The World Ocean Circulation Experiment Indian Ocean helium isotope data are mapped and features of intermediate and deep circulation are inferred and discussed. The 3He added to the deep Indian Ocean originates from (1) a strong source on the mid-ocean ridge at about 19°S/65°E, (2) a source located in the Gulf of Aden in the northwestern Indian Ocean, (3) sources located in the convergent margins in the northeastern Indian Ocean, and (4) water imported from the Indonesian Seas. The main circulation features inferred from the 3He distribution include (1) deep (2000–3000 m) eastward flow in the central Indian Ocean, which overflows into the West Australian Basin through saddles in the Ninetyeast Ridge, (2) a deep (2000–3000 m) southwestward flow in the western Indian Ocean, and (3) influx of Banda Sea Intermediate Waters associated with the deep core (1000–1500 m) of the through flow from the Pacific Ocean. The large-scale 3He distribution is consonant with the known pathways of deep and bottom water circulation in the Indian Ocean.
 Primordial helium trapped in the Earth's mantle is injected into the mid-depth ocean through submarine hydrothermal and volcanic activities. Typically, the injected helium is fivefold to tenfold enriched in the rare isotope 3He compared to atmospheric helium. The injected 3He is easily distinguished from the background 3He signature of atmospheric origin imprinted by air-sea gas exchange. Thus the mantle 3He signal can be traced over large distances in the ocean typically between 1000 and 3000 m depth. At these depths other tracer fields have no signal yet or frequently have small lateral gradients complicating their use for studies of flow patterns. 3He distributions have been used successfully to infer circulation patterns in the oceans such as the westward plume in the South Pacific at 15°S [Lupton and Craig, 1981], the southwestward plume in the northeastern Pacific Ocean and the eastward plume from Hawaii to the Mexican Coast [Lupton, 1998]. These plumes clearly define the regional circulation, which is not apparent in conventional tracer fields.
 The first helium measurements in the Indian Ocean were obtained from the Geochemical Ocean Sections Study (GEOSECS) during 1977–1978. These were made by W. B. Clarke and Z. Top at McMaster University, J. E. Lupton and H. Craig at the Scripps Institute of Oceanography, and W. J. Jenkins at the Woods Hole Oceanographic Institution. The results revealed moderate 3He enrichment with δ3He values of 12–15% [Östlund et al., 1987; Lupton and Craig, 1980] (δ3He [%] = 100 × (Rsample/Rair − 1); R = 3He/4He). The large-scale distribution of 3He is determined by the helium sources, the deep ocean circulation and ocean mixing. The GEOSECS 3He measurements revealed a north-south gradient with relatively high anomalies in the northern and central Indian Ocean (δ3He about 13–16%) and low anomaly in the southern Indian Ocean (δ3He about 8–9%). The measurements hinted at a prominent mid-depth eastward flow in the southern Indian Ocean and influx of Pacific waters enriched in 3He through the Indonesian Throughflow. The GEOSECS survey also illustrated the differences between the 3He distributions in the three major ocean basins due to differences in the processes controlling its distribution. These processes include the 3He injection rate and the overturning rate of the ocean basins. The deep Pacific Ocean is significantly enriched in 3He (δ3He about 25–30%) because of large-scale hydrothermal activity over the East Pacific Rise. In the Indian Ocean, δ3He excesses are on the order of 13–15% due to moderate input and possibly faster overturning. In contrast, the deep Atlantic Ocean is least enriched in 3He (δ3He about 5–8%) due to weak injection rates [Jenkins and Clarke, 1976; Top et al., 1987; Rüth et al., 2000] and relatively short turnover time [Stuiver et al., 1984; Smethie, 1993].
 A decade after the GEOSECS survey the French Indigo Cruises [Poisson et al., 1989] augmented the helium isotope data in the Indian Ocean. An analysis of the combined Indigo and GEOSECS helium data set by Jamous et al.  identified the spreading centers in the Gulf of Aden, at the vicinity of the triple junction of the Central, Southeast and Southwest Indian Ridges and the Indonesian Throughflow as the main sources of mantle helium in the Indian Ocean. The resulting 3He distribution revealed some features related to the characteristics of deep circulation in the Indian Ocean but could not resolve the interbasin exchanges due to the limited spatial coverage. More recent measurements in the eastern Indian Ocean and Indonesian Seas by Jean-Baptiste et al.  and Top et al.  indicate an influx of 3He into the Indian Ocean in the 1100–1400 m depth range from the Indonesian Seas.
