4.1. Neodymium as a Paleoceanographic Proxy
 Palmer  and Palmer and Elderfield  showed that formation of MnO2coatings around foraminifera occurs at the ocean surface. However, they also showed that these oxides, transferred at the bottom in less than a year, were affected by several dissolution re-precipitation processes at the sediment/water interface. They showed that, even if these processes change the Nd concentration in the oxides, their isotopic ratio is almost totally preserved. Indeed, we think that the finalεNd seawater signal is incorporated in oxides mainly at the sediment/water interface, but mainly results from the mixing of Nd isotopes that come from the surface via erosion processes [Burton and Vance, 2000; Gourlan et al., 2010; Horikawa et al., 2011; Le Houedec et al., 2012; von Blanckenburg and Nägler, 2001] and also from bottom currents [Piotrowski et al., 2004; Robinson et al., 2010; Scher and Martin, 2004; Thomas, 2005]. Thus, the final εNd signal depends on the relative influences of both of these processes which vary from place to place. These variations are illustrated by von Blanckenburg  for present seawater or by Le Houedec et al.  for the last 30 Ma. Le Houedec et al.  show that at the same location the εNd signal of seawater at intermediate depth can be influenced by oceanic transport as well as weathering inputs.
 All samples studied here come from cores drilled between 2000 to 1300 m and the measured εNd signal is a proxy of intermediate water masses. We therefore consider that the εNd seawater signal extracted from our samples reflects a mixture of Nd isotopes that come from products of surface erosion and from oceanic intermediate currents flowing at ∼2000 m depth.
4.2. Regional Nd Sources
 During and after the Cretaceous, the Indian Block broke up from Gondwana, separated from the African and Antarctic continents and migrated northward. The drift rate of India increased tremendously between 60 and 50 Ma and slowed down upon full frontal collision with Asia at ∼50 Ma [Dyment, 1998; Fisher et al., 1971; McKenzie and Sclater, 1971; Müller et al., 2008a; Patriat and Achache, 1984; Rowley, 1996; Schlich, 1982; Sclater and Fisher, 1974]. Indian motion strongly affected the location of Nd sources and the oceanic circulation pattern in the Tethys and in the Indian Ocean. These will be now discussed in detail dealing first with the Indian block.
 Between 90 and 65 Ma, erosion of India transferred unradiogenic neodymium (typical values of εNd = −20) from the Precambrian platform to the ocean [Beckinsale et al., 1980; Naqvi and Rodgers, 1983]. Sixty five million years ago, the activity of the Reunion hot spot brought a huge amount of basaltic lava to the Earth's surface, forming the Deccan traps: most of the volume was erupted in (possibly much) less than 1 Ma [Chenet et al., 2007; Courtillot et al., 1988; Duncan and Pyle, 1988]. During this short period, the Deccan volcanism probably contributed to the isotopic composition of seawater in the region inducing a Nd radiogenic spike, but the resolution of our records across the K/T boundary does not allow the detect this signal. This effect was very well documented for Os isotopes [Peucker-Ehrenbrink and Ravizza, 2000; Ravizza and Peucker-Ehrenbrink, 2003].
 After 65 Ma, evaluating the contribution of Nd from India is not straightforward. Even though lavas have an εNd isotopic signature between −5 and +5 [Allègre et al., 1982], the present river waters draining this plateau on both the eastern and western sides carry unradiogenic Nd due to the large extent of Precambrian rocks surrounding the basalt [Colin et al., 1999; Kessarkar et al., 2003]. We could thus consider that the Deccan Traps have had little influence on the εNd value of the Indian Ocean. Note that the Precambrian terrains to the north of the Greater India have now disappeared in the subduction zone. Therefore, the proportion of Precambrian rocks versus volcanic rocks before the collision was higher that now. However, Kent and Muttoni  suggested that erosional products and intensity changed with the transit of the Indian continent over different climatic belts. The equatorial humid belt was successively crossed over by the northern Precambrian plateau of the Greater India from 65 to ∼50/∼55 Ma and by highly weatherable Deccan Traps [Dessert et al., 2003; Louvat and Allègre, 1997] from ∼50/∼55 and 38 Ma. These authors suggest that after ∼50/∼55 Ma, the Deccan rocks must have suffered intense weathering. This effect could have added radiogenic Nd in seawater after ∼50/∼55 Ma.
 After 50 Ma, the Himalaya uplift did not occur immediately after the Indian-Asian collision, but only some 30–25 Ma ago after extrusion of Indochina to the Southeast [Tapponnier et al., 2001] and thus did not contribute to the εNd seawater signal during the 90–40 Ma interval.
