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

Contents lists available at ScienceDirect

Earth and Planetary Science Letters

j ourna l homepage: www.e lsev ie r.com/ locate /eps l

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

Chris J. Russo a,b,⁎, Ken H. Rubin b, David W. Graham a

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

pyroxenite

⁎ Corresponding author. School of Ocean EngineerUniversity of Hawaii, Honolulu, HI 96822 USA. Tel.: +1 8

E-mail address: crusso@soest.hawaii.edu (C.J. Russo)

0012-821X/$ – see front matter © 2008 Elsevier B.V. Aldoi:10.1016/j.epsl.2008.11.016

a b s t r a c t

a r t i c l e i n f o

Article history:

Intermediate-spreading Sou Received 19 May 2008Received in revised form 14 November 2008Accepted 14 November 2008Available online 30 December 2008

Editor: R.W. Carlson

Keywords:U-series disequilibriamid-ocean ridge basaltSoutheast Indian Ridgemantle meltingmantle lithology

theast Indian Ridge (SEIR) basalts display geographic gradients from 90°E to118°E in Th/U, (230Th/232Th) and 238U–230Th–226Ra disequilibria (230Th-excesses of 1 to 21% and 226Ra-excesses of 0 to 160%). Highly correlated (238U/232Th) and (230Th/232Th) (r2=0.94) span nearly the entireglobal MORB range; basalts from three of four ridge morphologies form subparallel, vertically stacked arrayson an equiline diagram. (226Ra/230Th) is inversely correlated with (230Th/238U) within sub-regions of thestudy area. There is a convex (230Th/238U) along-axis pattern (lower 230Th-excesses in the west and east, atshallowest and deepest depths), such that (230Th/238U) does not correlate with axial depth as describedelsewhere, despite the ~2300 m west-to-east axial depth increase. κ and κPb also vary along axis from κbκPbin the west to κ≅κPb in the east.Variations in axial depth, crustal thickness, U-series disequilibria, κ and κPb are best explained by forwardmodel 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 temperaturegradient and small variations in the underlying source lithology (pyroxenite veins). This leads to decreasedmelt 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 thewestern and central regions, decreasing eastward along the SEIR, but supply and rate vary inversely from thecentral to the eastern regions, because cooler eastern mantle contains a few percent of fusible pyroxeniteveins that melt productively to generate melts with high Th–U–Ra concentration but low 230Th-excess and226Ra-excess. The veins contribute only a small proportion of the total melt volume from mantle that is N95%peridotite. Th, Pb and He isotopic gradients are consistent with a changing proportion of enriched pyroxeniteveins 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 indirectprobes of upper mantle composition and melting conditions.Geographic variations in the conditions and extent of melting, andtheir linkages to mantle temperature, lithology and chemicalcomposition are central to an understanding of mantle evolutionand the origin ofmid-ocean ridge basalts (MORBs). Important patternsin melt generation are revealed by the relations between MORBchemistry 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 andSinton, 2007; Niu and O'Hara, 2008). Global studies of MORBcomposition must often make simplifying assumptions to relatecompositional variations to changes in a particular physical char-

ing Science and Technology,08 956 9328..

l rights reserved.

acteristic, such as axial depth or spreading rate (Klein and Langmuir,1987; Bourdon et al., 1996a; Lundstrom et al., 1998). This approachmakes it difficult to decipher the competing effects of severalindependent 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 keyvariables (e.g., spreading rate, axial depth), or where physical,morphologic and/or chemical gradients in ridge character resultfrom a coherent geologic history. Here we present results of a regionalscale U-series isotopic and geochemical study of melting conditions atthe intermediate spreading rate Southeast Indian Ridge (SEIR)between 90°E and 118°E, an ~2100 km long region displayingsignificant along-axis physical and chemical variations that can beused 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; Bourdonet 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 thusmelting rate; Beattie, 1993; LaTourette et al., 1993; Pertermann andHirschmann, 2003). There is a temporal aspect to observed U-seriesdisequilibria, because parent-daughter activity ratios relax to a state ofsecular equilibrium at a rate governed by their half-lives following U–Th–Ra fractionation. We use these attributes in conjunction withchemical (Sours-Page, 2000) and radiogenic isotope compositions(Graham et al., 2001, 2006; Mahoney et al., 2002) of the same SEIRsamples to test scenarios for physical variations in melting conditionsof a heterogeneous mantle source.

2. Regional geologic background

The modern SEIR (Fig. 1) marks the Australian-Antarctic plateboundary between the Rodrigues triple junction (25°S, 70°E) and theMacquarie triple junction (62°S, 151°E). This section of MOR containsno large transform offsets and exhibits a west-to-east gradient in axialdepth, from ~2300 m to N4500 m (Cochran et al., 1997). The SEIR fullspreading rate is nearly constant, from ~72 to 76 mm/yr across ourstudy area (Cochran et al., 1997). The SEIR has a bathymetric high atthe Amsterdam-St. Paul Plateau (ASP, 76°–78°E; Graham et al., 2001)and unusually deep, magma-starved ridges within the Australian-

Fig. 1. Satellite altimetry of the Indian Ocean basin between 40°E and 130°E and sample locaRidge (CIR), Rodrigues Triple Junction (RTJ), Amsterdam-St. Paul Plateau (ASP), Kerguelen PlatEnlarged panel defines ridge-axis trace of the SEIR from 40°S to 52°S and from 88°E to 118°Elocations for samples in this study are shown by black dots along the ridge-axis and are laindicates depth variationwith red shades representing the shallowest segments and blue theand eastern, EG, groups). The transform fault separating segments C13 and C12 forms a nadifferent Sr, Nd, Pb isotope compositions, suggesting different mantle geochemical historiemorphology between segments C3 and C4 forms the EG and CG divide.

Antarctic Discordance (AAD; e.g., Klein et al., 1991; Sempéré et al.,1991). The west to east axial depth increase has been attributed to along wavelength decrease in mantle temperature (Cochran et al.,1997; Sempéré et al., 1997), estimated to be ~80–150 °C using majorelement 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 indicatechanges in melt supply along this portion of the SEIR. Three distinctaxial morphological ‘zones’ were discovered by a detailed geophy-sical survey in 1994/1995 (Cochran et al., 1997; Sempéré et al.,1997). Most western zone ridge segments (C17–C13; Fig. 1) haveaxial-high morphology, similar in dimension and shape to manysegments of the fast-spreading East Pacific Rise (EPR). SegmentsC12–C4 have a ‘transitional’ morphology typical of other inter-mediate-spreading ridges (e.g., Canales et al., 1997). Cochran et al.(1997) subdivided the transitional morphologies into: (a) riftedaxial highs in the west (segments C12–C9), and (b) shallow axialvalleys in the east (segments C8–C4). SEIR segments C3 and C2 havea deep axial valley, similar to much of the mid-Atlantic ridge (MAR).We use the following abbreviations throughout this paper to refer

tion map. Labeled features include: the Southwest Indian Ridge (SWIR), Central Indianeau (KP), Southeast Indian Ridge (SEIR), and the Australian-Antarctic Discordance (AAD).. Individual ridge segments are identified above the axial trace (C17 through C2), dredgebeled below the axial trace (D71 through D145). Ridge-axis color in the online versiondeepest. We consider three geographical regions in this study (western,WG, central, CG,tural WG-CG divide. MORBs west of this transform tend to have higher 3He/4He, ands east and west of 100°E (Mahoney et al., 2002). The transition into deep-axial valley

57C.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 structurealong segments C12–C7 (Baran et al., 2005) and a crustal thicknessgradient from ~6.3 km at 101° E to ~3.8 km at 115° E (Holmes et al., inpress). The westernmost RH segment (C12) has the shallowest axialmagma chamber (AMC), thinnest layer 2A (Baran et al., 2005) andgreatest crustal thickness (~6.3 km between 100° and 116°E; Holmeset al., in press). Baran et al. (2005) divided segment C12 into a westernAH and eastern RH segment, which we adopt for the dredge 96 and100 morphological assignments (Table 1). There is currently noseismic data for the other AH segments (C17–C13). The AMC is ~1 kmdeeper and the crust is slightly thinner (~5.9 km) on RH segments. TheSV segments do not have detectable AMCs. The DV segments have thethinnest crust (~4.6 km). They are outside the Baran et al. (2005) studyarea but likely lack an AMC (like most of the slow-spreading MAR;Detrick et al., 1990).

2.2. Along-axis geochemical variations

All SEIR lavas discussed here are ‘Indian’-type MORBs (Mahoneyet al., 2002), with lower εNd, higher 87Sr/86Sr and elevated Δ208Pb/204Pb (after Hart, 1984) compared to Pacific and North AtlanticMORBs. Eastward decreases in Fe8 and increases in Na8 suggest agradient in the mean pressure and extent of melting (Sours-Page,2000; Graham et al., 2001; Mahoney et al., 2002; Fe8 and Na8 arefractionation corrected oxide abundances, Klein and Langmuir,1987). Broadly correlated 3He/4He and Fe8 suggest that high 3He/4He mantle is preferentially sampled by deeper melting in the west(Graham et al., 2001). Conversely, the most strongly ‘Indian’-typemantle is associated with deeper axial depths and by inference,lower degrees of melting in the east (Mahoney et al., 2002). Aheterogeneous upper mantle with lower 3He/4He and higher Δ208Pb/204Pb material concentrated in the shallowest mantle and preferen-tially sampled by shallower melting at lower mantle temperature inthe east, explains the Fe8–3He/4He–Δ208Pb/204Pb relationship (Gra-ham et al., 2001; Mahoney et al., 2002).

