Mantle melting and magma supply to the Southeast Indian Ridge: The roles of lithology and melting conditions from U-series disequilibria

Earth and Planetary Science Letters 278 (2009) 55–66

Contents lists available at ScienceDirect

Earth and Planetary Science Letters j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / e p s l

Mantle melting and magma supply to the Southeast Indian Ridge: The roles of lithology and melting conditions from U-series disequilibria Chris J. Russo a,b,⁎, Ken H. Rubin b, David W. Graham a a b

College of Oceanic and Atmospheric Sciences, Oregon State University, Corvallis, OR 97331 USA School of Ocean and Earth Science and Technology, University of Hawaii, Honolulu, HI 96822 USA

a r t i c l e

i n f o

Article history: Received 19 May 2008 Received in revised form 14 November 2008 Accepted 14 November 2008 Available online 30 December 2008 Editor: R.W. Carlson Keywords: U-series disequilibria mid-ocean ridge basalt Southeast Indian Ridge mantle melting mantle lithology pyroxenite

a b s t r a c t Intermediate-spreading Southeast Indian Ridge (SEIR) basalts display geographic gradients from 90°E to 118°E in Th/U, (230Th/232Th) and 238U–230Th–226Ra disequilibria (230Th-excesses of 1 to 21% and 226Raexcesses of 0 to 160%). Highly correlated (238U/232Th) and (230Th/232Th) (r2 = 0.94) span nearly the entire global MORB range; basalts from three of four ridge morphologies form subparallel, vertically stacked arrays on an equiline diagram. (226Ra/230Th) is inversely correlated with (230Th/238U) within sub-regions of the study area. There is a convex (230Th/238U) along-axis pattern (lower 230Th-excesses in the west and east, at shallowest and deepest depths), such that (230Th/238U) does not correlate with axial depth as described elsewhere, despite the ~ 2300 m west-to-east axial depth increase. κ and κPb also vary along axis from κ b κPb in the west to κ ≅ κPb in the east. Variations in axial depth, crustal thickness, U-series disequilibria, κ and κPb are best explained by forward model scenarios in which the physical characteristics of melting (such as melting rate, degree of melting, porosity and melt initiation depth) vary along axis in response to an inferred long wavelength temperature gradient and small variations in the underlying source lithology (pyroxenite veins). This leads to decreased melt supply from west to east, a concomitant change in axial morphology and axial magma chamber depth, and systematic long wavelength geochemical gradients. Melting rate covaries with melt supply in the western and central regions, decreasing eastward along the SEIR, but supply and rate vary inversely from the central to the eastern regions, because cooler eastern mantle contains a few percent of fusible pyroxenite veins that melt productively to generate melts with high Th–U–Ra concentration but low 230Th-excess and 226 Ra-excess. The veins contribute only a small proportion of the total melt volume from mantle that is N 95% peridotite. Th, Pb and He isotopic gradients are consistent with a changing proportion of enriched pyroxenite veins along axis having a roughly 0.5 to 1 Gyr reservoir age. © 2008 Elsevier B.V. All rights reserved.

1. Introduction Magmas produced beneath mid-ocean ridges (MORs) are indirect probes of upper mantle composition and melting conditions. Geographic variations in the conditions and extent of melting, and their linkages to mantle temperature, lithology and chemical composition are central to an understanding of mantle evolution and the origin of mid-ocean ridge basalts (MORBs). Important patterns in melt generation are revealed by the relations between MORB chemistry and physical attributes of MORs (e.g., Klein and Langmuir, 1987; Langmuir et al., 1992; Niu and Batiza, 1993; Shen and Forsyth, 1995; Bourdon et al., 1996b; Niu and Hekinian, 1997; Rubin and Sinton, 2007; Niu and O'Hara, 2008). Global studies of MORB composition must often make simplifying assumptions to relate compositional variations to changes in a particular physical char⁎ Corresponding author. School of Ocean Engineering Science and Technology, University of Hawaii, Honolulu, HI 96822 USA. Tel.: +1 808 956 9328. E-mail address: [email protected] (C.J. Russo). 0012-821X/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.epsl.2008.11.016

acteristic, such as axial depth or spreading rate (Klein and Langmuir, 1987; Bourdon et al., 1996a; Lundstrom et al., 1998). This approach makes it difficult to decipher the competing effects of several independent parameters on MOR magmatism. Regional scale studies are a useful complement to global studies, particularly in regions with a limited variation in one or more key variables (e.g., spreading rate, axial depth), or where physical, morphologic and/or chemical gradients in ridge character result from a coherent geologic history. Here we present results of a regional scale U-series isotopic and geochemical study of melting conditions at the intermediate spreading rate Southeast Indian Ridge (SEIR) between 90°E and 118°E, an ~ 2100 km long region displaying significant along-axis physical and chemical variations that can be used to constrain mantle melting and source lithology. Uranium series disequilibria in MORBs record variations in U–Th– Ra elemental fractionation due to the type (e.g., fractional, batch), extent, rate and depth of melting beneath MORs (e.g., McKenzie, 1985, 2000; Williams and Gill, 1989; Spiegelman and Elliott, 1993; Bourdon et al., 1996a,b; Stracke et al., 1999; Spiegelman, 2000; Sims et al.,

56

C.J. Russo et al. / Earth and Planetary Science Letters 278 (2009) 55–66

2002). U-series disequilibria also reflect MORB source mineralogy (e.g., garnet modal abundance, which dominates U–Th fractionation, and clinopyroxene mode, which largely controls fertility and thus melting rate; Beattie, 1993; LaTourette et al., 1993; Pertermann and Hirschmann, 2003). There is a temporal aspect to observed U-series disequilibria, because parent-daughter activity ratios relax to a state of secular equilibrium at a rate governed by their half-lives following U– Th–Ra fractionation. We use these attributes in conjunction with chemical (Sours-Page, 2000) and radiogenic isotope compositions (Graham et al., 2001, 2006; Mahoney et al., 2002) of the same SEIR samples to test scenarios for physical variations in melting conditions of a heterogeneous mantle source. 2. Regional geologic background The modern SEIR (Fig. 1) marks the Australian-Antarctic plate boundary between the Rodrigues triple junction (25°S, 70°E) and the Macquarie triple junction (62°S, 151°E). This section of MOR contains no large transform offsets and exhibits a west-to-east gradient in axial depth, from ~2300 m to N4500 m (Cochran et al., 1997). The SEIR full spreading rate is nearly constant, from ~72 to 76 mm/yr across our study area (Cochran et al., 1997). The SEIR has a bathymetric high at the Amsterdam-St. Paul Plateau (ASP, 76°–78°E; Graham et al., 2001) and unusually deep, magma-starved ridges within the Australian-

Antarctic Discordance (AAD; e.g., Klein et al., 1991; Sempéré et al., 1991). The west to east axial depth increase has been attributed to a long wavelength decrease in mantle temperature (Cochran et al., 1997; Sempéré et al., 1997), estimated to be ~80–150 °C using major element variations in N3,000 SEIR MORB glasses from the ASP (Douglas-Priebe, 1998) to the AAD (Klein et al., 1991; Pyle, 1994; Sours-Page, 2000). 2.1. Along-axis morphologic variations Ridge morphology and crustal thickness variations indicate changes in melt supply along this portion of the SEIR. Three distinct axial morphological ‘zones’ were discovered by a detailed geophysical survey in 1994/1995 (Cochran et al., 1997; Sempéré et al., 1997). Most western zone ridge segments (C17–C13; Fig. 1) have axial-high morphology, similar in dimension and shape to many segments of the fast-spreading East Pacific Rise (EPR). Segments C12–C4 have a ‘transitional’ morphology typical of other intermediate-spreading ridges (e.g., Canales et al., 1997). Cochran et al. (1997) subdivided the transitional morphologies into: (a) rifted axial highs in the west (segments C12–C9), and (b) shallow axial valleys in the east (segments C8–C4). SEIR segments C3 and C2 have a deep axial valley, similar to much of the mid-Atlantic ridge (MAR). We use the following abbreviations throughout this paper to refer

Fig. 1. Satellite altimetry of the Indian Ocean basin between 40°E and 130°E and sample location map. Labeled features include: the Southwest Indian Ridge (SWIR), Central Indian Ridge (CIR), Rodrigues Triple Junction (RTJ), Amsterdam-St. Paul Plateau (ASP), Kerguelen Plateau (KP), Southeast Indian Ridge (SEIR), and the Australian-Antarctic Discordance (AAD). Enlarged panel defines ridge-axis trace of the SEIR from 40°S to 52°S and from 88°E to 118°E. Individual ridge segments are identified above the axial trace (C17 through C2), dredge locations for samples in this study are shown by black dots along the ridge-axis and are labeled below the axial trace (D71 through D145). Ridge-axis color in the online version indicates depth variation with red shades representing the shallowest segments and blue the deepest. We consider three geographical regions in this study (western, WG, central, CG, and eastern, EG, groups). The transform fault separating segments C13 and C12 forms a natural WG-CG divide. MORBs west of this transform tend to have higher 3He/4He, and different Sr, Nd, Pb isotope compositions, suggesting different mantle geochemical histories east and west of 100°E (Mahoney et al., 2002). The transition into deep-axial valley morphology between segments C3 and C4 forms the EG and CG divide.