 Thus, although several features in the previously known 3He distribution revealed elements of intermediate and deep circulation, there still was the need for a more complete data set that would allow derivation of the circulation patterns. To address this need the tracer component of the World Ocean Circulation Experiment (WOCE) Hydrographic Program (WHP) included 3He measurements. The present analysis is based on the WOCE helium isotope data. Specifically the mantle-derived 3He is mapped mainly in the 1000–3000 m depth range. Intermediate and deep circulation patterns inferred from this distribution are discussed and compared with those obtained by other means such as hydrographic and nutrients analysis [e.g., Warren, 1981; Toole and Warren, 1993; Mantyla and Reid, 1995; Reid, 2003].
2. Data and Methodology
 The present helium isotope data are obtained from WHP lines I7N, I8N, I9N, I10, I1, I2, I3, and I4/I5W/I7S. The station locations are shown in Figure 1. Standard procedures for sampling, and sample processing were followed [Lott and Jenkins, 1998; Ludin et al., 1998; Clarke et al., 1976]. Water samples were collected in standard rosette systems equipped with 36 Niskin bottles each of volume 10 L. The samples were extracted at sea from 50 to 80 mL of water. The extracted gases were flame sealed in low-permeability glass ampoules for measurement at University of Miami (UM), Woods Hole Oceanographic Institution (WHOI), Lamont Doherty Earth Observatory (LDEO) and Pacific Marine and Environmental Laboratories (PMEL). The measurement procedures at UM, WHOI and LDEO are documented by Clarke et al. , Lott and Jenkins  and Ludin et al. , respectively. The procedure followed at UM is described below and the procedures followed in the other labs are similar.
 The gas samples collected at sea are admitted into a processing line where gases other than helium and neon are isolated by cold trapping. A fraction of the helium and neon is routed to a quadrupole analyzer for preliminary helium and neon measurement. The remaining helium and neon is then admitted through a cryogenic trap to separate neon. Helium is finally admitted into a MAP 215-50 isotope mass spectrometer, which measures the helium isotopes. Air samples and blanks are interspaced with the sample runs to calibrate and monitor the sensitivity changes of the instrument during operation. The precision in measurements of δ3He data is ±0.15–0.2%. The precision in measurements of 4He and total neon is about ±0.5% and ±2%, respectively.
 Since the helium samples were analyzed at four different labs (UM, WHOI, LDEO and PMEL) a preliminary test was performed to check for consistency between the data. This was done by comparing the helium measurements, in the depth range of 2000–4000 m at stations located close to the crossover point of any two WHP lines. The data from each station were first screened for any obvious outliers. Then a smooth spline was fitted through the data points (at least four measurements from each station were used) as a function of potential density referenced to 3000 db for each station in the crossover stations pairs identified in Table 1. For each station pair the δ3He difference between the curves was calculated at five points that covered the density range common to the stations. The mean and standard deviation of the differences are reported in Table 1. The I1 data when compared to I7N, I9N and I8N were found to be consistently higher in δ3He. To correct this offset we analyzed the difference between I1 samples measured by UM and WHOI at the same depth. It was found that the I1 data as measured by UM were systematically higher by about 1.7%. Therefore the I1 data used here were adjusted by subtracting 1.7%. The conclusion of this analysis is that the Indian Ocean δ3He measurements by four different laboratories are consistent. The measurements mostly agree within ±0.2–0.4%.
I1, I7 and I9N were analyzed at University of Miami, I8N was analyzed at Pacific Marine and Environmental Laboratories, and I3 and I2 were analyzed at Lamont-Doherty Earth Observatory.
WHP is World Ocean Circulation Experiment (WOCE) Hydrographic Program.
The I1 Rosenstiel School of Marine and Atmospheric Sciences (RSMAS) deep δ3He were found to be systematically higher by about 1.7%. RSMAS measurements were compared with Woods Hole Oceanographic Institution measurements between depths of 1000 and 1500 m obtained at the same station. The agreement shown above is after adjustment of the RSMAS deep helium data.
 The main aim of this study is to relate the observed 3He distribution to large-scale circulation patterns in the deep waters of the Indian Ocean. Profiles, sections and maps of δ3He are constructed to elucidate the distribution of 3He. The results of the individual WHP lines are presented as sections and the lateral distribution is contoured on σ1.5 = 34.35, σ2 = 36.92, σ2 = 37.00 and σ3 = 41.495 isopycnal surfaces. These surfaces capture the features of the 3He distribution at depths of roughly 1300, 2000, 2500 and 3000 m and are mostly identical with the ones used by Mantyla and Reid  and Reid . The contouring on these surfaces was done by objectively mapping the data with a Gaussian covariance and a horizontal scale of 300 km. The data on the isopycnals deeper than 2000 m were first mapped within individual deep basins and then combined. The results of the objective analysis were checked by comparing with hand-contoured plots.
 The δ3He maps are used to infer flow based on the spreading of the 3He signal on the density surfaces. It is common in the literature to use the term “spreading” or “extending” to describe the combined effects of advection and diffusion in shaping the, steady state, property fields. However, advection is thought to be important in determining the shape of property fields [e.g., Mantyla and Reid, 1995; Johnson et al., 1998; Reid, 2003]. We adopt this assumption in the discussion of 3He fields. Maps of salinity, oxygen and silica on the σ2 = 37.00 surface from Reid's  study are also included for reference.
