable trough at Nb and Ta (as well as a larger than usual peak at Pb), a characteristic commonly seen in lavas influenced by continental lithosphere, and in arc-related basalts, but rarely in either fresh or altered oceanic basalts (compare Site 248 pattern). Patterns of the visibly fresher N- and E-MORB samples (e.g. Site 236-33-3, Site 250A-26-5) are relatively smooth, and even the more altered samples lack the pronounced spikes or troughs seen for some elements in patterns of highly altered basalts (Bienvenu et al., 1990; Staudigel et al., 1995; Jochum & Verma, 1996). Alteration effects are most evident in a marked elevation of Rb in the N-MORB lavas; Ba is also elevated significantly in many of these lavas (e.g. the Site 223 sample). Small peaks or troughs, which may reflect alteration, are present at Pb in several patterns, whereas U peaks and/or low Th/U ratios indicate significant U uptake in several samples (e.g. Site 245-19-1). Figure
We determined Sr and Pb isotopes on both unleached and leached (or glass) splits of seven of the drill-core basalts studied and Nd isotopes on six pairs. The (87Sr/86Sr)t values of all but one of the leached splits are significantly lower (by as much as 0·0006) than those of the unleached splits, whereas only negligible differences are observed in [epsilon]Nd(t) (0-0·4 epsilon units). The differences in age-corrected Pb isotope ratios for the members of each unleached-leached (or glass) pair are also relatively small, with one exception (see linked symbols in Fig. 4a and b). For example, the difference in (206Pb/204Pb)t ranges from 0·04 to 0·14, except for sample 245-19-1, which shows a difference of 0·26, by far the largest observed for any of the samples in our study. As with Nd isotopes, the age-corrected Pb isotope ratios of an unleached split can be either slightly higher or lower than for the corresponding leached residue or glass separate. Values of (207Pb/204Pb)t are all within the range for MORB and ocean islands (Fig. 5a), except for sample 248-17-2, for which both leached and unleached splits have high (207Pb/204Pb)t (15·65, 15·63) relative to (206Pb/204Pb)t (18·40, 18·49). Abundances of Nd, Sm, Pb, U, Th, and Rb typically dropped substantially with leaching. Although the 147Sm/144Nd ratios of leached splits tend to be higher than for their unleached counterparts, consistent with a relative enrichment of clinopyroxene in most of these residues (e.g. Mahoney, 1987), the 238U/204Pb and 232Th/204Pb values can be either higher or lower, probably depending on the particular combination of altered and unaltered phases remaining in the leached residue. It is important to note that, except in the glasses, neither set of values necessarily corresponds to those in the pristine rock. The total range in [epsilon]Nd(t) (unless otherwise specified, in the text hereafter we use isotopic values for leached residues or glass, when available) is considerable: +10·3 to -0·6. That for (87Sr/86Sr)t is also large, from 0·70277 to 0·70431. In contrast to the Alexander Island basalts, most of the age-adjusted drillhole-basalt data plot within or very close to the MORB fields in the Nd-Sr isotope diagram (Fig. 3). An interesting exception is the chemically and petrographically unusual alkalic lava from Site 690C (Schandl et al., 1990) on Maud Rise, which may have formed in association with a jump of the ancestral Southwest Indian Ridge toward the Bouvet hotspot in the 85-100 Ma period (e.g. Barker et al., 1990); this sample has anomalously low (87Sr/86Sr)t (0·70376) for its low [epsilon]Nd(t) (-0·6). Pb isotopes also exhibit a wide range, with (206Pb/204Pb)t, varying from 17·40 (the Site 690C sample) to 19·28 (the Site 236 glass). Many of the old western Indian Ocean basalts resemble modern Indian MORB and hotspot islands in that their age-corrected Pb and Nd isotopic ratios place them on the low-206Pb/204Pb side of the Pacific-North Atlantic MORB-source field in Fig. 4a and b. The oldest lavas, from Site 249, were erupted at ~140 Ma during early spreading between Africa and southern Greater Indo-Madagascar, and display a strong Indian-Ocean-type signature in both Fig. 4a and b. Samples from Sites 690C, 239, 248, 235, 224, and 221, with ages between ~85 and 46 Ma, also