C.J. Russo et al. / Earth and Planetary Science Letters 278 (2009) 55–66

to SEIR basalts from these morphologic groups: Axial–High (AH), Rifted–Axial–High (RH), Shallow–Axial–Valley (SV), and Deep– Axial–Valley (DV). There are significant variations in upper crustal seismic structure along segments C12–C7 (Baran et al., 2005) and a crustal thickness gradient from ~6.3 km at 101° E to ~3.8 km at 115° E (Holmes et al., in press). The westernmost RH segment (C12) has the shallowest axial magma chamber (AMC), thinnest layer 2A (Baran et al., 2005) and greatest crustal thickness (~6.3 km between 100° and 116°E; Holmes et al., in press). Baran et al. (2005) divided segment C12 into a western AH and eastern RH segment, which we adopt for the dredge 96 and 100 morphological assignments (Table 1). There is currently no seismic data for the other AH segments (C17–C13). The AMC is ~ 1 km deeper and the crust is slightly thinner (~ 5.9 km) on RH segments. The SV segments do not have detectable AMCs. The DV segments have the thinnest crust (~ 4.6 km). They are outside the Baran et al. (2005) study area but likely lack an AMC (like most of the slow-spreading MAR; Detrick et al., 1990).

57

porosity (ϕ), and melting rate (M) for three mantle lithologies, as described in Section 5.3 and the supplement. 4. Results Analytical results are presented in Table 1; activity ratios are denoted by parentheses in the table and throughout the text. (234U/ 238 U) is effectively indistinguishable from unity in each of our samples (0.998 ± 0.004 to 1.005 ± 0.003), indicating that post-eruptive alteration/exchange with seawater has not affected U-series activities in this sample set. 4.1. Ba–Th–U concentrations Ba, Th, and U concentrations vary considerably in SEIR basalts (by factors of 8, 9 and 6, respectively) with much less variation in Ba/Th (2.3 times) and Th/U (1.6 times). Higher concentrations are more common in the east. Ba, Th, and U concentrations are each strongly correlated (r2 N 0.9) with (La/Sm)n over a threefold (La/Sm)n range.

2.2. Along-axis geochemical variations 4.2. (230Th/232Th) and (238U/232Th) All SEIR lavas discussed here are ‘Indian’-type MORBs (Mahoney et al., 2002), with lower εNd, higher 87Sr/86Sr and elevated Δ208Pb/ 204 Pb (after Hart, 1984) compared to Pacific and North Atlantic MORBs. Eastward decreases in Fe8 and increases in Na8 suggest a gradient in the mean pressure and extent of melting (Sours-Page, 2000; Graham et al., 2001; Mahoney et al., 2002; Fe8 and Na8 are fractionation corrected oxide abundances, Klein and Langmuir, 1987). Broadly correlated 3He/4He and Fe8 suggest that high 3He/ 4 He mantle is preferentially sampled by deeper melting in the west (Graham et al., 2001). Conversely, the most strongly ‘Indian’-type mantle is associated with deeper axial depths and by inference, lower degrees of melting in the east (Mahoney et al., 2002). A heterogeneous upper mantle with lower 3He/4He and higher Δ208Pb/ 204 Pb material concentrated in the shallowest mantle and preferentially sampled by shallower melting at lower mantle temperature in the east, explains the Fe8–3He/4He–Δ208Pb/204Pb relationship (Graham et al., 2001; Mahoney et al., 2002). Sr–Nd–Pb isotope data indicate little, if any, eastward mantle flow from the distant ASP or Kerguelen–Heard hotspots into the study region (Mahoney et al., 2002; Nicolaysen et al., 2007). Loose correlations of Sr, Nd, Pb and Hf isotopic compositions with their respective parent/daughter ratios have been interpreted as either mixing trends or 200–700 Myr reservoir ages (perhaps reflecting small amounts of trapped mantle melt from Gondwana breakup; Mahoney et al., 2002; Graham et al., 2006). 3. Samples and methods Sixteen basalt glass samples from 13 ridge segments, selected to represent the regional morphologic, geographic and geochemical variations within the study area (Fig. 1 and Table 1), were analyzed by thermal ionization mass spectrometry for Th–U–Ba concentrations and U and Th isotopic compositions; a 12 sample subset was analyzed for 226Ra concentration. The MORB U-series analytical protocols used in this study are described in Rubin et al., 2005 (and references therein). Alteration-free glass samples were handpicked under a binocular microscope (see Table S6 for sample conditions and weights; the letter “S” indicates tables or figures that are in the supplement). Analytical blank and standards meta-data are reported in the Table 1 footnotes. Scenarios for producing observed U-series disequilibria were forward modeled, using both dynamic melting (McKenzie, 1985, Williams and Gill, 1989) and equilibrium porous flow (Spiegelman and Elliott, 1993; Spiegelman, 2000) models. Multiple model simulations were run to examine the relative roles of melt fraction (F), mantle

SEIR basalts form a continuous array on an equiline diagram, from enriched (low) to depleted (high) (238U/232Th) values (Fig. 2) and span a larger range than any other ridge except the ultra-slow spreading Southwest Indian Ridge (SWIR), which has a similar range but noncontinuous data distribution (Standish, 2006). The SEIR (238U/232Th) range is greater than other intermediate spreading rate ridges such as the Gorda and Juan de Fuca (JDF) Ridges (Goldstein et al., 1992) and the eastern Galápagos Spreading Center (GSC; Kokfelt et al., 2005). Dividing the SEIR data into western, central and eastern geographic groups (WG, CG and EG; Table 1 and Fig. 2) reveals regional differences in our sample suite. SEIR basalts can also be divided into three slightly different relative source enrichment groups based on Th/U, forming distinct fields on an equiline diagram (Fig. 3A): enriched (≥3.4, n = 5), intermediate (2.8 b Th/U b 3.3, n = 7), and depleted (≈2.4, n = 4). Thorium concentrations in the intermediate and enriched groups overlap (0.17 to 0.58 and 0.25 to 1.16 µg/g, respectively), with implications for variations in melting style and mantle source composition (see section 5.4). 230 Th–238U disequilibria vary systematically along the SEIR. (230Th/ 232 Th) and (238U/232Th) are highly correlated in samples having 226Raexcesses (r2 = 0.94; Fig. 3A). Because the data lie to the left of the equiline (the locus of secular equilibrium), the linear regression slope of 0.90 indicates slightly higher average 230Th-excesses at the enriched end (~ 16%) than the depleted end (~6%) of this array. A similar relationship in other MORBs has been loosely interpreted to reflect a lithologic control on melting conditions (e.g., Goldstein et al., 1992; Rubin and Macdougall, 1992; Lundstrom et al., 1998). 4.3. Post-eruptive decay effects on (230Th/238U) and (226Ra/230Th) Sample age is particularly important in U-series studies because initial disequilibria decay away at a rate governed by the shorter halflife of a parent-daughter pair (75.69 kyr and 1.6 kyr for 230Th and 226 Ra, respectively). It is difficult to ensure that most MORBs are young enough to be unaffected by post-eruptive decay (especially those sampled by dredging, such as the SEIR basalts). Excess 226Ra, which is commonly observed in young MORBs, occurs for ~ 8 kyr following eruption (e.g., Rubin and Macdougall, 1988), making it a good indicator of sample youth (i.e., (230Th/238U) will not have been changed by post-eruptive decay in lavas with (226Ra/230Th) N 1). Assessing sample youth was the primary purpose of 226Ra analyses in this study. Eleven of twelve SEIR basalts analyzed for 226Ra have excess 226Ra (ranging from 15 ± 4% to 161 ± 4%) and lie on or within a computed error envelope around the Fig. 3A linear regression. The

58

Table 1 Southeast Indian Ridge U-series disequilibria and Th, U, Ra and Ba concentrations measured by thermal ionization mass spectrometry (TIMS) Sample (WW10-XX-x)

Lat (S)

Long (E)

Depth (m)

E or N MORB

Ridge morphology

[Th] ng/g

±

[U] ng/g

±

(234U/ U)

±

238

232 230

Th/ Th

±

(230Th/ Th)

±

232

(238U/ Th)

±

232

(230Th/ 238 U)

±

90.8 92.7 94.8 95.9 97.5

2350 2585 2719 2660 2470

N N N N E

AH AH AH AH RH

149.1 156.4 133.0 130.4 584.2

0.3 0.2 0.2 0.1 0.6

62.4 64.7 56.6 55.0 207.6

0.3 0.3 0.3 0.3 0.9

1.003 1.003 1.003 1.000 1.001

0.003 0.004 0.003 0.004 0.003

137500 139500 130500 142700 149200

1200 1300 1200 1300 1400

1.346 1.327 1.418 1.297 1.240 1.33

0.012 0.012 0.013 0.012 0.012 0.13

1.270 1.254 1.292 1.279 1.078

0.006 0.006 0.006 0.006 0.005

1.059 1.058 1.098 1.014 1.151 1.08

0.011 0.011 0.011 0.010 0.012 0.10

Central group (CG) 96-1 47.3 100-1† 47.6 105-1† 47.8 113-7† 48.8 118-1 48.4 125-1 49.5 126-7 49.5 † 132-1 50.2 141-1 50.4 CG averages

100.7 101.5 103.0 105.2 107.5 109.1 109.5 111.8 113.6

2465 2857 2783 3630 2673 3475 3240 3328 3002

E E E N E N E E E

AH RH RH SV RH SV SV SV SV

424.6 548.0 292.1 269.9 943.0 169.3 1158.3 279.7 432.3

0.4 0.5 0.3 0.3 1.1 0.2 1.3 0.3 0.6

149.6 168.4 92.3 82.7 267.3 58.0 312.8 81.6 142.4

0.7 0.8 0.4 0.4 1.2 0.3 1.4 0.4 0.6

0.999 1.001 1.000 1.002 1.002 0.998 1.001 1.003 0.998

0.004 0.004 0.004 0.004 0.002 0.003 0.003 0.004 0.004

162100 164000 161300 167300 178900 158100 192100 188600 165900

1700 2000 1500 1700 1700 1600 1700 2900 2000

1.142 1.129 1.148 1.106 1.035 1.170 0.963 0.981 1.116 1.09

0.012 0.014 0.011 0.011 0.010 0.012 0.009 0.014 0.013 0.15

1.069 0.933 0.959 0.930 0.860 1.038 0.819 0.885 0.975

0.005 0.004 0.004 0.004 0.004 0.005 0.004 0.004 0.005

1.068 1.210 1.197 1.190 1.203 1.127 1.176 1.108 1.145 1.16

0.012 0.016 0.012 0.013 0.013 0.012 0.012 0.018 0.015 0.10

Eastern group (EG) 144-4 50.0 145-1† 49.3 EG averages

115.2 116.7

3997 4665

N E

DV DV

247.3 694.4

0.3 0.7

72.7 183.7

0.3 0.8

1.005 1.002

0.003 0.003

183700 210700

1800 2600

1.007 0.878 0.94

0.010 0.011 0.18

0.892 0.803

0.004 0.004

1.129 1.094 1.11

0.012 0.014 0.05

±

(226Ra/ Th)