 Before a discussion of the results a few issues warrant attention. First, the analysis presented in this study does not consider a potential source of 3He to the oceans, namely, tritium decay. This decay could create a δ3He of 5% in the thermocline depending on the location in the northern hemisphere where most of the bomb tritium was added. The low natural tritium background precludes significant tritiogenic 3He in most of the deep ocean [Jenkins and Clarke, 1976]. Here the neglect of tritiogenic 3He is reasonable because most of the tritium was added to the northern hemisphere, whereas the Indian Ocean is largely a southern hemisphere ocean, and the present discussion deals with deep waters where tritium levels have been negligible [Östlund et al., 1987] even while oxygen and chlorofluorocarbons were high [Toole and Warren, 1993; Fine, 1993].
 Second, in describing the property fields we assume a steady state distribution. We feel that this assumption can be justified by comparing the turnover time for the deep ocean basins (∼103 years [e.g., Stuiver et al., 1984]) with the timescale of processes controlling the mantle degassing at tectonically active sites (∼106 years [e.g., Condie 1997]). Whereas this is a reasonable assumption at global scales, it may not be necessarily true at regional scales. Indeed, Jean-Baptiste et al.  observed anomalous 3He enrichment in the eastern Indian Ocean and attributed this to transient mantle input probably associated with volcanic activity in the vicinity of the Indonesian Seas. For the present discussion, the 3He distribution is considered to be in steady state in the absence of any noticeable anomaly, which would suggest otherwise.
 Finally, the abyssal Indian Ocean is divided into five major basins by a complex ridge system (Figure 1) [Fisher et al., 1982]. This ridge system severely constrains the path of bottom water circulation but is either nonexistent or extremely porous at the depth levels discussed in this study. Nevertheless, for most part of the following discussion it is convenient to divide the deep Indian Ocean into five different basins following the 3500 m isobath. It is well known that bathymetry exercises control on flow patterns and mixing even if there is no physical barrier by a ridge system. Flow patterns often follow bathymetric features that are located several thousand meters deep in the water column [e.g., Toole and Warren, 1993; Toole et al., 1994]. Therefore these basins are constantly referred to, while discussing the 3He distribution and inferring flow directions although they may not be physically present as a distinct entity.
3.1. Mantle 3He Sources in the Indian Ocean
 A map of maximum δ3He values observed during WOCE at each sampling location is shown in Figure 2. Four regions of elevated δ3He (marked S1 through S4) corresponding to the source areas of mantle 3He are seen in this map. Also included in this map are the present-day plate boundaries from Coffin et al. . From this map it is seen that two sources are located on the mid-ocean ridge systems and two sources are located in the convergent margins in the eastern Indian Ocean and Indonesian Seas.
 The strongest source is the mid-ocean ridge source located at roughly 19°S, 67°E on the spreading center in the vicinity of the triple junction of the Central, Southwestern and Southeast Indian Ridges (marked as S1 in Figure 2 [Jean-Baptiste et al., 1992]). The effects of this source are mainly seen in the southeastern Indian Ocean centered at a depth of 2500 m (Figure 3a).
 Another strong source is located in the spreading center located within the Gulf of Aden in the northwestern Indian Ocean at roughly 12°N, 45°E (marked as S2 in Figure 2). This source was identified previously by Jean-Baptiste et al. . Mixing and outflow from the Gulf of Aden transports 3He from the source region into the Indian Ocean. The 3He enrichment due to this source is mainly confined to the western Indian Ocean. The maximum anomaly observed is 22.6% at a depth of 2200 m (Figure 3b). It is also the maximum anomaly observed in the deep Indian Ocean during WOCE.
 Regional 3He enrichment observed in the northeastern Indian Ocean (marked as S3 in Figure 2) reveals a third 3He source. Profiles from this region show 3He enrichment in the 1500–2000 m depth range (Figure 3c). The maximum anomaly in the eastern Indian Ocean is in the Bay of Bengal region (δ3He > 20%). Anomalies as large as 19% are also found within the Andaman Basin which is a tectonically active back arc basin.
 Finally, elevated δ3He values (maximum ∼16%) observed at depths between 1000 and 1500 m in the Indonesian Throughflow region point to an indirect source of 3He to the Indian Ocean (Figure 3d). The origin of this 3He is a combination of both mid-ocean ridge derived 3He in the Pacific Ocean and additional input in the Indonesian Seas. The 3He rich deep Pacific waters enter the Indonesian Seas and pass through a series of deep basins before entering the Indian Ocean [e.g., Gordon et al., 2003]. While in transit in the Indonesian Seas they are further enriched in 3He by local sources [Jean-Baptiste et al., 1997; Top et al., 1997]. The deep component of the through flow transports this 3He into the Indian Ocean.