have clear Indian-Ocean-type isotopic signatures. In addition, dredged lavas from Afanasy-Nikitin Seamount (Fig. 2), erupted on or near the western Southeast Indian Ridge at ~80 Ma in a location far from continental landmasses, recently have been shown to possess lower [epsilon]Nd(t) (-8) and (206Pb/204Pb)t (16·77) values than any modern Indian Ocean (or any other oceanic) lavas (see Fig. 4). Thus, Indian-MORB-type isotopic compositions clearly were present in the old western Indian Ocean mantle, in good agreement with results for old lavas from the eastern Indian Ocean (Lanyon, 1995; Pyle et al., 1995; Weis & Frey, 1996). However, data for several sites overlap the Pacific-North Atlantic MORB-source array. As noted earlier, such characteristics appear to be very rare within the main part of the Indian Ocean domain today (Fig. 1). The Site 250A samples (including an N-MORB and an E-MORB), the two flows we analyzed from Site 240, and the upper of two petrographically distinct units (Fisher et al., 1974) at Site 236 have values that fall within the Pacific-North Atlantic MORB-source field in both Fig. 4a and b. The same is true for the Site 245 basalt, despite the significant difference in age-corrected Pb isotopic values between the leached and unleached splits of this sample. Moreover, the Site 250A-26-5 (90 Ma) E-MORB and the Site 236-33-3 glass-crystalline-rock pair (60 Ma) have age-corrected (206Pb/204Pb)t >19, significantly greater than seen for any modern Indian MORB (most of which have 206Pb/204Pb <18·4). [Although it was not age-corrected, an even higher present-day 206Pb/204Pb value of 19·580 was reported by
Hart, (1988) for a Site 250A lava deeper in the core than our E-MORB sample, with a value of 19·322.] Some modern Indian Ocean island lavas have values around 19, but they also have markedly lower [epsilon]Nd (~+4 or less vs +6·2 and +7·4), higher relative 208Pb/204Pb, and usually higher 87Sr/86Sr (e.g. ~0·704 for Réunion and Crozet). Two samples display notably ambiguous `mixed' isotopic characteristics, one from the lower unit at Site 236 and one from Site 223 (both ~60 Ma). In Fig. 4a, values for both the unleached and leached splits of these samples fall near the edge of but within the estimated Pacific-North Atlantic MORB-source array for 60 Ma. However, in Fig. 4b, data for both samples lie well above this field. Along the present-day spreading ridges, qualitatively similar mixed signatures have been found only in some lavas at the fringes of the Indian MORB domain (Mahoney et al., 1992; Pyle et al., 1992; Volker & et al., 1993). Masirah is an island off the coast of Oman that contains well-preserved exposures of uplifted abyssal oceanic crust. An older suite of MORB-type magmatic and ultramafic rocks is present, as well as a younger group of magmatic rocks, principally alkalic basalts and their differentiates but also amphibole-clinopyroxene gabbros and rare oceanic granites (e.g. Moseley & Abbotts, 1979; Abbotts, 1981; Moseley, 1990; Smewing et al., 1991; Gnos & Perrin, 1996; Nägler & Frei, 1997). Recent dating reveals that the MORB-type suite is ~150 Ma, whereas the younger suite is ~120 Ma (Smewing et al., 1991; Immenhauser, 1996; Nägler & Frei, 1997). The Masiran seafloor appears to have formed on the slow-spreading, transform-fault-dominated ridge system (e.g. Fisher et al., 1986) linking the main Tethyan and early western Indian Ocean (northeastern Somali Basin) spreading centers in the narrow basin between northwestern Greater Indo-Madagascar and the northeastern corner of Arabia-Africa (see Fig. 7) (e.g. Mountain & Prell, 1990; Smewing et al., 1991; G. Mountain, personal communication, 1993). (Note that the similar-age basement at Site 249 was formed in the Mozambique Basin farther southwest, to the southwest of Madagascar.) The cause of the later magmatism at ~120 Ma is uncertain but may be related to lithospheric fracturing and passage of the region near a hotspot (Meyer et al., 1996; Nägler & Frei, 1997). Figure