±

230

36.4

0.6

1.63

0.03

27.6 49.1 80.8

0.7 0.5 0.5

1.32 2.61 1.00 1.47

0.04 0.04 0.01 0.44

111.5

4.2

2.03

0.04

165.6 30.3 189.4 57.2 95.4

1.4 0.3 2.6 1.1 1.0

1.53 1.38 1.53 1.87 1.73 1.68

0.03 0.02 0.03 0.05 0.12 0.50

31.8 89.8

1.0 0.7

1.15 1.32 1.24

0.04 0.02 0.24

Ba µg/g 22.5 21.6 19.1 13.1 54.4

32.7 40.6 19.7 20.7 74.1 12.0 110.3 22.6 35.3

18.1 75.3

Table notes: 1. Isotope dilution concentrations and isotopic compositions were measured on the Univ. of Hawaii VG Sector 54 thermal ionization mass spectrometer equipped with a high abundance sensitivity (WARP) filter and ion-counting Daly detector (see also Rubin et al., 2005). 2. All measurements were made on splits from the same sample dissolution. Reported errors are 2σ. 3. Activities are denoted by ( ); Decay constants used to calculate activities are: λ226 = 4.33217 × 10− 4 yr− 1, λ230 = 9.1577 × 10− 6 yr− 1, λ232 = 4.948 × 10− 11 yr− 1, λ234 = 2.8263 × 10− 6 yr− 1, λ238 = 1.55125 × 10− 11 yr− 1. 4. Typical procedural blanks on concentration measurements were [Ba] = 2.5 × 10− 10 g; [Th] = 2.0 × 10− 12 g; [U] = 4.0 × 10− 12 g. 5. Duplicate 232Th/230Th measurements on the same dissolutions for samples denoted by † are given in the supplementary material. 6. Ridge morphologies are AH = axial high, RH = rifted high, SV = shallow valley, DV = deep valley; see section 2 of the text. 7. Sea water alteration of samples can be detected with (234U/238U), which is 1 in fresh, unaltered lavas and 1.14 in seawater (e.g., Thurber, 1967). (234U/238U) is effectively indistinguishable from unity in each of our samples. 8. Regional Group averages are given only for parameters that are discussed in the text. The WG (226Ra/230Th) average does not include the secular equilibrium sample. 9. Mean values for replicate analyses of Th (UCSC-A) and U (CRM112A) isotope standards are; 232Th/230Th = 170863 ± 2844 (2σ, n = 31) and 234U/238U = 5.2875 ×10− 5± 1.024 ×10− 7 (2σ, n = 10), respectively. Analyses of the in-house rock standard K1919 conducted during this study are reported in Rubin et al. (2005).

C.J. Russo et al. / Earth and Planetary Science Letters 278 (2009) 55–66

Western Group (WG) 71-1 42.9 75-4 43.6 78-2 44.8 84-7 45.1 89-4 47.5 WG averages

Ra fg/g

C.J. Russo et al. / Earth and Planetary Science Letters 278 (2009) 55–66

59

MORBs (~ 4). Instead, this basalt is likely substantially younger than 1600 years. (226Ra/230Th) is much lower in the other WG/EG basalts. It is difficult to tell a priori whether this is because these samples are from older lava flows or if lower (226Ra/230Th) reflects different melt generation conditions beneath these ridge segments. This latter possibility is considered below in conjunction with (230Th/238U) in the lavas and melt model predictions, but we stress that primary 226 Ra-excesses are not well-constrained in our data set, so we rely more heavily on 230Th-excesses to quantify melting conditions. 5. Discussion 5.1. U-series variations along the SEIR U-series signatures vary along the SEIR, both within and between the aforementioned WG, CG and EG geographic groups (Figs. 2, 3A, 4–6A).

Fig. 2. ‘Equiline’ diagram. (230Th/232Th)–(238U/232Th) diagram with SEIR basalts plotted by geographic groupings. The diagonal line in the center of the plot is the locus of secular equilibrium. Data that sit to the left of this line display excess 230Th. Global MORB data are shown for comparison from Elliott and Spiegelman (2003) with additions from Kokfelt et al. (2005), Standish (2006) and Tepley et al. (2004).

three CG and one WG samples not analyzed for 226Ra, plus the one sample (WW10-89-4) with (226Ra/230Th) = 1, lie within or above this envelope, indicating that post-eruptive decay has not significantly lowered their 230Th excesses compared to those samples with (226Ra/ 230 Th) N 1. Thus we believe that our measured 230Th–238U disequilibria record magmatic values at the time of eruption. One should also evaluate the potential of post-eruptive decay on (226Ra/230Th) to use it with (230Th/238U) to constrain magmatic processes and timing, as we attempt to do here. A few studies have used MORBs of known eruption age or geologic context from submersible collection to largely avoid post-eruptive 226Ra decay effects (e.g., Sims et al., 2002; Rubin et al., 2005). Most MORB studies, including this one, use dredged samples, for which age constraints and field relationships are limited, yet it is still possible that (226Ra/230Th) disequilibria retain aspects of disequilibria relationships at eruption, particularly when U-series patterns resemble those in age-constrained, submersible collected samples. There is an inverse relationship of (226Ra/230Th) and (230Th/238U) in age-constrained EPR basalts from the 9°–10° N (Lundstrom et al., 1999; Sims et al., 2002); our SEIR basalts exhibit the same inverse relationship, albeit at lower (226Ra/230Th) (Fig. 4). SEIR CG samples are correlated on this diagram and fall close to a line that represents 1 half life of decay (1600 years) from the EPR trajectory (although one datum is displaced an additional half life below this line). Such a wellcorrelated trajectory is unlikely to result from randomly dispersed initial ratios experiencing different amounts of decay. One can therefore speculate that this CG data array mimics the data dispersion as it was upon eruption, implying that the conditions that arose in the EPR trajectory likewise exist along this part of the SEIR. However, it is not possible with the present data set to determine whether these CG basalts initially followed the EPR array and have since decayed, or if zero-age SEIR lavas are offset to lower (226Ra/230Th) than the EPR. A similar explanation probably does not apply to WG and EG lavas, which form a separate, steeper array at generally lower 226Ra-excess for a given 230Th-excess (Fig. 4). Post-eruptive aging of a simple, EPRlike array could only produce this if sample aging increased smoothly from basalts having low (230Th/238U) to those having high (230Th/ 238 U). One WG sample (84–7) has 160% 226Ra-excess; if this sample has decayed by even one 226Ra half life, it would have erupted with (226Ra/230Th) = 4.2, making it one of the highest values observed in

Fig. 3. A) SEIR ‘Equiline’ diagram. SEIR basalts plotted by regional groupings; symbols are the same as in Fig. 2; black dots additionally indicate samples with measured 226Ra and thus age control. Regional groups are further divided into enrichment groups designated by dashed vertical lines. The solid black line is a least-squares regression through the data having measured 226Ra, following Lundstrom et al. (1998). Curved grey lines are the ±2 standard error envelope on the linear regression. B) ‘Equiline’ diagram and axial morphology. SEIR basalts grouped by axial morphology as described in Cochran et al. (1997) and Baran et al. (2005). Lines are least squares regressions through the data (with radium control) for each morphologic group; 4 samples that may have suffered 230Thdecay are excluded. Note the generally lower 230Th-excess of AH lavas and the progression to lower 230Th-excess for RH, SV and DV groups. AH morphologies are ‘EPR-like’, with 200– 400 m axial highs (segments C17-C14), RH segments have 100–200 m axial highs with ~2 km wide by 50–100 m deep rifts (segments C13-C9), SV segments have 10–15 km wide by 500–800 m deep rifts (segments C8-C4), and DV morphologies having well-defined (N 1000 m deep) axial valleys.

60

C.J. Russo et al. / Earth and Planetary Science Letters 278 (2009) 55–66

sample 84–7). Additionally, the six CG samples have the highest (226Ra/230Th) as a group, the WG samples have intermediate values, and the two EG samples have the lowest mean value (Table 1). Some of these differences may be due to post-eruptive 226Ra decay (section 4.3), but we also consider the implications of these regional Ra variations assuming some or all are magmatic. 5.2. Axial depth and

Fig. 4. (226Ra/230Th) versus (230Th/238U). (A) SEIR basalts separated into regional groups. Compare the inverse relationships in SEIR basalts to submersible collected, ageconstrained samples from the East Pacific Rise, plotted as black circles (Lundstrom et al., 1999; Sims et al., 2002). SEIR regional groupings fall along offset trends, with central group (CG) lavas lying along a trend at relatively higher values of (226Ra/230Th) than western group (WG) and eastern group (WG) lavas. Some of this difference is probably due to post-eruptive decay of 226Ra, although decay is not expected to move randomly distributed zero-age samples onto a well correlated line (such as the CG samples). The solid black line is a least squares regression through the EPR data. The dashed line is the slope of the EPR trend after one half-life of decay. (B) CG measured data, and data corrected for potential 226Ra decay. Much of the CG data follows this same relationship as the EPR if corrected for one half-life of decay (open symbols). Correction for two halflives would raise the lowest CG sample to the same trend as the others. (C) EG and WG measured data and data corrected for potential 226Ra decay. The WG and EG samples have generally lower 226Ra-excess, which may be due either to more extensive decay than in CG samples or to different melting conditions.

