Influence of subducted components on back-arc melting dynamics in the Manus Basin



Lavas erupted in back-arc basins afford the opportunity to explore the extent to which decompression and subduction-related components influence partial melting in this setting. We present U-Th-Ra disequilibria data from 24 well-characterized lavas from the Manus Basin behind the New Britain volcanic arc, supplemented by some additional trace element and Sr-Nd-Pb isotope data. The lavas range in composition from 49.6 to 57.7 wt % SiO2 and can be subdivided into those that are broadly like mid-ocean ridge basalts (MORB) with Ba/Nb < 16 and back-arc basin basalts (BABB) that are variably influenced by subduction components and have Ba/Nb > 16. Rifts closest to the arc are dominated by BABB, whereas both lavas types erupt further away at the Manus Spreading Center. The MORB have small 230Th excesses (up to 5%) and are displaced below the global correlation of (230Th/238U) with ridge depth. In most respects the BABB closely resemble lavas erupted along the New Britain arc front, including 238U excesses that reach 26%. The Pb isotope data can be explained by mixing of a subduction component into an Indian MORB mantle source. The Pb in the subduction component is derived from both the subducted sediment (5%) and fluids from the subducting altered Solomon Sea oceanic crust (95%), and these were mixed prior to addition to the mantle wedge. U/Th ratios, Fe3+/ΣFe, and H2O contents all increase with increasing 206Pb/204Pb. A model in which addition of the subduction component to the mantle wedge is followed by 230Th in-growth during decompression and dynamic melting all less than 140 kyr prior to eruption can simulate the data. However, our preferred model is one of dynamic decompression melting in which subduction-modified, more oxidized mantle had DU ≪ DTh leading to 238U excesses in contrast to unmodified mantle that yields 230Th excess. Large 226Ra excesses in some southern rift samples require addition of a fluid <8 kyr ago but elsewhere reflect melting under low-porosity conditions.

1. Introduction

Numerous studies have investigated the trace element and isotope heterogeneity in mid-ocean ridge basalts (MORB) and ocean island basalts (OIB) and the insights these provide into the dynamics of the Earth's mantle [e.g., Albarede and van der Hilst, 2002; Hart et al., 1992; Hofmann, 1997, 2003; Schilling, 1975; Schilling et al., 1983; Stracke et al., 2003, 2005; White and Hofmann, 1982; Zindler and Hart, 1986]. Much debate has centered on the extent to which such heterogeneities impact melting dynamics via differences in mineralogy and/or volatile contents with respect to normal peridotite [e.g., Maclennan, 2008; Stracke and Bourdon, 2009]. A substantial amount of heterogeneity is developed at convergent plate margins where sediment and hydrothermally altered oceanic crust are possible subducted components [Bercovici and Karato, 2003; Schubert et al., 2001; Stein and Hansen, 2008; van der Hilst et al., 1998; Hawkesworth et al., 1993]. Back-arc basin basalts (BABB) provide an ideal setting in which to assess the extent to which subducted components extend and influence melting beyond the arc.

Both decompression melting and flux melting are thought to play active roles in the genesis of back-arc magma, and these two styles of melting have fundamentally different relationships between water, fluid mobile trace elements, and percent melting [Kelley et al., 2005, 2006; Langmuir et al., 2006; Taylor and Martinez, 2003]. Because U series disequilibria also differ fundamentally between these two melting styles (see reviews by Lundstrom [2003] and Turner et al. [2003]), U series characteristics in back-arc magmas may tell much about how these different melting processes are distributed and interact. To date, only two such studies have been conducted, one in the Lau Basin [Peate et al., 2001] and one in the Scotia back arc [Fretzdorff et al., 2003]. However, the Manus Basin is thought to have the highest mantle potential temperature and, therefore, the greatest decompression component of all back-arc basins [Kelley et al., 2006], and its principal spreading center (Manus Spreading Center, MSC) lies farther behind the arc (∼275 km) than in other back-arc basins (Lau Basin < 250 km [Peate et al., 2001] and Scotia back arc ∼150 km [Fretzdorff et al., 2003]). Thus, if recent slab components occur even in MSC magmas, then a component of flux melting may be ubiquitous in all back arcs.

In this paper we present U-Th-Ra isotopic analyses of samples from the Manus back-arc basin in the Bismarck Sea, Papua New Guinea. We combine these with previously published major element data and both previously published and new trace element and Sr-Nd-Pb isotope data and compare these with published data from the arc front volcanoes to investigate the influence of mantle heterogeneities on mantle melting.

2. Geological Setting of the Manus Basin

The Manus Basin in the Bismarck Sea north of Papua New Guinea is a rapidly opening back-arc basin (Figure 1). It is bordered by the New Britain arc (south), New Ireland (east), the Manus Trench/New Hanover (north) and Papua New Guinea (west). Until <10 Ma ago the Caroline Plate to the north was subducted southward along the Manus Trench (Figure 1) [Cooper and Taylor, 1987; Falvey and Prichard, 1985]. About 10 Ma ago, the subduction direction reversed as a result of the collision of the Ontong Java Plateau with New Ireland and the North Solomon Arc [Cooper and Taylor, 1987] and since then the southerly Solomon Sea plate has been subducted in a northward direction along the New Britain trench at a rate of 15.4 cm yr−1 [Lee and Ruellan, 2006]. The Solomon Sea plate has maximum lithospheric ages of 24–44 Ma and thus represents relatively young hot oceanic lithosphere [Joshima and Honza, 1987] and may contain less pelagic sediment than other subducted plates in the Pacific area (see discussion below). Since <4 Ma New Britain and eastern Papua New Guinea have acted as a single rotating unit during the opening of the Bismarck Sea [Martinez and Taylor, 1996]. Spreading along the Extensional Transform Zone (EZT) and the Manus Spreading Centre (MSC), which are ∼ 275 km from the arc, started < 3.5 Ma ago simultaneously with rifting only ∼150 km behind the arc along the South and East Rifts (SR, ER; Figure 1) [Taylor, 1979]. Seismicity in the subducting plate reaches 600 km and underlies the ER and SR (at 350 km, Figure 1), but not the MSC and ETZ [Hall and Spakman, 2002]. Spreading rates in the Manus Basin decrease from a maximum of < 92 mm yr−1 in the southeast, approaching zero close to Manus Island in the northwest [Taylor, 1979; Woodhead et al., 1998]. Lavas from the New Britain arc have been the subject of detailed geochemical and U series studies by Woodhead and Johnson [1993], Woodhead et al. [1998], and Gill et al. [1993] while Sinton et al. [2003] presented a detailed geochemical study of the Manus Basin lavas.

Figure 1.

Map of the Manus Basin back-arc spreading centers using GMT [Wessel and Smith, 1991, 1998]. ETZ, Extensional transform zone; MSC, Manus spreading center. Sample numbers (e.g., 15-4) correspond to those of Sinton et al. [2003] also listed in Table 2, and the samples have been subdivided into MORB with Ba/Nb < 16 (indicated as black circles) and BABB with Ba/Nb > 16 (indicated as open circles). Dotted lines mark depth of Benioff Zone by Cooper and Taylor [1987].

3. Sample Selection and Analytical Methods

Manus Basin glass samples obtained by dredging were selected from localities along the Extensional Transform Zone and Manus Spreading Centre and the South and East Rifts (see Figure 1). For many of these, major and trace element concentrations along with Sr, Nd and Pb isotope ratios were presented by Sinton et al. [2003]. However, to ensure a more complete data set, additional trace element and/or radiogenic isotope analyses were acquired following methods described below.