Primitive-mantle-normalized element patterns of representative tholeiitic and alkalic Masiran lavas (the latter represented by an E suffix in the tables) are shown in Fig. 8a and b. Low, MORB-like abundances of incompatible elements characterize the 150 Ma tholeiitic basalts, and their patterns range from typical N-MORB type to transitional-MORB type (relatively flat). The patterns of the 120 Ma alkalic lavas slope generally upward to the left and broadly resemble those of many oceanic island basalts [compare the average OIB pattern in Fig. 8; see also
Meyer et al., (1996)]. Overall, the patterns are relatively smooth. However, Rb and Ba can be mildly to dramatically enriched or depleted relative to Th in the visibly more altered samples analyzed (e.g. MA-401, MSX-219E). Small to moderate troughs or peaks at Pb are present in some patterns as well, and a substantial negative Eu anomaly can be seen in the pattern for MSX-219E, a trachytic lava. Figure
As with the drillhole samples, acid-leaching of the Masiran rocks reduced their (87Sr/86Sr)t values while causing negligible changes in [epsilon]Nd(t) and only modest ones in age-corrected Pb isotope ratios. In several cases, a large reduction in (87Sr/86Sr)t occurred: for example, from 0·70466 to 0·70305 for MSX-171. In the Nd-Sr isotope diagram (Fig. 3), the data for most of the Masiran rocks plot in or very close to the estimated 150 Ma MORB-source field (in the broad area of this diagram where Pacific-North Atlantic and Indian fields overlap), although leaching failed to bring the Sr isotope values of MA-401 or MSX-219E into this field. The plagioclase separate-which we did not leach-of a gabbro (MSX-71g) also yielded higher (87Sr/86Sr)t (0·70335) than the leached splits of basalts MSX-75 and MSX-171 [with Sr isotopic values of ~0·7030 at similar [epsilon]Nd(t)], evidently indicating that the plagioclase was somewhat affected by interaction with seawater. For the 150 Ma samples, [epsilon]Nd(t) ranges from +10·5 to + 6·4, indicative of intrinsic heterogeneity in the mantle source; however, eight of the 11 samples have values between +9·0 and +7·6. Values of (206Pb/204Pb)t vary from 18·17 to 18·88, with eight of the samples having ratios between 18·32 and 18·64. Moreover, most of the Pb isotope data define good positive correlations close to 150 Ma reference isochrons in plots of present-day 206Pb/204Pb vs 238U/204Pb and 208Pb/204Pb vs 232Th/204Pb (Fig. 9a,b), consistent with alteration largely occurring within a few million years after eruption for most of these samples. In both Fig. 4c and d, data for the 150 Ma rocks lie within the Pacific-North Atlantic MORB-source field, except for the leached split of sample MA-401, which falls slightly to the left of this field. Figure
The 120 Ma alkalic lavas have lower [epsilon]Nd(t) than the 150 Ma rocks, from +6·0 to +2·9. Six of the ten samples show little variation in age-corrected (206Pb/204Pb)t ratios, which are between 18·81 and 19·00. In Fig. 4c, data points for these six samples plot toward the low-[epsilon]Nd(t) end of the Pacific-North Atlantic MORB-source array and in or near the Réunion-Crozet source field. In Fig. 4d, the data for these alkalic lavas lie beneath this field, and straddle the Pacific-North Atlantic MORB-source array. The remaining four samples analyzed have significantly lower (206Pb/204Pb)t, between 18·49 and 17·71, yet have [epsilon]Nd(t) in exactly the same range as the other alkalic lavas. In Fig. 4d, the data point for one of these four samples falls well below the Pacific-North Atlantic MORB field, unlike any modern ridge or oceanic island basalts, and in Fig. 9c and d data for these samples lie far from the array defined by the others. We infer that relatively recent alteration (including variable loss of Pb) has seriously disturbed the Pb isotope systematics of these four alkalic lavas. In addition to our work, Nägler & Frei, (1997) have recently analyzed 120 Ma Masiran oceanic granites and amphibole-bearing gabbros, as well as several 150 Ma samples, for Nd and Pb isotopic ratios and U, Pb, Nd, and Sm abundances. Their age-corrected [epsilon]Nd(t) and (206Pb/204Pb)t values are similar to ours in that their data fall largely within the estimated 150 Ma Pacific-North Atlantic MORB-source field of Fig. 4c or, for one granite, close to the Réunion-Crozet source field; also, as with our alkalic lavas, their 120 Ma gabbros and granites tend to have lower [epsilon]Nd(t) values than those of the 150 Ma rocks. The samples