All basalts have 230Th-excesses, ranging from 21 ± 1.2% to 1.4 ± 1.0%. This 20% 230Th-excess range is smaller than in young (i.e., with excess 226Ra), intermediate spreading rate MORBs from the JDF-Gorda Ridges (230Thexcess of 4% to 40%; Goldstein et al., 1992; Volpe and Goldstein, 1993), but slightly greater than GSC MORBs (5% to 21%; Kokfelt et al., 2005). The along-axis SEIR (230Th/238U) variation forms a slightly convex pattern, with lower values associated with the western and eastern ends of the study area (Fig. 5). The CG has much larger 230Th-excesses (16% on average) compared to lower WG and EG values (8% and 11%, respectively; Table 1). SEIR (230Th/232Th) shows a different geographic pattern, generally decreasing from WG to CG to EG segments and primarily reflecting mantle source composition before melting. Aspects of (226Ra/230Th) variations in the sample set likely reflect initial conditions at the time of eruption (e.g., the CG (230Th/238U) vs. (226Ra/230Th) inverse correlation and the high (226Ra/230Th) in WG

230

Th/238U disequilibria

Bourdon et al. (1996a) interpreted a weak negative correlation between ridge segment averaged axial depth and (230Th/238U) in a substantially filtered early global MORB data set to result from mantle temperature control on depth/degree of melting. Their model involves melting of hotter mantle beneath shallower ridges producing larger (230Th/238U) via a deeper peridotite solidus, higher overall F, thicker oceanic crust, and more melt generated in the garnet stability field (where Th is more incompatible then U; Beattie, 1993; LaTourrette et al., 1993; Blundy and Wood, 2003). The (230Th/238U) vs. axial depth relationship is less evident when newer data from the deep southern MAR and the shallow Reykjanes Ridge are included in the global compilation (Lundstrom et al., 1998; Peate et al., 2001; Lundstrom, 2003). MORBs from intermediate axial depths (~ 3000–3500 m) are highly variable and span nearly the entire MORB range (Fig. 6). The global data set still displays a weak negative covariation but the statistical significance of the global trend has been questioned, and it is largely defined by ‘end-member’ samples from the shallow Azores platform (MAR) and the unusually deep AAD (Elliott and Spiegelman, 2003). The two deepest samples from the AAD have not been analyzed for 226Ra, and since their (230Th/238U) is substantially different from our nearest EG samples, there are legitimate questions concerning whether they have experienced significant post-eruptive 230Th decay. Axial depth and (230Th/238U) are not negatively correlated along the SEIR, although SEIR samples do largely fit within the published MORB swath (Fig. 6). Yet (230Th/238U) displays the entire SEIR range in the narrow axial depth range from 2300 to 2700 m. In addition, many WG basalts have lower than expected 230Th-excesses for their axial depths and the two EG basalts erupted at the greatest depth have the same (230Th/238U) as CG and WG basalts erupted much shallower. If there truly is a weak global trend, the lack of a (230Th/238U)-axial depth correlation along the SEIR suggests that melt initiation depth does not respond to regional mantle temperature differences in the way envisioned to explain the global data set. Other variables, including M, ϕ and mantle composition likely vary along the SEIR in a way that obscures any simple potential (230Th/238U)-axial depth relationship. 5.3. Melt supply, axial morphology and U-series disequilibria 5.3.1. General observations and modeling Melt supply (i.e., the volumetric flux of melt from mantle to crust) affects ridge axis morphology, AMC depth and lava chemistry. Thermal models predict that AMC depth and melt supply are inversely related (Phipps Morgan and Chen, 1993; Chen 1996); ridge-axis morphology reflects interplays between melt supply, mantle temperature, and hydrothermal cooling (e.g., Chen, 2000); and strong correlations between lava chemistry, AMC depth and spreading rate demonstrate how such thermal conditions control magma accumulation conditions and MORB compositions (Rubin and Sinton, 2007). Coincident changes in AMC depth and ridge morphology at some ridges also support these thermal models (e.g., at EPR segments north and south of the Orozco transform; Chen, 2000, Carbotte et al., 1998). Intermediate spreading ridges such as the SEIR are generally more sensitive to magma supply variations than ridges with faster or slower spreading rates (e.g., Niu and Batiza 1993; Carbotte et al., 2006; Detrick et al., 2002), and small magma supply variations can produce large variations in ridge axis morphology (Chen, 1996).

C.J. Russo et al. / Earth and Planetary Science Letters 278 (2009) 55–66

61

Fig. 5. Longitudinal variations within the SEIR study area. (A) Axial-depth profile and smoothed axial depth curve (grey line) from Sempéré et al. (1997). (B) κ is the measured 232Th/ 238 U ratio for SEIR basalt glasses. (C) κPb calculated from the Pb isotope data of Mahoney et al. (2002); colored data points are samples with Th-U isotope dilution data from this study. κPb is the 232Th/238U ratio assuming closed system terrestrial evolution over 4.56 Ga, with Canyon Diablo troilite 206Pb/204Pb and 208Pb/204Pb as initial values (Tatsumoto et al., 1973). (D) (230Th/238U) for SEIR basalt glasses. Grey curve is the (230Th/238U) for the SEIR basalts, predicted from the smoothed axial depth in panel (A) using the ‘global’ (230Th/238U)-axial depth relationship of Bourdon et al. (1996a). (E) Measured (230Th/232Th). (F) Measured (226Ra/230Th).

Shallower axial depth, axial high morphology and greater crustal thickness all point to greater melt supply in the western portion of our SEIR study area. Systematic variations in these parameters as well as in U-series disequilibria (at nearly constant, intermediate spreading rate) provide a unique opportunity to quantify and investigate the underlying melt supply variations.

U-series disequilibria in volcanic rocks are often modeled to examine melting conditions and melt supply effects beneath MORs and mid-plate hotspot volcanoes at passive and active mantle upwelling conditions, respectively (e.g., Speigelman and Elliott, 1993, Sims et al., 1995, McKenzie, 2000; Stracke et al., 2006), and at hotspot-influenced ridges (e.g., Bourdon et al., 1996b; Kokfelt et al.,

62

C.J. Russo et al. / Earth and Planetary Science Letters 278 (2009) 55–66

of melting parameters have the biggest effect on 230Th-excess in MORB, with excesses being reduced by: 1) greater porosity, ϕ; 2) greater melt fraction, F; 3) higher melting rate, M; and in some instances, 4) shallower melt initiation depth. A wide range of conditions result in excess 226Ra, but the magnitude of the excess is highly dependent on melting and melt transport rates because the 226 Ra half life is so short.

Fig. 6. (230Th/238U) versus axial depth. (A) SEIR basalts are divided into regional groups. Small grey symbols are the ‘global’ dataset from Bourdon et al. (1996a), as updated by Lundstrom (2003). (B) SEIR basalts grouped by ridge morphology (Symbols are the same as in Fig. 3B). Small grey circles as in panel A. Other data include Galápagos Spreading Center samples with measured 226Ra-excesses (open circles; Kokfelt et al., 2005).

2005). Two general types of models have been developed that constrain the rate, duration and extent of mantle melting (e.g., Elliott, 1997); both are one-dimensional, consider variations with depth beneath the ridge axis, and produce radioactive disequilibria primarily by variations in M, ϕ and F. The main differences between the models are in specific melting conditions and whether or not melt and solid maintain chemical equilibrium throughout the melting column. Dynamic melting (e.g., McKenzie, 1985; Williams and Gill, 1989) predicts compositions of melts that remain with the solid residue until a porosity threshold, followed by melt extraction everywhere from the melt column at a rate balanced by continued melt production, resulting in constant ϕ. This model behaves like accumulated fractional melting for non-radioactive elements (e.g., Williams and Gill, 1989). The equilibrium porous flow (EPF) model predicts the combined effects of chemical fractionation, radioactive decay, and melt transport time on U-series disequilibria (Spiegelman and Elliott, 1993). M is not an input parameter but can be calculated from EPF model inputs (i.e., depths of melt initiation and cessation, plus the amount of melt present at each depth). Melt simulations described in this paper were run at the same conditions of M, ϕ and F for both types of models (see the Supplementary data). In general these melt models do not uniquely and quantitatively predict melting conditions that lead to specific disequilibria values (e.g., for 230Th–238U) because (a) model outputs are highly dependent on poorly constrained input parameters, (b) multiple combinations of parameters can generate the same result, and (c) such parameterizations have greater degrees of freedom when they are used to match just one U-series element pair. Here we use the results of model simulations as qualitative indicators of conditions that may result in disequilibria variations between groups of SEIR samples without emphasizing the absolute values of a particular model simulation. Many previous authors have used these same models to constrain mantle melt column parameters, some more literally than others (e.g., Williams and Gill, 1989; Spiegelman and Elliott, 1993; Lundstrom et al., 1995; Bourdon et al., 1996a,b, 1998; Sims et al., 1995, 2002; Stracke et al., 1999, 2003). These previous studies agree that a handful