Sr–Nd–Pb isotope data indicate little, if any, eastward mantle flowfrom the distant ASP or Kerguelen–Heard hotspots into the studyregion (Mahoney et al., 2002; Nicolaysen et al., 2007). Loosecorrelations of Sr, Nd, Pb and Hf isotopic compositions with theirrespective parent/daughter ratios have been interpreted as eithermixing trends or 200–700 Myr reservoir ages (perhaps reflectingsmall 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 torepresent the regional morphologic, geographic and geochemicalvariations within the study area (Fig. 1 and Table 1), were analyzed bythermal ionization mass spectrometry for Th–U–Ba concentrationsand U and Th isotopic compositions; a 12 sample subset was analyzedfor 226Ra concentration. The MORB U-series analytical protocols usedin this study are described in Rubin et al., 2005 (and referencestherein). Alteration-free glass samples were handpicked under abinocular microscope (see Table S6 for sample conditions andweights; the letter “S” indicates tables or figures that are in thesupplement). Analytical blank and standards meta-data are reportedin the Table 1 footnotes.

Scenarios for producing observed U-series disequilibria wereforward modeled, using both dynamic melting (McKenzie, 1985,Williams and Gill, 1989) and equilibrium porous flow (Spiegelman andElliott, 1993; Spiegelman, 2000) models. Multiple model simulationswere run to examine the relative roles of melt fraction (F), mantle

porosity (ϕ), and melting rate (M) for three mantle lithologies, asdescribed in Section 5.3 and the supplement.

4. Results

Analytical results are presented in Table 1; activity ratios aredenoted by parentheses in the table and throughout the text. (234U/238U) is effectively indistinguishable from unity in each of our samples(0.998±0.004 to 1.005±0.003), indicating that post-eruptive altera-tion/exchangewith seawater has not affected U-series activities in thissample set.

4.1. Ba–Th–U concentrations

Ba, Th, and U concentrations vary considerably in SEIR basalts (byfactors 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 morecommon in the east. Ba, Th, and U concentrations are each stronglycorrelated (r2N0.9) with (La/Sm)n over a threefold (La/Sm)n range.

4.2. (230Th/232Th) and (238U/232Th)

SEIR basalts form a continuous array on an equiline diagram, fromenriched (low) to depleted (high) (238U/232Th) values (Fig. 2) and spana larger range than any other ridge except the ultra-slow spreadingSouthwest Indian Ridge (SWIR), which has a similar range but non-continuous data distribution (Standish, 2006). The SEIR (238U/232Th)range is greater than other intermediate spreading rate ridges such asthe Gorda and Juan de Fuca (JDF) Ridges (Goldstein et al.,1992) and theeastern Galápagos Spreading Center (GSC; Kokfelt et al., 2005).

Dividing the SEIR data into western, central and eastern geo-graphic groups (WG, CG and EG; Table 1 and Fig. 2) reveals regionaldifferences in our sample suite. SEIR basalts can also be divided intothree slightly different relative source enrichment groups based onTh/U, forming distinct fields on an equiline diagram (Fig. 3A): enriched(≥3.4, n=5), intermediate (2.8bTh/Ub3.3, n=7), and depleted (≈2.4,n=4). Thorium concentrations in the intermediate and enrichedgroups overlap (0.17 to 0.58 and 0.25 to 1.16 µg/g, respectively), withimplications for variations in melting style and mantle sourcecomposition (see section 5.4).

230Th–238U disequilibria vary systematically along the SEIR. (230Th/232Th) and (238U/232Th) are highly correlated in samples having 226Ra-excesses (r2=0.94; Fig. 3A). Because the data lie to the left of theequiline (the locus of secular equilibrium), the linear regression slopeof 0.90 indicates slightly higher average 230Th-excesses at theenriched end (~16%) than the depleted end (~6%) of this array. Asimilar relationship in other MORBs has been loosely interpreted toreflect 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 becauseinitial disequilibria decay away at a rate governed by the shorter half-life of a parent-daughter pair (75.69 kyr and 1.6 kyr for 230Th and226Ra, respectively). It is difficult to ensure that mostMORBs are youngenough to be unaffected by post-eruptive decay (especially thosesampled by dredging, such as the SEIR basalts). Excess 226Ra, which iscommonly observed in young MORBs, occurs for ~8 kyr followingeruption (e.g., Rubin and Macdougall, 1988), making it a goodindicator of sample youth (i.e., (230Th/238U) will not have beenchanged by post-eruptive decay in lavas with (226Ra/230Th)N1).Assessing sample youth was the primary purpose of 226Ra analysesin this study. Eleven of twelve SEIR basalts analyzed for 226Ra haveexcess 226Ra (ranging from 15±4% to 161±4%) and lie on or within acomputed error envelope around the Fig. 3A linear regression. The

Table 1Southeast 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 NMORB

Ridgemorphology

[Th]ng/g

± [U]ng/g

± (234U/238U)

± 232Th/230Th

± (230Th/232Th)

± (238U/232Th)

± (230Th/238U)

± Rafg/g

± (226Ra/230Th)

± Baµg/g

Western Group (WG)71-1 42.9 90.8 2350 N AH 149.1 0.3 62.4 0.3 1.003 0.003 137500 1200 1.346 0.012 1.270 0.006 1.059 0.011 36.4 0.6 1.63 0.03 22.575-4 43.6 92.7 2585 N AH 156.4 0.2 64.7 0.3 1.003 0.004 139500 1300 1.327 0.012 1.254 0.006 1.058 0.011 21.678-2 44.8 94.8 2719 N AH 133.0 0.2 56.6 0.3 1.003 0.003 130500 1200 1.418 0.013 1.292 0.006 1.098 0.011 27.6 0.7 1.32 0.04 19.184-7 45.1 95.9 2660 N AH 130.4 0.1 55.0 0.3 1.000 0.004 142700 1300 1.297 0.012 1.279 0.006 1.014 0.010 49.1 0.5 2.61 0.04 13.189-4 47.5 97.5 2470 E RH 584.2 0.6 207.6 0.9 1.001 0.003 149200 1400 1.240 0.012 1.078 0.005 1.151 0.012 80.8 0.5 1.00 0.01 54.4WG averages 1.33 0.13 1.08 0.10 1.47 0.44

Central group (CG)96-1 47.3 100.7 2465 E AH 424.6 0.4 149.6 0.7 0.999 0.004 162100 1700 1.142 0.012 1.069 0.005 1.068 0.012 111.5 4.2 2.03 0.04 32.7100-1† 47.6 101.5 2857 E RH 548.0 0.5 168.4 0.8 1.001 0.004 164000 2000 1.129 0.014 0.933 0.004 1.210 0.016 40.6105-1† 47.8 103.0 2783 E RH 292.1 0.3 92.3 0.4 1.000 0.004 161300 1500 1.148 0.011 0.959 0.004 1.197 0.012 19.7113-7† 48.8 105.2 3630 N SV 269.9 0.3 82.7 0.4 1.002 0.004 167300 1700 1.106 0.011 0.930 0.004 1.190 0.013 20.7118-1 48.4 107.5 2673 E RH 943.0 1.1 267.3 1.2 1.002 0.002 178900 1700 1.035 0.010 0.860 0.004 1.203 0.013 165.6 1.4 1.53 0.03 74.1125-1 49.5 109.1 3475 N SV 169.3 0.2 58.0 0.3 0.998 0.003 158100 1600 1.170 0.012 1.038 0.005 1.127 0.012 30.3 0.3 1.38 0.02 12.0126-7 49.5 109.5 3240 E SV 1158.3 1.3 312.8 1.4 1.001 0.003 192100 1700 0.963 0.009 0.819 0.004 1.176 0.012 189.4 2.6 1.53 0.03 110.3132-1† 50.2 111.8 3328 E SV 279.7 0.3 81.6 0.4 1.003 0.004 188600 2900 0.981 0.014 0.885 0.004 1.108 0.018 57.2 1.1 1.87 0.05 22.6141-1 50.4 113.6 3002 E SV 432.3 0.6 142.4 0.6 0.998 0.004 165900 2000 1.116 0.013 0.975 0.005 1.145 0.015 95.4 1.0 1.73 0.12 35.3CG averages 1.09 0.15 1.16 0.10 1.68 0.50

Eastern group (EG)144-4 50.0 115.2 3997 N DV 247.3 0.3 72.7 0.3 1.005 0.003 183700 1800 1.007 0.010 0.892 0.004 1.129 0.012 31.8 1.0 1.15 0.04 18.1145-1† 49.3 116.7 4665 E DV 694.4 0.7 183.7 0.8 1.002 0.003 210700 2600 0.878 0.011 0.803 0.004 1.094 0.014 89.8 0.7 1.32 0.02 75.3EG averages 0.94 0.18 1.11 0.05 1.24 0.24

Table notes:

1. Isotope dilution concentrations and isotopic compositions weremeasured on the Univ. of Hawaii VG Sector 54 thermal ionizationmass 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 standardK1919 conducted during this study are reported in Rubin et al. (2005).

58C.J.Russo

etal./

Earthand

PlanetaryScience

Letters278

(2009)55

–66

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

Fig. 3. A) SEIR ‘Equiline’ diagram. SEIR basalts plotted by regional groupings; symbols arethe same as in Fig. 2; black dots additionally indicate samples with measured 226Ra andthus age control. Regional groups are furtherdivided into enrichmentgroupsdesignatedbydashed vertical lines. The solid black line is a least-squares regression through the datahaving measured 226Ra, following Lundstrom et al. (1998). Curved grey lines are the ±2standard error envelope on the linear regression. B) ‘Equiline’ diagram and axialmorphology. 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 (withradium control) for each morphologic group; 4 samples that may have suffered 230Th-decay are excluded. Note the generally lower 230Th-excess of AH lavas and the progressionto lower 230Th-excess for RH, SV andDV groups. AHmorphologies are ‘EPR-like’, with 200–400 m axial highs (segments C17-C14), RH segments have 100–200 m axial highs with~2 kmwide by 50–100m deep rifts (segments C13-C9), SV segments have 10–15 kmwideby 500–800 m deep rifts (segments C8-C4), and DV morphologies having well-defined(N1000 m deep) axial valleys.