 The 3He/4He ratios of the injected helium from the above regions are shown in Figure 4. The ratios were calculated based on a recent method developed by Roether et al. . The 3He/4He ratios of the injected helium reveal distinct differences between the individual sources in the Indian Ocean. The 3He/4He ratio in the vicinity of the Central Indian Ridge source is approximately 8–9 times the atmospheric 3He/4He ratio of 1.38 × 10−6. This is typical of mantle helium. Similar ratios have been observed at hydrothermal vents on mid-ocean ridge spreading centers [Craig and Lupton, 1981]. The 3He/4He ratios from the Gulf of Aden and the northeastern Indian Ocean are approximately 5 and 3 times the atmospheric ratio, respectively. Both these ratios are significantly lower than isotopic ratio of helium derived from the mantle. It is difficult to comment on the origin of this difference without a detailed analysis since the differing solubility of helium at different locations could create spurious results. However, if the difference is real the low ratios suggest a source of crustal helium. Low ratios of 3He/4He have been observed before in similar tectonic settings and it has been attributed to a flux of 4He from the crust and sediments [Belviso et al., 1987].
3.2. 3He Distribution
3.2.1. Western Indian Ocean
 WHP Line I7N (Figure 5a) captures the 3He distribution in the western Indian Ocean. 3He anomalies (δ3He > 14%) originating from the source located at the Gulf of Aden can be seen spreading southward from the western Indian Ocean in the 1500–3500 m depth range. This southward spreading is well correlated with other water properties which show a general southward movement of Indian Deep Water (IDW) in the 2000–3500 m depth range [e.g., Wyrtki, 1971; Spencer et al., 1982; Mantyla and Reid, 1995; Reid, 2003]. While this pattern was previously known from the GEOSECS and INDIGO data [Jamous et al., 1992], the new data show much greater detail in this region.
 In order to discuss the 3He distribution in the northwestern Indian Ocean it is convenient to refer to parts of the Arabian and Somali Basins north of the equator as the Arabian Sea following Warren and Johnson . In the deep Arabian Sea, higher 3He enrichment is found farther to the south at the western boundary than in the interior. Maps of δ3He on σ2 = 36.92, σ2 = 37.00 and σ2 = 41.495 isopycnal surfaces roughly at 2000 m, 2500 m and 3000 m depths, respectively (Figures 6b, 6c, and 6d), reveal a southwestward spreading of the 3He signal along the western boundary of the deep Arabian Sea. This is clearly seen in the 3He distribution on the σ2 = 37.00 surface, which shows δ3He values in excess of 14% close to the African Coast as far south as 10°S (Figure 6c). This indicates the presence of a southwestward flowing Deep Western Boundary Current (DWBC) in this region reported by several authors [e.g., Fieux et al., 1986; Dengler et al., 2002]. The southward flow close to the African Coast appears to be stronger at 2500 and 3000 m than at 2000 m since the 3He enrichment at 10°S is higher at these depths than at 2000 m, while the maximum enrichment near the Gulf of Aden is between 1700 and 2200 m. This southwestward spreading is associated with similar southwestward extending high-salinity (34.75–34.80), low-oxygen (140–160 μmol kg−1) and high-silica (130–140 μmol kg−1) tongues in the western Indian Ocean. These features are clearly seen in Mantyla and Reid  and Reid's  property maps on the σ2 = 37.00 surface (Figures 7a, 7b, 7c and 7d).
 In the interior Arabian Sea there is a thick layer (1500–3500 m) of water with high 3He anomalies (δ3He > 14%) north of 10°N which extents eastward to approximately 65°E (Figures 5a and 5b). South of 10°N the Carlsberg Ridge limits the vertical extent of this layer to a depth range between 1500 and 2500 m. From Figures 5a and 5b it is also seen that the bottom waters in the interior Arabian Basin have higher 3He than the bottom waters in the interior northern Somali Basin.
 South of the equator, in the interior Somali Basin, the 3He distribution shows a three-layer structure (Figure 5a). The waters in the depth range of 2000–3000 m are significantly enriched (δ3He > 13%) indicating continuation of southward spreading noted previously. Above the bottom water but below the nominal height of the Carlsberg Ridge (∼3000 m) the waters are moderately enriched in 3He (δ3He about 11–13%), probably due to mixing with waters above. The bottommost waters are characterized by δ3He values around 10%. In this layer potential temperature ranges between 0.9°C and 1.20°C, salinity is relatively low (about 34.72), oxygen values are fairly high (>175 μmol kg−1) and silica values range between 125 and 130 μmol kg−1. The low 3He content and the hydrographic characteristics of these waters are typical of northward flowing Circumpolar Deep Water (CDW) supplying bottom water to the Somali Basin.