5.3.2. Melt supply along the western portion of the study area The SEIR samples were subdivided into four groups (Table 1; Section 2.1) that reflect their axial morphology, AMC depth and crustal thickness. These morphologically grouped samples cluster discretely on the (230Th/238U) vs. axial depth diagram (Fig. 6B). The RH, SV and DV groups form an array that resembles the global trend of Bourdon et al. (1996a). However, the AH group (i.e., with highest melt supply) plots well below that trend at lower 230Th-excess for the same axial depth as RH samples. Shallower AMC depth beneath segment C12 (Baran et al., 2005) is one indication that higher melt supply may be the underlying cause of this RH-to-AH morphological and 230Thexcess shift. Observations at the intermediate spreading rate Galápagos Spreading Center provide context for these observations. The GSC has a strong axial depth gradient from interaction of the ridge and the Galápagos hotspot, and exhibits AH morphology in a ‘central bulge’ region nearest the hotspot due to excess magma supply, RH morphology near 94°W, and SV morphology west of 95°W (e.g., Detrick et al., 2002; Christie et al., 2005). MORBs erupted in the ‘central bulge’ (Fig. 6B) have lower 230Th-excesses than nearby nonAH segments (Kokfelt et al., 2005). The nearest hotspots to our study area are 700 to 1600 km away and do not directly influence magma supply, so something else must be increasing magma supply to account for the AH morphology, shallower ridge depth and lower 230Th-excesses than RH segments. In principle, all could result from some combination of larger F, greater ϕ, and faster M. Increased melt productivity (or melting rate) will produce larger F at constant melt column height if the mantle is more fertile or at higher temperature. Nothing from the trace element or isotope geochemistry implies that SEIR AH basalts were produced by more fertile mantle, but a mantle temperature difference with RH segments cannot be ruled out. Producing extra melt by only deepening the melt initiation depth beneath AH segments would lead to a greater proportion of melt generated in the garnet stability field, thereby increasing rather than lowering melt 230Th-excesses, which is inconsistent with the SEIR observations. Likewise, it is possible to increase melt supply at constant melt column properties if a greater proportion of melt were sampled from off-axis portions of the triangular melting region beneath the ridge, but smaller degree melts produced there would have greater 230Th-excess. Producing a lower proportion of melt within the garnet stability field would generate lower melt 230Th-excesses (i.e., if melt initiation depth were shallower beneath AH versus RH segments). For example, 230 Th-excess is reduced by ~25% when initiation depth is decreased from 40 kb to 25 kb (Fig. S4). Garnet no longer influences the melt composition if initial melting depth is reduced further, dramatically reducing melt 230Th-excess (e.g., compare dynamic and EPF melting simulations of spinel lherzolite and garnet lherzolite in Figs. S1 and S2). Some spinel lherzolite model simulations reproduce the 230Thexcesses observed in AH and RH samples, but only at unrealistically low values of M, ϕ and F (b1.0 × 10− 4 kg m− 3 yr− 1, b0.1% and b0.08, respectively). These conditions are also inconsistent with the relatively low 226Ra excess in some WG samples (i.e., except WW1084-7). Furthermore the very small degrees of melting in these simulations are inconsistent with ≥6.3 km thick crust in the western region (see calculations in the supplement). Thus shallow melt initiation depths are unlikely to have caused the low 230Th-excesses in the WG samples. However, the unusual characteristics of sample

C.J. Russo et al. / Earth and Planetary Science Letters 278 (2009) 55–66

WW10-84-7 (extremely low 1.4% 230Th-excess and very high 160% 226 Ra-excess) make it the one exception, in that spinel peridotite melting simulations can produce these characteristics and garnet peridotite simulations cannot. This sample was taken from a small dying ridge segment between a westward propagating segment and the transform near 96° E, a likely place for a shallow, low productivity melt column. It is unlikely that melt initiation depth varies significantly beneath most AH and RH segments since changing it cannot produce the observed decrease in average 230Th-excess at AH segments (6%) relative to nearby RH segments (19%). Assuming constant initial melting depth, one can forward model our U-series observations to qualitatively determine how much melt supply must increase from RH to AH segments to change the axial morphology and reduce the 230Thexcess. An increase in both M and ϕ will reduce 230Th-excess, whereas only an increase in M will also increase the amount of melt generated. A seven fold increase in M (from 1 × 10− 4 to 7 × 10− 4 kg m− 3 yr− 1) at constant ϕ (0.25%) and bulk Kd, will produce this 230Th-excess shift by dynamic melting with essentially no effect on 226Ra-excess (Table S3). This change in M can be reduced slightly to a five-fold increase (from 1 to 5 × 10− 4 kg m− 3 yr− 1) if ϕ is also varied from ~ 0.25% (RH) to ~ 1.0% (AH). In the latter case, 226Ra-excess would decrease along with 230Thexcess (Fig. S1), consistent with lower 226Ra-excess observed in all but one AH sample. The (230Th/238U) difference could also be made by increasing ϕ at constant M, although M would be relatively low (1 × 10− 4 kg m− 3 yr− 1) and the ϕ shift would be extreme (i.e., 12-fold, from ~0.2%, RH to ~2.5%, AH). It is difficult to imagine a scenario that would allow such large ϕ increase at constant upwelling and melting rate, so we consider model simulations requiring some change in both M and ϕ to be more realistic. We speculate that higher ϕ might increase melt transport efficiency and thus melt supply, but such melt transport mechanics are beyond the scope of this work. 5.3.3. Melt supply along the eastern portion of the study area It is conceptually straightforward to envision how an increase in M could supply the extra melt needed to produce the U-series, AMC depth and morphology of AH segments on the SEIR. However, this rationale does not explain the lower 230Th-excesses (and very low 226 Ra-excesses) in EG segments with morphology and crustal thickness indicative of decreased melt supply. SEIR basalts from lowest melt supply DV segments have 230Th-excesses between those of high melt supply AH segments and moderate melt supply RH and SV segments (i.e., the average 230Th-excess shift between SV-RH (15%) and DV (11%) segments is smaller than the AH (6%) to RH (19%) shift). Low melt supply morphology and thinner crust both imply that melting conditions beneath DV segments must suppress melt 230Thexcess without increasing the net melt flux from the mantle. If increased ϕ and/or M were the cause of reduced 230Th-excess, DV segments would require a shorter melt column than SV segments. For example, reducing the initial melting depth from 40 kb to 25 kb shifts (230Th/238U) from 1.16 to 1.11 with the EPF model at constant M (1.3 × 10− 4) and ϕ (0.5%) (Fig. S4), because less melt is produced within the garnet stability field. However, a shallower melt initiation depth alone cannot explain the mean 230Th-excess shift from SV to DV samples because the lavas also require a change to more enriched mantle source composition (e.g., Th/U, radiogenic isotope ratios). The difficulty with generating the combined variations in 230Thexcess, 226Ra-excess, Th/U and melt supply for the EG basalts with only physical changes to the melt column leads us to consider compositional changes that can influence melt productivity, Th and U bulk distribution coefficients, and source enrichment. 5.4. Mantle source lithology and

230

Th–238U disequilibria

Mantle composition might affect the magnitude of 230Th–238U disequilibrium in MORBs (e.g., Goldstein et al., 1992; Rubin and

63

Macdougall, 1992; Lundstrom et al., 1998). E-MORBs (which often have elevated Th/U) can have larger 230Th-excesses than N-MORBs from the same region, producing equiline diagram data arrays with slopes b1 (Lundstrom et al., 1998). Lundstrom et al. (1998) postulated that increasing the modal proportion of garnet (to lower the relative Th bulk distribution coefficient) and clinopyroxene (to increase fertility) in the E-MORB source increases 230Th–238U fractionation during melting. Melts of such a source would have larger 230Thexcesses than melts of more depleted mantle, and their mixtures would produce the observed equiline diagram arrays. Variations in mantle upwelling rate can be superimposed on this mantle composition effect to flatten the array slope at slower compared to faster spreading ridges because slower upwelling leads to relatively greater melt 230Th-excesses by 230Th production in the melting column. The slope of the full SEIR array (0.90) is significantly steeper than slopes for other intermediate spreading ridges (JDF slope = 0.42 and Gorda Ridge slope = 0.32; Lundstrom et al., 1998). The SEIR slope is, in fact, steeper than the steepest slope (0.82) for the superfast spreading EPR at 20°–27°S, where the high upwelling rate apparently counteracts any source lithologic effect on melt (230Th/ 238 U) (Lundstrom et al., 1998). Given the nearly constant intermediate spreading rate and large Th/U range (plus other isotopic variations), our Fig. 3A data array slope indicates that different mantle compositions beneath the SEIR do not melt in the manner predicted by Lundstrom et al. (1998). Furthermore, generally higher Fe8.0 in SEIR WG lavas (which also have lower Th/U) are inconsistent with requirements of the Lundstrom et al. model (i.e., shallower melt initiation depth of a less productive mantle source to produce lower 230Th-excess), as are the inflated axial morphology and shallow ridge depth in the area. N-MORBs and E-MORBs (defined as having (La/Sm)N N 0.80; Mahoney et al., 2002) occur throughout the SEIR study area with no systematic offset in (230Th/238U) or (230Th/ 232 Th) (Figs. S5a and S5b). Thus the origin of the large Th/U (and Ba/Th and La/Sm) range along the SEIR (e.g., Fig. 3A) must be different from that which manifests itself as Th/U and (230Th/238U) variations in N-MORBs and E-MORBs elsewhere. Although sample numbers are small in this study, there are systematic variations in source composition, melt supply and ridge morphology that suggest a link between variations in mantle lithology, mantle temperature and melting conditions along the SEIR axis. Average Th/U of each of our morphological groups indicate different mean mantle compositions beneath ridge sections of some morphologies (Th/U is 2.5 ± 0.2, 3.2 ± 0.3, 3.3 ± 0.3, 3.6 ± 0.3 along AH, RH, SV, DV segments, respectively). There are also 230Th–238U disequilibrium differences between our various morphologic groups. The AH segment has the lowest Th/U and 230Th-excess (6%). The other three morphologic groups (RH, SV and DV) form vertically stacked arrays on an equiline diagram, with decreasing average 230Th-excesses (i.e., 19%, 15% and 11%, respectively; Fig. 3B). Because it is difficult to fractionate Th from U during MORB mantle melting (e.g., Elliott, 1997), decreasing 230Th-excesses at increasing Th/U in these latter three groups implies a change in mantle composition (e.g., lithology, which can affect Th/U) and how that composition melts (which affects 230Thexcess). The trace element and isotopic gradients along this part of the SEIR also imply a gradient in mean mantle composition, such as would be produced by variable proportions of depleted and enriched lithologies. Least squares regression slopes for the individual morphological data arrays on Fig. 3B are ≤1 for the moderate to high melt supply groups (SV, RH, AH) and N1 for the DV group. The more enriched samples in the SV, RH and AH groups have slightly larger 230Th-excess than the depleted samples, whereas for the DV group the more enriched sample has smaller 230Th-excess. The “stacked” RH, SV and DV data arrays rotate toward steeper slopes as the average (238U/ 232 Th) decreases and as the SEIR deepens to the east, so that increasing the proportion of melts from mantle with higher (i.e., less depleted)