59C.J. Russo et al. / Earth and Planetary Science Letters 278 (2009) 55–66

three CG and one WG samples not analyzed for 226Ra, plus the onesample (WW10-89-4) with (226Ra/230Th)=1, lie within or above thisenvelope, indicating that post-eruptive decay has not significantlylowered their 230Th excesses compared to those samples with (226Ra/230Th)N1. Thus we believe that our measured 230Th–238U disequilibriarecord 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 magmaticprocesses and timing, as we attempt to do here. A few studies haveused MORBs of known eruption age or geologic context fromsubmersible collection to largely avoid post-eruptive 226Ra decayeffects (e.g., Sims et al., 2002; Rubin et al., 2005). Most MORB studies,including this one, use dredged samples, for which age constraints andfield 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-con-strained, submersible collected samples.

There is an inverse relationship of (226Ra/230Th) and (230Th/238U) inage-constrained EPR basalts from the 9°–10° N (Lundstrom et al.,1999; Sims et al., 2002); our SEIR basalts exhibit the same inverserelationship, albeit at lower (226Ra/230Th) (Fig. 4). SEIR CG samples arecorrelated on this diagram and fall close to a line that represents 1 halflife of decay (1600 years) from the EPR trajectory (although one datumis displaced an additional half life below this line). Such a well-correlated trajectory is unlikely to result from randomly dispersedinitial ratios experiencing different amounts of decay. One cantherefore speculate that this CG data array mimics the data dispersionas it was upon eruption, implying that the conditions that arose in theEPR trajectory likewise exist along this part of the SEIR. However, it isnot possible with the present data set to determine whether these CGbasalts initially followed the EPR array and have since decayed, or ifzero-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-excessfor a given 230Th-excess (Fig. 4). Post-eruptive aging of a simple, EPR-like array could only produce this if sample aging increased smoothlyfrom basalts having low (230Th/238U) to those having high (230Th/238U). One WG sample (84–7) has 160% 226Ra-excess; if this samplehas 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

MORBs (~4). Instead, this basalt is likely substantially younger than1600 years. (226Ra/230Th) is much lower in the other WG/EG basalts. Itis difficult to tell a priori whether this is because these samples arefrom older lava flows or if lower (226Ra/230Th) reflects different meltgeneration conditions beneath these ridge segments. This latterpossibility is considered below in conjunction with (230Th/238U) inthe lavas and melt model predictions, but we stress that primary226Ra-excesses are not well-constrained in our data set, so we relymore heavily on 230Th-excesses to quantify melting conditions.

5. Discussion

5.1. U-series variations along the SEIR

U-series signatures varyalong the SEIR, bothwithin and between theaforementioned WG, CG and EG geographic groups (Figs. 2, 3A, 4–6A).

Fig. 4. (226Ra/230Th) versus (230Th/238U). (A) SEIR basalts separated into regional groups.Compare the inverse relationships in SEIR basalts to submersible collected, age-constrained 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 centralgroup (CG) lavas lying along a trend at relatively higher values of (226Ra/230Th) thanwestern group (WG) and eastern group (WG) lavas. Some of this difference is probablydue to post-eruptive decay of 226Ra, although decay is not expected to move randomlydistributed zero-age samples onto a well correlated line (such as the CG samples). Thesolid black line is a least squares regression through the EPR data. The dashed line is theslope of the EPR trend after one half-life of decay. (B) CG measured data, and datacorrected for potential 226Ra decay. Much of the CG data follows this same relationshipas the EPR if corrected for one half-life of decay (open symbols). Correction for two half-lives would raise the lowest CG sample to the same trend as the others. (C) EG and WGmeasured data and data corrected for potential 226Ra decay. The WG and EG sampleshave generally lower 226Ra-excess, which may be due either to more extensive decaythan in CG samples or to different melting conditions.

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

All basalts have 230Th-excesses, ranging from 21±1.2% to 1.4±1.0%. This20% 230Th-excess range is smaller than in young (i.e., with excess 226Ra),intermediate spreading rate MORBs from the JDF-Gorda Ridges (230Th-excess 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 convexpattern, with lower values associated with the western and easternends 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 geographicpattern, generally decreasing from WG to CG to EG segments andprimarily reflecting mantle source composition before melting.

Aspects of (226Ra/230Th) variations in the sample set likely reflectinitial conditions at the time of eruption (e.g., the CG (230Th/238U) vs.(226Ra/230Th) inverse correlation and the high (226Ra/230Th) in WG

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 ofthese differences may be due to post-eruptive 226Ra decay (section4.3), but we also consider the implications of these regional Ravariations assuming some or all are magmatic.

5.2. Axial depth and 230Th/238U disequilibria

Bourdon et al. (1996a) interpreted a weak negative correlationbetween ridge segment averaged axial depth and (230Th/238U) in asubstantially filtered early global MORB data set to result frommantletemperature control on depth/degree of melting. Their model involvesmelting of hotter mantle beneath shallower ridges producing larger(230Th/238U) via a deeper peridotite solidus, higher overall F, thickeroceanic crust, and more melt generated in the garnet stability field(where Th is more incompatible then U; Beattie, 1993; LaTourretteet al., 1993; Blundy and Wood, 2003).

The (230Th/238U) vs. axial depth relationship is less evident whennewer data from the deep southern MAR and the shallow ReykjanesRidge are included in the global compilation (Lundstrom et al., 1998;Peate et al., 2001; Lundstrom, 2003). MORBs from intermediate axialdepths (~3000–3500m) are highly variable and span nearly the entireMORB range (Fig. 6). The global data set still displays a weak negativecovariation but the statistical significance of the global trend has beenquestioned, and it is largely defined by ‘end-member’ samples fromthe shallow Azores platform (MAR) and the unusually deep AAD(Elliott and Spiegelman, 2003). The two deepest samples from theAAD have not been analyzed for 226Ra, and since their (230Th/238U) issubstantially different from our nearest EG samples, there arelegitimate questions concerning whether they have experiencedsignificant post-eruptive 230Th decay.

Axial depth and (230Th/238U) are not negatively correlated alongthe SEIR, although SEIR samples do largely fit within the publishedMORB swath (Fig. 6). Yet (230Th/238U) displays the entire SEIR range inthe narrow axial depth range from 2300 to 2700 m. In addition, manyWG basalts have lower than expected 230Th-excesses for their axialdepths and the two EG basalts erupted at the greatest depth have thesame (230Th/238U) as CG and WG basalts erupted much shallower. Ifthere truly is aweak global trend, the lack of a (230Th/238U)-axial depthcorrelation along the SEIR suggests that melt initiation depth does notrespond to regional mantle temperature differences in the wayenvisioned to explain the global data set. Other variables, includingM, ϕ and mantle composition likely vary along the SEIR in a way thatobscures any simple potential (230Th/238U)-axial depth relationship.

5.3. Melt supply, axial morphology and U-series disequilibria

5.3.1. General observations and modelingMelt supply (i.e., the volumetric flux of melt from mantle to crust)

affects ridge axismorphology, AMC depth and lava chemistry. Thermalmodels predict that AMC depth and melt supply are inversely related(Phipps Morgan and Chen, 1993; Chen 1996); ridge-axis morphologyreflects interplays between melt supply, mantle temperature, andhydrothermal cooling (e.g., Chen, 2000); and strong correlationsbetween lava chemistry, AMC depth and spreading rate demonstratehow such thermal conditions control magma accumulation conditionsandMORB compositions (Rubin and Sinton, 2007). Coincident changesin AMC depth and ridgemorphology at some ridges also support thesethermal models (e.g., at EPR segments north and south of the Orozcotransform; Chen, 2000, Carbotte et al., 1998). Intermediate spreadingridges such as the SEIR are generally more sensitive to magma supplyvariations than ridges with faster or slower spreading rates (e.g., Niuand Batiza 1993; Carbotte et al., 2006; Detrick et al., 2002), and smallmagma supply variations can produce large variations in ridge axismorphology (Chen, 1996).

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/238U 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)-axialdepth relationship of Bourdon et al. (1996a). (E) Measured (230Th/232Th). (F) Measured (226Ra/230Th).

61C.J. Russo et al. / Earth and Planetary Science Letters 278 (2009) 55–66

Shallower axial depth, axial high morphology and greater crustalthickness all point to greater melt supply in thewestern portion of ourSEIR study area. Systematic variations in these parameters as well as inU-series disequilibria (at nearly constant, intermediate spreading rate)provide a unique opportunity to quantify and investigate the under-lying melt supply variations.

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

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 byLundstrom (2003). (B) SEIR basalts grouped by ridge morphology (Symbols are thesame as in Fig. 3B). Small grey circles as in panel A. Other data include GalápagosSpreading Center samples with measured 226Ra-excesses (open circles; Kokfelt et al.,2005).

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

2005). Two general types of models have been developed thatconstrain the rate, duration and extent of mantle melting (e.g., Elliott,1997); both are one-dimensional, consider variations with depthbeneath the ridge axis, and produce radioactive disequilibria primarilyby variations in M, ϕ and F. The main differences between the modelsare in specific melting conditions and whether or not melt and solidmaintain chemical equilibrium throughout the melting column.Dynamic melting (e.g., McKenzie, 1985; Williams and Gill, 1989)predicts compositions of melts that remainwith the solid residue untila porosity threshold, followed bymelt extraction everywhere from themelt column at a rate balanced by continued melt production,resulting in constant ϕ. This model behaves like accumulatedfractional melting for non-radioactive elements (e.g., Williams andGill, 1989). The equilibrium porous flow (EPF) model predicts thecombined effects of chemical fractionation, radioactive decay, andmelt transport time on U-series disequilibria (Spiegelman and Elliott,1993). M is not an input parameter but can be calculated from EPFmodel inputs (i.e., depths of melt initiation and cessation, plus theamount of melt present at each depth). Melt simulations described inthis paper were run at the same conditions of M, ϕ and F for bothtypes of models (see the Supplementary data).