Suitable glasses were crushed, washed in deionized water and hand-picked to avoid grains with visible evidence for hydrothermal alteration and/or Mn crusts. Samples were then ultrasonicated in 2.5N HCl:H2O2 (1:1 mix) for 20 min before being washed and ultrasonicated (20 min) in deionized water. The BBQ flow (sample 2392-9) was processed with three different leaching methods (no leaching, 20 min and 40 min leaching, see below and Table 1). Glasses were then reinspected for alteration and hand-picked a second time to obtain a clean set of glass chips. Trace elements were analyzed on an AgilentTM ICP-MS following standard techniques [Eggins et al., 1997]. Trace element data for samples 15-4, 18-4, 22-2, 24-9, 25-2, 30-2, 34-1, 35-7, 40-1, 42-1, 45-1 and 47-1 were performed by LA-ICP-MS at Macquarie University following the method described by Eggins [2003] and data reduction following Norman et al. [1996] with the exception of Ba for sample 15-4 that was taken from Sinton et al. [2003]. U-Th concentration data for all samples are from isotope dilution (see below).

Table 1. U Series Results for Rock Standards and the BBQ Flow
MaterialU (ppm)Th (ppm)(234U/238U)SE(230Th/232Th)SE(238U/232Th)SE(230Th/238U)SE226Ra (fg/g)(226Ra/230Th)SE
TML (n=5 for U-Th; n=4 for Ra-Th)10.31529.0341.0010.0041.0820.0031.0780.0031.0040.0053532.80.9920.006
BCR-1 (n=4)1.6675.6971.0010.0010.8790.0020.8880.0010.9900.003563.01.0090.006
BCR-2 (n=2)1.6655.7471.0000.0010.8760.0020.8790.0010.9960.003546.10.9740.038
2392-9 (no leach)0.0480.1191.0010.0021.3360.0031.2210.0011.0950.00348.92.7560.012
2392-9 (20 min leach)0.0470.1170.9960.0021.3460.0041.2170.0011.1060.00448.92.7760.011
2392-9 (40 min leach)0.0470.1200.9980.0021.3160.0031.1950.0011.1010.00448.82.7700.011

Typically, 0.5–1 g of samples were spiked with 236U−229Th and 228Ra tracers and dissolved in a HNO3-HCl-HF mix. Separation of U, Th and Ra followed methods described by Heyworth et al. [2007] and Turner et al. [2007]. Th and U were analyzed on a Nu InstrumentsTM MC-ICP-MS at Macquarie University following the approach described by Dosseto et al. [2006], Heyworth et al. [2007], and Sims et al. [2008]. Ra analyses were performed on a ThermoFinnigan TritonTM TIMS at Macquarie University following the procedures described by Turner et al. [2000].

Table 1 presents results for a number of equilibrium rock standards and a MORB glass supplied by M. Perfit from the BBQ flow (sample 2392-9) from the East Pacific Rise [Rubin et al., 1994] that were analyzed over the same 2 year period (2007–2008) as the Manus samples. The analyses of TML-3, BCR-1 and BCR-2 are generally within error of published values and within 2 standard errors of secular equilibrium for (230Th/238U) and (226Ra/230Th). Exceptions are (230Th/238U) for BCR-1 that deviates by 4‰ outside of 2 standard errors and (226Ra/230Th) for TML-3 that deviates by 2% from equilibrium. Accordingly, we estimate our precision during this period to be ≤ 5 ‰ for (230Th/238U) and ≤ 3% for (226Ra/230Th). U, Th and 226Ra concentrations from the BBQ flow are in good agreement with Rubin et al. [2005] and Sims et al. [2002] although we obtained a 3% lower (230Th/232Th) ratio that results in lower (230Th/238U) and (226Ra/230Th) ratios. Because, this is the opposite from the effect of seawater alteration or manganese crusts, it may reflect overcorrection for the 232Th tail on 230Th. Note, however, that both TML and BCR-2 are within ≤ 5 ‰ for (230Th/232Th) compared to the values reported by Sims et al. [2008]. A tail effect on (230Th/232Th) would lower the 230Th excesses of the MORB samples and increase the 238U excesses in the BABB, but not change any of the primary conclusions made below.

Sr and Nd isotope cuts were taken from the first Ra cationic columns and were prepared and analyzed at Macquarie University following methods described by Le Roux et al. [2009] and Heyworth et al. [2007]. Sr and Nd isotope ratios were obtained in static mode and corrected for instrumental mass bias to 86Sr/88Sr = 0.1194 and 146Nd/144Nd = 0.7219. Analyses of NIST SRM-987 gave 0.710210 ± 37 (2SD, n = 14) and BHVO-2 gave 0.703490 ± 52 (2 SD, n = 15) while the JMC Nd standard gave 0.511106 ± 2 (2SD, n = 8) and BHVO-2 yielded 0.512967 ± 6 (2SD, n = 17). The Sr data from both Sinton et al. [2003] and our new data were normalized to 0.710250 and all Nd isotope data were normalized to a La Jolla reference value of 0.511849.

Pb isotopes on hand picked glass chips were analyzed at the University of Melbourne following the double spike procedures described in detail by Woodhead et al. [1998]. Standard runs of NIST-SRM-981 (n = 5) gave average values of 16.9356 ± 27, 15.4891 ± 32 and 36.7006 ± 116 for 206Pb/204Pb, 207Pb/204Pb and 208Pb/204Pb, respectively. Blanks for Sr, Nd and Pb were generally lower than 0.1 ng, 0.2 ng and <100 pg, respectively.

4. Results

For completeness, the full data set for the samples analyzed for U series isotopes are presented in Table 2 including results from Sinton et al. [2003] and Shaw et al. [2004] (plain = legacy data, bold and italic = new measurements).

Table 2. Manus Basin Geochemical Dataa
SampleLocalityLatitude (°S)Longitude (°E)Depth (m)SiO2TiO2Al2O3FeOTFe2O3FeOMnOMgOCaONa2OK2OP2O5SO3H2O+H2OH2O (Sinton)H2O (Shaw)TotalLiBeScVCrCoNiCuZnGaRbSrYZrNbCsBaLaCePrNdSmEuGdTbDyHoErYbLuHfTaPbUTh87Sr/86Sr143Nd/144Nd206Pb/204Pb207Pb/204Pb208Pb/204Pb(234U/238U)(230Th/232Th)(238U/232Th)(230Th/238U)226Ra (fg/g)(226Ra/230Th)
  • a

    Settings: ER, East Manus Rifts; SR, Southern Rifts; ETZ, Extensional Transform Zone; MSC, Manus Spreading Center. Types: A, arc; E, enriched; B, BAAB; X, XBABB; M, MORB; subtypes discussed by Sinton et al. [2003]. H2O+ and H2O are from Sinton et al. [2003], and H2O data are for glass [Shaw et al., 2004]. Major elements are whole rock data [Sinton et al., 2003]. Trace elements and isotope data in normal type are for glasses by ICP-MS when Li is included or for whole rocks by XRF and INAA when not [Sinton et al., 2003]. Trace element data from 15-4, 18-4, 22-2, 24-9, 25-2, 30-2, 34-1, 35-7, 40-1, 42-1, 45-1, and 47-1 are LA-ICP-MS data on glass. U-Th concentrations are measured by isotope dilution for all samples. Values in bold and italics are our own new measurements. BCR-2 run during measurement period [Cunningham et al., 2009]. Errors are 1 SD.