 East of 60°E the 3He anomaly in the southern Somali Basin is relatively high and is greater than 14% (Figures 6c and 6d). The high 3He content is associated with relatively low salinity (ΔS ∼ −0.01) and higher oxygen (Δ02 > +5 μmol kg−1) concentrations (Figures 7b and 7c). This is due to a northwestward inflow of 3He-enriched waters from the boundary current at these depths in the Central Indian Basin. This boundary current inflow was first reported by Warren [1981, 1982] and is well supported by observations [Wyrtki, 1971; Mantyla and Reid, 1995; Dengler et al., 2002; Reid, 2003]. A tongue of relatively fresh and oxygen-rich water along the northeastern flank of the Mascarene Plateau can clearly be seen in Reid's maps (Figures 7b and 7c). In this depiction it also appears that this current is a part of the boundary current of the eastern flank of the Central Indian Ridge. The 3He anomaly in the vicinity of this current in the Central Indian Basin is >15%. The anomaly east of 60°E, close to the Mascarene Plateau is also >15%, which agrees well with the inflow of boundary current waters from the Central Indian Basin. However, the fate of this water in the Somali Basin is not very clear. In the southern Somali Basin between the 3He enrichment in the east due to the above inflow and the 3He enrichment farther west due to the southwestward flow is a region of relatively low 3He enrichment extending northward from the Mascarene Plateau to the equator. This low-3He region was first noted by Jamous et al. . It is possible that the 3He-enriched inflow from the east mixes with the 3He poor waters and the 3He signal is quickly diminished to surrounding values.
 The WHP line I7N covers the Amirante Passage and the Mascarene Basin meridionally, whereas the line I2 sampled the northern boundary of this basin zonally. The waters between 2000 and 3000 m are moderately enriched (δ3He about 12–13.5%) in the northern Mascarene Basin close to the Amirante Passage (Figures 5a and 5d). Again this is likely to be the continuation of the southward spreading water originating in the Arabian Sea. This suggests that the IDW from the deep Arabian Sea region is transported south through the gap between Madagascar Island and the Mascarene Plateau via the Amirante Passage. This is in good agreement with observations of IDW at 18°S in the Mascarene Basin by Warren . The northern Mascarene Basin (8°S to 10°S) appears to mark the southern limit of the helium signal originating in the Gulf of Aden. In the interior Mascarene Basin between 10°S and 15°S the 3He anomaly is relatively low (δ3He < 13%) on the σ2 = 36.92, σ2 = 37.00 and σ2 = 41.495 isopycnal surfaces (Figures 6b, 6c and 6d) which indicates influence of water of southern origin.
 The waters deeper than 3500 m in the Mascarene Basin are very uniform and have low 3He content (δ3He < 10%) at all stations in this basin. The low 3He content at depths greater than 3500 m is due to the inflow of waters of circumpolar origin. Numerous studies have observed CDW flowing northward close to the Madagascar Coast in this basin [e.g., Warren, 1981; Mantyla and Reid, 1995; Johnson et al., 1998]. The relatively low 3He anomaly of these waters is similar to observations farther south in the circumpolar waters by Jean-Baptiste et al. .
 Farther south in the Madagascar Basin the main feature is the westward spreading 3He anomaly centered on the σ2 = 37.00 surface at 20°S seen in the I3 section (Figures 5c and 6c). This westward spreading anomaly originates from the source located on the Central Indian Ridge probably due to isopycnal mixing. It appears to be limited to a very narrow zone of less than 4° meridional extent since the 3He enrichment measured at stations 710 and 700 at 18°S and 22°S during I7N and I7S is lower by almost 2%. At greater depths the bottom waters in this basin are characterized by low 3He content with δ3He of 7–8% typical of Circumpolar Deep Waters that ventilate the abyssal Indian Ocean.
 In the southwest corner of the Indian Ocean, in the Mozambique Basin, the δ3He values are the lowest measured anywhere in the Indian Ocean. The δ3He values are on the order of 6–9%. The low 3He content of the deep waters of this basin is mainly due to the influence of 3He-poor North Atlantic Deep water that fills this deep basin [Toole and Warren, 1993; Mantyla and Reid, 1995; Reid, 2003].
3.2.2. Central and Eastern Indian Ocean
 The WOCE Line I3 is a zonal section cutting across the Central Indian Basin along 20°S (Figure 5c). A mid-depth eastward spreading 3He anomaly is strikingly visible in this section. The eastward extending tongue of 3He anomaly is clearly seen emanating from the Central Indian Ridge (65°E and 2500 m). The anomaly spans the Central Indian Basin and spills into the West Australian Basin by overflowing through saddles in the Ninetyeast Ridge. An envelope of δ3He values greater than 14% at 2000–3000 m depth range is clearly seen in the I9N section (Figure 5f) and extends to about 100°E. This indicates a strong deep eastward flow and is in very good agreement with the recent discovery of eastward flow from the Central Indian Basin into the West Australian Basin at 28°S in the depth range between 2000 and 3400 m by Warren and Johnson .