In general these melt models do not uniquely and quantitativelypredict melting conditions that lead to specific disequilibria values(e.g., for 230Th–238U) because (a) model outputs are highly dependenton poorly constrained input parameters, (b) multiple combinations ofparameters can generate the same result, and (c) such parameteriza-tions have greater degrees of freedom when they are used to matchjust one U-series element pair. Here we use the results of modelsimulations as qualitative indicators of conditions that may result indisequilibria variations between groups of SEIR samples withoutemphasizing the absolute values of a particular model simulation.Many previous authors have used these same models to constrainmantle melt column parameters, somemore literally than others (e.g.,Williams and Gill, 1989; Spiegelman and Elliott, 1993; Lundstromet 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

of melting parameters have the biggest effect on 230Th-excess inMORB, with excesses being reduced by: 1) greater porosity, ϕ;2) greater melt fraction, F; 3) higher melting rate, M; and in someinstances, 4) shallower melt initiation depth. A wide range ofconditions result in excess 226Ra, but the magnitude of the excess ishighly dependent on melting and melt transport rates because the226Ra half life is so short.

5.3.2. Melt supply along the western portion of the study areaThe SEIR samples were subdivided into four groups (Table 1;

Section 2.1) that reflect their axial morphology, AMC depth and crustalthickness. These morphologically grouped samples cluster discretelyon the (230Th/238U) vs. axial depth diagram (Fig. 6B). The RH, SV andDV groups form an array that resembles the global trend of Bourdonet al. (1996a). However, the AH group (i.e., with highest melt supply)plots well below that trend at lower 230Th-excess for the same axialdepth as RH samples. Shallower AMC depth beneath segment C12(Baran et al., 2005) is one indication that higher melt supply may bethe underlying cause of this RH-to-AH morphological and 230Th-excess shift.

Observations at the intermediate spreading rate GalápagosSpreading Center provide context for these observations. The GSChas a strong axial depth gradient from interaction of the ridge and theGalápagos hotspot, and exhibits AH morphology in a ‘central bulge’region nearest the hotspot due to excess magma supply, RHmorphology 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 non-AH segments (Kokfelt et al., 2005).

The nearest hotspots to our study area are 700 to 1600 km awayand do not directly influence magma supply, so something else mustbe increasing magma supply to account for the AH morphology,shallower ridge depth and lower 230Th-excesses than RH segments. Inprinciple, all could result from some combination of larger F, greaterϕ, and faster M. Increased melt productivity (or melting rate) willproduce larger F at constant melt column height if the mantle is morefertile or at higher temperature. Nothing from the trace element orisotope geochemistry implies that SEIR AH basalts were produced bymore fertile mantle, but a mantle temperature difference with RHsegments cannot be ruled out.

Producing extra melt by only deepening the melt initiation depthbeneath AH segments would lead to a greater proportion of meltgenerated in the garnet stability field, thereby increasing rather thanlowering melt 230Th-excesses, which is inconsistent with the SEIRobservations. Likewise, it is possible to increase melt supply at constantmelt columnproperties if a greater proportionofmeltwere sampled fromoff-axis portions of the triangular melting region beneath the ridge, butsmaller degree melts produced there would have greater 230Th-excess.

Producing a lower proportion of melt within the garnet stabilityfield would generate lower melt 230Th-excesses (i.e., if melt initiationdepth were shallower beneath AH versus RH segments). For example,230Th-excess is reduced by ~25% when initiation depth is decreasedfrom 40 kb to 25 kb (Fig. S4). Garnet no longer influences the meltcomposition if initial melting depth is reduced further, dramaticallyreducing melt 230Th-excess (e.g., compare dynamic and EPF meltingsimulations of spinel lherzolite and garnet lherzolite in Figs. S1 andS2). Some spinel lherzolite model simulations reproduce the 230Th-excesses observed in AH and RH samples, but only at unrealisticallylow 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 therelatively low 226Ra excess in some WG samples (i.e., except WW10-84-7). Furthermore the very small degrees of melting in thesesimulations are inconsistent with ≥6.3 km thick crust in the westernregion (see calculations in the supplement). Thus shallow meltinitiation depths are unlikely to have caused the low 230Th-excessesin the WG samples. However, the unusual characteristics of sample

63C.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%226Ra-excess) make it the one exception, in that spinel peridotitemelting simulations can produce these characteristics and garnetperidotite simulations cannot. This sample was taken from a smalldying ridge segment between a westward propagating segment andthe transform near 96° E, a likely place for a shallow, low productivitymelt column.

It is unlikely that melt initiation depth varies significantly beneathmost AH and RH segments since changing it cannot produce theobserved decrease in average 230Th-excess at AH segments (6%)relative to nearby RH segments (19%). Assuming constant initialmelting depth, one can forward model our U-series observations toqualitatively determine howmuchmelt supplymust increase fromRHto AH segments to change the axial morphology and reduce the 230Th-excess. An increase in bothM and ϕwill reduce 230Th-excess, whereasonly an increase in Mwill 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) atconstant ϕ (0.25%) and bulk Kd, will produce this 230Th-excess shift bydynamic melting with essentially no effect on 226Ra-excess (Table S3).This change inM can be reduced slightly to a five-fold increase (from 1to 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 alongwith 230Th-excess (Fig. S1), consistent with lower 226Ra-excess observed in all butone AH sample. The (230Th/238U) difference could also be madeby 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 thatwould allow such large ϕ increase at constant upwelling and meltingrate, so we consider model simulations requiring some change in bothM and ϕ to be more realistic. We speculate that higher ϕ mightincrease melt transport efficiency and thus melt supply, but such melttransport mechanics are beyond the scope of this work.

5.3.3. Melt supply along the eastern portion of the study areaIt is conceptually straightforward to envision how an increase in M

could supply the extra melt needed to produce the U-series, AMCdepth and morphology of AH segments on the SEIR. However, thisrationale does not explain the lower 230Th-excesses (and very low226Ra-excesses) in EG segments with morphology and crustalthickness indicative of decreased melt supply. SEIR basalts fromlowest melt supply DV segments have 230Th-excesses between thoseof high melt supply AH segments and moderate melt supply RH andSV 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 thatmelting conditions beneath DV segments must suppress melt 230Th-excess without increasing the net melt flux from the mantle. Ifincreased ϕ and/or M were the cause of reduced 230Th-excess, DVsegments would require a shorter melt column than SV segments. Forexample, 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 withinthe garnet stability field. However, a shallower melt initiation depthalone cannot explain the mean 230Th-excess shift from SV to DVsamples because the lavas also require a change to more enrichedmantle source composition (e.g., Th/U, radiogenic isotope ratios).

The difficulty with generating the combined variations in 230Th-excess, 226Ra-excess, Th/U and melt supply for the EG basalts withonly physical changes to the melt column leads us to considercompositional changes that can influence melt productivity, Th and Ubulk distribution coefficients, and source enrichment.

5.4. Mantle source lithology and 230Th–238U disequilibria

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

Macdougall, 1992; Lundstrom et al., 1998). E-MORBs (which oftenhave elevated Th/U) can have larger 230Th-excesses than N-MORBsfrom the same region, producing equiline diagram data arrays withslopes b1 (Lundstrom et al., 1998). Lundstrom et al. (1998) postulatedthat increasing the modal proportion of garnet (to lower the relativeTh bulk distribution coefficient) and clinopyroxene (to increasefertility) in the E-MORB source increases 230Th–238U fractionationduring melting. Melts of such a source would have larger 230Th-excesses than melts of more depleted mantle, and their mixtureswould produce the observed equiline diagram arrays. Variations inmantle upwelling rate can be superimposed on this mantle composi-tion effect to flatten the array slope at slower compared to fasterspreading ridges because slower upwelling leads to relativelygreater melt 230Th-excesses by 230Th production in the meltingcolumn. The slope of the full SEIR array (0.90) is significantly steeperthan slopes for other intermediate spreading ridges (JDF slope=0.42and Gorda Ridge slope=0.32; Lundstrom et al., 1998). The SEIRslope is, in fact, steeper than the steepest slope (0.82) for the super-fast spreading EPR at 20°–27°S, where the high upwelling rateapparently counteracts any source lithologic effect on melt (230Th/238U) (Lundstrom et al., 1998).

Given the nearly constant intermediate spreading rate and largeTh/U range (plus other isotopic variations), our Fig. 3A data array slopeindicates that different mantle compositions beneath the SEIR do notmelt in themanner 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 toproduce lower 230Th-excess), as are the inflated axial morphology andshallow ridge depth in the area. N-MORBs and E-MORBs (defined ashaving (La/Sm)NN0.80; Mahoney et al., 2002) occur throughout theSEIR study area with no systematic offset in (230Th/238U) or (230Th/232Th) (Figs. S5a and S5b). Thus the origin of the large Th/U (and Ba/Thand La/Sm) range along the SEIR (e.g., Fig. 3A) must be differentfrom that which manifests itself as Th/U and (230Th/238U) variations inN-MORBs and E-MORBs elsewhere.

Although sample numbers are small in this study, there aresystematic variations in source composition, melt supply and ridgemorphology that suggest a link between variations in mantlelithology, mantle temperature and melting conditions along the SEIRaxis. Average Th/U of each of our morphological groups indicatedifferent mean mantle compositions beneath ridge sections of somemorphologies (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–238Udisequilibrium differences between our various morphologic groups.The AH segment has the lowest Th/U and 230Th-excess (6%). The otherthree morphologic groups (RH, SV and DV) form vertically stackedarrays on an equiline diagram, with decreasing average 230Th-excesses(i.e., 19%, 15% and 11%, respectively; Fig. 3B). Because it is difficult tofractionate Th from U duringMORBmantle melting (e.g., Elliott, 1997),decreasing 230Th-excesses at increasing Th/U in these latter threegroups implies a change in mantle composition (e.g., lithology, whichcan affect Th/U) and how that compositionmelts (which affects 230Th-excess). The trace element and isotopic gradients along this part of theSEIR also imply a gradient in meanmantle composition, such as wouldbe produced by variable proportions of depleted and enrichedlithologies.