   31-3ETZ (M1)3.502149.258216050.401.1913.8012.222.1510.290.207.0011.612.490. 99.723.270.35547.236149.650.956.814690.515.50.44268.433.054.50.7640.0064.671.425.091.015.762.360.8673.520.6374.400.9922.942.760.4171.500.0490.2500.0190.0690.7033350.51305518.00315.46137.8691.0120.8790.8391.04726.9621.249
   32-5ETZ (M1)3.420149.150236049.630.9814.849.951.548.560.178.1412.362.330.
   33-3ETZ (M1)3.528149.478210250.690.9714.3810.491.489.160.197.6012.412.
   31-1ETZ (E)3.502149.258216050.241.4614.3011.152.468.940.196.9411.852.650.
   24-1MSC (M1)3.545149.857217050.291.3813.6112.401.9110.680.206.7311.452.460. 99.446.900.46746.941517447.665.810989.516.90.84377.736.672.11.450.0149.132.157.421.358.
   25-2MSC (M1)3.608149.797223550.180.9514.989.941.788.340.178.2312.582.300. 100.044.690.71046.630125443.687.999.865.012.80.6079.
   39-1MSC (M1)3.290150.128232750.461.2313.7011.982.0010.180.206.9911.462.420.
   45-1MSC (M2)3.065150.455262050.601.1813.7111.691.8410.030.207.1611.632. 98.728.420.39748.642325.947.547.214310515.90.78581.035.577.71.31-9.402.327.611.559.353.331.244.740.8266.181.394.144.090.6232.400.0840.3370.0280.0880.7032990.51304218.08515.47037.9290.9990.9800.9681.01312.1091.260
   47-1MSC (M2)3.077150.563255250.671.7213.0114.442.9411.790.375.4310.202.720. 99.9110.40.71047.850235.853.648.410313618.51.3284.548.11091.58-15.33.5611.51.9912.04.781.686.
   42-1MSC (M2)3.160150.285248551.381.2713.9112.022.739.560.206.2710.872.610.
   28-1MSC (B)3.657149.673240550.381.0514.3510.622.188.660.187.4811.562.
   40-1MSC (B)3.248150.283230551.511.0714.7210.652.578.340.186.0910.792.630. 99.476.00-40.736657.638.949.711674.414.91.9615925.554.80.828-31.52.797.651.257.812.630.8693.370.6054.401.002.923.060.4191.550.0890.6000.0780.1740.7034840.51307818.54015.51138.2120.9971.2811.3640.93927.0791.085
   30-2ETZ (B)3.538149.408193053.011.0515.4110. 99.707.460.60038.83553.436.619.067.783.816.22.3018126.967.81.560.20340.04.2310.41.608.082.550.9313.240.5834.521.012.793.030.4751.900.1380.9630.1810.3730.7033050.513066---1.0041.3731.4750.93159.7041.041
   34-1ETZ (B)3.607149.732247750.020.8814.7310.141.748.570.187.7512.
   35-4MSC (X1)3.503149.897210754.410.8815.898.042.515.780.145.319.473.390.580. 99.864.290.79732.628376.332.549.711361.515.56.8747621.41084.310.27516917.233.14.0314.82.960.9443.040.4903.030.6551.911.830.2802.110.2252.730.9382.3000.7034390.51307118.74215.52538.3071.0171.0461.2380.845286.6191.064
   24-9MSC (X3)3.545149.857217057.720.9815.287.691.676.190.133.597.333.890.640.200.031.760.381.831.9499.629.990.82528.727417.523.824.784.259.317.57.4927119.91014.410.30416712.623.92.9911.42.650.9503.080.5753.610.7352.152.190.4022.380.2402.250.9562.2540.7032980.51304018.67415.51538.2440.9931.1441.2870.888303.4081.051
   35-7MSC (X3)3.503149.897210756.720.9115.677.411.785.810.124.097.943.720.660.200.021.430.65- 99.549.780.58025.925716.422.928.967.051.918.77.6434618.886.43.740.26016712.022.62.8111.02.400.8452.830.4553.080.6722.072.090.3262.070.1962.471.0862.2130.7034400.51307418.71715.52038.2791.0061.1851.4890.796341.1091.162
   19-12SR (Bs)3.752151.158263052.750.8715.289.781.988.000.165.8410.342.650.
   20-3SR (Bs)3.813150.693206251.310.5916.057.491.116.490.147.7912.792. 100.002.380.26241.824336035.311110849.312.22.6416317.828.00.3370.08437.81.574.130.7283.771.400.5422.010.3592.460.5551.651.580.2410.8120.0220.8440.0850.1200.7035080.51309418.67815.52438.2860.9891.7712.1360.82956.5292.367
   22-2SR (Bs)3.858150.430255549.880.5516.447.642.025.820.148.6312.572. 99.493.630.22436.723123936.710711373.414.52.1527112.723.40.5000.04842.01.994.840.7894.361.470.6461.870.3382.280.5041.431.470.2270.7410.0210.9700.0590.1110.7034880.51307718.63915.52538.2720.9991.5191.6260.93495.3505.050
   18-4ER (Be)3.705151.463265552.500.5116.497.801.786.200.146.9611.732.190.320.070.020.900.32- 99.956.030.31130.728529.333.830.413294.816.24.5718314.633.10.6000.16168.02.535.880.9034.731.560.6321.990.3722.510.5771.661.770.2771.030.0271.580.0920.1700.7034880.51299518.75115.53038.3580.9881.4221.6400.86752.8971.954
   14-6ER (A1)3.715152.017185252.010.5616.598.633.065.880.156.0211.032.670.430.100.020.840.31- 99.365.360.44535.936322.533.822.594.859.015.64.9144813.431.40.4720.2261212.876.911.035.281.560.5931.830.3332.020.4521.361.360.2060.8990.0341.880.1190.2200.7036920.51306118.79015.53238.3771.0081.5261.6460.92754.2581.446
   15-4ER (A2)3.778151.983186252.730.3813.719.063.076.300.167.1811.331.970.980.190.021.390.55- 99.656.730.85046.136910.640.319.212480.619.218.473312.635.30.6570.4481706.0512.51.959.092.130.7632.340.3441.890.4441.361.280.2020.850-2.920.2910.4610.7037690.51300818.80515.53938.4061.0141.7251.9180.89991.7281.032
   16-14ER (A2)3.702151.873202254.130.5615.717.401.865.730.135.639.992.900.760.170.041.350.301.62
   BCR-2 average -----------------------33.98±0.28392.8±1.7916.47±0.1036.10±0.2313.44±0.0827.30±0.14124.2±0.68-47.22±0.43332.2±3.4438.26±0.33193.2±1.8911.96±0.231.20±0.01667.6±6.5124.92±0.2652.04±0.436.88±0.0128.34±0.046.57±0.012.04±0.016.85±0.031.07±0.006.28±0.001.30±0.003.65±0.023.27±0.020.48±0.004.77±0.000.70±0.0210.32±0.011.63±0.005.99±0.02-----------

4.1. Major and Trace Elements

The compositional range of the samples is described in Sinton et al. [2003], to which the reader is referred for detail. Briefly, SiO2 ranges from 49.6 to 57.7 wt % while MgO ranges from 8.6 to 3.6 wt %. K2O ranges from 0.04 to 0.98 wt % and, with one exception, all samples have TiO2 > 0.5 wt %. On a total alkalis versus silica diagram, the samples are predominantly low-K, tholeiitic basalts and basaltic andesites but the data set does include one andesite (Figure 2). Mg #s extend from 0.70 to 0.44 and Ni and Cr range from 158 to 19 and 334 to 8 ppm, respectively. Thus, all of the lavas are significantly evolved from primary mantle melt compositions.

Figure 2.

Na2O + K2O versus SiO2 for Manus back-arc lavas. Representative field for the New Britain volcanic front (NBVF) is based on samples from Gill et al. [1993] and Cunningham et al. [2009] that have been analyzed for U series isotopes. The two samples with the highest SiO2 are XBABB that have extremely enriched trace elements (see section 4.1 and discussion by Sinton et al. [2003]).