 Between 90 Ma and 50 Ma, the erosion of the Australian and Antarctica Precambrian platforms brought very unradiogenic neodynium to the ocean [DePaolo et al., 1982; McCulloch, 1987]. Note that the separation of Gondwana in multiple pieces most likely enhanced the transfer of erosion products to the ocean, as has been shown to be the case for Rodinia [Donnadieu et al., 2004].
 The Southern Indian Ocean was also characterized by active submarine volcanism over this period of interest. It was initiated by the emplacement of the Kerguelen LIP with eruption ages ranging from 119 Ma in the southern part of the Kerguelen plateau to 34 Ma in its northern part [Coffin et al., 2002; Frey et al., 2002]. The peak of the Kerguelen LIP magmatic output occurred prior to 90 Ma (with the Rajmahal traps at ∼115 Ma; 0.9 km3/yr). Between ∼82 and 38 Ma, the Kerguelen hot spot generated the Ninetyeast Ridge (and the small Skiff Bank) at a much smaller rate of 0.1 km3/yr [Coffin et al., 2002; Frey et al., 2002]. The volume of volcanic rocks per 10 Myr period was evaluated by angle seismic data, gravity modeling and 40Ar/39Ar dating [Frey et al., 2002]. The Ninety East Ridge does not show large morphologic changes from South to North and thus Frey et al.  suggested that the activity of the Kerguelen hot spot was regular between 82 Ma and 38 Ma. Therefore, we will consider that the Kerguelen hot spot contribution to the regional Nd budget has remained constant over the time period studied here. This is likely true at first order only and the Kerguelen hot spot activity has probably fluctuated through time around the proposed value.
 All these tectonic changes had consequences on paleogeography and paleoceanography. The connection between the Indian and Pacific oceans remained open between 90 and 50 Ma in the equatorial area [Bush, 1997; Cogné and Humler, 2004; Poulsen et al., 1998]. An intermediate current flowed westward through this oceanic pathway, joining the Pacific and Atlantic oceans via the Equatorial Indian Basin [Bush, 1997; Pucéat et al., 2005; Stille et al., 1996]. However, the intermediate oceanic circulation between these Pacific and Indian basins occurred only near the equator, far away from our target region in the South which was, in addition, isolated from this equatorial current given the position of the Indian continent at the time (Figure 1). The situation changed after 50 Ma with the closure of the oceanic passage induced by the India-Asia collision. At that time,εNd equatorial intermediate seawater values from the Pacific Ocean ranged from −5 to −4 [Ling et al., 1997, 2005; Meynadier et al., 2008].
 The Nd composition of seawater in the southern Indian Ocean fluctuates between non-radiogenic (−10 to −11) and slightly more radiogenic (−9 to −8) composition. These variations are consistent at the scale of an entire oceanic basin and are interpreted as the result of regional weathering inputs as well as water mass mixing. At first order, we note that the ocean in this region was certainly well mixed (at least the first 2000 mbsl) over a large geographic distance since it generated a similar pattern in all four studied sites for each time period. The observed pattern leads us to divide the Indian Ocean Nd isotopic history into three episodes: (1) 90 to 65 Ma, (2) 65 to 50 Ma and (3) 50 to 40 Ma (Figure 5).
Figure 5. Schematic reconstructions of paleo oceanic currents and paleo Nd sources. (a) For the Late Cretaceous, the main intermediate oceanic current was the Equatorial Tethys Circumglobal Current (Tethys-CC). (c) For the Middle Eocene, after the India-Asia collision, we suggest a main oceanic influence of Pacific Equatorial water in South Indian Ocean. The Equatorial Circum Current was diverted by the Eastern coast of India to the Southern Indian Ocean as a result of India-Asia collision and the closure of the equatorial pathway. Blue arrows indicate surface to intermediate oceanic currents and dashed black arrows the main Precambrian continental Nd sources. Volcanic Nd sources are in red. (b) The interval of strong activity of the South East Indian Ridge (SEIR) from 60 to 50 Ma is underlined by a red broken line. The red zones (“D” and “K”) correspond to the Deccan Traps province and Kerguelen plateau (see text).