 Farther north, at 8°S, at depths of 2000–3500 m higher 3He enrichment is found in the trough between the Central Indian Ridge and the Mascarene Plateau than in the Central Indian Basin (Figure 5d). This suggests that the most of the DWBC flowing northward along the Central Indian Ridge [Warren, 1981, 1982] branches into a northwestward flowing DWBC along the Mascarene Plateau supplying water to the southern Somali Basin as noted previously.
 In the Central Indian Basin the main feature is a 1000 m thick layer with δ3He > 14% that shoals upward eastward. Warren and Johnson  have inferred westward inflow of waters from the West Australian Basin into the Central Indian Basin at 10°S and 5°S in the 2000–3500 m depth range. The 3He data in the vicinity of the Ninetyeast ridge at the overflow latitudes do not provide unambiguous evidence for overflow but the effects of this westward mid-depth flow are also seen in the I2 section. The shoaling of the lower 14% isopleth from 3500 m at the Central Indian Ridge to approximately 2200 m near the Ninetyeast Ridge is likely due to the inflow mentioned above. Additional indication of the branching of the DWBC westward and westward flow comes from the 3He distribution along 80°E in the Central Indian Basin (Figure 5e). In this section the 3He anomaly of magnitude greater than 15% extends to about 12°S which is the approximate latitude of branching and the westward inflow from the West Australian Basin. The absence of 3He enrichment greater than 15% in the Central Indian Basin north of this latitude is consonant with the branching of the DWBC and westward inflow from West Australian Basin. The extent of the mid-depth 3He anomaly, the eastward flow in the Southern Central Indian Basin and the branching of the DWBC are clearly visible in maps of δ3He on σ2 = 36.92, σ2 = 37.00 and σ3 = 41.495 isopycnal surfaces (Figures 6b, 6c and 6d). In the region of the 3He maximum, potential temperature ranges between 1.5°C and 1.6°C while salinity and oxygen are in the range of 34.72–34.73 and 170–175 μmol kg−1, respectively. The 3He anomaly north of the equator is about 13.5% compared to about 14.5% south of the equator.
 In the northeastern Indian Ocean, high 3He anomalies (δ3He about 15%) can be seen spreading southward from the Bay of Bengal region in the 1500–3000 m depth range (Figure 5f). This high 3He anomaly is due to sources located in the convergent margins discussed above and the southward spreading is probably due to mixing. Within the Bay of Bengal region (Figure 5a) the 3He anomaly decreases from east to west with a pronounced westward spreading near the bottom. Direct measurements in this region suggest an anticlockwise near bottom flow in this region [Gordon et al., 2002]. The 3He distribution is consistent with this circulation scheme.
 The bottom water in the Central Indian Ocean (δ3He about 12%) is relatively enriched compared to the bottom waters in the western and eastern Indian Ocean (δ3He about 9%). This distribution illustrates the unique deep circulation in this basin. The bottom waters deeper than 3800 m are derived from the West Australian Basin through saddles in the Ninetyeast Ridge at 10°S and 5°S [Warren, 1981, 1982; Warren and Johnson, 2002]. Therefore the bottom waters of the Central Indian Basin should have greater 3He enrichment than both the western and eastern basins since the bottom waters in the latter two basins are derived directly from the 3He depleted southern ocean waters, whereas the bottom water in the Central Indian Ocean have been enriched by mixing during transit through the West Australian Basin.
 Bottom water in the West Australian basin has a relatively low anomaly, indicating inflow of Circumpolar Deep Water (Figures 5c, 5d and 5f). This low-δ3He layer is thickest south of 10°S. Bottom water north of 10°S is relatively enriched in 3He due to the mixing with 3He rich waters above. In the West Australian basin the least enriched waters are found close to the eastern flank of the Ninetyeast Ridge in the region of the boundary current transporting 3He poor bottom waters northward (Figure 5c). Farther east this low-3He anomaly layer is less thick suggesting recirculation and southward flow of relatively 3He rich waters in the eastern part of the basin. This agrees well with other descriptions of abyssal flows in this region [Warren, 1981; Mantyla and Reid, 1995].