Incompatible trace element concentrations and ratios vary extensively from smooth, incompatible trace element-depleted, MORB-like patterns to samples that are strongly enriched in fluid mobile elements such as Rb, Ba, U, Pb and Sr and which have negative anomalies for fluid immobile elements such as Th, Nb, Ta and Ti (see end-member examples in Figure 3). The latter characteristics are inferred to reflect the variable influence of sediment and fluid components subducted beneath the adjacent New Britain volcanic front (NBVF). Using these trace element and other criteria, Sinton et al. [2003] subdivided the Manus Basin lavas into NMORB lavas (M1 in the southwest MSC and M2 in the northeast), back-arc basin basalts (BABB), and extremely enriched BABB (XBABB based on their high La/Sm and Ba/La. Note that these samples also have the highest SiO2, see Figure 2) all interspersed along the actively spreading MSC and ETZ, and more arc-like basalts (A and B types) in the currently rifting ER and SR. H2O contents in glass rims increase in that order [Shaw et al., 2004; Sinton et al., 2003]. Inferred degrees of flux melting do likewise, reaching the highest values (F: 0.2–0.3) combined with the highest H2O contents and lowest TiO2 contents and thus greatest prior depletion inferred for mantle sources (0.2–0.4 wt % H2O and 0.38–1.72 TiO2 wt %, respectively) [Kelley et al., 2006]. An ocean island basalt (OIB) component with high 3He/4He (up to 13.48) has also been recognized in the vicinity of the Extensional Transform Zone and the Manus Spreading Centre [Macpherson et al., 1998].

Figure 3.

Primitive mantle normalized incompatible element diagram contrasting the end-member examples of mafic MORB and BABB samples from the Manus basin. Grey field illustrates the range of mafic NBVF samples from Gill et al. [1993] and Cunningham et al. [2009].

For the purposes of our investigation, we have simply subdivided the samples into two groups on the basis of Ba/Nb ratio as a common index of the extent of influence of subduction components on back-arc lavas [Pearce and Stern, 2006]. Those with Ba/Nb < 16 are considered relatively unmodified by subduction components and referred hereafter to as MORB while those with Ba/Nb > 16 have variable subduction influence and are termed BABB. The southern lavas closest to the arc consist entirely of BABB whereas both MORB and BABB are found along the MSC and ETZ farther to the north (see Figure 1).

In Figure 2 the MORB are all basalts whereas the BABB extend from basalt to andesite in composition. In terms of both major and trace element compositions, the BABB from the ER and SR overlap the compositions of lavas erupted at the arc front (see Figures 2 and 3). The MORB are depleted in light rare earth elements (REE), whereas the BABB are light REE enriched. Both groups have flat chondrite-normalized heavy REE patterns [Sinton et al., 2003] with (Dy/Yb)N = 0.98–1.02 precluding a role for residual garnet during their genesis.

4.2. Radiogenic Isotopes

Combining our new radiogenic isotope data with those previously published by Sinton et al. [2003], 87Sr/86Sr ranges from 0.70322 to 0.70376, 143Nd/144Nd from 0.513021 to 0.513095, 206Pb/204Pb from 17.98 to 18.81, 207Pb/204Pb from 15.46 to 15.54 and 208Pb/204Pb from 37.86 to 38.41 (Table 2). As discussed by Sinton et al. [2003], the Manus MORB have relatively unradiogenic Sr and Pb combined with radiogenic Nd whereas the BABB extend to significantly more radiogenic Sr and Pb while Nd changes little. 87Sr/86Sr and 143Nd/144Nd largely overlap data for the arc front lavas (Figure 4a). In contrast, in Pb-Pb isotope space (Figure 4b), the Manus data form a tight linear array with the MORBs having 206Pb/204Pb = 18.10–18.30 and the BABB having 206Pb/204Pb = 18.30–18.75, reaching values of the arc. As noted by Sinton et al. [2003], the combined Sr-Nd-Pb isotopic characteristics of the least radiogenic Manus MORB lavas are consistent with a source region having an Indian MORB mantle composition.

Figure 4.

(a) 143Nd/144Nd versus 87Sr/86Sr and (b) 208Pb/204Pb versus 206Pb/204Pb for Manus basin lavas. The field for NBVF lavas incorporates data from Gill et al. [1993] and Cunningham et al. [2009] as well as from Woodhead et al. [1998].

4.3. Uranium Series Isotopes

Seawater has (234U/238U) = 1.14 and elevated concentrations of 230Th and 226Ra [Chen et al., 1986] and so the possibility of seawater alteration of glass is always a concern, especially at the low U contents of these MORB samples, despite the careful leaching and picking procedures employed in sample preparation. Our reproducibility of the (234U/238U) ratio was typically better than 4 ‰ (2σ) during the period of this project, yet the measured (234U/238U) ranges from 0.987 to 1.017 and 5 of the samples deviate from secular equilibrium by more than 1% (Figure 5). However, we observe no correlation of (230Th/238U) with (234U/238U) in Figure 5 or with Cl/K or (234U/238U) with 1/U (not shown). Accordingly, we have used the full data set in the following discussion, accepting that we cannot fully rule out the possibility of minor secondary alteration of some of the samples.

Figure 5.

Plot of (230Th/238U) versus (234U/238U) showing that while some of the samples lie outside of estimated analytical error from (234U/238U) = 1, there is no correlation with (230Th/238U) as would be anticipated from seawater alteration [Chen et al., 1986; Yokoyama et al., 2003]. Reproducibility of the (234U/238U) ratio was typically 4‰ (2σ) (see discussion in section 4.3).

The Manus Basin lavas encompass a large range in (238U/232Th) and (230Th/232Th) ratios from 0.839 to 2.136 and 0.879–1.771, respectively, leading to a range of 0.796–1.047 for (230Th/238U). The occurrence of both 238U and 230Th excesses is similar to observations in the Lau and East Scotia back-arc basins [Fretzdorff et al., 2003; Peate et al., 2001] but the range in (238U/232Th) and (230Th/232Th) ratios is larger in the Manus Basin. Although the samples come from areas with highly reflective side-scan sonar backscatter, indicating very young lava flows, their eruption ages are unknown. Nevertheless, all but 4 of the samples have 226Ra excesses exceeding 4% suggesting that they are younger than 8000 years and thus no age correction is required for the (230Th/232Th) ratios. Measured (226Ra/230Th) ratios range from to 1.01 to 5.55 with one MORB sample (33-3) having a 7% 226Ra deficit outside the analytical error that could be the result of unusual melt extraction conditions [Bourdon and Van Orman, 2006]. Note that sample 42-1 has the most extreme (230Th/238U) (0.826) of the MORB samples.

On the U-Th equiline diagram in Figure 6a, Manus Basin lavas form an array extending from the MORB which have low (230Th/232Th) and 1%–5% 230Th excesses (with the exception of MORB sample 42-1 having 238U excess), to the BABB which show a large range in (230Th/232Th) and 6%–26% 238U excesses. Both ratios change in the order “MORB” < “BABB in the MSC” < “lavas closer to the arc.” The former overlap data from Indian MORB [Tepley et al., 2004; Russo et al., 2009] while the latter extend into, but do not show as large a range in U/Th ratios, as the arc front lavas.

Figure 6.

Plots of (a) (230Th/232Th) versus (238U/232Th) and (b) (226Ra/230Th) versus (230Th/238U) for Manus basin lavas with fields for the NBVF lavas using samples with <60 wt % SiO2 from Gill et al. [1993] and Cunningham et al. [2009]. The dashed field in Figure 6a is for Indian MORB (data from Tepley et al. [2004] and Russo et al. [2009]). Inset in Figure 6b shows (226Ra/230Th) versus Ba/Nb for BABB only.

In Figure 6b, a plot of (226Ra/230Th) versus (230Th/238U), the MORB and BABB are clearly distinguished by their different sense of U-Th disequilibria. Note that an unknown amount of 226Ra decay may have occurred for these samples. However, a number of the BABB fall within the field for historic arc front lavas and two have larger 226Ra excesses and high Ba/Nb (see inset in Figure 6b). Because both Ra and Ba are fluid mobile and often correlated in arc front lavas [Turner et al., 2001], this raises the possibility that the (226Ra/230Th) ratios of some of these samples may be close to initial (i.e., erupted) values (but see below).