Download figure to PowerPoint
 At 90–80 Ma, the South Indian Ocean was surrounded by old continental blocks derived from Gondwana breakup (Antarctica-Australia, Africa, South America, India) and the seawaterεNd was on the order of −9. As discussed by Robinson et al. , this value is relatively high if we consider the very low values of surrounding old continental crust. Presently, the εNd value of the North Atlantic Ocean, which has a similar geological environment, is on the order of εNd = −13 [Goldstein and Hemming, 2003]. The Kerguelen Plateau and Hot spot in the center of the Southern Indian Ocean are the source of radiogenic Nd that probably explains this isotopic composition.
 From 90 to 65 Ma, the end of the supercontinent Gondwana splitting into several continental masses produced an increase of surface exchange between ocean and continent (Figures 1a and 5a), which enhanced erosion inputs to the ocean, mainly through transport by rivers [Donnadieu et al., 2004]. This increase of unradiogenic Nd inputs could explain the 2 to 3 units decrease of εNd values in ancient Indian seawater recorded at ODP sites 1135, 758 and 762 (Figure 2a). The lowest εNd values were reached at about 65 Ma in ODP sites 1135 (−10.5), 758 (−11.5) and 762 (−11.2).
 The trend toward more radiogenic values reversed sharply after 65 Ma, implying a strong radiogenic Nd input in the region which persisted until ∼50 Ma. From −60 to −50 Ma, εNd seawater value remained relatively constant at −9 ± 0.5. There is no evidence of abrupt tectonic, or oceanic circulation changes in the South Indian Basin around 65 Ma which could explain this sharp εNd increase. However, two major volcanic events that brought to the surface materials of mantle origin with high Nd radiogenic values occurred then in Indian Ocean (Figure 5b): (1) the gigantic eruption of Deccan Traps [Chenet et al., 2007; Courtillot et al., 1988], (2) the sudden increase of the South East Indian Ridge (SEIR) spreading rate [Cande et al., 2010; Cogné and Humler, 2004, 2006; Müller et al., 2008a; Patriat and Achache, 1984] as shown in Figure 2b. It was shown in the previous section that, with the probable exception of the first million year after the emplacement of the Traps, the Deccan traps might have had a small influence on the Indian Ocean seawater εNd value from ∼66 to 50/55 Ma. Interestingly, from 65 to 50 Ma, the εNd pattern follows the SEIR spreading rate pattern (Figure 2) suggesting that Nd from the ridge could have significantly influenced the seawater εNd value in this region. After 50 Ma, Indian Ridge activity dramatically dropped down from ∼16–18 cm/yr to a value lower than 7 cm/yr (Figure 2b). This was synchronously accompanied by a spectacular decrease of Indian Ocean εNd back to −11.
 The ridge crest contribution to the oceanic Nd budget has been the subject of several discussions. In an early paper, the ridge contribution was considered to be significant, as is the case for Sr and Pb [Elderfield and Greaves, 1982]. Later, it was shown that this flux was too small to influence the Nd isotopic composition of the global ocean [Bertram et al., 1992; Halliday et al., 1992; Michard et al., 1983; Piepgras and Wasserburg, 1985]. However, these discussions focused on the ridge influence on the average εNd of an entire ocean and not on the εNd signal at a regional scale. Today, the εNdsignal in the vicinity of the ultra-fast spreading East Pacific Rise (EPR) suggests a source of radiogenic Nd in this region. HighεNdvalues are found in Fe-Mn crusts in the EPR region [O'Nions et al., 1978; Piepgras and Wasserburg, 1985; Piepgras et al., 1979], which contrasts with other water data in the Pacific Ocean [Aplin et al., 1986; Frank et al., 1999; O'Nions et al., 1978; Piepgras and Wasserburg, 1985; Piepgras et al., 1979; van de Flierdt et al., 2004a]. The contribution of the ridge crest to the εNd signal of seawater can be a direct result of submarine volcanism on the ridge, hydrothermal activity but also seawater alteration of ridge crests and MnO2precipitation. Moreover, it has recently been shown that there is a local circulation of seawater toward the surface of the ocean close to ultra-fast spreading ridges. This circulation disperses chemical elements produced above the ridge and inhibits the fast scavenging near the ridge [Adams et al., 2011; Lupton and Craig, 1975; Von Damm et al., 1985]. ODP sites were located at 1500 mbsl, only ∼1000 m above high spreading rate ridge crests [Westphal et al., 2002]. The sites are therefore in the depth range affected by local low-density circulation. We suggest that the Nd isotopic composition of seawater in the region of ultra-high spreading ridge might be influenced by the ridge.