Least squares regression slopes for the individual morphologicaldata arrays on Fig. 3B are ≤1 for the moderate to high melt supplygroups (SV, RH, AH) and N1 for the DV group. The more enrichedsamples in the SV, RH and AH groups have slightly larger 230Th-excessthan the depleted samples, whereas for the DV group the moreenriched sample has smaller 230Th-excess. The “stacked” RH, SV andDV data arrays rotate toward steeper slopes as the average (238U/232Th) decreases and as the SEIR deepens to the east, so that increasingthe proportion of melts from mantle with higher (i.e., less depleted)

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=4.0and μ1=9.0. The grey band is the MORB and ocean island basalt data trend of Vlastelic etal. (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 evolutionmodels, using equations from Gast (1969), and assuming T=4.55 Ga, μ1=μ2=9.0, κ1=4.0and κ2=2.35, where the subscripts 1 and 2 refer to the evolution stages; the age ofsecond-stage fractionation (t) is given by numbers at the ends of the grey lines, andranges from 4.0 Ga to 0 (today).

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. Thiseffect cannot be produced by changing melting parameters in a singlelithology 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 fertilelithologies melt more productively than typical MORB-source peri-dotite under the same conditions, producing melts with lower 230Th-excess (Stracke et al., 1999; Elkins et al., 2008). Both dynamic and EPFmelting of garnet pyroxenite (45% cpx; 45% opx; 10% gnt; seesupplementary information) produce lower 230Th-excesses andlower 226Ra-excesses (~10% and 25%, respectively) at M that are 0.25to 4 times higher than peridotite and at ϕ=2.5% (i.e., the center of theporosity range required for melt compaction and segregation withoutdiffusive re-equilibration with surrounding peridotite matrix; Kogisoet al., 2004; Fig. S3 and Table S5). The high M and F for the pyroxene-rich lithology in these models exert a dominant control on themagnitude of 230Th-excesses in the partial melts, in part becauseproportionally more melt is produced, and in part because Thconcentration is higher in that (enriched) source lithology. Meltingof mantle that is variably enriched in garnet pyroxenite (and/oreclogite) can therefore account for both high Th/U and lower, moresimilar 226Ra-excesses and 230Th-excesses in SEIR DV segment lavas.Importantly, melting scenarios that produce low 230Th-excesses butdo not involve pyroxenite (Section 5.3) cannot produce all of the U-series, Th/U and radiogenic isotope characteristics observed along theSEIR.

5.5. Upper mantle lithology beneath the Indian Ocean

Isotope gradients along the SEIR indicate a significant regionalvariation in mantle source composition (Graham et al., 2001;Mahoney et al., 2002). 232Th/238U (κ) and (230Th/232Th) also varydramatically in these same samples, suggesting an upper mantle linkto all of these signatures.

The highest (230Th/232Th) and lowest κ values occur in the westernportion of the study area, progressing smoothly eastward to lower(230Th/232Th) andhigherκ (Fig. 5).κPb likewise varies smoothlyeastwardalong axis to higher values (κPb is the time-integrated Th/U ratio,computed from 208Pb/204Pb and 206Pb/204Pb, as discussed in the Fig. 5caption; e.g., Gast, 1969; Chase, 1981; Allegre et al., 1986; O'Nions andMcKenzie, 1993; Elliott et al., 1999). κ and κPb were only computed forthe subset of SEIR basalts having Pb isotopes (Mahoney et al., 2002) andisotope dilution Th and U concentrations (this study) because there issubstantial non-systematic variation in theMahoney et al. icp-ms Th–Udata. The κ and κPb ranges among SEIR basalts encompass a significantportion 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 thatκbbκPb in the west and κ≈κPb≈4.0 in the east. Globally, most MORBsdisplay κbκPb (Elliott et al., 1999; Kokfelt et al., 2005; Sims and Hart,2006; Vlastelic et al., 2006), as expected after one or more relativelyrecent (b0.5Ga)melt depletionevents have fractionatedTh fromU in themantle source. κ values in the easternmost SEIR study area are especiallynoteworthy because (a) they are close to the bulk silicate Earth value,implying a considerably more fertile mantle source than typical MORBsource mantle, and (b) κ≈κPb is rare in MORBs and implies long-timequasi-closed system evolution for this part of the mantle (109 years).

The linear SEIR κ–κPb data array (Fig. 7) could have been producedby magma generation from mantle regions having variable prior meltdepletion histories (via instantaneous or continuous scenarios; insetFig. 7), or by a variation in mixing proportions of melts from low-κupper mantle (i.e., like most global MORB) and a high-κ enrichedlithology. A detailed evaluation of how the SEIR κ–κPb array formed isbeyond the scope of this paper. Nevertheless, the array and systematicκ, κPb and (230Th/238U) variations record a longitudinal shift in mantle

composition, past evolutionary history and currentmelting conditionsbeneath 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 somemelting of garnet pyroxenite is required to explain the 226Ra–230Th–238U–232Th variations along the eastern SEIR. An enriched garnetpyroxenite lithology that has not been melted in the recent past willhave distinct isotopic characteristics and high Th, U and Pbconcentrations, so admixture of a small amount of this materialdispersed in the upper mantle could also produce the high κ≈κPb inthose samemagmas. Garnet pyroxenite melting simulations at highM(e.g., productivity of 0.5–1%/km) and high F (~25%) produce meltswith ~10% 230Th-excess and subdued 226Ra-excess (b50%). If easternSEIR basalts are mixtures of such pyroxenite melts with melts ofnormal depletedMORBmantle (F=7–10%, as discussed previously andin Fig. S6a), then no more than ~10% of the total melt producedbeneath the eastern SEIR is from garnet pyroxenite (assuming Th/U=4in the pyroxenite source; see the supplement). The same meltproportions explain the SEIR κ–κPb variations (Fig. S6b), and alsoindicate that a small amount of pyroxenite might be present in themantle beneath some of our CG sample sites. Mantle proportionscalculated from these melt mixtures and the respective pyroxeniteand peridotite melt productivities suggest a maximum of 3–4%pyroxenite by volume as veins (or blobs) within 96–97% peridotitebeneath the easternmost SEIR. This range is similar to volumeestimates of mafic veins observed in ophiolites and alpine massifs(e.g., Mukasa et al., 1991).

Sobolev et al. (2007) used trace element contents of olivinephenocrysts to suggest that MORBs globally are formed from mantlewith 10–15% pyroxenite compared to generally higher amounts inIceland (20%), Hawaii (60%) and the Siberian traps (100%). Such highMORB source proportions are difficult to reconcile with our muchlower (~5–25 fold) estimates for the SEIR. Regional MORB differencesare not discussed by Sobolev et al. (2007). Using their Mn/Femethodology and SEIR samples in their data supplement, wedetermined that mean mantle source pyroxenite fractions areessentially constant at 10–12% in sub-regions of our study area

65C.J. Russo et al. / Earth and Planetary Science Letters 278 (2009) 55–66

where U-series variations (this study) suggest a pyroxenite gradient. Itis unclear why these two studies differ in inferred pyroxenite fractionand distribution beneath the SEIR.

Sm–Nd and Lu–Hf systematics have been used to determine theproportion of MORB melting that occurs in the garnet stability field(e.g., Salters, 1996). Stracke et al. (1999) argued that thesesystematics coupled to (230Th/238U) could distinguish garnetpyroxenite and peridotite in Hawaiian mantle derived melts.However, Sm/Nd and Lu/Hf of accumulated fractional melts underour estimated SEIR melting conditions and pyroxenite proportionsdo not distinguish pyroxenite-free and pyroxenite-containinglithologies, indicating that these ratios are not as sensitive to thepresence of pyroxenite as (230Th/238U). The method employed bySalters and Hart (1989) and Stracke et al. (1999) uses δ(Sm/Nd) andδ(Lu/Hf), parameters that reflect relative offsets in the elementalratios from those predicted by 143Nd/144Nd and 176Hf/177Hf andassuming the same reservoir age for pyroxenite and peridotitesources. 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 ourEG and some CG basalts are best explained by melting of mantlecontaining a few percent of enriched garnet pyroxenite mixed into anupper mantle composed primarily of peridotite. We speculate that thehigher pyroxenite fraction in the eastern Indian Ocean, near theboundary with Pacific type mantle in the AAD may reflect thedynamics of mantle flow in the Indian Ocean; pyroxenite may havebeen preferentially depleted from the central part of the ocean basinand less so on the margins, perhaps when the SEIR passed over theKerguelen hotspot at 38 Ma (Duncan and Storey, 1992).

6. Summary

Variations in axial depth, crustal thickness, U-series disequilibria, κand κPb along the SEIR are best explained by models in which thephysical characteristics of melting (such as M, F, ϕ and melt initiationdepth) vary along axis in response to an inferred long wavelengthtemperature gradient and small variations in the underlying sourcelithology (pyroxenite veins). This leads to decreased melt supply fromwest to east, a concomitant change in axial morphology and axialmagma chamber depth, and systematic long wavelength geochemicalgradients. Melting rate covaries with melt supply in the western andcentral regions, decreasing eastward along the SEIR. However, meltsupply and melting rate vary inversely from the central to the easternregions, because cooler eastern mantle contains a larger proportion ofmore fusible pyroxenite veins. These veins melt productively togenerate melts with low 230Th-excess and 226Ra-excess and high Th,U and Ra concentrations, but still contribute only a small proportionof the total melt volume supplied to the ridge (which is dominatedby mantle that is N95% peridotite). Th, Pb and He isotopic gradientsare also consistent with a greater proportion of enriched pyroxenitewith a reservoir age of roughly 0.5 to 1 Gyr eastward along the SEIRaxis.