Finally, Bourdon et al. [1996b] noted a negative correlation between (230Th/238U) and axial depth for global MORB. This was inferred to reflect an increasing mean pressure of melting beneath the shallower ridges such that Th becomes increasingly less compatible than U [Blundy and Wood, 2003; Landwehr et al., 2001]. Figure 7 compares the Manus Basin data with a global MORB data set. The Manus MORB may display a slight sense of a negative array but are clearly displaced below the global array whereas the Manus BABB lie at significantly lower (230Th/238U) regardless of how inflated the ridge segment is [Russo et al., 2009]; nowhere except in back arcs is there 238U excess at these water depths. Although this may in part reflect our lower (230Th/232Th) ratios during period of analyses (see section 3), the difference between the global MORB data set and the Manus Basin lavas is too large to be solely explained by a 3% analytical shift. A similar observation was made by Peate et al. [2001] for samples in the Lau Basin.

Figure 7.

Plot of (230Th/238U) versus axial depth comparing Manus Basin lavas with a global MORB array (data from Bourdon et al. [1996a, 1996b], Goldstein et al. [1993], and Lundstrom et al. [1998]) and Lau Basin samples [Peate et al., 2001].

5. Discussion

It has long been known that back-arc lavas are variably influenced by H2O and other components inferred to be derived from the nearby subducting plate [Kelley et al., 2006; Langmuir et al., 2006; Pearce et al., 1994; Pearce and Stern, 2006; Taylor and Martinez, 2003] and the Manus basin lavas are no exception to this [Sinton et al., 2003]. For example, Langmuir et al. [2006] recently undertook a quantitative investigation of magma genesis in a number of back arcs including the Manus Basin. Their approach involved removing the effects of shallow level fractionation (including the suppression of plagioclase fractionation by H2O) back to a common reference point of 8 wt % MgO. Using these data, the extent of melting and the effects of prior depletion of the mantle wedge were assessed. Langmuir et al. [2006] concluded that back-arc magmas reflect mixing between high-temperature, high-Ti anhydrous melts and low-temperature, low-Ti hydrous melts.

5.1. Local Mantle Source

The MORB samples with the lowest 206Pb/204Pb represent the mantle source with the least slab component. Because most of them have Ce/Pb = ∼20, Nb/U ∼40, Δ208Pb > 40, 87Sr/86Sr ∼ 0.7033, and 143Nd/144Nd ∼0.51305, they are Indian-type MORB [Sinton et al., 2003]. Their Th/U is high ∼3.5 so that they are displaced to atypically low (238U/232Th) for NMORB (Figure 6a) as with other Indian MORB. Therefore we will assume an Indian depleted MORB mantle (IDMM) mantle source (compiled from Stracke et al. [2003]) in our melting models below.

Manus MORB also have very low Nb, U, and Th concentrations, indicating a higher percentage of melting as a result of the incompatible behavior of these elements and therefore little residual clinopyroxene. This may also account for their minimal enrichment in 230Th due to higher melting rates. High percent melting is consistent with evidence for high potential temperatures in the Manus Basin, including the high F intercept of Manus Basin basalts on a % melting versus H2O source diagram [e.g., Kelley et al., 2006, Figure 10b]. Depressed 230Th enrichment also might be caused by high rates of mantle upwelling but neither spreading rates nor ridge inflation are exceptional where MORB erupt along the Manus Spreading Centre.

5.2. Subduction Components in Manus Lavas

Our aim here is to explore the effects of subduction components on melting dynamics as recorded in U series isotope disequilibria. Thus, the initial assessment requires appraisal of the subduction components involved. These are likely to include subducted sediment and the underlying, possibly hydrothermally altered, oceanic crust. The Pb isotope data for Manus lavas form a tight array in Figure 8a which also shows the estimated composition of Indian depleted MORB mantle (IDMM) and fields for Solomon Sea sediments and altered basalts, which provide the closest estimates of likely subducted components. The Manus data form an array extending from IDMM to a more radiogenic component that has 208Pb/204Pb ≥ 38.4 and 206Pb/204Pb ≥ 18.8. As shown, a mixing line between the sediment and altered basalt having the highest 206Pb/204Pb ratios can form an appropriate end-member for the majority of the data. In practice, the Manus and NBVF data array suggests that more radiogenic end-members exist and this is supported by extrapolation of the Ba/Nb versus 206Pb/204Pb array in Figure 8b. In regard to this speculation, we note that altered Pacific oceanic crust from ODP Site 801 has significantly higher 206Pb/204Pb ratios than those currently measured from the Solomon Sea [Hauff et al., 2003]. Nevertheless, using the available Solomon Sea data, the relative contributions from sediment and altered basalt to the implied subduction component in Figure 8a are in the proportions 5:95.

Figure 8.

(a) Plot of 208Pb/204Pb versus 206Pb/204Pb showing Manus lavas and the fields for Solomon Sea sediment, altered Solomon Sea basalt, and the New Britain Volcanic Front (data from Woodhead et al. [1998]). The mixing line between selected sediment and altered basalt samples suggests that the Pb isotope composition of the subduction component (SC) could be a 5:95 mix of these compositions. A second mixing vector shows the effect of addition of a calculated fluid released from the subduction component to an Indian depleted MORB mantle (IDMM) wedge (compiled from Stracke et al. [2003]). (b) Plot of Ba/Nb versus 206Pb/204Pb with the same SC-IDMM mixing vector. Addition of 0–22‰ of SC fluid to the wedge can broadly describe the composition array of the Manus and NBVF lavas (see section 5.2 for further discussion and Table 3 for end-member compositions and partition coefficients).

In keeping with many previous studies and as we discuss further below, the contributions from the altered oceanic crust and sediment are most likely to have been added as a fluid. Thus, we have calculated the composition of a fluid derived from the 95:5 mix of altered basalt and sediment (see Table 3). We assumed 2% fluid produced in a manner analogous to a batch melting process, an eclogitic source mineralogy, and mineral/fluid partition coefficients from Brenan et al. [1995]. A mixing vector between the IDMM mantle wedge and this calculated subduction component fluid reasonably describes both the Manus and New Britain Volcanic Front data (NBVF) in Figure 8a. Therefore, to a first order, a single slab component pollutes the entire New Britain–Manus Basin system. The mass fraction of Pb from this component in the polluted mantle wedge is indicated by the tick marks in Figure 8a. The order of enrichment is “MORB” < “BABB” < “enriched BABB along the MSC and ETZ” < “BABB closer to the arc.” That is, up to 0.3% slab component is intimately mixed in the source of basalts (sampled within a single dredge) along the MSC and ETZ ∼275 km behind the arc, and ≥ 2.2% (overlapping the amount in the arc itself) is present in the ER and SR ∼150 km behind the arc.

Table 3. Input Parameters for Numerical Models
End-MemberData SourcesBa (ppm)Nb (ppm)Pb (ppm)U (ppm)Th (ppm)226Ra (fg/g)206Pb/204Pb208Pb/204Pb(238U/232Th)(230Th/232Th)(226Ra/230Th)
  • a

    Assumes secular equilibrium with 238U.

  • b

    Fluid is H2O + NaCl; Ra partition coefficients are based on DBa.

  • c

    Fluid is a batch fluid (see section 5.2).