64

C.J. Russo et al. / Earth and Planetary Science Letters 278 (2009) 55–66

Th/U leads to smaller melt 230Th-excess, rather than larger excess. This effect cannot be produced by changing melting parameters in a single lithology peridotite melting column. Clinopyroxene-rich lithologies (e.g., garnet pyroxenite or eclogite) potentially affect both mantle productivity and Th/U in partial melts (Hirschmann and Stolper, 1996; Bourdon et al., 1998; Sims et al., 2002; Elliott and Spiegelman, 2003; Stracke et al., 2003). These fertile lithologies melt more productively than typical MORB-source peridotite under the same conditions, producing melts with lower 230Thexcess (Stracke et al., 1999; Elkins et al., 2008). Both dynamic and EPF melting of garnet pyroxenite (45% cpx; 45% opx; 10% gnt; see supplementary information) produce lower 230Th-excesses and lower 226Ra-excesses (~ 10% and 25%, respectively) at M that are 0.25 to 4 times higher than peridotite and at ϕ = 2.5% (i.e., the center of the porosity range required for melt compaction and segregation without diffusive re-equilibration with surrounding peridotite matrix; Kogiso et al., 2004; Fig. S3 and Table S5). The high M and F for the pyroxenerich lithology in these models exert a dominant control on the magnitude of 230Th-excesses in the partial melts, in part because proportionally more melt is produced, and in part because Th concentration is higher in that (enriched) source lithology. Melting of mantle that is variably enriched in garnet pyroxenite (and/or eclogite) can therefore account for both high Th/U and lower, more similar 226Ra-excesses and 230Th-excesses in SEIR DV segment lavas. Importantly, melting scenarios that produce low 230Th-excesses but do not involve pyroxenite (Section 5.3) cannot produce all of the Useries, Th/U and radiogenic isotope characteristics observed along the SEIR. 5.5. Upper mantle lithology beneath the Indian Ocean Isotope gradients along the SEIR indicate a significant regional variation in mantle source composition (Graham et al., 2001; Mahoney et al., 2002). 232Th/238U (κ) and (230Th/232Th) also vary dramatically in these same samples, suggesting an upper mantle link to all of these signatures. The highest (230Th/232Th) and lowest κ values occur in the western portion of the study area, progressing smoothly eastward to lower (230Th/232Th) and higher κ (Fig. 5). κPb likewise varies smoothly eastward along axis to higher values (κPb is the time-integrated Th/U ratio, computed from 208Pb/204Pb and 206Pb/204Pb, as discussed in the Fig. 5 caption; e.g., Gast, 1969; Chase, 1981; Allegre et al., 1986; O'Nions and McKenzie, 1993; Elliott et al., 1999). κ and κPb were only computed for the subset of SEIR basalts having Pb isotopes (Mahoney et al., 2002) and isotope dilution Th and U concentrations (this study) because there is substantial non-systematic variation in the Mahoney et al. icp-ms Th–U data. The κ and κPb ranges among SEIR basalts encompass a significant portion of the global MORB range (κ = 2.43–3.91 and κPb = 3.79–4.02). κPb varies much less than κ in SEIR basalts (Figs. 5 and 7), such that κbbκPb in the west and κ ≈κPb ≈4.0 in the east. Globally, most MORBs display κ b κPb (Elliott et al., 1999; Kokfelt et al., 2005; Sims and Hart, 2006; Vlastelic et al., 2006), as expected after one or more relatively recent (b0.5 Ga) melt depletion events have fractionated Th from U in the mantle source. κ values in the easternmost SEIR study area are especially noteworthy because (a) they are close to the bulk silicate Earth value, implying a considerably more fertile mantle source than typical MORB source mantle, and (b) κ ≈κPb is rare in MORBs and implies long-time quasi-closed system evolution for this part of the mantle (109 years). The linear SEIR κ–κPb data array (Fig. 7) could have been produced by magma generation from mantle regions having variable prior melt depletion histories (via instantaneous or continuous scenarios; inset Fig. 7), or by a variation in mixing proportions of melts from low-κ upper mantle (i.e., like most global MORB) and a high-κ enriched lithology. A detailed evaluation of how the SEIR κ–κPb array formed is beyond the scope of this paper. Nevertheless, the array and systematic κ, κPb and (230Th/238U) variations record a longitudinal shift in mantle

Fig. 7. κ vs. κPb. SEIR basalts plotted in geographic groupings (same symbols as in Fig. 2). The κ = κPb line was calculated for a closed system (single stage) evolution using κ1 = 4.0 and μ1 = 9.0. The grey band is the MORB and ocean island basalt data trend of Vlastelic et al. (2006). Circled fields are representative MORB data sets from the Mid-Atlantic Ridge (Bourdon et al., 1996b) and East Pacific Rise (Sims et al., 2002) shown for comparison. Inset shows the SEIR data, along with isochron ages for two-stage mantle evolution models, using equations from Gast (1969), and assuming T = 4.55 Ga, μ1 = μ2 = 9.0, κ1 = 4.0 and κ2 = 2.35, where the subscripts 1 and 2 refer to the evolution stages; the age of second-stage fractionation (t) is given by numbers at the ends of the grey lines, and ranges from 4.0 Ga to 0 (today).

composition, past evolutionary history and current melting conditions beneath the SEIR. This is especially evident in the region east of 96°E, where (230Th/238U) decreases as κ and κPb converge. In the preceding sections we discussed the possibility that some melting of garnet pyroxenite is required to explain the 226Ra–230Th– 238 U–232Th variations along the eastern SEIR. An enriched garnet pyroxenite lithology that has not been melted in the recent past will have distinct isotopic characteristics and high Th, U and Pb concentrations, so admixture of a small amount of this material dispersed in the upper mantle could also produce the high κ ≈ κPb in those same magmas. Garnet pyroxenite melting simulations at high M (e.g., productivity of 0.5–1%/km) and high F (~25%) produce melts with ~10% 230Th-excess and subdued 226Ra-excess (b50%). If eastern SEIR basalts are mixtures of such pyroxenite melts with melts of normal depleted MORB mantle (F = 7–10%, as discussed previously and in Fig. S6a), then no more than ~ 10% of the total melt produced beneath the eastern SEIR is from garnet pyroxenite (assuming Th/U = 4 in the pyroxenite source; see the supplement). The same melt proportions explain the SEIR κ–κPb variations (Fig. S6b), and also indicate that a small amount of pyroxenite might be present in the mantle beneath some of our CG sample sites. Mantle proportions calculated from these melt mixtures and the respective pyroxenite and peridotite melt productivities suggest a maximum of 3–4% pyroxenite by volume as veins (or blobs) within 96–97% peridotite beneath the easternmost SEIR. This range is similar to volume estimates of mafic veins observed in ophiolites and alpine massifs (e.g., Mukasa et al., 1991). Sobolev et al. (2007) used trace element contents of olivine phenocrysts to suggest that MORBs globally are formed from mantle with 10–15% pyroxenite compared to generally higher amounts in Iceland (20%), Hawaii (60%) and the Siberian traps (100%). Such high MORB source proportions are difficult to reconcile with our much lower (~ 5–25 fold) estimates for the SEIR. Regional MORB differences are not discussed by Sobolev et al. (2007). Using their Mn/Fe methodology and SEIR samples in their data supplement, we determined that mean mantle source pyroxenite fractions are essentially constant at 10–12% in sub-regions of our study area

C.J. Russo et al. / Earth and Planetary Science Letters 278 (2009) 55–66

where U-series variations (this study) suggest a pyroxenite gradient. It is unclear why these two studies differ in inferred pyroxenite fraction and distribution beneath the SEIR. Sm–Nd and Lu–Hf systematics have been used to determine the proportion of MORB melting that occurs in the garnet stability field (e.g., Salters, 1996). Stracke et al. (1999) argued that these systematics coupled to (230 Th/ 238 U) could distinguish garnet pyroxenite and peridotite in Hawaiian mantle derived melts. However, Sm/Nd and Lu/Hf of accumulated fractional melts under our estimated SEIR melting conditions and pyroxenite proportions do not distinguish pyroxenite-free and pyroxenite-containing lithologies, indicating that these ratios are not as sensitive to the presence of pyroxenite as (230Th/238U). The method employed by Salters and Hart (1989) and Stracke et al. (1999) uses δ(Sm/Nd) and δ(Lu/Hf), parameters that reflect relative offsets in the elemental ratios from those predicted by 143Nd/144Nd and 176Hf/177Hf and assuming the same reservoir age for pyroxenite and peridotite sources. SEIR κ and κPb values indicate different source ages, precluding the use of this exact method on our samples. The U-series disequilibria and Th–U–Pb isotopic signatures of our EG and some CG basalts are best explained by melting of mantle containing a few percent of enriched garnet pyroxenite mixed into an upper mantle composed primarily of peridotite. We speculate that the higher pyroxenite fraction in the eastern Indian Ocean, near the boundary with Pacific type mantle in the AAD may reflect the dynamics of mantle flow in the Indian Ocean; pyroxenite may have been preferentially depleted from the central part of the ocean basin and less so on the margins, perhaps when the SEIR passed over the Kerguelen hotspot at 38 Ma (Duncan and Storey, 1992). 6. Summary Variations in axial depth, crustal thickness, U-series disequilibria, κ and κPb along the SEIR are best explained by models in which the physical characteristics of melting (such as M, F, ϕ and melt initiation depth) vary along axis in response to an inferred long wavelength temperature gradient and small variations in the underlying source lithology (pyroxenite veins). This leads to decreased melt supply from west to east, a concomitant change in axial morphology and axial magma chamber depth, and systematic long wavelength geochemical gradients. Melting rate covaries with melt supply in the western and central regions, decreasing eastward along the SEIR. However, melt supply and melting rate vary inversely from the central to the eastern regions, because cooler eastern mantle contains a larger proportion of more fusible pyroxenite veins. These veins melt productively to generate melts with low 230Th-excess and 226Ra-excess and high Th, U and Ra concentrations, but still contribute only a small proportion of the total melt volume supplied to the ridge (which is dominated by mantle that is N95% peridotite). Th, Pb and He isotopic gradients are also consistent with a greater proportion of enriched pyroxenite with a reservoir age of roughly 0.5 to 1 Gyr eastward along the SEIR axis. Acknowledgments This paper benefited greatly from discussions and long-term collaborations with D. Christie, J. Mahoney and D. Pyle, as well as two anonymous reviews and editorial comments from Rick Carlson. This study was supported by the National Science Foundation Marine Geology & Geophysics Program (grants 02-21069 and 02-21240 to KR and DG). Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.epsl.2008.11.016.