 Because the shift toward radiogenic values (from ∼−11 to ∼−8.5) is relatively small (2.5 εNd units), this contribution is clearly limited. With a first order mass balance calculation, we can estimate this contribution. Only 13% of radiogenic Nd (respectively 11%) is needed to shift the seawater value from −11 to −8.5, if we consider a value of εNdof +8 (respectively +12) for the ridge contribution. We can use this calculation to estimate the isotopic shift that would induced a similar contribution in the present-day Pacific Ocean seawater near the EPR. Seawater values would change from −4 (resp. −3) to ∼−2.3 (resp. ∼−1.4). This difference is compatible with the small seawaterεNd values in the region [Piepgras and Wasserburg, 1985; Piepgras et al., 1979], but it is difficult to go deeper in the comparison if we consider the analytical precision on seawater Nd isotopic data (±0.5 εNd units).
 Assuming that an ultra-fast spreading rate was a source of radiogenic Nd would explain whyεNdsuddenly rose after 65 Ma and strongly decreased to less radiogenic values after the India-Asia collision at ∼50 Ma, when the spreading rate dropped sharply. One could also suggest that the collision triggered theεNd decrease at ∼50 Ma through large inputs of eroded material from the Indian continent (i.e., rocks from the Himalayan belt). However, Métivier et al. showed that this erosion was associated with the uplift that occurred 20 to 25 Ma after the India-Asia collision [Tapponnier et al., 2001].
 The last 10 Myr of our record, from 50 to 40 Ma, are characterized by a large and rapid increase of εNd ancient seawater values, by ∼3 εNd units for ODP site 1135, and ∼2 εNd units for ODP sites 757 and 762 (Figure 2a). This could result from an increase of radiogenic input and/or be due to a decreasing contribution from unradiogenic sources. The radiogenic contribution could result from the erosion of the Deccan traps as they transited through the equatorial humid belt as suggested by Kent and Muttoni  (Figure 5c). The increase of εNd is precisely dated at 50 (±1) Ma and thus can be explain by the Deccan Traps weathering only if it started at 50 Ma. However, Kent and Muttoni  also suggested that the enhancement of the Deccan Traps weathering induced the decrease recorded in the seawater osmium isotopes at 55 Ma [Kent and Muttoni, 2008; Nielsen et al., 2009; Turekian and Pegram, 1997]. Changes in the Os and South Indian seawater εNd toward more volcanic values were not concomitant (5 Ma difference). Therefore, we suggest that only one of these two seawater records can reflect the weathering of Deccan Traps. considering the residence time of these two elements (∼30 Kyr for Os i.e., [Levasseur et al., 1999, 2000; Luck and Turekian, 1983; Oxburgh, 1998; Peucker-Ehrenbrink and Ravizza, 1996; Ravizza and Turekian, 1992] and ∼1 Kyr for Nd [Bertram and Elderfield, 1993; Tachikawa et al., 1999]) and the proximity of the Deccan Traps with our target region, it is difficult to envisage that the Deccan weathering affected the Os signal without disturbing the regional Nd record. Therefore, we suggest that the enhancement of the Deccan Traps weathering, if it ever occurred, started at 50 Ma.
 The second scenario, that does not exclude an enhancement of the Deccan Traps weathering at 50 Ma, involves the closure of the connection between the Pacific and North Atlantic basins and consequently the end of the Tethys Circumglobal Current (TCC). Due to the geostrophic oceanic circulation, the Indian-Pacific Equatorial current was diverted along the Eastern coast of India toward the South Indian Ocean (Figure 5c). The strong mixing of radiogenic Pacific seawater with less radiogenic Indian seawater induced an increase of εNd seawater values in our target region. This is also supported by the similarity of εNdvalues over a large geographic distance which suggests a well-mixed ocean.
 The establishment of this Southwest current is supported by the development of warm foraminifera faunas, interpreted as an increase of Pacific warm water input in the southern Indian Ocean [Berggren and Holister, 1977; Hocutt, 1987]. This current would have some similarity with the present South East current that prevails along the coast of Africa (M. Tomczak and S. J. Godfrey, Regional oceanography: An introduction, version 1.0, 2001, e-book available athttp://www.es.flinders.edu.au/∼mattom/regoc/pdfversion.html). Such a rapid rise of εNd was found in the Tasman Sea region on the East flank of the Australian Continent, after the narrowing of the Indonesian Gateway at ∼10 Ma [van de Flierdt et al., 2004b]. The Equatorial Pacific seawater with high εNd was there also diverted to the South and mixed with less radiogenic southern seawater.