Acknowledgments

This paper benefited greatly from discussions and long-termcollaborations with D. Christie, J. Mahoney and D. Pyle, as well astwo anonymous reviews and editorial comments from Rick Carlson.This study was supported by the National Science Foundation MarineGeology & Geophysics Program (grants 02-21069 and 02-21240 to KRand DG).

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.epsl.2008.11.016.

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 uppercrustal structure due to variable mantle temperature along the Southeast IndianRidge. Geochem. Geophys. Geosys. 6 (11), Q11002. doi:10.1029/2005GC000943.

Beattie, P., 1993. Uranium-thorium disequilibria and partitioning on melting of garnetperidotite. Nature 363, 63–65.

Blundy, J., Wood, B., 2003. Mineral-melt partitioning of uranium, thorium and theirdaughters. Rev. Mineral. Geochem. 52, 59–123.

Bourdon, B., Zindler, A., Elliott, T., Langmuir, C.H., 1996a. Constraints on mantle melting atmid-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 Mid-Atlantic 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 forthe Grande Comore volcanics: melting processes in an upwelling plume. EarthPlanet. 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 theinfluence 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 spreadingridge on the influence of magma supply and spreading rate on crustal magmabodies 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 theintermediate spreading Juan de Fuca Ridge. Geology 34, 209–212.

Chase, C.G., 1981. Oceanic island Pb: two-stage histories and mantle evolution. EarthPlanet. Sci. Lett. 52, 277–284.

Chen, Y.J., 1996. Dynamics of the mid-ocean ridge plate boundary: recent observationsand theory. PAGEOPH 146 (3/4), 621–648.

Chen, Y.J., 2000. Dependence of crustal accretion and ridge-axis topography onspreading rate, mantle temperature, and hydrothermal cooling. In: Dilek, U.,Moores, E.M., Elthon, D., Nicloas, A. (Eds.), Ophiolites and Oceanic Crust: NewInsights 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 andgeochemical 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 Ridgebetween 88°E and 118°E: gravity anomalies and crustal accretion at intermediatespreading 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 evidencefrom mulitichannel reflection data for a crustal magma chamber in the MARK areaon 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, andvolcanological manifestations of plume-ridge interaction along the GalápagosSpreading Center. Geochem. Geophys. Geosys. 3 10.1029200GC000350.

Douglas-Priebe, L.M.; 1998 Geochemical and petrogenetic effects of the interaction ofthe Southeast Indian Ridge and the Amsterdam-St.Paul hotspot, M.S. thesis, OregonState 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 theScientific Drilling in the Indian Ocean. Geophysical Monograph, vol. 70. AmericanGeophysical Union, pp. 91–103.

Elkins, L.J., Gaetani, G.A., Sims, K.W.W., 2008. Partitioning of U and Th during garnetpyroxenite partial melting: constraints on the source of alkaline ocean islandbasalts. 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 U-series perspective. Treatise Geochem. 3, 465–510.

Elliott, T., Zindler, A., Bourdon,B.,1999. Exploring thekappaconundrum: the roleof recyclingin 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 AscensionIslands. Earth Planet. Sci. Lett. 5, 353–359.

Goldstein, S.J., Murrell, M.T., Janecky, D.R., Delaney, J.R., Calgue, D.A., 1992. Geochronologyand petrogenesis of MORB from the Juan de Fuca and Gorda ridges by 238U–230Thdisequilibrium. Earth Planet. Sci. Lett. 109, 255–272.

Graham, D.W., Lupton, J.E., Spera, F.J., Christie, D.M., 2001. Upper-mantle dynamicsrevealed 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. Crypticstriations in the upper mantle revealed by hafnium isotopes in southeast Indianridge 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 originof 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 variationsalong 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 chemistrywith 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 theSoutheast Indian Ridge, 115° E–138° E. J. Geophys. Res. 96, 2089–2107.

Kogiso, T., Hirschmann, M.M., Pertermann, M., 2004. High-pressure partial melting ofmafic lithologies in the mantle. J. Petrol. 45, 2407–2422.

Kokfelt, T.F., Lundstrom, C., Hoernle, K., Hauff, F., Werner, R., 2005. Plume-ridgeinteraction studied at the Galápagos spreading center: evidence from 226Ra–230Th–238U 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 ridgebasalts: 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-OceanRidges. Geophysical Monograph, vol. 71. American Geophysical Union, pp. 183–280.

LaTourrette, T.Z., Kennedy, A.K., Wasserburg, G.J., 1993. Thorium–uranium fractionationby 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 basaltextraction by equilibrium porous flow. Science 270 (5244), 1958–1961.

Lundstrom, C.C., Williams, Q., Gill, J.B., 1998. Investigating solid mantle upwelling ratesbeneath mid-ocean ridges using U-series disequilibria, 1: a global approach. EarthPlanet. Sci. Lett. 157, 151–165.

Lundstrom, C.C., Sampson, D.E., Perfit, M.R., Gill, J., Williams, Q., 1999. Insights into mid-ocean ridge basalt petrogenesis: U-series disequilibria from the Siqueiros Trans-form, 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 IndianRidge asthenosphere, 86°E–118°E. J. Petrol. 43, 1155–1176.

McKenzie, D., 1985. 230Th–238U disequilibrium and the melting processes beneath ridgeaxes. Earth Planet. Sci. Lett. 72, 149–157.

McKenzie, D., 2000. Constraints onmelt generation and transport from U-series activityratios. Chem. Geol. 162, 81–94.

Mukasa, S.B., Shervais, J.W., Wilshire, H.G., Nielson, J.E., 1991. Intrinsic Nd, Pb, and Srisotopic heterogeneities exhibited by the Lherz Alpine Peridotite Massif, Frenchpyrenees. J. Petrol. 117–134 Special Lherzolite Issue.

Nicolaysen, K.P., Frey, F.A., Mahoney, J.J., Johnson, K.T.M., Graham, D.W., 2007. Influenceof 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 ridgesegmentation. Geochem. Geophys. Geosys. 8, Q09007. doi:10.1029/2006GC001540.

Niu, Y., Batiza, R., 1993. Chemical variation trends at fast and slow spreading mid-oceanridges. J. Geophys. Res. 98, 7887–7902.

Niu, Y., Hekinian, R., 1997. Spreading-rate dependence of the extent of mantle meltingbeneath ocean ridges. Nature 385, 326–329.

Niu, Y., O, M.J., ' Hara, 2008. Global correlations of ocean ridge basalt chemistry withaxial 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 theEarth. Philos. Trans., 342, no. 1663, pp. 65–77.

Peate,D.W.,Hawkesworth, C.J., vanCalsteren,P.W., Taylor, R.N.,Murton, B.J., 2001. 238U–230Thconstraints on mantle upwelling and plume-ridge interaction along the ReykjanesRidge. Earth Planet. Sci. Lett. 187, 259–272.

Pertermann, M., Hirschmann, M.M., 2003. Partial melting experiments on a MORB-likepyroxenite between 2 and 3 GPa: constraints on the presence of pyroxenite inbasalt 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 magmasupply and spreading rate. Nature 364, 706–708.

Pyle, D.G.; 1994, Geochemistry of mid-ocean-ridge basalt within and surrounding theAustralian-Antarctic Discordance; Ph.D. dissertation, Oregon State University,Corvallis.

Rubin, K.H., Macdougall, J.D., 1988. 226Ra excesses in mid-ocean-ridge basalts andmantle 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 magmaticstructure 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 limitfor 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 Ndisotope 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 MORBsource. Nature 342, 420–422.

Sempéré, J.-C., Cochran, J.R., SEIR Scientific Team, 1997. The Southeast Indian Ridgebetween 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. TheAustralian-Antarctic discordance. Geology 19, 429–432.

Shen, Y., Forsyth, D.W., 1995. Geochemical constraints on initial and final depths ofmelting 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 oceanicbasalts: 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 magmabeneath 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 lavasandmelt inclusions from the Southeast Indian Ridge, the Endeavour Segment of theJuan de Fuca Ridge, and the northern East Pacific Rise, Ph.D. thesis, Oregon StateUniversity.

Spiegelman, M., 2000. UserCalc: a web-based U-series calculator for mantle meltingproblems. Geochem. Geophys. Geosys. 1 (0030).

Spiegelman, M., Elliott, T., 1993. Consequences of melt transport for uranium seriesdisequilibrium in young lavas. Earth Planet. Sci. Lett. 118, 1–20.

Standish, J.J., 2006, The influence of Ridge Geometry at the Ultraslow-SpreadingSouthwest Indian Ridge (9°–25°E): Basalt Composition Sensitivitiy to Variation inSource 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 ofmelting 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 ofmeteorites 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 inMORB from the Garrett transform and implications for mantle melting. EarthPlanet. 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 differentiatedmantle domain. Earth Planet. Sci. Lett.248, 379–393.

Volpe, A.M., Goldstein, S.J., 1993. 226Ra–230Th disequilibrium in axial and off-axis mid-ocean 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.

Supplement to Russo et al., 2008, page 1

Contents of Supplementary Information for Russo et al., 2008 1. Melting Models

Description and Conditions of Melting Tests Modeling Results Fig. S1. Garnet Lherzolite Melting Fig. S2. Spinel Lherzolite Melting Fig. S3. Garnet Pyroxenite Melting Fig. S4. Deepening solidus effect on excess 230Th

2. Model Input Parameters and Results Table S1. Mineral-Melt Distribution Coefficients Table S2. Bulk Partition Coefficients Table S3. Results for Garnet Peridotite Table S4. Results for Spinel Peridotite Table S5. Results for Garnet Pyroxenite

3. Melt mixing from pyroxenite and peridotite lithologies Fig. S5. Occurrence of E-MORB and N-MORB throughout the SEIR study area Fig. S6. Example Mixing Trajectories for Pyroxenite and Peridotite Melts

4. SEIR sample characteristics Table S6. Sample characteristics and aliquot weights Table S7. Results of Th isotope ratio replicate analyses

1. Melting Models Description and Conditions of Melting Tests The extent to which different melting conditions (melt initiation depth, lithology and melt

productivity) are responsible for the observed variations in 226Ra and 230Th-excesses in SEIR

basalts was investigated using two different one-dimensional melting models: dynamic melting

(e.g., McKenzie, 1985, Williams and Gill, 1989) and equilibrium porous flow melting (hereafter

referred to as EPF, for simplicity; Spiegelman and Elliott, 1993). We used analytical solutions

from Williams and Gill (1989) in a spreadsheet for the dynamic melting simulations and

numerical solutions from the UserCalc website for the EPF (Spiegelman, 2000). Results of the

model simulations for different conditions are compiled here (tables with input parameters and

results appear in a group near the end of this discussion).