 At intermediate depths (1000–1500 m) in the latitude range of 5°S–15°S the 3He enrichment associated with the deep through flow from the Indonesian Seas is clearly discernable in the sections along 8°S, 95°E, 80°E and the δ3He map on the σ1.5 = 34.37 surface (Figures 5d, 5e, 5f and 6a). This 3He enrichment is associated with the influx of 3He rich intermediate depth relatively fresh water from the Banda Sea known as Banda Sea Water or Banda Sea Intermediate water [e.g., Wyrtki, 1961; Rochford, 1966; Warren, 1981; Wijffels et al., 2002]. The 3He anomaly due to this through flow is on the order of 14–15% close to the eastern boundary of the Indian Ocean. The δ3He map on σ1.5 = 34.37 surface at roughly 1300 m shows the extent of the 3He input from the Indonesian Seas. The core of the 3He signal is observed to spread westward between at 10°S and 12°S from the eastern Indian Ocean to about 70°E. The 3He maximum is associated with a potential temperature range between 4.7°C and 5.0°C going from east to west along this isopycnal. Salinity and oxygen are less than 34.70 and 90 μmol kg−1, respectively.
4. Discussion and Conclusions
 The vastly improved coverage obtained from the new measurements point out several interesting features of the nature of the 3He sources in the Indian Ocean. While the source areas of 3He were generally known from previous measurements, their strength and nature were not well known. Perhaps the most interesting result with respect to the sources is that the deep Indian Ocean receives input from sources both within and outside the Indian Ocean. Among the sources within the Indian Ocean are two essentially point sources located on the mid-ocean ridges and several distributed sources in the convergent margins in the eastern boundary. The deep core of the Indonesian Throughflow brings 3He into the Indian Ocean from sources located in the Pacific Ocean and the Indonesian Seas. This unique distribution of sources in the Indian Ocean has important implications for modeling studies using 3He. These studies require a source function for the release of 3He into the deep ocean. For example, in one of the first attempts to simulate the global oceanic 3He distribution, Farley et al.  linked a simple mid-ocean ridge source function to an oceanic general circulation model. The source function used in this study assumed an even distribution of sources along mid-ocean ridges only. Because of lack of evidence they did not include sources of 3He other than mid-ocean ridge spreading centers. As results from new data demonstrate, the assumption of uniformly located sources may not be appropriate in the Indian Ocean. The new data can be used to construct a more accurate source function.
 Another important result with respect to sources is the contribution of convergent margins to the Indian Ocean. The northeastern boundary of the Indian Ocean is a convergent margin associated with the Sunda Arc extending from the Indonesian Seas into the Bay of Bengal region (Figure 2). Here the Indian and the Australian Plates are subducted under the Eurasian Plate and this is a region of active volcanism [Condie, 1997]. Transient releases of energy and mantle volatiles have been observed at submarine volcanic areas at convergent margins [Belviso et al., 1987]. The observed anomaly in the eastern Indian Ocean suggests that convergent margins are a significant source of mantle 3He to the Indian Ocean. However, to quantify the contribution of convergent margins to the 3He budget in the Indian Ocean the contribution of the Pacific and Indonesian Seas origin 3He to the deep component of the Indonesian Throughflow must be known. It is likely that future studies will resolve this issue.
 Previous hydrographic and tracer studies have shown that in the Indian Ocean inflowing CDW is transformed into IDW and a portion of this IDW returns south between 2000 and 3500 m [Warren 1981; Broecker et al., 1985]. The 3He distribution supports the conclusion of these studies while pointing out additional pathways of deep water flow. The 3He distribution presents a long-term averaged view and complements descriptions of flow obtained from snapshots of the density fields. The deep circulation scheme derived from the 3He distribution is sketched in Figure 8 and is discussed below.
 In the western Indian Ocean, the 3He distribution confirms a pathway for the southward flow of IDW from the northern Indian Ocean. In the deep Arabian Sea the data indicate a southwestward flow of IDW in the 2000–3000 m depth range close to the African Coast. The deep flow in this area has been the subject of considerable attention. A number of surveys in the area have shown differing properties at different times of the year suggesting flows seemingly aligned with the upper ocean monsoon winds [Warren et al., 1966; Fieux et al., 1986; Schott et al., 1989; Warren and Johnson, 1992; Beal et al., 2000; Dengler et al., 2002]. Warren and Johnson  applied the Stommel and Arons  model to the deep water layer (2000–3500 m) in the Arabian Sea. In their model the upwelling from the bottom of the layer exceeded the upwelling from the top in order to sustain a southward flow in the layer. The resulting interior flow lines were directed southwestward and the model predicts a southwestward boundary current along the African Coast. The 3He distribution is in good agreement with these results. It may not be possible to comment on the flow reversal with the 3He data considering the range of the enrichment and the analytical uncertainties but the data do suggest that the mean flow is directed southwestward in this basin, which is consistent with other property distributions.