Indian mantle wedge (IDMM)Stracke et al. [2003]
Solomon Sea sediment averageWoodhead et al. [1993, 1998]10635.7117.631.154.334.5E+1118.71538.6252.052.051.00a
Altered Solomon oceanic crust 207.550.851.660.546.5E+1118.87238.1309.326.841.00a
Subduction mix = 95% AOC + 5% sediment 727.461.691.630.736.4E+1118.79038.3886.806.8065.46
Calculated subduction component (SC) BaNbPbUTh      
   Garnet (59%)/fluid partition coefficientsbBrenan et al. [1994, 1995]5.0E-040.060.0280.010.134.8E-05-----
   Clinopyroxene (40%)/fluid partition coefficientsbBrenan et al. [1994, 1995]5.0E-053.330.1150.313.130.00048-----
   Rutile (1%)/fluid partition coefficientsbBrenan et al. [1994, 1995]-164---------
2% fluid from subduction mixc 35542.510020.7111.090.553.2E+1318.79038.39061.026.8065
Dynamic melting at 1.5 GPa            
   Olivine (57%)/melt partition coefficientBlundy and Wood [2003]---6.0E-059.5E-065.8E-08-----
Orthopyroxene (28%)/melt partition coefficientBlundy and Wood [2003]---0.00780.00306.0E-07-----
Clinopyroxene (15%)/melt partition coefficientsBlundy and Wood [2003]---0.01760.01604.1E-06-----

This also means that all geochemical tracers that correlate with Pb isotopes can be attributed to this same slab component (see below). The following discussion is restricted to those which affect U series disequilibria but we note that many other slab-derived features observed in East Lau Spreading Center (ELSC) basalts [e.g., Escrig et al., 2009] also apply to those from the Manus Basin (e.g., elevated LILE and depleted HFSE concentrations relative to the REE). However, the mass fraction of slab component per km behind the arc is higher than in the Lau Basin (i.e., the slab signature is stronger in Manus Basin samples that are further from the trench than in ELSC samples). Earlier, we utilized Ba/Nb as a key discriminant between the Manus MORB and BABB. In practice, there is a continuum between the two and this is well illustrated in Figure 8b. Note that addition of the same subduction component to an IDMM source results in a curved Ba/Nb versus 206Pb/204Pb mixing array that also provides a reasonable approximation of both the Manus and NBVF data in Figure 8b.

Although correlations with Pb isotopes suggests a single slab component, the XBABB group is exceptional in some respects and seems to be the Manus Basin equivalent of the “damp” versus “wet” component discovered at the north end of the ELSC in the Lau Basin, farthest from the Tonga arc [Bézos et al., 2009]. Basalts of this type have less enrichment in fluid mobile elements such as K, Rb, Ba, and U relative to enrichments in the HFSE (Ti, Zr, Nb), and therefore are closer to conventional EMORB. XBABB are as enriched in fluid mobile elements as the “BABB” of the ER and SR, but differ from them by having H2O/Ce as low as MORB, high Ce/Pb and K2O/TiO2 relative to Nb/Yb, the highest Nb/(Zr,Yb) ratios, and suprachondritic Nb/Ta and Zr/Hf. Indeed, because Nb/Yb ratios exceed those in EPR EMORB, XBABB may reflect the OIB component in the Manus Basin [e.g., Macpherson et al., 1998].

5.3. Nature of the Subduction Component

An important aspect of Figure 8a is that the Manus lava array project from IDMM toward a tie line between the Solomon Sea sediment and the altered basalt. As argued previously for the arc front lavas [Woodhead and Johnson, 1993; Woodhead et al., 1998], this is consistent with a contribution from both components. However, because mixing is linear on such diagrams, the geometry in Figure 8a also requires that the two subducted components were mixed together prior to being transported out of the slab into the mantle wedge. This is in contrast to many studies that have suggest that at least for arc front lavas, the sediment component is added separately to the wedge and before fluid addition [Elliott, 1997; Turner et al., 1997]. However, if this were the case in the Manus Basin, the data should project from the fluid apex toward a point on the IDMM-sediment tie line and this is clearly not the case.

In Figure 9 we explore further the nature of the subduction component. Increases in the contribution from this component, as measured by increasing 206Pb/204Pb, are accompanied by increasing U/Th (Figure 9a) as well as Ba/Nb (Figure 8b) and, by analogy, Ra/Th. This is accompanied by increasing Fe3+/ΣFe (Figure 9b) suggesting increasing oxidation of those sources most contaminated by the subduction component. Finally, Figure 9c suggests that although there appears to be variable H2O in the Manus MORB, the BABB all have elevated H2O, reinforcing the findings of Shaw et al. [2004] and consistent with there being a positive correlation between H2O and Fe3+/ΣFe [Kelley and Cottrell, 2009]. However, H2O/Ce ratios (mostly 500–1500) are too low to be fluids in the conventional sense so that the subduction component either was a water-rich melt [e.g., Plank et al., 2009] or is stored in hydrous minerals before melting [Sinton et al., 2003]. Consequently we infer that the slab component, in addition to being Cl [cf. Sinton et al., 2003] and water-rich, was oxidized and preferentially transported U and Ra relative to Th. We will explore the time significance of this below.

Figure 9.

Plots of (a) U/Th, (b) Fe3+/ΣFe, and (c) H2O versus 206Pb/204Pb suggesting that those lavas with the greatest contribution from the subduction component have higher U/Th, are more oxidized, and have elevated H2O. H2O data are for glass from Sinton et al. [2003] and Shaw et al. [2004]. Note that some data are from glass and some are from whole rock analyses (see Table 2).

5.4. Coupled Flux and Decompression Melting

U series isotopes have the potential to provide insights into the time scales and physical processes, such as melting rate and residual porosity, of mantle melting [e.g., McKenzie, 2000]. Because U behaves as a fluid mobile element under oxidizing conditions it is widely thought that the 238U excesses commonly observed in arc front lavas reflect preferential addition of U in fluids from the slab [see Turner et al., 2003 for a review]. It is implicit in such models that the addition of such fluids induces partial melting beneath arc volcanoes. If the 238U excesses in the Manus lavas reflect addition of U without Th [Elliott, 1997; Turner and Hawkesworth, 1997] then the slope of the array in Figure 6 would indicate this occurred ∼ 140 kyr ago throughout the basin (but see below).

In theory, minor (up to 5%) 238U excesses can be produced by decompression melting in the absence of flux melting so long as the depth of melting is sufficiently shallow (<1 GPa) that DU < DTh [Landwehr et al., 2001] or the percent melting is high enough to exhaust clinopyroxene or both. Small 238U excesses have been found in some MORB from the Mid-Atlantic Ridge, the Kolbeinsey ridge and the Garrett Transform fault [Bourdon et al., 1996b; Sims et al., 2002; Tepley et al., 2004]. The majority of the Manus BABB 238U excesses are >5% and are associated with the presence of a slab component based on trace element and Pb isotope data.

MORB are typically characterized by 230Th excesses inferred to result from in-growth of 230Th during partial melting (see Elliott and Spiegelman [2003] for a review). The likelihood of decompression melting beneath actively spreading back-arc basins raises the possibility that some in-growth also occurs there and even beneath arcs [George et al., 2003]. Indeed, the occurrence of 231Pa excesses in the majority of arc front lavas provides compelling evidence for in-growth during melting in this environment as well [Turner et al., 2006]. However, it is less clear to what extent partial melting affects 238U-230Th disequilibria in arc lavas (relative to the effects of fluid addition of U), not least because it is possible that arc lavas are generated at pressures (∼ 1 GPa) where U and Th have very similar partition coefficients [Landwehr et al., 2001].

The progression from 230Th excesses in Manus MORB to 238U excesses in the Manus BABB is associated with increasing subduction component, water content, and oxidation (Figure 9) providing evidence for a link between the role of water, redox, and the production of 238U excesses in this back-arc basin. In Table 3, we have calculated the U-Th-Ra composition of the model fluid released from the slab (see also Figure 8a). Note that this subduction component contains a small amount of Th (see Table 3). The effect of addition of this component to the IDMM mantle wedge is to produce a very slightly inclined array extending to the right of the equiline as illustrated in Figure 10a.

Figure 10.

Plots of (a) (230Th/232Th) versus (238U/232Th) and (c) (226Ra/230Th) versus (230Th/238U) showing the effects of addition of the calculated subduction component to an IDMM mantle wedge in secular equilibrium followed by dynamic melting during which in-growth of both 230Th and 226Ra occurs (see section 5.4 and Table 3 for input parameters). (b) Illustrates the preferred model in which the subduction component is old such that the subduction modified mantle forms an array along the equiline. Subsequent dynamic melting in unoxidized mantle leads to 230Th excess for the MORB (partition coefficients as in Table 3) but 238U excess for the BABB coming from oxidized mantle where DU = DTh/10 [Lundstrom et al., 1994]. Note that mantle sources in Figure 10b start along the equiline in contrast to the fluid addition curve in Figure 10a.