Melting tests were run for three different lithologies: 1) garnet lherzolite (gnt lz), 2)

spinel lherzolite (sp lz), and 3) garnet pyroxenite (pyx). Mineral-melt partition coefficients (D)

for U, Th and Ra are summarized in Tables S1 and S2; ‘deep’ mineral-melt partition coefficients

apply to the gnt lz and pyx assemblages, and ‘shallow’ ones apply to the sp lz lithology.

Individual mineral-melt partition coefficients were chosen from literature values and D inter-

relationships discussed in Blundy and Wood (2003). Calculated bulk partition coefficients and

mineral modes (which were held constant for a given lithology) are given in Table S2.

Supplement to Russo et al., 2008, page 2

Tables S3, S4 and S5 summarize results of all the model simulations discussed in the

text, using alpha-numeric identifiers to distinguish different source lithologies and melting

conditions used in dynamic and EPF models. These three tables have results for melting of the

three lithologic types (models 1 through 3) and melting style (i.e., dynamic melting runs are

labeled a through c and EPF runs are labeled d through f). Predicted 226Ra and 230Th-excesses

using a range of M and residual porosity (φ) and that meet the melt duration criteria were

combined to make the grids in figures S1 through S3 (see the next section). Values of specific

model runs discussed in the text are also presented in Tables S3 (models 1a-1f) for the gnt lz

models, S4 (models 2a-2f) for the sp lz models, and S5 (models 3a-3f) for the pyroxenite models.

Additional conditions of the model tests were: Models 1 (gnt lz) and 2 (sp lz) assumed

the entire mantle that is melting consists of lherzolite (lz), bearing spinel (sp) at depths shallower

than 25 kbar (~75 km) with a sharp transition to the garnet (gnt) bearing lithology at greater

depths. Pyroxenite (pyx) models were run at the same upwelling rate as the peridotite models to

mimic melting behavior of both materials entrained in the same mantle upwelling flow field. The

high modal clinopyroxene content of this enriched lithology melts sufficiently differently (i.e.,

faster and to higher melting degree relative to the lz assemblages; Hirschmann and Stolper, 1996,

Elkins et al., 2007) so as to have a dramatic effect on 230Th-excess and 226Ra-excess in the melt.

Lz dynamic melt simulations ranged from 5 to 7% partial melting (F=0.05 to 0.07) and

10 to 15% (F=0.1 to 0.15), for sp lz and gnt lz lithologies respectively, where F is the degree of

melting. In dynamic melting, user defined degree of melting (X) and melting rate (M) determine

a timescale for the duration of melting, which we set to ≤2.0 Myr for sp lz models (e.g.,

assuming an ~70 km tall melting column, and a solid upwelling velocity of 0.036 m/yr, which is

the mean SEIR half-spreading rate). Gnt lz melt simulations were solved using the same criteria,

with the exception that melt duration was set to ≤2.6 Myr (e.g., an ~95 km tall melting column

and a solid upwelling velocity of 0.036 m/yr).

Melting rate per se is not an input parameter of the EPF model. Instead, simulations are

based on user-defined depths of initial melting and melt cessation, plus the amount (fraction) of

melt present at each depth. A transition depth between lithologies can be defined, although we

did not use a transition depth here so that the results would be directly comparable to the

dynamic melt tests. Our inputs for melt column height and porosity were determined according

to the relationship W~(hM/Xρmelt)

Supplement to Russo et al., 2008, page 3

where W is upwelling rate, h the height of the melt column, X the degree of melting and ρmelt is

the density of the melt (Asmerom, Y., 1999, EPSL, vol. 166). In ridge systems not influenced by

hotspots it is reasonable to assume the mantle upwelling rate is constant and equal to the half-

spreading rate, in which case M can be calculated from X or vice-versa. This fact can be used in

conjunction with the melt duration to determine reasonable limits on other variables such as X by

assuming an initial depth of melting (see Section 5.3 of the paper). One exceptional attribute of

the SEIR study area is that crustal thickness variations at near constant spreading rate give a

sense for how X varies along axis. This additional control provides a context in which to

evaluate how variations in M and porosity generate the observed disequilibria along the SEIR.

For instance, such constraints weigh into our conclusion regarding increased involvement of a

pyroxenitic source component beneath the eastern SEIR to form melts there (where crust is thin

but 230Th excesses are low; as discussed in Section 5 of the paper). Another example of how this

relationship weighs into our modeling is in the western portion of the study area (where crust is

thick and 230Th excesses are low): for a maximum spinel lherzolite melting column height of 70

km, M = 1.0x10-4 (i.e., the maximum value to produce ~7% 230Th excess in our models), and W

appropriate for the SEIR, 4.8 km of basaltic crust (i.e., F of 6.9% times 70 km) is produced if

100% of the melt produced on this one-dimensional mantle column is transferred to the crust

(which is probably an over-estimate). This generous maximum crustal production value is ≤75%

of the mean crustal thickness of ≥6.3 km for the western sector of the SEIR, implying that

melting of spinel lherzolite is incapable of producing our average WG basalt.

Modeling Results

Applicable ranges of 226Ra and 230Th-excess in the different melting tests were

determined by realistic bounds on melting conditions and measured values in the SEIR data set.

In general, EPF simulations result in much higher 226Ra and 230Th-excesses than dynamic runs at

the same conditions when melting rate is slow and porosity is low (the upper right corner of the

melting grids; Fig. S1, S2 and S3); both models return similar results when M is fast and φ is

high (the lower left corner of the melting grids).

Garnet Lherzolite Melting

Model output for melting of garnet lherzolite are given in Fig. S1. Conditions are

described in the figure caption. Both dynamic and EPF gnt lz simulations were fairly insensitive

to F over the range of M, from 1x10-4 to 1.1x10-3, when compared to the sp lz simulations, for

Supplement to Russo et al., 2008, page 4

which conditions of lower M at higher F do not fit the established melt duration criteria.

Figure S1 Garnet lherzolite results for dynamic (left panel) and EPF (right panel) model simulations. SEIR data are displayed on the dynamic grid using the same symbols as fig. 4 of the main text; red squares (Western Group), green circles (Central Group) and blue triangles (Eastern Group). Open green circles are CG data that have been corrected for 1 226Ra half-life of decay, as discussed in the text. In the EPF panel the two grids are for tests with different final F (10%, black grid; 15%, red grid) demonstrating the relative insensitivity of these models to F at these relatively high degrees of melting. Garnet lherzolite EPF simulations were assumed to begin melting at 40 kbar. Degrees of melting ranged from 10 to 15% and M ranged from 2.0x10-

4 to 1.1x10-3; column height was adjusted in the manner described in the previous section of this supplement, and porosity (φ) varied from 0.05 to 1.0%.

Spinel Lherzolite Melting

Spinel lherzolite melting results are in Fig. S2, with conditions described in the caption.

Figure S2 Model results for dynamic (left panel) and EPF (right panel) spinel lherzolite simulations. EPF simulations were assumed to begin melting at 25 kbar. Degree of melting and melting rate for the EPF runs were set to match those used in the corresponding dynamic simulations (F from 5% to 7%, M from 8.0x10-5 to 1.0x10-3 and φ from 0.02 to 1.0%). These values were then used in combination with the above equation for W to determine the appropriate column height.

Supplement to Russo et al., 2008, page 5

Garnet Pyroxenite Melting

Dynamic and EPF simulations for pyroxenite melting are given in Fig. S3. Models were

run using experimentally determined melting rates (0.4 to 2.0x10-3 kg m-3 yr-1) and F (25%), both

of which are higher than the corresponding lherzolite simulations In addition, porosity (φ) was

also increased in these model runs and was varied from 0.2 to 3.0%, similar to values required by

Kogiso et al. (2004) for pyroxenite melts to survive in a dominantly peridotite matrix.

Figure S3 Model results for dynamic (left panel) and EPF (right panel) garnet pyroxenite simulations.

Solidus depth

EPF tests were run to estimate the effect of a deepening solidus on (230Th/238U). In these

models the melting rate was held constant at 1.3x10-4 kg m-3 yr-1 and melting was assumed to

continue to the base of the lithosphere. The initial depth of melting was varied from 25 to 40

kbar while holding all other variables constant. All melt generated below 25 kbar used garnet

lherzolite partition coefficients, and all melting above 25 kbar used deep spinel lherzolite

partition coefficients (i.e., a transition depth of 0 km). The results are summarized in Figure S4

below. Note that the predicted 226Ra-230Th disequilibria are effectively the same in melts from all

four initiation depths.

Figure S4 EPF model results for (230Th/238U) at increasing depths of melt initiation.