from the Yarlung-Zangpo suture zone were collected from along the length of basalt outcrops to the southwest of Lhasa in the Xigaze area. Elemental and petrographic analyses of lavas from several of the same general areas show them to be chemically N-MORB (Pearce & Deng, 1988) altered to prehnite-pumpellyite or lower greenschist facies, with plagioclase replaced by albite and with abundant secondary groundmass actinolite and chlorite (Girardeau et al., 1985). The lavas were erupted at an eastern Tethyan spreading center north of Greater Indo-Madagascar and to the south of the Tibetan block (see Fig. 7; Pozzi et al., 1984). An age of 110 Ma was determined by
Marcoux et al., (1982) from Radiolaria in cherts interbedded conformably with pillow basalts. Previous isotopic work on magmatic rocks consisted of Pb isotope and Pb and U (but not Th) abundance measurements by
Göpel et al., (1984). Their results revealed a rough U-Pb whole-rock isochron (120 ± 10 Ma) which gave nearly the same age as the paleontologically derived age, indicating that the alteration affecting these rocks (including their U/Pb ratios) occurred fairly soon after eruption and that the rocks had remained nearly closed systems thereafter. As with the Alexander Island basalts, Sr isotope ratios of the Yarlung-Zangpo samples are elevated relative to [epsilon]Nd(t), and acid-leaching yielded only modest reductions in (87Sr/86Sr)t (to 0·70380-0·70406; Fig. 3). These results are consistent with the similar level of alteration in the two suites, specifically with the extensive replacement of original plagioclase by albite (in the Yarlung-Zangpo basalts, clinopyroxene is also partly replaced with secondary phases). Values of [epsilon]Nd(t) show very little variation in the Yarlung-Zangpo samples, all being between +8·0 and +8·5, indicating a nearly homogeneous mantle source. The age-corrected Pb isotope ratios also vary only slightly: the spread in (206Pb/204Pb)t, for example, is only 17·42-17·55 (in the same range as for Göpel et al.'s samples) and only 37·27-37·38 in (208Pb/204Pb)t. Our results confirm the good overall positive correlation of present-day 206Pb/204Pb with 238U/204Pb as well. Most important for our present purposes, the Yarlung-Zangpo basalts define a very small field with a clear Indian-MORB-type signature in both Fig. 4c and d. Consistent with the results for the Jurassic South Pacific lavas from Alexander Island, basalts of roughly similar age from several drill sites in the North Pacific recently have been shown to also possess Pacific-North-Atlantic-type age-adjusted isotopic signatures (Janney & Castillo, 1997). Allowing for relatively small changes in [epsilon]Nd and Pb isotope ratios resulting from radiogenic ingrowth in the source mantle during the last 150 my, the available data thus imply that the Pacific MORB mantle in the Jurassic and earliest Cretaceous was isotopically very similar to that of today. In the western Indian Ocean, the ~140 Ma Site 249 basalts reveal that Indian-Ocean-type isotopic compositions were present from almost the very beginning of the ocean itself-at least in some locations. However, the ~150 Ma rocks of Masirah and Site 261 (Weis & Frey, 1996) essentially lack normal Indian-MORB-type signatures. The crust at both sites formed near the southern boundary of the Tethyan Ocean, Masirah on the northwest and Site 261 on the northeast side of what would later become Greater India (see Fig. 7). Thus, the data for these sites provide no evidence that the Indian Ocean isotopic signature was inherited from Tethyan asthenosphere or, indeed, that the Indian Ocean mantle domain existed in anything like its present form north of East Gondwana in the Late Jurassic. On the other hand, the Yarlung-Zangpo basalts demonstrate the existence of Indian-MORB-type mantle in at least a part of the equatorial Tethys by 110 Ma. In addition, some high-quality, age-corrected Pb and Nd isotopic data have been published for the Samail ophiolite of Oman, the crust of which appears to have formed at a low northern latitude (Perrin et al., 1994) at ~95 Ma (e.g. Tilton et al., 1981), to the west of that preserved in the Yarlung-Zangpo suture (see Fig. 7). In Fig. 4d, the Samail Pb isotope data occupy a restricted field above the Pacific-North Atlantic MORB-source array, and in Fig. 4c the combined Nd and Pb isotopic results [which were obtained on different samples (Chen & Pallister, 1981; McCulloch et al., 1981)] define a rectangle that again largely falls outside the Pacific-North Atlantic MORB-source field (adjusted to a 95 Ma position; not shown in figure). Thus, the available data indicate an essentially Indian-Ocean-type mantle source for Samail crust at 95 Ma [see also