Some aspects of this model differ from those for other arcs such as the Marianas and Tonga [Elliott, 1997; Turner et al., 1997]. First, Solomon Sea sediment is U-rich so that its addition raises (230Th/232Th) ratios whereas the greater amount of pelagic sediment in the other arcs lowers (230Th/232Th). This difference makes it unnecessary to invoke addition of a U-enriched sediment melt >350 ka before fluid addition. Second, the slope of the mixing array depends on many assumed values. We used the partition coefficients of Brenan et al. [1995] for Cl-rich (Cl/K ranges from 2 to 6 in the Manus lavas) brines at high fO2, we did not include accessory minerals, we assumed a 2% fluid fraction and no updip loss of fluid.

Importantly, the Manus data form a much steeper array than the subduction component addition vector in Figure 10a. Conceivably, this could reflect ∼ 140 kyr of aging after addition of the subduction component although this would necessitate some subsequent trigger for partial melting, such as decompression due to back-arc spreading. However, when considered together the Manus MORB and BABB form an apparently contiguous array that crosses the equiline (Figure 10a) and simple addition of subduction components to the mantle wedge cannot account for this. A similar observation has been made in other back arcs such as the Lau Basin [Peate et al., 2001].

As mentioned above, one effect of partial melting is an increase in 230Th if U is more compatible than Th in the residue and if melting occurs over a time scale that is significant relative to the half-life of 230Th (75 kyr). The effect is most pronounced for sources with high U/Th ratios and so partial melting of variably metasomatized mantle also results in a positive sloped array on the U-Th equiline diagram [e.g., George et al., 2003; Turner et al., 2003]. In addition, the melt productivity (related to the melting rate) is lower for hydrous melting than for melting of anhydrous peridotite [Asimow et al., 2004]. We illustrate this for the Manus Basin using the dynamic melting model of Williams and Gill [1989] and the calculated composition of an IDMM mantle wedge variably enriched by the subduction component as shown in Figure 10a. The residual porosity (ϕ), or threshold at which melt is extracted from the matrix, was set at 10−3 and the extent of melting was 10% and 20% for the MORB and BABB, respectively [e.g., Kelley et al., 2006; Langmuir et al., 2006]. The melting rate (M) was set equal to an upwelling rate of 4.6 cm/yr, which is the half spreading rate in the Manus Basin [Sinton et al., 2003]. Melt productivities (Γ) were 0.5 and 0.35%/kbar, appropriate to anhydrous and hydrous peridotite melting [Asimow et al., 2004], for the MORB and BABB respectively. Source mineralogy and partition coefficients are given in Table 3 assuming a pressure of ∼1.5 GPa based on the absence of any residual garnet signature in the REE patterns (see also Figure 3).

Without attempting to reproduce the exact details of the data, the results show that such a model can provide a remarkably good first-order approximation of the Manus data (Figure 10a). Small 230Th excesses similar to those observed in the Manus MORB, are produced by dynamic melting of peridotite little affected by the subduction component as long as the pressure is ∼1.5 GPa so that DU is slightly higher than DTh for clinopyroxene. BABB with high (230Th/232Th) ratios are well simulated by dynamic melting of peridotite that had undergone variable prior enrichment by the subduction component. The array of data in Figure 10a reflects, first, the range in mass fraction of slab component added to the IDMM mantle and, second, decompression melting at 4.5 cm/yr upwelling. Thus, the Manus data can be explained by a combination of variable enrichment by subduction components (including H2O) that initiate coupled flux and decompression melting during which 230Th in-growth occurs (but see below). This combination of processes leads to a much steeper array on the U-Th equiline diagram than either flux melting or decompression melting alone suggesting that the combination also may be important in some arc fronts where strongly inclined U-Th isotope arrays have been observed (e.g., Kamchatka [Dosseto et al., 2003; Turner et al., 1998, 2007] and Nicaragua [Reagan et al., 1994]).

Some of the features that correlate with the slab component in Manus Basin lavas (e.g., the time-dependent (230Th/232Th)) also correlate with 10Be/9Be in the NBVF and western Bismarck arc [Gill et al., 1993]. This confirms that the slab component was derived in part from sediment and was added to the mantle wedge within the last several hundred thousand years.

Although Manus lavas have probably undergone an unknown amount of post eruptive decay of 226Ra, we show the same mixing and dynamic melting models on the plot of (226Ra/230Th) versus (230Th/238U) in Figure 10c for completeness. 226Ra-230Th disequilibria are mostly sensitive to the residual porosity and, since the observed (226Ra/230Th) ratios must be regarded as minima, some of the Manus BABB may require ϕ ≤ 10−3. Some basalts from both the ER and SR have (226Ra/230Th) > 5, as high as at the Tongan volcanic front [Turner et al., 2001] and higher than anywhere at the New Britain volcanic front [Cunningham et al., 2009; Gill et al., 1993]. Future work on 231Pa analyses of the Manus lavas will provide tighter constraints on ϕ and allow refinement to the melting models.

6. Decompression Melting of an Oxidized, Hydrous Mantle Wedge

Some implications that arise from the Manus data and the analysis above are rather problematic and warrant further discussion. The most obvious is the occurrence of subduction signatures so far from the arc and above the slab. Although this observation is not new, it is striking how similar (if not indistinguishable) many of the Manus samples from the ER and SR, 150 km behind the arc, are to the arc front lavas, and that the same component is present to a lesser extent along the MSC and ETZ 275 km behind the arc (see Figures 24, 6, and 8). Indeed, Sinton et al. [2003] questioned whether the Manus spreading center and extensional transfer zone lie above the slab at all. This raises important questions as to how the subduction signature reaches so far behind the arc so quickly. Derivation from the slab at 200–500 km depth seems incompatible with the absence of a garnet signature (Figure 3) and this is supported by major element arguments that the average pressure of melting is ∼ 1–2 GPa [e.g., Langmuir et al., 2006]. Some studies have suggested that slab roll-back might leave behind a subduction “polluted wedge” in its wake [e.g., Pearce et al., 1994] or the translation of hydrated peridotite far across the wedge [e.g., Davies and Stevenson, 1992; Sinton et al., 2003] while others have pondered the possible involvement of hydrated lithosphere [Langmuir et al., 2006]. However, the U series disequilibria data are not compatible with any of the models as discussed below.

As illustrated in Figure 9, if the subduction signal and 238U excesses start with flux melting, then a single episode of addition of an oxidized, hydrous component must occur throughout the Manus Basin and NBVF. In principle, this might result in the formation of pargasite, chlorite, and/or phlogopite in the mantle wedge [see Sinton et al., 2003]. However, under normal mantle redox conditions, amphibole is not capable of fractionating U from Th and phlogopite, and presumably chlorite, have such vanishingly small partition coefficients for U and Th that this phase cannot be responsible for the observed disequilibria either [Blundy and Wood, 2003]. Moreover, it is not at all clear that any hydrous phases could survive in the residue after 10%–20% flux melting [Kelley et al., 2006; Langmuir et al., 2006].

By analogy with arc front lavas, it seems more likely that the 238U excesses were started with the addition of a hydrous subduction component [see Turner et al., 2003 for a summary]. However, once created, any 238U excesses will decay away within ∼ 350 kyr and the models discussed above and shown in Figure 10a suggest that the maximum time since U addition is likely to be less than 140 kyr. As discussed by Turner and Hawkesworth [1997] this is too short for horizontal translation across the mantle wedge via the mechanism proposed by Davies and Stevenson [1992] and it seems improbable that fluids could hydrofracture hundreds of kilometers through the wedge [e.g., Davies, 1999].