Supplement to Russo et al., 2008, page 6

2. Model Input Parameters and Results.

Table S1 - Mineral-Melt Distribution Coefficients Mineral U Th Ra

Deep Shallow Deep Shallow Deep Shallow olivine 0.00038 0.00038 0.00005 0.00005 0 0 orthopyroxene 0.001 0.002 0.002 0.002 0.00001 0.00001 clinopyroxene 0.016 0.009 0.015 0.0065 0.0003 0.0003 garnet 0.025 0.009 0.00001 spinel 0 0 0

Table S2 - Bulk partition coefficients

Lithology Bulk partition

coefficient U Th Ra

Spinel Peridotite 0.0021 0.0016 0.00005 (ol/sp=0.6, opx=0.2, cpx=0.2)

Garnet Peridotite 0.0053 0.0036 0.00005

(ol=0.55, opx=0.22, cpx=0.13, gar=0.1)

Pyroxenite 0.0110 0.0050 0.00014

(opx=0.45, cpx=0.45, gar=0.1)

Supplement to Russo et al., 2008, page 7

Table S3 Dynamic Melting Equilibrium Porous Flow

model 1 1a 1b 1c 1d 1e 1f

Garnet Peridotite M=2.0x10-4 M=7.0x10-4 M=1.0x10-3 M=2.0x10-4 M=7.0x10-4 M=1.0x10-3

(226Ra/230Th) (230Th/238U) (226Ra/230Th) (230Th/238U) (226Ra/230Th) (230Th/238U) (226Ra/230Th) (230Th/238U) (226Ra/230Th) (230Th/238U) (226Ra/230Th) (230Th/238U)

φ= 1.0 1.404 1.090 1.387 1.048 1.378 1.038 1.414 1.150 1.422 1.099 1.425 1.079

0.5 1.794 1.117 1.733 1.055 1.702 1.041 1.831 1.225 1.868 1.117 1.886 1.090

0.1 4.397 1.153 3.534 1.061 3.200 1.045 5.102 1.346 5.944 1.135 5.808 1.101

Table S4 Dynamic Melting Equilibrium Porous Flow

model 2 2a 2b 2c 2d 2e 2f

Spinel Peridotite M=1.0x10-4 M=6.0x10-4 M=1.0x10-3 M=1.0x10-4 M=6.0x10-4 M=1.0x10-3

(226Ra/230Th) (230Th/238U) (226Ra/230Th) (230Th/238U) (226Ra/230Th) (230Th/238U) (226Ra/230Th) (230Th/238U) (226Ra/230Th) (230Th/238U) (226Ra/230Th) (230Th/238U)

φ= 1.0 1.190 1.029 1.182 1.013 1.176 1.009 1.371 1.008 1.369 1.005 1.367 0.994

0.5 1.376 1.043 1.347 1.015 1.327 1.010 1.746 1.021 1.753 1.018 1.739 1.007

0.1 2.697 1.068 2.245 1.018 2.026 1.011 4.806 1.048 4.714 1.044 4.804 1.033

Table S5 Dynamic Melting Equilibrium Porous Flow

model 3 3a 3b 3c 3d 3e 3f

Pyroxenite M=7.0x10-4 M=1.0x10-3 M=2.0x10-3 M=7.0x10-4 M=1.0x10-3 M=2.0x10-3

(226Ra/230Th) (230Th/238U) (226Ra/230Th) (230Th/238U) (226Ra/230Th) (230Th/238U) (226Ra/230Th) (230Th/238U) (226Ra/230Th) (230Th/238U) (226Ra/230Th) (230Th/238U)

φ= 3.0 1.179 1.108 1.177 1.091 1.171 1.059 1.186 1.198 1.186 1.173 1.191 1.150

2.0 1.267 1.129 1.263 1.106 1.252 1.066 1.282 1.251 1.283 1.210 1.295 1.180 1.0 1.520 1.161 1.507 1.127 1.470 1.074 1.572 1.327 1.577 1.260 1.633 1.222

Supplement to Russo et al., 2008, page 8

3. E-MORB and N-MORB distribution in the study area

The following two diagrams illustrate that (230Th/238U) and (230Th/232Th) variations in the SEIR

study area are not controlled by changes in the distribution of MORB subtypes (i.e., E-MORBs

and N-MORBs) within our geographical groups. There is no systematic difference in 230Th-238U

disequilibrium between E-MORBs and N-MORBs in our SEIR study area.

Fig. S5: Occurrence of E-MORB and N-MORB throughout the study area

Figure S5a. Equiline diagram showing E-MORB (open circles) and N-MORB (closed squares) in the SEIR data set. E-MORB and N-MORB occur in each geographic group (see table 1 in the main text) and are distributed across the most enriched and intermediate samples identified in Figure 3 of the manuscript. There is no systematic difference in extent of 230Th-238U disequilibrium between EMORB and NMORB in the study area.

Figure S5b. (230Th/232Th) in E-MORB (open circles) and N-MORB (closed squares) distributed along the SEIR. E-MORB and N-MORB are present in each of the three geographic groups (i.e. west of 100°E, 100°-114°E and east of 114°E) and both record a west-to-east decrease in (230Th/232Th).

Supplement to Russo et al., 2008, page 9

4. Melt Mixing from Peridotite and Pyroxenite Lithologies

Mixtures between melts of a single pyroxenite and two slightly different peridotite lithologies were calculated to examine likely fractions of each melt type that are consistent with our SEIR basalt observations. End member source compositions are internally consistent in Figs S6a and S6b (e.g., (238U/232Th) = 0.756 in the Gt Px source is equivalent to Th/U=4). Lower (238U/232Th) values for the Gt Px source would reduce the pyroxenite melt fractions that reproduce our sample data, so our estimates are likely maxima. We have not attempted to correct for the small anticipated reduction in peridotite melting from heat loss to the pyroxenite because the solid pyroxenite fraction is so small in our model (<4%).

Fig. S6: Example Mixing Trajectories for Pyroxenite and Peridotite Melts Figure S6a. Equiline diagram of modeled mixtures between MORB and Gt Px melts (shown as red and yellow lines). Percentage labels indicate the volume of Gt Px melt in the mixture (see also paper section 5.5). Open symbols are SEIR data. Mixtures were calculated from one Gt Px melt (black solid line, from a source having Th/U=4.0, [Th]=6.64 ppm and [U]=1.66 ppm) and two different N-MORB melt endmembers (filled circles). MORB#1 has Th/U=2.65, [Th]= 130 ppb and [U]= 49 ppb and MORB#2 has Th/U=2.35, [Th]=49 ppb and [U]=21.1 ppb. The solid black line represents a family of possible (230Th/238U) in Gt Px melts from one. We used the Gt Px melt composition was from the dynamic Gt Px model run at M=1.2 x 10-3 and φ= 2.5% in our mixing trajectories since this best fits the observations. Figure S6b. κ vs. κPb diagram of modeled mixtures between MORB and Gt Px melts (mixing trajectories are shown as red and yellow lines). Percentage labels indicate the fraction of Gt Px melt in the mixture (see text for further discussion). Open symbols are SEIR data. Mixtures were calculated from one Gt Px melt and two different end member N-MORB melt compositions (filled circles). N-MORB composition #1 has κ=2.65 and κPb=3.95, and N-MORB composition #2 has κ=2.35 and κPb=3.78. The Gt Px melt end member is indicated by the point labeled κ=κPb=4.0.

Supplement to Russo et al., 2008, page 10

4. Sample characteristics and aliquot weights

Table S6. Sample characteristics and aliquot weights

Sample Glass

Condition MnO coating sample mass

(g) ID-split mass

(g) IC-split mass

(g)

WW10-71-1 fresh none 1.66115 0.04542 1.61573 WW10-75-4* fresh none 1.27432 0.04392 1.23040 WW10-78-2 fresh none 1.53021 0.05340 1.47681 WW10-84-7 fresh none 1.28260 0.04160 1.24100 WW10-89-4 fresh none 0.49946 0.02215 0.47731 WW10-96-1 fresh none 0.65696 0.02480 0.63216

WW10-100-1* fresh none 0.56032 0.01052 0.54980 WW10-105-1* fresh none 1.14400 0.01998 1.12402 WW10-113-7* fresh none 1.02997 0.02124 1.00873 WW10-118-1 fresh none 0.35511 0.00646 0.34865 WW10-125-1 fresh none 1.44358 0.04313 1.40045 WW10-126-7 fresh none 0.23360 0.00625 0.22735 WW10-132-1 fresh none 0.96624 0.03070 0.93554 WW10-141-1 fresh none 0.85679 0.02681 0.82998 WW10-144-4 fresh none 0.85524 0.03069 0.82455 WW10-145-1 fresh none 0.56107 0.01086 0.55021

* These samples were not analyzed for 226Ra due to analytical difficulties; their measured 230Th excesses match those of similar samples with radium age control.

Table S7. Results of Th isotope ratio replicate analyses

232Th/230Th ± (230Th/232Th) ± replicate precision

2σ error (230Th/238U) ± (226Ra/230Th) ±

100-1 mean 164000 2000 1.129 0.014 0.7 1.2 1.210 0.016 100-1a 165200 1600 1.121 0.011 1.219 0.018 100-1b 162800 2300 1.137 0.016 1.202 0.013 105-1 mean 161300 1500 1.148 0.011 0.9 0.9 1.197 0.012 105-1a 162800 1600 1.137 0.011 1.208 0.012 105-1b 159700 1400 1.159 0.010 1.186 0.013 113-7 mean 167400 1700 1.106 0.011 0.4 1.0 1.190 0.013 113-7a 166700 1500 1.11 0.010 1.185 0.014 113-7b 168000 1800 1.102 0.012 1.194 0.012 132-1 mean 188700 2900 0.981 0.015 0.3 1.5 1.108 0.018 1.87 0.05 132-1a 189200 2800 0.978 0.014 1.112 0.018 1.87 0.05 132-1b 188100 3000 0.984 0.015 1.105 0.017 1.88 0.05 145-1 mean 210700 2600 0.8785 0.011 1.0 1.2 1.094 0.014 1.32 0.02 145-1a 212900 2400 0.869 0.010 1.106 0.015 1.28 0.02 145-1b 208500 2700 0.888 0.011 1.083 0.013 1.37 0.02 Results of repeat analyses for 5 samples. Replication precision exceeds the analytical error in each case. Values of other parameters that depend on the 232Th/230Th ratio also appear in this table. Mean values (in bold face font) are appear in Table 1 of the paper. all errors are 2σ