Benoit, (1997)]. Although very few locations have been studied as yet, the differences between the older and younger Tethyan sites suggest the possibility of a temporal change in asthenospheric composition. A period of continental lithospheric thinning was followed by spreading in the eastern Indian Ocean at ~135 Ma (e.g. Powell et al., 1988), and one possibility is that some Indian-MORB-type asthenosphere flowed northward out of the widening rift between Greater Indo-Madagascar and Australia-Antarctica, into parts of the Tethyan region. On the west side of Greater Indo-Madagascar, spreading and linkage with the Tethys began around 160-170 Ma (e.g. Lawver & Gahagan, 1993) (slightly before the 150 Ma Masiran rocks were formed), and northward flow of Indian-Ocean-type mantle may have begun earlier there. If so, to reach the paleolatitude of the Samail crust by 95 Ma, such asthenospheric flow must have occurred at rates of ~70-100 mm/yr; in comparison, rates of 25-40 mm/yr are indicated for westward flow of Pacific-type asthenosphere into the southeastern Indian Ocean between Antarctica and Australia since 43 Ma (Pyle et al., 1992, , 1995). Possible driving forces for significant asthenospheric outflow from the young, narrow Indian Ocean include (1) the ascent into the upper mantle of the large starting-plume heads of the Kerguelen, Marion, Bouvet, and Crozet hotspots [all of which may have reached the base of the lithosphere in the 200-120 Ma period; e.g. Storey, (1995) and references therein], and (2) the upward advection of the 660 km boundary between upper and lower mantle proposed to have occurred beneath East Gondwana early in the Cretaceous in response to accelerated circum-Pacific subduction of slabs into the lower mantle (Larson & Kincaid, 1996). However, dispersion of stringers of Indian-Ocean-type mantle into the Tethys is not a unique explanation of the existing isotopic data. Although evidence is lacking for a widespread, pre-110 Ma Tethyan upper mantle possessing Indian-MORB-type characteristics, an alternative possibility is that the Tethyan asthenosphere may have contained pockets of bothPacific-North-Atlantic-type and Indian-MORB-typecompositions (the latter presumably generated by the same types of processes as those acting in the Indian Ocean mantle) as far back as 150 Ma or even earlier. Study of Tethyan basalts from other locations along the southern Asian suture belt is required to evaluate these possibilities. The 120 Ma alkalic lavas of Masirah have ocean-island-like elemental signatures, and six have isotopic compositions rather similar to those of modern Réunion (21°S) and Crozet (46°S) hotspot basalts, particularly in Fig. 4c. Recent plate reconstructions in the hotspot reference frame (e.g. Curray & Munasinghe, 1991; Lawver & Gahagan, 1993; Müller et al., 1993) differ somewhat but suggest the Masiran region was situated between about 15°S and 25°S at 120 Ma, and thus appear to permit a Réunion connection of some sort (e.g. see Fig. 7) while ruling out an association with the Crozet hotspot, other than perhaps a `far-field' effect related to dispersion of Crozet plume-head material. Meyer et al., (1996) recently suggested that the Marion (Prince Edward) hotspot, located to the west of Crozet, could have been the source of the alkalic magmatism; however, like Crozet, the Marion hotspot is located at ~46°S. Also, the volcanoes of the Marion hotspot have different isotopic compositions (e.g. [epsilon]Nd = +5·7 to +7·4, 206Pb/204Pb = 18·5-18·6; Hart, 1988; Mahoney et al., 1992) from those of Réunion and Crozet. However, the origin of the 120 Ma Masiran rocks remains problematic, because although the isotopic signature of the main component in the Réunion plume appears to have changed little in the last 66 my (White