 IDW has been observed flowing southward at subtropical latitudes by a number of authors [e.g., Warren 1981; Toole and Warren, 1993]. On the basis of the 3He distribution in the Arabian Sea, Amirante Passage and northern Mascarene Basin it is clear that some portion of the IDW found in the subtropical South Indian Ocean owes its origin to southward flow in the Arabian Sea and that it enters the South Indian Ocean via the Amirante Passage. Within the Amirante Passage, Johnson et al.  found southward flow of IDW to be much larger than the northward flowing CDW beneath it. In a scheme where IDW is transformed CDW this results in a mismatch of volume fluxes in the Amirante Passage. Johnson et al. suggest that CDW upwelling in the eastern basins might flow westward north of the Amirante Passage at depths unconstrained by topography and join with the southward flow close to the African Coast which then flows south through the Amirante Passage. One possible pathway that supplies IDW to the Somali Basin is the westward flow into the Somali Basin that has been observed along the northeastern flank of the Mascarene Plateau [Wyrtki, 1971, Warren 1981, Dengler et al., 2002]. This westward flow is observed in the 3He distribution and it brings 3He rich waters into the southeastern Somali Basin. Whereas it is possible that the westward inflow along the Mascarene Plateau contributes to the southward flow in the Amirante Passage, its pathway in the Somali Basin is not clear. As noted previously, there is a region of low-3He anomaly between this helium enrichment and the 3He enrichment farther west at the African Coast. It is likely that this water mass is substantially modified before it reaches the western Somali Basin. These conclusions based on the 3He distribution are similar to the conclusions made by Dengler et al.  based on hydrographic data.
 In the Central Indian Basin an important result is evidence for prominent eastward flow in the south Indian Ocean. The eastward flow from the Central Indian Basin to the West Australian Basin deduced from the 3He data confirms the recent discovery of eastward flow at 28°S over the saddles in the Ninetyeast Ridge by Warren and Johnson . A question raised in their study is whether the eastward flow upon entering the West Australian Basin turns north as a boundary current along the Ninetyeast Ridge and then returns to the Central Indian Basin through the westward flows at 10°S and 5°S saddles and further does this westward flow at 10°S contribute to the southward flow in the Amirante Passage via the boundary current observed along the northern flank of the Mascarene Plateau. The 3He data do not resolve this question unambiguously but the 3He distribution in the Central Indian Basin and at 8°S can be reconciled with this scenario.
 Another finding is the westward spreading 3He signal in the intermediate layer in the eastern Indian Ocean. As noted previously this westward spreading is associated with the influx of Banda Sea Intermediate Water into the Indian Ocean. The Banda Sea Water has its origins in the intermediate waters of the Pacific Ocean and its presence in the Indian Ocean represents the deep component of the through flow between the Pacific and Indian Oceans [Warren 1981; Gordon, 1986; Wijffels et al., 2002]. Observations suggest that the deep component may contribute a significant fraction of the total through flow [Fieux et al., 1996; Molcard et al., 1996; Jean-Baptiste et al., 1997] while a coupled climate model study Banks  suggests that the magnitude of deep through flow is dependent on the magnitude of the deep inflow into the Pacific and Indian Oceans. The 3He data are likely to be a valuable tool for further study of the deep component of the through flow.
 In conclusion, this study presents an overview of a large helium data set in the Indian Ocean. The results point out new features with respect to the sources and deep circulation. The main points are summarized below:
 1. The deep Indian Ocean receives significant 3He input from sources located within and outside the Indian Ocean.
 2. The measurements clearly map the extent of 3He input from the previously known sources located in the Central Indian Ridge and at the Gulf of Aden and reveal new sources in the convergent margins in the eastern Indian Ocean.
 3. The deep component of the through flow from the Pacific Ocean transports 3He of deep Pacific and Indonesian Seas origin into the Indian Ocean.
 4. The 3He distribution reveals a deep eastward flow in the Central Indian Basin, in the 2000–3000 m depth range, which overflows into the West Australian Basin through the saddles in the Ninetyeast Ridge.
 5. In the western Indian Ocean the 3He distribution suggests southwestward flow of Indian Deep Water along the African Coast. Further, the data suggest that at least some of this water mass enters the south Indian Ocean through the Amirante Passage.
 6. In the eastern Indian Ocean, the influx of Banda Sea Intermediate Waters causes 3He enrichment in the 1000–1500 m depth range between 10°S and 12°S. The magnitude of the enrichment suggests a significant influx of this water mass.
 The WOCE data are likely to be used extensively in future studies. Further, as more of the Indian Ocean is mapped for hydrothermal activity, the source function will be better defined, resulting in a better resolved 3He field and the implied deep flow.
 The authors greatly appreciate the comments of two anonymous reviewers whose careful reading and insightful comments significantly improved this manuscript. We thank the personnel who collected and processed the helium isotope data at UM, LDEO, PMEL and WHOI. National Science Foundation support is acknowledged for the UM part of the work through grants OCE-9820131 and OCE-998150. Support for the LDEO portion of the work was obtained from the National Science Foundation through awards OCE 94-13162 and OCE 98-20130. This is LDEO contribution number 6598.