65

References Allegre, C.J., Dupre, B., Lewin, E., 1986. Thorium/uranium ratio of the Earth. Chem. Geol. 56, 219–227. Baran, J.M., Cochran, J.R., Carbotte, S.M., Nedimovic, M.R., 2005. Variations in upper crustal structure due to variable mantle temperature along the Southeast Indian Ridge. Geochem. Geophys. Geosys. 6 (11), Q11002. doi:10.1029/2005GC000943. Beattie, P., 1993. Uranium-thorium disequilibria and partitioning on melting of garnet peridotite. Nature 363, 63–65. Blundy, J., Wood, B., 2003. Mineral-melt partitioning of uranium, thorium and their daughters. Rev. Mineral. Geochem. 52, 59–123. Bourdon, B., Zindler, A., Elliott, T., Langmuir, C.H., 1996a. Constraints on mantle melting at mid-ocean ridges from global 238U–230Th disequilibrium data. Nature 384, 231–235. Bourdon, B., Langmuir, C.H., Zindler, A., 1996b. Ridge-hotspot interaction along the MidAtlantic Ridge between 37°30' and 40°30'N: the U–Th disequilibrium evidence. Earth Planet. Sci. Lett. 142, 175–189. Bourdon, B., Joron, J.-L., Claude-Ivanaj, C., Allegre, C.J., 1998. U–Th–Pa–Ra systematics for the Grande Comore volcanics: melting processes in an upwelling plume. Earth Planet. Sci. Lett. 164, 119–133. Canales, J.P., Danobeitia, J.J., Detrick, R.S., Hooft, E.E.E., Bartolome, R., Naar, D.F., 1997. Variations in axial morphology along the Galápagos spreading center and the influence of the Galápagos Hotspot. J. Geophys. Res. 102, 27,341–27,354. Carbotte, S.M., Mutter, J.C., Xu, L., 1998. Insights from a fast-intermediate spreading ridge on the influence of magma supply and spreading rate on crustal magma bodies and emplacement of the extrusive layer. Geology 26, 455–458. Carbotte, S.M., Detrick, R.S., Harding, A., Canales, J.P., Babcock, J., Kent, G., Van Ark, E., Nedimovic, M.R., Diebold, J., 2006. Rift topography linked to magmatism at the intermediate spreading Juan de Fuca Ridge. Geology 34, 209–212. Chase, C.G., 1981. Oceanic island Pb: two-stage histories and mantle evolution. Earth Planet. Sci. Lett. 52, 277–284. Chen, Y.J., 1996. Dynamics of the mid-ocean ridge plate boundary: recent observations and theory. PAGEOPH 146 (3/4), 621–648. Chen, Y.J., 2000. Dependence of crustal accretion and ridge-axis topography on spreading rate, mantle temperature, and hydrothermal cooling. In: Dilek, U., Moores, E.M., Elthon, D., Nicloas, A. (Eds.), Ophiolites and Oceanic Crust: New Insights from Field Studies and the Ocean Drilling Program: Boulder, Colorado. Geological Society of America Special Paper, vol. 349, pp. 161–179. Christie, D.M., Werner, R., Hauff, F., Hoernle, K., Hanan, B.B., 2005. Morphological and geochemical variations along the eastern Galápagos Spreading Center. Geochem. Geophys. Geosys. 6, Q01006. doi:10.1029/2004G000714. Cochran, J.R., Sempere, J.-C., SEIR Scientific Team, 1997. The Southeast Indian Ridge between 88°E and 118°E: gravity anomalies and crustal accretion at intermediate spreading rates. J. Geophys. Res. 102, 15,463–15,487. Detrick, R.S., Harding, A.J., Kent, G.M., Orcutt, J.A., Mutter, J.C., Buhl, P., 1990. No evidence from mulitichannel reflection data for a crustal magma chamber in the MARK area on the Mid-Atlantic Ridge. Nature 347, 61–64. Detrick, R.S., Sinton, J.M., Ito, G., Canales, J.P., Behn, M., Blacic, T., Cushman, B., Dixon, J.E., Graham, D.W., Mahoney, J.J., 2002. Correlated geophysical, geochemical, and volcanological manifestations of plume-ridge interaction along the Galápagos Spreading Center. Geochem. Geophys. Geosys. 3 10.1029200GC000350. Douglas-Priebe, L.M.; 1998 Geochemical and petrogenetic effects of the interaction of the Southeast Indian Ridge and the Amsterdam-St.Paul hotspot, M.S. thesis, Oregon State University, Corvallis. Duncan, R.A., Storey, M., 1992. The life cycle of Indian Ocean hotspots. In: Duncan, R.A., Rea, D.K., Kidd, R.B., von Rad, U., Weissel, J.K. (Eds.), Synthesis of Results from the Scientific Drilling in the Indian Ocean. Geophysical Monograph, vol. 70. American Geophysical Union, pp. 91–103. Elkins, L.J., Gaetani, G.A., Sims, K.W.W., 2008. Partitioning of U and Th during garnet pyroxenite partial melting: constraints on the source of alkaline ocean island basalts. Earth Planet. Sci. Lett. doi:10.1016/j.epsl.2007.10.034. Elliott, T., 1997. Fractionation of U and Th during mantle melting: a reprise. Chem. Geol. 139, 165–183. Elliott, T., Spiegelman, M., 2003. Melt migration in oceanic crustal production: a Useries perspective. Treatise Geochem. 3, 465–510. Elliott, T., Zindler, A., Bourdon, B.,1999. Exploring the kappa conundrum: the role of recycling in the lead isotope evolution of the mantle. Earth Planet. Sci. Lett. 169, 129–145. Gast, P.W., 1969. The isotopic composition of Lead from St. Helena and Ascension Islands. Earth Planet. Sci. Lett. 5, 353–359. Goldstein, S.J., Murrell, M.T., Janecky, D.R., Delaney, J.R., Calgue, D.A., 1992. Geochronology and petrogenesis of MORB from the Juan de Fuca and Gorda ridges by 238U–230Th disequilibrium. Earth Planet. Sci. Lett. 109, 255–272. Graham, D.W., Lupton, J.E., Spera, F.J., Christie, D.M., 2001. Upper-mantle dynamics revealed by helium isotope variations along the southeast Indian ridge. Nature 409, 701–703. Graham, D.W., Blichert-Toft, J., Russo, C.J., Rubin, K.H., Albarede, F., 2006. Cryptic striations in the upper mantle revealed by hafnium isotopes in southeast Indian ridge basalts. Nature 440, 199–202. Hart, S.R., 1984. A large-scale isotope anomaly in the Southern Hemisphere mantle. Nature 309, 753–757. Hirschmann, M.M., Stolper, E.M., 1996. A possible role for garnet pyroxenite in the origin of the ‘garnet signature’ in MORB. Contrib. Mineral. Petrol. 124, 185–208. Holmes, R.C., Tolstoy, M., Cochran, J.R., Floyd, J.S., in press. Crustal thickness variations along the Southeast Indian Ridge (100°–116°E) from 2-D body wave tomography. Geochem. Geophys. Geosyst. doi:10.1029/2008GC002152. Klein, E.M., Langmuir, C.H., 1987. Global correlations of ocean ridge basalt chemistry with axial depth and crustal thickness. J. Geophys. Res. 92, 8089–8115.