et al., 1990; Peng & Mahoney, 1995), the Réunion hotspot is believed by most workers to have appeared only shortly before 66 Ma, the Deccan Traps event being interpreted as the hotspot's initial, plume-head phase [e.g. Basu et al., (1993) and references therein]. Both Indian-Ocean-type and some Pacific-North-Atlantic-type isotopic signatures are preserved in the old western Indian Ocean drill sites. Moreover, when results for the drillholes are combined with those for Afanasy-Nikitin Seamount, a 3 [epsilon]Nd unit wider total spread of [epsilon]Nd(t) is encompassed than is found for present-day Indian MORB and oceanic islands (compare Figs 1a and 4a). This comparison excludes the Early Cretaceous Kerguelen and Naturaliste plateaux, which reach even lower [epsilon]Nd(t) values than Afanasy-Nikitin but, unlike Afanasy-Nikitin, probably contain blocks of continental lithosphere [e.g. see
Mahoney et al., (1996) and references therein]; it also excludes recent lavas (with less extreme values; e.g. Weis et al., 1992) erupted through the thick lithosphere of the Kerguelen Plateau. As such, the total range in [epsilon]Nd(t) for modern (0-10 Ma) Indian Ocean basalts is -4·0 to +11·3, whereas that for the old lavas is -8·0 to +10·3, with both maximum and minimum values seen in lavas formed at 60-80 Ma (note that Site 261 in the northeastern Indian Ocean, although formed at a Tethyan ridge, yielded an even higher value of +14·4; Weis & Frey, 1996). This difference is much greater than achievable by plausible isotopic evolution (i.e. `aging') in the source mantle since the Cretaceous. The larger isotopic range observed for the old basalts is remarkable in view of the very sparse sampling of old Indian Ocean crust relative to sampling of the present-day spreading centers and islands; further, it is unlikely that the few existing old sites have fortuitously sampled the full isotopic range present in old Indian Ocean seafloor. Admittedly, substantial sections of the modern spreading system remain unsampled or have only recently been dredged, and could potentially harbor more extreme isotopic compositions than found elsewhere. However, continuing work on the largest previously unsampled stretch of the system, the Southeast Indian Ridge between the Australian-Antarctic Discordance and St Paul Island, thus far reveals isotopic values that are all well within the previous range for modern Indian MORB (Hall et al., 1995; unpublished data, 1997). Nor does it appear that the difference in isotopic ranges can be ascribed in any simple way to differences in spreading rate. Greater isotopic heterogeneity along ridges generally is associated with slower spreading rates-for example, both the highest and lowest [epsilon]Nd values observed in the modern Indian Ocean (+11·3 and -4·0) are found on the central part of the very slowly spreading Southwest Indian Ridge (Mahoney et al., 1992). However, the old Indian Ocean samples with both the highest and lowest [epsilon]Nd values were formed (also on-ridge or very near-ridge) in the 60-80 Ma period of super-fast spreading (e.g. Fisher & Sclater, 1983). The most straightforward interpretation of the existing data is therefore that the Indian Ocean asthenosphere was isotopically more heterogeneous in the past and is gradually becoming better mixed on a time scale of tens of millions of years. In turn, greater heterogeneity in the not too distant past, including the presence in some locations of Pacific-North-Atlantic-type compositions, is most consistent with a relatively young origin for the Indian Ocean mantle domain; that is, one not considerably older than the age of the Indian Ocean itself. Reviews by R. Hickey-Vargas, P. Janney, and D. Pyle are gratefully acknowledged, as are helpful informal comments by P. Castillo and F. Frey. We thank F. Moseley, T.-Y. Guo, and the Ocean Drilling Program for providing samples, and N. Hulbirt, Z. Peng, K. Spencer, and D. VonderHaar for help with various aspects of the work. This work was funded by NSF grant EAR94-18168.INTRODUCTION
METHODS
A note on effects of alteration on isotopic ratios
RESULTS
Old Pacific MORB
Western Indian Ocean drill sites
Masirah
Yarlung-Zangpo suture zone
DISCUSSION AND CONCLUSIONS
ACKNOWLEDGEMENTS
REFERENCES