Langmuir et al. [2006] developed a model in which BABB reflect mixing between hydrous melts produced in the arc-side half of a ridge-like melting triangle and anhydrous peridotite melts produced on the far side. Melting in both halves is caused by decompression. These models provide reasonable fits to major and trace element data from a number of back-arc basins including the Manus Basin. Mixing is linear on the U-Th equiline diagram and so the Manus array in Figure 6a could be consistent with mixing between an anhydrous peridotite melts (represented by MORB with the lowest U/Th and 206Pb/204Pb) and hydrous peridotite melt (represented by the “BABB” from the ER and SR with the highest U/Th and 206Pb/204Pb). There is support for this model in the geographic distribution of the lava types in the Manus Basin. Closest to the arc, the SR and ER lavas are exclusively BABB whereas a much greater diversity of types ranging from MORB to BABB have erupted along the Manus Spreading Centre and Extensional Transfer Zone farther to the north (Figure 1). However, these models are less easily reconciled with the U series age constraints on the fluid component. As discussed above, this is between 0 and 140 kyr and so, particularly in the Manus case, it seems hard to see how this could reach regions undergoing decompression several 100 km from the arc front.

Another recent model involves diapiric upwelling of hydrated peridotite created above the slab surface [Gerya and Yuen, 2003; Grove et al., 2006; Hall and Kincaid, 2001; Tamura and Tatsumi, 2002]. These could translate the 238U excess created by fluid addition from the slab. However, partial melting (at 1–2 GPa) would have to occur before this signal had decayed away which corresponds to a vertical rise rate of >1 m/yr significantly faster than the numerical estimates of Gerya and Yuen [2003]. U-Th age constraints from the East Scotia back arc (∼100 kyr) and Lau Basin (∼50 kyr) are also incompatible with this model [Fretzdorff et al., 2003; Peate et al., 2001].

An alternative interpretation, that avoids many of the problems with the models discussed above, is that melting is caused solely by decompression under oxidizing conditions. A number of studies have shown that U changes from slightly more compatible than Th in the +4 valence state to highly incompatible in its +6 form as oxygen fugacity changes from ∼ QFM to QFM + 1 to 2 [Berry et al., 2008; Lundstrom et al., 1994; O'Neill et al., 2009]. If DU ≪ DTh, then dynamic melting leads to decreasing (230Th/232Th), as unsupported 230Th decays in the matrix, and results in 238U excesses. As illustrated in Figure 10b, this can produce an inclined array to the right of the equiline, simulating the BABB whilst the MORB obtain 230Th excesses via the normal process of 230Th in-growth due to dynamic melting of more reducing mantle where DTh < DU. In this model, the subduction component is likely stored in the form of hydrous phases indefinitely [Sinton et al., 2003] but it is the elevated oxygen fugacity in these regions that leads to the production of 238U excesses rather than direct fluid addition as normally invoked for arc lavas. Thus, this model starts with mantle sources along the equiline in Figure 10b unlike the recent fluid addition curve in Figure 10a. Note that because DTh and DRa are not sensitive to oxidation state, there is little difference between the two models for (226Ra/230Th).

Accordingly, we suggest that Figures 9b and 9c indicate that variable addition of the subduction component resulted in elevated U/Th ratio, variable oxygen fugacity and elevated H2O in the mantle beneath the Manus Basin. If this modification occurred >350 kyr ago and <8 Myr ago the result would be a wedge in secular equilibrium (i.e., mantle wedge samples will plot along the equiline) but with variable U/Th ratios, variable fO2 and live 10Be. Subsequent melting due to decompression would lead to 238U excesses in oxidized regions (BABB) and 230Th excesses (MORB) in areas minimally affected by subduction. This relaxes the U-Th time constraint on the timing and single-stage nature of fluid addition from the slab, and smearing out of slightly older solid heterogeneities may help explain the intimate juxtaposition of BABB and MORB along the MSC and ETZ (see Figure 1). It also provides a new potential explanation for the steep U-Th isotope arrays observed for Kamchatka [Dosseto et al., 2003; Turner et al., 1998, 2007] and Nicaragua [Reagan et al., 1994].

An important caveat to this interpretation is that all of the observed 226Ra disequilibria must be produced by in-growth during partial melting (see Figure 10b) since any fluid derived 226Ra excesses would have decayed away within 8 kyr. This is permissible from the present data since correlations between 226Ra excesses and slab fluid indices in the back-arc lavas are weak although the two samples with large 226Ra excesses (one each in ER and SR) require either very low porosity or recent fluid addition and rapid ascent (Figure 10b). In general, the situation at arc fronts appears to be very different where 226Ra excesses are strongly correlated with slab fluid indices and it seems likely that fluid addition induces partial melting on addition to the wedge above the slab [Turner et al., 2001]. Conversely, if such a correlation were found in back-arc lavas it would provide a powerful constraint on the competing models. Finally, we note that for elements enriched in the subduction component, the effects of flux melting outweigh the effects of decompression melting, even in the hottest back-arc basin. U series isotopes alone have the potential to distinguish decompression melting.

Although decompression-only melting in the Manus back arc has the fewest problems, it begs the question why percent melting and water correlate positively in subduction zones but negatively elsewhere [Kelley et al., 2006; Langmuir et al., 2006; Taylor and Martinez, 2003] if flux melting plays no role. The answer might be in the role of modal hydrous minerals that store the subduction component beneath arcs and back arcs versus nominally anhydrous minerals that store water elsewhere. If the modal proportion of metasomatic minerals controls percent melting in subduction environments but primarily affects only depth of melting elsewhere, then perhaps the difference reflects the integrated history of fluxing rather than the short-lived (millennial) process of “flux melting.”

If decompression-only melting occurs, then the ratio of water to other elements may reflect only the melting process and the hydrous phases being melted, not the flux. If that is the case, then ratios like H2O/Ce cannot be used to track slab surface temperatures and whether or not the slab component is a melt or fluid. Because the H2O/Ce ratios throughout the Manus Basin are typical of rear arcs globally (500–1500), our conclusion may apply widely.

7. Conclusions

The Pb and trace element systematics of the Manus Basin lavas suggest that a single slab component pollutes the entire New Britain–Manus Basin system and that linear mixing arrays and the geometry of mixing arrays in Pb-Pb isotope space indicate that this components is composed of 95:5 altered oceanic crust and subducted Solomon Sea sediment that mixed prior to being introduced into the mantle wedge. A progression from 230Th excesses in Manus MORB to 238U excesses in Manus BABB is associated with an increasing influence of the subduction component, water content and state of oxidation. In contrast to many arc front models whereby variable enrichment by subduction components (including H2O) initiate flux melting, we propose an interpretation in which melting is caused solely by decompression. Under oxidizing conditions DU < DTh results in 238U excesses whereas less modified mantle yields 230Th excesses in the normal way. Our model accounts for the observed U-Th array in the Manus Basin and relaxes U-Th time constraints which would otherwise require fluid velocities >1 m/yr in the mantle wedge.


We gratefully acknowledge the help of Peter Wieland, Norman Pearson, John Caulfield, and Craig Cook with the isotope analyses in Sydney and of Roland Maas and Jon Woodhead for the Pb isotope analyses. We acknowledge the very constructive and helpful reviews of David Peate and Thomas Kokfeld that helped improving the manuscript. Christoph Beier was supported by a Feodor Lynen fellowship of the Alexander von Humboldt Foundation and thanks Coopers Sparkling and the Durras Eco Point Resort for inspiration. Simon Turner was supported by an Australian Research Council Federation Fellowship, and this work was directly funded by Discovery Proposal 0771610. The analytical data were obtained using instrumentation funded by ARC LIEF and DEST, Systemic Infrastructure Grants, industry partners and Macquarie University. This study was funded by ARC Discovery grant 9211/2141. This is contribution 653 from the Australian Research Council National Key Centre for the Geochemical Evolution and Metallogeny of Continents and SOEST contribution 7929.