66

C.J. Russo et al. / Earth and Planetary Science Letters 278 (2009) 55–66

Klein, E.M., Langmuir, C.H., Staudigel, H., 1991. Geochemistry of basalts from the Southeast Indian Ridge, 115° E–138° E. J. Geophys. Res. 96, 2089–2107. Kogiso, T., Hirschmann, M.M., Pertermann, M., 2004. High-pressure partial melting of mafic lithologies in the mantle. J. Petrol. 45, 2407–2422. Kokfelt, T.F., Lundstrom, C., Hoernle, K., Hauff, F., Werner, R., 2005. Plume-ridge interaction studied at the Galápagos spreading center: evidence from 226Ra–230Th– 238 U and 231 Pa–235U isotopic disequilibria. Earth Planet. Sci. Lett. 234, 165–187. Langmuir, C.H., Klein, E.M., Plank, T., 1992. Petrology systematics of mid-ocean ridge basalts: constraints on melt generation beneath ocean ridges. In: Phipps Morgan, J., Blackman, D.K., Sinton, J.M. (Eds.), Mantle Flow and Melt Generation at Mid-Ocean Ridges. Geophysical Monograph, vol. 71. American Geophysical Union, pp. 183–280. LaTourrette, T.Z., Kennedy, A.K., Wasserburg, G.J., 1993. Thorium–uranium fractionation by garnet: evidence for a deep source and rapid rise of oceanic basalts. Science 261, 739–742. Lundstrom, C.C., 2003. Uranium-series disequilibria in mid-ocean ridge basalts: observations and models of basalt genesis. Rev. Mineral. Geochem. 52, 175–214. Lundstrom, C.C., Gill, J., Williams, Q., Perfit, M.R., 1995. Mantle melting and basalt extraction by equilibrium porous flow. Science 270 (5244), 1958–1961. Lundstrom, C.C., Williams, Q., Gill, J.B., 1998. Investigating solid mantle upwelling rates beneath mid-ocean ridges using U-series disequilibria, 1: a global approach. Earth Planet. Sci. Lett. 157, 151–165. Lundstrom, C.C., Sampson, D.E., Perfit, M.R., Gill, J., Williams, Q., 1999. Insights into midocean ridge basalt petrogenesis: U-series disequilibria from the Siqueiros Transform, Lamont Seamounts, and East Pacific Rise. J. Geophys. Res. 104, 13,035–13,048. Mahoney, J.J., Graham, D.W., Christie, D.M., Johnson, K.T.M., Hall, L.S., Vonderhaar, D.L., 2002. Between a hotspot and a cold spot: isotopic variation in the Southeast Indian Ridge asthenosphere, 86°E–118°E. J. Petrol. 43, 1155–1176. McKenzie, D., 1985. 230Th–238U disequilibrium and the melting processes beneath ridge axes. Earth Planet. Sci. Lett. 72, 149–157. McKenzie, D., 2000. Constraints on melt generation and transport from U-series activity ratios. Chem. Geol. 162, 81–94. Mukasa, S.B., Shervais, J.W., Wilshire, H.G., Nielson, J.E., 1991. Intrinsic Nd, Pb, and Sr isotopic heterogeneities exhibited by the Lherz Alpine Peridotite Massif, French pyrenees. J. Petrol. 117–134 Special Lherzolite Issue. Nicolaysen, K.P., Frey, F.A., Mahoney, J.J., Johnson, K.T.M., Graham, D.W., 2007. Influence of the Amsterdam/St. Paul hot spot along the Southeast Indian Ridge between 77° and 88°E: correlations of Sr, Nd, Pb, and He isotopic variations with ridge segmentation. Geochem. Geophys. Geosys. 8, Q09007. doi:10.1029/2006GC001540. Niu, Y., Batiza, R., 1993. Chemical variation trends at fast and slow spreading mid-ocean ridges. J. Geophys. Res. 98, 7887–7902. Niu, Y., Hekinian, R., 1997. Spreading-rate dependence of the extent of mantle melting beneath ocean ridges. Nature 385, 326–329. Niu, Y., O, M.J., ' Hara, 2008. Global correlations of ocean ridge basalt chemistry with axial depth: a new perspective. J. Petrol. 1–32. doi:10.1093/petrology/egm051. O'Nions, R.K., McKenzie, D., 1993. Estimates of mantle thorium/uranium ratios from Th, U and Pb isotope abundances in basaltic melts. Melting and Melt Movement in the Earth. Philos. Trans., 342, no. 1663, pp. 65–77. Peate, D.W., Hawkesworth, C.J., van Calsteren, P.W., Taylor, R.N., Murton, B.J., 2001. 238U–230Th constraints on mantle upwelling and plume-ridge interaction along the Reykjanes Ridge. Earth Planet. Sci. Lett. 187, 259–272. Pertermann, M., Hirschmann, M.M., 2003. Partial melting experiments on a MORB-like pyroxenite between 2 and 3 GPa: constraints on the presence of pyroxenite in basalt source regions from solidus location and melting rate. J. Geophys. Res. 108 (B2). doi:10.1029/2000JB000118. Phipps Morgan, J., Chen, Y.J., 1993. Dependence of ridge-axis morphology on magma supply and spreading rate. Nature 364, 706–708. Pyle, D.G.; 1994, Geochemistry of mid-ocean-ridge basalt within and surrounding the Australian-Antarctic Discordance; Ph.D. dissertation, Oregon State University, Corvallis. Rubin, K.H., Macdougall, J.D., 1988. 226Ra excesses in mid-ocean-ridge basalts and mantle melting. Nature 335, 158–161. Rubin, K.H., Macdougall, J.D., 1992. Th–Sr isotopic relationships in MORB. Earth Planet. Sci. Lett. 114, 149–157. Rubin, K.H., Sinton, J.M., 2007. Inferences on mid-ocean ridge thermal and magmatic structure from MORB compositions. Earth Planet. Sci. Lett. 260, 257–276.

Rubin, K.H., van der Zander, I., Smith, M.C., Bergmanis, E.C., 2005. Minimum speed limit for ocean ridge magmatism from 210Pb–226Ra–230Th disequilibria. Nature 437 (22), 534–538. Salters, V.J.M., 1996. The generation of mid-ocean ridge basalts from the Hf and Nd isotope perspective. Earth. Planet. Sci. Lett. 141, 109–123. Salters, V.J.M., Hart, S.R., 1989. The Hf-paradox, and the role of garnet in the MORB source. Nature 342, 420–422. Sempéré, J.-C., Cochran, J.R., SEIR Scientific Team, 1997. The Southeast Indian Ridge between 88°E and 118°E: variations in crustal accretion at constant spreading rate. J. Geophys. Res. 102, 15,489–15,506. Sempéré, J.-C., Palmer, J., Christie, D.M., Phipps-Morgan, J., Shor, A.N., 1991. The Australian-Antarctic discordance. Geology 19, 429–432. Shen, Y., Forsyth, D.W., 1995. Geochemical constraints on initial and final depths of melting beneath mid-ocean ridges. J. Geophys. Res. 100, 2211–2237. Sims, K.W.W., Hart, S.R., 2006. Comparison of Th, Sr, Nd and Pb isotopes in oceanic basalts: implications for mantle heterogeneity and magma genesis. Earth Planet. Sci. Lett. 245, 743–761. Sims, K.W.W., DePaolo, D.J., Murrell, M.T., Baldridge, W.S., Goldstein, S.J., Clague, D.A., 1995. Mechanisms of magma generation beneath Hawaii and Mid-Ocean Ridges: uranium/thorium and samarium/neodymium isotopic evidence. Science 267 (5197), 508–512. Sims, K.W.W., Goldstein, S.J., Blichert-Toft, J., Perfit, M.R., Kelemen, P., Fornari, D.J., Michael, P., Murrell, M.T., Hart, S.R., DePaolo, D.J., Layne, G., Ball, L., Jull, M., Bender, J., 2002. Chemical and isotopic constraints on the generation and transport of magma beneath the East Pacific Rise. Geochim. Cosmochim. Acta 66 (19), 3481–3504. Sobolev, A.V., 2007. The amount of recycled crust in sources of mantle-derived melts. Science 316, 412–417. Sours-Page, R.E.; 2000, Magmatic processes at mid-ocean ridges; evidence from lavas and melt inclusions from the Southeast Indian Ridge, the Endeavour Segment of the Juan de Fuca Ridge, and the northern East Pacific Rise, Ph.D. thesis, Oregon State University. Spiegelman, M., 2000. UserCalc: a web-based U-series calculator for mantle melting problems. Geochem. Geophys. Geosys. 1 (0030). Spiegelman, M., Elliott, T., 1993. Consequences of melt transport for uranium series disequilibrium in young lavas. Earth Planet. Sci. Lett. 118, 1–20. Standish, J.J., 2006, The influence of Ridge Geometry at the Ultraslow-Spreading Southwest Indian Ridge (9°–25°E): Basalt Composition Sensitivitiy to Variation in Source and Process, Ph.D. thesis, MIT/WHOI, 289 pages. Stracke, A., Salters, V.J.M., Sims, K.W.W., 1999. Assessing the presence of garnet– pyroxenite in the mantle sources of basalts through combined hafnium– neodymium–thorium isotope systematics. Geochem. Geophys. Geosys. 1. Stracke, A., Zindler, A., Salters, V.J.M., McKenzie, D., Gronvold, K., 2003. The dynamics of melting beneath Theistareykir, northern Iceland. Geochem. Geophys. Geosys. 4 (10), 8513. doi:10.1029/2002GC000347. Stracke, A., Bourdon, B., McKenzie, D., 2006. Melt extraction in Earth's mantle: constraints from U–Th–Pa–Ra studies in oceanic basalts. Earth Planet. Sci. Lett. 244, 97–112. Tatsumoto, M., Knight, R.J., Allegre, C.J., 1973. Time differences in the formation of meteorites as determined from the ration of lead-207 to lead-206. Science 180, 1279–1283. Tepley, F.J., Lundstrom, C.C., Sims, K.W.W., Hekinian, R., 2004. U-series disequilibria in MORB from the Garrett transform and implications for mantle melting. Earth Planet. Sci. Lett. 223, 79–97. Thurber, D., 1967. Anamolous 234U/238U in nature. J. Geophys. Res. 67, 4518. Vlastelic, I., Lewin, E., Staudacher, T., 2006. Th/U and other geochemical evidence for the Reunion plume sampling a less differentiated mantle domain. Earth Planet. Sci. Lett. 248, 379–393. Volpe, A.M., Goldstein, S.J., 1993. 226Ra–230Th disequilibrium in axial and off-axis midocean ridge basalts. Geochim. Cosmochim. Acta 57, 1233–1242. Williams, R.W., Gill, J.B., 1989. Effects of partial melting on the uranium decay series. Geochim. Cosmochim. Acta 53, 1607–1619.